Upload
others
View
3
Download
0
Embed Size (px)
Citation preview
Imperial College London
Department of Surgery and Cancer
Identification of potential novel
biomarkers in neuroendocrine
tumours of the
gastroenteropancreatic system.
Helen Cara Miller
June 2018
Submitted in part fulfilment of the requirements for the degree of
Doctor of Philosophy in Surgery and Cancer of Imperial College London
and the Diploma of Imperial College London
1
Declaration of Originality
The research presented herein is my own work except where the work of others is ac-
knowledged.
Copyright Declaration
The copyright of this thesis rests with the author and is made available under a Creative
Commons Attribution Non-Commercial No Derivatives licence. Researchers are free to
copy, distribute or transmit the thesis on the condition that they attribute it, that they
do not use it for commercial purposes and that they do not alter, transform or build upon
it. For any reuse or redistribution, researchers must make clear to others the licence terms
of this work.
3
Dissemination
Miller, H. C. Frampton, A. E. Malczewska, A. Ottaviani, S. Stronach, E. A.
Flora, R. Kaemmerer, D. Schwach, G. Pfragner, R. Faiz, O. Kos-Kudta, B.
Hanna, G. B. Stebbing, J. Castellano, L. Frilling, A. (2016). MicroRNAs
associated with small bowel neuroendocrine tumours and their metastases.
Endocrine-Related Cancer, 23(9), pp. 711-726.
Miller, H. C. Kidd, M. Modlin, I. M. Cohen, P. Dina, R. Drymousis, P.
Vlavianos, P. Kloppel, G. Frilling, A. (2015) Glucagon receptor gene muta-
tions with hyperglucagonemia but without the glucagonoma syndrome. World
Journal of Gastrointestinal Surgery, 7(4), pp. 60-66.
Miller, H. C. Kidd, M. Castellano, L. Frilling, A. (2015). Molecular genetic
findings in small bowel neuroendocrine neoplasms: a review of the literature.
International Journal of Endocrine Oncology, 2(2), pp. 143-150.
Miller, H. C. Drymousis, P. Flora, R. Goldin, R. Spalding, D. Frilling, A.
(2014). Role of Ki-67 proliferation index in the assessment of patients with
neuroendocrine neoplasias regarding the stage of disease. World Journal of
Surgery, 38(6), pp. 1353-1361.
5
Abstract
Gastroenteropancreatic neuroendocrine tumours (GEP-NET) are rare tumours arising
in the neuroendocrine cells of the digestive system. Chromosomal instability is rarely
observed in GEP-NET suggesting epigenetic changes, such as changes in microRNA
(miRNA) expression, may be drivers of disease pathology. There is an unmet clinical
need for novel prognostic biomarkers to enable further stratification of GEP-NET pa-
tients based on tumour behaviour and to inform treatment options.
In this thesis a study of 161 GEP-NET patients demonstrates that liver metastases
remain a common event despite the majority of tumours having low proliferation levels
as assessed by the proliferation marker Ki-67. 28 % of the GEP-NET patients with a
Ki-67 % of ≤ 2 % (G1) had stage IV disease. The results are even more striking for
patients with small bowel neuroendocrine tumours (SBNET) with 54 % of G1 SBNET
patients having stage IV disease.
In order to identify novel prognostic biomarkers for use in patients with SBNET, 800
miRNA are quantified in 90 different tissue samples from 37 SBNET patients. This
work represents the most comprehensive investigation of miRNA expression in SBNET
to date. Novel miRNA are identified that have not been previously associated with
SBNET tumourigenesis and disease progression. These miRNA warrant further study to
better understand their contribution to disease pathology in SBNET.
The most promising potential biomarkers associated with disease progression in SBNET
are validated in two independent populations of SBNET patients to ensure that the results
6
are reproducible. Further analysis demonstrates that miR-1 and miR-143-3p are the most
promising candidates for use as potential novel prognostic biomarkers in SBNET patients.
Further studies are warranted to determine the clinical utility of miR-1 and miR-143-3p
as prognostic biomarkers and to determine if they can be used to identify patients with
more aggressive disease subtypes and enable tailored treatment.
7
Acknowledgements
The work presented in this thesis was kindly supported by the Heinz-Horst Deichmann
Foundation.
There are a number of people who I would like to thank for their assistance and kindness
during my PhD project.
Firstly I would like to thank my supervisors Professor Andrea Frilling, Dr Euan
Stronach and Professor Robert Goldin for their help and guidance throughout my PhD.
I would also like to thank Panos Drymousis for his invaluable help and advice and for
answering all of my many clinical questions. I would like to thank Anne-Marie Feeney
for her kind encouragement throughout and for her help in tracking down clinical data.
Thank you also to Gule Hanid, Bernadette Khoshaba and Anna Malczewska.
Thank you to Rashpal Flora for his help and for always making the time to go over
slides with me despite a very heavy work load.
A big thank you to everyone in the Molecular Therapy Lab for making me feel so
welcome and for your encouragement and kindness. It made a huge difference. Thank
you Paula Cunnea, Elaina Maginn, Karen Menezes, Camila Henrique de Sousa, Phil
Lawton, Raj Burmi, Jamie Studd, Nona Rama, Matthias Pfeifer, Yuliana Astuti and
Ratri Wulandari.
Thank you to everyone within the Department of Histopathology at Hammersmith
Hospital and in particular to Roberto Dina and Patrizia Cohen. Thank you to Pritesh
Trivedi for always doing his best to fit me in on the busy IHC machines and to Patricia
Hoynes for her kind assistance in locating FFPE blocks and slides. Thank you also to
Anna Mroz and Hiromi Kudo for their advice and IHC instruction.
Thank you to Leandro Castellano and Adam Frampton for guidance on qPCR and
data analysis. Thank you also to Caoimhe Walsh who assisted me with some of the RNA
extractions and haematoxylin staining as part of her BSc project.
Thank you to Dr Mark Kidd for very kindly welcoming me into his lab at Yale Uni-
9
versity for a month at the start of my PhD. I had a wonderful time and learnt a lot of
new techniques. Thank you to Tarjei Dahl Svendsen, Jonas Jørandli, Andrew Taylor,
Brittany Davis, Daniele Alaimo, Wouter Hogendoorn and Takeshi Moriguchi for being so
friendly and making me feel at home.
I would like to thank my friends and family for supporting me and providing me with
some much needed respite and laughter. Thank you to my mother, Erika, and sister,
Amy, for the many times you have helped and supported me during my PhD.
Thank you to Laura, Becca, Lucy, Tasha, Abi, Bernadett, Madina and Florence for
always being there for me when I needed you and for the jokes and infectious laughter
whenever we get together. It wouldn’t have been possible without you!
Thank you to Max and John for trusting me with power tools and letting me loose on
crazy building projects! The banter was hilarious and I had so much fun.
Thanks also to all my awesome cinema friends, Adam, James, Cecilia, Luke, Sophie,
Aeolus, Emily, Marlen and the rest of the gang at Imperial Cinema for the camaraderie
and countless all-nighters and supermarket runs.
Thank you to Mary, Jeremy, Rebecca, Alasdair, Will, Geoffrey, Paul and Lil for your
friendship and kind words.
Finally thank you to my partner George for supporting me through thick and thin and
for putting up with me while I was writing up! I couldn’t wish for a more loving, kind
and accepting person to spend my life with. You help me be the best possible version of
myself and you’re always rooting for me when I go after my dreams.
Helen
June 2018
10
Contents
1. Introduction 29
1.1. Aim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
1.2. Research objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
1.3. Contribution to knowledge . . . . . . . . . . . . . . . . . . . . . . . . . . 33
1.4. Document outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2. Literature Review 36
2.1. Epidemiology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.1.1. Incidence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.1.2. Survival . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.1.3. Risk Factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
2.2. Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
2.2.1. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
2.2.2. Primary site . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
2.2.3. Grade . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
2.2.4. Stage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
2.2.5. Functioning syndromes . . . . . . . . . . . . . . . . . . . . . . . . 78
2.3. Treatment and imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
2.3.1. Treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
2.3.2. Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
11
2.4. Neuroendocrine cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
2.4.1. Development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
2.4.2. Differentiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
2.4.3. Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
2.4.4. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
2.5. MiRNA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
2.5.1. Regulation of gene expression . . . . . . . . . . . . . . . . . . . . 136
2.5.2. Dysregulation in cancer . . . . . . . . . . . . . . . . . . . . . . . 141
2.5.3. SBNET . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
2.5.4. PNET . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
2.5.5. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
2.6. Biomarkers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
2.6.1. Established biomarkers . . . . . . . . . . . . . . . . . . . . . . . . 171
2.6.2. Potential future biomarkers for use in patients with SBNET . . . 183
2.7. Gaps in the literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
3. Methods 193
3.1. Ethics Approval . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
3.2. Ki-67 % . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
3.2.1. Patient details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
3.2.2. Grade and stage . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
3.2.3. Data analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
3.2.4. Heterogeneity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
3.3. miRNA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
3.3.1. Patient Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
3.3.2. RNA extraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
3.3.3. Global miRNA quantification . . . . . . . . . . . . . . . . . . . . 204
3.3.4. Validation of candidate miRNA by qPCR . . . . . . . . . . . . . 206
12
3.3.5. IHC Ki-67 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
3.3.6. H&E . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
3.4. Bioinformatics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
3.4.1. Predicted gene targets of candidate miRNA . . . . . . . . . . . . 213
3.4.2. Gene expression datasets . . . . . . . . . . . . . . . . . . . . . . . 214
3.4.3. Data processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
3.4.4. Gene ontology and pathway analysis . . . . . . . . . . . . . . . . 218
4. Role of the Ki-67 proliferation index and disease stage in GEP-NET 220
4.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
4.1.1. Summary of results . . . . . . . . . . . . . . . . . . . . . . . . . . 221
4.2. Patients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
4.3. Grade and stage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
4.3.1. Metastases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
4.3.2. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
4.4. Tumour characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
4.4.1. Invasiveness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
4.4.2. Functionality and genetic status . . . . . . . . . . . . . . . . . . . 228
4.4.3. SBNET . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
4.4.4. PNET . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
4.5. Second primary malignancies . . . . . . . . . . . . . . . . . . . . . . . . . 230
4.6. Ki-67 % Heterogeneity . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
4.6.1. Patients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
4.6.2. Intertumoural heterogeneity . . . . . . . . . . . . . . . . . . . . . 232
4.6.3. Intratumoural heterogeneity . . . . . . . . . . . . . . . . . . . . . 233
4.6.4. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
4.7. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
13
5. Global miRNA expression profiling in SBNET, miRNA quantification
in matched tissue from the primary tumour and metastatic sites 244
5.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
5.1.1. Summary of results . . . . . . . . . . . . . . . . . . . . . . . . . . 245
5.2. Global miRNA expression profile . . . . . . . . . . . . . . . . . . . . . . 246
5.2.1. SBNET . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
5.2.2. Lymph node metastases . . . . . . . . . . . . . . . . . . . . . . . 252
5.2.3. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
5.3. Candidate miRNA validation by a second quantification method . . . . . 255
5.3.1. SBNET . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
5.3.2. Lymph Node metastases . . . . . . . . . . . . . . . . . . . . . . . 258
5.3.3. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
5.4. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
6. Validation of the global miRNA profiling in an independent group
of SBNET patients and the identification of miRNA dysregulated in
liver metastases. 263
6.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
6.1.1. Summary of results . . . . . . . . . . . . . . . . . . . . . . . . . . 265
6.2. SBNET patients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265
6.2.1. MiRNA expression, primary tumour . . . . . . . . . . . . . . . . 266
6.2.2. SBNET miRNA profile validation . . . . . . . . . . . . . . . . . . 269
6.2.3. MiRNA signature of SBNET . . . . . . . . . . . . . . . . . . . . . 276
6.2.4. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277
6.3. MiRNA implicated in metastatic disease . . . . . . . . . . . . . . . . . . 280
6.3.1. Liver metastases . . . . . . . . . . . . . . . . . . . . . . . . . . . 280
6.3.2. Lymph node metastases . . . . . . . . . . . . . . . . . . . . . . . 281
6.3.3. Disease progression . . . . . . . . . . . . . . . . . . . . . . . . . . 282
14
6.3.4. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286
6.4. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287
7. Bioinformatics 293
7.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293
7.1.1. Summary of results . . . . . . . . . . . . . . . . . . . . . . . . . . 295
7.2. Candidate miRNA and gene expression datasets . . . . . . . . . . . . . . 296
7.2.1. Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 298
7.3. Comparison of gene lists . . . . . . . . . . . . . . . . . . . . . . . . . . . 299
7.3.1. SBNET . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 300
7.3.2. Lymph node metastases . . . . . . . . . . . . . . . . . . . . . . . 302
7.3.3. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303
7.4. Enriched gene ontology terms and pathways . . . . . . . . . . . . . . . . 303
7.4.1. SBNET . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 304
7.4.2. Lymph node metastases . . . . . . . . . . . . . . . . . . . . . . . 306
7.4.3. Oncogene targets of downregulated miRNA in lymph node metastases325
7.4.4. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327
7.5. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 330
8. Discussion and Further Work 332
8.1. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 332
8.2. Further work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344
8.2.1. Experimental validation of bioinformatics results . . . . . . . . . . 344
8.2.2. Functional studies . . . . . . . . . . . . . . . . . . . . . . . . . . . 345
8.2.3. Future biomarker development . . . . . . . . . . . . . . . . . . . . 350
A. Sample ID dataset 1 422
B. Primers for qPCR 424
15
C. RNA extractions 425
D. Dysregulated miRNA 429
E. Bioinformatics 435
E.1. Genes list lymph node metastases . . . . . . . . . . . . . . . . . . . . . . 435
E.2. Enriched gene ontology terms SBNET . . . . . . . . . . . . . . . . . . . 436
E.3. Enriched gene ontology terms lymph node metastases . . . . . . . . . . . 440
F. Permission for reprints 441
F.1. Published papers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 443
16
List of Figures
2.1. Function of miRNA. A) Genes are transcribed into mRNA which are trans-
lated into protein (central dogma). B) miRNA regulate gene expression by
binding to the mRNA of certain genes and preventing their translation into
protein. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
2.2. The biogenesis of endogenous miRNA and their regulation of gene expres-
sion by RNA interference. Techniques for exogenous gene silencing which
utilise this biological pathway are also shown, with the introduction of small
interfering RNA (siRNA) or a short hairpin (shRNA) encoded in a viral
vector. Figure reproduced from (Bak and Mikkelsen, 2010), creative com-
mons licence: CC BY 2.0. . . . . . . . . . . . . . . . . . . . . . . . . . . 137
2.3. The effects of miRNA dysregulation during tumourigenesis. A) The role of
a miRNA in normal tissue. B) During tumourigenesis, different stages in
the miRNA biogenesis can become dysregulated or the miRNA gene may be
deleted/mutated leading to reduced levels of the miRNA and inappropriate
expression of the target oncogene. C) During tumourigenesis amplifica-
tion/overexpression of a miRNA can occur, so that it is expressed in the
wrong tissue or at an inappropriate time, it then prevents the expression of
the target tumour suppressor gene. Reprinted by permission from Macmil-
lan Publishers Ltd: [Nature Reviews Cancer] (http://www.nature.com/nrc)
(Esquela-Kerscher and Slack, 2006), ©(2006). . . . . . . . . . . . . . . . 142
17
2.4. A good biomarker should have both high sensitivity and high specificity,
this minimises the numbers of false negatives and false positives respec-
tively A) High sensitivity, low specificity, (many samples passed the test
that should have failed it) B) Low sensitivity, high specificity (many sam-
ples failed the test that should have passed it). Red circle: false positive,
blue circle: false negative, open circle: true negative/true positive. Images
from Rmostell, reproduced from (Rmostell, 2011a) and (Rmostell, 2011b),
creative commons licence: CC0 1.0 . . . . . . . . . . . . . . . . . . . . . 171
2.5. Intertumoural and intratumoural heterogeneity develops over time as addi-
tional mutations are acquired by the cells within tumours and their metas-
tases. This leads to metastasis 1 being made up of a different population of
cells with different mutation profiles and characteristics to those of metas-
tasis 2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
3.1. Study design, global miRNA expression . . . . . . . . . . . . . . . . . . . 198
3.2. Study design, bioinformatics . . . . . . . . . . . . . . . . . . . . . . . . . 213
4.1. Proportion of patients with metastases stratified by tumour grade . . . . . 224
4.2. Distribution of second primary malignancies by GEP-NET grade (n=14).
Reprinted by permission from the Licensor: Springer Nature [World Jour-
nal of Surgery] [(Miller et al., 2014)], ©(2014). . . . . . . . . . . . . . . 231
4.3. Ki-67 IHC at different tumour sites for patient 7 (Table 4.11), showing an
increase in the number of Ki-67 positive cells between the primary tumour
(G1) and the metastases (G2). Positive nuclei are stained in brown, X10
magnification. A: SBNET, Ki-67 %: 1 %. B Lymph node metastasis, Ki-
67 %: 3 %. C: Liver metastasis, Ki-67 %: 8 %. Reprinted by permission
from the Licensor: Springer Nature [World Journal of Surgery] [(Miller
et al., 2014)], ©(2014). . . . . . . . . . . . . . . . . . . . . . . . . . . . 233
18
4.4. A: Number of patients graded as G1, G2 or G3 based on Ki-67 % of the
primary tumour. B: Number of patients with a change in grade based on
the Ki-67 % at another tumour site. . . . . . . . . . . . . . . . . . . . . . 234
4.5. The minimum and maximum Ki-67 % are shown for the 5 different sites
assessed within each liver lesion. Grey circle: increased grade, yellow cir-
cle: same grade. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
4.6. Minimum Ki-67 % is in blue, maximum Ki-67 % is in red. #: lesion,
where 2 metastatic lesions were available for Ki-67 % assessment. The
dotted line indicates the G1/G2 boundary. . . . . . . . . . . . . . . . . . 235
5.1. miRNA with a significant increase in expression in SBNET relative to
adjacent normal small bowel tissue. * FDR < 0.05, ** FDR < 0.001, ***
FDR < 0.0001. For enlarged x axis labels please refer to Table 5.2 . . . . 247
5.2. miRNA with a significant decrease in expression in SBNET relative to
adjacent normal small bowel tissue. * FDR < 0.05, ** FDR < 0.001, ***
FDR < 0.0001. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250
5.3. miRNA that had a significant increase/decrease in expression in lymph
node metastases compared to the primary tumour. * FDR < 0.05, ** FDR
< 0.001, *** FDR < 0.0001. . . . . . . . . . . . . . . . . . . . . . . . . . 254
5.4. miRNA that had a significant increase/decrease in expression in lymph
node metastases compared to normal lymph nodes. * FDR < 0.05, **
FDR < 0.001, *** FDR < 0.0001. Log2FC: ≥ 1.5 or ≤ −1.5 . . . . . . 256
5.5. MiRNA with increased expression in small bowel primary (SBP) tumours
versus adjacent normal small bowel (SB N). The relative expression of each
miRNA is shown for each sample. Results are shown from normalisation
against both endogenous control genes, RNU6-1 and SNORD44. Error bars
show the mean +/- standard error of the mean (SEM). The scale of the y
axis varies between plots. * p < 0.05, ** p < 0.001, *** p < 0.0001. . . . 258
19
5.6. MiRNA with decreased expression in small bowel primary (SBP) tumours
versus adjacent normal small bowel (SB N). The relative expression of each
miRNA is shown for each sample. Results are shown from normalisation
against both endogenous control genes, RNU6-1 and SNORD44. Error bars
show the mean +/- standard error of the mean (SEM). The scale of the y
axis varies between plots. * p < 0.05, ** p < 0.001, *** p < 0.0001. . . . 259
5.7. MiRNA with decreased expression in lymph node metastases (LNM) tissue
versus small bowel primary (SBP) tissue. The relative expression of each
miRNA is shown for each sample. Results are shown from normalisation
against both endogenous control genes, RNU6-1 and SNORD44. Error bars
show the mean +/- standard error of the mean (SEM). The scale of the y
axis varies between plots. * p < 0.05, ** p < 0.001, *** p < 0.0001. . . . 260
6.1. A: Venn diagram showing miRNA that were significantly increased in SB-
NET relative to “normal” small bowel tissue. B: Venn diagram showing
miRNA that were significantly decreased in SBNET relative to “normal”
small bowel tissue. All miRNA had a FDR < 0.05 and a log2FC ≥ 1.5 or
≤ −1.5. D1: dataset 1, D2: dataset 2. . . . . . . . . . . . . . . . . . . . 272
6.2. Venn diagram showing the miRNA with increased expression in tumour
tissue relative to normal tissue. a) Small bowel primary (SBP)/ small
bowel “normal”(SB N), comprised of the intersection of dataset 1 (D1)
and dataset 2 (D2), see Figure 6.1. b) Lymph node metastases(LNM)/
lymph node normal tissue (LN N) c) Liver metastases(LVM)/ Liver adja-
cent normal tissue (LV N). All miRNA had a FDR < 0.05 and a log2FC
≥ 1.5 or ≤ −1.5. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 278
20
6.3. Venn diagram showing the miRNA with reduced expression in tumour tis-
sue relative to normal tissue. a) Small bowel primary (SBP)/ small bowel
“normal”(SB N), comprised of the intersection of dataset 1 (D1) and dataset
2 (D2), see Figure 6.1. b) Lymph node metastases(LNM)/ lymph node
normal tissue (LN N) c) Liver metastases(LVM)/ Liver adjacent normal
tissue (LV N). All miRNA had a FDR < 0.05 and a log2FC ≥ 1.5 or ≤ −1.5.279
6.4. Venn diagram showing all significantly dysregulated miRNA in lymph node
metastases (LNM) and/or liver metastases (LVM) relative to expression
in the primary tumour (SBP) (FDR: < 0.05). Italic text indicates miRNA
that had higher expression levels in metastatic tissue relative to the SBP
(all other miRNA had lower expression in the metastatic tissue.) . . . . . 283
6.5. Heatmap showing the miRNA that had significantly decreased/increased
expression in metastatic tissue, lymph node metastases (LNM) or liver
metastases (LVM), relative their expression in small bowel primary tu-
mours (SBP). Log2FC values are shown for each miRNA. A log2FC of
cut off of ≥ 1.5 or ≤ −1.5 was used (FDR of < 0.05). *: expression
of these miRNA were significantly reduced in LNM and LVM however the
log2FC values for the LVM were not of a high enough magnitude to meet
the ≤ −1.5 cut off, these values were nevertheless included to enable com-
parison with the values for LNM/SBP. Blank spaces indicate that there
was no significant change in the expression of that particular miRNA. . . 284
21
7.1. Reduced expression of miR-1 and miR-143 (datasets 1 and 2) may lead to
a reduced negative regulation of the expression of the KRAS and BCL-2
oncogenes in lymph node metastases and therefore could be contributing to
disease progression. A: Complementary base pairing between miR-1/miR-
143 and KRAS mRNA, gene expression data showing a significant reduc-
tion in KRAS expression in lymph node metastases compared to SBNET
(dataset a, GSE27162). B: Complementary base pairing between miR-
1/miR-143 and BCL-2 mRNA, gene expression data showing a significant
reduction in BCL-2 expression in lymph node metastases compared to SB-
NET (dataset a, GSE27162).Error bars show the mean +/- standard devi-
ation (* p < 0.05, ** p < 0.01, *** p < 0.001). Reprinted by permission,
©[2016] [BioScientifica Ltd.], (Endocrine-Related Cancer) (Miller et al.,
2016). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 326
7.2. Reduced levels of miR-1 and miR-143 in SBNET metastases (datasets 1
and 2) may result in reduced negative regulation of NUAK2 and FOSB
expression in SBNET metastases which could promote disease progres-
sion. A: Complementary base pairing between miR-1 and FOSB mRNA.
B: Complementary base pairing between miR-143 and FOSB mRNA. C:
Gene expression data showing a significant reduction in FOSB expres-
sion in lymph node and liver metastases compared to SBNET (dataset a,
GSE27162). D: Complementary base pairing between miR-1 and NUAK2
mRNA. Gene expression data showing a significant reduction in NUAK2
expression in lymph node metastases compared to SBNET (dataset a, GSE27162).
Error bars show the mean +/- standard deviation (* p < 0.05, ** p < 0.01,
*** p < 0.001). Reprinted by permission, ©[2016] [BioScientifica Ltd.],
(Endocrine-Related Cancer) (Miller et al., 2016). . . . . . . . . . . . . . 328
22
7.3. Reduced expression of miR-1 (datasets 1 and 2) may lead to a reduced
negative regulation of the expression of growth factors HGF and VEGFA
in SBNET metastases and could therefore could be contributing to dis-
ease progression. A: Complementary base pairing between miR-1 and HGF
mRNA, gene expression data showing a significant reduction in HGF ex-
pression in lymph node and liver metastases compared to SBNET (dataset
a, GSE27162). B: Complementary base pairing between miR-1 and VEGFA
mRNA, gene expression data showing a significant reduction in VEGFA ex-
pression in lymph node and liver metastases compared to SBNET (dataset
a, GSE27162). Error bars show the mean +/- standard deviation (* p
< 0.05, ** p < 0.01, *** p < 0.001). Reprinted by permission, ©[2016]
[BioScientifica Ltd.], (Endocrine-Related Cancer) (Miller et al., 2016). . 329
F.1. CC BY-NC (Creative Commons Attribution NonCommercial) for Table
2.1 and Table 2.4, for details see Table F.1 . . . . . . . . . . . . . . . . . 444
F.2. CC BY-NC (Creative Commons Attribution NonCommercial) for Table
2.1 and Table 2.2, for details see Table F.1 . . . . . . . . . . . . . . . . . 445
F.3. CC BY 2.0 (Creative Commons Attribution 2.0 Generic) for Table 2.2, for
details see Table F.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 445
F.4. Permissions for Figure 2.3, for details see Table F.1 . . . . . . . . . . . . 446
F.5. Permissions for Figure 2.4, for details see Table F.1 . . . . . . . . . . . . 447
F.6. Permissions for Figure 2.4, for details see Table F.1 . . . . . . . . . . . . 447
F.7. Reprint permission from Springer Nature, [World Journal of Surgery],
(Miller et al., 2014). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 448
F.8. Reprint permission from BioScientifica Ltd., [Endocrine-Related Cancer],
(Miller et al., 2016). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 449
23
List of Tables
2.1. GEP-NET grading according to ENETS guidelines, table reproduced from
Rindi et al. (2006) and Rindi et al. (2007), creative commons licence: CC
BY-NC. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
2.2. Disease staging SBNET, table reproduced from (Rindi et al., 2007), cre-
ative commons licence: CC BY-NC . . . . . . . . . . . . . . . . . . . . . 76
2.4. Disease staging PNET, table reproduced from (Rindi et al., 2006), creative
commons licence: CC BY-NC . . . . . . . . . . . . . . . . . . . . . . . . 77
2.6. Functioning tumours . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
2.7. Neuroendocrine cells of the GI tract and pancreas . . . . . . . . . . . . . 119
2.8. MiRNA expression studies in primary tumours and metastases of SBNET
patients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
3.1. Number of samples for miRNA quantification . . . . . . . . . . . . . . . 196
3.2. Samples dataset 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
3.3. Samples dataset 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
3.4. Antibodies used for Ki-67 IHC . . . . . . . . . . . . . . . . . . . . . . . . 210
3.5. Gene expression datasets for bioinformatics analysis . . . . . . . . . . . . 215
3.6. Comparison groups for gene expression datasets . . . . . . . . . . . . . . 216
4.1. Primary site of GEP-NET . . . . . . . . . . . . . . . . . . . . . . . . . . 222
25
4.2. Patient characteristics stratified by grade. Reprinted by permission from
the Licensor: Springer Nature [World Journal of Surgery] [(Miller et al.,
2014)], ©(2014). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
4.3. Summary of tumour stage . . . . . . . . . . . . . . . . . . . . . . . . . . 238
4.4. Location of distant metastases . . . . . . . . . . . . . . . . . . . . . . . . 239
4.5. Stage SBNET and PNET . . . . . . . . . . . . . . . . . . . . . . . . . . 239
4.6. Functioning syndromes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
4.7. SBNET . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240
4.8. PNET . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240
4.9. Second primary malignancies . . . . . . . . . . . . . . . . . . . . . . . . . 240
4.10. Primary sites in heterogeneity study . . . . . . . . . . . . . . . . . . . . 241
4.11. Ki-67 % at different disease sites, (patients 1-17 only, reprinted by per-
mission from the Licensor: Springer Nature [World Journal of Surgery]
[(Miller et al., 2014)], ©(2014)). . . . . . . . . . . . . . . . . . . . . . . 242
4.12. Liver metastases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
5.1. Samples, dataset 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
5.2. Enlarged x axis labels for Figure 5.1 . . . . . . . . . . . . . . . . . . . . 248
5.3. SBNET miRNA profile, most upregulated miRNA . . . . . . . . . . . . . 251
5.4. SBNET miRNA profile, most downregulated miRNA . . . . . . . . . . . 253
5.5. Significantly dysregulated miRNA in lymph node metastases versus SBNET253
6.1. Samples, dataset 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 266
6.2. SBNET miRNA profile, most dysregulated miRNA . . . . . . . . . . . . 267
6.3. MiRNA dysregulated in SBNET . . . . . . . . . . . . . . . . . . . . . . . 269
6.4. Dataset 2 profiling results for candidate miRNA . . . . . . . . . . . . . . 275
6.5. Significantly dysregulated miRNA in liver metastases . . . . . . . . . . . 290
6.6. Liver metastases, most dysregulated miRNA . . . . . . . . . . . . . . . . 291
26
6.7. Significantly dysregulated miRNA in Lymph node metastases . . . . . . 291
6.8. Lymph node metastases, most dysregulated miRNA . . . . . . . . . . . . 292
6.9. MiRNA dysregulated in both liver and lymph node metastases . . . . . . 292
7.1. Gene expression datasets for bioinformatics . . . . . . . . . . . . . . . . . 297
7.2. Potential gene targets of the candidate miRNA . . . . . . . . . . . . . . 298
7.3. Predicted gene targets of the candidate miRNA that were dysregulated in
all 3 gene expression datasets (b, c, d) . . . . . . . . . . . . . . . . . . . 301
7.4. Pathway analysis of downregulated genes in SBNET that are predicted
targets of the upregulated candidate miRNA . . . . . . . . . . . . . . . . 305
7.5. Significantly enriched gene ontology terms for upregulated genes in lymph
node metastases, predicted gene targets of miR-1 . . . . . . . . . . . . . 307
7.6. Top 30 enriched gene ontology terms for upregulated genes in lymph node
metastases, predicted gene targets of miR-1 . . . . . . . . . . . . . . . . 309
7.7. Enriched KEGG pathway terms for upregulated genes in lymph node
metastases, predicted gene targets of miR-1 . . . . . . . . . . . . . . . . 313
7.8. Significantly enriched gene ontology terms for upregulated genes lymph
node metastases, predicted gene targets of miR-143 . . . . . . . . . . . . 317
7.9. Enriched gene ontology terms for upregulated genes lymph node metas-
tases, predicted gene targets of miR-143 . . . . . . . . . . . . . . . . . . 319
7.10. Enriched KEGG pathway terms for upregulated genes in lymph node
metastases, predicted gene targets miR-143 . . . . . . . . . . . . . . . . . 323
8.1. Studies involving miRNA quantification in primary tumours and metas-
tases of SBNET patients . . . . . . . . . . . . . . . . . . . . . . . . . . . 338
A.1. Sample ID of FFPE tissue available for miRNA analysis (Dataset 1) . . . 422
A.2. Clinical details miRNA Dataset 1 . . . . . . . . . . . . . . . . . . . . . . 423
27
B.1. miRNA primers for qPCR . . . . . . . . . . . . . . . . . . . . . . . . . . 424
C.1. Dataset 1, RNA extractions for quantification NanoString . . . . . . . . . 425
C.2. Dataset 1, RNA extractions for quantification qPCR . . . . . . . . . . . 426
C.3. Dataset 2, RNA extractions for quantification NanoString . . . . . . . . . 427
D.1. Significantly dysregulated miRNA in lymph node metastases versus normal
tissue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 429
D.2. miRNA that were significantly dysregulated in SBNET relative to “nor-
mal” small bowel tissue . . . . . . . . . . . . . . . . . . . . . . . . . . . . 432
E.1. Top 10 enriched gene ontology terms for the predicted gene targets of
miR-7-5p, miR-204-5p and miR-375 . . . . . . . . . . . . . . . . . . . . . 437
F.1. Permissions for reprints . . . . . . . . . . . . . . . . . . . . . . . . . . . . 442
28
1. Introduction
Neuroendocrine tumours (NET) are tumours that arise in the neuroendocrine cells found
throughout the body. These neuroendocrine cells act as the interface between the en-
docrine and nervous systems by secreting hormones in response to neuronal input. Neu-
roendocrine tumours of the gastroenteropancreatic (GEP) system (GEP-NET) arise from
GEP neuroendocrine cells such as insulin secreting pancreatic beta cells and serotonin
secreting enterochromaffin cells.
GEP-NET are rare tumours, with an incidence in England of 1.32-1.33 per 100,000
people in 2006 (national cancer registry for England) (Ellis et al., 2010). The incidence
of GEP-NET is increasing globally this may be due to advances in imaging techniques and
better awareness of the condition, although autopsy studies show that some NET remain
undetected during the patient’s lifetime. At the start of 2008 there were an estimated
100,003 patients living with a NET diagnosis across the 27 European Union member
states, of these 63,691 had a well differentiated GEP-NET (Van Der Zwan et al., 2013).
GEP-NET can be locally advanced at diagnosis depending on the primary tumour site
and often metastasise to the liver, with considerably worse outcomes for patients. While
there are many treatment modalities available to these patients from surgery and control
of hormonal symptoms to targeted therapy, high level evidence is largely lacking due to
the inability to properly stratify patients to identify those who would most benefit from
these different approaches. Historically since these tumours are rare there has been less
funding for research in this area compared to other tumour types and this has led to a
29
lack of detailed knowledge about the process of NET tumourigenesis and weaker evidence
base for patient treatment.
GEP-NET have differing biological characteristics and clinical course depending on
the site of the primary tumour. In combination with the comparatively small overall
patient numbers this leads to challenges in designing effective experiments and clinical
trials. In addition there is both intertumoural and intratumoural heterogeneity within
the same patient leading to further challenges for understanding the biology of these
tumours. Small bowel NET (SBNET) arise from the enterochromaffin cells of the small
bowel while Pancreatic NET (PNET) arise from endocrine cells in the pancreatic islets
of Langerhans. The majority of GEP-NET are sporadic, however they can be associated
with hereditary mutations in the MEN1 gene resulting in MEN1 syndrome with multiple
tumours occurring in the pancreas (PNET) and other tissues.
SBNET have been associated with sporadic mutations in cyclin dependent kinase in-
hibitor 1B (CDKN1B) in 8 % of patients, while sporadic mutations in chromatin remod-
eling genes are instead common in PNET, present in 40 % of patients (Francis et al.,
2013; Marinoni et al., 2014). Mutations in the tumour suppressors TP53 and RB tran-
scriptional corepressor 1 (RB1) remain rare in low grade GEP-NET however they are a
common occurrence in high grade tumours (Garcia-Carbonero et al., 2016).
Biomarkers are biological markers which can be measured to provide information about
biological processes, including those that occur in a disease state or during treatment.
Advances in methods for the isolation and quantification of small amounts of biological
material and reductions in price are making it possible to objectively measure many
more different biological molecules in a clinical setting. Biomarkers can provide useful
information to the clinician about the likelihood of disease progression/disease recurrance,
patient survival and the presence/absence of a more aggressive tumour subtype. In
addition, increasing numbers of novel therapeutics are being developed in parallel with
companion biomarkers so that the patient population that would most benefit from these
30
therapies can be more precisely identified and treatment efficacy can be more closely
monitored.
The two main molecular biomarkers used in NET are Chromogranin A (CgA) which
is used to support a histopathological NET diagnosis and the Ki-67 proliferation marker
which is used for tumour grading. The Ki-67 proliferative index is used to categorise
GEP-NET patients as having either low grade tumours (Grade 1/Grade 2) or high grade
tumours (Grade 3) based on proliferation levels. Low grade tumours have low proliferation
levels while high grade tumours have high proliferation levels which are associated with a
worse prognosis. There are limitations however with this approach since liver metastases
are frequently present in patients with low grade GEP-NET such as those which arise
from the small bowel despite them having a Ki-67 % of ≤ 2 %. The presence of liver
metastases is associated with worse survival in GEP-NET patients. Recent consensus
conferences have called for novel biomarkers to be developed for the further stratification
of GEP-NET patients, particularly those with the more common low grade GEP-NET,
based on clinical behaviour and disease pathology (Frilling et al., 2014; Oberg et al.,
2015).
MicroRNA (miRNA) are short non-coding RNA molecules, approximately 22 nu-
cleotides long, which regulate gene expression by binding to complementary sequences
on their target mRNA. MiRNA are frequently dysregulated in cancer acting as either
tumour suppressors or in a role similar to that of oncogenes (oncomir). They have also
been developed as cancer biomarkers for example in pancreatic adenocarcinoma (PNAC),
where increased levels of miR-1290 in the serum could differentiate PNAC patients from
those with pancreatitis and PNET with a higher level of accuracy than the established
test, CA19–9 (Li et al., 2013a). The role of miRNA in breast cancer has been widely in-
vestigated with numerous circulating miRNA being developed as potential future breast
cancer biomarkers including the use of miR-195, miR-195, let-7a and miR-155 as a di-
agnostic biomarkers and miR-10b-5p as a prognostic biomarker (Hamam et al., 2017).
31
There is limited information on the potential role of miRNA in tumourigenesis in GEP-
NET. There have been quite a few miRNA studies PNET however, less is known about
the potential role of miRNA in SBNET.
Patient treatment in many disease areas is moving towards a more personalised medicine
approach, with the ability to treat patients based on their particular genetic and epige-
netic subtype. At the same time, methods for the isolation and analysis of circulating
microRNA, DNA and tumour cells are becoming more advanced, raising the possibility in
the future for the serial non-invasive monitoring of patients with GEP-NET via a serum
sample.
Novel molecular biomarkers are needed which can stratify GEP-NET patients into
clinically relevant subtypes to enable better prediction of disease course, early detection
of disease progression and treatment response monitoring. The development of novel
GEP-NET biomarkers would provide clinicians with more detailed information about
the disease pathology on which to base their treatment decisions, which could lead to
real patient benefits in terms of survival and quality of life.
1.1. Aim
To identify new potential prognostic biomarkers for use in GEP-NET.
1.2. Research objectives
The research aim of this thesis is broken down into five principal objectives:
1) Investigate the limitations of the existing prognostic biomarker in GEP-
NET.
2) Experimentally determine a global miRNA profile of SBNET.
3) Verify the reproducibility and robustness of the SBNET miRNA profile.
32
4) Identify miRNA associated with disease progression in SBNET.
5) Identify the most promising potential miRNA biomarkers for use in SB-
NET.
The first objective is to determine the efficacy of the current GEP-NET prognostic
biomarker in order to identify if additional prognostic biomarkers would be of clinical
utility in GEP-NET. The second objective is to identify a global profile of miRNA that
are dysregulated in SBNET. This would enable particular miRNA to be identified that
could be used as novel prognostic biomarkers in SBNET patients. The third objective is
involved in determining if the SBNET profile identified is reproducible. This would be
a necessary feature of any possible future biomarker. The fourth objective is to discover
miRNA that could be biomarkers of disease progression in SBNET. The final research
objective is to narrow down the possible future miRNA biomarkers identified to select
those with the most potential for use in the stratification of SBNET patients.
1.3. Contribution to knowledge
This thesis contributes three principal findings to the academic literature. The first of
these contributions is the demonstration that there is no level of Ki-67 % at which a
GEP-NET patient can be considered to be ‘safe’ from liver metastases. A large scale
study of 161 GEP-NET patients in which Ki-67 % was analysed with respect to disease
stage revealed that 28 % of the patients with G1 tumours had stage IV disease despite
having a Ki-67 % of ≤ 2 %. When further analysis was carried out, this effect was found
to be even more striking for SBNET patients. Stage IV disease was present in 54 % of the
SBNET patients with a G1 tumour and 100 % of the SBNET patients with a G2 tumour.
These results demonstrate an unmet clinical need for novel biomarkers for use alongside
Ki-67 % in patients with low grade tumours, to enable further patient stratification by
identifying patients with more aggressive disease subtypes.
33
The second contribution is that the work presented in this thesis represents by far
the largest and most comprehensive study of miRNA expression in SBNET and their
metastases to date. This study involved the quantification of 800 miRNA in 90 different
tissue samples taken from 37 patients with a SBNET. Novel miRNA were identified
that had not been previously associated with tumourigenesis and disease progression in
SBNET. The identification of these miRNA represents a compelling starting point for
further work, both to better understand their function in SBNET and to develop novel
biomarkers for use in SBNET patients.
The third contribution made by this thesis is the discovery and thorough validation of
6 miRNA related to disease progression in SBNET that have the potential to be used
as novel biomarkers in the future. To ensure that the results were reproducible, the
miRNA were extensively validated as being dysregulated in SBNET metastases in two
independent populations of SBNET patients. Further analysis led to the identification
of miR-1 and miR-143-3p as the most promising potential candidates for use as novel
prognostic biomarkers in patients with SBNET.
1.4. Document outline
This chapter has served as an introduction and set out the overall aim of the thesis, the
research objectives and the contribution to knowledge.
The next chapter, chapter 2, consists of a thorough review of the academic literature.
Firstly the epidemiology, classification and treatment of GEP-NET is examined, followed
by the different kinds of neuroendocrine cells these tumours can arise in. The literature on
miRNA and their role as oncomir and tumour suppressor miRNA in cancer and in GEP-
NET is explored. This is followed by a discussion of existing biomarkers and potential
future biomarkers. The gaps in the existing literature are identified and these form the
basis for the thesis aims and objectives.
34
Chapter 3 gives full details of the methodological approaches used in this work. In
chapter 4, an investigation of Ki-67 % with respect to disease stage in 161 patients
with GEP-NET is presented, this addresses the first research objective of the thesis. The
second research objective is addressed in chapter 5 in which results are presented from the
global miRNA expression profiling of matched tissue from 15 SBNET patients. Chapter
6 serves to address the third research objective of the thesis by presenting results from
the global miRNA expression profiling of tissue from an independent group of 22 SBNET
patients treated at a different institution. The fourth research objective is also addressed
in chapter 6 with the identification of miRNA associated with SBNET liver metastases.
In chapter 7, results are presented from the investigation of the potential novel miRNA
biomarkers identified in chapters 5 and 6 using bioinformatics. This serves to address the
final research objective of the thesis by identifying the most promising potential miRNA
biomarkers for use in SBNET patients so that these miRNA can be the focus of future
studies in this area.
Chapter 8 outlines the principle conclusions and describes further work to build upon
findings of the thesis.
35
2. Literature Review
2.1. Epidemiology
2.1.1. Incidence
The number of patients diagnosed with a well differentiated GEP-NET living in the EU
was estimated to be 64,762 at the start of 2008, in a study representing 65 % of EU
patients with a NET (n=100,003) (Van Der Zwan et al., 2013). Based on regression
analysis from historical epidemiological data, the projected incidence of GEP-NET in
the USA was estimated to be 10.9 and the prevalence 65 per 100,000 population in 2015
(Frilling et al., 2012; Yao et al., 2008; Modlin et al., 2008). In the UK, around 3000
new patients could be expected to be diagnosed with a GEP-NET each year (Davies and
Weickert, 2016; Yao et al., 2008).
Incidence increasing globally
Although GEP-NET remain rare entities, their incidence is rising globally. A study of
10,170 gastrointestinal NET patients from the National Cancer Registry for England
showed that there was a 3.8-4.9 fold overall rise in the incidence rate over 36 years (Ellis
et al., 2010). Ellis et al showed that the incidence rate in women increased from 0.35 to
1.33 between 1971 and 2006, while in men it increased from 0.27 to 1.32 (per 100,000
population per year). This increase in incidence was present across all primary sites.
Incidence has also increased in other European countries. A study in Germany for
36
example, showed that GEP-NET incidence increased around 5 fold in data from two
databases containing 2,821 GEP-NET patients in total (former East German National
Cancer Registry and the Joint Cancer Registry) (Scherubl et al., 2013). The authors
noted that the overall incidence rate increased from 0.57 to 2.38 in women and from 0.31
to 2.27 in men between the years of 1976 and 2006 (per 100,000 inhabitants per year).
The authors found that incidence rates increased over the time period studied across all
primary sites except for appendix NET in women, where incidence rates only increased
very slightly from 0.35 to 0.39.
In the USA, a study of the Surveillance, Epidemiology and End Results (SEER) registry
showed a 3.65 fold increase in the age adjusted incidence of GEP-NET from 1973-2007
(29,664 GEP-NET were included) (Lawrence et al., 2011; Modlin et al., 2003; Modlin
et al., 2008; Frilling et al., 2012; Fraenkel et al., 2014). The authors found that the
incidence rate increased from 1.0 in 1973 to 3.65 in 2007 (per 100,000 people, per year).
This incidence increase held true across all the GEP-NET primary sites.
The increase in GEP-NET incidence seems to be a global phenomenon observed in
additional National Cancer Registries and regional series, including in Norway, Sweden,
France (Burgundy), Canada (Ontario) and South Korea (Hauso et al., 2008; Sandvik
et al., 2016; Hemminki et al., 2001; Landerholm et al., 2010; Lepage et al., 2004; Lepage
et al., 2006; Fraenkel et al., 2014; Hallet et al., 2015; Cho et al., 2012).
Autopsy studies
Autopsy studies suggest that despite the observed increase in incidence around the world,
a large proportion of GEP-NET are failing to be diagnosed during the lifetime of the
patient. In a study of all autopsies carried out in Malmo, Sweden between 1958 and
1969, the incidence of digestive carcinoid tumours was around 8.4 per 100,000, which
was 7 times higher than in the Swedish National Cancer Registry at the time (Berge and
Linell, 1976). Small intestinal NET represented 73 % of the gastrointestinal NET in the
37
study with an annual incidence of 5.5 per 100,000. This figure is much higher than other
series where autopsy cases were not included. Around 90 % of the NET found in the
study were discovered incidentally during autopsy. In another study, pancreatic sections
of 800 autopsy patients in Japan were examined and pancreatic endocrine tumours were
identified in 2.5 % of the autopsy cases (Kimura et al., 1991).
This discrepancy could be due to the nature of GEP-NET. They are often asymp-
tomatic or they have non-specific symptoms. It is often the case that symptoms only
develop due to tumour mass effect when the tumour grows large enough. So tumours
may not be detected during the lifetime of patients if they die before developing symp-
toms. Many of the GEP-NET that are diagnosed are still discovered incidentally during
imaging for other indications. The autopsy findings suggest that large numbers of these
tumours will remain clinically silent during the patient’s lifetime. This represents a chal-
lenge for diagnosis and means that GEP-NET are frequently diagnosed at an advanced
stage, at which point the prognosis is poorer. The length of time between the onset of
symptoms and a GEP-NET diagnosis can be around 5-7 years which reduces the likeli-
hood that all patients with these tumours will be detected in time (Modlin et al., 2008;
Dıez et al., 2013).
Additional epidemiological studies would be helpful to identify in more detail the true
incidence. It would also be useful to have more autopsy studies of more recent patient
cohorts. These could determine if clinically silent GEP-NET levels remain similar to
those observed in these earlier studies or if numbers are decreasing as imaging techniques
become more advanced and there is better awareness of these tumours. Although some
of the tumours found in the autopsy series may be of little clinical relevance it should
be noted that very small GEP-NET lesions can still be metastatic. This is particularly
true in SBNET (NET of the jejunum/ileum) were 30 % of small, 1 cm, lesions can have
lymph node metastases at diagnosis and these metastases can also be identified in some
patients with 0.5 cm lesions (Scherubl et al., 2010; Eriksson et al., 2008). So these missed
38
tumours that are found at autopsy should not necessarily be considered to be benign in
nature.
Limitations
There are various limitations with these types of epidemiological studies. The national
cancer databases used are unlikely to represent the whole population of patients with a
GEP-NET diagnosis and may not be well set up for capturing GEP-NET patients. The
inevitable complexity involved in such a national undertaking results in incomplete data
collection at some sites with patients being missed. In addition, differences in methodol-
ogy for the data collection can impact on data accuracy and the amount of detail provided
for each patient. This may mean that some of the data must be excluded or will be of
limited usefulness. The available data is frequently scaled up to represent a larger popu-
lation/geographical region. In this case the patient demographics and environment in the
whole population may not be a close match to those in the original study, particularly
for the studies with small patient numbers or a small catchment area for inclusion.
These factors also present challenges when trying to compare the incidence figures be-
tween countries or continents. Additionally there are some quite large differences between
the national registries used, for example the SEER registry contains only malignant NET
which means that incidence figures may be conservative since localised tumours are likely
to have been excluded (Yao et al., 2008). Classification changes over the last 30-40 years
have also led to challenges for large scale epidemiological studies, since the categorisa-
tion and management of GEP-NET has changed dramatically in this time (see section
2.2). Therefore care must be taken to ensure that the same types of tumour are being
compared and it may be necessary to reclassify older cases based on new WHO/ENETS
guidelines, if there is sufficient clinical information available to do this.
39
Reasons for increasing incidence
The precise factors behind the global increase in the incidence of GEP-NET are less
clear. The findings of a recent Canadian study suggest that at least some of the increasing
incidence in NET could be caused by tumours being discovered at an earlier stage (Hallet
et al., 2015). Hallet et al, found that the percentage of patients presenting with NET
metastases decreased from 29 % in 1994 in to 13 % in 2009, while at the same time the
incidence of all NET was increasing. They found that the incidence of NET that were
metastatic at presentation stayed stable between 1994 and 2009, suggesting that more
tumours were being discovered at an early stage (the overall NET incidence increased
while the proportion of metastatic presenting NET decreased).
The increasing incidence of GEP-NET could be the result of improved availability
and quality of imaging techniques, such as computed tomography and gastrointestinal
endoscopy and a greater awareness of the condition, rather than reflecting a true increase
in the incidence of these tumours. Moreover, autopsy series suggest that large numbers
of GEP-NET patients are still not being diagnosed within their lifetime. Some of these
patients would have had very early stage tumours with little disease burden however
some patients are invariably being missed. This could be problematic in countries like
the UK and other European countries if the lower incidence figures observed reflect missed
opportunities to identify patients when compared to series from the USA and Canada
(Hallet et al., 2015). More recent autopsy studies would be needed however, to see if
this was the case. Alternatively the smaller incidence figures in Europe could be due to
a myriad of other factors such as differences in the environment or in the way healthcare
is organised in different countries.
As imaging methods continue to become more advanced, more GEP-NET may be
identified as incidental findings. This has the potential to provide substantial benefits for
patients by enabling more rapid diagnosis and treatment, resulting in better outcomes
as a greater proportion GEP-NET are discovered at an earlier stage. This may enable
40
the incidence rates to start to approach the numbers seen in the autopsy studies. This
will present further challenges for clinicians, with the need to predict which of these early
stage tumours will go on to develop metastases and which will remain localised.
Summary
These factors present challenges for obtaining a true picture of the epidemiology of GEP-
NET and demonstrate the importance of having standardised classification systems and
data collection methodologies so that the data collected is broadly comparable globally.
Since GEP-NET are rare, global consensus is particularly important as it may be difficult
to generate large enough datasets of these patients within a single country to understand
more the complex epidemiological factors that may be at play.
Nevertheless, while there may be limitations involved in the comparisons of individual
incidence rates between countries, the fact that so many different cancer registries showed
that GEP-NET incidence is increasing suggests that this can be considered to be a robust
phenomenon.
2.1.2. Survival
Despite the increasing incidence of GEP-NET, there have been only modest increases
in patient survival. Factors that can affect patient survival include the location of the
primary tumour, disease stage, tumour grade, surgical resection and age at diagnosis.
For more details on tumour grade and survival please see section 2.6.1.
In historical, population based data, from 10,878 NET patients in the USA (SEER
registry: 1973-1999) the overall 5 year survival rate for NET was 67.2 % (Modlin et
al., 2003). This figure was even lower, at 56.2 %, when bronchopulmonary NET were
excluded from the analysis. Distant metastases to sites such as the liver were found to
be associated with much worse survival, 38.5 %, compared to 71.7 % for NET with only
regional metastases.
41
Studies in France (1976-1999, n = 229) and the UK (1986-1999, n = 4104), also of ma-
lignant GEP-NET, showed lower 5 year survival rates of 50.4 % and 45.9 % respectively
(Lepage et al., 2004; Lepage et al., 2007). Data from population based registries from
12 European countries (1985-1994, n = 3,715) of malignant GEP-NET found that these
tumours had a 5 year survival rate of 47.5 % (Lepage et al., 2010). Geographical differ-
ences were revealed between Northern Europe, Western Continental Europe, the UK and
Eastern Europe, with 5 year relative survival rates of 60.3 %, 53.6 %, 42.5 % and 37.6
% respectively (p < 0.001). Another European population based study including 20,000
NET patients from 76 cancer registries (1978 to 2002) found an overall 5 year survival of
50 % for NET (Van Der Zwan et al., 2013).
Analysis including data up to 2007 from 29,664 GEP-NET (SEER registry: 1973-2007),
showed that 5 year survival rates were highly dependent on primary site ranging from
37.6 % for PNET to 88.5 % for rectal NET (Lawrence et al., 2011). The 5 year survival
for other tumour sites were as follows: small intestine (68.1 %), colon (54.6 %), stomach
(64.1 %) and appendix (81.3 %).
Functionality can also have an impact on survival, with insulinomas having a 5 year
survival rate of 97 % due to them very rarely having metastases in contrast to non-
functioning PNET which have much worse 5 year survival rate of 43 % (for more details
on functioning syndromes see section 2.2.5) (Oberg and Barbro, 2005; Falconi et al.,
2012). Increased age at diagnosis is associated with worse survival (Ahmed et al., 2009;
Lepage et al., 2007; Yao et al., 2008; Falconi et al., 2012; Lepage et al., 2010; Landerholm
et al., 2011).
Certain mutations have been associated with worse survival, these include DAXX/A-
TRX mutations. Loss of DAXX and ATRX is associated with chromosomal instability
and reduced survival in PNET (Minnetti and Grossman, 2016; Marinoni et al., 2014).
Low income and rural residency may also be associated with worse survival (see section
2.1.3) (Hallet et al., 2015).
42
Diagnosis of GEP-NET is typically delayed by 5-7 years (Modlin et al., 2008; Dıez
et al., 2013). This is probably due to a lack of awareness about NET amongst the general
public and primary care providers and the lack of specific symptoms in non functioning
tumours. Many GEP-NET are asymptomatic at early disease stages with patients not
being identified until their tumours are already at an advanced stage, when they develop
symptoms associated with mass effects such as abdominal pain.
This diagnosis delay means that a high proportion of GEP-NET arising from certain
primary sites, for example SBNET and PNET, are locally advanced at the time of diag-
nosis and many also have distant metastases. Distant metastases lead to worse outcomes
since curative resection becomes much less achievable. The majority of SBNET patients
have lymph node metastases at diagnosis and liver metastases are also common, occurring
in 60–80 % of patients (Norlen et al., 2012; Frilling et al., 2012; Clift et al., 2016).
Historically GEP-NET were treated conservatively since their behaviour and metastatic
potential was less well understood. It has now been shown that even small, low grade,
GEP-NET have the potential to be malignant and this is reflected in the latest ENETS
guidelines which recommend that these lesions should be removed surgically or endoscop-
ically with regular follow up to check for disease recurrence (O’Toole et al., 2016; Niederle
et al., 2016; Falconi et al., 2016; Ramage et al., 2016). In addition to this, historically
GEP-NET patients frequently died as a result of complications from poor management of
the hormone hypersecretion associated with GEP-NET functioning syndromes (see sec-
tion 2.2.5), however survival improved with the introduction of somatostatin analogues
(see section 2.3.1).
These changes in patient management could explain the modest increases in patient
survival over time. Analysis of the SEER registry showed a 23.6 % increase in GEP-NET
survival rates between periods 1973–1974 and 2001–2002 (relative change) (Lawrence et
al., 2011). A population based study of malignant GEP-NET from 1986 to 1999 in the
UK (n = 4104) however, showed no increase in survival over the time period studied
43
(Lepage et al., 2007). These differences may reflect true population differences or they
could be due to differences in the patients included in each database (see section 2.1.1,
limitations).
Recent data from specialist centres suggests that survival rates may have improved
relative to historical cohorts. For SBNET 5 year survival ranged from 89 % in one
study (1998-2015, n=84) and 88.9 % in another (1984-2008, n=270), to 67 % (1985-2010,
n=603) (Clift et al., 2016; Jann et al., 2011; Norlen et al., 2012). There was significantly
reduced 5 year survival in SBNET patients with liver metastases compared to those with
no liver metastases at 84 % and 100 % respectively (Clift et al., 2016). Improved survival
rates may be the result of more treatments becoming available in recent years, however
results from these more recent studies could be affected by selection bias. Population
based survival studies including patient cohorts from the last decade would be of interest
to shed light on this.
The presence of distant metastases has a large impact on survival. Data from a histor-
ical, population based, study of NET in the USA (n=35,618), showed that for SBNET
the median survival of patients with localised disease was 111 months compared to just
56 months for those with distant metastases (regional disease: 105 months) (Yao et al.,
2008). The difference in median survival was even more dramatic for PNET, 136 months
in patients with localised disease compared to only 24 months for patients with distant
metastases (regional disease: 77 months).
Overall 5 year survival post resection for liver metastases in GEP-NET is around 85–94
%, however less than 50 % of patients have 5 years of disease free survival due to early
disease recurrence (Frilling et al., 2012). In PNET, 5 year survival after liver resection is
47-76 % versus 30-40 % for untreated patients but up to 76 % of patients have tumour
recurrence (Falconi et al., 2012; Frilling et al., 2012; Ahmed et al., 2009). In a recent study
the 5 year survival was found to be 100 % in SBNET patients without liver metastases
compared to 84 % in patients with liver metastases (Clift et al., 2016).
44
More GEP-NET are being identified incidentally at an earlier disease stage due to
advances in imaging techniques and their more widespread use. This is likely to increase
patient survival in the future (Falconi et al., 2016). Imaging techniques such as 68Ga
DOTA-PET/CT enable smaller GEP-NET lesions and metastases to be identified leading
to more accurate disease staging information and primary tumour identification. This
has been shown to have the potential change patient management decisions in 20-30 % of
SBNET patients, which could lead to an improvements in survival (Niederle et al., 2016).
2.1.3. Risk Factors
Since GEP-NET are quite rare and there is less funding in this area when compared to
other cancers, there have been a historic lack of studies on potential risk factors for GEP-
NET development. In particular, there are very few studies in the literature investigating
potential environmental factors such as tobacco smoking, alcohol consumption and obe-
sity which may act as risk factors for tumourigenesis. Moreover, there are conflicting
findings from the handful of studies that have been done in these areas. In contrast, ad-
vanced age, inherited NET syndromes and gender have been investigated in some detail
as risk factors for GEP-NET. Studies into potential protective factors in GEP-NET are
even more scarce with only one study which suggested that Aspirin may be a significant
protective factor for SBNET (odds ratio: 0.20, 95 % confidence interval: 0.06 – 0.65)
(Rinzivillo et al., 2016).
The small numbers of patients with GEP-NET limits the scope of investigations into
risk factors. It also presents difficulties in investigating more subtle potential associations
due to the difficulties in obtaining sufficient patient numbers to be able to observe any
association, while attempting to control for potential confounding factors.
45
Family History of NET syndromes
NET usually represent a sporadic disease but familial syndromes associated with the
formation of multiple NET can occur due to inherited mutations in certain genes.
+++MEN1 The most common inherited mutations are in the multiple neoplasia type
1 gene (MEN1) and lead to MEN1 syndrome. MEN1 syndrome is an autosomal dominant
condition which is caused by inherited loss of function mutations in MEN1. MEN1
encodes the protein menin which has no homology to any other known proteins and is
thought to be involved in the regulation of DNA repair (Schernthaner-Reiter et al., 2016;
Scacheri et al., 2006). More rarely, MEN1 syndrome can occur sporadically due to de
novo mutations in MEN1 acquired during a patients lifetime.
Hundreds of different MEN1 mutations have been identified in patients with MEN1
syndrome, most of which result in the absence of menin or truncation of the protein
because of frameshift deletions or insertions (Schernthaner-Reiter et al., 2016). In around
10 % of MEN1 patients no mutation in MEN1 can be identified, this may be due to large
deletions of one exon or more of MEN1 and intron mutations which would not be detected
in routine sequencing of the gene (Schernthaner-Reiter et al., 2016). They may also be
the result of mutations in CDKN1B which causes MEN4 syndrome (see section 2.1.3,
MEN4) and other as yet unidentified genes.
A patient is considered to have MEN1 syndrome if they have endocrine tumours of two
out of the three types of tumours classically associated with MEN1, these are parathyroid
adenoma, pituitary tumour and entero-pancreatic NET (Brandi et al., 2001). If the
patient already has a 1st degree relative with MEN1 then they are considered to have
MEN1 once they develop one of these tumour types (Brandi et al., 2001). The prevalence
of MEN1 is estimated to be 2-3 per 100,000 live births (Sakurai et al., 2012).
MEN1 syndrome is associated with the development of multiple NET as well as the
development of other non-endocrine tumours such as facial angiofibromas (Brandi et al.,
46
2001). By 40 years of age, 90 % of MEN1 patients have developed parathyroid adenoma
(causing hyperparathyroidism), 40 % gastrinoma, 10 % insulinoma and 20 % a non-
functioning entero-pancreatic NET (Brandi et al., 2001). Anterior pituitary tumours are
quite common, with prolactinoma present in 20 % of patients by 40 years of age while
other tumours such as non-functioning bronchial and thymus NET are present at lower
rates in patients (Brandi et al., 2001). Around 5 % of pituitary adenomas are familial and
the majority of these are associated with MEN1 syndrome (Schernthaner-Reiter et al.,
2016).
Children of an individual with MEN1 syndrome have a 50 % chance of inheriting the
mutated MEN1 allele. MEN1 has an age related penetrance. A penetrance of 7 % under
10 years of age was found, rising to a high penetrance of over 90 % by 40 years of age
(Bassett et al., 1998; Crona and Skogseid, 2016). By 60 years of age, a penetrance of 100
% has been reported (Bassett et al., 1998).
In the past a leading causes of death in MEN1 patients was Zollinger-Ellison syndrome
(ZES) and hyperparathyroidism (Brandi et al., 2001). There are now effective treatments
for these conditions enabling them to be well controlled and increasing patient survival.
Despite this improvement two thirds of MEN1 patients still die of a MEN1 related cause
(Falconi et al., 2016). Many of the tumours associated with MEN1 are benign, however
tumours such as non-functioning PNET and gastrinomas can be malignant.
40-45 % of MEN1 patients have causes of death related to the presence of PNET
making this the leading cause of death in MEN1 patients (Ito et al., 2013; Falconi et al.,
2016). Therefore individuals with MEN1 require regular follow up to screen for PNET
so that they can be identified as early as possible.
+++MEN2 and MEN3 MEN2 (also known as MEN2A) and MEN3 (also known
as MEN2B) syndromes are autosomal dominant conditions caused by gain of function
mutations in the ret proto-oncogene (RET) gene which encodes a receptor tyrosine kinase,
47
RET (Thakker, 2016). RET is involved in cell growth and differentiation (Minnetti and
Grossman, 2016).
The most common feature in MEN2 and MEN3 syndrome is medullary thyroid cancer
(MTC) which develops in nearly 100 % of patients (Minnetti and Grossman, 2016).
Patients also have an approximately 50 % chance of developing a pheochromocytoma
(Minnetti and Grossman, 2016). Patients with MEN2 have a 15-50 % chance of developing
primary hyperparathyroidism, however MEN3 patients do not develop this condition but
instead may develop intestinal autonomic ganglion dysfunction (Thakker, 2014; Minnetti
and Grossman, 2016).
A third condition, familial MTC (FMTC) is where inherited RET mutations are
present in an adult patient with MTC but there is no primary hyperparathyroidism
or pheochromocytoma (Grajo et al., 2016).
Around 95-98 % of patients with MEN2, MEN3 and FMTC have a germline RET mu-
tation (Minnetti and Grossman, 2016). MTC is the most aggressive part of the disorder
while the pheochromocytomas tend to be benign. Family members who have inherited
the RET mutation are recommended to have prophylactic thyroidectomy between 5 and
10 years of age in MEN2 and FMTC (Grajo et al., 2016). This is done even earlier, before
6 months of age, in MEN3 patients due to the RET mutations present in these patients
causing a particularly aggressive form of MTC (RET mutations in exon 16, codon 918
or 883) (Grajo et al., 2016; Minnetti and Grossman, 2016).
+++MEN4 MEN4 syndrome (also known as MENX) is a rare inherited autosomal
dominant condition caused by loss of function mutations in cyclin dependent kinase in-
hibitor 1B (CDKN1B) (Thakker, 2016). CDKN1B transcription is regulated by menin
which could explain why patients with CDKN1B mutations and those with MEN1 exhibit
very similar clinical features (Schernthaner-Reiter et al., 2016).
Due to the small number of cases so far identified, there is a lack of detailed information
48
about the process of tumourigenesis in MEN4 and the precise clinical phenotype of this
condition.
Around 3 % of patients with MEN1 syndrome associated tumours in two or more
endocrine glands (parathyroid adenomas, pituitary tumours, PNET) have a mutation
in CDKN1B but no mutation in MEN1 (Lee and Pellegata, 2013). These patients are
considered to have MEN4 syndrome. The first human case was identified a decade ago
after extensive studies in rats with a mutation in the rat Cdkn1b gene (Pellegata et
al., 2006). 12 index cases of MEN4 have been described in the literature with germ
line mutations in CDKN1B (Lee and Pellegata, 2013). More recent case reports have
identified additional MEN4 cases (Tonelli et al., 2014; Pardi et al., 2014).
Since this condition was only discovered quite recently there is a lack of routine test-
ing for CDKN1B mutations in patients with the features of MEN1 syndrome who lack
mutations in MEN1. If testing for CDKN1B mutations was done more frequently in this
setting it is likely that more cases of MEN4 syndrome would emerge. A larger number
of cases would enable the clinical syndrome to be better characterised and the biological
effects of CDKN1B mutations to be better understood.
+++VHL Von Hippel-Lindau syndrome (VHL) is an inherited autosomal dominant
condition caused by mutations in the von Hippel-Lindau tumor suppressor (VHL) gene.
VHL regulates the hypoxia response by targeting the alpha subunits of hypoxia inducible
factor for degradation in normoxia but not when there are low cellular oxygen levels
(Schokrpur et al., 2016).
The incidence of VHL is estimated at 1 in 36,000 to 39,000 live births (Schunemann
et al., 2016). There is a high disease penetrance of 90 % by 65 years of age (Schunemann
et al., 2016).
Patients with VHL develop multiple tumours including endocrine neoplasms such as
paragangliomas, pheochromocytoma and in 10-17 % of patients, non-functioning PNET
49
(Minnetti and Grossman, 2016; Falconi et al., 2016). A study of 55 patients with VHL
and pancreatic lesions in South Korea found that the median age of onset was 33 years
of age (range 12-67 years) (Park et al., 2015).
VHL is also associated with the development of various non-endocrine tumours in-
cluding clear cell renal cell carcinoma (RCC), endolymphatic sac tumours, retinal and
central nervous system (CNS) haemangioblastomas, with RCC and CNS haemangioblas-
toma complications being the leading cause of death in these patients (Park et al., 2015).
During their lifetime, 60-90 % of VHL patients will develop multiple haemangioblastomas
(Schunemann et al., 2016).
+++NF1 Neurofibromatosis type 1 (NF1) (also known as von Recklinghausen’s dis-
ease) is an inherited autosomal dominant condition caused by mutations in the neurofi-
bromin 1 (NF1) gene (Minnetti and Grossman, 2016). This encodes neurofibromin which
has a role in proliferation and cell growth through the regulation of the RAS/MAPK
and the mTOR pathways (Minnetti and Grossman, 2016). NF1 occurs in around 1 in
2,500–3,000 live births, making it the most common autosomal dominant condition in
the nervous system (Nishi et al., 2012; Blakeley and Plotkin, 2016).
Patients with NF1 syndrome develop tumours in the central and peripheral nervous
system, skin lesions and can develop cognitive deficits and neuroendocrine tumours (Min-
netti and Grossman, 2016). Plexiform neurofibroma affects around 50 % of patients and
grows fastest during childhood, there are difficulties in attaining R0 in surgery due to
the location of these tumours (Blakeley and Plotkin, 2016). Patients have a lifetime risk
of 8-13 % of developing a malignant peripheral nerve sheath tumour which has a 5 year
survival rate of < 50 % due to limited treatment options (Blakeley and Plotkin, 2016).
NF1 patients have a 10-15 year lifespan decrease, 59 years of age is the median age of
death, with the malignancy being the most frequent cause of death (Jensen et al., 2008).
The most common NET observed in NF1 patients are somatostatinomas in the duode-
50
num or the periampullary region (Minnetti and Grossman, 2016). It has been reported
that 48 % of patients with duodenal somatostatinomas have NF1 syndrome (Jensen et
al., 2008). NET occurring rarely in NF1 include gastrinomas, pheochromocytomas and
VIPomas. PNET are also rare in NF1, with less than 10 cases reported in the literature,
however when they do occur they tend to be malignant (Nishi et al., 2012).
This condition is challenging to treat due to the wide range of possible disease man-
ifestations, with variation in the severity of the condition across the different stages of
development and with some patients having multiple aggressive tumours while others are
affected far less (Blakeley and Plotkin, 2016).
+++Tuberous Sclerosis Tuberous sclerosis is an inherited autosomal dominant con-
dition caused by mutations in tuberous sclerosis 1 (TSC1) or tuberous sclerosis 2 (TSC2)
genes (Jensen et al., 2008). TSC1 and TSC2 encode the proteins hamartin and tuberin
respectively (Jensen et al., 2008). TSC1 and TSC2 have a role in mTOR pathway reg-
ulation through interactions with the GTPase, Ras homolog enriched in brain (RHEB)
(Jensen et al., 2008). TSC1 and TSC2 mutations cause mTOR activation leading to cell
proliferation (Minnetti and Grossman, 2016).
Tuberous sclerosis is characterised by hamartomas, skin lesions, cerebral pathology and
renal angiomyolipomas (Minnetti and Grossman, 2016; Jensen et al., 2008).
A small proportion of patients with this condition have insulinomas, gastrinomas or
NF-PNET, these are normally seen in patients with TSC2 mutations (Minnetti and
Grossman, 2016; Jensen et al., 2008). Patients with low TSC2 expression were shown to
have shorter overall survival and time to disease progression (Missiaglia et al., 2010).
Family history of cancer
Various studies have found cancer in a first degree relative to be a possible risk factor for
an individual to develop a GEP-NET. It is considered to be one of the most relevant risk
51
factors for GEP-NET development (Leoncini et al., 2016).
Several studies have been done in PNET. A systematic review and meta-analysis was
done to identify case controlled studies from the literature up to October 2013, in order
to assess risk factors for PNET (Haugvik et al., 2015). 5 studies with a total of 827 cases
(2407 controls) met the inclusion criteria. Having a first degree relative with cancer was
associated with an increased risk of developing a sporadic PNET with a combined odds
ratio of 2.16 (95 % confidence interval: 1.64-2.85, p < 0.01).
In particular, several case controlled studies found an association with the presence
of oesophageal cancer, gall bladder cancer, sarcoma and ovarian cancer in first degree
relatives with the development of a PNET in a particular individual (Halfdanarson et al.,
2014; Hassan et al., 2008a; Leoncini et al., 2016). Cancers at other sites in first degree
relatives however, were not found to be associated with PNET (Halfdanarson et al., 2014;
Hassan et al., 2008a; Leoncini et al., 2016).
A recent Chinese case controlled study including 385 sporadic PNET patients and 614
age and sex matched controls separated functioning and non-functioning PNET in their
analysis (Ben et al., 2016). In this study, first degree family history of cancer was identified
as being associated with the development of non-functioning PNET but not functioning
PNET. It would be interesting to have more studies of this nature with functioning
and non-functioning tumours compared to see if this finding can be replicated in other
populations. Most of the other studies to date did not report tumour functionality or did
not conduct data analysis for these individual patient subgroups.
A family history of cancer has also been associated with the development of SBNET
(Leoncini et al., 2016). A recent Italian prospective case controlled study including 215
SBNET patients and 860 controls investigated family history of cancer as a risk factor
for SBNET (Rinzivillo et al., 2016). It was found that having a first degree relative
with colorectal cancer (odds ratio: 1.89, 95 % confidence interval: 1.07–3.33, p = 0.02)
or breast cancer (odds ratio: 2.25, 95 % confidence interval: 1.30–3.87, p = 0.003) was
52
associated with developing a SBNET.
Several other studies have investigated a family history of cancer as a risk factor for SB-
NET. A study carried out in Sweden and Finland found a significant association between
a first degree relative with kidney cancer or polycythemia vera and a patient developing
SBNET (Kharazmi et al., 2013; Leoncini et al., 2016). Other specific cancers in first
degree relatives that have been associated with the development of SBNET include, car-
cinoid tumours, nervous system cancers, oral cancer, endometrial cancer, non-Hodgkin’s
lymphoma, squamous cell skin cancer, colorectal cancer and prostate cancer (Hiripi et al.,
2009; Hassan et al., 2008a; Leoncini et al., 2016).
It should also be noted that patients with GEP-NET are themselves at risk of develop-
ing a non-NET second primary malignancy. As many as 15 % of GEP-NET patients have
a second primary malignancy with colorectal, breast and renal adenocarcinomas being
the most common (Clift et al., 2015).
More international case controlled studies are needed to increase the numbers of pa-
tients that can be included so that a more in depth analysis can be done on family
history of cancer and its relationship with patients developing GEP-NET. This will help
to identify if the differences in results seen between some of the existing studies are due
to population differences or selection biases. To this end it would be helpful if family his-
tory of cancer data was collected and included in the various national and international
databases being used to record the clinical details of GEP-NET patients. This would
enable larger studies to be done in the future and enable the risk factors for different
GEP-NET to be better understood.
Age
The likelihood of being diagnosed with a GEP-NET increases with age (Ellis et al., 2010).
This factor has been investigated in a number of large scale population based studies.
Increased age at diagnosis has also been linked to poorer survival (Ahmed et al., 2009;
53
Lepage et al., 2007; Yao et al., 2008).
Sporadic PNET are usually diagnosed between the ages of 50 and 80 (Yates et al.,
2015). Patients with inherited syndromes such as MEN1 however, usually develop PNET
at a much earlier age. For example, 25 % of MEN1 patients develop an insulinoma before
the age of 20 years (see section 2.1.3, MEN1) (Falconi et al., 2016).
In SBNET there is a peak in diagnosis frequency between the ages of 50 and 69 years
(Eriksson et al., 2008; Niederle et al., 2016).
A large European study was carried out investigating the epidemiology of NET includ-
ing over 20,000 patients (Van Der Zwan et al., 2013). The study found that the incidence
of NET was greatest amongst patients aged 65 years or older. NET incidence rates per
million population per year were found to be just 2 for the age group 0-24 years of age,
rising to 20 and 88 respectively for the age groups 25-64 and ≥ 65 years of age. For
the non-functioning, low grade GEP-NET included in the study, the incidence rates were
1.14 per million at age 0-24 years, 11 per million at age 25-64 years and 40 per million
at age ≥ 65 years.
There were similar findings from studies in the USA (n = 35,618) with increasing
age being associated with increased risk of a NET diagnosis (Yao et al., 2008). 50 %
of patients were 63 years or older when diagnosed with a NET in the study, while the
median age a diagnosis was 66 years for SBNET, 60 years for PNET and 65 years for
colon NET. In contrast to these results, appendiceal NET were more frequent in patients
a decade or so younger, with a median age at diagnosis of 47 years.
Appendiceal NET have a lower age adjusted incidence rate at advanced ages and are
uncommonly diagnosed in older patients, especially those over the age of 70 years, in
contrast to other sporadic GEP-NET (Ellis et al., 2010). This suggests that age may not
be a risk factor for these particular GEP-NET and it has been suggested that this may be
due to appendiceal NET being identified as incidental findings following appendectomy,
primarily carried out in patients under 40 years of age (Ellis et al., 2010).
54
Gender
In GEP-NET as a whole, data from the USA and the UK points towards a slightly higher
incidence in women than in men (Modlin et al., 2003; Yao et al., 2008; Hassan et al.,
2008b; Ellis et al., 2010). Other studies however, for example in Germany and Argentina
found no gender preference in GEP-NET (Ploeckinger et al., 2009; O’Connor et al., 2014).
Studies of GEP-NET patients usually record gender as a matter of course which adds
to the body of data available and in some of these studies, independent analysis of the
results is done for males and females. Data emerging from epidemiological studies such
as these reveal gender differences in the frequency of GEP-NET occurring at different
primary sites. These gender differences are however complex to investigate since they
may be particular to the populations being studied rather than a more overarching trend
in GEP-NET. Further studies will be needed, in particular studies which provide data
specific to each primary site, in a wider range of different populations to investigate this
further.
With respect to PNET, a study in the USA population found that PNET were more
frequently observed in males than in females (1.4:1), however the trend was reversed in
a Japanese study which found that in this population PNET were more frequently seen
in females than in males (1:1.6) (Yao et al., 2008; Ito et al., 2010).
In SBNET some studies have suggested that there is no change in risk associated with
gender, while other studies have suggested instead that there may be a slightly elevated
risk of SBNET in men (Niederle et al., 2016). A study in the UK (n = 10,324) for
example found that small intestinal NET were more common in males than females,
while the converse was true for appendix NET (Ellis et al., 2010). Studies in the USA
documented the same finding (Yao et al., 2008; Modlin et al., 2003). Ellis et al found no
gender preference however for gastric, colon and rectal NET in their UK study population
(Ellis et al., 2010). A study in Germany found no gender preference for SBNET, or NET
of other primary sites (Ploeckinger et al., 2009).
55
In a study in the USA, male patients were found to be significantly more likely than
female patients to have metastases at presentation, 29 % versus 25 % respectively (Yao
et al., 2008). This could be due to delays in diagnosis possibly exacerbated by men being
more reticent than women about visiting the doctor with health complaints that could
be related to cancer, this can be because of a myriad of reasons from embarrassment to a
lack of awareness about symptoms (Ozturk et al., 2015). Female patients also had better
survival durations in all stage categories than male patients (Yao et al., 2008).
Ethnicity
Ethnicity has been investigated with respect to the risks of developing different types of
GEP-NET in the USA as well as in Japan and Europe. While some general trends have
been identified within specific populations, it is not possible with the current data to
determine if the differences seen between certain ethnic groups are due to genetic factors
or may be instead be due to other differences such as diet, climate, socioeconomic factors
or healthcare systems in place in different geographical areas. Many of the epidemiological
studies that have been done did not record data on ethnic groups or did not have access
to this data, limiting the amount of information available.
In population based studies carried out in the USA, SBNET were found at the high-
est frequency in African Americans followed by white Americans, the incidence rate of
SBNET in African Americans was reported to be as much as 80 % higher than in white
Americans in one study (Hauso et al., 2008; Yao et al., 2008). SBNET were reported
to occur at a low frequency in Asian Americans, however rectal NET were seen at the
highest frequency in this ethnic group (Yao et al., 2008). These differences are likely to
reflect wider socioeconomic disparities between different ethnic groups within the USA
population in addition to any genetic differences. For example poverty levels in the USA
between 1980 and 2006 were 2-3 times higher for African Americans (Williams et al.,
2012). This reflects some of the challenges involved in unpicking these complexities to
56
understand how socioeconomic factors can interact with genetic factors and social factors
to affect health in certain ethnic groups in the way that has been observed. It is partic-
ularly hard to properly control for these many factors, such as income, education, access
to heathcare and other social factors in GEP-NET due to the relatively low numbers of
patients included in these studies compared to those done in other diseases such as breast
cancer where possible confounding factors can be more easily accounted for (Williams et
al., 2012).
There have been several studies in Asia, two in Japan and another in South Korea
(Ito et al., 2010; Cho et al., 2012). The study of the Japanese population in contrast to
the studies done in the USA population, found SBNET were rare whereas rectal NET
were the most frequent GEP-NET observed (Ito et al., 2010). These differences could be
due to genetic differences or differences in other environmental or socioeconomic factors.
Ito et al suggested that colonoscopy being included in periodical health examinations
in Japan could be behind the higher number of rectal NET being seen in the Japanese
population, as more might be discovered incidentally than in other populations (Ito et al.,
2010). Another Japanese study found that the frequency of insulinomas was higher in
Japan than in the USA and Europe (Sakurai et al., 2012).
The study in South Korea, found similar results to the Japanese study, that there was a
higher incidence of rectal NET and a lower incidence of SBNET compared to the studies
of people of white-European heritage in Europe and in the USA (Cho et al., 2012). Cho et
al investigated 4951 GEP-NET from 2000 to 2009 in a multicentre study in South Korea
with the findings that rectal NET were the most common, representing nearly half of the
GEP-NET (48.0 %) while SBNET were only present in 7.7 % of GEP-NET patients in
this population.
The vast majority of studies done in GEP-NET have been done in people of white-
European heritage in Europe, North America and Australia. There have also been several
GEP-NET epidemiological studies done in Asia, including studies in Japan and South
57
Korea and studies of the Han Chinese ethnic group in China (Zhan et al., 2013; Ben et al.,
2016). However there are very few studies of GEP-NET epidemiology in the continents
of South America and Africa. This prevents a more clear picture from emerging of the
epidemiology of GEP-NET in these geographical areas and amongst the ethnic groups
found in these continents, who remain under-represented in the studies to date.
National databases or continent wide databases of GEP-NET patients in Africa and
South America should be set up in order to capture clinical and epidemiological data from
patients in these populations and address the near complete lack data from these regions.
This may also help to raise awareness of GEP-NET amongst clinicians and the public so
that more patients can be identified. A consensus conference in 2008 on the management
of GEP-NET in Latin America called for a Latin-American Registry of patients with
NET of the gastrointestinal tract and pancreas to be set up (Costa et al., 2008).
Since then a study was done in 2014 of GEP-NET in Argentina and found that the most
common primary sites were the small bowel and the pancreas with NET in the appendix
and stomach being rarer (O’Connor et al., 2014). The authors noted that this matches
findings in Spain and other European populations, which is unsurprising given the large
proportion of Argentina’s population which is of white-European heritage (O’Connor et
al., 2014).
Socioeconomic Factors
There have been limited studies that have recorded data on socioeconomic factors such
as education level, income level, poverty in childhood and social background. The studies
that do exist vary greatly in methodology, the types of GEP-NET included in the study
and how and if certain socioeconomic factors are examined. This probably explains the
conflicting results seen in different countries, as to whether or not there is an association
between certain socioeconomic factors and increased risk of certain types of GEP-NET or
worse prognosis. This is further exacerbated by the vast differences in healthcare systems
58
and the availability of different levels of healthcare in different countries. For example
universal healthcare that is free at the point of use in the EU, compared to insurance
based healthcare in countries like the USA, where individuals may be more likely to fall
through the cracks if they are uninsured.
A study of well differentiated NET in the UK between 1986 and 2001 (n = 3233) found
that there was no difference in survival between different socioeconomic groups (Lepage
et al., 2007). Two studies focusing on PNET have suggested that certain socioeconomic
factors were not associated with the risk of PNET or differences in survival. A study
in PNET in China (n = 385), found that education levels were not associated with an
increased risk of PNET or with having an advanced ENETS stage at diagnosis (Ben
et al., 2016). A study of PNET in the USA (n = 3851), found that median income and
insurance status were not patient survival predictors (Stewart et al., 2008).
A study in Sweden of 5184 NET found that socioeconomic factors including birth in a
large city and a well educated social background were risk factors for NET (Hemminki
et al., 2001). Perhaps this is linked to a higher likelihood of the GEP-NET being diag-
nosed in this setting due to better awareness and access to more specialist centres for
investigation and identification of the condition. This study did not investigate any pos-
sible associations between socioeconomic factors and patient outcomes. A case controlled
study in China of insulinoma patients (n = 196) found that living in a rural area was
associated with an increased risk of insulinoma occurrence (Zhan et al., 2013).
A Canadian study of 5619 NET (1994 – 2009) found that low socioeconomic status
(lowest income quintile) and rural living were significant independent predictors of worse
overall survival (Hallet et al., 2015). This is likely due to healthcare delivery disparities
in these patients leading to delays in diagnosis and access to effective treatments. For
example a lack of awareness amongst patients and rural doctors who may see very few
such cases, as well as restricted access to specialist treatment centres which are usually
centred around large conurbations.
59
Blood type
A small number of retrospective studies have been done in GEP-NET to investigate if
there is an association between ABO blood group and GEP-NET. They were prompted
by findings that blood type O was associated with a reduced risk of developing pancreatic
adenocarcinoma, while blood group A was associated with an increased risk (Nell et al.,
2015; Weisbrod et al., 2012).
A retrospective study in the USA was done on 181 VHL syndrome patients (those for
whom blood type was known) (Weisbrod et al., 2012). The study found an association
between the presence of blood type O in VHL patients and the presence of solid pancreatic
tumours. More studies will be needed to determine if this association is also present in
other global populations of VHL patients.
There was conflicting information from two separate studies done in MEN1 patients.
The authors of the VHL study, Weisbrod et al, did a study investigating blood type O in
MEN1 patients (Weisbrod et al., 2013). The study included 105 MEN1 patients in the
USA to identify any association between ABO blood group and GEP-NET development.
The presence of a GEP-NET was more common in the group of patients with blood type
O than in those with a non-O blood type, with 53 % of the blood type O patients having
a GEP-NET compared to 28 % of the non-O blood type patients. This suggested that
there could be an association between the O blood type and the development of a primary
GEP-NET.
A study in the Netherlands found conflicting results. This study used the Dutch
national MEN1 database to look for any association between blood type O and GEP-
NET occurrence (all MEN1 patients in the database were included for whom there was a
record of their ABO blood group) (Nell et al., 2015). In this study (n=200) no difference
was found between the type O blood group and the non-O blood type group with respect
to survival and metastases, with similar clinical and demographic characteristics across
the two groups.
60
These differences between the two studies could be caused by the different population
of MEN1 patients being studied in each case. It could be that in the MEN1 population in
the USA, there is an association between blood type O and NET development while this
pattern is not present in the Dutch MEN1 population. The Dutch study was population
wide (national database) where as the study in the USA was a small subset of the MEN1
population so the results are more likely to be affected by selection bias. It would be
interesting to see data from additional studies of other populations of GEP-NET patients
both in the USA and in other countries to determine if there is a true association between
patients having blood type O and the development of these tumours in certain MEN1
populations.
Tobacco smoking and alcohol consumption
Case controlled studies investigating tobacco smoking and alcohol consumption as poten-
tial risk factors in GEP-NET have presented conflicting data. Due to this and the limited
number of studies that have been done, there are no clearly established environmental
risk factors or exposures for the development of GEP-NET in contrast to other more
common cancers (Du et al., 2016).
Though the emerging data suggests that smoking and alcohol consumption could be
risk factors in GEP-NET there is not yet enough reproducible evidence to draw definitive
conclusions. The studies that have been done have varied study design with differences
in patient inclusion and how and what patient subgroups, if any, are considered in the
analysis (for example: primary site, functionality, presence of a familial syndrome). These
factors could explain the conflicting results found in some of the investigations in this
area. More studies are needed, with a careful study design and if possible larger patient
numbers from international rather than single centre or national studies to gain a more
accurate picture of whether tobacco smoking and alcohol consumption really do represent
a risk factor for GEP-NET.
61
+++Tobacco Smoking There have been mixed results from studies of tobacco smok-
ing as a risk factor for GEP-NET. In studies looking at GEP-NET as a whole, a study
in the USA did not find smoking to be a risk factor, whereas a study in Italy found that
smoking rates in GEP-NET were double those found in the general population (Hassan
et al., 2008b; Faggiano et al., 2012). These results could represent population differences
or may be the result of patient selection biases.
Due to the heterogeneity of GEP-NET it is likely to be more appropriate to investigate
risk factors based on primary site, although this too is associated with problems of being
able to include sufficient patients to be able to identify such associations. In studies
where tobacco smoking was investigated in relation to the primary site of the tumour,
there was a significant association of heavy smoking with the development of SBNET but
not PNET, as described below.
A meta analysis of 4 case controlled studies in PNET (studies up to June 2014 were
included) found no significant association between ever smoking tobacco and PNET
(Leoncini et al., 2016). There was also no significant association between heavy smoking
and PNET (Leoncini et al., 2016). A subsequent case controlled study in 385 sporadic
PNET patients had opposing findings, identifying that ever or heavy smoking were in-
dependent risk factors for non-functioning PNET but interestingly not for functioning
PNET (Ben et al., 2016). The study also found that ever or heavy smoking was associ-
ated with a more advanced stage of disease at diagnosis (ENETS stage: III or IV).
Earlier studies did not investigate smoking with respect to the functionality of the
tumour so this could explain the differing findings, if the functioning PNET included in
the earlier studies were masking an association of smoking with non-functioning PNET.
Large well designed studies will be needed to understand this in more detail and to
investigate if there are differences in the risk factors for different subgroups of patients,
in particular smoking for functioning versus non-functioning PNET.
A case controlled study in the USA including 325 SBNET patients found that smoking
62
(≥ 100 cigarettes during lifetime) or heavy smoking (≥ 1 cigarette pack per day for > 20
years) were not significant risk factors for SBNET (Hassan et al., 2008b). A European
multicentre population based case controlled study in SBNET (n = 84) found contrasting
results based on data from Denmark, Sweden, France, Germany and Italy (Kaerlev et
al., 2002b). The study found that ever smoking and heavy smoking were associated with
SBNET, this held true even for individuals who had stopped smoking > 10 years before
the study. A recent case controlled study in Italy (n = 215) also found that smoking and
heavy smoking were significantly associated with SBNET (Rinzivillo et al., 2016).
Given the differing findings in the current studies of cigarette smoking in GEP-NET
more investigations will be needed to make reliable conclusions as to whether it is a risk
factor in GEP-NET as a whole and or in specific subgroups or populations of patients.
+++Alcohol There is a lack of good quality data on alcohol consumption in GEP-
NET patients and the data that is available presents a conflicting picture. For the data
that does exist, there are differing findings depending on primary tumour site. There
are conflicting results from a small number of studies done in SBNET, however several
studies have suggested an association between alcohol consumption and the development
of NET of the pancreas and possibly the rectum, as described below.
A meta analysis of the available 4 case controlled studies was done in PNET investigat-
ing alcohol consumption as a risk factor for PNET (Leoncini et al., 2016). Leoncini et al
found that drinking and heavy drinking in particular were associated with an increased
risk in PNET. The odds ratio for ever drinking was 1.09 (95 % confidence interval: 0.67-
1.77, p = 0.001) and there was an even stronger effect for heavy drinking with an odds
ratio of 2.44 (95 % confidence interval: 1.07-5.59, p = 0.054). A recent study in the
Han Chinese ethnic group in China found that heavy drinking was associated with an
increased risk of developing a functioning PNET (Ben et al., 2016)
There are even fewer studies looking into alcohol consumption in rectal NET. A recent
63
South Korean study found that heavy drinking was associated with rectal NET, with an
adjusted odds ratio 1.56 (95 % confidence interval: 1.01-2.42, p = 0.045) (Jung et al.,
2014). An earlier study in the USA however, found no significant association between
alcohol consumption and rectal NET (Hassan et al., 2008b).
In SBNET, two case controlled studies comparing never drinkers with drinkers showed
that alcohol consumption was not associated with a significantly increased risk of SBNET
(Hassan et al., 2008b; Chen et al., 1994; Leoncini et al., 2016). The Hassan et al study
also investigated heavy drinking but still found this factor was also not associated with
a significantly increased risk of SBNET (Hassan et al., 2008b).
A more recent study found similar findings for never drinkers versus moderate drinkers,
with no significant increase in the risk of SBNET (Rinzivillo et al., 2016). For heavy
drinkers however, those who consumed > 21 drinks per week (each containing 12 g of
alcohol), this study did find a significant association with the development of a SBNET.
More case controlled studies will be needed in order to definitively show one way or
another if drinking or heavy drinking can be considered to be a risk factor for SBNET.
More studies in this area are also warranted for other GEP-NET sites to increase the
evidence base.
Other risk factors
Other possible risk factors that have been investigated as possible risk factors for de-
veloping a GEP-NET include some lifestyle related risk factors such as obesity, type II
diabetes and certain occupational risk factors.
+++Diabetes There have been a number of studies done into diabetes as a possible
risk factor for GEP-NET. There are consistent results suggesting that diabetes may be a
risk factor for PNET but not SBNET. Diabetes type II is highly correlated with obesity
which is another possible risk factor for GEP-NET (Hassan et al., 2008b).
64
In PNET, three case controlled studies all found that there was a significant association
between diabetes and the development of PNET, the estimates of the size of the effects
were 2.80 (95 % confidence interval: 1.50–5.20), 4.80 (95 % confidence interval: 2.30–9.90),
and 1.91 (95 % confidence interval: 1.26–2.91) (Hassan et al., 2008b; Halfdanarson et al.,
2014; Capurso et al., 2012; Leoncini et al., 2016). A more recent study also found that
diabetes was associated with non-functioning PNET but not functioning PNET (Ben
et al., 2016).
Diabetes in gastric and rectal NET was also investigated by the Hassen et al study
in the USA, in which it was found that patients with diabetes were at increased risk of
Gastric NET but not rectal NET (Hassan et al., 2008b).
Two studies in SBNET found no significant association between diabetes and develop-
ment of SBNET (Hassan et al., 2008b; Cross et al., 2013).
+++Obesity Obesity, as assessed by a body mass index (BMI) ≥ 30, has been in-
vestigated in several studies in GEP-NET. There are conflicting results from the studies
that have been done as to whether obesity is associated with GEP-NET or whether in
fact there is an inverse association with these tumours. Therefore at this time there is too
little reproducible data to draw any conclusions as to if there is any positive or negative
association between obesity and GEP-NET.
There have been several case controlled studies investigating obesity in PNET patients
with conflicting results. A study in China of insulinoma patients (n = 196) and a study in
the USA of PNET patients (n = 309) both found that obesity (BMI ≥ 30) was associated
with a small increased risk of these tumours (Halfdanarson et al., 2014; Zhan et al., 2013).
Another study in the USA that included PNET patients (n=160) however identified the
opposite relationship, with being obese or overweight being inversely associated with
PNET (Hassan et al., 2008b).
The study by Hassan et al also included SBNET patients (n=325) in the analysis, and
65
in SBNET the study found an inverse association for obese and for overweight individuals
with SBNET, the size of the effect was 0.40 (95 % confidence interval 0.20 – 0.50) matching
the findings of this study for PNET (Hassan et al., 2008b). A more recent study, also
conducted in the USA, of SBNET (n=124) however found the opposite effect, that being
obese rather than normal weight increased the risk of SBNET with a hazard ratio of 1.95
(95 % confidence interval: 1.06 – 3.48) (Cross et al., 2013).
In a case controlled study of rectal NET in 102 patients in South Korea, no association
was found between rectal NET and BMI, physical activity or waist circumference, however
higher cholesterol levels were significantly associated with the occurrence of rectal NET
(Pyo et al., 2016). A study in the USA also found no significant association between BMI
and rectal NET or gastric NET (Hassan et al., 2008b).
+++Occupational risk factors There is very scarce data on occupation as a possible
risk factor for GEP-NET, with only one study to date in the literature. This was a
European population based case control study by Kaerlev et al carried out in 84 SBNET
patients (from Denmark, Sweden, France, Germany and Italy) (Kaerlev et al., 2002a).
The occupations most associated with SBNET (with a two fold or greater odds ratio) were
women working in the wholesale food/beverage industry (odds ratio: 8.2; 95 % confidence
interval: 1.9 – 34.9) and men working in the manufacturing of footwear (odds ratio: 3.9; 95
% confidence interval, 0.9 – 16.1), motor vehicle bodies (odds ratio: 5.2; 95 % confidence
interval: 1.2 – 22.4) and metal structures (odds ratio: 3.3; 95 % confidence interval: 1.0
– 10.4). High risk occupations (odds ratio > 2) included shoemakers, welders, machine
fitters and construction workers. The authors note that the findings of their explorative
study are tentative and the associations could be due to chance (Kaerlev et al., 2002a).
These findings indicate that certain occupations with certain exposures could be linked to
SBNET and this is an area that it would be interesting to investigate further in additional
studies both in SBNET patients and patients with other types of GEP-NET.
66
Summary
A family history of an inherited GEP-NET syndrome such as MEN1, MEN2/3, MEN4,
NF1 or tuberous sclerosis in a first degree relative is well established as a cause of GEP-
NET if the causative mutation has been inherited by an individual. These syndromes are
autosomal dominant and are characterised by a high penetrance, with nearly complete
penetrance by 60 years of age in MEN1 for example.
Other potential risk factors for GEP-NET development have been investigated in a
relatively small number of studies so there is a lack of high level evidence in this area.
The most promising potential risk factors, with the most data to support them, include
family history of cancer in a first degree relative, increasing age and diabetes (in PNET
only) as being associated with an increased risk of GEP-NET.
Further epidemiological studies are warranted to expand the data available for the
other possible risk factors so far identified. These include gender, ethnicity, socioeco-
nomic factors, blood type, tobacco smoking, alcohol consumption, obesity and occupa-
tion. Currently there is insufficient data in these areas and the data that is available
presents conflicting results.
There were differences in the risk factors found for PNET and SBNET and for function-
ing and non-functioning NET, suggesting that they may have different tumourigenesis
pathways. This should be a consideration when designing future epidemiological studies
of this kind.
Coming to clear conclusions on whether certain possible risk factors are truly associated
with GEP-NET remains challenging since there is still very little data on this. It is hard to
determine if the data is reproducible due to study design differences, low patient numbers
and differences in risk factors between the different GEP-NET primary sites pointing to
different tumourigenesis pathways. Therefore data from studies of GEP-NET as a whole
is likely to mask important differences in the risk factors between SBNET and PNET for
example. Low sample numbers inherent in the study of a rare disease means that studies
67
usually lack the statistical power needed to identify true risk factors that are associated
with only a modest increase in the risk of GEP-NET. It should also be noted that when
a particular genetic or environmental factor is identified in association studies as a risk
factor for GEP-NET, this does not necessarily imply causation, mechanistic studies would
be needed to demonstrate if the risk factor had a true biological effect on tumourigenesis.
The data that is available is often not unanimous, with some studies finding a significant
association, while others may not be able to reproduce this or occasionally even find
an inverse association. This conflicting picture is exacerbated by large differences in
study design, and the unquantified influence of genetic, environmental, socioeconomic
and medical treatment access differences inherent in the populations being studied. These
confounding factors remain challenging to control for when investigating one particular
risk factor or another, particularly when the patient numbers in each study remain quite
low.
Studies rarely represent the full GEP-NET population (or subpopulation of interest
eg: SBNET) present in a given country as they often represent only certain regions or
certain medical centres so there may be regional differences or certain communities that
are missed from the figures. In South America and Africa, there are hardly any published
epidemiological studies on GEP-NET and these leaves gaps in our knowledge of potential
differences in risk factors and in the incidence of certain primary tumours that may exist
in these continents.
More studies will be needed to address these weaknesses and to establish a higher qual-
ity evidence base for the risk factors that impact on GEP-NET. In particular collaborative
efforts to maximise the numbers of GEP-NET patients and to collect as much data on
potential risk factors as possible (for example: family history of cancer, exposures, BMI)
within existing cancer and specific GEP-NET registries and databases will enable future
studies to investigate risk factors in more depth.
68
2.2. Classification
The first GEP-NET case to be documented in the literature was described by Theodor
Langhans in 1867 (Langhans, 1867; Modlin et al., 2004). Several additional small in-
testinal cases were subsequently identified over 20 years later by Otto Lubarsch in 1888
and William Bramwell Ransom in 1890 (Lubarsch, 1888; Ransom, 1890). Ransom also
observed for the first time symptoms in one of his patients which were likely to be caused
by what was later known as carcinoid syndrome (Ransom, 1890; Wardlaw and Smith,
2008). Although the neoplasms were described in these few documented cases from the
1800s as having unusual histological features their status as a distinct pathological entity
was not confirmed.
The word ’karzinoide’ (carcinoid) was first used to describe these tumours in 1907
by Siegfried Oberndorfer (Oberndorfer, 1907). He considered them to be “carcinoma-
like” and described the histology of small ileal lesions made up of nests of polymorphic
cells (Oberndorfer, 1907; Modlin et al., 2004). On the basis of 6 ileal cases Oberndorfer
started to characterise the features of these tumours (Oberndorfer, 1907). He made
the important conclusion that, on the basis of their histology, they were distinct from
other gastrointestinal neoplasms, initially thinking that they had benign behaviour on
the basis of their slow growth rate when compared to carcinomas. In 1929, on the basis
of his studies of 36 additional patients, Oberndorfer revised his original opinion of their
benign behaviour and concluded that these tumours did in fact have the potential to
metastasise (Oberndorfer, 1928).
The endocrine-related nature of these tumours was first identified by Gosset and Mas-
son in 1914 who suggested, based on their silver impregnation studies in the tumours,
that they might arise from the enterochromaffin cell (Gosset and Masson, 1914; Modlin
et al., 2004; Wardlaw and Smith, 2008). For more details on types of neuroendocrine
cells please see section 2.4.
Initial attempts at developing classification systems for GEP-NET were started in
69
1963 on the basis of the percieved embryonic origin of the cells in the GEP system,
so the tumours were classified very broadly, based on if they arose from the foregut,
midgut or hindgut (Williams and Sandler, 1963; Modlin et al., 2004). This classification
system was of limited usefulness since it did not take into account the high levels of
heterogeneity in these tumours and it was also designed prior to a proper characterisation
of the different cell types which make up the diffuse neuroendocrine system. As more
became known about the tumour biology of GEP-NET and how tumourigenesis and
clinical course differed depending on the site of the primary tumour, the primary tumour
location was instead used as the basis for classification systems and treatment guidelines
for GEP-NET. These guidelines were updated over the years as more became known
about the tumour biology and more clinical trials were done providing a better quality
of evidence to support various treatment options.
Since NET are frequently slow growing they were historically considered benign with
low malignant potential. They are however often identified at a late stage and subsequent
studies showed that GEP-NET arising from certain sites such as the small bowel can be
aggressive and many tumours have a high chance of liver metastases (see section 2.1.2).
These findings have been reflected in various WHO and ENETS classification systems in
recent years. These now acknowledge the importance of considering the primary site and
features of the primary tumour in order to better assess the prognosis for a particular
GEP-NET patient. These considerations are reflected in the latest ENETS guidelines for
the management of GEP-NET published in 2016 (O’Toole et al., 2016; Niederle et al.,
2016; Falconi et al., 2016; Delle Fave et al., 2016; Garcia-Carbonero et al., 2016; Pavel
et al., 2016; Ramage et al., 2016; Pape et al., 2016).
2.2.1. Terminology
Since the discovery of GEP-NET they have been classified in a number of different ways.
In 1980 this type of tumour was still classified by the WHO as a carcinoid tumour (Merola
70
et al., 2016). By 2004 the term NET was adopted by the WHO to describe these tumours
in preference to the term carcinoid and different types of tumours were more precisely
classified into distinct subtypes based on their features and typical clinical behaviour
(Ramage et al., 2005). The 2010 WHO classification used grading to further classify
GEP-NET based on the Ki-67 index (Merola et al., 2016).
Terminology in this document
The term ‘carcinoid’ can be ambiguous due to historical changes in how this term was
used, thereofore the use of the term ‘carcinoid’ will be avoided wherever possible and
instead the term GEP-NET will be used, or the site of the primary tumour specified.
The term SBNET refers to NET of the ileum and or the jejunum only. This reflects
the latest ENETS guidelines which consider tumours of the ileum and jejunum to have
a similar etiology whereas the tumours of the duodenum much more closely resemble
gastric NET (Niederle et al., 2016).
2.2.2. Primary site
One of the most important ways that NET are characterised is based on the site of the
primary tumour within the body. Primary site is a key prognostic factor in GEP-NET
in addition to Ki-67 % grading and the presence or absence of metastases. Analysis of
historical GEP-NET data from the USA (SEER registry, 1973–2007) showed that 5 year
survival rates were lowest at 37.6 % for PNET and highest at 88.5 % for rectal NET,
which have much better prognosis (Lawrence et al., 2011). Other sites had intermediate
5 year survival rates in the study of 54.6 % for colon NET, 64.1 % for gastric NET, 68.1
% for small intestinal NET and 81.3 % for appendiceal NET.
GEP-NET are highly heterogeneous neoplasms, based not only on the primary site,
the peptide hormones/amines they secrete, the presence or absence of functional and/or
inherited syndromes but also on the specific biological pathways involved in tumourige-
71
nesis. This results in clinical heterogeneity which presents a key challenge for clinical
managment and for those running clinical trials. Of particular concern is whether it is
better to divide patients into smaller and smaller subgroups based on characteristics such
as primary site or to design trials based on broader biochemical pathways which may be
held in common across some of the different tumour subgroups (Cives et al., 2016).
Some of the biological and clinical heterogeneity of GEP-NET is due to the differing
tumour biology that these tumours have depending on their primary site or more specif-
ically the type of neuroendocrine cell the tumour arose in (see section 2.4). For example
a NET of the pancreas may overproduce insulin if the tumour arose in a pancreatic beta
cell (insulinoma) while a SBNET arising in an enterochromaffin cell may overproduce
serotonin (functioning SBNET, carcinoid syndrome). For more details on functioning
GEP-NET see section 2.2.5.
These differences in GEP-NET based on primary site are becoming better recognised.
In more recent ENETS guidelines, including those issued in 2016, duodenal NET are
classified alongside gastric NET in recognition that they more closely resemble biological
and clinical features of gastric NET than they do NET of the ileum or jejunum (Delle
Fave et al., 2016; Niederle et al., 2016).
Risk factors also have a different effect on GEP-NET arising from different primary
sites. For example diabetes and heavy alcohol consumption were possible risk factors for
PNET but not SBNET, while heavy smoking was found to be a possible risk factor for
SBNET but not PNET. This suggests that different biological pathways may be involved
in the development of these tumours (see risk factors, section 2.1.3).
When it comes to treatment it should be noted that despite their biological heterogene-
ity common features do remain across a wide range of GEP-NET. This means that there
are some treatment targets in common between biologically divergent GEP-NET due to
the same pathways being dysregulated (see section 2.3.1, treatment). These include the
SSTR which is expressed in the majority of well differentiated GEP-NET despite dif-
72
fering primary sites and functional status, making SSA an effective treatment across a
broad spectrum of GEP-NET (Cives et al., 2016). Another example is mTOR pathway
inhibition, which has shown efficacy even in treating non pancreatic GEP-NET since
it is upregulated in many of these tumours despite the absence of the mTOR pathway
mutations seen in PNET (Zatelli et al., 2016; Yao et al., 2011; Yao et al., 2016; Cives
et al., 2016). These findings suggest that in SBNET, epigenetic factors may be instead be
responsible for changes in the mTOR pathway observed in these tumours, in the absence
of mutations in mTOR pathway genes (see section 2.5).
2.2.3. Grade
Ki-67 is expressed in active cell cycle stages but not in resting cells, therefore it can be
used to identify proliferating cells (Scholzen and Gerdes, 2000). High proliferation rates
often occur in tumours due to dysregulation of the cell cycle and apoptosis pathways.
The Ki-67 index (Ki-67 %) is used as a prognostic biomarker in GEP-NET and in other
cancers such as prostate cancer (Rindi et al., 2011; Bullwinkel et al., 2006; Scholzen and
Gerdes, 2000). For more details on the role of Ki-67 % as a biomarker in GEP-NET
please see section 2.6.1.
During tumour grading, 2000 tumour cells are assessed to find the percentage of cells
staining positive on immunohistochemistry (IHC) for Ki-67 in areas with the highest
nuclear staining (Rindi et al., 2006; Rindi et al., 2007; Niederle et al., 2016).
This is used to categorise GEP-NET into low grade tumours, Grade 1 (G1) and Grade 2
(G2) tumours, which are well differentiated and high grade, Grade 3 (G3) tumours, (also
called neuroendocrine carcinomas) which are poorly differentiated and have high levels
of proliferation (see Table 2.1). Mitotic count is a parallel grading system which may be
used in addition to Ki-67 % or where Ki-67 immunohistochemistry is not available, it is
used mainly in broncho-pulmonary carcinoids.
Well differentiated, G1 and G2 tumours are far more common than poorly differentiated
73
Table 2.1.: GEP-NET grading according to ENETS guidelines, table reproduced fromRindi et al. (2006) and Rindi et al. (2007), creative commons licence: CCBY-NC.
Grade Ki-67 index (%) Mitotic countG1 ≤ 2 < 2G2 3-20 2-20G3 > 20 > 20
G3 tumours with the later representing only around 5 % of gastrointestinal NET (Garcia-
Carbonero et al., 2016).
The biological pathways and clinical behaviour of G1 and G2 tumours are different to
those underlying G3 tumours. Patients with G3 tumours have an aggressive and more
deterministic disease course with much worse outcomes than those with low grade G1/G2
tumours.
For G3 GEP-NET with metastases at diagnosis, median survival ranges from 1 month
with only the best supportive care to 12-19 months if treated with the best available
therapy, 85 % of G3 GEP-NET are metastatic at diagnosis (Garcia-Carbonero et al.,
2016). While patients with G1 and G2 tumours do have much better prognosis than
those with G3 tumours, there is a large amount of heterogeneity within low grade GEP-
NET. This means that tumour behaviour varies greatly between one patient and another
in patients with low grade tumours making the disease course, survival and response to
treatment in these tumours more challenging to predict. Novel biomarkers would be very
useful in this setting particularly if they could further stratify patients with low grade
tumours into clinically useful subgroups and identify which of these patients might have
a more tumours with more aggressive behaviour.
High grade tumours (G3) are characterised by different mutations and biological path-
ways, in addition to their differing clinical behaviour to that of low grade tumours. G3
tumours have a far higher mutation rate than low grade tumours with inactivating muta-
tions in tumour suppressors, tumour protein p53 (TP53) and RB transcriptional corepres-
sor 1 (RB1) (encodes the retinoblastoma-associated protein) being a common occurrence
74
(Garcia-Carbonero et al., 2016). Inactivating TP53 mutations are present in 50 % of can-
cers and cause genome instability through an impaired DNA damage response (Reinhardt
and Schumacher, 2012).
Mutations in TP53 and RB1 remain rare however, in G1 and G2 GEP-NET which
have a low mutation rate characteristic of stable cancer (Garcia-Carbonero et al., 2016;
Miller et al., 2015b). These differences mean that it is important to consider G3 tumours
as a separate entity to G1 and G2 tumours when doing investigations in this area.
There are key morphological differences between G3 GEP-NET and G1/G2 GEP-NET.
G3 tumours are poorly differentiated containing pleomorphic cells with atypical nuclei,
abundant mitoses and tumour necrosis, while G1 and G2 tumours are well differentiated
and are comprised of uniform neoplastic cells usually organised into an organoid archi-
tecture with secretory granules containing neuroendocrine markers such as chromogranin
and synaptophysin (Fazio and Milione, 2016).
The focus of this thesis will be on low grade (G1/G2) tumours since although these
tumours are much more common than G3 tumours, it is remains difficult to predict their
disease course due to gaps in our understanding of the biological processes underpinning
the disease pathology of these tumours.
2.2.4. Stage
Disease staging in GEP-NET is done based on the site of the primary tumour. For the
TNM classification and disease staging of SBNET and PNET see Tables 2.2 and 2.4
(Rindi et al., 2007; Rindi et al., 2006). Within Europe the European Neuroendocrine
Tumour society (ENETS) TNM staging system devised in 2006/2007 is used to stage
GEP-NET (Rindi et al., 2006; Rindi et al., 2007). Classification takes into account the
size of the primary tumour and its local extent which is represented by the T while the
presence or absence of locoregional metastases and distant metastases is represented by
the N and M respectively (see imaging, section 2.3.2).
75
Table 2.2.: Disease staging SBNET, table reproduced from (Rindi et al., 2007), creativecommons licence: CC BY-NC
A: TNM classification
TNMPrimary tumor, T
TX Primary tumor cannot be assessedT0 No evidence of primary tumorT1 Tumor invades mucosa or submucosa and size ≤ 1 cmT2 Tumor invades muscularis propria or size > 1 cmT3 Tumor invades subserosaT4 Tumor invades peritoneum/other organs
For any T add (m) for multiple tumorsRegional lymph nodes, N
NX Regional lymph nodes cannot be assessedN0 No regional lymph node metastasisN1 Regional lymph node metastasis
Distant metastasis, MMX Distant metastasis cannot be assessedM0 No distant metastasesM1 Distant metastasis
B: Disease stage
Stage T N MI T1 N0 M0IIA T2 N0 M0IIB T3 N0 M0IIIA T4 N0 M0IIIB Any T N1 M0IV Any T Any N M1
76
Table 2.4.: Disease staging PNET, table reproduced from (Rindi et al., 2006), creativecommons licence: CC BY-NC
A: TNM classification
TNMPrimary tumor, T
TX Primary tumor cannot be assessedT0 No evidence of primary tumorT1 Tumor limited to the pancreas and size < 2 cmT2 Tumor limited to the pancreas and size 2–4 cmT3 Tumor limited to the pancreas and size > 4 cm or invading
duodenum or bile ductT4 Tumor invading adjacent organs or the wall of large vessels
For any T add (m) for multiple tumorsRegional lymph nodes, N
NX Regional lymph nodes cannot be assessedN0 No regional lymph node metastasisN1 Regional lymph node metastasis
Distant metastasis, MMX Distant metastasis cannot be assessedM0 No distant metastasesM1 Distant metastasis
B: Disease stage
Stage T N MI T1 N0 M0IIA T2 N0 M0IIB T3 N0 M0IIIA T4 N0 M0IIIB Any T N1 M0IV Any T Any N M1
77
2.2.5. Functioning syndromes
GEP-NET can secrete a variety of different hormones. When these hormones are secreted
at low levels the tumour is classed as a non-functioning GEP-NET. If however, a particular
hormone is secreted into the bloodstream at sufficient levels to cause a clinical syndrome,
the patient is said to have a functioning GEP-NET. Approximately 25-35 % of GEP-NET
are considered to be functioning tumours (Merola et al., 2016). Functioning GEP-NET
are classified based on the type of hormone they overproduce which results in a particular
syndrome associated with various symptoms in the patient (see Table 2.6).
Carcinoid syndrome, caused by serotonin hypersecretion, is present in around 18 %
of SBNET patients (Modlin et al., 2008). It is characterised by diarrhoea, flushing
and in around 20 % of cases, carcinoid heart disease (Dıez et al., 2013; Merola et al.,
2016). Carcinoid syndrome is rarer in NET arising in other gastrointestinal sites such as
the colon, appendix and rectum. Carcinoid syndrome is present in 2-5 % of pulmonary
carcinoid tumours (Caplin et al., 2015). Carcinoid syndrome is associated with metastatic
disease, about 95 % of patients with carcinoid syndrome have liver metastases (Niederle
et al., 2016).
Insulinomas, characterised by the hypersecretion of insulin (causing hypoglycemia), are
the most common type of functioning NET of the pancreas. The incidence of insulinomas
is 2-4 per million population per year (Oberg and Barbro, 2005; Kizilgul and Delibasi,
2015). They represent 1-2 % of pancreatic neoplasms as a whole (Kizilgul and Delibasi,
2015). Patients with insulinomas experience symptoms such as confusion, sweating and
weakness due to hypoglycemia (Kizilgul and Delibasi, 2015). Approximately 4-6 % of
insulinomas are associated with MEN1 syndrome (Oberg and Barbro, 2005).
Gastrinomas, characterised by gastrin hypersecretion, are the next most common func-
tioning PNET. The incidence of gastrinomas is 0.5-4.0 per million population per year
(Oberg and Barbro, 2005). 20 % of gastrinomas are associated with MEN1 syndrome
(Oberg and Barbro, 2005). The remaining functioning pancreatic NET are very rare.
78
Table 2.6.: Functioning tumours
Tumour Mostcommonprimary site
Hormonehyper-secreted
Syndrome Common symptomsand complications
Functionalcarcinoid
small bowel,lung
serotonin carcinoidsyndrome
diarrhoea, flushing,carcinoid heartdisease, livermetastases
Insulinoma pancreas,duodenum
insulin hypoglycemia confusion, sweating,amnesia, blurredvision, hypoglycemia
Gastri-noma
pancreas,duodenum
gastrin Zollinger-Ellisonsyndrome
diarrhoea,indigestion, ulcers(stomach/duodenal),liver metastases
Somato-statinoma
pancreas,duodenum
somato-statin
somatostati-nomasyndrome
diarrhoea, jaundice,hyperglycaemia ,gallstones, diabetesmellitus, livermetastases
Glucago-noma
pancreas glucagon glucagonomasyndrome
necrolytic migratoryerythema, anaemia,diabetes mellitus,deep vein thrombosis,liver metastases
VIPoma pancreas VIP Verner-Morrisonsyndrome
watery diarrhoea,dehydration,hypokalaemia, livermetastases
VIP: vasoactive intestinal polypeptide
79
These include somatostatinoma, vasoactive intestinal polypeptideoma (VIPoma) and
glucagonoma (see Table 2.6).
The functionality of GEP-NET can impact on patient survival. Insulinomas, for exam-
ple, have a much better prognosis than non-functioning pancreatic tumours (NF-PNET).
The 5 year survival rate of insulinomas has been reported to be as high as 97 % (Oberg
and Barbro, 2005). Conversely NF-PNET are quite aggressive with a median overall sur-
vival of only 38 months, distant metastases are often present at diagnosis in patients with
NF-PNET (Yao et al., 2008; Falconi et al., 2012). While insulinomas rarely metastasise
to the liver, the presence of liver metastases at the time of diagnosis is common in the
other types of functioning GEP-NET such as functioning SBNET and gastrinomas.
The presence of a functional syndrome in GEP-NET patients increases the complexity
involved in the clinical management of these patients, however hormonal symptoms are
well controlled in the majority of patients with the use of somatostatin analogues (see
section 2.3.1).
2.3. Treatment and imaging
2.3.1. Treatment
The therapeutic options in GEP-NET are varied, however there is a low level of evidence
for which treatments should be given to which groups of patients. A recent consensus
conference identified the need for novel biomarkers to be identified to enable the par-
ticular patients who would most benefit from these different treatment modalities to be
identified (Frilling et al., 2014). There is a lack of therapies being developed in GEP-NET
with companion biomarkers which could predict treatment response and allow treatment
monitoring during the patient journey.
Large prospective randomised placebo controlled trials available for only a handful
of the treatments offered in GEP-NET (somatostatin analogues, targeted therapies and
80
peptide receptor radionuclide therapy). Small patient numbers and the heterogeneous
nature of these tumours represent challenges for setting up clinical trials and mean that
international studies and collaborations are usually a necessity to achieve suitable patient
numbers. This presents a conundrum for the study design with a decision needed on
whether to maximise patient numbers by having a more heterogeneous set of patients in
the trial or to have a narrowly defined set of inclusion criteria (Cives et al., 2016).
Surgery
Surgical resection remains the only potentially curative treatment in GEP-NET if the
tumour is localised. Surgery is carried out in both localised and metastatic disease if it is
believed that R0 can be achieved. Surgery can also be used palliatively in patients with
a large disease burden with substantial liver metastases or hormonal syndromes which
can not be controlled medically (Dıez et al., 2013; Modlin et al., 2010).
+++Low grade GEP-NET In G1 and G2 GEP-NET patients, surgical resection
with curative intent is the first treatment consideration even in the presence of liver
metastases (Pavel et al., 2016). In cases where R0 can not be achieved, debulking surgery
is indicated in functional GEP-NET with predominant liver disease since symptom control
can be improved even with a reduction in liver tumour burden of < 90 % (Pavel et al.,
2016). In patients with non-functioning GEP-NET where patients are suffering from
symptoms of mass effect, palliative surgery may be considered if the disease does not
progress over a period of 6 months (Pavel et al., 2016).
In patients with low grade SBNET, surgical resection with the dissection of lymph
nodes along the superior mesenteric root has a high chance of being curative particularly
in the absence of distant metastases (Niederle et al., 2016). Patient survival rate has
been shown to be improved when the regional lymph nodes, that are at risk of metastatic
disease, are dissected in addition to the primary tumour (Landry et al., 2013). Radical
81
surgery results in good 5 and 10 year survival rates in SBNET patients treated at an
early stage in the disease. In one study, the 5 year survival rates for localised (stage I/II)
tumours was 100 % (Tamburrino et al., 2016). In stage III disease the 5 year survival
rate was found to be 97.1 % and in stage IV disease it was 84.8 % (Tamburrino et al.,
2016).
Low grade rectal NET and gastroduodenal NET (ECLomas) are less aggressive than
other GEP-NET, therefore conservative management with endoscopic resection and follow
up (rather than surgical resection) is considered to be sufficient in the majority of cases
(Ramage et al., 2016; Delle Fave et al., 2016).
There is a risk of carcinoid crisis during surgery and post-surgery in patients with
carcinoid syndrome, therefore it is important that the syndrome is controlled with so-
matostatin analogues prior to surgery (Tamburrino et al., 2016). In other functional
GEP-NET, symptoms should also be controlled prior to the surgical resection of liver
metastases (Pavel et al., 2016).
Liver transplantation is a consideration in highly selected, young patients, with a func-
tioning low grade GEP-NET that is resistant to medical therapy, with unresectable liver
metastases but resectable extrahepatic disease (Pavel et al., 2016; Frilling et al., 2014).
Studies in this area are hampered by the very low numbers of GEP-NET patients receiv-
ing liver transplants, they represent 0.2-0.3 % of transplants carried out (Frilling et al.,
2014; Fan et al., 2015).
+++High grade GEP-NET G3 GEP-NET are rare, so there is a paucity of data
on which treatment modalities are most effective with information coming from small
retrospective studies and non-controlled trials (Garcia-Carbonero et al., 2016). Most of
the available data on treatment that is available comes from the more extensive studies of
small cell lung carcinoma (which are much less rare) however it is unclear how relevant this
data may be with respect to the treatment of high grade GEP-NET (Garcia-Carbonero et
82
al., 2016). Curative surgery is usually attempted in localised high grade tumours, however
there is a high rate of relapse so platinum based adjuvant therapy is usually carried out
after surgery (Garcia-Carbonero et al., 2016). In high grade tumours debulking surgery or
surgical resection of distant metastases is not recommended, so treatment in this setting
is usually chemotherapy and radiotherapy (Garcia-Carbonero et al., 2016).
+++Carcinoid heart disease Carcinoid heart disease occurs in around 20 % of
patients with carcinoid syndrome with fibrosis of the right heart valve (Dıez et al., 2013;
Merola et al., 2016). This is associated with poor prognosis and can lead to right sided
heart failure. Historically right sided heart failure was responsible for around 1/3 of
carcinoid syndrome related deaths (Norheim et al., 1987; Druce et al., 2010).
Treatment of carcinoid heart disease is by cardiac surgery with valve replacement along-
side treatment with somatostatin analogues. The 5 year survival rates for carcinoid heart
disease have improved from < 30 % in the 1980s to current levels of around 55 % (Niederle
et al., 2016). This is likely to be the result of successful cardiac surgery and better control
of the carcinoid syndrome.
Somatostatin analogues
Somatostatin analogues (SSA) are analogues of the peptide hormone somatostatin. SSA
are used in patients with functioning GEP-NET to help control the symptoms of syn-
dromes associated with hormone hypersecretion such as carcinoid syndrome. More re-
cently SSA have started to be used in GEP-NET patients with non-functioning tumours
as well due to the discovery that they can slow tumour growth.
Somatostatin was first discovered in 1973 as an inhibitor of the release of growth
hormone (GH) (Brazeau et al., 1973). Somatostatin is a neurotransmitter expressed
in the central nervous system, the gastrointestinal tract, the thyroid glands, adrenal
glands and the endocrine pancreas (Modlin et al., 2006). Somatostatin has an inhibitory
83
effect on many different physiological processes including inhibiting pituitary gland GH
secretion and the secretion of gastrointestinal and pancreatic hormones such as insulin
and glucagon (Oberg and Lamberts, 2016).
Native somatostatin and SSA act by binding to the somatostatin receptor (SSTR),
which is a G protein-coupled seven transmembrane receptor (Baldelli et al., 2014). Ap-
proximately 80 % of GEP-NET express one or more of the 5 human subtypes of the
somatostatin receptor (SSTR1-5) (Baldelli et al., 2014). SSTR are expressed less fre-
quently in high grade GEP-NET and are found on the cell surface at a lower density (if
present) than in low grade GEP-NET (Modlin et al., 2006).
In 1982 the first biologically stable SSA, Octreotide, was synthesised (Bauer et al.,
1982). It was more potent than the native somatostatin at inhibiting GH and insulin
secretion and had a much longer half life (Octreotide: 1.5-1.9 hours, native somatostatin:
3 minutes) (Bauer et al., 1982; Oberg and Lamberts, 2016). Octreotide was subsequently
approved for use in GEP-NET, with multiple daily injections being given to control the
symptoms of carcinoid syndrome (Oberg and Lamberts, 2016). Newer versions of SSA,
Octreotide LAR and Lanreotide Autogel, are far longer acting and allow less intrusive
fortnightly or monthly injections (Baldelli et al., 2014).
It is thought that that the different binding affinities of SSA to the different SSTR
changes the clinical activity of the treatment, so this is being considered in the develop-
ment of novel SSA (Baldelli et al., 2014). Pasireotide (SOM230) is a newer SSA which
has a higher binding affinity to SSTR1-3 and SSTR5 (but not SSTR4) in contrast to Oc-
treotide and Lanreotide which interact primarily with SSTR2 and SSTR5 (Baldelli et al.,
2014). A Phase II (n=44) and a Phase III (n=110) trial have shown that Pasireotide
LAR could control symptoms in GEP-NET patients who were refractory to treatment
with other SSA (Kvols et al., 2012; Wolin et al., 2015; Oberg and Lamberts, 2016).
84
+++Control of symptoms SSA are used to control the symptoms of functioning
syndromes in GEP-NET and so improve the quality of life of patients. As analogues
of somatostatin, they dampen down the biological effect of the hormone hypersecretion
responsible for the symptoms of functioning syndromes in patients.
SSA are used in as a first line therapy in patients with carcinoid syndrome to control
symptoms caused by serotonin hypersecretion (Pavel et al., 2016). In pooled data from
studies of SSA treatment 71 % of patients on octreotide and 75-80 % of patients on
lanreotide had resolution of their diarrhoea and flushing symptoms (Modlin et al., 2006;
Oberg and Lamberts, 2016). SSA are also used to treat the symptoms of functioning
PNET (Pavel et al., 2016). Approximately 80-90 % of patients with glucagonoma or
VIPomas treated with SSA recovered from their diarrhoea and skin rashes (Jensen et al.,
2012; Falconi et al., 2016). SSA are rarely needed in G3 GEP-NET since > 90 % of these
patients have non-functioning tumours (Garcia-Carbonero et al., 2016).
+++Increase in progression free survival More recently studies have shown that
SSA can slow the time to disease progression in GEP-NET patients, in addition to their
role in the control of hormone hypersecretion in patients with functioning tumours (Rinke
et al., 2009; Caplin et al., 2014). These trails have led to SSA being used to control tumour
growth in non-functioning as well as in functioning GEP-NET.
In a double-blind, randomised study of 85 patients with well differentiated, metastatic,
midgut NET, those receiving octreotide had a significant improvement in their median
time to disease progression of 14.3 months compared to 6 months on the placebo arm of
the trial (prospective, phase III, PROMID trial) (Rinke et al., 2009).
A subsequent clinical trial demonstrated that SSA could also improve progression free
survival (PFS) in patients with non-functioning GEP-NET (Caplin et al., 2014). There
were 204 patients with non-functioning, somatostatin receptor–positive GEP-NET in-
cluded in the trail, all with a Ki-67 of < 10 % (double-blind, prospective, phase III,
85
CLARINET trial). The study found a significant increase in PFS in patients treated
with SSA, with a median PFS of 18 months for the placebo arm compared to median not
reached for the lanreotide arm of the trail (crossover on disease progression from placebo
to treatment arm of trail).
Incorporating the results of these phase III trails, the latest ENETS guidelines recom-
mend the use of SSA (octreotide LAR /lanreotide autogel) as antiproliferative first line
systemic therapies to control tumour growth in low grade (G1/G2) intestinal NET (Pavel
et al., 2016). In PNET the ENETS guidelines recommend that SSA can also be used to
control tumour growth if Ki-67 % is < 10 % (Pavel et al., 2016). SSA may also be used
for low grade GEP-NET at other sites if there is a positive SSTR status on imaging (see
section 2.3.2) (Pavel et al., 2016).
There is an ongoing debate about the appropriate cut off level of Ki-67 % for treatment
with SSA. The cut off in the CLARINET trial was < 10 % Ki-67 % but there have been
suggestions that the cut off levels for SSA treatment to control tumour growth should be
set even lower at < 5 % (Pavel et al., 2016; Caplin et al., 2014). Additional clinical trials
will be needed to establish a more defined subgroup of patients who would most benefit
from these the use of SSA to improve PFS.
Second line therapies for the treatment of carcinoid syndrome
SSA remain the first line therapy for the treatment of the symptoms caused by hormone
hypersecretion in carcinoid syndrome (see section 2.2.5). In patients who do not respond
to SSA treatment, options for second line therapies include treatment with interferon-α
or inhibitors of serotonin synthesis.
+++Interferon-α Interferon-α is used as a second line therapy for symptom con-
trol in patients with carcinoid syndrome or functioning PNET which are refractory to
SSA treatment (Pavel et al., 2016; Jensen et al., 2012). Interferon-α leads to symp-
86
tom control in 40-70 % of carcinoid syndrome patients, but it is associated with adverse
events, including fever and weight loss (Dimitriadis et al., 2016). It may also be used
for its antiproliferative effects in advanced G1/G2 intestinal NET (Pavel et al., 2016).
Interferon-α binds to interferon receptors which are expressed by GEP-NET, signalling
via these receptors activates interferon inducible genes leading to the control of hormone
hypersecretion and the inhibition of cell proliferation (Patel et al., 2016).
+++Inhibition of serotonin synthesis The recent development of the oral sero-
tonin synthesis inhibitor, telotristat, adds another possibility to the armamentarium for
symptom control in carcinoid syndrome when it is refractory to SSA treatment. Serotonin
is a major mediator in carcinoid syndrome with a particular role in causing diarrhoea and
carcinoid heart disease (Ito et al., 2016). Small molecule telotristat acts in the periphery
(it can not cross the blood-brain barrier) to inhibit serotonin synthesis by inhibiting the
enzyme tryptophan hydroxylase which converts tryptophan to serotonin (Ito et al., 2016).
Early results from a placebo controlled randomised double blind phase III trial (n=135)
in metastatic carcinoid syndrome patients refractory to SSA, showed a significant decrease
in bowel movement frequency of 35 % on 500 mg of telotristat with minimal side effects
(TELESTAR trial) (Kulke et al., 2015). If approved, this treatment could represent an
additional option for patients with carcinoid syndrome who still experience symptoms of
hormone hypersecretion while on the highest dosage of SSA (Dimitriadis et al., 2016).
PRRT
Peptide receptor radionuclide therapy (PRRT) is a type of radiotherapy used in the
treatment of inoperable or metastasised GEP-NET (Kwekkeboom et al., 2009). PRRT
uses radiolabeled SSA to target β radiation to the GEP-NET. The SSA binds to the
SSTR on the tumour cells where the radionuclide, usually Yttrium-90 (90Y) or lutetium-
177 (177Lu), which is coupled to the SSA, emits beta radiation directly at the site of the
87
tumour (Pusceddu et al., 2016).
For PRRT to be effective there needs to be strong SSTR expression on imaging in all
lesions to be targeted (Pavel et al., 2016).
90Y has an 11 mm pathway in soft tissue and so can be used for larger tumours compared
to the 2 mm pathway of 177Lu which is more useful for irradiating smaller lesions, allowing
less energy to escape into the surrounding tissue (Pusceddu et al., 2016). If there is
a range of tumour sizes a combination of the two radiopharmaceuticals can be used
(Pusceddu et al., 2016). Renal toxicity is less often seen in patients treated with 177Lu and
the radionuclide also emits low energy γ radiation, allowing for scintigraphy to monitor
treatment response, so this radionuclide is becoming more popular than 90Y for GEP-
NET treatment (Pusceddu et al., 2016; Pavel et al., 2016; Zwan et al., 2015).
PRRT is usually used as a second line therapy after medical therapy has failed in
patients with low grade intestinal NET with progressive disease which are SSTR positive
(Pavel et al., 2016). It may also be used in patients with low grade PNET if there is
disease progression after treatment with SSA, targeted drugs or chemotherapy, however
high quality data on the use of PRRT in PNET patients is lacking (Pavel et al., 2016).
Prospective randomised clinical trails will be needed to establish a better evidence base
and recommendations for where PRRT would be best placed in the treatment pathway
for PNET. G3 GEP-NET rarely express SSTR however PRRT treatment is an option in
the small subset of patients who do have SSTR positive tumours.
The first phase III prospective randomised controlled trial in patients with G1/G2
advanced intestinal NET (NETTER-1), showed that treatment with 177Lu-DOTATATE
(octreotate radiolabeled with 177Lu) led to fewer patients having disease progression than
those treated with a high dose of octreotide LAR (60 mg) (Strosberg et al., 2016; Oberg
and Lamberts, 2016). Of the 230 patients included in the trail just 23 patients had
disease progression on radiopharmaceutical treatment (177Lu-DOTATATE) compared to
67 in the octreotide LAR 60 mg treatment arm, with the median PFS not being reached
88
and being 8.4 months respectively (Strosberg et al., 2016).
Systemic chemotherapy
Platinum based systemic chemotherapy is used in G3 GEP-NET as an adjuvant treat-
ment in localised tumours, and in unresectable advanced G3 GEP-NET with distant
metastases, systemic chemotherapy is usually the first line therapy (Garcia-Carbonero
et al., 2016). In this setting, cisplatin or carboplatin and etoposide are usually used
(Garcia-Carbonero et al., 2016). As a second line treatment, irinotecan or oxaliplatin-
based regimens may be used but more studies are needed to assess the effectiveness of
this approach (Garcia-Carbonero et al., 2016).
A study of 305 patients with G3 GEP-NET showed that tumours with a higher Ki-67
% of > 55 % responded better to platinum based chemotherapy (response rate: 42 %)
than tumours at the bottom of the G3 Ki-67 % range with a Ki-67 % of 20-55 % (response
rate: 15 %) (Sorbye et al., 2013). Despite this, the patients who responded less well (but
had lower Ki-67 %) survived 4 months longer (14 months versus 10 months) suggesting
that G3 GEP-NET should not be considered a uniform group.
Systemic chemotherapy is not usually offered to patients with G1/G2 GEP-NET except
in the case of PNET with diffuse liver metastases (Pavel et al., 2016). Systemic streptozo-
tocin (STZ) based chemotherapy therapy, alongside SSA and targeted therapies, is used
in G1/G2 PNET patients with a high tumour burden or rapid tumour progression (≤
6-12 months) (Pavel et al., 2016). Chemotherapy regimens are usually STZ/5 fluorouracil
(5FU) or STZ/doxorubicin (Pavel et al., 2016).
High level evidence is still lacking for the use of different chemotherapy regimes in
advanced GEP-NET due to an absence of the large scale randomised trails done for
SSA and targeted therapies (everolimus/sunitinib). A recent systematic review called
for randomized trials to address this gap and provide better quality data on the relative
efficacy of different chemotherapy regimens, and how they compare to other treatment
89
modalities available to patients with advanced GEP-NET (Lee et al., 2016).
Liver directed locoregional treatments (non-surgical)
Cytoreductive therapies such as transarterial chemoembolisation (TACE), transarterial
embolisation (TAE), radiofrequency ablation (RFA) and radioembolisation with selective
internal radiation therapy (SIRT) are directed at liver metastases may be used in certain
patients with advanced G1/G2 GEP-NET to control tumour growth.
Liver metastases of GEP-NET patients are highly vascular and are fed only by the
hepatic artery, while normal liver tissue has a dual vascular supply (inflow: 70 % portal
vein, 30 % hepatic artery) which makes these tumours well suited to intra-arterial treat-
ment delivery via the hepatic artery (De Baere et al., 2015). Embolisation of arterial
tumour feeders with particles can be combined with chemotherapy (chemoembolisation)
or radiotherapy (radioembolisation) injected into the hepatic artery (De Baere et al.,
2015).
There is no randomised data from trials comparing patients on these treatments to
those receiving debulking surgery or other kinds of cytoreductive therapies (Patel et al.,
2016). This has led to a lack of high level evidence for the use of one type of cytoreductive
therapy over another, or information on which subset of patients would most benefit from
each type of treatment.
+++Chemoembolisation Chemoembolisation, TACE, may be used in the treatment
of liver metastases in G1/G2 GEP-NET patients with unresectable lesions to control tu-
mour growth and to control symptoms (in functioning tumours refractory to SSA treat-
ment) (Patel et al., 2016). Doxorubicin is the most common chemotherapeutic agent
used in this setting (De Baere et al., 2015). 52-86 % of GEP-NET patients have symp-
tomatic response upon chemoembolisation with overall survival being 3-4 years (median
survival: 38.6 months) (De Baere et al., 2015). Further improvements in CT imaging for
90
image guidance such as 3D vascular imaging (cone-beam CT imaging) are likely to allow
improve TACE further (De Baere et al., 2015).
+++Radioembolisation Radioembolisation, SIRT, may be given to patients with
G1/G2 GEP-NET liver metastases if surgery is contraindicated (Pavel et al., 2016). The
latest ENETS guidelines considered SIRT as an investigational method, in the absence
of studies investigating the efficacy of SIRT when compared to TAE alone (Pavel et al.,
2016). SIRT overcomes the limitations of external radiation therapy for the treatment
of liver metastases by delivering, directly to the tumour, sufficient β radiation (usually
from 90Y bound to microspheres) to be tumoricidal while at the same time minimising
the damage to normal liver tissue (De Baere et al., 2015).
+++Thermal ablation RFA can be used for thermal tumour destruction as a sole
therapy to treat small G1/G2 GEP-NET tumours (< 5 cm) in certain non-surgical can-
didates or in combination with surgical treatment of liver metastases (Patel et al., 2016).
Imaging (ultrasound, CT) is required during treatment to guide the probe to the tumour
and deliver the thermal ablation (De Baere et al., 2015). The most suitable patients for
RFA may be those with a low tumour volume, in particular those with a small number of
small metastases that in order to be removed would require a large resection (De Baere
et al., 2015).
Targeted Therapy
Targeted therapies, everolimus and sunitinib, are antiproliferative therapies that are used
in the treatment of GEP-NET patients. The use of oral everolimus and sunitinib is
approved for G1/G2 progressive PNET and they are usually used as a second line therapy
after the failure of SSA or systemic chemotherapy (Pavel et al., 2016). They can also
be used as a first line therapy in patients where systemic chemotherapy or SSA are not
appropriate (Pavel et al., 2016).
91
In G1/G2 intestinal NET everolimus is used as a second line therapy in patients who
are refractory to SSA treatment or as a third line therapy in patients that do not respond
to PRRT (Pavel et al., 2016). The potential efficacy of sunitinib in progressive advanced
intestinal NET is being still being investigated (ongoing SUNLAND randomised placebo
controlled trial, NCT01731925) (Pavel et al., 2016).
The highest level evidence on treatment comes from the large prospective randomised
placebo controlled trials, these have been done for targeted therapies (everolimus, suni-
tinib) and SSA (octreotide LAR /lanreotide autogel) in GEP-NET but are lacking for
other treatment modalities (Pavel et al., 2016).
+++Everolimus Everolimus works by inhibiting mammalian target of rapamycin
complex 1 (mTORC1) thus preventing the mTOR pathway, which is unregulated in
a high proportion of GEP-NET, from promoting proliferation (Briest and Grabowski,
2014).
The treatment of G1/G2 PNET with everolimus was assessed in a randomised, double-
blind, placebo controlled phase III clinical trial (RADIANT-3, NCT00510068) (Yao et al.,
2011). The trial included 410 patients with G1/G2 advanced PNET and disease progres-
sion. There was significantly increased PFS for patients in the everolimus treatment arm
of 11 months compared to 4.6 months for the placebo arm of the trail. At 18 months it
was estimated that the percentage of patients who were alive and had not progressed was
34 % for those treated with everolimus compared to 9 % for those on placebo (patients
on placebo who had disease progression were switched to everolimus treatment). With
respect to adverse events, stomatitis was a common occurrence in the everolimus treat-
ment group (64 % of patients) as were rashes and diarrhoea, more serious adverse events
were rarer and included pneumonitis and anaemia.
Everolimus treatment of G1/G2 advanced progressive NET of the gastrointestinal tract
(n=175), lung (n=90) and unknown primary (n=36) was assessed in a randomised, double
92
blind, phase III clinical trial (RADIANT-4, NCT01524783) (Yao et al., 2016). Median
PFS was higher in the treatment arm of the trial (11 months) than the placebo arm (3.9
months).
The effects of everolimus treatment in G3 GEP-NET patients as a second line therapy
after platinum based systemic chemotherapy are being investigated in an ongoing phase II
clinical trial (EVINEC trial, NCT02113800) since this subgroup of patients were excluded
from earlier everolimus trials (Merola et al., 2016).
There is an ongoing randomised phase III study into systemic chemotherapy (STZ/5-
FU) versus everolimus treatment in advanced progressive PNET, the crossover study
design switches the patients to the alternative treatment modality when they experience
disease progression (SEQTOR, NCT02246127) (Pavel et al., 2016). This trial should help
to address the questions regarding the best sequence for these therapies to be given in
and also identify which treatment or sequence of treatment may be the most potent.
+++Sunitinib Sunitinib is a multiple receptor tyrosine kinase (RTK) inhibitor (Pusceddu
et al., 2016). It inhibits vascular endothelial growth factor receptors (VEGFR) 2 and
3 which drive angiogenesis, stem cell growth factor receptor Kit (SCFR or c-kit) and
platelet-derived growth factor receptors (PDGF-R) α and β which drive cell prolifera-
tion, all of which are highly expressed in metastatic PNET (Raymond et al., 2011).
Sunitinib treatment was investigated compared to placebo in a randomised, double-
blind phase III clinical trial of 171 advanced G1/G2 PNET patients with disease progres-
sion (NCT00428597) (Raymond et al., 2011). Median PFS was higher in the treatment
arm of the trial at 11.4 months, compared to 5.5 months for the placebo arm. 10 %
of patients had died at the end of the study in the sunitinib treatment arm of the trial
compared to 25 % on placebo. Adverse side effects associated with treatment included
diarrhoea and nausea, and the trial was ended early due to the higher number of deaths
and serious adverse events in the placebo arm of the trial.
93
+++Novel targeted therapies Investigations into the effectiveness of additional tar-
geted therapies in GEP-NET are the focus of ongoing prospective randomised clinical
trials. These include, multiple RTK inhibitors (axitinib, sorafenib, pazopanib), a mon-
oclonal antibody against VEGF (bevacizumab) and an mTOR inhibitor (temsirolimus)
(Abdel-Rahman, 2014; Pusceddu et al., 2016; Hobday et al., 2015; Phan et al., 2015).
Oncolytic virus
One future possible treatment for GEP-NET is virotherapy. An oncolytic virus was
engineered which could selectively kill neuroendocrine tumour cells in-vitro and in-vivo,
while leaving healthy cells intact (Leja et al., 2007; Leja et al., 2010; Leja et al., 2011; Yu
et al., 2011). An added advantage of this approach is that the viral lysis of the tumour
cells stimulates the immune system, usually repressed during tumourigenesis, to attack
the tumour.
Various modifications were made to the Adenovirus serotype 5 (Ad5) to enable it to
target only tumour cells, while attenuating any effects on normal cells. A gene required
for the replication of the virus (adenoviral E1A) was modified so that it was under the
control of the Chromogranin A (CgA) promoter (Leja et al., 2007). This meant that the
virus could only produce the E1A protein products in cells expressing CgA. Since these
proteins are required for viral replication, the Ad5 could replicate and lyse the tumour
cells but not the normal surrounding tissue.
Since liver metastases are common in GEP-NET patients, Ad5 was also modified to
enable it to successfully target and lyse the tumour cells in liver metastases, while repli-
cation was arrested in healthy hepatocytes, minimising the liver toxicity of the treatment
(Leja et al., 2010; Yu et al., 2011).
Promising cell line and mouse studies led to an ongoing human stage I/IIa clinical trial
of the AdVince therapy at Uppsala University Hospital, Sweden (Randle et al., 2013;
Essand et al., 2016). The clinical trial was partially paid for by crowdfunding, with
94
donations raised from thousands of individuals through the oncolytic virus fund and the
rest of the money provided by a high-net-worth individual with a neuroendocrine tumour
(Essand et al., 2016). It was the largest amount of money raised by crowdfunding for a
randomised controlled clinical trial (Sharma et al., 2015).
The online crowdfunding of the clinical trial raises interesting questions for the future
with respect to how clinical trials are paid for in rare diseases where there may be a
public interest in the therapy becoming available but the therapy is not backed by a
pharmaceutical company. It is likely that in order for these kinds of approaches to
succeed in the future widespread media coverage will be needed to raise public awareness
among potential donors (as indeed was the case with the oncolytic virus fund).
2.3.2. Imaging
Morphological imaging with computed tomography (CT) and/ or Magnetic resonance
imaging (MRI) is typically used in the diagnostic work up for patients with GEP-NET
and for TNM staging and follow up. Nuclear medicine imaging techniques are used to
provide additional disease staging information and to identify if a patient would benefit
from PRRT therapy with radiolabelled SSA (Essen et al., 2014).
Morphological imaging
+++CT Triple-phase contrast-enhanced multi-slice CT is recommended for GEP-
NET imaging (Niederle et al., 2016). Hundreds of 2D transverse x-ray images per second
are produced and assembled into clinical anatomical images, with images being taken
both prior to and during intravenous injection of iodine based contrast medium (Essen
et al., 2014). Normal liver tissue receives around 75 % of its blood supply from the
portal vein whereas liver metastases receive only arterial blood from the hepatic artery
(Essen et al., 2014). Liver lesions are frequently hypervascular on MRI (hyperattenuat-
ing, appearing bright in the late arterial phase) but they can also be hypovascular (low
95
attenuating, appearing dark in the portal venous phase) (Essen et al., 2014).
The mean sensitivity and specificity of CT for detecting liver metastases is 82 % and 92
% respectively (Essen et al., 2014). For the detection of abdominal and thorax GEP-NET
metastases, CT has a mean sensitivity 83 % and the mean specificity 76 % (Essen et al.,
2014).
+++MRI MRI with gadolinium-based contrast medium is recommended for GEP-
NET imaging (Niederle et al., 2016). MRI uses a strong magnetic field (1.5 and 3.0
teslas) to align the spin of the protons in the body in the direction of the magnetic field
(Essen et al., 2014). Varying radio waves are applied which change the alignment of the
protons and when these are turned off, the protons in the body return to their normal
spin at different rates, producing a radio signal that can be reconstructed to produce
images of the body (Essen et al., 2014). As with CT, images are taken prior to and after
intravenous injection of contrast medium (Essen et al., 2014).
For PNET detection the mean sensitivity and specificity of MRI is 93 % and 88 %
respectively (Essen et al., 2014). For liver metastases MRI has been found to be superior
to CT for detection and follow up, with the added benefit of a better safety profile due
to the lack of ionising radiation (Pavel et al., 2012; Tan, 2011). MRI is also considered
to be the superior imaging modality for the detection of tumours in solid visceral organs
such as the pancreas (Tan, 2011).
+++Ultrasound Conventional ultrasound uses high frequency sound waves (10 or 12
MHz) to produce images of internal organs (Niederle et al., 2016). A key drawback of
ultrasound when compared to MRI/CT is that it is investigator dependant and contrast
agents are rarely used, which limits the reliability and sensitivity respectively (Niederle
et al., 2016). The use of ultrasound is limited in GEP-NET due to a lower sensitivity
than that of MRI and CT for the detection of liver metastases, with a sensitivity of 68
% (Maxwell and Howe, 2015).
96
Endoscopic ultrasound (EUS) involves the addition of an ultrasound transducer to an
endoscope so that it can be used to produce ultrasound images from inside the gas-
trointestinal tract (Essen et al., 2014). For PNET imaging the pancreas is examined by
pressing the endoscope against the walls of the stomach and duodenum (Essen et al.,
2014). EUS is the most sensitive imaging modality for PNET diagnosis with a detection
rate of around 90 %, however its use is limited by lack of availability in many centres
(Pavel et al., 2012; Essen et al., 2014).
Nuclear medicine imaging
During nuclear medicine imaging, small amounts of radioactive tracers are injected or
ingested, the tracers accumulate in particular areas of the body and release radiation at
these locations which is detected to produce an image (Essen et al., 2014). This can be
superimposed onto CT scans so that the areas of tracer accumulation can be mapped
onto an anatomical image of the body.
Radionuclides used in the functional imaging of GEP-NET include indium-111 (111In)
which emits γ radiation and gallium-68 (68Ga) which emits positrons, β radiation. The
radionuclide is linked to a biologically active molecule which is usually a SSA for GEP-
NET functional imaging. The resulting radiotracer or radiolabelled ligand is injected
intravenously and radiation is emitted from the sites where the ligand binds. The radia-
tion detected is processed to produce an image identifying areas with high uptake of the
radiotracer.
In somatostatin receptor based imaging the radionuclide is linked to a SSA which binds
with high affinity to SSTR2 receptors. SSTR are present in 60-100 % of NET with the
most highly expressed SSTR subtype being SSTR2 (85 % are SSTR2) (Frilling et al.,
2014). They are expressed on the cell surface of the majority of well differentiated GEP-
NET and their metastases but rarely in G3 tumours (Niederle et al., 2016; Maxwell and
Howe, 2015). The uses of somatostatin receptor based imaging include the detection and
97
localisation of G1/G2 GEP-NET primaries and their metastases, tumour staging, follow
up and to identify patients who would benefit from PRRT treatment (Maxwell and Howe,
2015).
In G3 GEP-NET, radionuclide fluorine-18 (18F) linked to glucose analogue FDG can
be used for functional imaging (18F-FDG PET/CT) (Dıez et al., 2013). This technique
is of limited usefulness in well differentiated lesions since these usually lack the increased
glucose metabolism present in G3 lesions (Essen et al., 2014; Niederle et al., 2016).
+++111In Radionuclide 111In linked to a SSA is used in GEP-NET for somatostatin
receptor scintigraphy (SRS) (also called octreoscan). This is done with single photon
emission computed tomography (SPECT) which is processed to produce a 3D image of
the gamma radiation (rather than the planar image from SRS alone) or SPECT/CT to
improve the localisation of tumours (Maxwell and Howe, 2015). Intravenous injection of
the radiotracer is followed by image acquisition after 4 hours and 24 hours.
SRS can identify lesions missed on CT or MRI, a study showed it could identify new
lesions in 28 % of patients that were missed by morphological imaging (Maxwell and
Howe, 2015). Sensitivities range from 46 % to 100 % for imaging abdominal NET with
a wide variation between studies, probably due to differing protocols and selection of
patients, and an overall sensitivity of 78 % (Essen et al., 2014). For liver metastases
sensitivities range from 49 to 91 % (Maxwell and Howe, 2015).
Smaller tumours are more difficult to detect which limits the sensitivity of this imaging
technique due to the occurrence of false negative results for small lesions. In a compara-
tive study of SRS to MRI and CT for the identification of metastases, SRS sensitivities
were significantly correlated with median metastasis size (Dromain et al., 2005). A high
sensitivity SRS result was seen in the study in only 22 % of patients with a small median
metastasis size (< 7 mm) compared to 64 % of patients with a larger median metastasis
size (> 15 mm).
98
+++68Ga The radionuclide 68Ga linked to a SSA is a more recently developed imag-
ing modality in GEP-NET. The β radiation (positrons) released at the sites that the
radiotracer binds to are imaged by positron emission tomography (PET). This technique
is combined with CT images for better anatomical resolution (Niederle et al., 2016).
In 68Ga DOTA-PET/CT the radiotracer binds to SSTR2 and the 68Ga in the radio-
tracer undergoes positive β decay emitting a positron. The positron travels a very short
distance through the tissue, then looses energy and annihilates when it hits an electron,
producing a pair of γ photons. The gamma photons pass out of the body and are detected
by the PET scanner.
A key advantage of 68Ga DOTA-PET/CT is the increased sensitivity when compared
to SRS, with fewer false negative results. For 68Ga DOTA-PET/CT the limitation of
detection is in millimeters compared to 1 cm or more for SRS (Maxwell and Howe,
2015). This is because for 68Ga DOTA-PET/CT radiation is measured from two photons
simultaneously, resulting in superior spatial resolution to (SRS)-SPECT which directly
measures the gamma radiation emitted by only one photon (Maxwell and Howe, 2015).
Another benefit of 68Ga DOTA-PET/CT is the single time point for image acquisition,
60 minutes after the intravenous injection of the radioactive tracer.
Multiple comparative studies have demonstrated the superiority of 68Ga DOTA-PET/CT
over SRS (Maxwell and Howe, 2015; Gabriel et al., 2007; Niederle et al., 2016). 68Ga-
DOTA-TOC was found to have significantly better sensitivity at 97 %, than (SRS)-
SPECT with a sensitivity of 52 % (the specificity was 92 % for both imaging modalities)
(Gabriel et al., 2007). In SBNET 68Ga imaging could change management in 20-30 % of
patients and was especially useful for the detection of small lesions (Niederle et al., 2016).
It also has a high chance of determining the primary tumour in cases of GEP-NET with
unknown primary (Maxwell and Howe, 2015).
The use of 68Ga DOTA-PET/CT remains limited due to reduced availability compared
to SRS (Dıez et al., 2013). A study investigating the costs involved in these two imaging
99
modalities showed that 68Ga DOTA-PET/CT was rather more cost effective than 111In-
octreotide, with total costs of 548 euros and 827 euros respectively, when material and
personnel costs were considered (Schreiter et al., 2012).
Summary
MRI and/or CT imaging is recommended to assess if GEP-NET liver metastases can
be resected (Pavel et al., 2012). Despite improvements in imaging techniques for liver
metastases, 50 % still remain undetected on preoperative imaging when compared to thin
slice pathological examination (Frilling et al., 2014).
In a study comparing different types of imaging for GEP-NET liver metastases, MRI
was found to be the superior modality for detection, with a 95.2 % sensitivity compared
to 78.5 % for CT and 49.3 % for octreoscan (Dromain et al., 2005). The study also found
that MRI detected additional liver lesions in patients which were missed by the other
imaging modalities.
MRI is the superior imaging modality for the detection of GEP-NET primaries and
metastases in solid visceral organs such as the liver and pancreas, however, CT is more
effective for detecting tumours in hollow organs such as SBNET and their lymph node
metastases (Maxwell and Howe, 2015). Despite this, CT remains cheaper and more widely
available than MRI so it is more frequently performed in patients with a GEP-NET (Essen
et al., 2014).
In SBNET to localise the primary tumour, CT and/or MRI is recommended followed
by 68Ga DOTA-PET/CT (or SRS SPECT/CT if 68Ga DOTA-PET/CT is not available)
(Niederle et al., 2016). 68Ga DOTA-PET/CT is recommended for the staging and locali-
sation of non-insulinoma PNET patients where it can change the management in 20-55 %
of cases (Falconi et al., 2016). 68Ga DOTA-PET/CT is also recommended when available
to assess the resectability of liver metastases (Frilling et al., 2014).
68Ga DOTA-PET/CT outperforms SRS on many metrics including, increased sensitiv-
100
ity, with the ability to identify smaller lesions, and increased convenience due to the single
time point for image acquisition. In the future this imaging modality should become more
widely available especially if it does indeed prove to be more cost effective.
2.4. Neuroendocrine cells
Neuroendocrine cells secrete bioactive peptides and amines in response to neuronal, chem-
ical or mechanical input. Neuroendocrine cells are scattered throughout the body form-
ing the diffuse neuroendocrine system and are found in virtually all organs in vertebrates
(Hofmann et al., 2013).
There are thought to be at least 17 different types of neuroendocrine cells within the
GEP system alone (Schimmack et al., 2011). They are found interspersed with epithelial
cells or scattered in subepithelial linings or stroma (Hofmann et al., 2013).
Neuroendocrine cells have mixed morphological and physiological features in common
with both neuronal and endocrine regulatory systems (Schimmack et al., 2011). For
example, neuroendocrine cells express synaptophysin (also found in synaptic vesicles at
neuronal synapses) as well as being involved in the synthesis and secretion of bioactive
peptides and amines.
Multiple different products can be produced by individual neuroendocrine cells. These
are stored in dense core secretory granules which contain high concentrations of peptides
for future secretion, when the correct signal is received. These products can have diverse
functions depending on location where the bioactive peptides or amines are released and
the type of receptor present on the cell membrane of target cells.
The type of signalling can also vary and a single signalling molecule released by a neu-
roendocrine cell can have a plethora of different effects. These include paracrine signalling
(local signalling between cells within the same organ), endocrine signalling (long distance
signalling between organs), autocrine signalling (feedback loop, with signalling via re-
101
ceptors on the neuroendocrine cell of origin) and neurotransmitter and neuromodulatory
roles (Tischler, 1989).
Somatostatin, released from neuroendocrine δ cells for example, can act as both a
paracrine and an endocrine signalling mediator. Somatostatin secreted by pancreatic δ
cells binds to SSTR on local pancreatic α and β cells (paracrine signalling) leading to an
inhibitory effect on glucagon and insulin secretion (Hauge-Evans et al., 2009; Caicedo,
2013). In contrast somatostatin released by neuroendocrine cells is also secreted from
the pancreas in pancreatic juice. This is released into the lumen of the duodenum where
it signals in an endocrine manner to suppresses the hormone secretion from other neu-
roendocrine cells and the nutrient absorption activity of the gut (Arimura and Fishback,
1981).
Neuroendocrine cells are able to integrate signals both from other neuroendocrine cells
and from neurones as well as from physical and chemical changes in the gut. This enables
the synthesis and secretion of bioactive peptides and amines to occur in an intricate and
finely tuned manner. These cellular products in turn enable digestion to take place
and regulate this process. A network of intercellular feedback pathways and autocrine
signalling helps to maintain homoeostasis.
2.4.1. Development
Neuroendocrine cells were first identified as a distinct entity during the 1960s. Studies
to identify the cell responsible the production of the peptide hormone calcitonin led to
the discovery that it was produced by the thyroid follicular cells, subsequently named C
cells (Bussolati and Pearse, 1967; Tischler, 1989; Cutz, 1982). These studies identified
the shared characteristics of the C cells with markers expressed that were also present in
pancreatic islets and membrane bound secretory granules. Markers of peptide production
were present, and in some cell types there were mechanisms for aromatic amine precursor
uptake and decarboxylation (Pearse, 1966; Tischler, 1989).
102
In 1966, Pearse hypothesised, due to these characteristics, that these types of cells
could arise from a common ancestral cell that might have migrated to the gut from the
neural crest (Pearse, 1966; Tischler, 1989). This led to controversy for many years as
different groups tried to demonstrate the validity of Pearse’s hypothesis.
The findings of these studies were that the C cells of the thyroid glands described as
producing calcitonin by Pearse do indeed arise from the neural crest and this is similar
to the differentiation of some other amine and polypeptide producing endocrine cells
(Schonhoff et al., 2004). These cells differentiate at an early stage in development and
turnover very slowly (Schonhoff et al., 2004). Investigations into the development of amine
and polypeptide secreting cells of the gastrointestinal and pancreatic system however,
showed that this was not the case for neuroendocrine cells of the digestive system.
Studies were done in which the neural crest of the Japanese quail was grafted into
chick embryos with excised neural crests (Tischler, 1989). Cells that originated from
the Japanese quail could be identified from the chick cells at different stages in chick
development due to their distinctive nuclear morphology (target-like nuclei caused by a
dense central mass of heterochromatin that is absent in the chick cells) (Tischler, 1989).
These embryonic cell tracing studies demonstrated that the only amine and polypeptide
producing cells that originated from the neural crest (ectoderm) were the thyroid C cells,
the adrenal medulla, the extra adrenal paraganglia and cells of the myenteric plexus and
sympathetic ganglia (Tischler, 1989).
The amine and polypeptide producing cells of the gastrointestinal tract and the pan-
creas were shown instead to be derived from the endoderm (analogous to enterocytes)
(Schonhoff et al., 2004). Neuroendocrine cells of the gastrointestinal tract, rather than
migrating from the neural crest during development, instead arise from local, tissue spe-
cific, multipotential stem cells (Schimmack et al., 2011). These are found at the base of
the crypts of the intestine or in the neck of gastric glands and give rise to all regional
epithelial cell types (Schimmack et al., 2011; Jenny et al., 2002).
103
Subsequent studies identified the presence of cycling columnar stem cells at the very
base of intestinal crypts which express the protein, leucine rich repeat containing G
protein-coupled receptor 5 (LGR5), with 4-6 of these cells being present per crypt (Barker
et al., 2007; May and Kaestner, 2010). LGR5 is a target of Wnt signalling which maintains
the proliferative activity of the intestinal crypt under physiological conditions (May and
Kaestner, 2010).
Lineage tracing studies in mice showed that all intestinal epithelial lineages including
neuroendocrine cells are derived from crypt base stem cells positive for Lgr5 (Barker et
al., 2007). In further experiments, individual Lgr5+ cells tagged with green fluorescent
protein (GEP) from transgenic mice, sorted by flow cytometry into wells (1 cell per well)
were able to grow into organoids in vitro (Sato et al., 2009). These organoids contained
crypt-villus structures and all terminally differentiated small intestinal cell types.
Cell lineage tracing studies have shown that pancreatic neuroendocrine cells differ-
entiate from transient endocrine progenitor cells of the proximal trunk domain of the
pancreatic epithelium during development (Kim et al., 2015b). The proximal trunk do-
main also gives rise to pancreatic ductal cells and more studies are needed to identify if
there is a common pancreatic progenitor cell for pancreatic ductal and neuroendocrine
cells or if the region contains a heterogeneous population of cells with a predefined lineage
(Kim et al., 2015b).
2.4.2. Differentiation
New intestinal neuroendocrine cells are produced throughout life by differentiation from
a reservoir of stem cells in the crypts and migrate up the villi to replace the turnover
of mature neuroendocrine cells (Schonhoff et al., 2004; Schimmack et al., 2011). Mature
cells at the villi tips are thought to undergo apoptosis and be extruded into the gut lumen
(Wang et al., 2016).
In kinetic studies adult mice were injected with 3H-thymidine so that the radioactive
104
thymidine was incorporated into the DNA with each cell division, to produce radiolabelled
cell nuclei (Cheng and Leblond, 1974). The mice were sacrificed at different time points so
that the intestinal crypts could be examined. Crypt base columnar stem cells are able to
phagocytose nearby non-viable cells resulting in phagosomes appearing in their cytoplasm
which contain nuclear material from the phagocytosed cell (in the presence of radioactive
thymidine these phagosomes become radiolabelled). In their elegant experiment Cheng et
al were able to utilise this cellular process and the radiolabeling of phagosomes to observe
the differentiation of the crypt cells into enterocytes, paneth, goblet and neuroendocrine
cells, the 4 main types of mature intestinal cells.
In these experiments, exposure to radiation caused many more phagosomes to be ob-
served in the cells at the base of the crypts 6 hours after injection than were observed
under physiological conditions (Cheng and Leblond, 1974). At this time point the vast
majority of the labelled phagosomes were in the crypt bases with only a single phago-
some in the mid crypt region, enabling the fate of the crypt cells to be followed over
time. At 12 hours after injection in addition to the radiolabelled phagosomes at the base
of the crypts, these were also present in partially differentiated mid crypt columnar cells
and oligomucous cells (oligomucous cells differentiate into goblet cells). By time point
30 hours, the phagosomes were observed in terminally differentiated enterocytes, paneth
and neuroendocrine cells. This led the authors to conclude that neuroendocrine cells dif-
ferentiate from common precursor pluripotential stem cells at the base of the intestinal
crypts to partially differentiated cells in the mid crypt region and then fully differentiated
cells.
It was shown that the turnover process undergone by goblet, enterocytes and neuroen-
docrine cells lasts 3-4 days, with the cells migrating upwards from the crypts to the mid
crypt region where they differentiate into mature fully differentiated cells which migrate
gradually up to the villi tips (Cheng and Leblond, 1974; Schonhoff et al., 2004). Paneth
cells instead migrated downwards, and persisted for a longer period of around 16 days in
105
total (Cheng and Leblond, 1974).
Some of the transcription factors needed to commit a cell during embryonic devel-
opment to becoming a neuroendocrine cell, as opposed to the other intestinal cell types
originating from the common crypt precursor cells, have been identified. More studies are
however needed to identify the precise sequence of events required for the differentiation
from the common precursor into the many individual types of terminally differentiated
intestinal neuroendocrine cells such as enterochromaffin cells and δ cells. More work is
also needed to identify the transcription factors and processes required for neuroendocrine
differentiation of individual intestinal neuroendocrine cell types both during development,
and for their replenishment as they turnover in adulthood.
Much of the studies that have been done on the differentiation of neuroendocrine cells
have been in murine models, particularly in mouse knockout models to determine gene
function and carry out lineage tracing studies. The transcription factors involved in this
process are highly conserved between mammals and patterns of their expression between
mice and humans are similar, making the mouse a useful model for investigating devel-
opment and cell fate during differentiation. Limitations remain however, since mouse
studies do not represent the whole picture and differences remain which could confound
experimental findings from these studies and limit their usefulness in understanding hu-
man biology.
Where available, data is included on human diseases arising from loss of function
mutations in these transcription factors, in order to provide a more detailed picture so
that comparisons can be made between the findings in human studies and mouse models.
Complete or near complete loss of function mutations in these transcription factors in
humans have been very infrequently identified in the literature, they are likely to be
extremely rare due to being deleterious to the survival of the foetus during an early stage
of development. This limits the possibilities for functional investigations of these genes
in humans.
106
The reduction in the cost of whole genome sequencing and projects to sequence large
human populations, for example the 100,000 genomes project sequencing the genomes of
100,000 NHS patients with rare diseases in the UK are likely to lead to the identification
of additional as yet unidentified gene mutations (Genomics England, 2017). This will
identify areas of interest for further study and lead to a better understanding of the
function of these genes in human biology and various disease states.
Early observations that neuroendocrine cells had certain features in common with
neurones have gained a new significance with more recent discoveries. These showed
that despite digestive tract neuroendocrine cells being shown to arise from the endoderm
(rather than the neuroectoderm) one of the characteristics they do share with neurones
is that their differentiation is regulated by the same transcription factor gene family
as the differentiation of neuronal cells (Li et al., 2011; Srivastava et al., 2013). These
transcription factors come from the basic helix-loop-helix transcription factor family and
contain two α helices with a loop connecting them and a DNA binding region.
Important transcription factors involved in neuroendocrine differentiation from the
basic helix-loop-helix family are Protein atonal homolog 1 (MATH1) encoded by the
MATH1 gene (also know as ATOH1), Neurogenin-3 (NGN3) encoded by the NEUROG3
gene and protein neurogenic differentiation factor 1 encoded by NEUROD1 (also known
as BETA2 ).
These three transcription factors are expressed sequentially during intestinal neuroen-
docrine differentiation, with negative regulation of this process being provided by the
Notch signalling pathway.
Early Transcription Factors
+++MATH1 During development, expression of transcription factor Math1 in pre-
cursor cells within the intestinal crypts of mice directs them to a secretory lineage. This
commits the cells to differentiate into either paneth (secrete antimicrobial peptides),
107
goblet (secrete gel forming mucins) or neuroendocrine cells (secrete amines/polypetides)
(Schonhoff et al., 2004). Math1 expression was absent however in the pancreas and
stomach, suggesting that it is not important for the development of neuroendocrine cells
within these organs (Yang et al., 2001).
Cell lineage studies demonstrated in mice that when the β-galactosidase gene (LacZ )
was under the control of the Math1 promoter, cells expressing Math1/LacZ became
paneth, goblet or intestinal neuroendocrine cells (Yang et al., 2001; Schonhoff et al.,
2004; Hsu, 2015). Enterocyte development was independent of this transcription factor
and instead is promoted by Notch signalling (May and Kaestner, 2010). These findings
suggest the presence of a common progenitor cell which expresses Math1 and differentiates
to produce the intestinal secretory lineages.
Math1 knockout mice (-/-) demonstrated that in the absence of this transcription
factor, paneth, goblet and intestinal neuroendocrine cells failed to develop (Yang et al.,
2001). The development of the fourth main intestinal epithelial cell type, absorptive
enterocytes, was unaffected. MATH1 does not appear to have a role in the differentiation
of pancreatic neuroendocrine cells and it is not expressed in the pancreas (Yang et al.,
2001). The pancreatic and duodenal homeobox 1 transcription factor (PDX1 ) is required
for the development of the pancreas and is later involved in the maturation pancreatic
β and δ cells by transactivating the genes for insulin and somatostatin (Ohisson et al.,
1993; Miller et al., 1994).
Neurog3 (-/-) mice, do express Math1 while the converse is not true of Math1 knockout
mice (-/-) which do not express Neurog3 (Li et al., 2011). This suggests that these
genes are expressed sequentially with Neurog3 being one of the downstream targets of
Math1 during the differentiation of neuroendocrine cells. Growth factor independent 1
transcriptional repressor (GFI1) is another transcription factor downstream of MATH1
which acts as a negative regulator of neuroendocrine differentiation, instead promoting
the production of goblet and paneth cells (Shroyer et al., 2005; Li et al., 2011).
108
Taken together these findings suggest that MATH1 expression in a common precursor
cell represents an essential first step in a sequence of events leading to development of
the intestinal secretory cell lineages including the intestinal neuroendocrine cells. Notch
signalling is an important negative regulator of neuroendocrine differentiation within the
digestive tract.
+++Notch and HES1 The Notch signalling pathway is involved in the regulation
of the development of neuroendocrine cells by providing a brake on neuroendocrine dif-
ferentiation. Differentiating neuroendocrine cells increase their expression of the Notch
ligand which binds to Notch receptors on the cell surface of adjacent cells, thus activat-
ing downstream Notch signalling in those adjacent cells and providing a limiting factor
for neuroendocrine differentiation, since these adjacent cells will then go on to become
non-neuroendocrine cell types (Schonhoff et al., 2004).
Knockout mice for one of the downstream targets of activated Notch, hairy enhancer
of split 1 (Hes1 ) (-/-) mice had an 3-7 fold increase in the number of neuroendocrine
cells in the stomach and small intestine (Jensen et al., 2000; Schonhoff et al., 2004).
There was an increase in the expression of Math1 in the intestine and Neurog3 in the
pancreas of the mice (Jensen et al., 2000). This provides further evidence of the role of
the Notch signalling pathway and HES1 as a negative regulator of neuroendocrine differ-
entiation. It also demonstrates how the signalling cascades generated by the expression
of the different basic helix-loop-helix transcription factors involved in the differentiation
of neuroendocrine cells are nuanced and tissue specific.
+++NGN3 NGN3, like MATH1, is a member of the family of basic helix-loop-helix
transcription factors. It is also negatively regulated by the Notch signalling pathway (Li
et al., 2011). Knockout mice for the Neurog3 gene (-/-) (which encodes NGN3) lack
all pancreatic and intestinal neuroendocrine cells and die postnatally of diabetes and
malabsorptive diarrhoea (Kim et al., 2015b; Gradwohl et al., 2000).
109
In the stomach of Neurog3 (-/-) mice however, only some of the neuroendocrine cell
types are missing. Neurog3 expression is required for the development of stomach somato-
statin and gastrin secreting cells but not for the development of serotonin and histamine
secreting cells within the stomach (Heller et al., 2005; Li et al., 2014). This suggests that
differing regulatory processes governing the development of these particular neuroen-
docrine cells exist based on location, with intestinal but not stomach serotonin producing
cells requiring NGN3.
In humans, loss of function mutations in the gene encoding NGN3 lead to the loss
of intestinal neuroendocrine cells and congenital malabsorptive diarrhoea. Congenital
malabsorptive diarrhoea caused by autosomal recessive mutations in NEUROG3 is a rare
condition in humans that demonstrates the important role of NGN3 in the development
of digestive neuroendocrine cells in humans. It was first described in 3 patients in 2006
(Wang et al., 2006). Additional cases have since been identified (Aksu et al., 2016; Pinney
et al., 2011). The 3 patients investigated by Wang et al had homozygous point mutations
in NEUROG3 predicted to cause the amino acid change R107S in one patient and R93L
in the other two (Wang et al., 2006). These are mutations in important regions of the
protein for DNA binding and the activation of various genes downstream of NGN3.
NGN3 transactivates the transcription of NEUROD1/BETA2 another transcription
factor from the basic helix-loop-helix transcription factor family. Site directed mutage-
nesis studies by Wang et al, with transient transfections in HeLa cells using a luciferase
reporter under the control of the Neurod1/Beta2 promoter, demonstrated that cells har-
bouring the mutations seen in patients (R107S or R93L) were unable to activate Neu-
rod1/Beta2 expression (Wang et al., 2006).
Small bowel biopsies taken from the 3 patients revealed normal villi architecture, paneth
cells, goblet cells and enterocytes but a lack of pancreatic and intestinal neuroendocrine
cells. Only a single cell was found, out of the 350 small bowel cypts examined, with posi-
tive immunohistochemical staining for CgA, and also for serotonin, the cell had abnormal
110
morphology). No neuroendocrine cells expressing synaptophysin, gastrin, somatostatin
or vasoactive intestinal polypeptide were found. In contrast 5 or 6 neuroendocrine cells
were found per small intestinal crypt in control normal mucosa included in the study.
Two of the patients went on to type 1 diabetes during childhood by 8 years of age
(the 3rd patient died unexpectedly of sepsis at 3 years of age) (Wang et al., 2006). An
additional patient was identified with a homozygous mutation causing a truncation of
the NGN3 protein (E123X) and this patient developed diabetes at an even earlier age
(neonatally) (Pinney et al., 2011). These findings suggest that NGN3 may be important
in humans for proper islet development and function, as is the case in mice, however,
additional studies would be needed to determine if this was the case.
To date, the studies of congenital malabsorptive diarrhoea with NEUROG3 muta-
tions have not investigated the detailed islet morphology, or the presence or structure of
neuroendocrine cells in the pancreatic islets, since the focus of the studies has been on
intestinal abnormalities. Studies of this nature, and those that identify the presence or
absence of diabetes in additional cases of NEUROG3 mutations would be very beneficial
to investigate the differences that seem to be present between the human and mouse
studies.
It may be that while NGN3 appears to be essential for the development of functional
neuroendocrine cells in the intestine, there is some redundancy in its function in the
determination of the fate of pancreatic neuroendocrine cells (with this same redundancy
not being present in mice). An alternative theory is that a complete absence of NGN3 (as
with the Neurog3 -/- mice) is required to prevent human islets from developing all together
while the mutations seen in humans may still retain some low level of NGN3 functionality
sufficient for some islet function to be present. Human foetuses with more deleterious
NEUROG3 mutations may not survive to term and if so would not be identified in
the literature. Possible support for this theory comes from the truncation mutation
causing a much earlier onset of the diabetes (Pinney et al., 2011). More work needs to
111
be done to investigate the nuances of the role of NGN3 for both intestinal and pancreatic
neuroendocrine development and function.
Lineage tracing studies in transgenic mice enable terminally differentiated daughter
cells to be traced from their earlier parent progenitor cells. These studies have shown
that local NGN3+ cells are early progenitors of pancreatic and intestinal neuroendocrine
cells (Jensen et al., 2000; Gradwohl et al., 2000; Pinney et al., 2011; Jenny et al., 2002;
Gu et al., 2002).
One such experiment used a tamoxifen inducible Cre-ER system in transgenic mice
to investigate which cells expressed Neurog3 during development (Gu et al., 2002).
The study showed that NGN3+ cells that were present at mouse embryonic days E8.5
and E12.5 later developed into cells that expressed either insulin, glucagon, pancreatic
polypeptide or somatostatin.
The presence of NGN3 is required for the expression of later important transcription
factors involved in digestive neuroendocrine differentiation including NEUROD1/BETA2,
with Neurog3 (-/-) mice lacking Neurod1/Beta2 expression (Jenny et al., 2002; Schonhoff
et al., 2004).
Overall the findings of studies investigating NGN3 suggest that it is required in humans
to commit local stem cells in the intestine (and possibly to a lesser extent in the pancreas)
to a neuroendocrine fate. It is not however required for intestinal goblet cell, paneth cell or
enterocyte development. These studies suggest that there is a common NGN3+ intestinal
progenitor cell which differentiates into the diverse range of neuroendocrine cells of the
intestine.
Later transcription factors
More recent studies have identified other transcription factors required later on in the
differentiation of gastrointestinal and pancreatic neuroendocrine cells. These tend to have
a high level of temporal and site specificity, with their function depending on the context
112
in which they are expressed.
Neurod1/Beta2 is needed for the terminal differentiation of intestinal neuroendocrine
cells secreting cholecystokinin (I cells) and secretin (S cells), with Neurod1/Beta2 (-/-)
mice lacking these cell types (Naya et al., 1997). The mice also were found to have very
few pancreatic β cells and died perinatally.
Studies in mice have shown that winged helix transcription factors, forkhead box A1
(Foxa1 ) and forkhead box A2 (Foxa2 ), are functional transactivators of the glucagon gene
and are needed for the proper functioning of mature pancreatic α and β cells, Foxa2 is
required for the terminal differentiation of α cells (Masson et al., 2014). These studies had
to be done using conditional knockouts in particular tissues rather than global knockouts
due to the early lethality in null mice. (Foxa1 (-/-) mice die from hypoglycaemia and
Foxa2 (-/-) mice die from neural tube patterning defects (Ye and Kaestner, 2009).
Mouse studies were done investigating a conditional knockout of both Foxa1 and Foxa2
in the small bowel and the colon only (Ye and Kaestner, 2009). The null mice were found
to have no cells expressing glucagon like peptide 1 and 2, and had reduced numbers of cells
expressing somatostatin and peptide YY in the small bowel and colon. There was also
a reduction in goblet cell numbers with the aberrant expression of the different mucin
genes. The investigators found reduced levels of another neuroendocrine transcription
factor, paired box 6 (PAX6) mRNA, suggesting that Foxa1/a2 act upstream of Pax6
expression in the transcription factor cascade in regulating the differentiation of intestinal
neuroendocrine cells.
In murine studies, the NK2 homeobox 2 transcription factor (Nkx2.2 ) was found to be
expressed in the intestine, pancreas and central nervous system both during development
and adulthood (Mastracci et al., 2013). Nkx2.2 acts downstream of Neurog3 and is needed
during embryonic development for cell fate determination of intestinal and pancreatic
neuroendocrine cells (Gross et al., 2016). It is also required for the terminal differentiation
and function of serotonin secreting enterochromaffin cells (Gross et al., 2016).
113
Mice with global Nkx2.2 deletions die postnatally due to hyperglycaemia, they lack
pancreatic β cells and have fewer α and pancreatic polypeptide cells (Mastracci et al.,
2013). The study showed that lineage specification was disrupted in the intestines of the
mice, with reduced levels of cells secreting serotonin, somatostatin and glucose-dependent
insulinotropic peptide (also called gastric inhibitory polypeptide).
In another study of Nkx2.2 knockout mice, null mice had a significant reduction in
the expression of gastrin, glucagon, cholecystokinin, gastric inhibitory polypeptide, neu-
rotensin, somatostatin and serotonin (Desai et al., 2008). The expression of peptide YY
however, was similar to that in the wild type mice and there was only a small reduction
in secretin expression suggesting that the presence of Nkx2.2 may not be needed for the
embryonic linage specification of these cell types.
Global knockout Nkx2.2 mice had massively increased numbers of ghrelin+ cells in
both the pancreas and the intestine, suggesting that these cells are being promoted at
the expense of the other neuroendocrine cell types (under physiological conditions, the
pancreas only contains a subpopulation of ghrelin+ cells during embryogenesis) (Mas-
tracci et al., 2013; Desai et al., 2008).
Conditional knockouts enabled the function of Nkx2.2 to be investigated in adult mice
(Gross et al., 2016). When the gene was deleted in the the duodenum and colon of adult
mice, serotonin secreting cells were severely reduced in number, however cholecystokinin
and secretin secreting cells were unaffected and ghrelin secreting cells were increased.
This suggests that Nkx2.2 expression has an ongoing function in adults regulating the
specification of intestinal neuroendocrine cell subtypes, in particular promoting the dif-
ferentiation of serotonin+ cells as the intestinal mucosa turns over. This is in contrast
to its broader role during embryogenesis, where it regulates the cell fate determination
of the majority of intestinal neuroendocrine cells.
A transcription factor acting downstream of Nkx2.2 was identified in the study, LIM
homeobox transcription factor 1 alpha (Lmx1a), which had reduced expression in the
114
Nkx2.2 null mice (Gross et al., 2016). Lmx1a regulates the expression of the rate limiting
enzyme for serotonin synthesis, tryptophan hydroxylase 1 (Tph1 ), and this enzyme was
also found to be downregulated in the mutant mice.
PDX1 is needed for the maturation and function of β and δ cells. It transactivates
the genes for insulin, glucose transporter 2 and islet amyloid polypeptide in β cells and
the gene for somatostatin in δ cells (Ohisson et al., 1993; Zhou et al., 2014; Miller et al.,
1994). In addition to being necessary for the formation of the pancreas from the proximal
duodenum during the development of the foetus, it also needed for gastrin cell (G cell)
maturation (Schonhoff et al., 2004). Pdx1 (-/-) mice lack these gastrin secreting cells
(Schonhoff et al., 2004).
Pdx1 (-/-) mice were found to have an approximately 60 % reduction in the numbers of
neuroendocrine cells expressing secretin (S cells), serotonin (enterochromaffin cells) and
cholecystokinin (I cells) in the proximal duodenum when compared to Pdx1 (+/+) mice
(Offield et al., 1996). Interestingly, the numbers of these 3 different cell types present in
rest of the intestine of the Pdx1 (-/-) mice was found to be normal. This illustrates one
of the ways in which the development and differentiation of neuroendocrine cells is highly
context specific, with the differentiation of neuroendocrine cells with the same secretion
products in different parts of the gastrointestinal tract being regulated differently.
This is likely to be due to different early transcription factors being expressed during
the fetal development of these organs, with each triggering the expression of different
downstream cascades of later transcription factors. Pdx1 expression is important for
proximal duodenal and pancreatic development while conversely Math1 expression directs
early cell lineage specification in the small bowel and the colon.
NK6 Homeobox 1 (NKX6-1) has been shown in mice to be both required and sufficient
to specify β cell lineage, with the lack of Nkx6-1 expression in mice being sufficient to con-
vert β cells and their precursor cells into δ like cells (Schaffer et al., 2013; DiGruccio et al.,
2016). Conversely, ectopic expression of Nkx6-1 in endocrine precursor cells ensured that
115
they differentiated into β cells only, at the expense of other pancreatic neuroendocrine
cells which were absent (Schaffer et al., 2013).
NKX6-1 was also shown to transcriptionally repress aristaless related homeobox (Arx )
expression (Schaffer et al., 2013). ARX is involved in α cell differentiation and is also
expressed in mature α cells, with null mice lacking these cells (Courtney et al., 2013;
Heller et al., 2004).
ARX has an opposing function to paired box 4 (PAX4) in the determination of the
destiny of neuroendocrine precursor cells in the pancreas, with Pax4 expression triggering
differentiation into a β or δ cell lineage rather than an α cell lineage (Courtney et al.,
2013). This differentiation process is regulated by the antagonistic function of these
two proteins which compete with each other by triggering the down regulation of the
expression of the alternative gene in pancreatic neuroendocrine precursor cells.
Knockout mice for Pax4 lack pancreatic β and δ cells and have fewer duodenal sero-
tonin, cholecystokinin, peptide YY, gastric inhibitory polypeptide and secretin expressing
cells (Heller et al., 2005; May and Kaestner, 2010). There are no changes however in the
numbers of neuroendocrine cells in the ileum or colon (May and Kaestner, 2010). In
the stomach, gastrin producing cells were found to be unaffected but there were reduced
numbers of somatostatin and serotonin expressing cells (May and Kaestner, 2010).
Another paired box transcription factor Pax6, has also been shown in mice to be
involved in duodenal and stomach neuroendocrine development. Knockout mice have
reduced numbers of neuroendocrine cells expressing gastric inhibitory polypeptide in the
duodenum, somatostatin and gastrin in the stomach and insulin, glucagon, pancreatic
polypeptide and somatostatin in the pancreas (May and Kaestner, 2010; Heller et al.,
2005). One of the functions of Pax6 is as a transcriptional transactivator of the genes for
glucagon, insulin and somatostatin (Heller et al., 2004).
The function of the basic helix-loop-helix transcription factors, NGN3 and MATH1 in
the early differentiation of neuroendocrine cells is quite well understood however this is
116
not the case for the plethora of later transcription factors being identified in different
neuroendocrine cells. Studies have shown that these transcription factors have negative
and positive regulatory effects on each other in a very location specific manor to enable
different cellular lineages to emerge. The roles of later transcription factors and their
interactions with each other both during embryonic development as well as their on
going function in subsets of mature neuroendocrine cells needs to be further investigated.
The development of further mouse models with conditional transcription factor knock-
outs, to knockout genes at particular time points during development in different in cell
types, will provide useful functional information about the role of these transcription
factors both during neuroendocrine development and in mature neuroendocrine cells.
Traditional gene knockouts for these transcription factors usually exhibit embryonic or
perinatal lethality, therefore the majority of earlier studies have not investigated their
function in mature neuroendocrine cells. This is of particular importance in the small
intestine where neuroendocrine cells are rapidly turning over and being replaced. The
development of 3D organoid cultures could also provide an additional useful model for
studying this process and if developed from human tissue could prove very useful in
determining if the findings in mouse models hold true in human tissues.
Further studies will enable the identification of the timings, locations and cell clusters
in which different transcription factors need to be expressed, for neuroendocrine cells
secreting a particular main secretory product to become terminally differentiated and
fully functional. A better understanding of the biology underpinning these processes
could be used in the future for the development of novel treatments for GEP-NET,
diabetes, obesity and other gastrointestinal diseases.
Neuroendocrine plasticity
Considerable plasticity exists in neuroendocrine cells, even after they have undergone ter-
minal differentiation they can retain the ability to change their morphology and hormone
117
secretion patterns in response to certain signals in their microenvironment (Tischler,
1989). An example of this is Roux-en-Y gastric bypass surgery, with regrowth of the in-
testine, with the remaining part increasing 2-3 fold in size and demonstrating very little
change in the numbers and density of neuroendocrine cells (Engelstoft et al., 2013).
Recent studies in adult mice have shown that when β cells are destroyed, α cells will
spontaneously convert into β cells (Schaffer et al., 2013). Selective inhibition in adult α
cells of the Arx gene alone was sufficient to convert them into β like cells, so this gene may
be behind the functional plasticity demonstrated in neuroendocrine cells demonstrated
when the β cells were ablated in the previous study (Courtney et al., 2013; Friedman-
Mazursky et al., 2016).
The plasticity of neuroendocrine cells and their diverse functions adds to the complexity
involved in understanding the details of how the diffuse neuroendocrine system operates
under physiological conditions and in different disease states.
2.4.3. Function
At least 17 types of neuroendocrine cells have been identified in the GEP system (Schim-
mack et al., 2011). They secrete either peptide hormones or monoamine neurotransmit-
ters, which are important signalling molecules involved in the regulation of digestion.
Functions range from causing the contraction/relaxation of smooth muscle to regulate
peristalsis and the rate that food travels through the stomach and intestines to stimulat-
ing/inhibiting the production of gastric acid and digestive enzymes (see Table 2.7).
118
Table 2.7.: Neuroendocrine cells of the GI tract and pancreasCell type Main cell
locationSecretionproduct
Type, length inamino acids
Function
EC cell GI tract,pancreas
serotonin* monoamineneurotransmitter,1∼
stimulates smooth muscle in the gut to contractaround food, increases intestinal motility,stimulates mucus secretion
ECL cell stomach histamine* monoamineneurotransmitter,1∼
stimulates gastric acid secretion by parietal cells
δ cell GI tract,pancreas
somato-statin
peptide hormone,14; 28
inhibits hormone secretion (see *), slows digestionby reducing smooth muscle contractions andintestinal blood flow
β cell pancreaticislets
insulin* peptide hormone,51
produced in response to high blood glucose,promotes glucose absorption andglycogenesis/lipogenesis
α cell pancreaticislets
glucagon* peptide hormone,29
produced in response to low blood glucose,promotes glycogenolysis in the liver and glucoserelease
PP cell pancreaticislets
PP* peptide hormone,36
produced in response to food intake, reduces rateof gastric emptying and appetite
VIP cell GI tract,pancreas
VIP* peptide hormone,28
increases glycogenolysis, relaxes smooth muscle inthe stomach, inhibits secretion of gastric acid
G cell stomach,duodenum
gastrin* peptide hormone,14; 17; 34
stimulates gastric acid secretion by parietal cells,reduces rate of gastric emptying
ghrelin cell stomach,duodenum
ghrelin* peptide hormone,28
produced prior to a meal/during fasting,increases appetite, increases GI motility, reducesinsulin secretion
S cell duodenum,jejunum
secretin* peptide hormone,27
inhibits secretion of gastric acid and gastrin,regulates water homeostasis
119
Continuation of Table 2.7Cell type Main cell
locationSecretionproduct
Type, length inamino acids
Function
K cell duodenum,jejunum
GIP* peptide hormone,42
incretin, increases insulin secretion (when bloodglucose high), increases glucagon secretion (whenblood glucose low)
L cell duodenum,jejunum,
GLP-1* peptide hormone,30; 29
incretin, increases insulin secretion, inhibitsglucagon secretion, reduces rate of gastricemptying and acid secretion
ileum,colon
PYY* peptide hormone,36, 34
produced after a meal, reduces appetite, inhibitsgastric motility,
I cell duodenum,jejunum
cholecys-tokinin*
peptide hormone,8; 22; 33; 58
produced after a meal, inhibits gastric emptying,stimulates pancreatic digestive enzyme andgallbladder bile salt release
M cell duodenum,jejunum
motilin* peptide hormone,22
produced during fasting, promotes interdigestivemigrating contractions clearing the GI tract ofdebris
N cell ileum neu-rotensin*
peptide hormone,13
produced in response to dietary fat, increasesfatty acid absorption, stimulates histamine release
GI: gastrointestinal, EC: enterochromaffin, ECL: enterochromaffin like, GIP: gastric inhibitorypolypeptide, PP: pancreatic polypeptide, VIP: vasoactive intestinal peptide, GIP:glucose-dependent insulinotropic peptide, *: secretion inhibited by somatostatin, ∼: decarboxylated
120
The majority of neuroendocrine cells are known to secrete multiple bioactive products
and the same neuroendocrine cell type can have differing functions depending on their
anatomical location and different physiological conditions (Tischler, 1989). For example
in addition to secreting the peptide hormone insulin, pancreatic β cells also secrete an-
other peptide hormone, islet amyloid polypeptide (also known as amylin), from the same
secretory granules (Zhang et al., 2014). The concentration of islet amyloid polypeptide is
around 1-2 % that of insulin, and it is thought to slow gastric emptying and suppresses
glucagon secretion (Cao et al., 2013).
Depending on the type of neuroendocrine cell and the location of that particular cell,
it will release hormones into the extracellular space and/or the gut lumen and capillary
network. The hormones will then bind to and activate receptors on the surface of local
and/or distant target cells, leading to intracellular signal transduction. Many hormones
will bind to several types or families of receptors on their target cells, for example sero-
tonin and somatostatin bind to multiple different receptors. δ cells negatively regulate
the secretions of all of the other gastrointestinal and pancreatic neuroendocrine cells by
secreting the peptide hormone somatostatin (see Table 2.7).
Many of these signalling molecules are pleiotropic, having functions within the nervous
system as neurotransmitters and neuromodulators, while within the endocrine system
they act as hormones (Alzugaraya et al., 2016). An example of a molecule with this
dual functionality is serotonin. The vast majority of serotonin, 95 %, is produced by
gut neuroendocrine cells, where it has many different functions involved in the regulation
of digestion (Berger et al., 2009). In contrast, less than 1/1,000,000 central nervous
system (CNS) neurones make serotonin, nevertheless all brain areas express receptors for
serotonin allowing it to modulate the majority of brain functions (Berger et al., 2009).
Depending on the location of the target cell within the body and the specific receptors
expressed by that cell, the same hormone can have different effects, for example sero-
tonin signalling can have either excitatory or inhibitory effects depending on the specific
121
receptor the serotonin binds to.
Secretion products and receptors
The secretion products of neuroendocrine cells are processed from the trans golgi network
into large dense core vesicles and small synaptic like vesicles where they are stored prior
to calcium dependant exocytosis (Schimmack et al., 2011). The peptide hormones and
amines are usually be processed into individual secretory granules however in some types
of neuroendocrine cell, secretory products co-localise within the same secretory granule
(Schimmack et al., 2011).
Several components of the secretory machinery of neuroendocrine cells are utilised in
the histopathological diagnosis of a GEP-NET, since tumours arising from neuroendocrine
cells usually retain some neuroendocrine features. These include CgA and synaptophysin
(see section 2.6.1). CgA forms the soluble core of dense core secretory granules and
regulates their biogenesis by inducing budding from the trans golgi network (Giovinazzo
et al., 2013; D’amico et al., 2014). It is found across the neuroendocrine and nervous
system. Chromogranin B (CgB) promotes the aggregation mediated sorting of peptides
into secretory granules (Schimmack et al., 2011). Synaptophysin is a synaptic vesicle
glycoprotein found in neuroendocrine cells and neurones. The biochemical function of
this protein remains elusive since there appears to be some redundancy in its function
which has confounded functional investigations, despite it being highly evolutionarily
conserved (Adams et al., 2015).
+++Peptide hormones The main secretion product of neuroendocrine cells is most
frequently a peptide hormone, for example α, β, and δ cells of the pancreas secrete the
peptide hormones glucagon, insulin and somatostatin respectively (see Table 2.7).
Peptide hormones are a common type of hydrophilic signalling molecule. They are
made up of a short chain of amino acids, usually less than 50 amino acids long, with an
122
amine group at one end of the chain and a carboxyl group at the other end.
Peptide hormones are made by the cell during mRNA translation in the rough en-
doplasmic reticulum (ER). The initial inactive form, preprohormone, is made up of a
larger chain of amino acids containing an N terminal signalling sequence, the hormone
itself and linking amino acids. Further processing occurs in the rough ER including the
removal of the signalling sequence and in the golgi apparatus where some also undergo
glycosylation. The resulting prohormone is packaged into vesicles and superfluous amino
acids are cleaved prior to secretion, producing the active peptide hormone (Schimmack
et al., 2011).
For example, the inactive form of insulin is preproinsulin, this is translocated to the
rough ER lumen where the signalling molecule is cleaved. Proinsulin is folded and disul-
phide bond formation occurs within the rough ER to generate the correct tertiary struc-
ture (Weiss et al., 2000). Proinsulin is transported to the golgi apparatus where it is
packaged into vesicles (Weiss et al., 2000). Prior to secretion the linker C-peptide region
of proinsulin which joins together the two insulin chains (A chain, B chain) is cleaved
with the release of the C-peptide (Weiss et al., 2000). This produces the mature, active,
form of insulin, a heterodimer with the A and B chains now linked together only by two
disulphide bonds between cysteine amino acids (Weiss et al., 2000).
Studies in hydra and other metazoa (animalia) have suggested that neuropeptides
were the first type of transmitters of intracellular signals and that they appeared early
in evolution, they are present in a wide range of different animals (Alzugaraya et al.,
2016). Studies in the phylum Cnidaria (which includes species of jellyfish) found that
neuropeptides act on epithelial muscle cells to enable coordinated muscle movements and
in the hyrda gastrovascular cavity they induced peristalsis movements (Alzugaraya et al.,
2016)
123
+++Monoamine neurotransmitters Certain types of neuroendocrine cells secrete
small hydrophilic signalling molecules known as monoamine neurotransmitters instead
of the larger peptide hormones (see Table 2.7). A monoamine neurotransmitter is much
smaller than a peptide and they are usually made from a single amino acid that has been
decarboxylated to remove the -CO2 group. They are called amines due to having an
amine group (-NH2).
The enterochromaffin like (ECL) cells of the stomach synthesise and secrete histamine,
a monoamine neurotransmitter that is made by the decarboxylation of the amino acid
histidine by the enzyme histidine decarboxylase.
Enterochromaffin (EC) cells are found throughout the gastrointestinal tract where they
synthesise and secrete serotonin (5-hydroxytryptamine (5-HT)), a monoamine neuro-
transmitter. Serotonin is synthesised by EC cells from the amino acid tryptophan. This
involves an enzymatic reaction catalysed by the enzymes tryptophan hydroxylase and
aromatic amino acid decarboxylase (Best et al., 2010).
+++Receptors on target cells The receptors for the signalling molecules secreted
by GEP neuroendocrine cells are mostly 7 transmembrane domain, G protein coupled
receptors (GPCR), as is the case for glucagon, histamine, somatostatin, gastrin, secretin
pancreatic polypeptide (PP) and VIP (Alzugaraya et al., 2016).
In GPCR, the ligand binding triggers a conformational change in the receptor which in
turn activates the attached G protein by facilitating the exchange of guanosine diphos-
phate (GDP) for guanosine triphosphate (GTP). The activated G protein (with GTP
attached) disassociates from the GPCR and activates further intracellular signal trans-
duction pathways via second messengers such as cyclic AMP (cAMP). GPCR remain
a popular drug target, with 36 % of all therapeutics targeting these receptors (Rask-
Andersen et al., 2011).
In the case of serotonin there are 6 families of serotonin receptors (5-HT1−6), one of
124
these families, 5-HT3, are ligand gated cation (Na+ and K+) channels while the other
families are GPCR (Berger et al., 2009). The binding of a ligand to a ligand gated cation
channel causes the channel to open, allowing the cations to enter the cell and plasma
membrane depolarisation.
Insulin binds to insulin receptors (IRA-B) and with a lower affinity, to insulin like
growth factor 1 (IGF-1) receptors (Boucher et al., 2014). These are receptor tyrosine
kinases containing a single transmembane domain that can form homodimers (IRA, IRB)
or heterodimers (IRA/B) which are linked together by disulphide bonds. The binding of
insulin to the α extracellular chains of the receptor triggers a conformational change in
the receptor and autophosphorylation of tyrosine residues in the intracellular β subunits
(Lee and Pilch, 1994). This initiates a chain of intracellular phosphorylation events with
the activation of the Ras-MAPK pathway, or the PI3K/Akt/mTOR pathway (Boucher
et al., 2014).
It is common for the same signalling molecule to bind two or more different types of
receptor. This greatly increases the number and complexity of the different biological
processes that can be triggered by the release of peptide hormones or biogenic amines
from neuroendocrine cells. It also enables the signalling effects to vary depending on the
location within the GEP system, since different types of target cells at various locations
will express different receptors or different numbers of a particular receptor on the cell
surface. In addition, the composition of the receptors on the surface of a particular cell
is not static but can also change over time as the receptors turnover and are recycled in
cellular endosomes (Bowman et al., 2016).
EC cell
EC cells are found scattered throughout the gut mucosa, particularly in the crypts. They
represent 0.25 – 0.50 % of the total mucosa volume (Schimmack et al., 2011). They are
responsible for the secretion of 95 % of the serotonin found in the body (Berger et al.,
125
2009).
Serotonin was discovered and characterised by Vittorio Erspamer in Italy in the 1930s,
he gave it the name enteramine and described it as being the main secretory product of
the chromium staining EC cells (Erspamer, 1957; Whitaker-Azmitia, 1999; Wang et al.,
2017).
Serotonin released by EC cells regulates a wide range of gastrointestinal processes
including peristalsis, visceral pain and the regulation of blood flow (O’Mahony et al.,
2015). Serotonin also has a role in mucosa protective mechanisms for example stimulating
mucosal bicarbonate secretion in response to the lumen acidification in the duodenum
(Hansen and Witte, 2008).
Other neurotransmitters synthesised and secreted by EC cells in addition to sero-
tonin include, melatonin and substance P (Hansen and Witte, 2008; Grun et al., 2015).
Melatonin (N-acetyl-5-methoxytryptamine) is synthesised by EC cells from serotonin. It
functions to protect the intestines from damage by endogenously produced oxygen free
radicals it does this by activating their reduction and inhibiting nitric oxide synthesis
(Chojnacki et al., 2013). Substance P is a peptide hormone involved in the regulation
of smooth muscle contractions, vascular permeability and intestinal immune function
(O’Connor et al., 2004).
+++Mechanosensors and chemosensors EC cells act as mechanosensors and chemosen-
sors within the digestive system. They detect changes in the intestinal milieu and respond
to these chemical changes and to mechanical changes (such as lumenal distension) by se-
creting serotonin.
Recent studies suggest that the mechanogated channel, piezo-type mechanosensitive
ion channel component 2 (PIEZO2) could be the primary mechanotransducer in EC
cells, with serotonin being released in response to distension of the intestinal wall being
detected by PIEZO2 on the luminal surface of the EC cell (Wang et al., 2017; Linan-Rico
126
et al., 2016; Galligan, 2017).
Chemical signals in the gut lumen that are detected by EC cells include the pres-
ence of free monosaccharides, amino acids, fatty acids as well as peptides and nucleotide
triphosphates such as adenosine triphosphate (ATP) and uridine triphosphate (UTP)
(Linan-Rico et al., 2016).
+++Serotonin release When food is ingested, bolus induced pressure on the intesti-
nal wall and the presence of glucose or other nutrients in the gut lumen are both directly
detected by EC cells which have a border made up of microvilli that project into the lu-
men (Linan-Rico et al., 2016; Gershon, 2004; Hansen and Witte, 2008). Local EC cells in
the area then secrete serotonin from their basolateral membrane into the lamina propria
beneath the intestinal epithelium.
Serotonin that is released in response to mechanical stimulation binds to 5-HT3 and/or
5-HT4 receptors on the nerve processes of neurones of the enteric nervous system (Furness
et al., 2004). Neurones stimulated by serotonin include the intrinsic primary afferent
neurons embedded in the submucosa and neurones embedded in the smooth muscle of
the gut lining and these stimulate the motility reflexes of the bowel in response to the
ingested food (Furness et al., 2004).
+++Sensory transducers EC cells (and other neuroendocrine cells) are needed as
sensory transducers. This is because there are no intraluminal or intraepithelial enteric
nerve endings, therefore EC cells, with their mucosal location, represent an essential
intermediary for the detection of changes in the gut lumen which require a response from
the enteric nervous system (Gershon, 2004).
EC cells constitutively secrete serotonin, however they secrete it at much higher levels
after a meal (Gershon, 2004). Specific functions of serotonin in a particular context are
dependent on the type and number of serotonin receptors present on target cells in the
local area where serotonin is released. Due to the rapid turn over of cells in the intestinal
127
epithelium, with cells being lost from the tips of the villi into the gut lumen, neurons are
separated by a variable distance from EC cells and therefore do not form morphologically
recognisable junctions with them (Gershon, 2004).
Accumulation of serotonin and excessive overflow into the portal venous system and
gut lumen is prevented by mucosal epithelial cells and platelets, which take up serotonin
via the sodium dependent serotonin transporter (SERT) on their cell surface (Gershon,
2004; Hansen and Witte, 2008; Costedio et al., 2007). This enables the inactivation of
serotonin signalling and modulates the bioavailability of this key signalling molecule.
+++Serotonin receptors To date, 15 different serotonin receptors have been identi-
fied in humans. Of these, 7 have been identified as having a gastrointestinal localisation
and function (all except for 5-HT1P are also expressed in the CNS). These serotonin
receptors are 5-HT1A (enteric nervous system), 5-HT1P (jejunum), 5-HT2A (gut smooth
muscle), 5-HT2B (stomach fundus, myenteric nerves, colon smooth muscle), 5-HT3 (en-
teric neurones, smooth muscle cells, primary afferent neurons), 5-HT4 (enteric neurones,
smooth muscle cells) and 5-HT7 (smooth muscle cells) (O’Mahony et al., 2015).
Serotonin binding can trigger differing responses depending on the receptor, with differ-
ent receptors having have differing functions. When activated, 5-HT2B stimulates smooth
muscle contractions, in contrast 5-HT7 triggers the relaxation of smooth muscle and 5-
HT1A stimulates mast cells to release histamine (O’Mahony et al., 2015). The number of
different receptors increases the complexity involved in serotonin signalling within differ-
ent gastrointestinal localisations and adds nuance to the regulatory role of serotonin in
digestion.
When it comes to functional studies of serotonin and investigating and treating pathol-
ogy arising in EC cells, this can prove challenging. This is because exogenous serotonin
receptor agonists and antagonists will frequently have cross target reactivity (Bornstein,
2012; Costedio et al., 2007). This means that there are difficulties in ensuring that the
128
receptor of interest is targeted during experimental studies due to several different recep-
tor subtypes being present in similar localisations or on the same cells. This represents
a key challenge for determining the different functions serotonin has when it binds to
particular receptors. Abnormal serotonin secretion has been implicated as a contribut-
ing factor in inflammatory bowel disease, irritable bowel syndrome and diarrhoea in the
setting of bacterial toxin induced enterocolitis and in diarrhoea triggered by platinum
based chemotherapy in addition to the carcinoid syndrome seen in GEP-NET patients
(Linan-Rico et al., 2016).
Tumours that arise from EC cells or their precursor cells have the ability to produce
large amounts of serotonin and other hormones. As the tumour grows, these serotonin
producing tumour cells increase in number. This can lead to the development of car-
cinoid syndrome which has symptoms caused by excessive serotonin secretion including
diarrhoea, abdominal pain and flushing.
δ cell
δ cells are found scattered throughout the gastrointestinal system and are also found
in the pancreatic islets. They release somatostatin which has a dampening effect on
hormone secretion from other types of neuroendocrine cells. For example, the release of
somtatostatin from δ cells in pancreatic islets inhibits the secretion of both insulin and
glucagon from β and α cells respectively (DiGruccio et al., 2016). The colocolisation of
these different neuroendocrine cell types within the pancreatic islets allows for coordinated
secretion or inhibition via efficient paracrine signalling and feedback pathways.
+++Inhibition of neuroendocrine cell secretion Somatostatin released from δ
cells negatively regulates the secretion of the main secretory products of many other
neuroendocrine cells. These include EC cells (serotonin), G cells (gastrin), ECL cells
(histamine), β cells (insulin), α cells (glucagon), digestive VIP producing cells (VIP), S
129
cells (secretin) small intestinal M cells (motilin) and I cells (cholecystokinin) (Schimmack
et al., 2011). This enables somatostatin to have a widespread inhibiting effect on di-
gestive processes by inhibiting endocrine and exocrine secretion, gastrointestinal motility
(reduced smooth muscle contractions) and absorption (reduced intestinal blood flow).
These effects of somatostatin form the basis of the treatment of the symptoms of
functioning GEP-NET with SSA. For example, in patients with carcinoid syndrome SSA
bind to SSTR2 receptors on tumour cells attenuating the excessive serotonin secretion
which causes the syndrome (Schimmack et al., 2011).
Somatostatin is released from the basolateral side of δ cells and travels to other nearby
neuroendocrine cells or alternatively enters the capillaries to travel to more distant gas-
trointestinal cell targets. Somatostatin secretion is regulated by neuronal signalling. In
the stomach for example, the neuropeptide gastrin releasing peptide (GRP) is released by
postganglionic fibres of the vagus nerve and binds to GRP receptors on the basolateral
side of δ cells, triggering somatostatin release (Watson et al., 2006). Conversely post-
ganglionic cholinergic muscarinic nerves release the neurotransmitter acetylcholine onto
the δ cells where it binds to muscarinic acetylcholine receptors which triggers negative
regulation of somatostatin secretion (Takeuchi et al., 2016).
Somtatosatin secretion by stomach δ cells leads to an inhibition of gastic acid secretion
through somatostatin binding to SSTR2 on both ECL cells and parietal cells (Takeuchi
et al., 2016). This inhibits the release of hydrogen ions into the stomach. The release
of histamine from ECL cells is inhibited by somatostatin, so that it is unable to signal
to parietal cells to promote hydrocloric acid secretion. Somatostatin also acts directly
on the parietal cells themselves, inhibiting the release of hydrogen ions into the stomach
(Takeuchi et al., 2016).
+++Negative feedback loops Somatostatin promotes homoeostasis by providing
negative feedback loops. These work to ensure that there is a brake on the secretion of
130
bioactive peptides and amines in the digestive system. This ensures that as food moves
through the gastrointestinal tract the areas posterior to this can return to their relaxed
state once the chyme is no longer present.
In addition to neuronal input signals to promote or inhibit somatostatin secretion δ cells
also have receptors on their basolateral surface which respond to paracrine and endocrine
signals from other GEP neuroendocrine cells. Signalling via these receptors on the δ cell
promotes the secretion of somatostatin. This in turn travels to local neuroendocrine cells
or via the capillary bed to more distant neuroendocrine cells where it binds to the cell
surface and triggers a reduction of the secretion of the original bioactive peptide or amine.
This negative feedback loop via δ cells provides an important brake on the secretion of
other neuroendocrine cells, whereby neuroendocrine cell signalling and the effects of their
secretion products can be gradually turned off when the signal gets stronger.
For example δ cells have receptors for both gastrin (secreted by G cells) and histamine
(secreted by ECL cells) on their cell surface. When gastrin and histamine are present
they bind to these receptors triggering δ cells to release somatostatin. The somatostatin
then travels to the G and ECL cells enabling a regulatory negative feedback loop whereby
it inhibits the secretion of these two signalling molecules leading to a gradual attenuation
of their biological effects. This reduces the levels of gastrin thus preventing it from
promoting hydrogen ion transport into the stomach lumen by parietal cells (Watson et
al., 2006). Reduced levels of histamine also prevents it from binding to parietal cells and
promoting the same process (Watson et al., 2006).
Tumours that arise from δ cells, or their precursor cells, have the ability to produce large
amounts of somatostatin leading to the development of the very rare somatostatinoma
syndrome. Somatostatinoma syndrome is present in less than 10 % of somatostatinoma
patients, with symptoms including diabetes mellitus, gall stones, weight loss and diarrhea
(Schimmack et al., 2011; Anderson and Bennett, 2016). Somatostatinomas are themselves
extremely rare, data from the USA showed an annual incidence of 1 per 40 million
131
population (Anderson and Bennett, 2016).
Pancreatic β cell
Pancreatic β cells are found in the endocrine part of the pancreas, the islets of Langerhans.
In adult humans islets the majority of the 3000 cells in a pancreatic islet, 54 %, are β
cells (Schimmack et al., 2011). The remaining cells in the islets are α cells (34 %), δ cells,
(10 %), and very small numbers of VIP cells, PP cells, and EC cells (Schimmack et al.,
2011).
Insulin secreted into the bloodstream by β cells regulates glucose homeostasis. High
levels of glucose in the circulation triggers β cells to release insulin into the blood and
also inhibits the release of glucagon from pancreatic α cells. Conversely low blood glucose
concentrations cause an inhibition of β cell insulin secretion and instead promote α cell
glucagon secretion. Glucagon then works to increase blood glucose levels by promoting
glycogenolysis and gluconeogenesis by hepatocytes.
Circulating insulin regulates metabolism by promoting glucose absorption and storage
by the liver (glycogenesis and lipogenesis), skeletal muscle (glycogenesis) and fat cells
(lipogenesis), reducing blood glucose concentrations. Insulin also inhibits gluconeogenesis
by hepatocytes.
Tumours that arise from β cells, or their precursor cells, have the ability to produce
large amounts of insulin leading to the development of an insulinoma. Insulinoma pa-
tients experience hypoglycaemia due to their high levels of insulin secretion, with symp-
toms that can include confusion, mood swings, weakness, sweating, blurred vision, heart
palpitations and dizziness.
2.4.4. Summary
The different types of neuroendocrine cells scattered throughout the body have a key role
in maintaining homoeostasis and regulating key processes such as digestion via the secre-
132
tion of bioactive peptides and hormones. The differing characteristics of neuroendocrine
cells as a result of their biological function and location within the body has an effect
on the types of tumours which may then develop if these cells become neoplastic. For
example different tumours may overproduce different hormones, depending on the par-
ticular characteristics and biology of the type of neuroendocrine cell they arose from. A
single neuroendocrine tumour will usually produce at least 2 different peptide hormones
(Tischler, 1989). Tumour cells may secrete both their usual secretory products and ec-
topic hormones that would not usually be produced by that neuroendocrine cell type or
hormones that are developmentally inappropriate. This variability is a key contributing
factor to the biological and clinical heterogeneity observed in GEP-NET patients.
2.5. MiRNA
MiRNA are small non-coding RNA, 19-24 nucleotides long, which act as post-transcriptional
regulators of endogenous gene expression (Ling et al., 2013). They do this by binding
to the 3’UTR of mRNA with a complementary nucleotide sequence, thus preventing the
mRNA from being translated into protein (see Figure 2.1). This process is known as
RNA interference.
MiRNA were first identified in the nematode worm, Caenorhabditis elegans in 1993
(Lee et al., 1993). However miRNA were not investigated in humans until 2001, when
miRNA let-7 was found to be part of a large family of miRNA genes some of which were
evolutionarily conserved between C. elegans and humans (Lagos-Quintana et al., 2001;
Lau et al., 2001; Lee and Ambros, 2001; Li and Kowdley, 2012; Kincaid and Sullivan,
2012). MiRNA have now been identified in many other animals, in plants, in amoeba,
and even encoded in viral genomes to control host gene expression (Avesson et al., 2012;
Kincaid and Sullivan, 2012).
MiRNA expression is tissue specific and they regulate diverse cellular processes in
133
Figure 2.1.: Function of miRNA. A) Genes are transcribed into mRNA which are trans-lated into protein (central dogma). B) miRNA regulate gene expression bybinding to the mRNA of certain genes and preventing their translation intoprotein.
134
humans from the cell cycle to immune system development to fat metabolism (Liu et al.,
2008; Bueno and Malumbres, 2011; Thai et al., 2007; Chen et al., 2014). More than 60%
of human protein coding genes are predicted to be regulated by one or more miRNA based
on the evolutionary conservation of miRNA binding sites within their 3’UTR (Friedman
et al., 2009).
MiRNA dysregulation is found in many disease states including diabetes and cancer
(Ling et al., 2013; Catalanotto et al., 2016). The role of miRNA in cancer was first
discovered in chronic lymphocytic leukemia in 2002 (Calin et al., 2002). Studies of this
disease showed that 69 % of patients showed the deletion or knockdown of miR-15a and
miR-16-1 (Iorio and Croce, 2012).
Regulation of gene expression by miRNA is an epigenetic mechanism whereby miRNA
provide post-transcriptional regulation of gene expression. Other epigenetic mechanisms
include histone modifications and DNA methylation (see section 2.6.2). The role of
epigenetic mechanisms including miRNA in tumourigenesis has not been well studied in
GEP-NET, with a lack of miRNA studies in SBNET in particular.
Nomenclature
With respect to nomenclature, the mature form of the miRNA is usually referred to
as miR-X, while the gene encoding that miRNA is referred to as mir-X (where X is a
number representing a particular miRNA). Nomenclature guidelines for the naming of
miRNA were laid out by Ambros et al in 2003 and these have been broadly adopted
within miRNA databases and the literature (Ambros et al., 2003). These rules state
that if the same mature miRNA is encoded at different loci in the genome then it is
referred to as miR-X-1, miR-X-2, miR-X-3 etc. Conversely similar but not non-identical
miRNA would be denoted miR-Xa, miR-Xb and miR-Xc etc. A prefix may be added
to denote the organism for example hsa-miR-X for a mature human miRNA (Kozomara
and Griffiths-Jones, 2014).
135
When two mature miRNA sequences have been identified in cloning studies as being
produced from the same miRNA hairpin precursor these are denoted miR-X-3p (3′ arm)
and miR-X-5p (5′ arm) (Kozomara and Griffiths-Jones, 2014). An asterisk may be used
to indicate the less common product of the precursor if this is known eg: miR-X (common
product), miR-X∗ (less common product) (Kozomara and Griffiths-Jones, 2014).
The recognition that both arms of the miRNA precursor frequently have a biological
function has led to the proposal in 2014 that the asterisk term be dropped in favour of
the universal use of the -3p, -5p notation which does not imply any relative biological
importance between the two arms of the precursor (Kozomara and Griffiths-Jones, 2014).
Older terms include antisense and sense strands.
Nomenclature changes over the years as more miRNA have been identified has led to
problems within the field due to the names of certain miRNA evolving and changing as
more information becomes available. The extent of the problem was revealed in a study
which identified that 12 % of mature human miRNA publications used erroneous or out
of date miRNA naming conventions for the time in which they were published (Van Peer
et al., 2014). This has led to difficulties in assessing the relevance of some experimental
findings due to ambiguity in miRNA names. This is being partially addressed with the
introduction of tools which enable researchers to track the changes over time in the
names of particular miRNA or miRNA platforms (Van Peer et al., 2014). Whether this
will be successful in mitigating the effects of the changes in miRNA annotations over
time remains to be seen and is likely to be dependant on the level of awareness amongst
researchers of these naming ambiguities.
2.5.1. Regulation of gene expression
MiRNA provide a complex epigenetic mechanism for the post-transcriptional regulation
of gene expression under physiological conditions. They help to maintain cellular ho-
moeostasis by ensuring that mRNA are translated in a spatially, temporally and develop-
136
Figure 2.2.: The biogenesis of endogenous miRNA and their regulation of gene expres-sion by RNA interference. Techniques for exogenous gene silencing whichutilise this biological pathway are also shown, with the introduction of smallinterfering RNA (siRNA) or a short hairpin (shRNA) encoded in a viral vec-tor. Figure reproduced from (Bak and Mikkelsen, 2010), creative commonslicence: CC BY 2.0.
mentally appropriate manor. The human genome is now thought to contain around 1900
evolutionary conserved miRNA based on miRBase, the public repository of published
miRNA sequences identified from deep sequencing experiments (Kozomara and Griffiths-
Jones, 2014; Meijer et al., 2014). This number is gradually increasing as more miRNA
are identified and as more becomes known about their structure and function.
Endogenous miRNA genes are transcribed by RNA polymerase II as long primary tran-
scripts (pri-miRNA). These are processed in the nucleus into individual double stranded
pre-miRNA which are 50-70 nucleotides long and have a stem loop (also known as hairpin
137
loop) structure (Kurzynska-Kokorniak et al., 2015). The pre-miRNA translocate out of
the nucleus for further processing into mature miRNA in the cytoplasm (see Figure 2.2).
An important protein for the biogenesis of functional miRNA is enzyme endoribonu-
clease Dicer, encoded by the gene dicer 1, ribonuclease III (DICER1 ). Dicer cleaves the
pre-miRNA to remove the loop and process it into a short duplexed miRNA which is
around 19-24 nucleotides long (Ling et al., 2013). These have 3′ (hydroxyl) end over-
hangs, 2 nucleotides long, and there are usually several base pairing mismatches present
(Meijer et al., 2014).
The gene silencing function of miRNA is carried out after it binds to the RNA induced
silencing complex (RISC) loading complex. The RISC loading complex is made up mul-
tiple proteins including Dicer, protein argonaute 2 (AGO2) and RISC-loading complex
subunit trans-activation-responsive RNA-binding protein (TARBP2).
The duplexed miRNA is loaded through the action of Dicer and TARBP2 onto AGO2.
AGO2 unwinds the duplex and identifies the guide and passenger strands according to the
thermodynamic properties of the 5′ (phosphate) end of each strand. The strand chosen
to be the guide strand of the mature miRNA is usually the one with the least stable 5′
end (Meijer et al., 2014). The guide stand is then incorporated into the activated RISC
complex with the passenger strand being discarded (Meijer et al., 2014). It was assumed
that the passenger strand was degraded, however there is a growing body of evidence that
suggests that the passenger strand may also frequently have a functional role (Kozomara
and Griffiths-Jones, 2014).
The mature miRNA is incorporated into the AGO2 protein to form the miRISC com-
plex and is then able to bind to mRNA that have a partially complementary sequence.
It is very rare in mammals (although common in plants) for there to be a perfect or near
perfect base pairing match between the miRNA and the mRNA target. When perfect
complementarity is present it triggers site specific endonucleolytic cleavage of the mRNA
by AGO2 (Iorio and Croce, 2012).
138
The usual mechanism of translational repression that occurs in mammals is non-
cleavage repression (Iorio and Croce, 2012). This occurs when there are missmatches
in the base pairing between the miRNA and the target mRNA. This triggers the re-
cruitment of further proteins to the complex leading to gene silencing through either
degradation or translation inhibition of the target mRNA. These miRNA mediated si-
lencing mechanisms ensure that only small amounts of the target mRNA are available
for translation into protein.
The acceleration of target mRNA degradation is thought to be achieved by the re-
cruitment of proteins such as the PAN2–PAN3 deadenylation complex by the miRISC
complex (Jonas and Izaurralde, 2015). This catalyses the removal of the poly-adenosine
(poly-A) tail from the 3′ end of the mRNA which destabilises the mRNA and promotes
degradation by exoribonucleases. More recent studies have suggested that there is a se-
quential process of miRNA silencing. First with the repression of translation and then
later with the promotion of mRNA decay, possibly due to kinetic differences between
these two processes (Jonas and Izaurralde, 2015; Catalanotto et al., 2016). More studies
are needed however to confirm if this is the case and to identify any interactions between
these two pathways. In the mean time the precise mechanism of miRNA silencing remains
poorly understood.
MiRNA most often bind to complementary sequences within the 3′ untranslated region
(UTR) of mRNA. They can also bind to miRNA response elements (MRE) in other parts
of the mRNA including the coding region and the 5′ UTR (Catalanotto et al., 2016). Each
miRNA can have over 100 different potential mRNA targets based on the evolutionarily
conserved miRNA binding sites present within genes (Meijer et al., 2014). The number
of potential targets involved presents a challenge for functional experiments investigating
the precise function of a given miRNA. This is due to the complexity involved in miRNA
mediated gene silencing and the large numbers of potential miRNA-mRNA interactions
requiring in vitro validation to determine their physiological importance.
139
The most important part of the miRNA for mRNA target recognition is the seed site
(also known as the seed region). A key feature of the seed site is that contrast to the
imperfect complementarity of the rest of the miRNA for its target mRNA, the seed site
has approximately 6 adjacent nucleotides with complementary Watson-Crick base pairing
with the MRE on the target mRNA (Catalanotto et al., 2016). The seed site nucleotides
were usually found at positions 2-8, 2-7 and 3-8 from the 5′ end of the miRNA (Seok
et al., 2016).
A more recent study using high-throughput sequencing of RNA isolated by cross linking
immunoprecipitation (HITS-CLIP), targeting mRNA cross linked to argonaute protein
in mouse brain tissue showed that this region is not always required for successful gene
silencing (Chi et al., 2012; Seok et al., 2016). Chi et al found that 27 % of the mRNA
argonaute clusters could not be accounted for by canonical seed sites (Chi et al., 2012).
This led to the recognition that many important miRNA silencing effects may well be me-
diated by non-canonical seed sites with looser rules for base pairing with sites containing
for example a mismatched base pair or a bulged nucleotide. Similar studies found that
these sites were present in various cell lines and in human brain tissue (Helwak et al.,
2013; Seok et al., 2016).
Studies have also shown that in addition to their function as post-transcriptional regu-
lators of gene expression, mature miRISC can also translocate back into the nucleus and
bind to DNA to promote either transcriptional gene activation or silencing (Iorio and
Croce, 2012; Catalanotto et al., 2016).
The study of miRNA is still at a relatively early stage and 17 years after miRNA
were first investigated in humans in 2001, many unanswered questions remain as to
their function both under physiological conditions and in different disease states. Future
experiments in this area should establish the function of miRNA further with respect to
their particular targets and their regulatory role.
140
2.5.2. Dysregulation in cancer
MiRNA have been found to be dysregulated in virtually all types of cancer and these
changes in miRNA expression promote tumour development and disease progression (Ling
et al., 2013). Studies have shown that miRNA genes are found in areas of the chromo-
somes that are prone to the deletions and amplifications that are frequntly found in
human tumours (Iorio and Croce, 2012). The changes in miRNA expression that occur
in cancer upset the balance of gene expression by promoting the translation of oncogene
transcripts and inhibiting the expression of tumour suppressor transcripts.
A single miRNA has the potential to target 100s of different mRNA thereby regulating
multiple different signalling pathways at the same time (Kuninty et al., 2016). This makes
them promising therapeutic targets since a single miRNA could be used to target multiple
oncogenes involved in several different tumourigenesis pathways. Due to their widespread
dysregulation in cancers miRNA also have the potential to be used as biomarkers in this
setting.
During tumourigenesis chromosomal loci encoding miRNA which can silence the ex-
pression of tumour suppressor transcripts can become amplified (Iorio and Croce, 2012).
This prevents the tumour suppressor genes from being translated into protein, see Figure
2.3. MiRNA which are instead involved in the repression of the expression of oncogene
transcripts are frequently found at chromosomal loci where mutations and deletions are
common in tumourigenesis (Iorio and Croce, 2012). This leads to an absence of the nega-
tive regulation of oncogene expression usually provided by the miRNA so that oncogenes
become overexpressed.
An example of an oncogenic miRNA or oncomir is miR-21 which is overexpressed in
several cancers including breast and colon cancers (Iorio and Croce, 2012; Ling et al.,
2013). When miR-21 is overexpressed in cancer it prevents apoptosis by targeting PTEN
and programmed cell death 4 (PDCD4) transcripts for gene silencing, stopping the trans-
lation of these tumour suppressors (Iorio and Croce, 2012). Conversely tumour suppressor
141
Figure 2.3.: The effects of miRNA dysregulation during tumourigenesis. A) The roleof a miRNA in normal tissue. B) During tumourigenesis, different stagesin the miRNA biogenesis can become dysregulated or the miRNA gene maybe deleted/mutated leading to reduced levels of the miRNA and inappropri-ate expression of the target oncogene. C) During tumourigenesis amplifi-cation/overexpression of a miRNA can occur, so that it is expressed in thewrong tissue or at an inappropriate time, it then prevents the expression ofthe target tumour suppressor gene. Reprinted by permission from Macmil-lan Publishers Ltd: [Nature Reviews Cancer] (http://www.nature.com/nrc)(Esquela-Kerscher and Slack, 2006), ©(2006).
142
miRNA such as miR-15a and miR-16-1 which negatively regulate oncogene BCL-2, or let-
7 which negatively regulates oncogene RAS have reduced expression in cancers such as
prostate cancer (Iorio and Croce, 2012; Ling et al., 2013). Global suppression of miRNA
expression may also occur in cancer, with dysregulation of the proteins involved in the
miRNA biogenesis pathway (Lin and Gregory, 2015).
MiRNA dysregulation is not just a feature of tumour cells, it is also found in the stromal
cells that make up the tumour microenvironment, such as cancer associated fibroblasts,
tumor-associated macrophages and other immune cells and endothelial cells. The tumour
microenvironment plays an important role in the growth and metastatic spread of cancer
(Bell and Taylor, 2017). MiRNA which have been found to be dysregulated in stromal
cells include increased miR-21 expression in colorectal cancer associated fibroblasts, re-
duced levels of miR-155 and miR-214 in ovarian cancer associated fibroblasts and reduced
miR-155 expression in hepatocellular carcinoma associated macrophages (Kuninty et al.,
2016).
Therapeutics
MiRNA have the potential to be used as therapeutics in the future. Small interfering
RNA (siRNA) could be chosen or designed to act as therapeutics by binding to certain
mRNA to reduce the expression of certain genes, for example to reduce the expression
of an oncogene (see Figure 2.3). Another use of exogenous siRNA is RNA interference
(RNAi) studies, where specific genes of interest are knocked down in cell lines or model
organisms as a molecular method for determining gene function.
In addition, siRNA can also be designed as therapeutics which act on the miRNA
themselves. This means that it would be possible to downregulate an oncomir in cancer
by using a siRNA with a complementary sequence, thus preventing the repression of the
target gene (a tumour suppressor). An important consideration will be the avoidance of
off target effects due to the large number of different genes a single miRNA can regulate.
143
Therapeutics based on siRNA would need to be designed so as to ensure that the siRNA
developed was specific to the gene target of interest (or several gene targets within the
same pathway with redundant functions).
Challenges that remain are how to successfully deliver the siRNA to tumour cells,
while avoiding targetting other cell types, however studies in this area will benefit from
recent advances in gene therapy techniques using adenovirus associated vectors for gene
delivery. Various precinical studies and phase I clinical trials are being carried out in
different diseases, for example in primary liver cancer (miR-34 mimic), in atherosclerosis
(anti-miR-33) and a phase II trial for use in the treatment of hepatitis C viral infections
(anti-miR-122) (Christopher et al., 2016; Ling et al., 2013).
MiR-122 is needed for hepatitis C replication as it binds to the viral genome enhancing
translation and replication (Christopher et al., 2016). The introduction of an artifi-
cial oligonucleotide with a complementarity sequence to the endogenous miR-122 (an
antagomir) can be used to sequester the miR-122, preventing it from binding to viral
genome (Christopher et al., 2016). More clinical trials will be needed to determine if
miRNA mimics and inhibitors can be effective in treating cancer and other diseases. No
studies of miRNA therapeutics have been done thus far in GEP-NET patients.
Biomarkers
Dysregulation of miRNA expression is a common event in cancers and around 50 % of
miRNA are located in regions of cancer associated chromosomal abnormalities (Bell and
Taylor, 2017). MiRNA profiling studies have shown that miRNA signatures enable the
accurate identification of the tissue or tumour type (Lu et al., 2005; Volinia et al., 2006;
Ludwig et al., 2016; Guo et al., 2015). This is in contrast to mRNA which are inaccurate
predictors of tissue or tumour type (Iorio and Croce, 2012). In another study, the miRNA
signature of metastases was able to correctly predict the site of the primary tumour for
77 % of the tumours included in the study (Rosenfeld et al., 2008). MiRNA therefore
144
represent promising candidates as cancer biomarkers.
MiRNA dysregulation seems to be a very early event in cancer development suggesting
that the identification of a specific miRNA signature might enable earlier cancer diagno-
sis. For example in pancreatic ductal adenocarcinoma, increased expression of miR-21
preceded phenotypic changes in duct morphology in a conditional KRAS (G12D) mouse
model and miR-21 was increased in human tumour tissues with increasing tumour grade
(Du Rieu et al., 2010). In lung cancer the miRNA signature found in plasma samples
collected 1-2 years before the onset of the disease was able to predict the likelihood of
lung cancer development and the aggressiveness of the future disease (Boeri et al., 2011).
These miRNA biomarkers could improve patient survival rates if they could be used to
identify patients at an earlier disease stage.
An advantage of the use of miRNA as biomarkers is that they have a very stable struc-
ture. In particular, miRNA are much more stable than the far longer and more readily
degraded mRNA. MiRNA precursors have a stable stem loop secondary structure while
mature miRNA are sequestered within the miRISC complex where they are stabilised by
the argonaute protein and base pairing with sequences in the 3′ UTR of target mRNA.
MiRNA have an average half life of around 5 days compared to just 9 hours for mRNA
(46 hours for proteins) (Meijer et al., 2014; Schwanhausser, 2011).
MiRNA have been found to be resistant to degradation at high temperatures and when
subjected to multiple cycles of freezing and thawing and when left for long periods of time
at room temperature (Mitchell et al., 2008; Jung et al., 2010; Bell and Taylor, 2017). A
study was done to compare the heat stability of miRNA and mRNA in samples incubated
at 80◦C (Jung et al., 2010). This showed that the mRNA became degraded so that they
could not be reliably quantified by qPCR, in contrast the miRNA remained stable and
could be reliably quantified by qPCR even when the RNA integrity scores (calculated
based on the 28S:18S ribosomal RNA ratio) were low. These results suggest that RNA
integrity scores are not a reliable indicator of miRNA integrity.
145
This means that unlike mRNA, miRNA are found intact in FFPE tissue. This is
advantageous for the development of biomarkers since the majority of tissue samples from
patients are stored as FFPE tissue within local and national tissue archives in contrast
to fresh frozen tissue which is much less common within such archives. MiRNA studies
can therefore benefit from this larger pool of tumour tissue available for investigation.
Limitations of the use of miRNA as biomarkers are that functional information on
many miRNA and their particular gene targets under physiological conditions and in
different disease states is still emerging. In GEP-NET in particular there have been very
few studies done of miRNA expression compared to other more common types of cancer
such as breast cancer, lung cancer and prostate cancer and virtually no in vitro functional
studies or circulating miRNA studies.
MiRNA can be found in serum/plasma and in other bodily fluids. The majority of
circulating miRNA is bound to argonaute proteins (around 90 %), with the remainder
being inside circulating exosomes (30-100 nm vesicles released by cells) (Sato-Kuwabara
et al., 2015; Bell and Taylor, 2017; Larrea et al., 2016; Witwer, 2015). MiRNA are quite
stable in blood, since they are more resistant than mRNA to degradation by ribonucleases
(RNases), present in blood and other bodily fluids (Aryani and Denecke, 2015).
Circulating miRNAs have been proposed as biomarkers in a diverse range of different
types of cancer. This liquid biopsy approach is far less invasive than taking a tissue biopsy
and would enable miRNA biomarkers to be measured at multiple time points during the
patient journey to monitor genetic/epigenetic changes in the tumour (Crowley et al.,
2013; Diaz and Bardelli, 2014). Examples of possible biomarkers include, serum miR-
141 in prostate cancer, where it was found to be 46 fold overexpressed in patient serum
compared to control serum and melanoma, where a signature of 5 serum miRNA (miR-
150, miR-15b, miR-199a-5p miR-33a and miR-424) stratified patients into high and low
recurrence risk groups (Mitchell et al., 2008; Friedman et al., 2012).
In the future, miRNA biomarkers could be used to stratify GEP-NET patients into
146
subgroups based on clinical and pathological behaviour, such as prognosis, the likeli-
hood future disease progression/distant metastases and response to particular therapies.
MiRNA biomarkers could also be used to identify patients with low grade GEP-NET that
have a more aggressive, metastatic phenotype to enable tailored treatment. Circulating
miRNA could have the potential in the future to be used as part of a non invasive liquid
biopsy taken at multiple time points for the early detection of disease recurrence or for
the monitoring treatment response. Clinical trials would be needed to validate any po-
tential miRNA biomarkers and to ensure that they would be of benefit to patients with
GEP-NET.
Endogenous controls
Endogenous controls (or reference genes) are used for data normalisation of the raw qPCR
data prior to comparison of expression levels between different sample groups of interest
(Reboucas et al., 2013). An endogenous control is an internal control that is usually
constitutively expressed. Expression levels of the endogenous control should be stable
across the experimental comparison groups. Endogenous controls are used to minimise
experimental errors caused by variations between samples introduced by factors such
RNA extraction and cDNA synthesis efficiency, RNA quality and quantity and pipetting
errors (Reboucas et al., 2013).
Normalisation against tissue or serum volume as an alternative to using endogenous
control genes does not appear to be biologically representative as it does not reflect
differences in sample conditions (Song et al., 2012; Reboucas et al., 2013).
Studies have shown that the choice of endogenous control can have a large effect on the
outcome of qPCR data analysis and the reliability of results (Song et al., 2012; Reboucas
et al., 2013). It is very important therefore that the chosen endogenous control has been
experimentally validated as being stable in a particular experimental setting prior to
study commencement.
147
In qPCR studies investigating mRNA expression, endogenous controls (or housekeeping
genes) such as β actin have been well validated in multiple tissue types, nevertheless no
single endogenous control will be suitable across all experiential models and tissue types
(Reboucas et al., 2013). The expression levels of the chosen endogenous control gene
should be checked in the particular tissue types to be included in a study and prior to
using a new experimental design, since they could still be altered in certain disease states
or biological conditions.
This is less straight forward for miRNA studies, where the endogenous controls are less
well established, especially those for use in serum samples. MiRNA have been far less
studied than mRNA due to their much more recent discovery and their expression levels
in different types of tissues and disease states is still being investigated. This means that
in the case of miRNA the experimental validation of the endogenous control prior to the
commencement of qPCR studies is even more crucial.
2.5.3. SBNET
There have been limited studies of miRNA in SBNET. This is in contrast to PNET
where miRNA have been more extensively studied both in humans and in a PNET mouse
model. There are as yet no mouse models of SBNET. The results from the studies that
have been done suggest that SBNET and PNET have a different miRNA profile, with
different miRNA being dysregulated during tumourigenesis.
There have been 3 studies published to date which include at least some element of
miRNA analysis in tumour tissue from SBNET patients (Ruebel et al., 2010; Li et al.,
2013b; Nieser et al., 2016). The largest study quantified the expression of 847 miRNA in
SBNET and metastases, another study quantified the expression of a smaller panel of 85
miRNA in SBNET and metastases, while the remaining study quantified just 15 miRNA
in primary tumours only. A single study investigated the expression of 9 miRNA in the
serum of SBNET patients (Li et al., 2015).
148
MiRNA in tissue
The earliest study of miRNA in SBNET was by Ruebel et al. (2010). The expression
pattern of a small number of cancer related miRNA, 95 miRNA, was investigated in tissue
from matched SBNET, lymph node metastases and liver metastases from 14 patients,
Table 2.8. The final analysis included 85 miRNA in total, due to 10 miRNA being
excluded from the study due to non-consistent amplification. There was a significant
reduction in the expression of miR-133a in the metastases compared to the primary
tumour.
149
Table 2.8.: MiRNA expression studies in primary tumours and metastases of SBNET patients
Paper Ruebel et al. (2010) Li et al. (2013b) Nieser et al. (2016)Number of
Samples 28 24 20Patients 14 22 20MiRNA 85 847 15
Study type cancer panel,SBNET/metastases
global profiling,SBNET/metastases
15 miRNA, Chr18 (+/-) or(+/+)
Sample type fresh frozen tissue fresh frozen tissue FFPE tissueMethods
MiRNA profiling assay QuantiMir Cancer qPCRArray
GeneChip miRNA 1.0Array
-
Profiling assay provider System Biosciences, CA,USA
Affymetrix, CA, USA -
Profiling normalisation miR-197 quantile normalisation -Validation assay qPCR qPCR qPCRValidation endogenous
controlSNORD48 SNORD48 SNORD61
MiRNA profiling studyNumber of patients 8 15 -Tissue types: SBNET,
LNM, LVM8, 1, 7 5, 5*, 5 -
Validation studyNumber of patients 6 7 20PT, LNM, LVM 6, 5, 1 3, 3*, 3 -Other samples (normal) normal ileal tissue EC cells -PT Chr18 (+/-), (+/+) - - 10, 10
Validated miRNA
150
Continuation of Table 2.8
Paper Ruebel et al. (2010) Li et al. (2013b) Nieser et al. (2016)Comparison groups LNM/LVM v PT LNM/LVM v PT (and all v
EC cells)Chr18 (+/-) v Ch18 (+/+)
Significant expressiondifferences
(-) miR-133a (-) miR-133a None
(-) miR-31(-) miR-129-5p
(-) miR-215(+) miR-96(+) miR-182(+) miR-183(+) miR-196a
(+) miR-200a (NS:LNM/LVM v PT)
PT: primary tumour, LNM: lymph node metastasis, LVM: liver metastases, *: mesentericmetastases, NS: non-significant, HD: healthy donor
151
The expression pattern of miR-133a was also investigated in normal ileum FFPE sam-
ples by in situ hybridisation which showed that miR-133a was expressed by EC cells in
normal mucosa. MiR-133a was expressed in the cytoplasm of the primary tumour cells
but not in the adjacent connective tissue. The cytoplasm of tumour cells in the liver
metastases were also positive for miR-133a on in situ hybridisation but not the adja-
cent normal liver tissue. No comparisons were made by qPCR of the expression levels of
miR-133a in the tumour samples compared to the normal ileum.
In 2013 the only global study of miRNA expression in SBNET was published by Li
et al. (2013b). This included a much larger panel of around 800 miRNA, validated from
the miRBase database. The expression of the miRNA was investigated in tissue from
the primary tumour, lymph node metastases and liver metastases. There were 24 tissue
samples included, a similar number to the earlier study of 28 samples, see Table 2.8 (Li
et al., 2013b; Ruebel et al., 2010). In contrast to the Ruebel et al study, the primary and
metastasis samples were not matched but instead came from different patients (except
for two patients with both a SBNET and a mesenteric metastasis sample included).
Li et al. (2013b) identified 9 dysregulated miRNA in their global profiling study that
were chosen for validation by a second quantification method, qPCR, and in a second set
of samples, see Table 2.8. The 9 miRNA were significantly upregulated or downregulated
during tumour progression (from primary to mesenteric or liver metastases) and also
when compared to miRNA extracted from laser capture microdissected EC cells from
normal small bowel tissue.
There were 4 miRNA, miR-96, miR-182, miR-183 and miR-196a, that were signifi-
cantly upregulated in tumour metastases compared to the SBNET and 4 miRNA, miR-
133a, miR-31, miR-129-5p and miR-215, that were significantly downregulated in tumour
metastases compared to the SBNET. A comparison of the 9 miRNA in the EC cells and
the primary tumour, lymph node metastasis and liver metastasis tissue showed that the
upregulated/downregulated miRNA were also significantly differentially expressed in the
152
EC cells compared to the SBNET, Table 2.8.
An additional miRNA, miR-200a, was identified as being upregulated in the initial
global profiling study but was not found to be significantly differentially expressed in
the validation study. This miRNA was also investigated in EC cells and was found to
be significantly upregulated in SBNET and metastatic tissue compared to its expression
in EC cells. Interestingly the size of the change in expression of miR-200a between the
EC cells and the tumour tissue, at only one order of magnitude difference, is quite a
lot smaller than for the other 8 dysregulated miRNA which had a relative change in
expression of at least 2 orders of magnitude in the tumour samples compared to the EC
cells (miR-183 had a similar size of change in expression to miR-200a). It would have been
interesting to see the exact fold changes in miRNA expression between sample groups,
however these were not presented.
Possible gene targets of the dysregulated miRNA were predicted based on matching the
miRNA seed site to mRNA sequences of genes previously identified as being differentially
expressed in SBNET using the bioinformatics program TargetScan (Li et al., 2013a; Lewis
et al., 2005; Friedman et al., 2009). These included genes involved in a wide range of
functions from GPCR involved in signal transduction such as FZD3 and CALCR which
were predicted to be possible gene targets of miR-31, to transcription factors involved
in differentiation such as NKX2-2, and genes involved in membrane transport such as
SLC8A2, SLC7A2 and GPM6A predicted as possible gene targets of miR-133a. These
finding will need to investigated experimentally to confirm these predicted miRNA-mRNA
interactions.
+++Comparison of miRNA tissue studies Interestingly miR-133a was identified
as being significantly downregulated in metastases compared to the primary tumour in
both the Li et al. (2013b) and Ruebel et al. (2010) study (in the latter study it was the
only miRNA found to be significantly dysregulated). This suggests that the change in
153
miR-133a expression is particularly robust in SBNET since it was identified in both the
studies to date, despite the studies being were carried out at different centres and the
differing methodologies used.
The remaining 8 miRNA validated from the global miRNA expression study were not
mentioned as having been dysregulated in the study by Ruebel et al. (2010). One possible
explanation for this is that miR-133a is the only one of the 9 miRNA validated in the
global study that was investigated in both studies, since Ruebel et al. (2010) study only
had 95 miRNA in their array.
The Ruebel et al. (2010) paper did not contain a list of the 95 miRNA quantified in
their study or of the 10 miRNA they later excluded from the analysis. In addition, the
single figure representing the results from the panel of miRNA (Ruebel et al. (2010),
Figure 1), which shows the relative expression of 62/85 miRNA for 8 cases, is challenging
to interpret due to the low resolution of the figure, in particular of the x axis labels
(making it difficult to interpret the miRNA names).
This means that from the information provided in the Ruebel et al. (2010) paper alone
it is difficult to determine if any of the 9 miRNA validated in the global Li et al. (2013b)
study were also investigated in this study (in addition to miR-133a). Instead examination
of a miRNA list available online from System Biosciences, the company that produced
the QuantiMir Cancer qPCR Array assay used in the study (System Biosciences, CA,
USA), shows that some of these 9 miRNA were in the array. This showed that in addition
to miR-133a the array also included miR-215, miR-183, miR-196a and miR-200a. There
were 4/9 miRNA validated in the Li et al study that were however absent from the Ruebel
et al study, thus definitively explaining why these particular miRNA were not identified
in both studies as being significantly dysregulated.
Using the well identifiers from the array (available online from the suppliers) and
referring back to Ruebel et al. (2010) Figure 1, it is possible to identify with reasonable
certainty two of the miRNA shared between the two studies within Figure 1, D9 miR-183
154
and E7 miR-196a. MiR-183 appears to be upregulated in the metastases compared to the
primary tumour in the initial profiling, in keeping with the Li et al. (2013b) results, while
miR-196a shows mixed results, with increased expression in some metastasis samples
and decreased expression in others (contrasting with the Li et al results) (Ruebel et al.,
2010). These differences may be due to differences in methodology such as normalisation
methods however since neither of these miRNA were chosen for validation in the Ruebel
et al study, it is difficult in the absence of this data to draw strong conclusions as to these
similarities and differences.
MiR-215 and miR-200a could not be identified in the x axis labels of Ruebel et al.
(2010), Figure 1, suggesting that they may have been amongst the 10 miRNA excluded
from the profiling analysis or amongst the 23 miRNA excluded from the figure (only
results from 62/85 of the miRNA analysed are presented).
Differences and similarities between these two studies are summarised in Table 2.8. Key
similarities include that both studies used fresh frozen tissue and qPCR for the validation
study and both studies found miR-133a to be significantly downregulated in metastases
compared to the primary tumour.
Key differences between the two studies are that Ruebel et al. (2010) analysed the
expression of a smaller number of miRNA, 85 miRNA, compared to the global study
by Li et al. (2013b) which quantified 847 miRNA, they selected fewer miRNA for val-
idation, just 1 miRNA compared to 9 for the Li et al study (see Table 2.8). Different
normalisation methods were also used for miRNA expression profiling, with Ruebel et
al. (2010) normalising against miR-197 expression while Li et al. (2013b) used quantile
normalisation.
There were also differences in the types of samples used since Ruebel et al. (2010) used
matched patient samples while Li et al. (2013b) used non-matched patient samples. No
comparison was made of the miR-133a levels between the normal ileum samples and the
tumour tissue samples in the Ruebel et al study in contrast to the Li et al study where
155
the expression levels of 9 miRNA were compared between EC cells and tumour samples.
In conclusion despite the large methodological differences between the two studies,
including differences in study design and miRNA quantification and analysis, miR-133a
was found to be significantly downregulated in metastases compared to the primary
tumour in both studies (Li et al., 2013b; Ruebel et al., 2010). This suggests that miR-133a
dysregulation in SBNET metastases is a reproducible phenomenon which warrants further
study in vitro. This could enable the identification of biologically relevant gene targets
and the investigation of possible functions of miRNA-133a in promoting tumourigenesis
and/or metastatic growth. Future experiments using samples from different centres would
be beneficial to confirm the reliability of the findings and those for the 8 other miRNA
identified as differentially expressed in SBNET in the global study (Li et al., 2013b).
Functional studies would also be of interest to investigate the role of these miRNA in the
development of SBNET and their metastases.
+++Loss of chromosome 18 miRNA in primary tumours A recent study, in
2016, investigated the loss of one allele of chromosome 18 (Chr18) in SBNET, a common
genetic event in this type of neuroendocrine tumour (see section 2.6.2), with respect to
the gene and protein expression levels of 7 tumour suppressor genes (Nieser et al., 2016).
Only one of these genes, DCC netrin 1 receptor (DCC) (previously known as deleted in
colorectal carcinoma), was reduced at the protein level in one third of the SBNET studied
with loss of one allele of Chr18.
While tumour suppressor gene expression was the focus of the study, possible differences
in the expression pattern of 15 of the miRNA located on Chr18, which included miR-
133a, were also investigated in FFPE tissue from the primary tumour (Nieser et al., 2016).
Only tissue from the primary tumour was included in the study, with FFPE tissue from
Chr18 (+/-) SBNET (n=10) and Chr18 (+/+) SBNET (n=10) being the two comparison
groups. No significant differences were found in the expression levels of the 15 miRNA
156
between the two study groups, this included miR-133a (miR-133a fold change: -1.87, p
value: 0.22).
It is not possible to make a direct comparison of the findings in this study to the
findings of the two earlier studies by Ruebel et al. (2010) and Li et al. (2013b), since the
comparison groups were different (Nieser et al., 2016). It would have been interesting if
this latest study had in addition to the primary tumour tissue, also included metastatic
tissue from SBNET patients in the analysis. This would have enabled comparison with the
two other miRNA studies and would have identified if miR-133a had reduced expression in
the metastases compared to the primary tumour in the patients included in this study. It
would also have been interesting to see if any possible reduction in miR-133a in metastatic
tissue was affected in the presence or absence of Chr18 deletions in the primary tumour.
MiR-133a is the only miRNA validated in the SBNET miRNA studies to date that is
located on Chr18. MiR-96, miR-182, miR-183, miR-196a, miR-200a, miR-31, miR-129-
5p and miR-215 are all located elsewhere in the genome and so were not amongst the 15,
Chr18 miRNA quantified by Nieser et al. (2016).
The findings of the study, with no significant change in miR-133a expression with the
loss of one allele of Chr18, suggests that this common deletion in SBNET patients may
not be responsible for the significant reductions in miR-133a observed in these tumours
(Ruebel et al., 2010; Li et al., 2013b; Li et al., 2015). Alternatively the loss of Chr18
may have a different effect in metastatic tissue in contrast to the primary tumour, future
studies including metastatic tissue would be needed to determine if this is the case.
MiRNA in serum
There has only been one study of circulating miRNA in SBNET patients to date, and no
global studies of circulating miRNA expression in SBNET patients in contrast to PNET
(see section 2.5.4). This study was published by the same group that carried out the
global miRNA study in tumour tissue Li et al. (2015). The aim of the study was to
157
investigate a small number of miRNA (a miRNA signature) in serum to determine if
these miRNA could be detected in the circulation and if their levels changed with SSA
treatment.
The levels of 9 miRNA (validated in the study in tissue Li et al. (2013b)) were quantified
in serum from 48 SBNET patients, at different tumour stages, and 10 healthy donors (Li
et al., 2015). The serum samples included in the study came from Uppsala University
Hospital (42 patients, 7 healthy donors) and University College London (UCL) (6 patients,
3 healthy donors). This represents a much larger total number of samples than the earlier
studies in tissue (58 serum samples) however only 9 miRNA were investigated.
The majority of the analysis was done in the serum samples from Uppsala (Li et
al., 2015). These were 21 untreated patients (7 primary, 7 lymph node metastases, 7
liver metastases), 21 SSA treated patients (7 primary, 7 lymph node metastases, 7 liver
metastases) and 7 healthy donors.
All 9 miRNA that were differentially regulated in tumour tissue were detectable in
the serum from the 21 patients (untreated) at all different stages of disease and in the
serum from the 7 healthy donors. The relative expression levels of the serum miRNA
at different stages of disease in the 21 patients compared to the healthy donors was
investigated. This revealed that the 4 miRNA that were significantly downregulated in
SBNET and metastasis tissue, miR-133a, miR-31, miR-129-5p and miR-215, were also
significantly downregulated in the serum of SBNET patients at all disease stages when
compared to the control serum.
The picture with respect to the miRNA that were upregulated in tissue samples was less
clear, with no significant difference in miR-183 expression between the SBNET serum and
the healthy donor serum (for all disease stages), while miR-182, miR-196a and miR-200a
(upregulated in SBNET/metastasis tissue samples) were only significantly upregulated
in liver metastasis patient serum but not serum from patients with only the primary
tumour or lymph node metastases (versus healthy donor serum). MiR-96 was significantly
158
upregulated in the serum from patients with the primary tumour only and liver metastases
but not in patients with lymph node metastases.
The results for the downregulated miRNA in primary tumours and metastases com-
pared to normal controls, miR-133a, miR-31, miR-129-5p and miR-215 were comparable
between the study in tissue samples and the study in serum. This suggests that a liquid
biopsy using these 4 miRNA could be useful as a future biomarker, as levels of these
miRNA in serum appear to be indicative of the levels in tumour tissue.
For the upregulated miRNA in tumour tissue samples the results were more mixed
with serum from patients with liver metastases having significant upregulation of serum
miR-182, miR-196a, miR-200a and miR-96 but with this not being observed for earlier
disease stages. This may mean that serum levels of the miRNA that were upregulated in
tissue samples may be less useful as a biomarkers, since circulating levels of these miRNA
appear to be less representative of the miRNA changes in the tumour tissue.
A possible explanation for this could be that patients with liver metastases may be
more effective at releasing the upregulated miRNA into the blood stream than those
with only localised disease, with the SBNET not releasing sufficient levels of miRNA on
its own to be detected in serum, despite the tumour cells themselves overexpressing these
miRNA. The choice of endogenous control and the types of samples included for each
stage of the disease could also have affected the results (possible factors that may not
have been controlled for in the study are described in the next section).
There is conflicting information in the literature about the level of similarity in the
miRNA profiles between tumour tissue and circulating miRNA, with many studies iden-
tifying differences (Witwer, 2015; Larrea et al., 2016). This may be because systemic
changes caused by the presence of the tumour and/or metastases and changes in the
tumour microenvironment may be contributing to the total levels of circulating miRNA.
More studies are needed to establish the possible functions of circulating miRNA and
which cell types they originate from both under physiological conditions and in disease
159
states such as cancer.
For the downregulated miRNA, miR-133a, miR-31, miR-129-5p and miR-215 further
studies including additional tissue samples would be helpful to determine if this is a stable
signature in SBNET primary tumours and metastases and to investigate if these miRNA
could have potential as future biomarkers. It would be interesting to see more studies of
these miRNA in further tissue and serum samples from SBNET patients, in particular
more global studies of miRNA expression in SBNET patients. Ideally studies would be
carried out with matched tumours (primary/metastases) and serum samples taken from
the same SBNET patients to give a clearer idea of how well circulating miRNA reflect
the miRNA that are dysregulated in tumour tissue in SBNET patients and if they can
be used to predict useful clinical factors such as survival or future metastases.
The second part of the serum miRNA study by Li et al. (2015) focused on a comparison
of their 9 miRNA signature in serum from 21 untreated patients, compared to 21 SSA
treated patients at different stages of disease.
Treatment with SSA had no effect on the serum levels of the 4 miRNA that were
downregulated with disease progression in the tissue study. The 5 miRNA that were
found to be upregulated in tumour tissues, miR-96, miR-182, miR-183, miR-196a and
miR-200a were found at significantly higher levels in serum from SSA treated patients
than untreated patients. This held true for all stages of disease (except for miR-200a in
patients with liver metastases which as not significant).
If these results with respect to SSA treatment are confirmed in tumour tissue samples
in SBNET this is a rather unexpected finding as it would suggest that the 5 miRNA that
were upregulated in SBNET metastases are upregulated still further with SSA treatment,
despite SSA treatment increasing PFS (Cives and Strosberg, 2015; Li et al., 2013b). This
might suggest that despite the upregulated miRNA being increased in metastases they
could possibility have a protective role. Alternatively, since the profile of circulating
miRNA may differ from that found in tumour tissue therefore the increased levels of the
160
5 circulating miRNA with SSA treatment may be limited to serum and may not reflect
changes in the tumour tissue itself. This would need to be confirmed by future studies
including both tissue and serum from the same patients.
Another possible explanation of these findings is that they could be caused by other
uncontrolled for factors in the study rather than being the direct result of SSA treat-
ment. The SSA treated patients may have had a functioning SBNET while the untreated
patients had a non-functioning SBNET, the functionality of the tumours included in the
study was not mentioned (Li et al., 2015). This difference in tumour pathology might
have been responsible for the increased serum levels of miR-96, miR-182, miR-183, miR-
196a and miR-200a observed in the SSA treated patients rather than the SSA treatment
itself. It would also have been interesting to know what other treatments the patients
received (for example surgical resection) and if the SSA untreated patients were newly
diagnosed/treatment naive since these factors could also have affected the results.
A limitation of the study is that Li et al. (2015) do not mention if they validated the
stability of their chosen endogenous control, miR-16, in serum from SBNET patients and
controls. Instead they state that miR-16 was chosen based on its use as an endogenous
control in gastric cancer studies (Li et al., 2015; Zhang et al., 2015). The validation of
miR-16 in the particular sample types included in the study would have been necessary
to ensure expression levels were stable across SBNET patient samples and normal serum
samples to avoid the results being skewed by inter-sample variability rather than being
a true reflection of changes in the miRNA under investigation.
Experimental validation of the stability of an endogenous control miRNA in the partic-
ular sample types being used in an experiment is of particular importance in the emerging
area of circulating miRNA quantification, since there is a lack of consensus in this field
about the appropriate endogenous controls for qPCR (Song et al., 2012; Reboucas et al.,
2013).
Additional studies are warranted to further investigate the effects of SSA treatment
161
on miRNA levels both in serum and tissue, ideally with matched samples. This would
determine if these results are reproducible across multiple studies and could investigate
the effect of SSA treatment and other types of treatment on miRNA expression.
2.5.4. PNET
In PNET there have been quite a few studies of miRNA. These include comprehensive
global miRNA expression studies in tumour tissue and serum, studies including the quan-
tification of both miRNA and mRNA from the same patients and functional studies of
certain miRNA in in vitro and in vivo models. MiRNA have therefore been much better
characterised in PNET patients than in SBNET patients where there have been fewer
studies carried out with no functional studies in in vitro and in vivo models.
The earliest study of miRNA in PNET was in 2006 by Roldo et al. (2006). The levels
of the human miRNA known to exist at the time (235 miRNA) were quantified using
a custom microarray. The study included frozen primary tumour tissue from sporadic
pancreatic tumours, 12 insulinomas, 28 non-functioning PNET, and 4 acinar tumours
and 12 adjacent normal pancreas samples.
The study found that there was an increase in miR-103 and miR-107 expression and a
reduction in miR-155 expression in the tumour tissue compared to the adjacent normal
tissue. A set of 10 miRNA enabled the neuroendocrine tumours to be distinguished from
the exocrine acinar tumours. Roldo et al. (2006) suggest that these differences may be
the result of either differences in tumourigenesis pathways or differences that originate
during normal endocrine differentiation. Other miRNA of interest that were identified
in the study included miR-204 which correlated with insulin expression and was mainly
expressed in the insulinomas rather than in the other tissue groups.
MiR-21 was the only miRNA in the study that was found to be able to distinguish
the G3 tumours and those with liver metastasis from the remaining PNET, with the
levels of this miRNA being increased in the primary tumour in these patient groups.
162
Overexpression of miR-21 is a common feature in cancers including ovarian cancer, lung
cancer, cervical cancer and colorectal cancer where it is associated with invasive tumours,
high proliferation rates, worse clinical outcomes and reduced apoptosis (Pfeffer et al.,
2015; Mima et al., 2016; Buscaglia and Li, 2011).
There have been many functional studies done on miR-21 as a result of its common
association with cancer and its ability to target genes playing key roles in virtually all
the different hallmarks of cancer (Buscaglia and Li, 2011; Hanahan and Weinberg, 2011).
Many of the experimentally validated gene targets of miR-21 are tumour suppressors
including phosphatase and tensin homolog (PTEN) which inhibits anti-apoptotic AKT
signalling (Buscaglia and Li, 2011). PTEN, was found to be mutated in 7.3 % of sporadic
PNET in a study of 68 PNET patients (Jiao et al., 2011). This suggests that in G3
PNET, miR-21 may play an important role in silencing PTEN expression in patients
lacking PTEN somatic mutations or in silencing the unaffected allele in patients with
heterozygous mutations.
+++ RIP-Tag2 mouse model There is a transgenic mouse model of PNET, this
was used to study miRNA expression in a study by Olson et al. (2009). This is the
transgenic RIP-Tag2 mouse strain developed in 1985, by introducing the simian virus 40
(SV40) large T antigen oncogene, under the control of the insulin promoter (expressed
exclusively in pancreatic β cells) (Hanahan D., 1985). The mice develop multiple β cell
tumours which proceed through well defined stages and acquire the different hallmarks
of cancer, including the “angiogneic switch” occurring in hyperplasic islets which will go
on to become tumours (Hanahan D., 1985; Akerblom et al., 2012; Olson et al., 2009).
There have been no such studies of miRNA expression and function in SBNET due to an
absence of murine models of SBNET.
In the Olson et al. (2009) study the mouse miRNA known at the time of the study,
430 miRNA, were quantified at different points in tumourigenesis in the mouse model
163
(Olson et al., 2009). This included normal islets, hyperplastic islets, angiogenic islets, tu-
mours (n=39) and liver metastases (n=6). Each tumourigenesis step was associated with
differences in the expression of a large number of different miRNA. These included the
miR-200 family of miRNA, the downregulation of which was correlated with a metastatic
profile. The expression of the miR-200 family was low in most of the tumours but up-
regulated in liver metastases and in a small subset of the primary tumours which had a
metastasis-like miRNA profile.
There were 34 miRNA that were differentially expressed between normal islets and
the primary tumours, whereas for the angiogenic islets compared to the primary tumours
there were only 10 miRNA that were differentially expressed, suggesting that the majority
of miRNA differences between normal tissue and the primary tumour occurred during
the pre-tumour stages in tumourigenesis (Olson et al., 2009). Increased miR-155 and
miR-142-3p expression was associated with hyperproliferative hyperplasic islets, in which
these miRNA were upregulated 5.0 and 6.4 fold respectively compared to normal islets.
Expression of these two miRNA was decreased at later disease stages however, including
in the primary tumour. MiR-155 was also identified in the human miRNA study by Roldo
et al. (2006), where it was reduced in the PNET compared to normal tissue.
RIP-Tag2 mice were also treated daily for 7 days with sunitinib or a vehicle and their
tumours were then dissected (Olson et al., 2009). Interestingly, several miRNA that had
been found to be upregulated previously in angiogenic islets, miR-424, miR-126, and
miR-21 were found to have their expression reduced by the anti-angiogenic treatment.
A recent study reanalysed previously published data to enable comparisons to be made
between the miRNA and mRNA profiles of human PNET and those of the RIP-Tag2 mice
(Sadanandam et al., 2015). This was to determine the usefulness of the mouse model by
investigating how closely tumours in the mouse resembled human PNET. Sadanandam
et al. (2015) used miRNA data from 40 human PNET (Roldo et al. (2006)), data from
the miRNA profile of the RIP-Tag2 mice (Olson et al. (2009)) and data from a study of
164
the mRNA profile of 86 PNET samples (Missiaglia et al. (2010)).
Based on this expression data the study found that the RIP-Tag2 mouse tumours could
be classified into two subtypes, based on having a miRNA and mRNA profile resembling
that of primary tumour tissue from either human insulinomas or human G3 primary
tumours (Sadanandam et al., 2015). The G3 primary tumours were found to have a
miRNA/mRNA expression profile similar to that of liver metastases and expressed genes
involved in early pancreatic development, whereas the insulinomas expressed mature islet
cell genes, suggesting that different tumourigenisis mechanisms may be involved. These
findings suggest that the RIP-Tag2 mouse model is of particular usefulness in the study
of these two types of human PNET and may conversely be of limited usefulness for the
study of non-functioning low grade PNET.
+++Tissue and serum comparison A global profiling study in serum and tissue
samples was done to investigate the levels of 754 miRNA in PNET patients (Thorns
et al., 2014). Samples included were, FFPE tissue from 37 PNET patients, 9 patients
with non-neoplastic pancreas morphology, 7 normal microdissected pancreatic islets and
serum samples from 27 PNET patients and 15 healthy volunteers. Each tissue type had
a distinct miRNA profile. Mi-642 expression was positively correlated with Ki-67 %
(p=4.0 x 10−6) while miR-210 was upregulated in the primary tumour of patients with
metastases (p=7.4 x 10−5). There was little overlap between the dysregulated miRNA
in PNET tissue and those dysregulated in serum, however miR-193b was upregulated
in both tissue and serum from PNET patients and so could be a potential biomarker if
further validated.
Another study, primarily investigating pancreatic ductal adenocarcinoma to identify
biomarkers for the early diagnosis of cancer in individuals with a family history of the
disease, included serum samples from PNET patients as one of the comparison groups
(Slater et al., 2014). The study identified increased levels of two circulating miRNA,
165
miR-196a and miR-196b, which together had a sensitivity of 1.0 and specificity of 0.9 for
the diagnosis of pancreatic cancer (or the high grade multifocal pancreatic intraepithelial
lesions (PanIN) which proceed it) (Slater et al., 2014). A Receiver Operating Character-
istic (ROC) curve showed an area under the curve value of 0.99. MiR-196a and miR-196b
miRNA were significantly higher in serum samples, taken pre-operatively, of histologi-
cally confirmed sporadic and familial pancreatic cancer (n=19)/high grade PanIN (n=5)
than in PNET patients (n=10) healthy controls (n=10) and individuals with a family
history of pancreatic cancer but no lesions (n=5). These results suggest that circulating
miR-196a and miR-196b may have potential as future biomarkers for pancreatic cancer
and could be used for early diagnosis in individuals at risk of developing pancreatic can-
cer, particularly since they appear to be unaffected by the presence of other types of
pancreatic pathology such as PNET.
+++MEN1 and miR-24-1 A study carried out detailed functional characterisation
of miR-24-1 in a PNET cell line (Luzi et al., 2012). There have been no such in vitro
studies carried out in SBNET. The study by Luzi et al. (2012) showed that this miRNA
could be acting as a mimic for the second hit of the Knudson’s hypothesis in NET
patients who have one germ line mutation in MEN1 but have not yet had undergone loss
of heterozygousity (Knudson, 1971; Knudson, 1974; Luzi et al., 2012). The study also
demonstrated in vitro in the BON1 cell line that the 3’UTR of the mRNA for MEN1 was
targeted by miR-24-1 (highly conserved seed region: 599-605). An antisense based loss of
function assay showed that binding via this seed region was required for miRNA-mRNA
binding (luciferase reporter assay).
Northern blot (mature miR-24-1) and western blot (menin) experiments showed that
introduction of a selective inhibitor of miR-24-1 (antisense oligonucleotides specific to
miR-24-1) in BON1 cells, caused reduced miR-24-1 expression and increased menin levels
compared to controls (due to inhibited endogenous miR-24-1 expression) (Luzi et al.,
166
2012). In contrast overexpression of miR-24-1 pre-miRNA caused increased miR-24-1
expression and reduced menin levels compared to controls (due to increased mature miR-
24-1 expression). Controls were mutated versions of pre-miR-24-1 antisense miR-24-1.
The study also investigated miR-24-1 and menin levels in parathyroid adenoma tissue
from MEN1 patients with (n=4) and without (n=4) loss of heterozygosity for MEN1,
sporadic parathyroid adenomas (n=3) and normal parathyroid tissue from patients op-
erated on for thyroid carcinoma (n=3) (Luzi et al., 2012). The MEN1 (-/-) parathyroid
adenomas had no expression of MEN1 or miR-24-1 while the MEN1 (+/-) parathyroid
adenomas had overexpression of miR-24-1, reduced MEN1 mRNA (qPCR) and no menin
protein (western blot). The expression levels of miR-24-1 and MEN1 in the sporadic
parathyroid adenomas were the same as in the normal parathyroid tissue, with low miR-
24-1 expression and high menin levels suggesting that this may not be an important
pathway in these patients.
These findings suggest that miR-21-1 is silencing the expression of the functional copy
of the MEN1 gene in the heterozygous patients (+/-), causing these patients to have the
same phenotype with respect to menin protein levels as the patients who had lost both
alleles of MEN1 (-/-). The authors then did a chromatin immunoprecipitation assay for
menin, DNA that co-precipitated with menin was analysed by qPCR of the genomic locus
containing the promoter region for miR-21-1. The promoter region was only occupied
by menin in the MEN1 (+/-) parathyroid adenoma tissue but not in the MEN1 (-/-)
tissue or with control IgG. The authors proposed based on these findings that there is
a “negative feedback loop” with miR-21-1 expression being activated by menin binding
to the miR-21-1 promoter, while the mature miR-21-1 goes onto suppress MEN1 mRNA
expression thus having a knock on effect in reducing menin levels (Luzi et al., 2012).
167
2.5.5. Summary
MiRNA have an important physiological role as epigenetic regulators of gene expression.
The expression of miRNA is altered in different physiological states and disease states.
During tumourigenesis the levels of certain miRNA become dysregulated triggering al-
tered expression of their gene targets.
Studies investigating miRNA in SBNET are scarce compared to PNET where there
has been a more extensive characterisation of the role of miRNA in tumourigenesis.
Nevertheless, in SBNET, a reduction in miR-133a expression in SBNET metastases was
consistently identified (in the two existing studies) despite differences in study design
(Ruebel et al., 2010; Li et al., 2013b). Increased understanding of miRNA function and
dysregulation in different disease states is needed to enable the tumourigenesis pathways
that occur in GEP-NET patients to be better understood. This could lead to the discovery
of novel therapeutic targets, either the miRNA themselves or the genes they are silencing.
Since miRNA are very stable and can be quantified in FFPE tissue and in serum/-
plasma as well as in frozen samples they have great potential for use as future biomarkers
in SBNET and PNET patients. Future miRNA biomarkers could be used for patient
stratification, prediction of the disease course and treatment response. Serum miRNA
biomarkers having the potential to be used throughout the patient journey, to monitor
factors such as treatment response and to give an early indicator of disease progression/re-
currence.
2.6. Biomarkers
Biomarkers are biological markers that can be measured accurately and reproducibly, to
provide information about a particular physiological state, disease state or response to
therapy (Strimbu and Tavel, 2011).
Tumour biomarkers may have been identified as early as 1965 with tumour specific
168
antigens being identified in colon carcinomas (Gold and Freedman, 1965; Foroutan, 2015;
Foroutan, 2015). Interest in the use of tumour biomarkers greatly accelerated in the
1990s, with the early use of tumour biomarkers in the clinic including prostate specific
antigen in prostate cancer (Barry, 1998; Foroutan, 2015).
There are now many different cancer biomarkers being used in a clinical setting. Ex-
amples of well established cancer biomarkers include the estrogen receptor (ER), the
progesterone receptor (PgR) and human epidermal growth factor receptor 2 (HER2) as-
sessed in breast cancer (Weigel and Dowsett, 2010; James et al., 2007). These biomarkers
are used to stratify patients into clinically useful groups and are both prognostic and pre-
dictive of treatment response (Weigel and Dowsett, 2010).
Single tumour biomarkers rarely provide sufficiently powerful information to warrant
their clinical use in a particular disease due to the inherent complexity of the tumouri-
genesis pathways involved which a single molecule is unable to recapitulate (Weigel and
Dowsett, 2010; Modlin et al., 2016). Biomarker signatures or panels of biomarkers specific
to a disease are better able to stratify patients into clinically relevant subgroups based
on the underlying tumour biology and have the potential to enable tailored treatment.
IHC has been the mainstay of biomarker detection in the past however newer molecular
techniques using qPCR, microarrays and fluorescent in situ hybridization (FISH) are be-
coming more readily available for use in the clinic. These techniques can use much smaller
amounts of starting material than IHC and enable the parallel detection and quantifica-
tion of multiple biomarkers. These biomarkers can then be used to better stratify patients
into distinct disease subgroups based on biological differences in tumourigenesis enabling
more reliable prediction of clinical factors such as survival or treatment response.
Newer classes of biomarkers being used in the clinic include the Oncotype DX signa-
ture, which uses qPCR to quantify the expression of 16 breast cancer related genes to
predict the risk of disease recurrence and the CellSearch system which quantifies circulat-
ing tumour cell numbers to predict survival in metastatic prostate, breast and colorectal
169
cancer (Goldstein et al., 2008; Weigel and Dowsett, 2010; De Bono et al., 2008; Cristo-
fanilli et al., 2004; Cohen et al., 2008a). Increasingly companion biomarkers are being
developed alongside new targeted therapies due to these therapies only being effective in
small subgroups of patients to avoid unnecessary toxicity and expense in treating patients
who are unlikely to benefit from a particular therapy (Duffy and Crown, 2013).
Advances in ‘omics’ technologies, including genomics, transcriptomics, epigenomics,
microbiomics and metabonomics aimed at a better understanding of disease biology have
also identified a plethora of new biomarker signatures representing particular tumour sub-
types (Vargas and Harris, 2016). Biomarkers identified should be subjected to extensive
analytical validation, clinical validation and an assessment of clinical utility with prospec-
tive randomised clinical trials to prove benefit over current approaches prior to widespread
adoption (Henry and Hayes, 2012). Analytical validation of a potential biomarker in-
cludes an assessment of the sensitivity and specificity or accuracy of the biomarker for
its intended purpose, for example the diagnosis of a disease or prediction of a particular
disease outcome or treatment response (see Figure 2.4) (Bossuyt et al., 2003). The per-
formance of the biomarker must also be reproducible both within a single laboratory and
in other laboratories (Henry and Hayes, 2012).
Biomarkers can be used in national screening programs for cancer, for example a popu-
lation based screening program for colorectal cancer (age: 60-69), has been in place since
2010 in the UK (Logan et al., 2012). This involves the detection of haemoglobin in faeces
using a guaiac-based faecal occult blood test (Young et al., 2015). Early results showed
a screening uptake of 55-60 % and suggested that so far the screening programme was on
track to achieve a 16 % reduction in overall bowel cancer mortality in keeping with the
findings of European clinical trails (Logan et al., 2012).
National screening programs are of limited usefulness for rare cancers. This is because
the risks of a potential screening program in terms of false positives and public anxiety,
with the expense and potential harm involved in further investigations, are unlikely to
170
Figure 2.4.: A good biomarker should have both high sensitivity and high specificity, thisminimises the numbers of false negatives and false positives respectively A)High sensitivity, low specificity, (many samples passed the test that shouldhave failed it) B) Low sensitivity, high specificity (many samples failed thetest that should have passed it). Red circle: false positive, blue circle: falsenegative, open circle: true negative/true positive. Images from Rmostell,reproduced from (Rmostell, 2011a) and (Rmostell, 2011b), creative commonslicence: CC0 1.0
be outweighed by the benefits to a small number of individuals (see Figure 2.4).
2.6.1. Established biomarkers
There are two main biomarkers in clinical use for GEP-NET, these are CgA measured in
the serum/plasma and Ki-67 % used for tumour grading (Niederle et al., 2016). Consensus
conferences held in 2012 in London, UK and in 2014 in Nashville, USA both identified
biomarkers as a crucial area of unmet need for the management of GEP-NET patients
(Frilling et al., 2014; Oberg et al., 2015). In particular the need for the development and
validation of new biomarkers or panels of biomarkers in tissue, blood and urine which
could indicate metastatic growth, predict prognosis, predict/monitor treatment efficacy
and be used for the early identification of disease progression/recurrence (Frilling et al.,
2014). For potential future biomarkers being developed for use in patients with GEP-
NET please see sections 2.5.2 (for miRNA biomarkers) and 2.6.2 (for other classes of
biomarker).
171
CgA
Serum and plasma CgA levels are measured in biochemical tests in the clinic and can
indicate the presence of a neuroendocrine tumour. There are limitations with this ap-
proach since serum/plasma CgA lacks specificity with false positive results being caused
by other conditions, for example chronic atrophic gastritis (Niederle et al., 2016).
CgA is expressed in most normal cells of the diffuse neuroendocrine system and also in
the tumours arising from these cells, but not in normal tissue or in tumours that do not
arise form the diffuse neuroendocrine system (Helman et al., 1988). It is a component of
dense core secretory vesicles and regulates the budding of these granules from the golgi
apparatus (Giovinazzo et al., 2013; D’amico et al., 2014). It secreted from neuroendocrine
cells upon certain stimuli along with the hormones produced by these cells (Wollam et
al., 2017). CgA IHC is used alongside synaptophysin IHC and haematoxylin and eosin
staining for the pathological diagnosis of a SBNET following a biopsy or surgical resection
of the tumour (Niederle et al., 2016).
Functional studies have shown that there is some redundancy in the role of CgA, in
β cells of the mouse at least, with mice with an islet specific CgA knockout showing
overexpression of two other members of the granin family, chromogranin B (CgB) and
secretogranin II (Wollam et al., 2017). Wollam et al found that while CgB and secre-
togranin II were able to compensate in part for the production of secretory granules, there
were larger numbers of mature secretory granules at the expense of immature ones and
a higher granule insulin content. This suggests that CgA is important for regulating the
relative numbers of different vesicle types and regulating the concentration of vesicle con-
tents. More studies will need to be done with knockouts in other cell types to determine
if these findings hold true for other types of neuroendocrine cells.
There are limitations in the use of serum/plasma CgA as a diagnostic biomarker in
GEP-NET. This is in part due to there being no universally validated diagnostic method-
ology for the measurement of this biomarker (Gut et al., 2016). There are a number of
172
different methodological approaches available. One study used 46 GEP-NET patients
and 31 controls to compare the sensitivity and specificity of plasma CgA measurements
using these different methods (Stridsberg et al., 2003). The findings from this study
were that the radioimmunoassay performed the best (sensitivity of 93 %, specificity: 85
%), followed by the enzyme-linked immunosorbent assay (ELISA) (sensitivity: 85 %,
specificity: 85 %) and the immunoradiometric assay (sensitivity: 67 %, specificity 96 %)
(Stridsberg et al., 2003).
The commercially available assays each use different antibodies and CgA can be mea-
sured in serum or plasma, with higher levels being found in plasma (Gut et al., 2016).
All these factors make establishing proper cut off levels and making comparisons between
experimental studies extremely challenging.
In addition to this methodological variability in the measurement of serum/plasma
CgA there is a lack of overall specificity due to the high frequency of false positives. High
serum/plasma CgA can be caused by other conditions and by the use of proton pump
inhibitors leading to serum/plasma CgA being elevated in the absence of a NET (Gut et
al., 2016). Chronic atrophic gastritis, liver cirrhosis, impaired kidney function, congestive
heart failure, inflammatory bowel disease and the presence of non-GEP-NET tumours
such as hepatocellular carcinoma are some of the conditions that trigger increased CgA
levels (Gut et al., 2016; Niederle et al., 2016). These differential diagnoses increase the
complexity involved in the interpretation of serum/plasma CgA results and limits the
clinical utility of serum/plasma CgA as a diagnostic biomarker.
A recent study compared serum CgA to serum CgB measurements in PNET (n=91) pa-
tients to patients with pancreatic cancer (n=52), chronic pancreatitis (n=54) and healthy
age and sex matched controls (n=104) (Miki et al., 2017). Miki et al found that CgB
had a sensitivity and specificity of 72 % and 77 % respectively, similar to the values for
CgA (sensitivity: 79 %, specificity: 64 %). Unlike CgA, CgB levels were not found to be
increased by the use of proton pump inhibitors or with increasing age, renal impairment
173
however, increased the levels of both analytes. CgB was better at distinguishing patients
with PNET from those with other pancreatic pathologies. These findings suggest that
serum CgB may represent a more specific biomarker for the detection of PNET.
Not much work has been done to investigate the biological variation in serum/plasma
CgA, this would be important for the establishment of a physiological baseline for this
biomarker. Results from one study of 22 healthy volunteers (5 samples taken each) showed
that serum CgA levels are significantly higher (p=0.01) in women than in men (Braga
et al., 2013).
A recent meta analysis investigated the diagnostic utility of circulating CgA for NET,
13 studies met their inclusion criteria, with a total of 1260 NET patients and 967 healthy
controls (Yang et al., 2015). Yang et al found that circulating CgA had an overall figures
for sensitivity of 0.73 (95 % confidence interval: 0.71 to 0.76) and a specificity of 0.95 (95
% confidence interval: 0.93 to 0.96).
It should be noted that the controls used in the studies included in the meta analysis
were healthy people (Yang et al., 2015). So they lacked the conditions which could have
been confounding factors in the case of a ‘false’ positive result. This means that the true
specificity for the test in a clinical setting could be lower as some of the patients that
would undergo circulating CgA biochemical tests are likely have the conditions that could
cause a differential diagnosis.
Increasing levels of plasma CgA has been found to be associated with the accumulation
of metastases in the liver (Gut et al., 2016). Higher CgA levels are associated with
metastatic disease and worse survival in SBNET and PNET but not in some other types
of NET such as gastrinoma (Gut et al., 2016). This raises the possibility for plasma
CgA to be used as a prognostic biomarker. A study of plasma CgA in well differentiated,
metastatic GEP-NET (n=344) found that elevated CgA levels were associated with worse
median and 5 year survival (Arnold et al., 2008).
Increasing plasma CgA levels may also indicate disease recurrence after radical surgery
174
(Niederle et al., 2016). A retrospective study was carried out of SBNET patients (n=56)
followed up after radical surgery with plasma CgA being measured 1-3 times per year
(Welin et al., 2009). The study found that 33 patients had disease recurrence after a
median time of 32 months (range: 6–217), with 28/33 having their recurrence being
detected by elevated plasma CgA prior to being detectable by CT/MRI imaging. In
3/28 patients the recurrence was also simultaneously visible on somatostatin receptor
scintigraphy or PET with 5- hydroxytryptophan. The authors suggest that plasma CgA
could represent a less expensive option for the follow up of these kinds of patients, with
plasma CgA measurements twice a year and an annual ultrasound with further imaging
only being done if clinical symptoms emerge or CgA becomes elevated (Welin et al.,
2009).
An earlier study of GEP-NET patients (n=127) showed that in 83.3 % of cases, in-
creased plasma CgA during follow up was associated with disease progression (Bajetta
et al., 1999). In the presence of liver lesions increases in CgA were associated with pro-
gressive disease in 100 % of the cases. The GEP-NET were analysed as a whole without
individual analyses being done based on primary site.
These results suggest that plasma CgA is useful for the follow up of patients with
SBNET to identify disease recurrence early and this is represented in the current ENETS
guidelines (Niederle et al., 2016). Further studies are still needed to increase the body of
evidence in this area, with investigations to see if the results can be replicated in larger
numbers of patients, in other centres and for other types of GEP-NET.
The clinical utility of serum/plasma CgA as a diagnostic biomarker in GEP-NET is
currently limited by low specificity and a lack of standardised measurement methodolo-
gies. Early results suggest that increased plasma CgA has the potential to be used in the
future as an early indicator of disease recurrence in SBNET patients who have undergone
radical surgery. Despite the use of plasma CgA biochemical tests being recommended
by ENETS for the follow up of GEP-NET, further work needs to be done to validate
175
the utility of this biomarker for identifying disease recurrence in larger patient cohorts
and different primary sites and to investigate plasma CgA with respect to response to
different treatment modalities.
Synaptophysin
Synaptophysin is a membrane glycoprotein found in presynaptic vesicles that is expressed
in both normal neuroendocrine cells and in NET (Wiedenmann et al., 1986). IHC for
synaptophysin is used alongside CgA in the pathological diagnosis of a GEP-NET to
establish that a tumour is neuroendocrine in nature (Falconi et al., 2012; Niederle et al.,
2016; Lam and Lo, 1997). G3 GEP-NET usually express synaptophysin but may lack CgA
expression, while G1/G2 tumours are usually positive for both of these neuroendocrine
markers (Sorbye et al., 2013; Rindi et al., 2007).
5-hydroxy indole acetic acid
5-hydroxy indole acetic acid (5-HIAA) is a major metabolite of serotonin that can be
readily measured in urine samples. Urinary 5-HIAA is used as a biomarker in patients
with SBNET since these tumours produce serotonin and exhibit elevated levels of 24
hour urinary 5-HIAA on biochemical tests (Niederle et al., 2016). Serotonin itself is an
unreliable biomarker due to large biological variations in serum serotonin levels between
individuals due to serum serotonin being rapidly absorbed by platelets (Gedde-Dahl et
al., 2013). In contrast to serum serotonin, 24 hour urinary 5-HIAA has a high sensitivity
and specificity for detecting the presence of carcinoid syndrome of up to 100 % and 85-90
% respectively (Niederle et al., 2016).
24 hour urinary 5-HIAA is assessed as part of the biochemical tests during the diag-
nostic work up for a suspected SBNET and for patient follow up including response to
treatment and if disease recurrence/progression is suspected (Niederle et al., 2016). To
avoid false positive results it is necessary for patients to avoid foods rich in serotonin
176
prior to and during testing (Ardill and Erikkson, 2003; Niederle et al., 2016). It has been
suggested that 8 hour urine sampling of 5-HIAA may be as accurate as 24 hour urinary
5-HIAA for patient follow up (Gedde-Dahl et al., 2013). This could have the potential
to make the follow up tests for patients with carcinoid syndrome less time consuming if
these results can be reproduced in controlled clinical trials.
Ki-67 %
Ki-67 was first identified in 1983, when a mouse monoclonal antibody was produced that
bound to a nuclear protein that was expressed during cell proliferation but not in cells
that were in a resting state (Gerdes et al., 1983). Ki-67 was proposed as a biomarker of
proliferation on the basis of it being expressed during cell cycle phases G1, S and G2 but
not G0 (Weigel and Dowsett, 2010). Little is known about the function of Ki-67 although
there have been some studies suggesting that it is present in the mitotic chromosome
periphery where it may have a role in stabilising chromosomes after nuclear envelope
disassembly (Booth et al., 2014; Cuylen et al., 2016).
The Ki-67 proliferation index, is the only clinically approved prognostic biomarker used
in GEP-NET. It is used to grade patients from low to high grade based on increasing
proliferation levels, G1 (≤ 2 %), G2 (3-20 %) and G3 (> 20 %) (Rindi et al., 2006; Rindi
et al., 2007; Niederle et al., 2016; Falconi et al., 2016). For details of Ki-67 % grading in
GEP-NET please see section 2.2.3.
+++Heterogeneity GEP-NET are heterogeneous neoplasms both in their tumour bi-
ology, signalling pathways, expression patterns and clinical behaviour (Cives et al., 2016;
Cortez et al., 2016; Wang et al., 2013; Sadanandam et al., 2015; Briest and Grabowski,
2014). During tumourigenesis tumours acquire new mutations as tumour cells proliferate
leading to a heterogeneous population of cells within the tumour and within any sub-
sequent metastases (see Figure 2.5) (Gerlinger et al., 2012; Navin et al., 2011). This
177
Figure 2.5.: Intertumoural and intratumoural heterogeneity develops over time as addi-tional mutations are acquired by the cells within tumours and their metas-tases. This leads to metastasis 1 being made up of a different population ofcells with different mutation profiles and characteristics to those of metastasis2.
heterogeneity extends to Ki-67 expression which shows considerable heterogeneity both
when assessed at different points within a single lesion or metastasis and also between
different tumours taken from the same patient (Yang et al., 2011; Shi et al., 2015; Singh
et al., 2014; Couvelard et al., 2009).
This presents a challenge for grading since this is currently done only in one lesion for
each patient and in either the primary or a metastasis. This snapshot Ki-67 % assessment
from a single lesion may not be representative of the true score for that patient and
may not be comparable between patients. These differences can be sufficient to change
the grade of a tumour with the potential to affect patient management since treatment
decisions may be based on tumour grade. More studies are needed to assess the extent
of intratumoural and intertumoural heterogeneity in Ki-67 % and to determine in what
proportion of cases this is sufficient to lead to a change in grade.
178
+++Prognostic biomarker Ki-67 is currently the only prognostic biomarker used
clinically in GEP-NET. It was first recommended for use in these tumours by ENETS in
2006 and this was adopted in the WHO classification of 2010 (Rindi et al., 2006; Rindi et
al., 2007; Niederle et al., 2016; Falconi et al., 2016; Garcia-Carbonero et al., 2010). Grade
as assessed by Ki-67 % corresponds to differing prognosis, with G3 tumours having much
worse survival than G1 and G2 tumours (Rindi et al., 2012). As such, Ki-67 % is the
most extensively validated and widely adopted indicator of prognosis in these tumours.
As more data emerges from newer studies about the molecular pathways that are dis-
rupted in GEP-NET, novel prognostic biomarkers are likely to be identified that provide
further more detailed information for patient prognostic stratification, beyond that cur-
rently provided by Ki-67 %. These will need to be extensively tested and validated
however in randomised clinical trials before they can be adopted in the clinic for use
alongside Ki-67 %.
Ki-67 expression can be assessed by IHC on FFPE tissue as part of the routine
histopathlogical work up for GEP-NET using equipment in readily available in clinical
laboratories, it does not rely on more sophisticated equipment such as qPCR machines.
The efficacy of Ki-67 % grading in providing useful prognostic information in GEP-
NET has been supported by multiple single centre retrospective studies with small patient
numbers and in several large retrospective multicentre national and international studies
(Jamali and Chetty, 2008; Pape et al., 2008; Pelosi et al., 1996; La Rosa et al., 1996;
Klimstra et al., 2010; Khan et al., 2013a; Yamaguchi et al., 2013; Cherenfant et al., 2013;
Ahmed et al., 2009; Garcia-Carbonero et al., 2010; Jann et al., 2011; Rindi et al., 2012).
These studies established tumour grade, as assessed by Ki-67 %, and disease stage as
independent predictors of survival in patients with GEP-NET.
The largest study was of 1072 PNET patients and found that grade (Ki-67 %), stage,
and curative surgery were independent predictors of survival (Rindi et al., 2012). Patients
in the study were followed up for a minimum of 2 years and during this period there were
179
tumour related deaths in 63.5 % of the G3 PNET patients, compared to 26.6 % and 6.6
% for patients with G2 and G1 tumours respectively. A multicentre study including 270
intestinal NET found that the relative risk of death for G3 tumours was 11 fold higher
than that of G1 tumours (Jann et al., 2011). The 5 year survival rates for the jejunoileal
NET included in the study (n=214) were 50.0 % for G3 tumours compared to 83 % and
93.8 % for G2 and G1 tumours respectively.
The studies of Ki-67 % in GEP-NET patients had differences in study design and in the
reported data, with some studies not following recommendations made for the reporting
of tumour marker prognostic studies (REMARK guidelines) (Harris et al., 2007; McShane
et al., 2005; Altman et al., 2012). This leads to challenges when making comparisons
between studies if information is lacking or there are large differences in methodology,
for example Ki-67 % cut off levels. The studies were also universally retrospective and
most had small patient numbers which limits the level of evidence for the clinical rec-
ommendations that can be made. Nevertheless the finding that Ki-67 % grading is an
independent predictor of GEP-NET survival has remained robust.
Ki-67 % is very useful in GEP-NET for identifying the rare, poorly differentiated, G3
tumours (Ki-67 > 20 %) from the well differentiated, G1/G2 tumours. G3 tumours make
up only approximately 5 % of gastrointestinal NET and 7 % of PNET, they are however
much more common in the lung (small cell carcinoma) (Garcia-Carbonero et al., 2010;
Garcia-Carbonero et al., 2016). G3 tumours are associated with very poor survival, < 30
% at 5 years, with only 5 % of patients being long-term survivors (Garcia-Carbonero
et al., 2010; Rindi et al., 2012; Garcia-Carbonero et al., 2016).
Many studies now suggest that G3 GEP-NET have a different tumourigenesis path-
way to G1/G2 tumours which could explain their much more aggressive behaviour. G3
tumours have different mutations to well differentiated tumours, for example 57 % of G3
NET have somatic inactivating mutations in tumour suppressor TP53 and this is 95 %
for G3 PNET (Yachida et al., 2012; Vijayvergia et al., 2016; Garcia-Carbonero et al.,
180
2016). TP53 mutations are very rare in the well differentiated G1/G2 tumours. Gene
expression studies indicate that in PNET G3 tumours express pancreatic progenitor spe-
cific genes while G1/G2 PNET express mature β cell genes, suggesting that G3 tumours
may arise from neuroendocrine progenitor cells rather than terminally differentiated cells
(Sadanandam et al., 2015; Schimmack et al., 2011).
These differences are reflected in the most recent ENETS guidelines (2016), where G3
GEP-NET are referred to as neuroendocrine carcinoma (NEC), while the G1/G2 GEP-
NET are classed as neuroendocrine neoplasms (NEN) (Garcia-Carbonero et al., 2016;
Niederle et al., 2016; Falconi et al., 2016).
Well differentiated G1/G2 tumours represent the majority of SBNET and PNET
(Niederle et al., 2016; Garcia-Carbonero et al., 2016). They are very heterogeneous
tumours with less predictable angiogenesis behaviour than G3 tumours (which are vir-
tually all metastatic) and this presents challenges for their clinical management. Liver
metastases are a common occurrence in well differentiated SBNET and in non-functioning
PNET including those with a Ki-67 % of ≤ 2 % (Jann et al., 2011; Shi et al., 2015; Norlen
et al., 2012; Frilling et al., 2014). Distant metastases can occur even when the primary
lesions are small (≤ 2 cm) and are a strong predictor of survival (Haynes et al., 2011;
Cherenfant et al., 2013; Tamburrino et al., 2016; Ahmed et al., 2009; Rindi et al., 2012;
Jann et al., 2011).
While Ki-67 % is effective at identifying well differentiated tumours (G1/G2), from the
poorly differentiated G3 tumours, Ki-67 % alone provides little additional information
on the behaviour of these tumours. In particular Ki-67 % is unable to determine which
patients with G1 and G2 tumours will have a more aggressive tumour phenotype or to
predict distant metastases (associated with worse survival) or disease recurrence (Frilling
et al., 2014; Cherenfant et al., 2013; La Rosa et al., 1996; Yamaguchi et al., 2013). It
would be interesting to have more studies which investigate the relationship between
grade and stage in GEP-NET to provide further information about the proportions of
181
well differentiated tumours which might have a more aggressive phenotype.
If new biomarkers could stratify patients further within the well differentiated (G1/G2)
tumour group, this information would help with patient management decisions since there
are a large number of possible treatment modalities available for these patients but there
is a low level of evidence for which treatments might be of most benefit to a particular
patient.
Although there have been a large number of studies on Ki-67 %, limitations remain.
These include a lack of consensus about cut off values, intertumoural and intratumoural
heterogeneity (with the potential for undergrading), differences in IHC methods, intra-
observer error and the use of inaccurate eyeballing estimates in some centres, rather than
assessing 2000 tumour cells for positive staining as recommended by ENETS guidelines
(see section 2.2.3), causing reproducibility problems (Kim and Hong, 2016; Niederle et al.,
2016; Rindi et al., 2007; Oberg et al., 2015; Khan et al., 2013a).
Automated counting for cells staining positive for Ki-67 % using image processing
software is becoming more widely available and may reduce intra-observer errors, however
these techniques need to be further validated in GEP-NET (in particular to ensure that
lymphocytes which show Ki-67 positivity on IHC are excluded from counts) (Oberg et al.,
2015; Yamaguchi et al., 2013; Kim and Hong, 2016; Fujimori et al., 2012).
Although G2 tumours have worse survival than G1 tumours, this is much less dramatic
than for G3 tumours, with some studies finding no significant difference in survival be-
tween G1 and G2 tumours in SBNET and PNET when staging data was not considered
(Jann et al., 2011; Rindi et al., 2012). This suggests that Ki-67 % should not be assessed
in isolation of other factors such as disease stage and primary site when determining
patient prognosis. There is considerable debate about whether the Ki-67 % cut off for
G1 and G2 tumours should be modified, since several studies have shown that a cut off
value of ≤ 5 % (rather than ≤ 2 %) is a better predictor of prognosis (Pelosi et al., 1996;
Khan et al., 2013a; Pape et al., 2008; Yamaguchi et al., 2013).
182
Despite Ki-67 % being widely used and validated as a prognostic biomarker in GEP-
NET, it remains a single biomarker representing an indication of proliferation levels but
not of other pathways dysregulated in these tumours. In the future novel GEP-NET
biomarkers may help to provide a more nuanced interpretation of the Ki-67 score. These
could enable the more common well differentiated tumours to be stratified further based
on clinically useful characteristics such as tumour aggressiveness or treatment response.
2.6.2. Potential future biomarkers for use in patients with
SBNET
Genetics
The vast majority of SBNET are sporadic although there have been a small number of
case reports of families with several affected members (Cunningham et al., 2011). SBNET
patients rarely have inherited mutations in genes such as MEN1 or VHL in contrast to
PNET patients (see section 2.1.3). Somatic mutations resulting in genomic instability
such as inactivating mutations in DAXX or ATRX are a common occurrence in PNET
but not in SBNET (Minnetti and Grossman, 2016; Marinoni et al., 2014).
A study of 48 SBNET, in which massive parallel exome sequencing was carried out, con-
firmed that the mutation rate was low in SBNET with 0.1 somatic single nucleotide vari-
ants on average per 106 nucleotides and mainly C>T and A>G transitions, characteristics
of a stable cancer (Banck et al., 2013; Miller et al., 2015b). Somatic copy number analysis
showed that on average tumours had 12.6 amplifications and 8.7 deletions, in particular,
gains of chromosomes 4, 5, 19 and 20 and losses of chromosomes 11 and 18 (Banck et al.,
2013; Miller et al., 2015b). 33 % of patients had mutations in AKT1/AKT2/MTOR path-
way genes and 46 % of patients had mutations in SMAD/TGFβ pathway genes (Banck
et al., 2013; Miller et al., 2015b). 197 somatic, nonsilent, single nucleotide variants
were identified in a large number of different genes including SMAD1, FGFR2, MEN1,
183
HOOK3, EZH2, MLF1, VHL, NONO and CARD11 (Banck et al., 2013; Miller et al.,
2015b).
Many studies have demonstrated loss of Chr18 in SBNET, with this occurring in 61-78
% of tumours (Banck et al., 2013; Kulke et al., 2008; Kim et al., 2008b; Francis et al.,
2013; Hashemi et al., 2013; Kim et al., 2008a; Cunningham et al., 2011; Miller et al.,
2015b). Interestingly a study showed that there was an association between the loss of
Chr18 and global reductions in methylation in SBNET (Fotouhi et al., 2014; Miller et
al., 2015b). These findings suggest that the loss of Chr18 in SBNET may be triggering
epigenetic changes in tumour cells. DNA methylation changes that occur in SBNET are
described in more detail later in this section. Interestingly the gain of Chr14 in SBNET
without the loss of Chr18 was associated with worse survival suggesting that this could
be a useful prognostic biomarker to identify this subset of SBNET patients with more
aggressive behaviour (Andersson et al., 2009; Miller et al., 2015b).
A study in 30 patients with SBNET found that 23 % of these patients had mutations
in APC which is part of the Wnt pathway (Bottarelli et al., 2013; Miller et al., 2015b).
A recent investigation of 180 SBNET patients by whole genome and exome sequencing
by Francis et al. (2013) revealed that 8 % of SBNET had somatic frame-shift mutations
or hemizygous deletions of CDKN1B (Banck and Beutler, 2014; Miller et al., 2015b).
CDKN1B encodes the cyclin dependent kinase inhibitor p27Kip1 (p27) which is involved
in cell cycle regulation by inhibiting cyclin dependent kinases required for the progression
of the cell cycle (Quraish et al., 2016). In mouse studies, null mice, Cdkn1b (-/-), develop
organomegaly, pituitary adenomas and grow to twice the size of their normal counterparts
while induced overexpression of Cdkn1b in the cells of adult transgenic mice suppressed
cellular proliferation in all tissue types examined (Pruitt et al., 2013).
Interestingly, in contrast to other cyclin dependent kinase inhibitors, p21 and p16Ink4a,
p27 is activated to suppress cell cycle progression by extracellular antiproliferative signals
not by TP53 or RB1 signalling pathways (DNA damage response) (Pruitt et al., 2013).
184
This could potentially explain how cell cycle check points can still be overruled in the
absence of mutations in TP53 or RB1 in SBNET patients.
Different genetic subtypes of SBNET are beginning to emerge. As these subtypes and
their clinical characteristics become better understood it will enable more targetted ther-
apies and novel biomarkers to be developed that can be used in SBNET patients. Indeed
72 % of the mutated genes identified by Banck et al can be therapeutically targetted
(Banck et al., 2013; Miller et al., 2015b). Novel biomarkers are needed that can success-
fully stratify patients based on clinical behaviour and provide a more tailored treatment
approach by identifying which therapies would be most effective in which SBNET pa-
tients.
Circulating DNA, mRNA, miRNA
Circulating nucleic acids including cell free DNA and RNA (mRNA/miRNA) in the
peripheral blood represent appealing candidates for future biomarkers for use in SBNET
patients due to the possibility of using a non-invasive liquid biopsy approach. This
allows the genetic changes occurring in a tumour to be sampled in a non-invasive manner
(Crowley et al., 2013; Diaz and Bardelli, 2014; Miller et al., 2015b). This approach
could be particularly useful in GEP-NET due to the considerable heterogeneity in these
tumours (see section 2.2.2), therefore circulating DNA/RNA is likely to represent a more
comprehensive picture of the tumour biology than a single biopsy taken at a single time
point (Murtaza et al., 2013; Miller et al., 2015b). The use of circulating nucleic acids as
biomarkers has the potential to make clinical desisions more effective by accounting for
tumour evolution (Miller et al., 2015b).
Nucleic acid biomarkers extracted from plasma/serum have the potential to be used
to provide clinically useful information in SBNET such as the early identification of
micrometastases or treatment response monitoring. This approach has been used in
prostate cancer with certain changes in circulating tumour DNA and mitochondrial DNA
185
being associated with poor prognosis (Sita-Lumsden et al., 2013b; Miller et al., 2015b).
There have been no studies of circulating DNA in SBNET in the academic literature to
date. There has been one study of circulating mRNA (transcripts) in GEP-NET patients
(Modlin et al., 2014). 51 circulating mRNA, originally identified in tissue microarrays,
were extracted from the peripheral blood of patients with GEP-NET (n=41) and were
able to successfully identify 38 out of 41 patients with a GEP-NET (Modlin et al., 2014).
The study used qPCR for transcript quantification and was able to distinguish peripheral
blood samples from GEP-NET patients from age and sex matched control samples in 95
% of cases (5 % of control samples gave a false positive result, controls included blood
from healthy persons as well as patients with cysts or gastroesophageal reflux disease but
no NET). The study showed that elevated levels of these 51 transcripts had a sensitivity
of 92.8 % and a specificity 92.8 %, outperforming the current single analyte biomarker
CgA (ELISA assay) in this study, which was found to have a sensitivity and specificity
of 76 % and 59 % respectively. Limitations remain for methods based on circulating
mRNA since they are inherently less stable than miRNA and DNA and therefore more
susceptible to degradation.
MiRNA regulate the expression of mRNA by gene silencing. Cell free miRNA have
been isolated from many different bodily fluids including serum, plasma, urine, faeces,
saliva and tears (Sita-Lumsden et al., 2013b; Sita-Lumsden et al., 2013a; Mall et al.,
2013; Miller et al., 2015b). MiRNA have a much more stable structure than mRNA
so they are promising candidates for a non-invasive liquid biopsy approach (see section
2.5.2). MiRNA are frequently dysregulated in cancer but there have been very few stud-
ies of miRNA expression in SBNET patients. The potential for miRNA to be used as
future novel biomarkers for the stratification of patients with SBNET is more extensively
discussed in the miRNA section, section 2.5.
186
Circulating tumour cells
Recent technological advances have enabled the reproducible and robust detection of
circulating tumour cells (CTC) in peripheral blood (Krebs et al., 2010; Krebs et al., 2014).
There are a number of different approaches to CTC quantification including the validated
CellSearch platform which is based on epithelial adhesion, this has been approved by the
FDA in the USA for use in prognosis prediction in colorectal and prostate cancer (Khan
et al., 2011; Krebs et al., 2014; Bono et al., 2008; Cohen et al., 2008b; Miller et al.,
2015b). Challenges still remain with this approach due to the very small numbers of
CTC present in peripheral blood compared to leukocytes (Miller et al., 2015b).
The use of CTC, as with the use of circulating cell free DNA/RNA, enables the
non-invasive monitoring of tumour biology and thus represents an advantage over sin-
gle time point tumour biopsies by providing information about tumour evolution and
intra-tumoural heterogeneity (Krebs et al., 2014).
There are ongoing studies of CTC cell in SBNET patients including a phase IV clinical
trial (NCT02075606) in functioning SBNET investigating CTC cell counts for monitoring
Somatuline Autogel treatment response as an alternative to CT scans and for predicting
PFS (Miller et al., 2015b). A prospective study of 176 patients with metastatic NET
previously showed that the presence of ≥ 1 CTC in NET patients was associated with
worse PFS and overall survival (Khan et al., 2013b). Another study by the same group
also found that the absence of CTC was associated with stable disease in SBNET (n=26)
and PNET (n=19) patients while their presence was associated with tumour progression
(Khan et al., 2011). These findings suggest that CTC could be useful as prognostic
biomarkers in SBNET and PNET patients.
DNA methylation and histone modification
DNA methylation and histone modifications along with non-coding RNA such as miRNA
(see section 2.5) represent epigenetic mechanisms for the regulation of gene expression.
187
Epigenetic changes are stable molecular changes that are heritable during somatic cell
divisions but do not change the sequence of the DNA, they convey genomic adaptations
to the environment (Jovanovic et al., 2010).
Methylation of gene promoter regions and post-translational histone modifications af-
fect the accessibility of genes to transcription factors, changes in these epigenetic processes
frequently occur during tumourigenesis (Miller et al., 2015b). Epigenetic changes have
the potential to act as early biomarkers since they are thought to precede genetic events
in tumourigenesis such as the mutation of tumour suppressor genes, the activation of
oncogenes and genomic instability (Karpathakis et al., 2013; Miller et al., 2015b). In low
grade GEP-NET, which usually lack mutations in commonly mutated tumour suppres-
sors TP53 and RB1, it has been suggested that epigenetic changes could be the main
drivers of disease pathology (Karpathakis et al., 2013; Miller et al., 2015b).
DNA methylation of GSTP1 and FOXC1 was found to be predictive of survival in
breast cancer patients and the absence of ABCB1 methylation was found to be associated
with disease progression during doxorubicin treatment (Dejeux et al., 2010; Miller et al.,
2015b).
There have been quite a few studies of DNA methylation that included SBNET pa-
tients, however there is little information in the literature about the histone modification
status of these patients. A study in SBNET patients (n=44) identified changes in the
methylation status of various genes including the methylation of WIF1, NKX2-3 and
CXCL14 promoter regions which led to reduced expression of these genes in SBNET
compared to controls and even further reduced expression in SBNET metastases (Fo-
touhi et al., 2014; Miller et al., 2015b). This study and others identified RASSF1A and
CTNNB1 methylation as being associated with reduced overall survival in SBNET pa-
tients (Fotouhi et al., 2014; Liu et al., 2005; Zhang et al., 2006). Future clinical trials
could determine if methylation changes in the promoters of these genes could be used as
prognostic biomarkers in SBNET. These findings also suggest that treatment with DNA
188
demethylators may be of benefit to SBNET patients since if this treatment approach was
able to rescue the expression of these genes it could potentially improve patient survival.
There is one study in the literature of histone modification in patients with intestinal
NET (FFPE tissue), this showed that in the majority of patients, 13/14 patients, histone
H3K4 was dimethylated this was a rare occurrence however in the hepatocellular carci-
noma patients also included in the study 8/51 (Magerl et al., 2010; Miller et al., 2015b).
Magerl et al suggest that their findings may be helpful to aid differential diagnosis in
certain patients (Magerl et al., 2010). More studies are warranted to determine what
genes might become overexpressed in SBNET patients due to the dimethylation of H3K4
and to identify what effects this might have on tumour pathology.
In addition to the potential for the use of DNA methylation and histone modification
as tumour biomarkers, these molecular changes also have the potential to be targeted in
novel therapeutic approaches. Therapy with DNA demethylators and histone deacetylase
inhibitors can reset the global epigenetic status of patients with cancer (Miller et al.,
2015b). Treatment with the hypomethylating agent decitabine for example, was shown
to increase PFS in ovarian cancer by causing desensitisation to carboplatin (Matei et al.,
2012; Miller et al., 2015b). There have been no studies to date on the use of such therapies
in GEP-NET.
2.7. Gaps in the literature
There is currently only one prognostic biomarker used in GEP-NET, the Ki-67 prolifera-
tive index. Ki-67 % has been widely adopted for tumour grading in GEP-NET patients,
with the classification of tumours based on proliferation levels as either G1 or G2 (well
differentiated, low grade tumours) or G3 (poorly differentiated, high grade tumours).
Low grade GEP-NET have differing biological and clinical behaviour to the far rarer
high grade tumours, virtually all of which have metastases to distant sites at presentation.
189
Despite this, a substantial number of patients with low grade GEP-NET exhibit a more
aggressive metastatic phenotype, and liver metastases remain a common occurrence even
in patients with very low proliferation levels and small tumours. This is particularly
true of patients with low grade SBNET and NF-PNET. These low grade tumours have
a heterogeneous disease pathology and clinical course which presents a challenge for the
management of GEP-NET patients.
Limitations remain with the use of Ki-67 % for prognostic prediction in patients with
low grade tumours. A particular challenge is the inability of Ki-67 % to identify which
low grade GEP-NET patients have a more aggressive metastatic tumour subtype (section
2.6.1). Another limitation is the heterogeneity of Ki-67 %, both within a single tumour
and between different tumours in the same patient, and the lack of consensus in the
literature about suitable cut off levels for Ki-67 %.
The need for the development of novel biomarkers for use in GEP-NET patients, partic-
ularly those with low grade tumours, has been outlined in several consensus conferences
(Frilling et al., 2014; Oberg et al., 2015). In particular biomarkers are needed which can
further stratify patients with low grade tumours into clinically useful subgroups. These
biomarkers could be used for the prediction/early detection of disease progression/recur-
rence and to identify patients who would most benefit from more aggressive treatment
approaches, with the potential to improve patient outcomes.
There are a number of areas that have been less well explored in the existing academic
literature. There have been few large studies investigating what proportion of low grade
patients have liver metastases, since many studies focus on malignant GEP-NET, with
grade and stage being considered separately (section 2.6.1). Also lacking is detailed
information on the extent of Ki-67 % intertumoural and intratumoural heterogeneity in
patients with GEP-NET.
Despite low grade GEP-NET being far more common than high grade tumours, there
remains a lack of information in the academic literature about the biological processes
190
which are disrupted in these tumours (sections 2.5.2 and 2.6.2). Information is also lacking
on what might cause a subset of patients with low grade GEP-NET to have tumours with
more aggressive pathological and clinical features (section 2.6.1).
Novel cancer biomarkers have been explored in a number of different ways in GEP-
NET, for example there have been studies investigating genetic changes and the use of
CTC and circulating mRNA as biomarkers. However, studies of epigenetic changes in
GEP-NET patients such as changes in miRNA expression, DNA methylation and histone
modifications are still in their infancy (section 2.6.2). Epigenetic studies may be of
particular importance for understanding the disease pathology of patients with low grade
GEP-NET, since it has been suggested that epigenetic changes may be key drivers of
disease pathology in these patients (Karpathakis et al., 2013). This is because patients
with low grade GEP-NET usually lack large scale chromosomal deletions/insertions and
mutations in key tumour suppressors that are frequently mutated in cancer, such as TP53
and RB1.
MiRNA are frequently dysregulated in cancer, and their stability both in FFPE tissue
and in bodily fluids, such as peripheral blood and urine, makes them well suited for
use as tumour biomarkers (section 2.5.2). MiRNA expression has been investigated in
PNET patients and in a murine PNET model, with further functional studies of miRNA
in PNET cell lines. Studies are, however, particularly lacking in the academic literature
about the role of miRNA in tumourigenesis and disease progression in patients with
SBNET, with only a few miRNA expression studies to date in small patient cohorts
(section 2.5.3). What is absent from the academic literature is a comprehensive miRNA
expression profiling of tissue from SBNET and their metastases in a large patient cohort,
validated with samples from an independent group of SBNET patients.
Gaps in the academic literature remain, particularly around the potential role of epige-
netics in SBNET with few existing studies on miRNA expression changes, DNA methy-
lation changes and histone modifications that could be contributing to tumourigenesis
191
and the development of liver metastases in patients with SBNET. The lack of knowledge
around miRNA expression in SBNET and their metastases and their potential use as
novel biomarkers forms the basis for the aim of this thesis.
192
3. Methods
3.1. Ethics Approval
The studies included in this thesis are a part of the project R12025: Genetic signature,
metabolic phenotyping and integrative biology of neuroendocrine tumours. This project
was given ethical approval by the ICHTB Tissue Management Committee, REC number:
07/MRE09/54.
FFPE tissue samples were provided by the Imperial College Healthcare NHS Trust
Tissue Bank (London, UK). Other investigators may have received samples from these
same tissues. The research was supported by the National Institute for Health Research
(NIHR) Biomedical Research Centre based at Imperial College Healthcare NHS Trust
and Imperial College London. The views expressed are those of the author and not
necessarily those of the NHS, the NIHR or the Department of Health.
41 frozen tissue samples from SBNET tumours and their metastases for the miRNA
study (dataset 2) came from Zentralklinik Bad Berka (Bad Berka, Germany) and were
provided by Dr Daniel Kaemmerer.
Normal small bowel samples were obtained from 2 patients undergoing a right hemi-
colectomy procedure at Imperial College Healthcare NHS Trust and were provided by
Mr Paul Ziprin. Patient consent was given for the small bowel tissue that would be re-
moved anyway during the course of a normal right hemicolectomy procedure to be used
for research. These samples were used as a “normal” small bowel comparison group in
193
the miRNA study (dataset 2), for more details on these samples please see section 3.3.1.
Unstained slides from 6 GEP-NET patients with liver metastases who were candidates
for liver transplantation from the Institute of Pathology at Essen University Hospital
(Essen, Germany) were provided by Professor Kurt Schmid for the Ki-67 % heterogeneity
study.
This research was supported by the Dr Heinz-Horst Deichmann Foundation.
3.2. Ki-67 %
3.2.1. Patient details
There were 161 GEP-NET patients included in this retrospective study. They were from
a prospectively maintained database of GEP-NET patients seen at Imperial Heathcare
NHS Trust. Clinical data was obtained from the database. Appendix NET patients were
excluded, since low grade appendix NET are usually benign (Griniatsos and Michail,
2010; Pape et al., 2016; Pawa et al., 2018). GEP-NET diagnosis was confirmed by H&E
and IHC from a surgical specimen and/or biopsy.
This study was published in the World Journal of Surgery in 2014 (Miller et al., 2014).
3.2.2. Grade and stage
Clinical tumour stage (TNM) and grade was determined according to the ENETS guide-
lines for GEP-NET, see Literature review, section 2.2.3, Table 2.1 (Rindi et al., 2006;
Rindi et al., 2007). Grading was done in either resected tissue or a biopsy from the
primary site or a metastasis. For IHC staining methods see section 3.3.5.
Patients with a G1/G2 GEP-NET had either 68Ga-DOTATATE PET/CT or Oc-
treoscan while patients while patients with a G3 GEP-NET had 18F-fludeoxyglucose
PET/CT. This information was used to determine clinical tumour stage and further data
was obtained from CT, MRI, EUS and endoscopy (Miller et al., 2014). For the study
194
analysis, we adapted the TNM so that any T was permitted, this enabled tumour stage
to be compared across different primary sites.
3.2.3. Data analysis
Data analysis was done using Microsoft Excel 2010. Patients were stratified based by
tumour grade to enable comparisons to be made based on different clinical characteristics.
The Fisher’s exact test (two tailed) was used to test the statistical significance of group
comparisons, p values of < 0.05 were considered to be statistically significant.
3.2.4. Heterogeneity
The Ki-67 % heterogeneity part of the study included 30 GEP-NET patients in total.
Patients were selected based on the availability of tumour tissue from more than 1 tu-
mour loci or site of disease manifestation, this tissue was removed as part of the routine
treatment of these patients for their GEP-NET.
17 such patients from Imperial College Heathcare NHS Trust were included in our
World Journal of Surgery publication (patients 1-17, see chapter 4.6, Table 4.11) (Miller
et al., 2014). After the publication of the paper a further 7 such patients were identified
at Imperial College Heathcare NHS Trust and added to the study (patients 18-24, Table
4.11).
Unstained slides from 6 GEP-NET patients with liver metastases who were candidates
for liver transplantation were kindly provided by Professor Kurt Schmid from the Institute
of Pathology at Essen University Hospital (Essen, Germany) (patients 25-30, Table 4.11).
Ki-67 % was assessed in the primary site and in all lesions for which FFPE tissue was
available according to ENETS guidelines (see section 2.2.3, Table 2.1). 2000 cells were
assessed and the percentage of cells staining positive for Ki-67 (brown nuclear staining)
was determined for tumour grading (Rindi et al., 2006; Rindi et al., 2007). For antibody
and IHC staining protocols see section 3.3.5.
195
For the 6 explanted liver patients the primary tumour was graded according to ENETS
guidelines at the Institute of Pathology at Essen University Hospital (Essen, Germany).
To determine the extent of intratumoural heterogeneity in the liver metastases from these
6 patients, the Ki-67 % was determined in 5 different sites within each liver lesion.
3.3. miRNA
3.3.1. Patient Details
This study was published in Endocrine Related Cancer in September 2016 (Miller et al.,
2016). Clinical details from patients S1-S13 and S15 of dataset 1 were also included in
the earlier World Journal of Surgery publication on 161 GEP-NET patients (Miller et al.,
2014) (see chapter 5, Table 3.2).
Table 3.1.: Number of samples for miRNA quantification
dataset 1 dataset 2Primary SBNET 15 13Lymph node metastases 9 15Liver metastases 2 13Lymph node normal 7 0Adjacent normal small bowel 12 0Adjacent normal liver 2 0Small bowel “normal” tissue 0 2Total number 47 43
The study design is shown in Figure 3.1. The study involved the global quantification of
miRNA in SBNET and their metastases. The study included 90 tissue samples in total,
Table 3.1. It included 2 independent sets of SBNET patients treated at two separate
institutions, dataset 1 from Imperial College Healthcare NHS Trust (London, UK) and
dataset 2 from Zentralklinik Bad Berka (Bad Berka, Germany).
The most dysregulated miRNA were validated by a second quantification method
(qPCR). The types of samples included in dataset 1 and dataset 2 are shown in Ta-
196
ble 3.1.
Dataset 1
This part of the study included 15 patients with low grade (G1/G2) SBNET from Imperial
College Healthcare NHS Trust. FFPE tissue was used for miRNA expression quantifi-
cation in all sites for which tissue was available, resulting in a total of 47 samples being
included (dataset 1).
Tumour tissue was available from the primary SBNET for all patients (n=15), matched
lymph node metastases (n=9) and matched liver metastases (n=2). Matching adjacent
normal tissue was also available from adjacent normal small bowel (n=12), normal lymph
nodes (n=7) and adjacent normal liver (n=2).
All the patients had a low grade SBNET with the majority, (12 patients), being classed
as G1 and 3 patients being G2, see Table 3.2. Most of the patients, 87 %, had locoregional
metastasis (13 patients, 13/15) and 60 % had distant metastases (9 patients, 9/15). Of
the 9 patients with distant spread of their disease 8 patients had liver metastases and 1
patient had peritoneal metastases. There was only one patient with no metastatic spread
of the SBNET, patient number S11 (tumour stage: T3N0M0). Most of the patients had
a non-functioning SBNET, with just 4 patients having carcinoid syndrome. None of the
patients had MEN1 mutations. For tissue sample numbers and further clinical details
see Appendix, section A, Tables A.1 and A.2.
197
Figure 3.1.: Study design, global miRNA expression
198
Table 3.2.: Samples dataset 1
Patient Gender Age Ki-67 Grade Tumour Available FFPE tissue?no. (%) stage SBNET SB adj nor LN met LN nor Liver met Liver adj norS1 F 76 3 G2 T3N1M1 yes yes yes yes - -S2 M 81 1 G1 T4N1M1 yes yes yes yes yes yesS3 F 75 1-2 G1 T3N1M1 yes yes - - - -S4 M 38 < 2 G1 T2N1M0 yes yes yes yes - -S5 F 59 < 1 G1 T2N1M1 yes yes yes - - -S6 F 69 < 1 G1 T3N1M0 yes - - - - -S7 F 57 < 0.5 G1 T4N1M1 yes yes yes yes - -S8 F 84 < 1 G1 T4N1M0 yes yes yes yes - -S9 M 83 2 G1 T3N1M1 yes yes yes yes yes yesS10 M 69 < 2 G1 T3N1M0 yes yes yes - - -S11 M 75 < 2 G1 T3N0M0 yes - - - - -S12 M 59 < 2 G1 T4N1M0 yes - - - - -S13 M 77 4-5 G2 T4N1M1 yes yes yes yes - -S14 M 60 2-3 G2 T3N1M1 yes yes - - - -S15 F 61 1 G1 T1N0M1 yes yes - - - -
SB: small bowel, LN: lymph node, adj nor: adjacent normal, met: metastasis
199
Dataset 2
This part of the study included 43 frozen tissue samples in total, see Table 3.3. The
tumour tissue consisted of 41 frozen tumour tissue samples from Zentralklinik Bad Berka.
The samples came from 22 SBNET patients and included primary SBNET (n=13), lymph
node metastasis (n=15) and liver metastases (n=13). The samples were stored at -80◦C.
2 “normal” small bowel samples from Imperial College Healthcare NHS Trust were
used as a comparison group in the second miRNA gene expression profiling experiment
(dataset 2). This was unaffected small bowel tissue that was removed during a normal
right hemicolectomy procedure (see section 3.1). This tissue was collected from the
operating room and snap frozen in liquid nitrogen before being stored at -80◦C.
A limitation of this approach is that although the “normal” small bowel samples were
not expected to have any small bowel disease pathology, morphological changes or disease
pathology could still potentially exist in these samples. In order to control for this,
histology of the “normal” small bowel samples was checked for normal morphology, with
H&E staining being done on tissue sections sliced immediately before and immediately
after the tissue sections that were used for RNA extraction (for H&E staining see section
3.3.6).
3.3.2. RNA extraction
FFPE tissue (dataset 1)
The FFPE tissue was cut in 10, 10 µm thick sections per block. A rotary microtome
(Olympus, CUT 4060) was used to cut tissue onto membrane slides (MembraneSlide NF
1.0 polyethylene naphthalate, 415190-9081-000, Carl Zeiss Ltd., Cambridge, UK) to allow
the areas of interest to be easily dissected. The slides were air dried for 2 days at room
temperature. Deparaffinisation was done, with the sides being incubated in a xylene bath
for 5 minutes, a second xylene bath for 5 minutes, followed by an incubation in 100 %
200
ethanol (2 minutes), and then in a second 100 % ethanol bath (2 minutes). Slides were
rinsed in running water for 2 minutes followed by staining with coles haematoxylin (2
minutes). The slides were rinsed in running water (2 minutes), placed in acid alcohol and
gently agitated up and down a few times, rinsed in running water and then incubated
for a few seconds in scots tap water with gentle agitation. Slides were rinsed in running
water before being dried in an oven at 40◦C (20 minutes).
Tumour tissue was marked up on corresponding H&E slides by an experienced histopathol-
ogist. H&E slides corresponding to the adjacent normal tissue were marked up in the
same way. A scalpel was used to cut out the tissue areas of interest from each membrane
slide and the rolls of tissue were placed in a microcentrifuge tube (KC124, Appleton
Woods Ltd., Birmingham, UK). Approximately 5 x 5 mm2 areas of tissue were cut out
for RNA extraction per sample.
Total RNA was extracted from the samples using the miRNeasy FFPE kit (217504,
QIAGEN, Manchester, UK). 150 µL of buffer PKD and 10 µL of proteinase K was added
to the microcentrifuge tubes with the tissue which was then placed overnight in a 56◦C
waterbath to enable cell lysis and to digest cellular proteins. Samples were vortexed then
heated at 80◦C in a heating block for 15 minutes followed by a 3 minute incubation on
ice. 16 µL DNase Booster Buffer and 10 µL of DNase 1 stock solution were added and
the tubes incubated for 15 minutes at room temperature.
The lysate was transferred to a new tube and 320 µL of binding buffer RBC was added
and mixed by pipetting up and down. 720 µL 100 % ethanol was added to the sample
and mixed by pipetting up and down. 700 µL of each sample was transferred to a spin
column and centrifuged at for 15 seconds at 8000 x g. The remainder of each sample was
then added to the spin columns and the spin step repeated. 500 µL of Buffer RPE was
added to wash the column and the spin step was repeated. 500 µL of Buffer RPE was
added and centrifuged for 2 minutes at 8000 x g. The spin columns were centrifuged for
5 minutes at full speed (13,000 x g) then placed in new tubes. 14 µL of RNase free water
201
(10977035, Life Technologies Ltd.) was added to each spin column for RNA elution and
centrifuged for 1 minute at full speed. RNA quality and quantity was checked using a
NanoDrop 1000 spectrophotometer (Thermo Scientific, Loughborough, UK). The RNA
samples were stored at -80◦C.
RNA extractions were done twice from the original FFPE blocks. First RNA was
extracted from the tissue for the global miRNA quantification and then a separate RNA
extraction was done from the same FFPE blocks for the later miRNA hit validation
by qPCR (for RNA concentrations see Appendix, section C, Tables C.1 and C.2). The
normal lymph node sample (2.6), from patient number S4 (see chapter 5, Table A.1) was
excluded from the qPCR validation experiment due to failing RNA quantity and quality
checks for the second RNA extraction.
Frozen tissue (dataset 2)
Tissue was removed from the -80◦C freezer and transported to the cryostat (OTF model,
containing a 5040 microtome, Bright Instruments Ltd., Luton, UK) on dry ice. The
tissue was allowed to equilibrate to the cryostat temperature (-20◦C) for at least 20
minutes prior to sectioning. The tissue was placed in Cryo-M-Bed (Bright Instruments
Ltd.) on the metal cryostat mount and allowed to cool until the embedding material had
hardened. Excess embedding material was trimmed off. Tissue was sectioned at 20 µm
and the sections put into a fresh microcentrifuge tube. The cryostat was cleaned with 70
% ethanol before each individual tissue sample was cut and a fresh microtome blade (type
S 35, 720-1998, VWR International Ltd., Lutterworth, UK) was used for each sample.
The sectioned tissue was put on dry ice until TRIzol™ reagent (15596-018, Life Tech-
nologies Ltd., Paisley, UK) could be added. 500 µL of trizol was added to each sample
for cell lysis and the samples were vortexed for 20 seconds. A further 500 µL of trizol
was added and the lysate was pipetted up and down and vortexed again to homogenise
the samples. The samples were then either frozen at -80◦C for RNA extraction the next
202
day or the RNA extraction was performed immediately.
Samples that had been frozen (-80◦C) were removed from the freezer and kept on
the bench until they reached room temperature. All samples were left for 5 minutes at
room temperature to enable nucleoprotein dissociation. Then 200 µL of chloroform was
added and the samples were incubated at room temperature for 3 minutes before being
shaken for 15 seconds. Samples were centrifuged for 15 minutes at 4◦C at 12,000 x g to
separate the upper aqueous layer from the lower organic layer. The top clear aqueous
layer (containing RNA) was transferred into a fresh microcentrifuge tube and used for
RNA precipitation.
500 µL of isopropanol was added and the samples were incubated at room temperature
for 10 minutes. The samples were then centrifuged for 10 minutes at 4◦C at 12,000 x g.
The supernatant was pipetted out leaving only the pellet. 1000 µL of 75 % ethanol was
added to wash the RNA and the sample was briefly vortexed followed by centrifugation
for 5 minutes at 7500 x g at 4◦C. The supernatant was entirely pipetted out and the
pellet was left to dry for 5-10 minutes. The pellet was resuspended in 30 µL of RNase
free water (10977035, Life Technologies Ltd.). Very small pellets were resuspended in a
smaller volume (15-20 µL), to maintain a sufficient concentration of RNA. The samples
were incubated in a 55◦C waterbath for 10 minuets. The samples were then transferred
into fresh microcentrifuge tubes and put on ice. The sample concentration was checked
(NanoDrop, Thermo Scientific) and the samples were stored at -80◦C.
Spin columns from the miRNeasy kit (217504, QIAGEN) were used to further clean the
RNA and to remove any traces of contaminants such as phenol (trizol) which could inhibit
downstream reactions. The RNA was defrosted and vortexed to mix it thoroughly then
15 µL of RNA was added to 500 µL of RBC binding buffer and mixed by pipetting up
and down. The protocol was then followed according to the manufacturers instructions
(for protocol details see section 3.3.2, FFPE tissue RNA extraction, steps following the
addition of buffer RBC). The RNA was eluted from the spin columns with 30 µL of
203
RNase free water, unless the starting RNA concentration (prior to being added to the
spin column) was < 400 ng/µL in which case RNA was eluted with 14 µL of RNase
free water instead. RNA quality and quantity was checked using a NanoDrop 1000
spectrophotometer (Thermo Scientific). One sample was excluded due to there being little
or no RNA present (RNA concentration: 3.2 ng/µL) (for sample RNA concentrations see
Appendix, section C, Table C.3).
3.3.3. Global miRNA quantification
The extracted total RNA from the FFPE tissue (dataset 1) and the frozen tissue (dataset
2) was diluted to a concentration of 100 ng/µL. 5 µL of each diluted sample was sent to
NanoString Technologies (Seattle, USA) for global miRNA quantification. The miRNA
were quantified using the NanoString nCounter Human miRNA Expression Assay V2,
with 100 ng of input RNA according to the manufacturers instructions (Geiss et al.,
2008; Kulkarni, 2011).
The NanoString nCounter assay provides a direct count of the number of each indi-
vidual miRNA present in a sample based on miRBase version 18. This is a hybridisation
based method which enables multiplexed quantification of 800 known miRNA. The assay
does not involve reverse transcription (to produce cDNA) or PCR amplification, reducing
the chances of errors being introduced by avoiding these enzymatic steps.
The detection of each miRNA species is enabled through the binding of two sequence
specific probes, a reporter probe (fluorescent labelled) and a capture probe (biotin linked)
(Kulkarni, 2011). The probes and the miRNA for each sample are hybridised in solution
by complementary base pairing (Geiss et al., 2008). The reporter probe is labelled with
a unique fluorescent tag which represents a particular order of 4 differently coloured
fluorophores at 6 different positions on the reporter probe (Kulkarni, 2011). The capture
probe has a biotin tag which enables it to bind to the streptavidin coated surface of a
slide. After the capture probe has bound each miRNA/capture probe/reporter probe
204
complex to the slide, a voltage is applied to align the complex so that the fluorescent
label on the reporter probe is in the correct orientation on the slide. Each reporter probe
is detected by a microscope and camera attached to a computer which produces of a
count for the number of times a specific miRNA appears in a sample.
Data analysis
Data analysis was done independently for dataset 1 and dataset 2, using R/Biocon-
ductor (versions: R 3.1.1, Bioconductor 3.0). The R packages used were edgeR 3.8.6,
DESeq2 1.6.3 and limma 3.22.7. The edgeR package was used to filter out poorly ex-
pressed miRNA prior to data analysis (Robinson et al., 2009; McCarthy et al., 2012).
The DESeq2 package was used for normalisation and statistical analysis of the profiling
data (Love et al., 2014). The mean expression of each miRNA was calculated for each
sample group.
The fold change and log2 fold change values were calculated using DESeq2 so that the
magnitude of the differences in miRNA expression could be compared between sample
groups. To ensure that only miRNA with relatively large changes in expression between
sample groups were considered a log2 fold change cut off of ≥ 1.5 or ≤ −1.5 was used.
A log2 fold change of 1.5 is equivalent to a fold change of approximately 3, (21.5 = 2.828,
conversely: log22.828 = 1.5). This would exclude miRNA with only small changes in
expression between the comparison groups. For these miRNA even though the change in
expression is statistically significant between comparison groups the impact of the change
is likely to be negligible. This is because the magnitude of the change is so small that it
is unlikely to represent a true biological effect.
T tests were done and the False Discovery Rate (FDR), also known as the Ben-
jamini–Hochberg adjusted p value, was used to adjust for multiple-testing, with an FDR
of < 0.05 being considered statistically significant.
The data from both global miRNA profiling studies, dataset 1 and dataset 2, was
205
uploaded to the publicly available repository, Gene Expression Omnibus (GEO), provided
by the National Center for Biotechnology Information (NCBI), with the GEO accession:
GSE70534.
3.3.4. Validation of candidate miRNA by qPCR
Selection of miRNA candidates
To validate the results from the global miRNA profiling using the NanoString nCounter
miRNA Expression Assay, a second method, qPCR, was used to quantify the expression
levels of the most dysregulated miRNA (or “top hits”) from dataset 1 (those with the
highest magnitude of fold change in expression). This was 11 miRNA in total, comprising
of 7 miRNA in dysregulated in SBNET compared to adjacent normal small bowel tissue
and 4 miRNA dysregulated in lymph node metastases.
Reverse transcription
Specific Taqman® primers (4427975, Life Technologies) for each candidate miRNA and
endogenous control were used with the miRNA Reverse Transcription Kit (4366596, Life
Technologies) for the reverse transcription, according to the manufacturer’s guidelines.
For primer details see Appendix, section B, Table B.1. The total RNA (including miRNA)
was diluted to 2 ng/µL with RNase free waster and kept on ice. The master mix was
made up on ice. 1.5 µL of reverse transcription primers, 2.5 µL of diluted RNA and 3.5
µL of reverse transcription master mix was added to each well of the PCR plate (48-Well
Semi-Skirted Plates: 11771198, lids: 11751188, Fisher Scientific UK Ltd., Loughborough,
UK). Master mix consisted of 0.075 µL of dNTPs, 0.5 µL of reverse transcriptase, 0.75
µL of reverse transcription buffer, 0.095 µL of RNase inhibitor and 2.08 µL of nuclease
free water per well. The plate was sealed and spun down (2000 rpm, 30 seconds) before
being transferred to a 96-Well Thermal Cycler (Applied Biosystems by Life Technologies
Ltd.) for reverse transcription to produce cDNA. Thermal cycling conditions were: 16◦C
206
for 30 minutes, 42◦C for 30 minutes, 85◦C for 5 minutes then 4◦C (indefinitely).
qPCR
The qPCR assay was done using TaqMan® Universal PCR Master Mix and specific qPCR
probes for each miRNA and endogenous control (4324018 and 4427975, Life Technologies)
according to the manufacturer’s guidelines. For primer details see Appendix, section B,
Table B.1. All reagents and samples were kept on ice until it was time to add them to
the qPCR plate and tubes containing primers were also wrapped in foil to protect the
fluorophore from ambient light. For the qPCR, 17.5 µL of nuclease free water was added
to each well of the reverse transcription plate to dilute the cDNA. 4.43 µL of the diluted
sample was then added to each well of the qPCR plate (4346907, Life Technologies Ltd.).
1 µL of specific miRNA primers, 10 µL of PCR master mix and 4.57 µL of nuclease free
water were mixed together and added to each well. The plate was sealed with Optical
Adhesive Film (4360954, Life Technologies Ltd.) and spun down (2000 rpm, 30 seconds).
Samples were run in duplicate on the qPCR plates using a StepOnePlus™ Real-Time PCR
System (Applied Biosystems by Life Technologies Ltd.). Thermal cycling conditions were
1 cycle of 95◦C for 10 minutes (polymerase activation), followed by 40 cycles of 95◦C for
15 seconds and 60◦C for 1 minute (denaturation and annealing/extension).
Data analysis
Two different endogenous control genes were used for qPCR normalisation to ensure
that any differences in candidate miRNA expression were not affected by the choice of
endogenous control. The endogenous control genes were U6 small nuclear 1 (RNU6-1)
(previously known as U6) and small nucleolar RNA, C/D box 44 (SNORD44) (previously
known as RNU44). Both RNU6-1 and SNORD44 were verified as having stable expression
across the samples.
Data analysis was done using Microsoft Excel 2010. The mean threshold cycle (Ct)
207
was determined for each sample (samples run in duplicate). This was used to calculate
the relative miRNA expression, delta Ct. Normalisation was done using RNU6-1 and
SNORD44 expression in each sample. Similar results were obtained for both of these
endogenous controls (see Chapter 5, section 5.3).
The mean delta Ct was calculated for each of the sample groups to enable the expression
of each miRNA to be compared between different tissue types (eg: tumour tissue versus
“normal” tissue). An unpaired, one-tailed t test was done and a p value of < 0.05 was
considered to be satistically significant.
3.3.5. IHC Ki-67
FFPE blocks were sectioned at 2.5 µm onto microscope slides (631-0107, VWR Interna-
tional Ltd.) using a rotary microtome (Olympus, CUT 4060). Antigen retrieval and IHC
were done using a Leica BOND-MAX™ automated IHC machine (Leica Biosystems, New-
castle upon Tyne, UK) according to the manufacturer’s recommendations. The epitope
retrieval step was citrate for 30 minutes. The bond polymer refine detection kit (DS9800,
Leica Biosystems) was used for the IHC. For details of the primary, secondary and horse
radish peroxidase (HRP) conjugated antibodies used please see Table 3.4. Primary anti-
bodies were diluted with Bond primary antibody diluent (AR9352, Leica Biosystems).
Slides were removed from the IHC machine and rinsed in running tap water for 2
minutes, then incubated in 0.5 % CuSO4 in dH2O for two minutes. The slides were rinsed
again in running tap water for 2 minutes. Slides were dehydrated by immersing them in
baths of 70 % ethanol (30 seconds), 95 % ethanol (3 minutes), 100 % ethanol (3 minutes)
and 100 % ethanol (10 seconds), followed by washing in 3 xylene baths with a 30 second
incubation in each. The slides were then coverslipped using pertex mounting medium
(3808706E, Leica Biosystems) and a Leica Coverslipper (CV5030, Leica Biosystems). All
steps were carried out at room temperature. Normal human tonsil tissue was used as a
positive control, PBS was added rather than the primary antibody as a negative control.
208
Table 3.3.: Samples dataset 2
Patient no. Gender Age Grade Tissue available? Sample IDB9 M 60 G1 Liver MTS 8.0B53 M 63 G1 SBNET 8.1B57 F 86 G1 Liver MTS 10.1B60 M 63 G1 Lymph node MTS 8.3B65 M 64 G1 SBNET 10.2B66 M 88 G1 SBNET 10.4
Lymph node MTS 10.6Liver MTS 10.7Liver MTS 10.8
B74 F 75 G1 SBNET 10.9Lymph node MTS 11.0
B76 F 79 G1 Lymph node MTS 8.6Lymph node MTS 8.7
B77 M 72 G2 SBNET 8.8Liver MTS 8.9Lymph node MTS 11.2
B84 M 74 G3 Lymph node MTS 9.0Liver MTS 9.1Liver MTS 7.7
B86 F 75 G1 SBNET 11.3Lymph node MTS 9.2
B89 M 62 G2 SBNET 9.3Lymph node MTS 9.4
B103 F 62 G1 SBNET 11.4Lymph node MTS 11.5
B117 M 61 G1 SBNET 11.6Lymph node MTS 11.7
B118 M 70 G2 Liver MTS 9.5B119 M 64 G1 SBNET 11.8B121 M 75 G2 Liver MTS 9.6B121.1 M 59 G1 SBNET 12.0
Lymph node MTS 12.1B124 F 74 G1 SBNET 12.2
Lymph node MTS 12.3Liver MTS 12.4
B125 M 66 G1 SBNET 12.5Liver MTS 12.6
B133 M 54 G1 Liver MTS 9.7Liver MTS 9.8Lymph node MTS 9.9
B140 M 58 G1 Lymph node MTS 7.8C1 F 69 N/A “normal” small bowel* 0.1C2 F 63 N/A “normal” small bowel* 0.2
*“normal” small bowel tissue was from Imperial College Healthcare NHS Trust
209
Table 3.4.: Antibodies used for Ki-67 IHC
Antibody Dilution Timeapplied(mins)
Details Epitoperetrieval,time(mins)
Positivecontrol
Provider Cataloguenumber
Ki-67 1/100 30 Mouse anti-humanKi-67
Citrate, 30 tonsil LeicaBiosys-tems
NCL-L-Ki-67-MM1
Secondary n/a 15 Rabbit anti-mouseIgG (in 10 % animalserum)
n/a n/a LeicaBiosys-tems
DS9800
HRPconjugated
n/a 15 Anti-rabbitpoly-HRP-IgG (in 10% animal serum)
n/a n/a LeicaBiosys-tems
DS9800
210
The Ki-67 index was determined in the primary tumour according to ENETS guidelines
(see Literature review, section 2.2.3, Table 2.1), in the area of the primary tumour with
the highest nuclear staining and expressed as the percentage of positive (brown) tumour
nuclei out of 2000 tumour nuclei (Rindi et al., 2006; Rindi et al., 2007).
3.3.6. H&E
Dataset 1, FFPE tissue
A rotary microtome (Olympus, CUT 4060) was used to cut 2.5 µm sections of the FFPE
tissue for H&E staining for each FFPE block. The slides used were Surgipath® pre-
cleaned microscope slides (3808122GCE, Leica Biosystems, Newcastle upon Tyne, UK).
Deparaffinisation was done, with the sides incubated in a xylene bath for 5 minutes, a
second xylene bath for 5 minutes, followed by an incubation in 100 % ethanol (2 minutes),
and then in a second 100 % ethanol bath (2 minutes). Slides were rinsed in running tap
water for 2 minutes followed by staining with haematoxylin (2 minutes). The slides were
rinsed in running tap water (2 minutes), placed in 1 % acid alcohol and gently agitated up
and down a few times, rinsed in running tap water and then incubated for a few seconds
in scots tap water with gentle agitation. Slides were rinsed in running tap water and
then stained in 1 % eosin for 5 minutes before being rinsed in running tap water again.
Slides were dehydrated by immersing them in baths of 70 % ethanol (30 seconds), 95%
ethanol (3 minutes), 100 % ethanol (3 minutes) and 100 % ethanol (10 seconds), followed
by washing in 3 xylene baths with a 30 second incubation in each. The slides were then
coverslipped using pertex mounting medium (3808706E, Leica Biosystems) and a Leica
Coverslipper (CV5030, Leica Biosystems).
Dataset 2, Frozen tissue
A cryostat (OTF model, containing a 5040 microtome, Bright Instruments Ltd.) was
used to cut 10 µm sections of the frozen tissue for H&E staining. The sections were cut
211
at the start and end of the portion of tissue that was used for RNA extraction in the
miRNA study. The slides were left at room temperature for > 1 hour after they had
been cut to allow the sections to dry. Slides were placed in 70 % ethanol for 1 minute
followed by distilled water for 1 minute. The slides were incubated in haematoxylin for 3
minutes. The slides were rinsed in running tap water until the water was clear. The slides
were placed in 1 % alcohol to destain them for a few seconds. The slides were incubated
in aqueous 1 % eosin for 3 minutes. The slides were rinsed in running tap water until
the water was clear. The slides were drained to remove excess water. The slides were
then dehydrated; 1 minute incubation in 70 % ethanol, 1 minute incubation in 100 %
ethanol, 30 seconds in a second bath of 100 % ethanol. The slides were then transferred
into Histo-Clear™ (AGR1345, Agar Scientific, Stansted, UK) for two minutes followed by
a fresh Histo-Clear™ bath for a further 2 minutes. The slides were mounted using DPX
resin (44581, Sigma-Aldrich Company Ltd., Dorset, UK) and coverslips. The slides were
allowed to dry for 2 hours.
3.4. Bioinformatics
Bioinformatics was done to predict the gene targets of the candidate miRNA that might
have an important role in SBNET. A bioinformatics approach was used to predict gene
targets of the candidate miRNA (miR-7-5p, miR-204-5p, miR-375, miR-1 and miR-143-
3p) identified in the miRNA profiling experiments. Publicly available gene expression
datasets were identified to identify genes that were dysregulated in SBNET tissue. The
genes that were dysregulated in SBNET were compared to the predicted gene targets
of each individual candidate miRNA. This was done to determine if the expression of
these genes could be being regulated by processes such as gene silencing by the candidate
miRNA in SBNET. Further bioinformatics approaches were done to identify the key
biochemical pathways that might be implicated in SBNET. This was also to identify
212
Figure 3.2.: Study design, bioinformatics
the the most promising miRNA-mRNA interactions for future study including future in
vitro experiments to confirm particular miRNA-mRNA interactions. The study design is
shown in Figure 3.2.
3.4.1. Predicted gene targets of candidate miRNA
The candidate miRNA investigated in the bioinformatics study were miR-7-5p, miR-
204-5p, miR-375, miR-1 and miR-143-3p. MiR-7-5p, miR-204-5p and miR-375 were
upregulated in SBNET relative to “normal” small bowel tissue, while miR-1 and miR-
143-3p were downregulated in lymph node metastases relative to SBNET tissue.
These miRNA had been validated by two different miRNA quantification methods
(NanoString miRNA Expression Assay and qPCR, see section 3.3.4) and were found to be
213
dysregulated in both datasets 1 and 2. The 4 candidate miRNA that were downregulated
in SBNET versus “normal” small bowel were excluded since these miRNA were not found
to be dysregulated in dataset 2.
The biological gene targets of the candidate miRNA were predicted using TargetScan
(Human version 6.2) (Friedman et al., 2009; Lewis et al., 2005). TargetScan works by
using the conserved seed region on the miRNA and searching for corresponding conserved
sites on mRNA (transcripts) that would have complementary base pairing to the miRNA
seed region (Friedman et al., 2009). TargetScan looks for sites on mRNA that share 6
adjacent Watson-Crick base paring matches to the seed region on the miRNA (for more
information on seed regions see Literature review, section 2.5.1) (Friedman et al., 2009).
The output of the TargetScan analysis was a list of predicted gene targets for each
candidate miRNA.
3.4.2. Gene expression datasets
The gene expression datasets used for the bioinformatics analysis are shown in Table
3.5. Datasets a-d consisted of available data from expression studies previously done in
SBNET patients. The results of these microarray experiments were publicly available
either through either the NCBI GEO or the European Bioinformatics Institute (EBI) Ar-
rayExpress platforms under the accession codes GSE27162, GSE6272, E-TABM-389 and
GSE9576 (Edfeldt et al., 2011; Kidd et al., 2014; Leja et al., 2009). Further details in-
cluding the URLs for accessing the data from these gene expression profiling experiments
are shown in Table 3.5.
The gene expression data from dataset b and dataset c for SBNET relative to “normal”
small bowel tissue was available in supplementary tables (S1 and S2) of the paper by Kidd
et al. (2014). This showed the genes that were significantly dysregulated in SBNET (p
value: < 0.05).
This was not available for dataset a and dataset d, so for these datasets analysis was
214
Table 3.5.: Gene expression datasets for bioinformatics analysis
Geneexpression
Publicdatabase ID
Publications Method
dataset a GSE271621 Edfeldt et al.(2011)
gene expressionprofiling,microarray
dataset b GSE62722 Kidd et al. (2014) gene expressionprofiling,microarray
dataset c E-TABM-3893 Leja et al. (2009)and Kidd et al.(2014)
gene expressionprofiling,microarray
dataset d GSE95764 Leja et al. (2009) gene expressionprofiling,microarray
1 https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE271622 https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE62723 https://www.ebi.ac.uk/arrayexpress/experiments/E-TABM-389/4 https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE9576
carried out using the GEO2R platform provided by NCBI GEO for the analysis of gene
expression data contained in the repository (Sean and Meltzer, 2007). GEO2R was used
to carry out data analysis on gene expression levels of the tissue types of interest to
identify mRNA that were dysregulated.
For dataset d the gene expression in SBNET samples was compared to that in “normal”
small bowel tissue. For dataset a lymph node metastasis samples were included in the
data, therefore data analysis was done to compare the gene expression in lymph node
metastases to that in the primary tumour, see Table 3.6. The genes that were significantly
upregulated and genes that were significantly downregulated were identified, with p value
of < 0.05 being considered to be statistically significant.
Once the data had been analysed to determine dysregulated genes in the tissues of in-
terest, gene lists were prepared of genes that were upregulated/downregulated in SBNET
(versus “normal” small bowel) for datasets b-d and those that were upregulated/down-
215
Table 3.6.: Comparison groups for gene expression datasets
Gene Comparison Number of samplesexpression data groups SBP SB N LNMdataset a (Edfeldtet al., 2011)
LNM/SBP 18 17
dataset b (Kiddet al., 2014)
SBP/SB N 9 3
dataset c (Lejaet al., 2009; Kiddet al., 2014)
SBP/SB N 3 3
dataset d (Lejaet al., 2009)
SBP/SB N 3 3
LNM: lymph node metastasis, SBP: small bowel primary, SB N: small bowel “normal”
regulated in lymph node metastases (versus SBNET) for dataset a. This was to enable
the list of predicted gene targets from TargetScan for each candidate miRNA (see section
3.4.1) to be compared to genes that had been found experimentally to be dysregulated
in tissue from SBNET patients.
3.4.3. Data processing
The output of TargetScan was a gene list of all predicted gene targets of each candidate
miRNA. The output of the analysis of the publicly available gene expression datasets was
a gene list of genes that were upregulated or downregulated in the tissues of interest.
In order to identify experimentally dysregulated genes that might be being targeted by
the candidate miRNA (experimentally identified as dysregulated in the miRNA profiling
experiments) comparisons were done of the gene lists generated in sections 3.4.1 and
3.4.2.
Experimentally upregulated genes were compared to the lists of predicted gene targets
(TargetScan) of downregulated miRNA and the converse was done for the downregulated
genes which were compared to the predicted gene targets of upregulated miRNA, the
antiregulation paradigm (Frampton et al., 2014; Miller et al., 2016). For further details
216
of how miRNA regulate gene expression and how they can act as oncomir and tumor
suppressor miRNA during tumourigenesis see the Literature review, sections 2.5.1 and
2.5.2.
Genes that were downregulated in SBNET versus “normal” small bowel (datasets b-
d) were compared to the gene targets of the candidate miRNA that were upregulated
in SBNET versus “normal” small bowel (miR-7-5p, miR-204-5p, miR-375). Genes that
were upregulated in lymph node metastases versus SBNET (dataset a) were compared to
the gene targets of candidate miRNA that were downregulated in lymph node metastases
versus SBNET (miR-1 and miR-143-3p).
Since the gene lists came from different sources the gene symbols used to represent a
particular gene were not always consistent. To ensure that every instance of the same gene
was represented by the same gene symbol (to enable effective gene list comparisons) all
gene symbols were checked against the HUGO Gene Nomenclature Committee (HGNC)
database and replaced with the official gene symbol for that gene.
The gene lists of predicted gene targets of the candidate miRNA (identified using the
NanoString nCounter Human miRNA expression assay and qPCR) were compared to the
genes (mRNA) that had been found to be dysregulated in SBNET experimentally (in the
publicly available gene expression datasets). This was done to find the genes in common
between the gene lists. The particular comparisons are described below.
SBNET
There were 3 gene expression datasets for this comparison, datasets b-d. There were
3 candidate miRNA, miR-7-5p, miR-204-5p, miR-375. For each miRNA that was up-
regulated in SBNET (versus “normal” small bowel) the list of predicted gene targets
(TargetScan) was compared to the lists of genes (mRNA) that were downregulated in
the SBNET (versus “normal” small bowel) from each of the 3 gene expression datasets
(datasets b-d). This was to find genes that might be being regulated by the candidate
217
miRNA in SBNET.
The genes that were downregulated in 2 or more of the gene expression datasets
(datasets b-d) and that were also predicted gene targets for the candidate miRNA were
taken forwards to the gene ontology and pathway analysis (section 3.4.4).
Lymph node metastases
There was 1 gene expression dataset for this comparison, dataset a. There were 2 candi-
date miRNA, miR-1 and miR-143-3p. For each miRNA that was downregulated in lymph
node metastases (versus SBNET) the list of predicted gene targets (TargetScan) was com-
pared to the list of genes (mRNA) that were upregulated in the lymph node metastases
(versus SBNET), dataset a. This was to find genes that might be being regulated by the
candidate miRNA in the lymph node metastases.
The genes in common between the list of genes that were upregulated in lymph node
metastases and the list of genes that were predicted targets for each miRNA were taken
forwards to the gene ontology and pathway analysis (section 3.4.4).
3.4.4. Gene ontology and pathway analysis
Gene ontology and pathway analysis was done to identify using bioinformatics approaches
the biological mechanisms that might be affected by the dysregulation of particular can-
didate miRNA and their mRNA targets in SBNET. This was in order to identify which
of the predicted miRNA-mRNA interactions might be of particular interest for future ex-
perimental work to experimentally confirm the gene silencing in vitro and develop novel
prognostic biomarkers for use in SBNET.
The database for annotation, visualization and integrated discovery (DAVID) bioinfor-
matics tool was used for the functional annotation of dysregulated genes that could be
targets of the candidate miRNA (Huang et al., 2009b; Huang et al., 2009a).
First DAVID was used for gene set enrichment analysis, to identify gene ontology terms
218
that were over represented in the functional annotation of the gene targets of interest for
each miRNA (Subramanian et al., 2005; Ashburner et al., 2000).
Next pathway enrichment analysis was done using the Kyoto encyclopedia of genes
and genomes (KEGG) databases within DAVID to identify signalling pathways and in-
teractions that might be particularly important for SBNET tumourigenesis and disease
progression (Kanehisa and Goto, 2000; Kanehisa et al., 2010; Huang et al., 2009b; Huang
et al., 2009a).
A FDR correction for multiple testing was done to reduce false positive results (Ben-
jamini–Hochberg adjusted p value), with a FDR of < 0.05 being considered to be statis-
tically significant (Benjamini and Hochberg, 1995; Bleazard et al., 2015).
219
4. Role of the Ki-67 proliferation
index and disease stage in
GEP-NET
4.1. Introduction
Results presented in this chapter, chapter 4, were published in the World Journal of
Surgery in 2014 (Miller et al., 2014).
In this chapter the limitations in the use of Ki-67 % in GEP-NET are investigated
further (see Literature review, section 2.6.1). This follows on from chapter 3, in which
the methods are described including patient details, imaging and tumour grading/staging.
The findings of chapter 4 are built upon in the next results chapter, chapter 5. In chap-
ter 5 miRNA expression in low grade SBNET is investigated to identify novel biomarkers
with the potential to provide additional clinically useful information for patient stratifi-
cation over the use of Ki-67 % alone.
Ki-67 % was assessed in relation to the disease stage for 161 GEP-NET including
84 PNET and 37 SBNET. The location of any distant metastases, the presence of an-
gio/lymphovascular invasion and the presence of second primary malignancies were also
investigated. For a subset of patients, 30 patients, Ki-67 % was investigated at multiple
tumour loci to assess the extent of Ki-67 % heterogeneity.
220
This chapter addresses the first research objective of this thesis:
“1) Investigate the limitations of the existing prognostic biomarker in GEP-
NET.”
This objective was addressed in two principle ways. Firstly, the effectiveness of Ki-67
% in GEP-NET with respect to disease staging was investigated. Secondly, the extent of
Ki-67 % heterogeneity was investigated.
4.1.1. Summary of results
The findings were that there was no level of Ki-67 % at which a patient could be considered
to be safe from liver metastases. Metastases were a common occurrence even in G1
tumours with Ki-67 % of ≤ 2 %. An assessment of the extent of Ki-67 % heterogeneity in
30 patients led to the finding that both intertumoural and intratumoural heterogeneity
are a problem in GEP-NET leading to undergrading in some patients.
4.2. Patients
The study included 161 GEP-NET patients. The median age was 61 years (range: 21-91
years) and there were slightly more male patients (88, 55 %) than female patients (73,
45 %). The most common site for the primary tumour was the pancreas in just over half
of the patients (84, 52 %), see Table 4.1. Gastrointestinal tract primaries represented 43
% of the patients (69 patients). Of these tumours the small bowel was the most common
primary site with 37 patients having a SBNET (23 % of the GEP-NET cohort).
Other primary sites were less well represented in the cohort, with 12 patients with a
duodenal NET (8 %), 12 patients with a stomach NET (8 %), 5 patients with a rectal
NET (3 %), 2 patients with a colon NET (1 %), and 1 patient with an oesophageal NET
(1 %). There were 8 patients with an unknown primary (CUP) (5 %), with extensive
221
abdominal neuroendocrine metastases but where the site of the primary was investigated
but not found (Miller et al., 2014). Ki-67 % heterogeneity was investigated in 30 patients
(section 4.6).
Table 4.1.: Primary site of GEP-NET
Primary site Number of patientsPancreas 84Small bowel 37Duodenum 12Stomach 12Rectum 5Colon 2Oesophagus 1CUP 8
4.3. Grade and stage
Of the 161 GEP-NET patients included in the study there were 115 G1 tumours, 36 G2
tumours and 10 G3 tumours. For each grade the proportion of tumours with each disease
stage was considered to determine how common local and distant metastases were in
these tumours, these results appear in Table 4.2. The data was further analysed based
on the site of tumour metastases and on the primary site for patients with SBNET and
PNET. For the purposes of this study, any T was permitted for the tumour stage, this
was to enable tumour stage to be compared across different GEP-NET primary sites (see
Methods, section 3.2).
Characteristics including functionality, tumour invasiveness, genetic status and the
presence or absence of second, non-NET, primary malignancies were also investigated.
These results are shown by tumour grade in Table 4.2, with further details being presented
in later sections: 4.4 and 4.5).
222
4.3.1. Metastases
In the cohort as a whole (n=161), 42 % of the GEP-NET patients had distant metastases
(67 patients) while 15 % had locoregional disease only (24 patients). In 44 % of patients
there was no observed locoregional or distant spread of the disease (70 patients).
The proportion of patients with metastases was determined for each tumour grade
to determine how common an occurrence tumour metastases were at different ENETS
tumour grades, see Table 4.3 and Table 4.2.
A high proportion of the patients with G1 tumours, 46 %, had metastases, either to
local lymph nodes or to a distant site (53 patients), Figure 4.1, Tables 4.3 and 4.2. The
proportion of G2 tumours with locoregional and or distant metastases was even higher
at 78 %, (28 patients). Unsurprisingly all of the G3 tumours (10 patients, 100 %) had
lymph node metastases and/or distant sites, (p value: 0.000021).
Nearly a third of the G1 patients had stage IV disease, 28 %, (32 patients) despite
having a Ki-67 % of ≤ 2 %. Liver metastases were present in 24 % of G1 patients (27
patients) and were by far the most common site of distant disease manifestation. 84 %
of the G1 patients with stage IV disease had liver metastases (27/32 patients).
72 % of the G2 patients had stage IV disease (26 patients), with only 2 patients having
lymph node metastases in the absence of distant disease manifestation (Figure 4.1). Liver
metastases were present in 69 % of the G2 patients (25 patients) and were present in all
but one of the G2 patients with stage IV disease (25/26 patients, 96 %).
For the G3 tumours, 90 % of patients had spread of the disease to a distant site (9
patients). While none of the G3 patients were free from metastases, 1 patient had only
locoregional metastases but no distant manifestation of the disease. Liver metastases
were present in 80 % of the G3 patients, (8 patients).
223
Figure 4.1.: Proportion of patients with metastases stratified by tumour grade
224
Site of metastases
The most common metastatic site was the lymph nodes, with lymph node metastases
being present in 44 % of patients (71 patients). For 15 % of patients locoregional lymph
node metastases were their only site of metastatic disease manifestation (24 patients).
The liver was by far the most common site for distant metastasis. Liver metastases
were present in 37 % of patients overall (60/161 patients) and in 90 % of the patients
with stage IV disease (60/67 patients).
Liver metastases were a common feature in patients with low grade tumours with 24 %
the patients with G1 tumours (27 patients) and 69 % of the patients with G2 tumours (25
patients) having liver metastases, Table 4.2. 80 % of patients with high grade tumours,
G3, had liver metastases (8 patients) (p value: 0.000000036).
The majority of the liver metastases 82 %, were synchronous with respect to the pri-
mary tumour (49 patients), with the remainder being metachronous (11 patients). There
were quite similar rates of synchronous/metachronous liver metastases across the different
tumour grades although G2 tumours had a slightly higher proportion of metachronous
liver metastases, 24 %, than did G1 and G3 tumours at 15 % and 13 % respectively,
Table 4.2.
The patients were assessed morphologically to determine the type of growth in the
liver metastasis, this was classified as type I (one single metastasis), type II (isolated
metastatic bulk with smaller deposits alongside it) or type III (disseminated metastatic
spread) (Frilling et al., 2009; Frilling and Clift, 2015). Type III growth was the most
common, present in 45 % of liver metastases (27 patients), while type II growth and type
I growth was present in 32 % and 23 % of the liver metastases respectively (19 patients,
14 patients). When the tumour grade was taken into account type III growth remained
the most common for the G1 and G3 tumours representing 44 % (12 patients) and 75 %
(6 patients) of the liver metastases for these grades respectively, however type II growth
was most common in the G2 tumours, 48 % of which had type II growth (12 patients),
225
Table 4.2.
While the liver was the most common site for distant metastatic spread, 17 % of patients
(28 patients) had metastasis of their GEP-NET to one or more other distant sites. This
was either in addition to their liver metastasis, in 21 patients, or in the absence of any
hepatic disease manifestation in 7 patients.
Overall, 8 % of patients had bone metastases (13 patients), 7 % had lung metastases (11
patients) and 6 % had peritoneal metastases (10 patients), after the liver these represented
the most common sites of distant disease manifestation, Table 4.4. 89 % of the patients
with distant non-hepatic disease manifestation had a metastasis to one or more of these
sites (25/28 patients). Rarer sites for distant metastases included metastases in the neck,
stomach, pericardium, uterus and adnexa, brain, rectum, large bowel and spleen, see
Table 4.4.
SBNET and PNET
There were 37 patients with a SBNET and 84 patients with a PNET were included in
the study, these patients were stratified by tumour grade to determine the proportions
of metastases occurring for each grade, Table 4.5. Other patient characteristics such
as genetic status and functionality were also investigated in SBNET and PNET, these
results are presented in sections 4.4.3 and 4.4.4. All of the 37 SBNET patients included
in the study had low grade (G1/G2) tumours and most of the 84 PNET patients included
in the study also had low grade tumours (82/84 patients) with the exception of 2 PNET
patients who had G3 tumours.
The majority of the patients with low grade SBNET, 92 %, had metastatic disease, (34
patients), with 65 % having stage IV disease (24 patients). Stratified by grade, of the G1
SBNET, 89 % had metastases (25 patients) and 54 % had stage IV disease (15 patients),
Table 4.5. All of the G2 SBNET had metastatic disease manifestation to locoregional
and distant sites (9 patients). There were no high grade (G3) SBNET included in the
226
study, Table 4.2.
Of the 82 low grade PNET included in the study a high proportion, 42 %, had
metastatic disease and 32 % had stage IV disease (34 and 26 patients respectively). When
considered by grade, the G1 PNET had metastases in 30 % of patients (18 patients) and
stage IV disease in 20 % of patients (12 patients). The proportion of metastatic patients
for the G2 PNET was higher than for G1 patients, with 73 % of G2 patients having
metastases (16 patients) and 64 % having stage IV disease (14 patients). Both of the
G3 PNET (n=2) had locoregional metastases but only 1 of these patients had stage IV
disease (see Table 4.5).
4.3.2. Summary
These results demonstrate that metastases, particularly to the liver are a common occur-
rence even in low grade GEP-NET, with 46 % of G1 tumours and 78 % of G2 tumours
being metastatic, although these rates are still lower than for G3 tumours (100 %). There
were similar findings for the presence of stage IV disease which was also found at rela-
tively high rates in patients with low grade tumours, it was present in 28 % of G1 patients
and 72 % of G2 patients, the figure for G3 patients was 90 %. When the primary site
was considered, 65 % of patients with a low grade SBNET and 32 % of patients with a
low grade PNET were found to have stage IV disease.
4.4. Tumour characteristics
4.4.1. Invasiveness
Angio/lymphovascular invasion and perineural invasion was assessed by tumour grade,
where the data was available. The presence or absence of angio/lymphatic invasion was
recorded for 99 of the 161 patients, of these patients, 43 % had angio/lymphovascular
invasion (43 patients, 43/99). Stratified by grade, 35 % of patients with G1 tumours
227
for whom data was available had angio/lymphovascular invasion (25 patients, 25/72),
with an even higher figure for G2 tumours of 65 % (15 patients, 15/23). For the G3
tumours angio/lymphovascular was present in 75 % of tumours for which it was recorded
(3 patients, 3/4).
The presence or absence of perineural invasion was recorded for 54 of the 161 patients,
of these patients 33 % had perineural invasion (18 patients, 18/54). For the G1 tumours,
perineural invasion was present in 28 % of tumours from patients were data was available
(12 patients, 12/43) compared to 50 % for the G2 tumours (4 patients, 4/8) and 67 %
for G3 tumours (2 patients, 2/3).
4.4.2. Functionality and genetic status
Overall 39 % of the tumours were functioning tumours (63 patients), Table 4.2. A
functioning syndrome was most frequently observed in patients with G1 tumours of whom
44 % had a functioning syndrome (50 patients) compared to 33 % for G2 tumours (12
patients) and 10 % for G3 tumours (1 patient).
The most common functioning syndromes were insulinomas, present in 19 % of patients
(30 patients) followed by carcinoid syndrome which was present in 10 % of patients (16
patients) and gastrinomas present in 7 % of patients (11 patients), Table 4.6 (see also
sections 4.4.3 and 4.4.4).
The majority of the GEP-NET in the study were the result of sporadic disease, 94
%, (152 patients), however a few patients, 6 %, had the familial MEN1 syndrome with
mutations in MEN1 (9 patients), all of them PNET.
4.4.3. SBNET
There were 37 SBNET included in the study, these were predominantly G1 tumours, 76
% (28 patients), with the remainder being G2 SBNET (9 patients, 24 %), Table 4.2. All
of the SBNET patients had sporadic disease. Most of the primary SBNET were unifocal,
228
84 % (31 patients), while 16 % of patients had a multifocal primary tumour (6 patients).
Carcinoid syndrome was present in 38 % of the patients with a SBNET (14 patients)
and represented the only functioning syndrome observed in the SBNET patients, Table
4.7. Subdivided by grade, 36 % of the G1 SBNET had carcinoid syndrome (10 patients),
while 44 % of the G2 SBNET had carcinoid syndrome (4 patients).
4.4.4. PNET
There were 84 PNET included in the study. The PNET were predominantly G1 tumours,
71 % (60 patients), with 26 % being G2 tumours (22 patients) and 2 % being G3 tumours
(2 patients), Table 4.2. The majority of the PNET had sporadic disease, 92 %, (77
patients), however 8 % had MEN1 syndrome with mutations in MEN1 (7 patients), Table
4.8. 86 % of the primary tumours were unifocal (72 patients) however in a minority of
patients, 14 %, there was a multifocal primary (12 patients) (Table 4.8).
There was a nearly equal split between functioning and non-functioning PNET, 49 %
were functioning while 51 % were non-functioning (41 patients and 43 patients respec-
tively). When stratified by grade, for the G1 tumours, functioning tumours were more
common than non-functioning tumours while the reverse was true for G2 tumours where
non-functioning tumours were the most represented for these PNET (see Table 4.8). 55
% of the G1 PNET patients (33 patients) had a functioning tumour while 64 % of the
patients with a G2 PNET (14 patients) had a functioning tumour. Neither of the 2
patients with G3 PNET had a functional syndrome.
Of the functioning tumours, insulinomas were the most common, present in 73 % of
the PNET (30 patients), while 15 % and 10 % of the PNET was either a gastrinoma or a
somatostatinoma respectively (6 patients, 4 patients). Subdivided by grade, insulinomas
were by far the most common in the G1 PNET, present in 82 % of the G1 PNET (27
patients) compared to just 9 % for gastrinomas (3 patients) and 6 % for somatostatinomas
(2 patients). For the G2 PNET however, insulinomas and gastrinomas were equally
229
common with 38 % of G2 PNET patients having each of these syndromes (3 patients, 3
patients) and 25 % of G2 PNET suffering from a somatostatinoma (2 patients). One of
the G1 patients had a VIPoma with no MEN1 syndrome.
4.5. Second primary malignancies
Second primary malignancies were observed in 9 % of the GEP-NET patients (14 pa-
tients), Table 4.9. This was when in addition to their GEP-NET, patients also had a
separate non-NET malignancy. These second primary malignancies included colon, skin,
lymphatic system, breast, pancreas, kidney and tonsil carcinomas.
Second primary malignancies were most common in the G3 tumours, where 30 % of
patients had a second primary malignancy (3 patients, 3/10) compared to 11 % for G2
tumours (4 patients, 4/36) and 6 % for G1 tumours (7 patients, 7/115) (p value: 0.036).
The second primary malignancies had been diagnosed prior to the GEP-NET in 9 pa-
tients and had a synchronous diagnosis with the GEP-NET in 5 patients. Synchronous
diagnosis of the GEP-NET was less common than metachronous diagnosis for all tumour
grades with 43 % of G1 GEP-NET having synchronous diagnosis (3 patients, 3/7), com-
pared to 25 % for G2 GEP-NET (1 patient, 1/4) and 33 % for the G3 GEP-NET (1
patient, 1/3) (see Table 4.9 and Figure 4.2).
4.6. Ki-67 % Heterogeneity
4.6.1. Patients
There were 30 GEP-NET patients included in the Ki-67 % heterogeneity part of the
study. Ki-67 % was assessed for these patients at multiple sites either within the same
tumour/metastasis (intratumoural heterogeneity) or between the primary tumours and
metastases (intertumoural heterogeneity). Patients number 1-17 were included in our
230
Figure 4.2.: Distribution of second primary malignancies by GEP-NET grade (n=14).Reprinted by permission from the Licensor: Springer Nature [World Journalof Surgery] [(Miller et al., 2014)], ©(2014).
231
World Journal of Surgery publication, for more details and patient selection see Methods,
section 3.2.4 (Miller et al., 2014).
To investigate the extent intertumoural heterogeneity, Ki-67 % was assessed for 24
patients in all lesions with tissue available. To investigate intratumoural heterogeneity,
Ki-67 % was assessed in 5 different sites within liver metastases from 6 explanted livers
from candidates for liver transplantation.
The median age of the 30 GEP-NET patients was 57 (range: 35-81). There were
more male patients in the study, 18 patients, than female patients, 12 patients. The
most common primary site was the small bowel, in 16 patients, followed by the pancreas
in 11 patients, see Table 4.10. One patient had a GEP-NET liver metastasis where the
primary location was investigated but remained unknown. There were three patients with
multifocal PNET (patients number 14, 15 and 22, Table 4.11). There were two metastatic
liver lesions available for Ki-67 % assessment in two patients (patients number 25 and
28, Table 4.11).
4.6.2. Intertumoural heterogeneity
For 24 patients, the Ki-67 % was assessed in two or more different lesions for each patient
according to ENETS guidelines, patients number 1-24, Table 4.11 (Methods, section
3.2.4). There was a change in grade between the different sites assessed in 54 % of the
patients when the second site was taken into account (13 patients), Figure 4.4 and Figure
4.3. The grade increased in 42 % of the patients (10 patients). For 9/10 patients with
an increase in grade between the primary and the second site the change was from G1 to
G2, Figure 4.4. There was only one patient (patient number 5, Table 4.11) with a shift
in grade from G2 to G3.
Of the patients with liver metastases, 46 % (5 patients, 5/11) had a higher Ki-67 %
in their liver metastasis leading to a higher grade (G2) than that of the primary tumour
(G1). There were 2 patients with a lower grade in the liver metastasis than in the
232
Figure 4.3.: Ki-67 IHC at different tumour sites for patient 7 (Table 4.11), showing anincrease in the number of Ki-67 positive cells between the primary tumour(G1) and the metastases (G2). Positive nuclei are stained in brown, X10magnification. A: SBNET, Ki-67 %: 1 %. B Lymph node metastasis, Ki-67%: 3 %. C: Liver metastasis, Ki-67 %: 8 %. Reprinted by permission fromthe Licensor: Springer Nature [World Journal of Surgery] [(Miller et al.,2014)], ©(2014).
primary tumour. These findings suggest that many patients would be undergraded if
only the Ki-67 % of the primary tumour was taken into account.
4.6.3. Intratumoural heterogeneity
The Ki-67 % was assessed for GEP-NET liver metastases from explanted livers from 6
patients who were candidates for liver transplantation (patients number: 25-30, Table
4.11). There were two explanted livers where two lesions were assessed (see methods
section 3.2.4). The results for the 5 different sites assessed for Ki-67 % within each liver
lesion are shown in Table 4.12.
Of the 6 patients, 67 % (4 patients) had a different grade depending on the region of
the liver metastasis assessed, with the difference in the Ki-67 % being large enough to
shift the patient’s grade from G1 to G2 (Figures 4.5 and 4.6). All grade changes were
from G1 to G2.
233
Figure 4.4.: A: Number of patients graded as G1, G2 or G3 based on Ki-67 % of theprimary tumour. B: Number of patients with a change in grade based on theKi-67 % at another tumour site.
Figure 4.5.: The minimum and maximum Ki-67 % are shown for the 5 different sitesassessed within each liver lesion. Grey circle: increased grade, yellow circle:same grade.
234
Figure 4.6.: Minimum Ki-67 % is in blue, maximum Ki-67 % is in red. #: lesion, where2 metastatic lesions were available for Ki-67 % assessment. The dotted lineindicates the G1/G2 boundary.
235
4.6.4. Summary
These findings suggest that there is considerable intratumoural and intertumoural het-
erogeneity in Ki-67 % in GEP-NET which is sufficient to change the tumour grade in
a high proportion of patients. This could lead to under grading of GEP-NET with the
potential to affect the treatment that patients receive. The majority of the changes in
grade were from G1 to G2. Currently G1 and G2 GEP-NET patients receive the same
treatment, however this may change in the future as new therapies become available.
4.7. Conclusions
This chapter has addressed the 1st research objective of this thesis by investigating the
limitations of the existing prognostic biomarker used in GEP-NET, Ki-67 %. Ki-67 %
was assessed and disease staging was carried out in 161 GEP-NET patients including 84
patients with PNET and 37 patients with SBNET.
The principle finding was that there is no Ki-67 % at which a GEP-NET patient can be
considered to be safe from liver metastases. Stage IV disease was a common occurrence
even in patients with a G1 GEP-NET. Stage IV disease was present in 28 % of G1
patients, with a Ki-67 % of ≤ 2 %, and in 72 % of G2 patients with a Ki-67 % of 3-20
%. The most common site of distant metastasis was the liver with 24 % the G1 tumours
and 69 % of G2 tumours having liver metastases.
A similar pattern was found when the data was analysed by primary site with distant
metastases being a common occurrence even when the Ki-67 % was low. All of the
SBNET patients had low grade tumours (G1/G2). Stage IV disease was present in a
high proportion of the G1 SBNET patients, 54 %, and in all of G2 SBNET patients (100
%). 82/84 of the PNET patients had low grade tumours. Stage IV disease was present
in 20 % of the G1 PNET patients, 64 % of the G2 PNET patients and 50 % of the G3
patients.
236
Ki-67 % heterogeneity was assessed in 30 patients revealing considerable intertumoural
and intratumoural heterogeneity which was sufficient to change the tumour grade in 54
% and 67 % of patients respectively. This could lead to undergrading with the potential
to affect the treatment that GEP-NET patients receive.
These results demonstrate that despite expressing low levels of the proliferation marker
Ki-67, low grade GEP-NET frequently metastasise to distant sites. It would therefore
be useful to have additional prognostic biomarkers for use alongside Ki-67 % that could
predict which patients with low grade GEP-NET might have a more aggressive disease
course. The chapters that follow, chapters 5 and 6 are concerned with the identification
of potential novel biomarkers for use in patients with low grade GEP-NET that could be
used for further patient stratification.
237
Table 4.2.: Patient characteristics stratified by grade. Reprinted by permission from theLicensor: Springer Nature [World Journal of Surgery] [(Miller et al., 2014)],©(2014).
Grade 1 Grade 2 Grade 3 TotalNo. of patients 115 36 10 161Median age 62 58 61 61Age range 21-91 35-83 36-86 21-91Gender
Male 64 [56%] 20 [56%] 4 [40%] 88 [55%]Female 51 [44%] 16 [44%] 6 [60%] 73 [45%]
Site of originPancreas 60 [52%] 22 [61%] 2 [20%] 84 [52%]Jejunum/ileum 28 [24%] 9 [25%] 0 37 [23%]Duodenum 12 [10%] 0 0 12 [8%]Stomach 7 [6%] 3 [8%] 2 [20%] 12 [8%]Rectum 4 [4%] 1 [3%] 0 5 [3%]Colon 0 0 2 [20%] 2 [1%]Oesophagus 0 0 1 [10%] 1 [1%]CUP 4 [4%] 1 [3%] 3 [30%] 8 [5%]
Functioning tumour 50 [44%] 12 [33%] 1 [10%] 63 [39%]ENETS stage
Any T N0M0 62 [54%] 8 [22%] 0 70 [44%]Any T N1M0 21 [18%] 2 [6%] 1 [10%] 24 [15%]Any T N1M1 20 [17%] 18 [50%] 9 [90%] 47 [29%]Any T N0M1 12 [10%] 8 [22%] 0 20 [12%]
Liver metastasis 27 [24%] 25 [69%] 8 [80%] 60 [37%]Synchronous 23 [85%] 19 [76%] 7 [88%] 49 [82%]Metachronous 4 [15%] 6 [24%] 1 [13%] 11 [18%]Type I growth 9 [33%] 4 [16%] 1 [13%] 14 [23%]Type II growth 6 [22%] 12 [48%] 1 [13%] 19 [32%]Type III growth 12 [44%] 9 [36%] 6 [75%] 27 [45%]
Status at studyAlive 100 [87%] 30 [83%] 5 [50%] 135 [84%]Dead from disease 10 [9%] 5 [14%] 5 [50%] 20 [12%]Dead other cause 4 [4%] 0 0 4 [3%]Lost to follow up 1 [1%] 1 [3%] 0 2 [1%]
Table 4.3.: Summary of tumour stage
Grade 1 Grade 2 Grade 3Metastasis∗ 46% 78% 100%Stage IV 28% 72% 90%Any T N0M0 54% 22% 0%
∗ Lymph nodes and/or distant site
238
Table 4.4.: Location of distant metastases
Distant metastasis Number of patientsLiver 60Other sites (non-liver) 28
Bone 13Lung 11Peritoneum 10Large bowel 2Brain 2Neck 2Mediastinum 2Uterus/adnexa 2Spleen 1Pericardium 1Stomach 1Subcutaneous 1
Table 4.5.: Stage SBNET and PNET
Stage Grade 1 Grade 2 Grade 3 TotalSBNET
Any T N0M0 3 [11%] 0 0 3 [8%]Any T N1M0 10 [36%] 0 0 10 [27%]Any T N1M1 12 [43%] 9 [100%] 0 21 [57%]Any T N0M1 3 [11%] 0 0 3 [8%]
PNETAny T N0M0 42 [70%] 6 [27%] 0 48 [57%]Any T N1M0 6 [10%] 2 [9%] 1 [50%] 9 [11%]Any T N1M1 5 [8%] 7 [32%] 1 [50%] 13 [16%]Any T N0M1 7 [12%] 7 [32%] 0 14 [17%]
Table 4.6.: Functioning syndromes
Number of patientsInsulinoma 30Carcinoid syndrome 16Gastrinoma 11Somatostatinoma 4Functioning gastric NET 1Vipoma 1
239
Table 4.7.: SBNET
Grade 1 Grade 2 TotalNo. of patients 28 9 37Functionality
Functioning 10 [36%] 4 [44%] 14 [38%]Non functioning 18 [64%] 5 [56%] 23 [62%]
SyndromeCarcinoid syndrome 10 [36%] 4 [44%] 14 [38%]Other 0 0 0
Focality (primary)Unifocal 23 [82%] 8 [89%] 31 [84%]Multifocal 5 [18%] 1 [11%] 6 [16%]
Table 4.8.: PNET
Grade 1 Grade 2 Grade 3 TotalNo. of patients 60 22 2 84Genetic status
Sporadic 54 [90%] 21 [95%] 2 [100%] 77 [92%]MEN1 6 [10%] 1 [5%] 0 7 [8%]VHL 0 0 0 0
FunctionalityFunctioning 33 [55%] 8 [36%] 0 41 [49%]Non functioning 27 [45%] 14 [64%] 2 [100%] 43 [51%]
SyndromeInsulinoma 27 [82%] 3 [38%] 0 30 [73%]Gastrinoma 3 [9%] 3 [38%] 0 6 [15%]Somatostatinoma 2 [6%] 2 [25%] 0 4 [10%]Verner Morrison 0 0 0 0Carcinoid syndrome 0 0 0 0Other 1 [3%] 0 0 1 [2%]
Focality (primary)Unifocal 51 [85%] 19 [86%] 2 [100%] 72 [86%]Multifocal 9 [15%] 3 [14%] 0 12 [14%]
Table 4.9.: Second primary malignancies
Grade 1 Grade 2 Grade 3No. of patients, non-NET malignancy 7 [6%] 4 [11%] 3 [30%]
Synchronous 3 1 1Metachronous 4 3 2
240
Table 4.10.: Primary sites in heterogeneity study
Primary NET Number of patientsSmall bowel 16Pancreas 11Rectum 1Stomach 1Unknown 1
241
Table 4.11.: Ki-67 % at different disease sites, (patients 1-17 only, reprinted by permissionfrom the Licensor: Springer Nature [World Journal of Surgery] [(Miller et al.,2014)], ©(2014)).
PatientNo.
Age(years)
GenderPrimaryNET site
Ki-67 Index (%)Primary site Metastatic site
Liver Lymphnodes
Peri-toneum
1 59 F small bowel 1 1.5 32 40 F stomach 40 303 81 M small bowel 2 8 14 64 M pancreas 15 35 64 F pancreas 20 236 73 F small bowel 1-2 4-5 <17 79 M small bowel 1 8 38 55 F small bowel <2 2 2-59 40 M pancreas 10 15-2010 69 M small bowel <2 1011 74 F small bowel 3 <212 57 F small bowel <1 <113 67 M small bowel <2 <114 58 M pancreas lesion 1: 3-4
lesion 2: < 115 42 M pancreas lesion 1: 4
lesion 2: 116 57 M pancreas 3 <217 48 F pancreas 3-4 818 58 M small bowel 4.5 119 53 F small bowel 1 120 64 M small bowel <1 <121 50 M small bowel <1 122 35 M pancreas lesion 1: <2
lesion 2: <223 46 M rectum <2 3 524 47 M small bowel <2 <225 57 M small bowel <2 lesion 1:
3.5lesion 2:2.4
26 52 F unknown n/a 1.827 53 F pancreas <2 428 67 M small bowel <2 lesion 1:
1lesion 2:6
29 62 M pancreas <2 5.630 55 F pancreas <2 5.6
242
Table 4.12.: Liver metastases
Patient No. Ki-67 % in livermetastasis regionsA B C D E
25 (lesion 1) 3.0 3.0 3.5 1.5 2.5(lesion 2) 2.4 1.6 2.3 1.3 2.3
26 1.7 0.3 1.8 0.2 1.627 4.0 4.0 3.0 3.6 2.428 (lesion 1) 0.6 0.5 0.6 0.6 1.0
(lesion 2) 3.6 6.0 4.0 0.1 6.029 3.0 1.6 1.6 3.0 5.630 5.6 3.3 3.1 3.8 1.6
243
5. Global miRNA expression
profiling in SBNET, miRNA
quantification in matched tissue
from the primary tumour and
metastatic sites
5.1. Introduction
Results presented in chapter 5, were published in Endocrine Related Cancer in 2016
(Miller et al., 2016).
In this chapter results are presented from the global miRNA expression profiling of
FFPE tissue from SBNET patients treated at Imperial College Healthcare NHS Trust
(London, UK). These results form dataset 1 (n=47). This chapter follows on from chapter
4 in which limitations of current GEP-NET biomarker, Ki-67 %, were identified.
On the basis of the results that are presented in this chapter (chapter 5), primary and
metastatic tissue was sought from SBNET patients treated at a separate institution. This
was to determine if the results could be validated in an independent group of SBNET
patients. A secondary goal was to include tissue from a larger number of liver metastases
244
than was available at Imperial College Healthcare NHS Trust. To this end miRNA
expression was quantified in tumour tissue from SBNET patients treated at Zentralklinik
Bad Berka (Bad Berka, Germany). These results form dataset 2 (n=43) and are presented
in the next chapter, chapter 6.
MiRNA were quantified in FFPE tissue from matched primary tumour, metastases
and adjacent normal tissue from 15 patients with SBNET treated at Imperial College
Healthcare NHS Trust. Overall, 800 miRNA were quantified in 47 tissue samples using the
NanoString nCounter Human miRNA expression assay. Comparisons were made between
tissue types to determine the miRNA that were most dysregulated in SBNET. This was
to determine the miRNA expression profile of a SBNET in order to better understand
the potential role of miRNA in SBNET tumourigenesis and to identify possible novel
biomarkers for patient stratification. Potential miRNA biomarkers, candidate miRNA,
were quantified by a second quantification method (qPCR) to confirm the results.
This chapter addresses the second research objective of this thesis:
“2) Experimentally determine a global miRNA profile of SBNET.”
5.1.1. Summary of results
The quantification of miRNA in matching tissue from SBNET and their metastases and
adjacent normal tissue resulted in the identification a global miRNA profile of SBNET.
Many miRNA were identified for the first time as being differentially regulated in primary
tumours and with disease progression. Candidate miRNA were identified and changes in
their expression levels was confirmed by qPCR. These miRNA have the potential to be
used as biomarkers in SBNET patients in the future.
245
5.2. Global miRNA expression profile
Global miRNA quantification was done in matched primary tumour, metastatic and
adjacent normal tissue samples from 15 SBNET patients with low grade (G1/G2) tumours
treated at Imperial College Healthcare NHS Trust, see Table 5.1. Dataset 1 was comprised
of the results from this miRNA analysis. MiRNA was extracted from all tumour sites for
which tissue was available.
There were 47 FFPE samples included in the study, for clinical details see Methods,
section 3.3.1, Table 3.2 and for ethics approval see Methods, section 3.1.
The median age of the SBNET patients was 69 (range: 38-84 years). The study
included almost equal numbers of female and male patients (7 and 8 respectively).
Table 5.1.: Samples, dataset 1
Tissue type NumberPrimary SBNET 15Lymph node metastasis 9Liver metastasis 2Adjacent normal small bowel 12Normal lymph node 7Adjacent normal liver 2
5.2.1. SBNET
There were 212 miRNA with a significant change in expression in SBNET compared to
adjacent normal small bowel tissue, using a FDR of < 0.05 (Benjamini–Hochberg adjusted
p value, see Methods, section 3.3.3). There were 72 miRNA that were significantly
downregulated in SBNET and 140 that were significantly upregulated in SBNET, these
are shown in Figure 5.2 and Figure 5.1 respectively.
246
Fig
ure
5.1.
:m
iRN
Aw
ith
asi
gnifi
can
tin
crea
sein
expr
essi
onin
SB
NE
Tre
lati
veto
adja
cen
tn
orm
alsm
all
bow
elti
ssu
e.*
FD
R<
0.05
,**
FD
R<
0.00
1,**
*F
DR<
0.00
01.
For
enla
rged
xax
isla
bels
plea
sere
fer
toT
able
5.2
247
Table 5.2.: Enlarged x axis labels for Figure 5.1
Label miRNA Label miRNAon graph on graph1 miR-2110 71 miR-374b-5p2 miR-615-5p 72 let-7f-5p3 miR-1255a 73 miR-374a-5p4 miR-139-5p 74 miR-423-5p5 miR-125a-3p 75 miR-99a-5p6 miR-491-5p 76 miR-12067 miR-193a-3p 77 miR-744-5p8 miR-320d 78 miR-24-3p9 miR-33a-5p 79 miR-378b10 miR-550b-3p 80 miR-191-5p11 miR-210 81 miR-32812 miR-598 82 miR-30d-5p13 miR-574-3p 83 miR-361-3p14 miR-452-5p 84 miR-342-5p15 miR-504 85 miR-26a-5p16 miR-509-5p 86 miR-26b-5p17 miR-548b-3p 87 miR-141-3p18 miR-769-5p 88 miR-454-3p19 miR-664-3p 89 miR-23b-3p20 miR-335-5p 90 miR-29a-3p21 miR-508-3p 91 miR-505-3p22 miR-501-3p 92 miR-13723 miR-30b-5p 93 miR-425-5p24 miR-627 94 let-7g-5p25 miR-593-3p 95 miR-15a-5p26 miR-577 96 miR-615-3p27 miR-193a-5p 97 miR-660-5p28 miR-379-5p 98 miR-652-3p29 miR-199a-5p 99 miR-42130 miR-151a-3p 100 miR-12831 miR-93-5p 101 miR-34a-5p32 miR-25-3p 102 let-7d-5p33 miR-484 103 miR-181b-
5p+181d34 miR-130a-3p 104 miR-361-5p35 miR-106b-5p 105 miR-129-5p36 miR-153 106 miR-125a-5p37 miR-1301 107 let-7e-5p38 miR-663b 108 miR-29c-3p39 let-7b-5p 109 miR-132-3p
248
Continuation of Table 5.2
Label miRNA Label miRNAon graph on graph40 miR-365a-3p 110 miR-532-5p41 miR-378f 111 miR-29b-3p42 miR-15b-5p 112 miR-148b-3p43 miR-100-5p 113 let-7i-5p44 miR-219-5p 114 miR-200b-3p45 miR-330-5p 115 miR-9846 miR-200c-3p 116 miR-301a-3p47 miR-500a-
5p+501-5p117 miR-331-3p
48 miR-1468 118 miR-10749 miR-185-5p 119 miR-99b-5p50 miR-130b-3p 120 miR-42951 miR-423-3p 121 miR-135a-5p52 miR-197-3p 122 miR-486-3p53 let-7a-5p 123 miR-551b-3p54 let-7c 124 miR-330-3p55 miR-323a-3p 125 miR-642a-5p56 miR-23a-3p 126 miR-342-3p57 miR-320a 127 miR-324-5p58 miR-532-3p 128 miR-181c-5p59 miR-362-5p 129 miR-200a-3p60 miR-582-5p 130 miR-96-5p61 miR-708-5p 131 miR-196a-5p62 miR-151a-5p 132 miR-118063 miR-186-5p 133 miR-9564 miR-542-5p 134 miR-182-5p65 miR-362-3p 135 miR-183-5p66 miR-340-5p 136 miR-48967 miR-125b-5p 137 miR-129-2-3p68 miR-16-5p 138 miR-204-5p69 miR-27b-3p 139 miR-37570 miR-874 140 miR-7-5p
In order to generate a profile of differentially expressed miRNA in SBNET, only those
miRNA with a log 2 fold change (log2FC) in expression of ≥ 1.5 or ≤ −1.5 were con-
sidered. This corresponds to a 3 fold increase or decrease in expression (fold change of
approximately 3, see Methods, section 3.3.3).
249
Figu
re5.2.:
miR
NA
with
asign
ifican
tdecrease
inexpression
inS
BN
ET
relativeto
adjacent
norm
alsm
allbow
eltissu
e.*
FD
R<
0.05,**
FD
R<
0.001,***
FD
R<
0.0001.
250
These more stringent criteria, produced a signature of 63 miRNA which were differ-
entially expressed in SBNET compared to adjacent normal small bowel tissue (log2FC
≥ 1.5 or ≤ −1.5, FDR < 0.05). There were 55 miRNA that were upregulated in SBNET
and 8 miRNA that were downregulated, see Tables 5.3 and 5.4.
The miRNA that had the greatest increase in expression was miR-7-5p with a log2FC
of 6.4 (FDR: 8.0 x 10-109) while the miRNA with the greatest decrease in expression was
miR-215 with a with a log2FC of -3.3 (FDR: 8.1 x 10-20).
Table 5.3.: SBNET miRNA profile, most upregulated miRNA
Upregulated miRNA (dataset 1)miRNA log2FC FDRmiR-7-5p 6.4 8.0039E-109miR-375 5.7 2.20335E-68miR-204-5p 5.2 1.9373E-62miR-129-2-3p 4.6 5.5816E-26miR-489 4.0 8.4092E-28miR-183-5p* 3.9 1.55415E-24miR-182-5p* 3.9 3.85719E-20miR-95 3.8 1.46191E-39miR-1180 3.6 1.61452E-42miR-196a-5p* 3.2 1.99056E-09miR-96-5p* 3.2 2.16357E-17miR-200a-3p* 3.2 1.45664E-22miR-181c-5p 3.1 2.63525E-38miR-324-5p 3.1 1.07522E-22miR-342-3p 2.9 1.17464E-16miR-642a-5p 2.9 3.28763E-37miR-330-3p 2.8 1.39674E-34miR-551b-3p 2.7 5.30586E-10miR-486-3p 2.6 1.09829E-13miR-135a-5p 2.6 8.08028E-08miR-429 2.6 4.19338E-19miR-99b-5p 2.6 1.8234E-16miR-107 2.5 9.83865E-14miR-331-3p 2.5 5.60415E-21miR-301a-3p 2.5 6.16426E-21miR-98 2.4 5.31477E-15miR-200b-3p 2.4 2.77176E-12
251
Continuation of Table 5.3
Upregulated miRNA (dataset 1)miRNA log2FC FDRlet-7i-5p 2.3 6.03222E-15miR-148b-3p 2.3 9.65803E-22miR-29b-3p 2.3 1.42688E-14miR-532-5p 2.3 5.87916E-13miR-132-3p 2.2 9.82012E-11miR-29c-3p 2.2 1.72535E-14let-7e-5p 2.2 5.14319E-10miR-125a-5p 2.2 1.63693E-12miR-129-5p* 2.0 3.83482E-08miR-361-5p 2.0 5.71651E-12miR-181b-5p+181d 2.0 1.156E-13let-7d-5p 1.9 5.12672E-10miR-34a-5p 1.9 2.54637E-08miR-128 1.9 2.90003E-20miR-421 1.7 8.46111E-16miR-652-3p 1.7 2.50032E-10miR-660-5p 1.6 1.07427E-12miR-615-3p 1.6 7.86281E-15miR-15a-5p 1.6 2.51866E-07let-7g-5p 1.6 2.53956E-06miR-425-5p 1.6 1.08054E-08miR-137 1.6 3.08393E-07miR-505-3p 1.6 1.07919E-11miR-29a-3p 1.5 6.0698E-08miR-23b-3p 1.5 4.40622E-06miR-454-3p 1.5 2.04489E-10miR-141-3p 1.5 2.78064E-06miR-26b-5p 1.5 1.22241E-06
MiRNA had a FDR < 0.05 and a log2FC ≥ 1.5 or ≤ −1.5, *: miRNAalso identified by Li et al. (2013b).
5.2.2. Lymph node metastases
There were 8 miRNA with a significant change in expression in lymph node metastases
compared to the primary SBNET, using a FDR of < 0.05 (see Methods, section 3.3.3).
There were 4 miRNA that were significantly downregulated in the lymph node metastases
252
Table 5.4.: SBNET miRNA profile, most downregulated miRNA
Downregulated miRNA (dataset 1)miRNA log2FC FDRmiR-215* -3.3 8.09619E-20miR-378a-3p+378i -2.1 1.82391E-20miR-4516 -1.9 4.34411E-05miR-148a-3p -1.7 2.76117E-12miR-451a -1.7 4.96531E-05miR-378g -1.6 1.74596E-15miR-1915-3p -1.5 0.001850829miR-31-5p* -1.5 1.62969E-13
MiRNA had a FDR < 0.05 and a log2FC ≥ 1.5 or ≤ −1.5, *: miRNA also identified byLi et al. (2013b).
and 4 that were significantly upregulated in the lymph node metastases, these are shown
in Figure 5.3 and Table 5.5.
The miRNA that had the greatest increase in expression was miR-142-3p with lymph
node metastases having double the levels of this miRNA found in the SBNET (log2FC:
1.0, FC: 2.0, FDR: 3.8 x 10 -5). The miRNA with the greatest decrease in expression
was miR-133a with a with a log2FC of -1.0 (FC: 0.5, FDR: 5.6 x 10 -5). MiR-133a was
also identified as being significantly downregulated in lymph node metastases in the two
earlier studies by Li et al. (2013b) and Ruebel et al. (2010), see Literature review, section
2.5.3.
Table 5.5.: Significantly dysregulated miRNA in lymph node metastases versus SBNET
miRNA log2FC FDRmiR-142-3p 0.995252927 3.84919E-05miR-146a-5p 0.903204118 0.000320864miR-150-5p 0.822877597 0.000320864miR-548m 0.53991659 0.00650879miR-145-5p -0.709042891 0.018083561miR-1233 -0.777980086 0.000320864miR-1 -0.784645293 0.000396118miR-133a* -0.974871704 5.67702E-05
*: miRNA also identified by Li et al. (2013b) and Ruebel et al. (2010).
253
Figure 5.3.: miRNA that had a significant increase/decrease in expression in lymph nodemetastases compared to the primary tumour. * FDR < 0.05, ** FDR <0.001, *** FDR < 0.0001.
254
There were 106 miRNA with a significant change in expression in lymph node metas-
tases compared to matched normal lymph nodes, using a FDR of < 0.05 (for a full list
see Appendix, section D Table D.1). Using the log2FC in expression cut off of ≥ 1.5
or ≤ −1.5 resulted in 53 dysregulated miRNA in lymph node metastases compared to
normal lymph node tissue, these are shown in Figure 5.4.
5.2.3. Summary
This global screen of miRNA in matched tissue from SBNET and metastases identified
novel miRNA which had not been previously identified in SBNET and could have a role
in SBNET pathology and tumourigenesis. The miRNA identified as being differentially
expressed in the lymph node metastases could be involved in promoting disease progres-
sion. It would have been interesting to have been able to study miRNA expression in
a larger number of liver metastases since there were only 2 liver metastasis samples in-
cluded in dataset 1. This was addressed in the second miRNA profiling study (dataset
2) which included 13 tissue samples taken from SBNET liver metastases, see chapter 6,
section 6.3.1. The miRNA identified in this global profiling study have the potential to be
used as future biomarkers or therapeutic targets in SBNET if they are further validated
and prove to be clinically useful. The most differentially expressed miRNA were taken
forwards for validation (section 5.3).
5.3. Candidate miRNA validation by a second
quantification method
To confirm the findings of the profiling study in which miRNA were quantified using the
NanoString nCounter human miRNA expression assay, qPCR experiments were done to
determine if the most dysregulated miRNA from the profiling experiment could be con-
firmed as dysregulated using this different method of miRNA quantification (see Methods,
255
Figure 5.4.: miRNA that had a significant increase/decrease in expression in lymph nodemetastases compared to normal lymph nodes. * FDR < 0.05, ** FDR <0.001, *** FDR < 0.0001. Log2FC: ≥ 1.5 or ≤ −1.5
256
sections 3.3.4 and 3.3.3). The expression levels of 7 miRNA identified in the profiling ex-
periment were quantified by qPCR in SBNET compared to adjacent normal tissue, miR-
215-5p, miR-378i, miR-378a-3p, miR-451a, miR-7-5p, miR-204-5p and miR-375. The
expression levels of 3 miRNA were investigated in SBNET tissue compared to lymph
node metastases, miR-1, miR-143-3p and miR-1233. These miRNA have the potential to
be useful prognostic biomarkers in the future if found to be consistently dysregulated in
SBNET and their metastases.
5.3.1. SBNET
The 7 candidate miRNA with differential expression in SBNET compared to adjacent
normal tissue in the profiling study were confirmed as being significantly up or downreg-
ulated in SBNET in the validation study. Two separate endogenous control genes were
used for the qPCR normalisation, SNORD44 and RNU6-1 (see Methods, section 3.3.4).
The results for each candidate miRNA were very similar regardless of which endogenous
control was used for the normalisation (results for both are shown in Figures 5.5 and 5.6).
The relative expression levels of miR-7-5p, miR-204-5p and miR-375 were significantly
increased in the SBNET tissue compared to the adjacent normal small bowel tissue which
had very low levels of these particular miRNA, Figure 5.5. This confirms the findings
from the profiling study (see Figure 5.1).
The relative expression levels of miR-215-5p, miR-378i and miR-378a-3p were signifi-
cantly reduced in SBNET compared to adjacent normal tissue, Figure 5.6. These results
confirm those of the profiling study, in which these miRNA were also significantly reduced
in SBNET (Figure 5.2).
MiR-451a was significantly reduced in SBNET when normalized using SNORD44, p
value: 0.0080, but missed significance when normalised using RNU6-1, p value: 0.0600
(p value < 0.05 considered significant, Figure 5.6).
These qPCR results confirm the findings of the profiling experiment and show that the
257
Figure 5.5.: MiRNA with increased expression in small bowel primary (SBP) tumoursversus adjacent normal small bowel (SB N). The relative expression of eachmiRNA is shown for each sample. Results are shown from normalisationagainst both endogenous control genes, RNU6-1 and SNORD44. Error barsshow the mean +/- standard error of the mean (SEM). The scale of the yaxis varies between plots. * p < 0.05, ** p < 0.001, *** p < 0.0001.
changes in expression with respect to candidate miRNA in SBNET remain robust across
different experimental methods of miRNA quantification.
5.3.2. Lymph Node metastases
The expression levels of the 3 candidate miRNA, miR-1, miR-143-3p and miR-1233 that
were differentially expressed in the profiling study in SBNET tissue compared to their
lymph node metastases were investigated in qPCR experiments. MiR-1 and miR-143-
3p both had a significant reduction in relative expression in lymph node metastases
compared to SBNET confirming the results of the profiling study. These results held
true when normalised against both SNORD44 and RNU6-1, see Figure 5.7. For miR-
1233 there was however no significant change in relative expression between the lymph
258
Figure 5.6.: MiRNA with decreased expression in small bowel primary (SBP) tumoursversus adjacent normal small bowel (SB N). The relative expression of eachmiRNA is shown for each sample. Results are shown from normalisationagainst both endogenous control genes, RNU6-1 and SNORD44. Error barsshow the mean +/- standard error of the mean (SEM). The scale of the yaxis varies between plots. * p < 0.05, ** p < 0.001, *** p < 0.0001.
node metastases and the SBNET so the profiling results could not be confirmed for this
miRNA (SNORD44 p value: 0.2328, RNU6-1 p value: 0.2383).
These findings suggest that miR-1 and miR-143-3p may be promising candidates to
take forwards for further studies since they have a robust reduction in expression in lymph
node metastases compared to SBNET across different miRNA quantification methods.
5.3.3. Summary
There was a very dramatic difference in the expression of the upregulated miRNA, miR-
7-5p, miR-204-5p and miR-375 in SBNET with virtually no expression of these miRNA in
the adjacent normal small bowel tissue. This suggests that these miRNA would be good
candidates for a possible biomarker in SBNET. There was also a robust downregulation in
SBNET of miR-215-5p, miR-451a, miR-378a-3p and miR-378i and of miR-1, miR-143-3p
in lymph node metastases which confirmed the results from the global profiling study.
259
Figure 5.7.: MiRNA with decreased expression in lymph node metastases (LNM) tissueversus small bowel primary (SBP) tissue. The relative expression of eachmiRNA is shown for each sample. Results are shown from normalisationagainst both endogenous control genes, RNU6-1 and SNORD44. Error barsshow the mean +/- standard error of the mean (SEM). The scale of the yaxis varies between plots. * p < 0.05, ** p < 0.001, *** p < 0.0001.
260
These miRNA could be promising candidates for further studies into the mechanisms
of SBNET tumourigenesis and disease progression and for the development of future
biomarkers.
5.4. Conclusions
This chapter has fulfilled the second research objective of this thesis, by experimentally
determining a global miRNA profile of SBNET. The expression levels of 800 miRNA were
assessed in matched tissue from 15 patients with low grade SBNET treated at Imperial
College Healthcare NHS Trust.
Global miRNA expression profiling using the NanoString nCounter human miRNA
expression assay determined the miRNA profile of SBNET. Novel miRNA were revealed
that had not been previously implicated in SBNET tumourigenesis. The SBNET profiling
identified 140 miRNA that were significantly upregulated in SBNET and 72 miRNA
that were significantly downregulated in SBNET relative to adjacent normal small bowel
tissue. There were 8 miRNA identified that were significantly dysregulated in lymph node
metastases compared to the SBNET and these miRNA could be involved in promoting
disease progression in SBNET.
Further experiments, using a second method of miRNA quantification (qPCR), con-
firmed the miRNA profiling results for potential novel miRNA biomarkers, candidate
miRNA. These results identified a particularly dramatic change in expression for miR-
7-5p, miR-204-5p and miR-375. These miRNA were greatly upregulated in SBNET but
were hardly expressed at all in adjacent normal small bowel tissue and would therefore
be good candidates for a possible future biomarkers for use in SBNET.
The qPCR experiments also confirmed the profiling results for candidate miRNA, miR-
215-5p, miR-451a, miR-378a-3p and miR-378i which had reduced expression in SBNET
versus adjacent normal small bowel. 2 of the 3 candidate miRNA investigated for lymph
261
node metastases, miR-1 and miR-143-3p, were confirmed by qPCR as having reduced
expression in lymph node metastases relative to the SBNET tissue. These miRNA could
be implicated in disease progression.
A limitation of this study was the small number of liver metastasis samples which nar-
rowed the scope of the investigation of miRNA that could be linked to disease progression
to those dysregulated in lymph node metastases. This is addressed in the next chapter,
chapter 6, in which 13 liver metastasis samples were included to enable a more in depth
investigation of miRNA expression changes that might be involved in disease progression.
The candidate miRNA with the largest magnitude of changes in expression in SB-
NET, miR-7-5p, miR-204-5p, miR-375, miR-215-5p, miR-451a, miR-378a-3p, miR-378i,
miR-1 and miR-143-3p represent promising candidates for studies to develop future novel
biomarkers in SBNET. The expression of these miRNA in SBNET and their metastases
and the suitability of these miRNA as potential biomarkers for use in SBNET is investi-
gated further in the chapters that follow, chapter 6 and chapter 7.
262
6. Validation of the global miRNA
profiling in an independent group
of SBNET patients and the
identification of miRNA
dysregulated in liver metastases.
6.1. Introduction
Results presented in this chapter, chapter 6, were published in Endocrine Related Cancer
in 2016 (Miller et al., 2016).
In this chapter results are presented from the miRNA expression profiling of fresh frozen
tissue from the primary tumours and liver and lymph node metastases of SBNET patients
treated at Zentralklinik Bad Berka (Bad Berka, Germany). These results form dataset
2 (n=43). These experiments were done to validate the results of the global miRNA
expression profile of SBNET that were presented in the previous chapter, chapter 5, in
an independent group of SBNET patients that were treated at a different institution and
to investigate miRNA expression in SBNET liver metastases.
The results presented in this chapter are built upon in the final results chapter, chapter
263
7. In chapter 7, key predicted gene targets of the dysregulated miRNA are identified
using bioinformatic approaches in order to in order to determine the potential biological
functions of these miRNA.
MiRNA were quantified in fresh frozen tissue samples from 22 SBNET patients treated
at Zentralklinik Bad Berka. Liver metastasis tissue was available for miRNA quantifica-
tion from 13 of the SBNET patients. These samples were used to identify miRNA that
could be involved in metastatic growth and disease progression. In total, 800 miRNA
were quantified in 43 different tissue samples.
This was to determine if the findings of the first miRNA profiling experiment could
be validated in tissue from an independent set of SBNET patients, treated at a separate
institution. The purpose of this was to determine if the results were robust to differences
in SBNET patient populations, sample collection methods and storage conditions, neces-
sary qualities for a potential future prognostic biomarker. This would enable a SBNET
miRNA signature to be developed from the miRNA that were found to be reproducibly
dysregulated in SBNET which could provide the basis for further studies into SBNET
tumourigenesis and for the development of potential new SBNET biomarkers.
This chapter addresses the third and fourth research objectives of this thesis:
“3) Verify the reproducibility and robustness of the SBNET miRNA profile.”
“4) Identify miRNA associated with disease progression in SBNET.”
In order to fulfil the third research objective, global miRNA quantification was done
using the NanoString nCounter Human miRNA Expression Assay on SBNET tissue from
a separate population of SBNET patients treated at a different institution.
In order to fulfil the fourth research objective, global miRNA quantification was done
using the NanoString nCounter Human miRNA Expression Assay on liver and lymph
node metastasis tissue from SBNET patients.
264
6.1.1. Summary of results
Overall a total of 90 tissue samples were included across both dataset 1 (n=47) and
dataset 2 (n=43) in which 800 miRNA were quantified. This is by far the largest and
most thorough study of miRNA expression in SBNET to date. The results of the global
miRNA expression profile of SBNET presented in chapter 5 were validated in an indepen-
dent group of SBNET patients treated at a different institution. This suggests that the
results are reproducible and identified a 40 miRNA SBNET signature of miRNA that are
of particular interest for future study to investigate their potential function as oncomir
and tumour suppressor miRNA in SBNET. Novel miRNA were identified that were dys-
regulated in SBNET liver metastases and these miRNA could be involved in promoting
disease progression. These newly identified miRNA in SBNET and their metastases give
an important insight into the biological pathways that become disrupted in SBNET and
could be promising targets for the development of future prognostic biomarkers in this
disease.
6.2. SBNET patients
In order to validate the miRNA expression results from dataset 1 in an independent
group of SBNET patients, global miRNA quantification was done in tumour tissue from
22 SBNET patients treated at Zentralklinik Bad Berka (Bad Berka, Germany), see Table
6.1. Fresh frozen tumour tissue was available from 13 SBNET primary tumours, 15
lymph node metastases and 13 liver metastases. The larger number of liver metastasis
samples (13 samples) enabled the potential role of miRNA in progressive disease to be
investigated more thoroughly than was possible in dataset 1.
Two samples of “normal” small bowel tissue were included in the study and were
obtained at Imperial College Healthcare NHS Trust (London, UK). The samples were
from patients undergoing a normal right hemicolectomy procedure, patient consent was
265
given for the small bowel tissue that would be removed anyway during the course of a
normal right hemicolectomy to be used for research.
There were 43 fresh frozen tissue samples included in the study. For clinical details
and ethical approval see Methods, sections 3.3.1 and 3.1.
The median age of the SBNET patients was 65 (range: 54-88 years). The study
included a higher proportion of male patients, 73 % (16 male patients, 6 female patients).
Table 6.1.: Samples, dataset 2
Tissue type NumberPrimary SBNET 13Lymph node metastases 15Liver metastases 13Small bowel “normal” tissue 2
6.2.1. MiRNA expression, primary tumour
The miRNA expression of the SBNET primary tumour tissue was compared with that
of the “normal” small bowel tissue. Of the 800 miRNA that were quantified, there were
106 miRNA that were significantly dysregulated in the SBNET relative to the “normal”
small bowel tissue using a FDR of < 0.05 (Benjamini–Hochberg adjusted p value, see
Methods, section 3.3.3). There were 76 miRNA with a significant increase in expression
in the SBNET tissue and 30 miRNA with a significant decrease in expression in the
SBNET tissue. For a full list of these miRNA including p values and log2FC values see
Appendix, section D, Table D.2.
In order to generate a profile of dysregulated miRNA in SBNET to enable comparison of
these profiling results with those from dataset 1, miRNA were selected that had a log2FC
in expression of ≥ 1.5 or ≤ −1.5. Using this more stringent criteria there were 57 miRNA
that were significantly upregulated (FDR < 0.05, log2FC ≥ 1.5) in the SBNET compared
to the “normal” small bowel tissue and 2 miRNA that were significantly downregulated
(FDR < 0.05, log2FC ≤ −1.5). These miRNA are shown in Table 6.2.
266
Table 6.2.: SBNET miRNA profile, most dysregulated miRNA
Upregulated miRNA (dataset 2) Downregulated miRNA (dataset 2)miRNA log2FC FDR miRNA log2FC FDRmiR-489 3.3 2.6165E-05 miR-3180 -3.2 2.42985E-08miR-137 3.3 2.23998E-05 miR-31-5p* -1.5 0.000622114miR-375 3.1 0.000154715miR-95 2.9 5.55994E-05miR-7-5p 2.7 0.001217902miR-301a-3p 2.6 0.0002353miR-204-5p 2.5 0.003106112miR-642a-5p 2.5 0.00068917miR-129-2-3p 2.4 0.003903987miR-181c-5p 2.3 0.000712133miR-183-5p* 2.2 0.011899449miR-107 2.2 0.002905876miR-26a-5p 2.2 0.003106112miR-148b-3p 2.1 0.003106112miR-34a-5p 2.1 0.003566272miR-454-3p 2.1 0.001168798miR-98 2.1 0.003106112miR-429 2.1 0.008470798miR-1206 2.0 0.017439505miR-129-5p* 2.0 0.012327891miR-660-5p 2.0 0.005681993miR-582-5p 2.0 0.003173223miR-551b-3p 2.0 0.017439505miR-96-5p* 2.0 0.019746543miR-182-5p* 2.0 0.017184072miR-340-5p 2.0 0.003173223miR-200a-3p* 2.0 0.011945207miR-128 1.9 0.003173223miR-342-3p 1.9 0.013795817miR-374b-5p 1.9 0.003903987miR-324-5p 1.9 0.004989686miR-505-3p 1.9 0.003135006miR-30c-5p 1.9 0.005875075miR-29c-3p 1.9 0.009462954miR-99b-5p 1.9 0.004989686miR-4284 1.9 0.003710895let-7f-5p 1.9 0.012158394miR-362-3p 1.9 0.00649524miR-132-3p 1.8 0.0123323miR-135a-5p 1.8 0.039768879
267
Continuation of Table 6.2
Upregulated miRNA (dataset 2) Downregulated miRNA (dataset 2)miRNA log2FC FDR miRNA log2FC FDRmiR-196a-5p* 1.8 0.038155946miR-30b-5p 1.8 0.021083682let-7i-5p 1.8 0.015664337miR-421 1.8 0.00068917miR-27b-3p 1.7 0.021083682miR-24-3p 1.7 0.011899449miR-23b-3p 1.6 0.035290968miR-181b-5p+181d
1.6 0.003106112
miR-16-5p 1.6 0.039458926miR-361-5p 1.6 0.003106112miR-1180 1.6 0.019746543miR-664-3p 1.6 0.00305366miR-22-3p 1.5 0.015547491miR-335-5p 1.5 0.00494862miR-615-3p 1.5 0.009462954let-7c 1.5 0.041537624miR-1468 1.5 0.021277504
MiRNA had a FDR < 0.05 and a log2FC ≥ 1.5 or ≤ −1.5, *:miRNA also identified by Li et al. (2013b).
A limitation of these findings is that they are based on data from the miRNA expression
of only a small number of “normal” small bowel samples (n=2) compared to the SBNET
samples (n=13). It would have been advantageous to have had a larger number of fresh
frozen “normal” small bowel samples for comparison with the SBNET samples however
this was not possible due to cancellations on the day of surgery and time restraints.
Initially there were 4 potential patients that were identified who were undergoing suitable
procedures and patient consent was given for the small bowel tissue that would be removed
anyway during the course of a “normal” right hemicolectomy procedure to be used for
research. Unfortunately it was only possible to obtain samples from 2 of these patients.
The primary purpose of this part of the study was to generate a second dataset of
results from an independent group of SBNET patients treated at a separate institution
268
(dataset 2) that could be used to determine if it was possible to validate the results of
the SBNET miRNA profile identified in the previous chapter of this thesis, chapter 5
(dataset 1). This is considered in the next section, section 6.2.2.
6.2.2. SBNET miRNA profile validation
The results from dataset 2 (section 6.2.1) were compared to the miRNA expression profile
of a SBNET generated in the previous chapter, chapter 5, section 5.2 (dataset 1). Only
miRNA with a significant change in expression (FDR of < 0.05) and a log2FC in expres-
sion of ≥ 1.5 or ≤ −1.5 were considered for the comparison. This corresponds to a 3 fold
increase or decrease in expression (see Methods, section 3.3.3). This stringent cut off was
designed to exclude miRNA where there were only small fold changes in expression in
SBNET. Although statistically significant, the impact of the change in miRNA expression
is likely to be negligible since the magnitude of the change is so small that it is unlikely to
represent a true biological effect. The miRNA with a log2FC: ≥ 1.5 or ≤ −1.5 conversely
represent the most promising candidates for future study and the development of future
biomarkers.
Using this approach, there were 63 miRNA which were significantly dysregulated in
SBNET in dataset 1 and 59 miRNA that were significantly dysregulated in SBNET in
dataset 2 (log2FC: ≥ 1.5 or ≤ −1.5, FDR: < 0.05). There were 40 of these miRNA that
were dysregulated in SBNET in both dataset 1 and dataset 2, see Table 6.3 and Figure 6.1.
These findings show a high degree of overlap between dataset 1, SBNET patients treated
at Imperial College Healthcare NHS Trust and dataset 2, SBNET patients treated at
Zentralklinik Bad Berka, particularly for those miRNA that were upregulated in SBNET.
Table 6.3.: MiRNA dysregulated in SBNET
Upregulated miRNA (dataset 2) Downregulated miRNA (dataset 2)miRNA log2FC FDR miRNA log2FC FDRmiR-489 3.3 2.6165E-05 miR-3180 -3.2 2.42985E-08
269
Continuation of Table 6.3
Upregulated miRNA (dataset 2) Downregulated miRNA (dataset 2)miRNA log2FC FDR miRNA log2FC FDRmiR-137 3.3 2.23998E-05 miR-31-
5p*-1.5 0.000622114
miR-375$ 3.1 0.000154715miR-95 2.9 5.55994E-05miR-7-5p$ 2.7 0.001217902miR-301a-3p 2.6 0.0002353miR-204-5p$ 2.5 0.003106112miR-642a-5p 2.5 0.00068917miR-129-2-3p 2.4 0.003903987miR-181c-5p 2.3 0.000712133miR-183-5p* 2.2 0.011899449miR-107 2.2 0.002905876miR-26a-5p 2.2 0.003106112miR-148b-3p 2.1 0.003106112miR-34a-5p 2.1 0.003566272miR-454-3p 2.1 0.001168798miR-98 2.1 0.003106112miR-429 2.1 0.008470798miR-1206 2.0 0.017439505miR-129-5p* 2.0 0.012327891miR-660-5p 2.0 0.005681993miR-582-5p 2.0 0.003173223miR-551b-3p 2.0 0.017439505miR-96-5p* 2.0 0.019746543miR-182-5p* 2.0 0.017184072miR-340-5p 2.0 0.003173223miR-200a-3p* 2.0 0.011945207miR-128 1.9 0.003173223miR-342-3p 1.9 0.013795817miR-374b-5p 1.9 0.003903987miR-324-5p 1.9 0.004989686miR-505-3p 1.9 0.003135006miR-30c-5p 1.9 0.005875075miR-29c-3p 1.9 0.009462954miR-99b-5p 1.9 0.004989686miR-4284 1.9 0.003710895let-7f-5p 1.9 0.012158394miR-362-3p 1.9 0.00649524miR-132-3p 1.8 0.0123323miR-135a-5p 1.8 0.039768879
270
Continuation of Table 6.3
Upregulated miRNA (dataset 2) Downregulated miRNA (dataset 2)miRNA log2FC FDR miRNA log2FC FDRmiR-196a-5p* 1.8 0.038155946miR-30b-5p 1.8 0.021083682let-7i-5p 1.8 0.015664337miR-421 1.8 0.00068917miR-27b-3p 1.7 0.021083682miR-24-3p 1.7 0.011899449miR-23b-3p 1.6 0.035290968miR-181b-5p+181d
1.6 0.003106112
miR-16-5p 1.6 0.039458926miR-361-5p 1.6 0.003106112miR-1180 1.6 0.019746543miR-664-3p 1.6 0.00305366miR-22-3p 1.5 0.015547491miR-335-5p 1.5 0.00494862miR-615-3p 1.5 0.009462954let-7c 1.5 0.041537624miR-1468 1.5 0.021277504
MiRNA had a FDR < 0.05 and a log2FC ≥ 1.5 or ≤ −1.5, boldfont: miRNA also identified in dataset 1 (same cut off), $:candidate miRNA dataset 1 (qPCR), *: miRNA also identified byLi et al. (2013b).
This suggests that the results of the SBNET miRNA expression profile identified re-
mains robust despite the differences between the two populations of SBNET patients
being investigated and differences in sample preservation method (fresh frozen/FFPE)
and sample handling. The 40 miRNA signature of a primary tumour identified contains
miRNA that may be acting as oncomir or tumour suppressor miRNA in SBNET and
thus contributing to tumourigenesis. These miRNA represent promising candidates for
future work to better understand SBNET tumourigenesis and for development as poten-
tial new prognostic biomarkers in SBNET, however there are some limitations with this
comparison.
271
Figure 6.1.: A: Venn diagram showing miRNA that were significantly increased in SBNETrelative to “normal” small bowel tissue. B: Venn diagram showing miRNAthat were significantly decreased in SBNET relative to “normal” small boweltissue. All miRNA had a FDR < 0.05 and a log2FC ≥ 1.5 or ≤ −1.5. D1:dataset 1, D2: dataset 2.
272
Limitations
Although within each miRNA expression profiling study the tissue preservation method
was the same for each sample, different methods of tissue preservation were used for the
tissue samples from Zentralklinik Bad Berka (dataset 2) to those from Imperial College
Healthcare NHS Trust (dataset 1). The miRNA quantified in dataset 2 were extracted
from fresh frozen tissue samples whereas the miRNA quantified in dataset 1 were ex-
tracted from FFPE tissue samples. Ideally the two studies would have used tissue sam-
ples treated with the same tissue preservation method to make sure that this factor did
not affect the results of the comparison between datasets 1 and 2. Unfortunately this was
not possible since fresh frozen tissue was not available from the SBNET patients treated
at Imperial College Healthcare NHS Trust, only FFPE tissue and conversely only fresh
frozen tissue was available from the SBNET patients treated at Zentralklinik Bad Berka
and not FFPE tissue.
The difference in tissue preservation method might not have had too large an effect
on the results since miRNA remain stable even in FFPE tissue due to their secondary
structure and even remain stable after multiple freeze thaw cycles (unlike long mRNA
molecules which degrade readily, see Literature review, section 2.6.2). It is however
possible that certain miRNA may be more susceptible to differences in tissue preservation
methods and this could have affected the results. Further studies designed to enable a
direct comparison between miRNA expression in frozen or FFPE in tissue from the same
SBNET patients would be needed to determine the extent of the effect of this factor on
miRNA expression. The results of this would be of interest for future SBNET miRNA
studies since FFPE tissue is much more readily available in tumour tissue archives than
fresh frozen tissue.
The “normal” small bowel tissue used for comparison with the SBNET primary tumour
tissue in order to generate the SBNET miRNA expression profile was different in the two
different studies. The “normal” small bowel comparison group for dataset 1 was adjacent
273
normal small bowel tissue from the 15 patients with SBNET, this tissue was available
from 12 of the SBNET patients. Ideally adjacent normal small bowel tissue would have
been used for comparison in the study of SBNET patients treated at Zentralklinik Bad
Berka (dataset 2) however adjacent normal small bowel tissue samples were not available
from these patients, only tumour tissue samples. MiRNA were quantified in all available
samples from the 22 SBNET patients treated at Zentralklinik Bad Berka and for the 15
SBNET patients treated at Imperial College Healthcare NHS Trust.
In order to enable the miRNA expression in the SBNET from Zentralklinik Bad Berka
to be compared to that in normal small bowel for the identification of miRNA that were
dysregulated during tumourigenesis, fresh frozen samples of “normal” small bowel were
obtained at Imperial College Healthcare NHS Trust. The samples were from patients
undergoing a normal right hemicolectomy procedure (patient consent was given for the
small bowel tissue that would be removed anyway during the course of a normal right
hemicolectomy procedure to be used for research).
This difference in the “normal” small bowel comparison tissue used in dataset 1 and
dataset 2 could have affected the results. This is because cells in the stroma surrounding a
tumour although initially tumour suppressing can change over time and begin to promote
disease progression (Bremnes et al., 2011). There may have been fewer tumour associated
stromal changes in the “normal” small bowel tissue from the right hemicolectomy patients
than in the adjacent “normal” small bowel tissue from the SBNET patients which may
have had an effect on miRNA expression. Differences in sample collection and in sample
preservation could also have affected the miRNA expression in the samples.
Candidate miRNA
In the previous chapter, chapter 5, section 5.3, 7 candidate miRNA were identified as
potential future biomarkers in SBNET. A further 2 candidate miRNA, miR-1 and miR-
143-3p, were identified as potential future biomarkers of metastatic disease, these are
274
discussed in a later section, section 6.3.3.
MiR-7-5p, miR-204-5p, miR-375, miR-215-5p, miR-378i, miR-378a-3p and miR-451a
were confirmed as being dysregulated in SBNET relative to adjacent normal small bowel
by two different miRNA quantification methods, the NanoString nCounter human miRNA
expression assay and qPCR. MiR-7-5p, miR-204-5p and miR-375 were found to be signif-
icantly upregulated in SBNET while miR-215-5p, miR-378i, miR-378a-3p and miR-451a
were found to be significantly downregulated in SBNET. These candidate miRNA were
investigated in dataset 2.
The candidate miRNA that were significantly upregulated in SBNET using two dif-
ferent miRNA quantification methods in dataset 1, miR-7-5p, miR-204-5p and miR-375,
were also found to be significantly upregulated in dataset 2, Table 6.4, Figure 6.1. All of
the miRNA had a log2FC of at least 2.5 and they were amongst the top 7 most upregu-
lated miRNA in SBNET in dataset 2, Table 6.3. These findings suggest that miR-7-5p,
miR-204-5p and miR-375 do indeed represent promising candidate biomarkers for further
study in SBNET, with reproducible results in two independent populations of SBNET
patients.
Table 6.4.: Dataset 2 profiling results for candidate miRNA
miRNA log2FC FDRmiR-375 3.1 0.000154715miR-7-5p 2.7 0.001217902miR-204-5p 2.5 0.003106112miR-451a -0.4 0.664579109miR-378a-3p+378i -0.7 0.437906242miR-215-5p -1.4 0.095478296
In contrast, the candidate miRNA that were found to be significantly downregulated in
SBNET by two different miRNA quantification methods in the SBNET samples in dataset
1, miR-215-5p, miR-378i, miR-378a-3p and miR-451a were not significantly downregu-
lated in the SBNET samples in dataset 2, Table 6.4. This suggests that although these
miRNA could be quantified as being downregulated in SBNET in the dataset 1 samples
275
by 2 different methods of miRNA quantification, the results were not reproducible in an
independent group of SBNET patients.
This could be due to differences in sample collection or handling or differences in
the samples themselves or the different patient groups (see limitations above). Further
studies would be needed to determine the reason why these miRNA were not significantly
downregulated in the SBNET samples in dataset 2.
There was only one miRNA that was found to be significantly downregulated in SBNET
relative to the “normal” small bowel tissue in both datasets 1 and 2 this was miR-
31-5p, Figure 6.1. MiR-31-5p therefore represents the best downregulated candidate
miRNA for further studies into possible tumour suppressor miRNA in SBNET since it was
downregulated in both datasets. Only 2 miRNA in total were significantly downregulated
in SBNET in dataset 2 and met the log2FC cut off criteria (log2FC: ≤ −1.5) so there
could only have been a maximum of 2 miRNA in common between datasets 1 and 2 with
respect to downregulated miRNA (see Figure 6.1 B).
6.2.3. MiRNA signature of SBNET
In order to identify miRNA that could have an important role in the cancer biology
of SBNET across all different disease stages, a comparison was made of the miRNA
expression in tumour/metastatic tissue versus the respective “normal” tissue for each
site. Only miRNA with a significant change in expression in tumour tissue (FDR: < 0.05)
and a log2FC of ≥ 1.5 or ≤ −1.5 were considered for the analysis. The miRNA that were
dysregulated in SBNET (relative to “normal” small bowel tissue) in both datasets 1 and
2 were included in the analysis (the intersection of each of the Venn diagrams in Figure
6.1). This resulted in 40 miRNA being selected for the primary tumour versus “normal”
tissue group (log2FC: ≥ 1.5 or ≤ −1.5, FDR: < 0.05). These miRNA were compared to
the miRNA that were dysregulated in lymph node metastases versus normal lymph node
tissue and to those dysregulated in liver metastases versus adjacent normal liver tissue,
276
these results were available from dataset 1 only.
There was quite a high degree of overlap in the miRNA that were upregulated in
SBNET and their metastases compared to their respective “normal” tissues. Overall,
there were 29 miRNA in common, with increased expression in both SBNET and their
lymph node and liver metastases relative to their respective “normal” tissues, Figure 6.2.
When the same comparison was done for miRNA that were downregulated in tumour
and metastatic tissue relative to “normal” tissue however, there was no overlap, see
Figure 6.3. This was not surprising since none of the miRNA had a large enough log2FC
to satisfy the cut off criteria for the lymph node metastases versus normal lymph node
group and only one miRNA satisfied the cut off for the SBNET versus “normal” small
bowel group.
6.2.4. Summary
This confirms the SBNET miRNA profile identified in chapter 5 in an independent set
of SBNET patients treated at a different institution suggesting that the results were re-
producible particularly for those miRNA that were upregulated in SBNET. A 40 miRNA
signature of SBNET was identified made up of miRNA that were dysregulated with large
magnitude changes in expression in both dataset 1 and dataset 2. These miRNA would be
promising candidates for future research into tumourigenesis in SBNET and for the devel-
opment of future molecular biomarkers. Furthermore, 29 of these miRNA had increased
expression in both local and distant metastases relative to “normal” tissues as well as
being upregulated in the primary tumour. These miRNA appear to have a potential role
in the tumour biology of SBNET across all different disease stages.
The 3 candidate miRNA that were confirmed as upregulated in SBNET by two sepa-
rate quantification methods in chapter 5, miR-7-5p, miR-204-5p and miR-375, were also
significantly upregulated in SBNET in dataset 2 and would be particularly promising
candidates for future studies to determine if they can be detected in the serum of SB-
277
Figure 6.2.: Venn diagram showing the miRNA with increased expression in tumour tis-sue relative to normal tissue. a) Small bowel primary (SBP)/ small bowel“normal”(SB N), comprised of the intersection of dataset 1 (D1) and dataset2 (D2), see Figure 6.1. b) Lymph node metastases(LNM)/ lymph node nor-mal tissue (LN N) c) Liver metastases(LVM)/ Liver adjacent normal tissue(LV N). All miRNA had a FDR < 0.05 and a log2FC ≥ 1.5 or ≤ −1.5.
278
Figure 6.3.: Venn diagram showing the miRNA with reduced expression in tumour tissuerelative to normal tissue. a) Small bowel primary (SBP)/ small bowel “nor-mal”(SB N), comprised of the intersection of dataset 1 (D1) and dataset 2(D2), see Figure 6.1. b) Lymph node metastases(LNM)/ lymph node nor-mal tissue (LN N) c) Liver metastases(LVM)/ Liver adjacent normal tissue(LV N). All miRNA had a FDR < 0.05 and a log2FC ≥ 1.5 or ≤ −1.5.
279
NET patients. They could be used as potential future biomarkers in SBNET if studies
showed that they were able to stratify SBNET patients into clinically useful subgroups
or enable the monitoring of factors such as tumour burden.
6.3. MiRNA implicated in metastatic disease
6.3.1. Liver metastases
There were 13 fresh frozen SBNET liver metastasis samples included in the study. The
global miRNA expression of the liver metastases was determined to identify miRNA that
could be involved in disease progression.
There were 60 miRNA with a significant change in expression in the liver metasta-
sis samples compared to the primary tumour samples (FDR: < 0.05). The expression
of 32 miRNA was significantly increased in the liver metastases while 28 miRNA had
significantly decreased expression in the liver metastases, these are shown in Table 6.5.
The dysregulation in the expression of these miRNA could be involved in promoting
metastatic growth in SBNET liver metastases.
A log2FC cut off of ≥ 1.5 or ≤ −1.5 was applied to select only those miRNA with a
larger magnitude of change in expression. There were 12 miRNA with a significant change
in expression in liver metastases relative to SBNET using this log2FC cut off, these are
shown in Table 6.6). These miRNA could be useful as potential future biomarkers of
progressive disease, particularly if their expression could be detected in serum, enabling
a non invasive liquid biopsy approach to be used (see Literature review, section 2.6.2).
Further work would be needed to determine if these miRNA could provide clinically
useful information, for example if serum levels of these miRNA could be used for the
early detection of liver micrometastases.
280
6.3.2. Lymph node metastases
There were 15 lymph node metastasis samples included in the study. The global miRNA
expression of the lymph node metastases was determined in order to identify miRNA
that could be involved in promoting locoregional spread of SBNET.
There were 25 miRNA with a significant change in expression in the lymph node metas-
tasis samples compared to the primary tumour samples (FDR: < 0.05). The expression
of 6 miRNA was significantly increased in the lymph node metastases while 19 miRNA
had significantly decreased expression in the lymph node metastases, these are shown
in Table 6.7. The dysregulation in the expression of these miRNA could be involved in
promoting local metastatic growth.
A log2FC cut off of ≥ 1.5 or ≤ −1.5 was applied to select miRNA with a larger
magnitude change in expression. Using this criteria, there were 10 miRNA with a change
in expression in lymph node metastases relative to SBNET (see Table 6.8). These miRNA
could be particularly important for maintaining and promoting the growth of lymph node
metastases. Some of these miRNA are the same as those identified in the liver metastases
suggesting that they could have a role in metastatic growth and disease progression in
both local and distant metastases. This is investigated further in the next section, section
6.3.3.
These results from the patients treated at Zentralklinik Bad Berka (dataset 2) were
compared to those from the miRNA previously identified as being dysregulated in the
lymph node metastasis (relative to the SBNET tissue) in the samples from patients
treated at Imperial College Healthcare NHS Trust (dataset 1, chapter 5, section 5.2.2,
Table 5.5). There were 4 miRNA that were found to have reduced expression in the lymph
node metastases of both independent groups of SBNET patients. These were miR-133a,
miR-1, miR-145-5p and miR-1233. These miRNA would be of particular interest for
future study to determine their possible function as tumour suppressor miRNA in lymph
node metastases of SBNET patients.
281
6.3.3. Disease progression
There were 13 liver metastasis and 15 lymph node metastasis samples included in the
tissue from SBNET patients treated at Zentralklinik Bad Berka. This meant that it
was possible to identify possible patterns in the relative expression of the miRNA during
disease progression. Analysis was carried out to identify if miRNA that were reduced in
lymph node metastases relative to the SBNET were even further suppressed in the liver
metastasis samples relative to the SBNET. This might suggest that these miRNA could
have a protective role in suppressing disease progression or suppressing metastatic growth
(tumour suppressor miRNA, see Literature review, section 2.5.2), with a reduction in the
expression of this miRNA leading to metastases in SBNET patients.
In order to identify miRNA that might be involved in promoting or suppressing tumour
progression, miRNA that had a significant change in expression in SBNET metastases
relative to the primary tumour were selected (FDR: < 0.05). There were 16 miRNA with
a significant increase/decrease in expression in both liver and in lymph node metastatic
tissue relative to the primary tumour. These miRNA are shown in Table 6.9.
As well as certain miRNA being dysregulated in both liver and lymph node metastases
some miRNA were only dysregulated in either liver or lymph node metastases. These are
indicated in the Venn diagram in Figure 6.4. These miRNA may have a role in promoting
metastatic growth that is tissue specific.
MiRNA that were significantly dysregulated in liver and lymph node metastases with
a log2FC of ≥ 1.5 or ≤ −1.5 relative to primary tumour tissue were identified in earlier
sections (sections 6.3.1 and 6.3.2). These miRNA are shown in the heatmap in Figure
6.5. Using this log2FC cut off, there were 7 miRNA in common between the lymph node
metastases and the liver metastases. All 7 miRNA had reduced expression in both the
liver metastases and in the lymph node metastases relative to the primary tumour.
The heatmap, Figure 6.5, shows that for 6/7 miRNA, miR-1, miR-143-3p, miR-145-
5p, miR-139-3p, miR-139-5p and miR-1233, the relative expression of these miRNA was
282
Figure 6.4.: Venn diagram showing all significantly dysregulated miRNA in lymph nodemetastases (LNM) and/or liver metastases (LVM) relative to expression inthe primary tumour (SBP) (FDR: < 0.05). Italic text indicates miRNA thathad higher expression levels in metastatic tissue relative to the SBP (all othermiRNA had lower expression in the metastatic tissue.)
283
Figure 6.5.: Heatmap showing the miRNA that had significantly decreased/increased ex-pression in metastatic tissue, lymph node metastases (LNM) or liver metas-tases (LVM), relative their expression in small bowel primary tumours (SBP).Log2FC values are shown for each miRNA. A log2FC of cut off of ≥ 1.5 or≤ −1.5 was used (FDR of < 0.05). *: expression of these miRNA weresignificantly reduced in LNM and LVM however the log2FC values for theLVM were not of a high enough magnitude to meet the ≤ −1.5 cut off, thesevalues were nevertheless included to enable comparison with the values forLNM/SBP. Blank spaces indicate that there was no significant change in theexpression of that particular miRNA.
284
reduced further in the liver metastasis than in the lymph node metastasis (relative to
expression in the primary tumour). This suggests that these miRNA might be acting
as tumour suppressor miRNA with a role in suppressing disease progression, since their
expression decreases during tumour progression from the primary tumour, to the local
metastases to the liver metastases. Amongst these miRNA were miR-1 and miR-143-3p
which were both identified as candidate miRNA in chapter 5, (section 5.3.2).
If future work was able to show that these differences in miRNA expression also changed
the circulating levels of these miRNA, with serum levels of changing with increased tu-
mour burden or disease progression, then these miRNA could potentially be used in the
early identification of disease progression. Further studies would be needed to deter-
mine if this was the case and controlled clinical trials to determine the clinical utility of
any future biomarker, to identify liver metastases early or predict future metastases for
example.
Interestingly, miR-133a was equally reduced in both lymph node and liver metastases,
relative to the SBNET, with a log2FC of -2.9 (see Figure 6.5). This suggests that while
a reduction in miR-133a levels may also have a role in promoting metastatic growth it
appears to be equally downregulated in both local and distant metastases compared to
the primary tumour. It may therefore be less useful as a predictor of disease progression
from lymph node to liver metastases. MiR-133a might be more useful as a potential
future biomarker in identifying patients with local metastases although the majority of
SBNET patients have lymph node metastases at presentation, as demonstrated in chapter
4, where 89 % of the patients with G1 SBNET and all of the patients with G2 SBNET
had metastatic disease (chapter 4, section 4.3.1).
The levels of miR-133a were measured in the serum of SBNET patients in a study by
Li et al (Li et al., 2015) and were significantly reduced in the serum from SBNET patients
compared to healthy control serum (see Literature review, section 2.5.3). This suggests
that miRNA expression results identified originally in tissue samples (Li et al., 2013b)
285
may also hold true in serum, which suggests that miRNA identified in this way could be
used as potential as future biomarkers in a liquid biopsy setting if further work was able
to identify their clinical utility. Of particular interest would be miRNA that showed a
change in expression in serum with respect changes in tumour burden, invasiveness or
disease progression.
MiR-494, miR-144-3p, miR-885-5p, miR-451a and miR-122-5p were all significantly
reduced in the liver metastases but had no significant change of expression in the lymph
node metastases relative to the primary tumour samples, Figure 6.5 (FDR: < 0.05,
log2FC: ≥ 1.5 or ≤ −1.5). This could suggest that the reduction in the expression
of these miRNA could have a particular role in promoting metastatic growth that is
specific to liver metastases. These miRNA could be good candidates to investigate in
future studies both to better understand the disease pathology of liver metastases and
for studies of circulating miRNA to see if they might have the potential for the early
identification of the presence of liver micrometastases in SBNET patients.
6.3.4. Summary
The availability of 13 liver metastasis samples from the SBNET patients treated at Zen-
tralklinik Bad Berka enabled a global analysis of miRNA expression in liver metastases
to be carried out. This revealed 60 miRNA that were significantly dysregulated in liver
metastases and could therefore be involved in promoting and maintaining metastatic
growth. The majority of these miRNA have never been previously identified as dys-
regulated in SBNET liver metastases. The miRNA with the largest magnitude changes
in expression would be promising candidates for future studies to determine their gene
targets and to elucidate their function in SBNET liver metastases. They would also be
promising candidates for potential future biomarkers for the identification of patients
with more aggressive SBNET.
Of particular interest as future biomarkers were miR-1, miR-143-3p, miR-145-5p, miR-
286
139-3p, miR-139-5p and miR-1233 since the expression of these miRNA was decreased in
the lymph node metastases and was even further decreased in the liver metastases.
These miRNA are promising candidates for use as potential future prognostic biomark-
ers in SBNET patients, since their expression decreases with tumour progression. Further
studies would be needed to validate these results and determine if these miRNA are able
to stratify patients with low grade SBNET into clinically useful subgroups based on
clinical and pathological behaviour, for example to indicate a subgroup of patients with
more aggressive tumours. If these results were confirmed in circulating miRNA from
SBNET serum samples, these miRNA could potentially be used the prediction or early
identification of disease progression in SBNET patients.
6.4. Conclusions
This chapter has fulfilled the third and fourth research objectives of this thesis by verify-
ing the reproducibility and robustness of the SBNET miRNA profile and by identifying
miRNA that are associated with disease progression in SBNET. The global miRNA ex-
pression levels were assessed in tissue from 22 SBNET patients treated at Zentralklinik
Bad Berka.
Global miRNA expression profiling was done using the NanoString nCounter human
miRNA expression assay. This confirmed that it was possible to reproduce the results of
the global miRNA expression profile of SBNET presented in chapter 5 in an independent
group of SBNET patients treated at a separate institution.
Of particular interest for future work was the 40 miRNA signature identified for SB-
NET. This primary tumour miRNA signature consisted of significantly dysregulated
miRNA with large changes in expression between the SBNET and the “normal” small
bowel tissue in both profiling experiments. These large, reproducible changes in miRNA
expression are likely to represent biologically important changes that occur during SB-
287
NET tumourigenesis. These miRNA therefore represent promising candidates for the
development novel SBNET biomarkers and for future work to investigate the biological
pathways that become disrupted in SBNET.
Moreover, 29 of these miRNA were upregulated not only in SBNET but also in lymph
node and liver metastases relative to their respective “normal” tissues. These 29 miRNA
would be of interest for future studies since their change in expression across all SBNET
tumour stages compared to “normal” tissue suggests that they may be necessary for the
growth and survival of primary tumours and metastases in patients with SBNET. Further
studies would be needed to determine if this was the case.
All 3 candidate miRNA selected in chapter 5 that were upregulated in SBNET, miR-
7-5p, miR-204-5p and miR-375, had large, significant increases in expression in SBNET
versus “normal” small bowel in the second profiling experiment. The 4 candidate miRNA
that were downregulated in SBNET were not found to be significantly dysregulated in
the second profiling experiment, these candidate miRNA were therefore excluded from
the bioinformatics analysis carried out in the next chapter, chapter 7.
Global miRNA quantification was done in the liver and lymph node metastasis sam-
ples in order to identify miRNA that were associated with disease progression in SB-
NET. MiRNA quantification in the 13 liver metastasis samples identified 60 miRNA that
were significantly dysregulated in the liver metastases compared to the primary tumour.
Novel miRNA were identified that had not been previously associated with SBNET liver
metastases. These miRNA could be involved in promoting metastatic growth and disease
progression.
Of particular interest was the pattern of expression of miR-1, miR-143-3p, miR-145-5p,
miR-139-3p, miR-139-5p and miR-1233. The expression of these miRNA was reduced in
lymph node metastases and then even further reduced in liver metastases. This pattern of
expression would be particularly useful for a future SBNET prognostic biomarker if future
studies were able to demonstrate the clinical utility of these miRNA for the prediction or
288
early identification of disease progression for example.
Interestingly miR-133a was equally reduced in both lymph node and liver metastases
and may therefore be of less interest as a marker of disease progression from local to
distant metastases.
In the chapter that follows, chapter 7, candidate miRNA miR-7-5p, miR-204-5p, miR-
375, miR-1 and miR-143-3p are taken forwards for bioinformatics analysis to select the
most promising potential miRNA biomarkers for use in SBNET and to identify miRNA-
mRNA interactions of interest for future studies.
289
Table 6.5.: Significantly dysregulated miRNA in liver metastases
Upregulated miRNA Downregulated miRNAmiRNA log2FC FDR miRNA log2FC FDRmiR-122-5p 3.8 1.25147E-12 miR-1 -3.1 3.51478E-11miR-451a 2.2 6.36872E-05 miR-133a -2.9 2.9984E-08miR-885-5p 2.1 9.49406E-05 miR-145-5p -2.5 3.10117E-07miR-144-3p 2.0 0.000331623 miR-143-3p -2.5 3.69016E-07miR-494 1.8 0.000940358 miR-139-5p -2.5 1.83097E-07miR-96-5p 1.3 0.04148076 miR-139-3p -2.4 3.08973E-07miR-1206 1.3 0.026470936 miR-1233 -2.2 1.02826E-06miR-182-5p 1.3 0.027506518 miR-490-3p -1.4 0.00989529miR-142-3p 1.1 0.012357616 miR-378a-
3p+378i-1.3 0.010565391
miR-424-5p 1.1 0.00487345 miR-28-3p -1.3 1.79589E-05miR-20a-5p+20b-5p
1.0 0.010169799 miR-28-5p -1.3 5.8881E-05
miR-663a 1.0 0.018091806 miR-1246 -1.2 0.049233296miR-548g-3p 1.0 0.015926894 miR-125b-5p -1.2 0.034241259miR-93-5p 0.9 0.009974824 miR-195-5p -1.1 0.002755741miR-219-5p 0.9 0.01553825 miR-378g -1.1 0.022064011miR-106b-5p 0.9 0.014061011 miR-30a-5p -1.1 0.016794119miR-499a-5p 0.9 0.018386742 miR-663b -1.1 0.041360633miR-455-5p 0.8 0.039185572 miR-214-3p -1.0 0.020777689miR-329 0.8 0.038924891 miR-21-5p -1.0 0.015926894miR-25-3p 0.8 0.001531826 miR-497-5p -0.9 0.022064011miR-219-1-3p 0.7 0.021569968 miR-27a-3p -0.9 0.004274135miR-548ae 0.7 0.022064011 miR-130a-3p -0.8 0.04829487miR-3184-5p 0.7 0.018386742 miR-296-5p -0.8 0.015926894miR-19b-3p 0.7 0.039185572 miR-656 -0.7 0.035891647miR-507 0.6 0.022064011 miR-940 -0.7 0.036612026miR-550b-3p 0.6 0.021569968 miR-331-5p -0.7 0.034241259miR-449c-5p 0.6 0.002755741 miR-1281 -0.7 0.033340791miR-302b-3p 0.6 0.027506518 miR-1825 -0.6 0.041456837miR-130b-3p 0.6 0.030035211miR-618 0.6 0.015926894miR-1181 0.5 0.022064011miR-191-5p 0.5 0.031320004
290
Table 6.6.: Liver metastases, most dysregulated miRNA
Upregulated miRNA Downregulated miRNAmiRNA log2FC FDR miRNA log2FC FDRmiR-122-5p 3.8 1.25147E-12 miR-1 -3.1 3.51478E-11miR-451a 2.2 6.36872E-05 miR-133a -2.9 2.9984E-08miR-885-5p 2.1 9.49406E-05 miR-145-5p -2.5 3.10117E-07miR-144-3p 2.0 0.000331623 miR-143-3p -2.5 3.69016E-07miR-494 1.8 0.000940358 miR-139-5p -2.5 1.83097E-07
miR-139-3p -2.4 3.08973E-07miR-1233 -2.2 1.02826E-06
Table 6.7.: Significantly dysregulated miRNA in Lymph node metastases
Upregulated miRNA Downregulated miRNAmiRNA log2FC FDR miRNA log2FC FDRmiR-15b-5p 0.7 0.01427061 miR-133a -2.9 4.66852E-13miR-330-5p 0.7 0.044509549 miR-1 -2.9 3.90001E-14miR-455-3p 0.7 0.048124664 miR-143-3p -2.2 8.03288E-09miR-455-5p 0.6 0.048124664 miR-145-5p -2.1 1.835E-08miR-764 0.6 0.046526091 miR-139-3p -2.0 9.71775E-08miR-191-5p 0.5 0.044509549 miR-139-5p -1.9 4.40343E-07
miR-1233 -1.9 9.03129E-07miR-378a-3p+378i
-1.7 2.30357E-05
miR-187-3p -1.7 2.42623E-06miR-378g -1.5 0.000105601miR-10a-5p -1.3 0.001434782miR-30a-5p -1.1 0.001434782miR-9-5p -1.1 0.044509549miR-28-5p -1.0 0.000747229miR-574-5p -0.9 0.01427061miR-28-3p -0.9 0.001434782miR-152 -0.9 0.029448112miR-331-5p -0.7 0.01427061miR-1825 -0.5 0.048124664
291
Table 6.8.: Lymph node metastases, most dysregulated miRNA
miRNA log2FC FDRmiR-133a -2.9 4.66852E-13miR-1 -2.9 3.90001E-14miR-143-3p -2.2 8.03288E-09miR-145-5p -2.1 1.835E-08miR-139-3p -2.0 9.71775E-08miR-139-5p -1.9 4.40343E-07miR-1233 -1.9 9.03129E-07miR-378a-3p+378i -1.7 2.30357E-05miR-187-3p -1.7 2.42623E-06miR-378g -1.5 0.000105601
Table 6.9.: MiRNA dysregulated in both liver and lymph node metastases
Lymph node metastases Liver metastaseslog2FC log2FC
miR-133a -2.9 -2.9miR-1 -2.9 -3.1miR-143-3p -2.2 -2.5miR-145-5p -2.1 -2.5miR-139-3p -2.0 -2.4miR-139-5p -1.9 -2.5miR-1233 -1.9 -2.2miR-378a-3p+378i -1.7 -1.3miR-378g -1.5 -1.1miR-30a-5p -1.1 -1.1miR-28-5p -1.0 -1.3miR-28-3p -0.9 -1.3miR-331-5p -0.7 -0.7miR-1825 -0.5 -0.6miR-191-5p 0.5 0.5miR-455-5p 0.6 0.8
292
7. Bioinformatics
7.1. Introduction
Results presented in chapter 7, were published in Endocrine Related Cancer in 2016
(Miller et al., 2016).
In this chapter results are presented from a bioinformatics study to identify predicted
gene targets of dysregulated miRNA in SBNET to determine the potential function of
these miRNA in the tumours of SBNET patients. This builds upon the work of results
chapters 5 and 6 in which candidate miRNA, miR-7-5p, miR-204-5p, miR-375, miR-1 and
miR-143-3p, were identified as potential future biomarkers. These miRNA were found
to be dysregulated in SBNET and their metastases both in the SBNET patients treated
at Imperial College Healthcare NHS Trust and those treated at Zentralklinik Bad Berka
and their change in expression was confirmed by qPCR.
A bioinformatics approach was used to identify the predicted gene targets of each of the
candidate miRNA, miR-7-5p, miR-204-5p, miR-375, miR-1 and miR-143-3p. These gene
targets were compared against 4 publicly available gene expression databases containing
genes (mRNA) that were previously identified as being dysregulated in SBNET patients.
This was to determine if dysregulation in expression of the candidate miRNA in SBNET
could be correlated with opposing changes in the mRNA levels in SBNET of the predicted
gene targets of the candidate miRNA.
The bioinformatics was carried out to identify the gene targets of the candidate miRNA
293
that might be of particular importance in the disease pathology of SBNET. This was in
order to indicate the possible function of the candidate miRNA in SBNET and the biolog-
ical pathways they might be regulating (see Literature review, sections 2.5.1 and 2.5.2).
This was done to select particularly promising predicted miRNA-mRNA interactions for
future experimental investigation in SBNET cell lines.
This chapter addresses the final research objective of this thesis:
“5) Identify the most promising potential miRNA biomarkers for use in SB-
NET.”
In order to fulfill this objective bioinformatics approaches were used to predict the
potential miRNA-mRNA interactions that were most likely to be important for SBNET
tumourigenesis and therefore the most promising candidates for potential future biomark-
ers. This was broken down into the following items:
1 - Identify the predicted gene targets of each individual candidate miRNA
to predict potential miRNA-mRNA interactions.
The candidate miRNA were identified experimentally in results chapters 5 and 6 as
being dysregulated in SBNET. A bioinformatics approach was used to predict the gene
(mRNA) targets of these miRNA in order to identify potential miRNA-mRNA interac-
tions.
2 - Identify experimentally dysregulated mRNA in the tissue of SBNET pa-
tients using publicly available gene expression datasets.
Some of the mRNA that were found to be upregulated or downregulated in SBNET
patients are likely to be miRNA targets and changes in their levels of expression could
be due to gene silencing by the candidate miRNA (see Literature review, section 2.5.1).
294
3 - Select mRNA that are both predicted targets of the candidate miRNA
(experimentally dysregulated in SBNET patients) and have also been demon-
strated experimentally to be dysregulated themselves in SBNET patients.
This was done in order to select predicted miRNA-mRNA interactions that were most
likely to be biologically important for SBNET tumourigenesis and disease progression.
This would enable the most promising predicted miRNA-mRNA interactions to be se-
lected for further in vitro work to experimentally confirm these interactions.
4 - Use these mRNA (selected above) to carry out gene ontology and pathway
enrichment analysis to identify biological processes that could be important
for disease pathology in SBNET patients.
This was done to identify the most promising miRNA-mRNA interactions and bio-
logical pathways for future experimental studies in SBNET to investigate the function
of these miRNA in SBNET patients and to develop novel biomarkers and for patient
stratification.
7.1.1. Summary of results
Novel predicted miRNA-mRNA interactions were identified that could play an important
role in SBNET tumourigenesis and disease progression. Experimental data strength-
ened the bioinformatics approach enabling the identification of dysregulated genes in
SBNET that were also predicted targets of the dysregulated candidate miRNA in SB-
NET (datasets 1 and 2). MiRNA-mRNA interactions of particular interest in SBNET
were identified for future in vitro studies and biomarker development. Gene ontology
analysis revealed enriched gene ontology terms related to apoptosis and cell death and
identified important oncogenes that are overexpressed in the absence of miR-1 and miR-
143-3p negative regulation in SBNET metastases. This reduction in oncogene silencing
could be contributing to disease progression in SBNET.
295
7.2. Candidate miRNA and gene expression datasets
In order to determine miRNA-mRNA interactions that might be of particular importance
in SBNET tumourigenesis and to identify the most promising biomarkers for further work
bioinformatics analysis was carried out on the candidate miRNA identified in the preced-
ing two chapters, (chapter 5 and chapter 6. For the full methods of the bioinformatics
study and the study design see Methods, section 3.4 and Figure 3.2.
The candidate miRNA investigated by bioinformatics were miR-7-5p, miR-204-5p,
miR-375, miR-1 and miR-143-3p. MiR-7-5p, miR-204-5p and miR-375. These miRNA
were upregulated in SBNET relative to the small bowel “normal” tissue in datasets 1 and
2. MiR-1 and miR-143-3p were downregulated in lymph node metastases relative to the
SBNET in datasets 1 and 2. MiR-1 and miR-143-3p were also downregulated in the liver
metastasis samples relative to the SBNET.
These candidate miRNA were selected due to being found to be significantly dysreg-
ulated in the tumour tissue, with a high magnitude change in expression in both the
SBNET patients treated at Imperial College Healthcare NHS Trust and those treated at
Zentralklinik Bad Berka (datasets 1, dataset 2) by two different miRNA quantification
techniques (NanoString nCounter miRNA Expression Assay, qPCR).
Bioinformatics approaches (TargetScan) were used to predict gene (mRNA) targets of
each of the candidate miRNA to identify genes that the miRNA could be regulating by
gene silencing (see Methods, section 3.4.1).
In order to narrow down the list of predicted gene targets of each candidate miRNA
to those that were most likely to have an important function in SBNET, the predicted
gene targets of each miRNA were compared to experimental data on mRNA expression
in the same tissue types that the experimental miRNA results were obtained in.
Publicly available gene expression datasets containing the appropriate tissue types
were identified and the mRNA expression data was analysed for use in the bioinformatics
study. Details of the 4 available expression datasets are that were included in the study are
296
shown in Table 7.1. For full methodological details and NCBI GEO/ EBI ArrayExpress
identification numbers, see Methods, section 3.4.2.
Table 7.1.: Gene expression datasets for bioinformatics
Gene expression data Com-parisongroups
No. of samples Dysregulated genesfor bioinformatics
SBP SB N LNM No.of
genes
Gene expression
dataset a, Edfeldtet al. (2011)
LN-M/SBP
18 17 4787 upregulated inLNM
dataset b, Kiddet al. (2014) SBP/SB N
9 3 368 downregulatedin SBP
dataset c, Leja et al.(2009) and Kiddet al. (2014)
SBP/SB N3 3 605 downregulated
in SBP
dataset d, Leja et al.(2009) SBP/SB N
3 3 4230 downregulatedin SBP
The bioinformatically predicted gene targets of each candidate miRNA (TargetScan)
were compared against the mRNA that were dysregulated in SBNET (gene expression
datasets). This was done in order to select the most promising biomarkers for future
study in SBNET including the most promising predicted miRNA-mRNA interactions for
future in vitro verification of gene silencing. The results of this comparison are shown in
the next section, section 7.3.
Table 7.2 shows the number of predicted mRNA targets (TargetScan) for each of
the candidate miRNA (experimentally dysregulated in datasets 1-2) and the number of
mRNA that were identified in the publicly available gene expression datasets (experimen-
tally dysregulated in datasets a-d). Lists of the predicted gene targets for each miRNA
of interest were compared with the mRNA that were found to be upregulated in SBNET
(for the downregulated miRNA) or mRNA found to be downregulated in SBNET (for
the upregulated miRNA). For details on the mechanisms of gene expression regulation
by gene silencing and the role of miRNA in cancer as oncomir and tumour suppressor
297
miRNA see the Literature review, sections 2.5.1 and 2.5.2.
Table 7.2.: Potential gene targets of the candidate miRNA
MiRNA expression No. predicted Gene expression No.gene targets genes
Decreased LNM/SBP Increased LNM/SBPmiR-1 3115 dataset a 4787miR-143-3p 3287
Increased SBP/SB N Decreased SBP/SB NmiR-7-5p 3574 dataset b 368miR-204-5p 3985 dataset c 605miR-375 2267 dataset d 4230
7.2.1. Limitations
Bioinformatics was used in order to narrow down the large number of possible miRNA-
mRNA interactions to those that were most likely to be important in SBNET tumourigen-
esis so that these could be the focus of future experimental work. Due to the large number
of possible miRNA-mRNA interactions it would not have been practical or economical
to test all of these experimentally.
There limitations with bioinformatics approaches since these rely on predicting biologi-
cal interactions, in this case predicting that a particular miRNA will silence the expression
of particular genes (TargetScan). The algorithms used by TargetScan are based on com-
plementary base pairing and known biological ’rules’ about the structure and length of
sequences that have been shown experimentally to be required for miRNA binding and
for subsequent silencing to occur (for more details see Methods, section 3.4.1 and Lit-
erature review, section 2.5.1). Despite this, to ensure that any bioinformatics findings
represent a true biological interaction that is important in the setting of SBNET, fu-
ture in vitro experiments in cell lines would need to be carried out. These experiments
would experimentally validate the most promising predicted miRNA-mRNA interactions
(Luciferase assay) and could confirm if overexpression of the miRNA of interest directly
298
triggers reduced expression of a particular mRNA in vitro. Western blot experiments
could then be used to confirm if reduced mRNA expression is also matched by reduced
protein levels of the gene of interest.
In order to mitigate against the risk that the bioinformatics results might not be
reproducible in experimental studies and to ensure that any particular miRNA-mRNA
interactions identified were likely to be gene silencing events occurring in SBNET tumour
tissue, experimental gene expression data was used in the bioinformatics study. This was
done to strengthen the bioinformatics approach by only including genes in the study if
they had been found to be experimentally dysregulated in the same tissue types that the
candidate miRNA were dysregulated in. To narrow this down further to possible gene
silencing events, miRNA that were upregulated were only compared to genes that were
downregulated and vice versa for miRNA that were downregulated.
7.3. Comparison of gene lists
In order to identify genes that might be being silenced by the candidate miRNA in
SBNET, genes that were found experimentally to be upregulated in the tissues of inter-
est (dataset a) were compared to the gene targets of the candidate miRNA that were
found experimentally to be downregulated in the tissues of interest (datasets 1, 2). Con-
versely, genes that were found experimentally to be downregulated in the tissues of inter-
est (datasets b, c, d) were compared to the gene targets of candidate miRNA that were
found experimentally to be upregulated in the tissues of interest (datasets 1, 2). The
datasets for this comparison are described in the previous section, section 7.2 and shown
in Table 7.2.
299
7.3.1. SBNET
The predicted gene targets of miR-7-5p, miR-204-5p and miR-375 (upregulated in SB-
NET versus “normal” small bowel tissue) were compared to the genes that were signifi-
cantly downregulated (at the mRNA level) in SBNET versus “normal” small bowel tissue
(datasets b-d).
There were 19 genes in common for miR-7-5p, 23 genes for miR-204-5p and 14 genes for
miR-375. These genes were downregulated in SBNET versus “normal” small bowel tissue
in all 3 gene expression datasets and were also predicted gene targets of the candidate
miRNA, Table 7.3.
Interestingly there were quite a few dysregulated genes in SBNET that were predicted
to be regulated by more than one of the candidate miRNA (see Table 7.3). There
were 4 downregulated genes that were predicted gene targets of all 3 of the candidate
miRNA, miR-7-5p, miR-204-5p and miR-375. These genes were FZD5, ACOX1, PTER
and SLC31A2. These genes might be of particular importance in SBNET and represent
promising targets for future experimental study of miRNA-mRNA interactions since the
bioinformatics results seem to suggest a redundancy in the negative regulation of these
genes by multiple miRNA that are upregulated in SBNET tissue. This might suggest
that the suppression of these genes is particularly important for SBNET tumourigene-
sis, however experimental studies would be needed to determine if this was indeed the
case. This would include “in vitro” studies to confirm these predicted miRNA-mRNA
interactions and functional studies to investigate any effects of gene silencing.
FZD5 encodes a seven transmembrane domain protein that is a receptor for Wnt pro-
teins and was previously found to be upregulated in renal cell carcinoma and pancreatic
ductal adenocarcinoma and found in a cell line study to be required for cellular prolif-
eration (Listing et al., 2015; Ueno et al., 2013; Steinhart et al., 2017). The opposite
pattern was found in SBNET with the expression of FZD5 being reproducibly downreg-
ulated in SBNET versus “normal” small bowel tissue in all 3 of the publicly available
300
Table 7.3.: Predicted gene targets of the candidate miRNA that were dysregulated in all3 gene expression datasets (b, c, d)
Gene targets of Gene targets of Gene targets ofmiR-7 ∩ b ∩ c ∩ d miR-204 ∩ b ∩ c ∩ d miR-375 ∩ b ∩ c ∩ d
FZD5 FZD5 FZD5ACOX1 ACOX1 ACOX1PTER PTER PTER
SLC31A2 SLC31A2 SLC31A2GK PCK1 CEBPG
SLC1A1 HSD17B2 SLC46A3MAOA SLC1A1 NR5A2
RETSAT ASS1 DFNA5TMPRSS2 ZG16 LPGAT1
ACO2 CYBRD1 SLC31A1LPGAT1 SLC46A3 ERBB2
EVI2B RETSAT FDX1MTUS1 NR5A2 CASP7
TGFA SDC1 HNMTCASP7 TMPRSS2 GAREMGALE FGL2 RMDN3
PPARGC1A PTP4A1LDHA SUCLG2
MICALL1 MTUS1ADTRP FDX1
DNMBPPPARGC1A
P4HBGAREM
bold font: genes targeted by more than 1 candidate miRNA.
gene expression datasets. This suggests that there may be something different occurring
with respect to Wnt-FZD5 signalling in SBNET than has been observed for renal cell
carcinoma and pancreatic ductal adenocarcinoma. This might not perhaps be surprising
given that SBNET patients with low grade tumours have low proliferation levels, so in
this case overexpression of miR-7-5p, miR-204-5p and miR-375 in SBNET could be pro-
tective. This would need to be investigated further in SBNET cell line studies in order to
determine the potential effects of gene silencing by miR-7-5p, miR-204-5p and miR-375
on FZD5 expression and any impact on cellular proliferation.
301
ACOX1 encodes an enzyme involved in the metabolism of fatty acids, increased expres-
sion of ACOX1 has been associated with the HER2 breast cancer subtype (Kim et al.,
2015a). There have been relatively few studies of SLC31A2 which is thought to encode
a copper transporter involved in copper uptake in intracellular organelles, in contrast to
the better characterised copper transporter SLC31A1 which is primarily localised on the
plasma membrane (Wee et al., 2013). PTER encodes a hydrolase enzyme that hydrolyses
esters and is thought to be proinflammatory (Cheng et al., 2014). Further work would be
needed to experimentally confirm that miR-7-5p, miR-204-5p and miR-375 interact with
FZD5, ACOX1, PTER and SLC31A2 in vitro and investigate any phenotype changes as
a result of gene silencing.
Genes were identified that were downregulated in 2 or more of the gene expression
datasets and were predicted gene targets of the candidate miRNA, these genes were
taken forwards for gene ontology and pathway analysis (see section 7.4).
7.3.2. Lymph node metastases
The predicted gene targets of miR-1 and miR-143-3p (downregulated in lymph node
metastases versus SBNET tissue) were compared to the genes that were significantly
upregulated in lymph node metastases versus SBNET tissue (dataset a).
There were 805 genes in common for miR-1, and 904 genes in common for miR-143-3p.
These genes were upregulated in lymph node metastases versus SBNET in the available
gene expression dataset containing lymph node metastasis tissue (dataset a) and were
also predicted gene targets of the candidate miRNA. There were 278 upregulated genes in
lymph node metastases that were predicted gene targets of both miR-1 and miR-143-3p
(for a full list see Appendix, section E.1).
Genes that were upregulated in gene expression dataset and were also predicted gene
targets of the candidate miRNA were taken forwards for gene ontology and pathway
analysis in section 7.4.
302
7.3.3. Summary
This work revealed novel predicted miRNA-mRNA interactions in SBNET and lymph
node metastases which could be important for SBNET tumourigenesis. The bioinformat-
ics approach was strengthened by using experimental data to identify mRNA that were
predicted miRNA targets and were also dysregulated in SBNET. These results suggest
that gene silencing is likely to be occuring during tumourigenesis in SBNET patients with
miR-7-5p, miR-204-5p and miR-375 acting as oncomir and miR-1 and miR-143-3p acting
as tumour suppressor miRNA. Further experimental work would be needed to confirm
particular miRNA-mRNA interactions and to determine the functional significance of
these for disease pathology.
Of particular interest were the 4 mRNA, FZD5, ACOX1, PTER and SLC31A2, which
were downregulated in all 3 gene expression datasets and were predicted gene targets of
all 3 upregulated candidate miRNA, miR-7-5p, miR-204-5p and miR-375. These mRNA
may be of particular importance in SBNET since they are consistently experimentally
downregulated (at the mRNA level) in SBNET and are predicted targets of 3 different
miRNA that are consistently experimentally upregulated in SBNET (datasets 1 and 2).
7.4. Enriched gene ontology terms and pathways
Gene ontology and pathway analysis was carried out to determine which of the predicted
gene targets of each of the candidate miRNA might be the most important in SBNET tu-
mourigenesis and to identify potential functional/molecular pathway implications of these
predicted miRNA-mRNA interactions. This was done to narrow down the genes of inter-
est for future in vitro work to experimentally prove that a particular gene was targeted
by a particular miRNA (gene silencing) and to validate potential miRNA biomarkers.
The gene lists for the gene ontology analysis and pathway analysis were identified in
the previous section, for more details see Methods, sections 3.4.3 and 3.4.4.
303
7.4.1. SBNET
Bioinformatics approaches (DAVID) were used to identify over represented gene ontology
terms and molecular pathways amongst the genes identified for each candidate miRNA
(see Methods, section 3.4.4). The candidate miRNA were miR-7-5p, miR-204-5p and
miR-375.
There were a number of different enriched gene ontology terms and enriched pathways
for the dysregulated gene targets of miR-7-5p, miR-204-5p and miR-375 (see Table 7.4
and Appendix, section E.2, Table E.1). Enriched pathways included the MAPK signalling
pathway, various cellular metabolism pathways and gap junctions, Table 7.4. Unfortu-
nately none of the enriched gene ontology terms or the enriched pathways identified for
miR-7-5p, miR-204-5p and miR-375 reached statistical significance (FDR < 0.05).
304
Table 7.4.: Pathway analysis of downregulated genes in SBNET that are predicted targets of the upregulated candidate miRNA
Downregu-lated genetargets
KEGG pathway term GenesCount
%associatedwith thisterm
No. ofgenes ingene list
Popula-tionhits
Popula-tiontotal
Fold En-richment
P value FDR
miR-7 hsa00512: O-Glycanbiosynthesis
GALNT3, GALNT7,ST6GALNAC1
3 1.91083 63 30 5085 8.07143 0.05114 45.05863
miR-7 hsa00270: Cysteine andmethionine metabolism
LDHA, AHCYL2, CBS 3 1.91083 63 34 5085 7.12185 0.06395 52.95304
miR-7 hsa04514: Cell adhesionmolecules (CAMs)
F11R, CD8A, HLA-DMA,CLDN23, CLDN15
5 3.18471 63 132 5085 3.05736 0.07637 59.60066
miR-7 hsa04530: Tight junction F11R, NRAS, YES1, CLDN23,CLDN15
5 3.18471 63 134 5085 3.01173 0.07974 61.25089
miR-204 hsa00071: Fatty acidmetabolism
ACOX1, ADH5, ALDH3A2,ACOX3, ACSL5
5 3.22581 59 40 5085 10.77331 0.00103 1.12185
miR-204 hsa03320: PPAR signalingpathway
ACOX1, SLC27A2, PCK1,ACOX3, ACSL5
5 3.22581 59 69 5085 6.24539 0.00760 8.03973
miR-204 hsa00592: alpha-Linolenic acidmetabolism
ACOX1, PLA2G12B, ACOX3 3 1.93548 59 18 5085 14.36441 0.01740 17.54828
miR-204 hsa00010: Glycolysis /Gluconeogenesis
ALDOB, ADH5, ALDH3A2,PCK1
4 2.58065 59 60 5085 5.74576 0.03046 28.82578
miR-204 hsa04950: Maturity onsetdiabetes of the young
SLC2A2, HNF4G, NR5A2 3 1.93548 59 25 5085 10.34237 0.03243 30.39756
miR-204 hsa00020: Citrate cycle (TCAcycle)
SUCLG2, SDHD, PCK1 3 1.93548 59 31 5085 8.34062 0.04813 41.85607
miR-204 hsa04540: Gap junction CDK1, PLCB3, PDGFRA,LPAR1
4 2.58065 59 89 5085 3.87355 0.08040 60.20191
miR-204 hsa04010: MAPK signalingpathway
RPS6KA3, DUSP3, RELB,PLA2G12B, PDGFRA,PTPRR, GNG12
7 4.51613 59 267 5085 2.25957 0.08189 60.90412
miR-375 hsa00071: Fatty acidmetabolism
ACOX1, CPT1A, ACSL5 3 3.09278 38 40 5085 10.03618 0.03379 29.55523
miR-375 hsa00983: Drug metabolism XDH, NAT2, TPMT 3 3.09278 38 43 5085 9.33599 0.03859 33.04272miR-375 hsa00232: Caffeine metabolism XDH, NAT2 2 2.06186 38 7 5085 38.23308 0.04986 40.62672miR-375 hsa04920: Adipocytokine
signaling pathwayPRKAG2, CPT1A, ACSL5 3 3.09278 38 67 5085 5.99175 0.08484 59.48654
miR-375 hsa03320: PPAR signalingpathway
ACOX1, CPT1A, ACSL5 3 3.09278 38 69 5085 5.81808 0.08922 61.41987
miR-375 hsa00230: Purine metabolism XDH, GDA, AK2, NT5E 4 4.12371 38 153 5085 3.49845 0.09880 65.36166
305
7.4.2. Lymph node metastases
Bioinformatics approaches (DAVID) were used to identify over represented gene ontology
terms and molecular pathways amongst the gene lists identified for each candidate miRNA
(see Methods, section 3.4.4). The candidate miRNA were miR-1 and miR-143-3p.
miR-1
For miR-1 there were 3 significantly enriched gene ontology terms, from the gene ontology
analysis using a FDR of < 0.05 as a statistically significant result. All of the gene
ontology terms were related to the regulation of apoptosis and cell death (gene ontology
terms: GO:0042981, GO:0043067 and GO:0010941). There were 65 genes that were
associated with these gene ontology terms. All these genes had been previously found to
be upregulated in lymph node metastases compared to the SBNET (dataset a) and to be
predicted gene targets of miR-1 (TargetScan), which was downregulated in lymph node
and liver metastases (datasets 1 and 2). The significantly enriched gene ontology terms
for miR-1 and the genes associated with these terms are shown in Table 7.5.
306
Table 7.5.: Significantly enriched gene ontology terms for upregulated genes in lymph node metastases, predicted gene targetsof miR-1
Gene ontologyterm
Genes Count %associatedwith thisterm
No. ofgenes ingene list
Popu-lationhits
Popu-lationtotal
Foldenrich-ment
P Value FDR
GO:0042981 reg-ulation ofapoptosis
Seegene list
65 8.53018 605 804 135281.80774
4.26E-06 0.00757
GO:0043067 reg-ulation ofprogrammed celldeath
Seegene list
65 8.53018 605 812 135281.78993
5.76E-06 0.01025
GO:0010941 reg-ulation of celldeath
Seegene list
65 8.53018 605 815 135281.78334
6.51E-06 0.01159
Gene list (n=65): NUAK2, PREX1, RBM5, TLR4, NR2E1, KCNIP3, CUL3, ZFP91, BAG4, G2E3, NOD1, PAX7, CHST11,RARB, ALX4, NQO1, MKL1, EGFR, ARHGEF7, ARHGEF18, PIM1, ACTN2, PRKCE, STK4, STK3, BCL2L11, CARD10,
MAPK1, SERPINB9, TRIM35, TNFRSF10D, VEGFA, NAIP, NGFR, YWHAZ, BCLAF1, PRKDC, PLEKHG2, TUBB,KRAS, SH3GLB1, BCL2, BCL11B, HSPE1, INPP5D, BMF, RASA1, PHLDA1, STAMBP, B4GALT1, COL4A3, CFLAR,
IL2RB, CARD8, IL2RA, SPHK1, NR4A2, IGF1, HGF, ATP7A, NRAS, ATF5, CASP10, ETS1, BMP7
307
The genes associated with the significantly enriched gene ontology terms for miR-1 in-
cluded several members of the BCL-2 protein family, the oncogene BCL-2 (inhibits apop-
tosis) and two other BCL-2 protein family members, BCL2L11 and BCLAF1. Oncogene
KRAS, part of the RAS/MAPK signalling pathway, and oncogenes NUAK2 and FOSB
were associated with the significantly enriched gene ontology terms for miR-1 as well
as growth factors HGF and VEGFA. These genes had been found to be upregulated in
lymph node metastases relative to the SBNET in gene expression studies (dataset a) and
were also predicted gene targets of miR-1.
The results of the top 30 enriched gene ontology terms for miR-1 are shown in Table
7.6.
308
Table 7.6.: Top 30 enriched gene ontology terms for upregulated genes in lymph node metastases, predicted gene targets ofmiR-1
Gene ontology term Genes Count % associated
with this term
Length of
gene list
Population
hits
Population
total
Fold
enrichment
P Value FDR
GO:0042981 regulation of
apoptosis
NUAK2, PREX1, RBM5,
TLR4, NR2E1. . .
65 8.53018 605 804 13528 1.80774 4.26E-06 0.00757
GO:0043067 regulation of
programmed cell death
NUAK2, PREX1, RBM5,
TLR4, NR2E1. . .
65 8.53018 605 812 13528 1.78993 5.76E-06 0.01025
GO:0010941 regulation of
cell death
NUAK2, PREX1, RBM5,
TLR4, NR2E1. . .
65 8.53018 605 815 13528 1.78334 6.51E-06 0.01159
GO:0006793 phosphorus
metabolic process
CDK17, NUAK2,
NUAK1, PASK, SYNJ1...
67 8.79265 605 973 13528 1.53971 3.99E-04 0.70728
GO:0006796 phosphate
metabolic process
CDK17, NUAK2,
NUAK1, PASK, SYNJ1...
67 8.79265 605 973 13528 1.53971 3.99E-04 0.70728
GO:0012501 programmed
cell death
NUAK2, GULP1, PREX1,
RBM5, PRKDC. . .
45 5.90551 605 611 13528 1.64683 0.00119 2.10250
GO:0043066 negative
regulation of apoptosis
YWHAZ, NUAK2,
NR2E1, BAG4, ZFP91. . .
30 3.93701 605 354 13528 1.89494 0.00122 2.14450
GO:0045165 cell fate
commitment
ERBB4, PAX6, PRKDC,
PAX3, VSX2. . .
16 2.09974 605 139 13528 2.57385 0.00137 2.40268
GO:0043069 negative
regulation of programmed
cell death
YWHAZ, NUAK2,
NR2E1, BAG4, ZFP91. . .
30 3.93701 605 359 13528 1.86855 0.00149 2.61984
GO:0060548 negative
regulation of cell death
YWHAZ, NUAK2,
NR2E1, BAG4, ZFP91. . .
30 3.93701 605 360 13528 1.86336 0.00158 2.77239
GO:0006915 apoptosis NUAK2, GULP1, PREX1,
RBM5, RFFL. . .
44 5.77428 605 602 13528 1.63431 0.00159 2.79398
GO:0034613 cellular
protein localization
COPA, GRPEL2,
YWHAZ, HPS4, SNX2. . .
33 4.33071 605 411 13528 1.79536 0.00161 2.81883
GO:0008284 positive
regulation of cell
proliferation
NAMPT, ERBB4, IL6ST,
NAP1L1, PAX6. . .
33 4.33071 605 414 13528 1.78235 0.00181 3.17675
GO:0070727 cellular
macromolecule
localization
COPA, GRPEL2,
YWHAZ, HPS4, SNX2. . .
33 4.33071 605 414 13528 1.78235 0.00181 3.17675
GO:0035295 tube
development
B4GALT1, IGF1, CFTR,
PAX3, HECA. . .
21 2.75591 605 220 13528 2.13440 0.00204 3.57028
309
Continuation of Table 7.6
Gene ontology term Genes Count % associated
with this term
Length of
gene list
Population
hits
Population
total
Fold
enrichment
P Value FDR
GO:0016265 death MICB, NUAK2, PREX1,
RBM5, TBP. . .
50 6.56168 605 724 13528 1.54422 0.00232 4.04515
GO:0016310 phosphoryla-
tion
CDK17, NUAK2,
NUAK1, PASK, FXN. . .
54 7.08661 605 800 13528 1.50932 0.00247 4.30478
GO:0007169 transmem-
brane receptor protein
tyrosine kinase signaling
pathway
EGFR, MTSS1, MPZL1,
ERBB4, STAP1. . .
21 2.75591 605 224 13528 2.09628 0.00252 4.39360
GO:0042127 regulation of
cell proliferation
NAMPT, IL6ST, PTGS1,
RBM5, PAX6. . .
53 6.95538 605 787 13528 1.50584 0.00281 4.88939
GO:0008219 cell death MICB, NUAK2, PREX1,
RBM5, TBP. . .
49 6.43045 605 719 13528 1.52386 0.00336 5.82114
GO:0006357 regulation of
transcription from RNA
polymerase II promoter
ELF1, PAX6, PAX3,
NR2E1, KCNIP3. . .
49 6.43045 605 727 13528 1.50709 0.00420 7.20891
GO:0009792 embryonic
development ending in
birth or egg hatching
PAX6, PRKDC, PAX3,
SOX8, CUL3. . .
27 3.54331 605 334 13528 1.80757 0.00425 7.29583
GO:0010628 positive
regulation of gene
expression
GLIS3, ELF1, PAX6,
PRKDC, NUFIP1. . .
41 5.38058 605 581 13528 1.57792 0.00433 7.42853
GO:0045944 positive
regulation of transcription
from RNA polymerase II
promoter
GLIS3, ELF1, PAX6,
PRKDC, NUFIP1. . .
29 3.80577 605 371 13528 1.74784 0.00473 8.09518
GO:0006350 transcription ZNF451, TBP,
APOBEC3F, NR2E1,
PGR...
118 15.48556 605 2101 13528 1.25584 0.00540 9.19058
GO:0007167 enzyme
linked receptor protein
signaling pathway
FMOD, MTSS1, MPZL1,
ERBB4, IL6ST. . .
27 3.54331 605 342 13528 1.76529 0.00578 9.80516
GO:0008104 protein
localization
COPA, GRPEL2,
SLC15A2, RAB5C,
HPS4. . .
56 7.34908 605 882 13528 1.41970 0.00726 12.16733
310
Continuation of Table 7.6
Gene ontology term Genes Count % associated
with this term
Length of
gene list
Population
hits
Population
total
Fold
enrichment
P Value FDR
GO:0043588 skin
development
FRAS1, ATP7A, LEF1,
NGFR, COL5A2. . .
6 0.78740 605 29 13528 4.62628 0.00851 14.10317
GO:0006468 protein
amino acid
phosphorylation
STK16, CDK17, ERBB4,
NUAK2, NUAK1. . .
44 5.77428 605 667 13528 1.47504 0.00944 15.52678
GO:0043065 positive
regulation of apoptosis
BCLAF1, PREX1, RBM5,
PRKDC, TLR4. . .
31 4.06824 605 430 13528 1.61202 0.01050 17.12224
For unabridged gene lists see (Miller et al., 2016), Supplementary Table 4
311
The enriched KEGG pathways for the upregulated predicted gene targets of miR-1 are
shown in Table 7.7. These included KEGG pathways associated with cancer (hsa05200)
and various signalling pathways implicated in tumourigenesis including the VEGF signal-
ing pathway (promotes angiogenesis, hsa04370), the ErbB signalling pathway (promotes
cell proliferation and inhibits apoptosis, hsa04012) and the mTOR signalling pathway
(promotes proliferation, hsa04150). None of the enriched enriched pathways identified
for miR-1 reached statistical significance using a FDR of < 0.05.
312
Table 7.7.: Enriched KEGG pathway terms for upregulated genes in lymph node metastases, predicted gene targets of miR-1KEGG pathway term Genes Count % associated
with this term
No. of genes
in gene list
Population
hits
Population
total
Fold
Enrichment
P value FDR
hsa05200: Pathways in
cancer
TPM3, CDC42, KRAS,
BCL2, PIK3R5. . .
28 3.67454 237 328 5085 1.83158 0.00230 2.75292
hsa05223: Non-small cell
lung cancer
EGFR, NRAS, MAPK1,
RASSF5, KRAS. . .
9 1.18110 237 54 5085 3.57595 0.00312 3.71529
hsa04666: Fc gamma
R-mediated phagocytosis
ARPC1A, CDC42,
MAPK1, ARPC3,
PIKFYVE. . .
12 1.57480 237 95 5085 2.71019 0.00421 4.98673
hsa05211: Renal cell
carcinoma
CDC42, NRAS, MAPK1,
KRAS, ETS1. . .
10 1.31234 237 70 5085 3.06510 0.00469 5.53088
hsa04144: Endocytosis EGFR, STAMBP,
ARFGAP1, IL2RB,
IL2RA. . .
18 2.36220 237 184 5085 2.09893 0.00482 5.68398
hsa04510: Focal adhesion EGFR, COL4A4,
ITGA11, IGF1,
ACTN2. . .
18 2.36220 237 201 5085 1.92141 0.01145 13.02074
hsa04062: Chemokine
signaling pathway
ITK, GNAI3, PREX1,
NFKBIB, ADRBK2. . .
17 2.23097 237 187 5085 1.95052 0.01258 14.21961
hsa00760: Nicotinate and
nicotinamide metabolism
NAMPT, ENPP1,
NT5C2, NADK, PNP
5 0.65617 237 24 5085 4.46994 0.02311 24.66218
hsa04320: Dorso-ventral
axis formation
NOTCH3, EGFR,
MAPK1, KRAS, ETS1
5 0.65617 237 25 5085 4.29114 0.02654 27.80259
hsa04722: Neurotrophin
signaling pathway
CDC42, NRAS, MAPK1,
YWHAZ, KRAS. . .
12 1.57480 237 124 5085 2.07636 0.02818 29.26266
hsa04360: Axon guidance SEMA5A, CDC42, NRAS,
MAPK1, SEMA6A. . .
12 1.57480 237 129 5085 1.99588 0.03625 36.05893
hsa04960:
Aldosterone-regulated
sodium reabsorption
MAPK1, KRAS, ATP1B4,
IGF1, PIK3R5. . .
6 0.78740 237 41 5085 3.13986 0.03961 38.70720
hsa05210: Colorectal
cancer
EGFR, MAPK1, KRAS,
BCL2, LEF1. . .
9 1.18110 237 84 5085 2.29882 0.03995 38.97139
hsa05216: Thyroid cancer NRAS, MAPK1, KRAS,
LEF1, TPM3
5 0.65617 237 29 5085 3.69926 0.04316 41.40041
hsa04650: Natural killer
cell mediated cytotoxicity
NRAS, IFNAR2, MAPK1,
MICB, CD244. . .
12 1.57480 237 133 5085 1.93585 0.04378 41.85808
313
Continuation of Table 7.7
KEGG pathway term Genes Count % associated
with this term
No. of genes
in gene list
Population
hits
Population
total
Fold
Enrichment
P value FDR
hsa04914:
Progesterone-mediated
oocyte maturation
PGR, MAPK1, GNAI3,
KRAS, IGF1. . .
9 1.18110 237 86 5085 2.24536 0.04497 42.72507
hsa05218: Melanoma EGFR, NRAS, MAPK1,
KRAS, IGF1. . .
8 1.04987 237 71 5085 2.41754 0.04523 42.91843
hsa04670: Leukocyte
transendothelial migration
ITK, CDC42, RASSF5,
GNAI3, CXCR4. . .
11 1.44357 237 118 5085 2.00011 0.04630 43.68744
hsa04012: ErbB signaling
pathway
EGFR, NRAS, MAPK1,
KRAS, ERBB4. . .
9 1.18110 237 87 5085 2.21955 0.04762 44.62634
hsa05212: Pancreatic
cancer
EGFR, CDC42, MAPK1,
KRAS, VEGFA. . .
8 1.04987 237 72 5085 2.38397 0.04821 45.03861
hsa05221: Acute myeloid
leukemia
NRAS, MAPK1, KRAS,
FLT3, PIM1. . .
7 0.91864 237 58 5085 2.58948 0.05068 46.74382
hsa05215: Prostate cancer EGFR, NRAS, MAPK1,
KRAS, BCL2. . .
9 1.18110 237 89 5085 2.16968 0.05324 48.45653
hsa04070:
Phosphatidylinositol
signaling system
PIK3C2A, PIK3C2B,
SYNJ1, PIKFYVE,
PIK3R5. . .
8 1.04987 237 74 5085 2.31953 0.05454 49.30337
hsa04370: VEGF
signaling pathway
CDC42, NRAS, MAPK1,
KRAS, VEGFA. . .
8 1.04987 237 75 5085 2.28861 0.05789 51.43602
hsa04060:
Cytokine-cytokine
receptor interaction
EGFR, IL2RB, IL1R1,
IL2RA, FLT3. . .
19 2.49344 237 262 5085 1.55595 0.06003 52.75898
hsa04664: Fc epsilon RI
signaling pathway
NRAS, MAPK1, KRAS,
PIK3R5, MAPK10. . .
8 1.04987 237 78 5085 2.20058 0.06868 57.76426
hsa04010: MAPK
signaling pathway
EGFR, IL1R1, CACNG8,
TAOK3, MAPK10. . .
19 2.49344 237 267 5085 1.52681 0.06900 57.93836
hsa05214: Glioma EGFR, NRAS, MAPK1,
KRAS, IGF1. . .
7 0.91864 237 63 5085 2.38397 0.07048 58.73986
hsa04810: Regulation of
actin cytoskeleton
EGFR, SSH1, ARHGEF7,
ITGA11, ACTN2. . .
16 2.09974 237 215 5085 1.59670 0.07319 60.17589
hsa05213: Endometrial
cancer
EGFR, NRAS, MAPK1,
KRAS, LEF1. . .
6 0.78740 237 52 5085 2.47566 0.09169 68.80543
314
Continuation of Table 7.7
KEGG pathway term Genes Count % associated
with this term
No. of genes
in gene list
Population
hits
Population
total
Fold
Enrichment
P value FDR
hsa04150: mTOR
signaling pathway
MAPK1, VEGFA,
STRADA, IGF1,
PIK3R5. . .
6 0.78740 237 52 5085 2.47566 0.09169 68.80543
hsa05120: Epithelial cell
signaling in Helicobacter
pylori infection
EGFR, ATP6V1C1,
CDC42, ATP6V1A,
ADAM10. . .
7 0.91864 237 68 5085 2.20867 0.09415 69.81296
For unabridged gene lists see (Miller et al., 2016), Supplementary Table 4
315
miR-143
There were 3 significantly enriched gene ontology terms for miR-143 (FDR: < 0.05).
These were gene ontology terms related to the regulation of apoptosis and cell death
(GO:0042981, GO:0043067 and GO:0010941). These were the same gene ontology terms
that were significantly enriched in the gene ontology analysis for miR-1. These gene
ontology terms were associated with 67 genes in the input data for the DAVID bioinfor-
matics analysis. These genes were upregulated in lymph node metastases compared to
SBNET (dataset a) and were also predicted gene targets of miR-143 (TargetScan), which
was downregulated in lymph node and liver metastases of SBNET patients (datasets 1
and 2), see section 7.3. The significantly enriched gene ontology terms and the 67 genes
associated with them from the bioinformatics analysis are shown in Table 7.8.
316
Table 7.8.: Significantly enriched gene ontology terms for upregulated genes lymph node metastases, predicted gene targets ofmiR-143
Gene ontologyterm
Genes Count %associated
Lengthof
Popula-tion
Popula-tion
Fold P Value FDR
with thisterm
gene list hits total enrich-ment
GO:0042981 reg-ulation ofapoptosis
See gene list 67 7.86385 658 804 13528 1.71327 1.74E-05
0.03115
GO:0043067 reg-ulation ofprogrammed celldeath
See gene list 67 7.86385 658 812 13528 1.69639 2.34E-05
0.04192
GO:0010941 reg-ulation of celldeath
See gene list 67 7.86385 658 815 13528 1.69015 2.67E-05
0.04784
Gene list (n=67): MEF2C, IER3, NUAK2, SNCA, GDNF, CIAPIN1, MAP3K7, CUL3, ZFP91, G2E3, PROP1, CD44,TIAM1, PAX7, CHST11, CASP8, VNN1, NQO1, ALX4, CASP2, ALX3, RAB27A, EGFR, PIK3CG, ARHGEF7, TP53,ACTN2, DAPK2, PRKCE, NLRP1, MAPK1, SMO, TRIM35, PSEN1, TNFRSF10D, MAPK9, NGFR, UBA52, MCL1,
PLEKHG2, TUBB, KRAS, ERCC6, BCL2, DYRK2, STAMBP, B4GALT1, CFLAR, IL2RA, VAV3, MCF2, CREB1,YWHAB, IGF1, BIRC5, NFKBIL1, ATM, NRAS, BFAR, P2RX7, UACA, SARM1, CASP14, BBC3, NLRP12, APAF1,
BMP7
317
The three significantly enriched gene ontology terms were associated with various genes
that were upregulated in lymph node metastases relative to SBNET (dataset a) including
the oncogenes BCL-2, KRAS, NUAK2 and FOSB. These oncogenes had also been iden-
tified amongst the significantly enriched gene ontology terms for miR-1. In total there
were 26 genes associated with significantly enriched gene ontology terms in common for
miR-1 and miR-143 (for a full list of these genes see the Appendix, section E.3).
For the full gene ontology results for miR-143 including gene ontology terms that missed
statistical significance (FDR: < 0.05) see Table 7.9.
318
Table 7.9.: Enriched gene ontology terms for upregulated genes lymph node metastases, predicted gene targets of miR-143Gene ontology term Genes Count % associated
with this term
Length of
gene list
Population
hits
Population
total
Fold
enrichment
P Value FDR
GO:0042981 regulation of
apoptosis
MEF2C, IER3, NUAK2,
SNCA, GDNF. . .
67 7.86385 658 804 13528 1.71327 1.74E-05 0.03115
GO:0043067 regulation of
programmed cell death
MEF2C, IER3, NUAK2,
SNCA, GDNF. . .
67 7.86385 658 812 13528 1.69639 2.34E-05 0.04192
GO:0010941 regulation of
cell death
MEF2C, IER3, NUAK2,
SNCA, GDNF. . .
67 7.86385 658 815 13528 1.69015 2.67E-05 0.04784
GO:0006915 apoptosis MEF2C, STEAP3, IER3,
NUAK2, CIAPIN1. . .
52 6.10329 658 602 13528 1.77588 7.32E-05 0.13112
GO:0012501 programmed
cell death
MEF2C, STEAP3, IER3,
NUAK2, CIAPIN1. . .
52 6.10329 658 611 13528 1.74973 1.07E-04 0.19076
GO:0001932 regulation of
protein amino acid
phosphorylation
EGFR, NF2, ENPP1,
IL6ST, MAP4K1. . .
21 2.46479 658 173 13528 2.49563 2.80E-04 0.49987
GO:0008219 cell death MEF2C, STEAP3, IER3,
NUAK2, CIAPIN1. . .
57 6.69014 658 719 13528 1.62987 3.01E-04 0.53780
GO:0016265 death MEF2C, STEAP3, IER3,
NUAK2, CIAPIN1. . .
57 6.69014 658 724 13528 1.61862 3.52E-04 0.62927
GO:0032268 regulation of
cellular protein metabolic
process
NCBP1, METAP1,
ENPP1, IL6ST, SNCA. . .
41 4.81221 658 474 13528 1.77833 4.67E-04 0.83343
GO:0043403 skeletal
muscle regeneration
MTPN, PAX7, IGF1,
PLAU, PLAUR
5 0.58685 658 10 13528 10.27964 9.15E-04 1.62674
GO:0043066 negative
regulation of apoptosis
MEF2C, IER3, MCL1,
NUAK2, SNCA. . .
32 3.75587 658 354 13528 1.85847 0.00110 1.95591
GO:0000278 mitotic cell
cycle
HAUS6, DBF4, USP9X,
CUL3, TUBB. . .
33 3.87324 658 370 13528 1.83366 0.00113 2.01215
GO:0043069 negative
regulation of programmed
cell death
MEF2C, IER3, MCL1,
NUAK2, SNCA. . .
32 3.75587 658 359 13528 1.83258 0.00138 2.43991
GO:0006796 phosphate
metabolic process
PDP1, ATP6V0E1,
NUAK2, SNCA, EIF2A. . .
69 8.09859 658 973 13528 1.45795 0.00142 2.51574
GO:0006793 phosphorus
metabolic process
PDP1, ATP6V0E1,
NUAK2, SNCA, EIF2A. . .
69 8.09859 658 973 13528 1.45795 0.00142 2.51574
GO:0060548 negative
regulation of cell death
MEF2C, IER3, MCL1,
NUAK2, SNCA. . .
32 3.75587 658 360 13528 1.82749 0.00144 2.55341319
Continuation of Table 7.9
Gene ontology term Genes Count % associated
with this term
Length of
gene list
Population
hits
Population
total
Fold
enrichment
P Value FDR
GO:0050730 regulation of
peptidyl-tyrosine
phosphorylation
LIF, EGFR, ZFP91, NF2,
CD80. . .
11 1.29108 658 68 13528 3.32576 0.00155 2.73636
GO:0051240 positive
regulation of multicellular
organismal process
MAVS, EGFR, B4GALT1,
COL4A4, NOS1. . .
24 2.81690 658 244 13528 2.02222 0.00181 3.19835
GO:0043065 positive
regulation of apoptosis
CUL3, TUBB, PLEKHG2,
ERCC6, CD44. . .
36 4.22535 658 430 13528 1.72124 0.00194 3.42406
GO:0014910 regulation of
smooth muscle cell
migration
IL6ST, BCL2, IGF1,
TRIB1, IGFBP5
5 0.58685 658 12 13528 8.56636 0.00200 3.51486
GO:0043068 positive
regulation of programmed
cell death
CUL3, TUBB, PLEKHG2,
ERCC6, CD44. . .
36 4.22535 658 433 13528 1.70932 0.00216 3.79837
GO:0010942 positive
regulation of cell death
CUL3, TUBB, PLEKHG2,
ERCC6, CD44. . .
36 4.22535 658 435 13528 1.70146 0.00236 4.13752
GO:0031399 regulation of
protein modification
process
ENPP1, IL6ST, SNCA,
PAX5, MAP4K1. . .
27 3.16901 658 295 13528 1.88170 0.00247 4.32744
GO:0042327 positive
regulation of
phosphorylation
EGFR, IL6ST, IGF1,
RICTOR, LIF. . .
13 1.52582 658 97 13528 2.75537 0.00253 4.43155
GO:0010604 positive
regulation of
macromolecule metabolic
process
MEF2C, NCBP1,
CHURC1, IL6ST,
FAM175A. . .
61 7.15962 658 857 13528 1.46338 0.00259 4.53157
GO:0001817 regulation of
cytokine production
MAVS, TNFSF4, IL27RA,
IL6ST, CREB1. . .
19 2.23005 658 181 13528 2.15816 0.00310 5.40713
GO:0045937 positive
regulation of phosphate
metabolic process
EGFR, IL6ST, IGF1,
RICTOR, LIF. . .
13 1.52582 658 100 13528 2.67271 0.00326 5.68903
GO:0010562 positive
regulation of phosphorus
metabolic process
EGFR, IL6ST, IGF1,
RICTOR, LIF. . .
13 1.52582 658 100 13528 2.67271 0.00326 5.68903
320
Continuation of Table 7.9
Gene ontology term Genes Count % associated
with this term
Length of
gene list
Population
hits
Population
total
Fold
enrichment
P Value FDR
GO:0001775 cell
activation
SNCA, KLRK1, TLR6,
CBFB, ZFP91. . .
26 3.05164 658 287 13528 1.86251 0.00343 5.97173
GO:0045321 leukocyte
activation
EGR1, ADAM10,
TNFSF4, SNCA,
KLRK1. . .
23 2.69953 658 242 13528 1.95398 0.00351 6.10081
For unabridged gene lists see (Miller et al., 2016), Supplementary Table 4
321
There were various enriched KEGG pathways for the predicted gene targets of miR-143
that were also upregulated in lymph node metastases (dataset a), Table 7.10. These in-
cluded the p53 signalling pathway (hsa04115) and the ErbB signalling pathway (hsa04012,
also identified for miR-1). None of these enriched KEGG pathways identified for miR-143
achieved statistical significance (FDR: < 0.05).
322
Table 7.10.: Enriched KEGG pathway terms for upregulated genes in lymph node metastases, predicted gene targets miR-143KEGG pathway term Genes Count % associated
with this term
No. of genes
in gene list
Population
hits
Population
total
Fold
Enrichment
P value FDR
hsa05214:Glioma EGFR, PIK3CG,
MAPK1, NRAS, KRAS. . .
12 1.40845 279 63 5085 3.47158 5.16E-04 0.62160
hsa04660:T cell receptor
signaling pathway
PIK3CG, PDK1, VAV3,
CBL, CTLA4. . .
16 1.87793 279 108 5085 2.70012 6.96E-04 0.83839
hsa05215:Prostate cancer PIK3CG, EGFR, TCF7,
CREB1, TP53. . .
14 1.64319 279 89 5085 2.86698 9.74E-04 1.17053
hsa05218:Melanoma EGFR, PIK3CG,
MAPK1, NRAS, KRAS. . .
12 1.40845 279 71 5085 3.08042 0.00145 1.74144
hsa05219:Bladder cancer EGFR, MAPK1, NRAS,
KRAS, TP53. . .
9 1.05634 279 42 5085 3.90553 0.00167 2.00290
hsa05210:Colorectal
cancer
PIK3CG, EGFR, TCF7,
MSH3, TP53. . .
13 1.52582 279 84 5085 2.82066 0.00184 2.20723
hsa05200:Pathways in
cancer
WNT5B, MMP2, SUFU,
TPM3, KRAS. . .
30 3.52113 279 328 5085 1.66699 0.00626 7.31143
hsa05213:Endometrial
cancer
EGFR, PIK3CG,
MAPK1, NRAS, TCF7. . .
9 1.05634 279 52 5085 3.15447 0.00664 7.73776
hsa04012:ErbB signaling
pathway
EGFR, PIK3CG,
MAPK1, NRAS, KRAS. . .
12 1.40845 279 87 5085 2.51390 0.00731 8.49026
hsa05223:Non-small cell
lung cancer
EGFR, PIK3CG, MAPK1,
NRAS, RASSF5. . .
9 1.05634 279 54 5085 3.03763 0.00835 9.64430
hsa04115:p53 signaling
pathway
STEAP3, BBC3, RRM2,
CASP8, TP53. . .
10 1.17371 279 68 5085 2.68027 0.01102 12.53316
hsa04722:Neurotrophin
signaling pathway
PIK3CG, PDK1,
YWHAB, TP53, NRAS. . .
14 1.64319 279 124 5085 2.05775 0.01719 18.90426
hsa05216:Thyroid cancer MAPK1, NRAS, TCF7,
KRAS, TP53. . .
6 0.70423 279 29 5085 3.77086 0.01897 20.67109
hsa04510:Focal adhesion EGFR, COL4A4,
PIK3CG, VAV3,
DIAPH1. . .
19 2.23005 279 201 5085 1.72284 0.02546 26.78071
hsa04144:Endocytosis EGFR, STAMBP, IL2RA,
ERBB4, VTA1. . .
17 1.99531 279 184 5085 1.68391 0.04287 41.12134
hsa04810:Regulation of
actin cytoskeleton
EGFR, PIK3CG,
ARHGEF1, VAV3,
SSH1. . .
19 2.23005 279 215 5085 1.61065 0.04509 42.74601
323
Continuation of Table 7.10
KEGG pathway term Genes Count % associated
with this term
No. of genes
in gene list
Population
hits
Population
total
Fold
Enrichment
P value FDR
hsa04210:Apoptosis PIK3CG, CFLAR,
TNFRSF10D, BCL2,
IL1RAP. . .
10 1.17371 279 87 5085 2.09492 0.04680 43.97785
hsa04662:B cell receptor
signaling pathway
PIK3CG, MAPK1, NRAS,
VAV3, KRAS. . .
9 1.05634 279 75 5085 2.18710 0.05081 46.76171
hsa05220:Chronic myeloid
leukemia
PIK3CG, MAPK1, NRAS,
KRAS, CBL. . .
9 1.05634 279 75 5085 2.18710 0.05081 46.76171
hsa04664:Fc epsilon RI
signaling pathway
PDK1, PIK3CG, MAPK1,
NRAS, VAV3. . .
9 1.05634 279 78 5085 2.10298 0.06149 53.56179
hsa04512:ECM-receptor
interaction
COL4A4, CD36, CD44,
ITGA6, ITGA11. . .
9 1.05634 279 84 5085 1.95276 0.08677 66.61767
hsa05222:Small cell lung
cancer
PIK3CG, COL4A4,
ITGA6, BCL2, TP53. . .
9 1.05634 279 84 5085 1.95276 0.08677 66.61767
hsa03018:RNA
degradation
PATL1, DCP2, PNPT1,
TTC37, XRN1. . .
7 0.82160 279 57 5085 2.23826 0.08913 67.64646
hsa05212:Pancreatic
cancer
EGFR, PIK3CG,
MAPK1, KRAS, TP53. . .
8 0.93897 279 72 5085 2.02509 0.09675 70.76995
For unabridged gene lists see (Miller et al., 2016), Supplementary Table 4
324
7.4.3. Oncogene targets of downregulated miRNA in lymph
node metastases
Key oncogenes were identified as being associated with SBNET metastases in the signifi-
cantly enriched gene ontology terms from the DAVID analysis. The significantly enriched
gene ontology terms were the regulation of apoptosis and cell death (see section 7.4.2).
Oncogenes BCL-2, KRAS, FOSB, NUAK2, HGF and VEGFA were investigated further.
This was to determine if a reduction in the expression of miR-1 and miR-143 in SBNET
metastases was preventing the negative regulation of oncogene expression (gene silencing,
see Literature review, sections 2.5.1 and 2.5.2). If this was the case then this reduction in
the repression of these key oncogenes could be contributing to disease progression in SB-
NET. This would suggest that miR-1 and miR-143 may be acting as tumour suppressor
miRNA in SBNET.
BCL-2 and KRAS
Oncogenes BCL-2 and KRAS were identified in the gene ontology analysis for both miR-
1 and miR-143 (section 7.4.2). Analysis was done to determine if miR-1 and miR-143
could be tumour suppressor miRNA that regulate the mRNA levels of these oncogenes,
with this regulation being disrupted in SBNET metastases due to miR-1 and miR-143
expression being reduced (datasets 1 and 2).
The relative expression of BCL-2 and KRAS was significantly increased in lymph node
metastases of SBNET, Figure 7.1. The complementary base pairing between miR-1/miR-
143 and BCL-2 /KRAS is shown in Figure 7.1. For details on miRNA-mRNA binding
see Literature review section 2.5.1 and Methods, section 3.4.1.
NUAK2 and FOSB
Oncogenes NUAK2 and FOSB were identified in the gene ontology analysis that was
done for miR-1 and miR-143 (section 7.4.2). The relative expression of NUAK2 was
325
Figure 7.1.: Reduced expression of miR-1 and miR-143 (datasets 1 and 2) may lead to areduced negative regulation of the expression of the KRAS and BCL-2 onco-genes in lymph node metastases and therefore could be contributing to dis-ease progression. A: Complementary base pairing between miR-1/miR-143and KRAS mRNA, gene expression data showing a significant reduction inKRAS expression in lymph node metastases compared to SBNET (dataseta, GSE27162). B: Complementary base pairing between miR-1/miR-143and BCL-2 mRNA, gene expression data showing a significant reduction inBCL-2 expression in lymph node metastases compared to SBNET (dataset a,GSE27162).Error bars show the mean +/- standard deviation (* p < 0.05, **p < 0.01, *** p < 0.001). Reprinted by permission, ©[2016] [BioScientificaLtd.], (Endocrine-Related Cancer) (Miller et al., 2016).
326
significantly increased in lymph node but not in liver metastases, Figure 7.2. The relative
expression of FOSB was significantly increased in both lymph node and liver metastases
of SBNET, Figure 7.2. NUAK2 and FOSB are predicted gene targets of miR-1 and
miR-143 based on complementary base pairing between miR-1 and miR-143 and the
transcripts of these genes, Figure 7.2.
HGF and VEGFA
Growth factors HGF and VEGFA were identified in the gene ontology analysis for miR-1
(section 7.4.2). Analysis was done to determine if miR-1 could be regulating the mRNA
levels of HGF and VEGFA by gene silencing, with this negative regulation being disrupted
in SBNET metastases which have reduced miR-1 expression (datasets 1 and 2).
HGF and VEGFA expression was significantly increased in lymph node and liver metas-
tases of SBNET, Figure 7.3. HGF and VEGFA are predicted gene targets of miR-1 with
complementary base pairing between the mRNA of these genes and miR-1, Figure 7.3.
7.4.4. Summary
Gene ontology analysis revealed gene ontology terms related to apoptosis and cell death
that were significantly enriched amongst the genes with increased expression in the lymph
node metastases of SBNET (dataset a) that were also predicted gene targets of the
candidate miRNA with reduced expression in lymph node metastases, miR-1 and miR-
143-3p (datasets 1 and 2). Key oncogenes BCL-2, KRAS, FOSB and NUAK2 were
amongst the genes associated with the enriched gene ontology terms that are targeted
by miR-1 and miR-143-3p. Growth factors HGF and VEGFA were also identified for
miR-1. These newly identified miRNA-mRNA interactions in SBNET metastases related
to enriched gene ontology terms for apoptosis and cell death could play an important role
in disease progression in SBNET patients. MiR-1 and miR-143-3p therefore represent
the most promising candidates for the development of future novel miRNA biomarkers
327
Figure 7.2.: Reduced levels of miR-1 and miR-143 in SBNET metastases (datasets 1 and2) may result in reduced negative regulation of NUAK2 and FOSB expressionin SBNET metastases which could promote disease progression. A: Comple-mentary base pairing between miR-1 and FOSB mRNA. B: Complementarybase pairing between miR-143 and FOSB mRNA. C: Gene expression datashowing a significant reduction in FOSB expression in lymph node and livermetastases compared to SBNET (dataset a, GSE27162). D: Complementarybase pairing between miR-1 and NUAK2 mRNA. Gene expression data show-ing a significant reduction in NUAK2 expression in lymph node metastasescompared to SBNET (dataset a, GSE27162). Error bars show the mean +/-standard deviation (* p < 0.05, ** p < 0.01, *** p < 0.001). Reprintedby permission, ©[2016] [BioScientifica Ltd.], (Endocrine-Related Cancer)(Miller et al., 2016).
328
Figure 7.3.: Reduced expression of miR-1 (datasets 1 and 2) may lead to a reduced negativeregulation of the expression of growth factors HGF and VEGFA in SBNETmetastases and could therefore could be contributing to disease progression.A: Complementary base pairing between miR-1 and HGF mRNA, gene ex-pression data showing a significant reduction in HGF expression in lymphnode and liver metastases compared to SBNET (dataset a, GSE27162). B:Complementary base pairing between miR-1 and VEGFA mRNA, gene ex-pression data showing a significant reduction in VEGFA expression in lymphnode and liver metastases compared to SBNET (dataset a, GSE27162). Er-ror bars show the mean +/- standard deviation (* p < 0.05, ** p < 0.01,*** p < 0.001). Reprinted by permission, ©[2016] [BioScientifica Ltd.],(Endocrine-Related Cancer) (Miller et al., 2016).
329
for use in SBNET.
Further work is warranted to experimentally confirm silencing of these oncogenes by
miR-1 and miR-143-3p and to determine the phenotypic effects of the absence of oncogene
silencing on apoptosis in studies in SBNET cell lines. Once this is confirmed further
work would be needed to determine the efficacy of miR-1 and miR-143-3p as potential
prognostic biomarkers for use in patients with low grade SBNET.
7.5. Conclusions
This chapter has addressed the fifth research objective of this thesis by identifying the
most promising potential miRNA biomarkers for use in SBNET. This was done using
bioinformatics approaches and publicly available gene expression data.
The global miRNA profiling studies, in results chapters 5 and 6 of this thesis, identified
hundreds of dysregulated miRNA in SBNET and their metastases. Many of these miRNA
are likely to be regulating the expression of different genes in SBNET through gene
silencing. These potential miRNA-mRNA interactions would be far too many to test
experimentally, therefore bioinformatics was used to narrow the interactions down to
those most likely to be of particular biological importance for SBNET tumourigenesis
and disease progression. These miRNA-mRNA interactions could then be the focus of
further work in this area to develop novel miRNA biomarkers.
Novel miRNA-mRNA interactions were identified in SBNET that could be contributing
to tumourigenesis. Of particular interest were the genes FZD5, ACOX1, PTER and
SLC31A2 which were found to be downregulated in SBNET in all 3 gene expression
datasets and were also predicted gene targets of miR-7-5p, miR-204-5p and miR-375
which were upregulated in SBNET in the miRNA profiling experiments presented in
chapters 5 and 6. This level of redundancy in gene silencing suggests that these miRNA-
mRNA interactions may be particularly important in primary tumours of SBNET and
330
may play a role in tumourigenesis. These miRNA-mRNA interactions would be promising
candidates for future experimental work to determine if miR-7-5p, miR-204-5p and miR-
375 might be acting as oncomir in primary SBNET tumours.
Gene ontology analysis for the upregulated predicted gene targets of the candidate
miRNA which were downregulated in lymph node and liver metastases in the miRNA
profiling experiments, miR-1 and miR-143-3p, revealed significantly enriched gene ontol-
ogy terms related to cell death and apoptosis. There were 65 upregulated genes associated
with these terms for miR-1 and 67 upregulated genes associated with these terms for miR-
143-3p. This included growth factors HGF and VEGFA which were predicted targets of
miR-1 but not miR-143-3p.
Of particular interest for future studies was the identification of key oncogenes BCL-2,
KRAS, FOSB, NUAK2 which were predicted gene targets of both miR-1 and miR-143-
3p. These findings suggest that the absence of gene silencing in SBNET metastases by
miR-1 and miR-143-3p enables the overexpression of these oncogenes which could be
contributing to tumour progression. These miRNA-mRNA interactions would therefore
be of particular importance for further study, both to better understand the disease
pathology of SBNET metastases and to develop novel prognostic biomarkers for use in
SBNET patients.
This work identified miR-1 and miR-143-3p as the most promising potential miRNA
biomarkers for predicting prognosis in SBNET patients. Further work is warranted to
take these miRNA forwards into future experimental and clinical studies. This would
include in vitro experiments to experimentally confirm gene silencing of BCL-2, KRAS,
FOSB and NUAK2 by miR-1 and miR-143-3p and functional studies to determine any
changes in phenotype related to apoptosis response to the presence or absence of this
gene silencing. The efficacy of any potential new biomarker would then need to be tested
in further studies and in clinical trials to determine if it would be of benefit to SBNET
patients.
331
8. Discussion and Further Work
8.1. Discussion
The aim of this thesis was the following:
“To identify new potential prognostic biomarkers for use in GEP-NET.”
This overall aim was broken down into five principle research objectives:
“1) Investigate the limitations of the existing prognostic biomarker in GEP-
NET.”
“2) Experimentally determine a global miRNA profile of SBNET.”
“3) Verify the reproducibility and robustness of the SBNET miRNA profile.”
“4) Identify miRNA associated with disease progression in SBNET.”
“5) Identify the most promising potential miRNA biomarkers for use in SB-
NET.”
The first research objective was addressed in chapter 4 with an investigation of Ki-67
% grading and staging in 161 GEP-NET patients. This demonstrated that there was no
level of Ki-67 % at which patients could be considered to be ‘safe’ from distant metastases.
A high proportion of patients with G1 tumours (Ki-67: ≤ 2 %) and G2 tumours (Ki-67:
332
3-20 %), had stage IV disease, 28 % and 72 % of patients respectively, despite having a
low Ki-67 %. These findings held true when the data was analysed by primary site with
65 % of patients with low grade SBNET and 32 % of patients with low grade PNET
having distant metastases (G1/G2).
A study in 30 patients revealed that there was considerable heterogeneity in Ki-67
expression both within a single lesion and between different lesions from the same patient
which resulted in a change of grade in 67 % and 54 % of patients respectively (the majority
of the changes in grade were from G1 to G2).
These results demonstrated that there were limitations with the use of Ki-67 % as a
GEP-NET prognostic biomarker, particularly for low grade tumours. This indicated that
it would be useful to identify additional prognostic biomarkers for use alongside Ki-67 %
for further stratification of patients with low grade GEP-NET.
The second research objective was tackled in chapter 5, in which miRNA profiling was
done on matched tissue from 15 patients with low grade SBNET treated at Imperial
College Healthcare NHS Trust. This determined the global miRNA expression profile of
SBNET and revealed novel miRNA that had not been previously linked to SBNET tu-
mourigenesis. Further miRNA were identified with a potential role in tumour progression
due to being dysregulated in lymph node metastasis tissue.
Candidate miRNA (n=10) were selected as the most promising candidates for the
development of future biomarkers, based on having large changes in expression. The
expression changes of 9/10 candidate miRNA were confirmed by a second miRNA quan-
tification method and these miRNA were taken forwards for further experimental and
bioinformatics investigations, presented in chapters 6 and 7 respectively.
The third research objective was addressed in chapter 6, in which a second global
miRNA profiling experiment was carried out on tissue from 22 SBNET patients treated
at a separate institution, Zentralklinik Bad Berka. These experiments determined that
the global SBNET miRNA profile was reproducible, particularly for those miRNA that
333
were upregulated in SBNET, in an independent population of SBNET patients.
A 40 miRNA SBNET signature was identified made up of those miRNA with the
largest changes in expression between SBNET and “normal” small bowel tissue from
both profiling experiments. There were 29 of these miRNA that were also found to
be dysregulated in lymph node and liver metastases relative to their respective “normal”
tissues. These particular miRNA appear to have a role in the disease pathology of SBNET
that spans different disease stages and would therefore be of interest for future studies
into SBNET tumourigenesis.
The fourth research objective was also addressed in chapter 6. Global miRNA ex-
pression levels were quantified in the 13 liver metastasis samples and 15 lymph node
metastasis samples from the SBNET patients treated at Zentralklinik Bad Berka.
The miRNA expression profiling in liver metastases revealed 60 miRNA that were sig-
nificantly dysregulated in liver metastases relative to the primary tumour. Novel miRNA
were identified that had not been previously associated with liver metastases in SBNET.
These miRNA could be involved in promoting tumour progression and metastatic growth.
Of particular interest for the development of future prognostic biomarkers were miR-
1, miR-143-3p, miR-145-5p, miR-139-3p, miR-139-5p and miR-1233. These miRNA had
dramatically reduced expression levels in the lymph node metastasis samples and their ex-
pression was even further reduced in the liver metastasis samples (relative to the primary
tumour). The expression of these miRNA appears to be reduced during disease progres-
sion. These miRNA therefore represent particularly promising candidates for prognostic
biomarker development in SBNET. Further studies and clinical trials would be needed to
determine if these miRNA could successfully stratify patients with low grade SBNET into
clinically useful subgroups based on clinical and pathological behaviour. If these results
could be confirmed in circulating miRNA, a non-invasive liquid biopsy approach could
be used to quantify these miRNA in SBNET patients. It would be particularly helpful if
these miRNA could be used for the early detection of liver micrometastases and disease
334
progression. In contrast to these miRNA, miR-133a may be of less use for the detection
of disease progression since it was found to be equally repressed in both lymph node and
liver metastases.
The global miRNA profiling results presented in chapter 5 and in chapter 6 represent the
most comprehensive study of miRNA expression in SBNET and their metastases to date.
The novel miRNA identified as being associated with SBNET and their metastases are
likely to have important functions in promoting tumourigenesis and disease progression
through gene silencing. Further studies investigating the functional significance of these
results is likely to lead to a better understanding of SBNET tumourigenesis, particularly
since it is thought that epigenetic changes, such as changes in miRNA expression, may be
the main drivers of disease pathology in patients with low grade SBNET in the absence
of mutations in key tumour suppressors such as TP53 and RB1 (see Literature review,
section 2.6.2).
The final research objective was addressed in chapter 7 in which bioinformatics analysis
was carried out on the candidate miRNA that were confirmed as dysregulated by two
quantification methods and in both profiling experiments. Bioinformatics was done for
miR-7-5p, miR-204-5p, miR-375 (increased expression in SBNET relative to the “nor-
mal” small bowel tissue) and for miR-1 and miR-143-3p (reduced expression in lymph
node and liver metastases relative to SBNET). This was done to identify miRNA-mRNA
interactions that were most likely to be of biological importance for tumourigenesis and
disease progression in SBNET so that these interactions could be the focus of future
experimental studies to develop the most promising novel miRNA biomarkers.
The bioinformatics analysis identified miR-1 and miR-143-3p as the most promising
candidate miRNA for the development of future biomarkers to predict prognosis in SB-
NET patients. Bioinformatics analysis of predicted gene targets of miR-1 and miR-143-3p
with increased expression in lymph node metastases revealed significantly enriched gene
ontology terms for apoptosis and cell death. The upregulated genes associated with these
335
enriched gene ontology terms included key oncogenes BCL-2, KRAS, FOSB, NUAK2
which were predicted gene targets of miR-1 and miR-143-3p silencing. Growth factors
HGF and VEGFA were also identified and were predicted gene targets of miR-1 but not
miR-143-3p. These results suggest that in lymph node and liver metastases of SBNET
patients, the absence of gene silencing by miR-1 and miR-143-3p enables the overexpres-
sion of these oncogenes which could be contributing to metastatic growth and disease
progression.
MiR-1 and miR-143-3p had a reproducible reduction in expression in metastatic tissue
compared to the primary tumour in both miRNA profiling experiments and also had
promising results in the bioinformatics study, they therefore represent the most promis-
ing potential miRNA biomarkers for use in patients with low grade SBNET. Further
experimental and clinical studies are warranted to confirm the predicted miRNA-mRNA
interactions and to determine the clinical utility of miR-1 and miR-143-3p as prognostic
biomarkers. Further work that builds upon the findings of this thesis is explored in the
next section, section 8.2.
This represents the largest and most comprehensive study of miRNA expression in
SBNET patients to date, with global miRNA expression being quantified in 90 patient
samples in total across both profiling studies (datasets 1 and 2). The SBNET miRNA
profile identified in patients treated at Imperial College Healthcare NHS Trust was vali-
dated in an independent population of SBNET patients treated at a separate institution,
Zentralklinik Bad Berka. This showed that the results were reproducible in a separate
population of SBNET patients treated at a different institution. This study represents a
non-biased approach since 800 confirmed human miRNA were quantified based on miR-
Base version 18 (see Methods, section 3.3.3). Many novel miRNA were identified that
had not been previously associated with SBNET tumourigenesis and metastases. Table
8.1 demonstrates how the miRNA profiling results presented in this thesis compare to the
other studies to date that have included at least some element of miRNA quantification
336
in SBNET tumour tissue.
337
Table 8.1.: Studies involving miRNA quantification in primary tumours and metastases of SBNET patients
Paper Ruebel et al.(2010)
Li et al. (2013b) Nieser et al.(2016)
Miller et al.(2016)
Number ofSamples 28 24 20 90Patients 14 22 20 37MiRNA 85 847 15 800
Study type cancer panel,SBNET/metas-
tases
global profiling,SBNET/metas-
tases
15 miRNA,Chr18 (+/-) or
(+/+)
global profiling,SBNET/metas-
tasesSample type fresh frozen
tissuefresh frozen
tissueFFPE tissue FFPE/fresh
frozenMethods
MiRNA profiling assay QuantiMirCancer qPCR
Array
GeneChipmiRNA 1.0
Array
- nCountermiRNA
ExpressionAssay
Profiling assay provider SystemBiosciences, CA,
USA
Affymetrix, CA,USA
- NanoStringTechnologies,
WA, USAProfiling normalisation miR-197 quantile
normalisation- quantile
normalisationValidation assay qPCR qPCR qPCR qPCRValidation endogenous control SNORD48 SNORD48 SNORD61 SNORD44,
RNU6-1MiRNA profiling study
Number of patients 8 15 - 15Tumour tissue: SBNET, LNM, LVM 8, 1, 7 5, 5*, 5 - 15, 9, 2Adjacent normal: SB, LN, LV - - - 12, 7, 2
338
Continuation of Table 8.1
Paper Ruebel et al.(2010)
Li et al. (2013b) Nieser et al.(2016)
Miller et al.(2016)
Validation studyNumber of patients 6 7 20 22PT, LNM, LVM 6, 5, 1 3, 3*, 3 - 13, 15, 13Other samples (normal) normal ileal
tissueEC cells - normal ileal
tissuePT Chr18 (+/-), (+/+) - - 10, 10 -
Validated miRNAComparison groups LNM/LVM v
PTLNM/LVM vPT (and all v
EC cells)
Chr18 (+/-) vCh18 (+/+)
LNM/LVM vPT or #[T v
N]#Significant expression differences (-) miR-133a (-) miR-133a None (-) miR-133a
(-) miR-31 (-) miR-1(-) miR-129-5p (-) miR-143-3p
(-) miR-215 (-) miR-145-5p(+) miR-96 (-) miR-139-3p(+) miR-182 (-) miR-139-5p(+) miR-183 (-) miR-1233(+) miR-196a #[ (+)
miR-7-5p ]#(+) miR-200a
(NS: LNM/LVMv PT)
#[ (+)miR-204-5p ]#
#[ (+) miR-375]#
PT: primary tumour, LNM: lymph node metastasis, LVM: liver metastases, *: mesenteric metastases, NS: non-significant, HD:healthy donor, SB: “normal” small bowel, LN: lymph node, LV: liver, FFPE: initial study, fresh frozen: validation study. #[Tv N]#: indicates miRNA are upregulated in all tumour tissue types v their respective “normal” tissues.
339
There has only been one previous global study of miRNA expression in SBNET. This
study included only 24 samples in total, unmatched SBNET, mesenteric metastasis and
liver metastasis samples taken from SBNET patients, Table 8.1 (Li et al., 2013b). The
other main study of miRNA in SBNET and their lymph node and liver metastases was
by Ruebel et al, this study was limited in scope compared to a global analysis of miRNA
expression since it was focused on the expression of a panel of 85 miRNA and may
therefore have missed important changes in miRNA expression that were not covered
by the miRNA panel (Ruebel et al., 2010). For more details on these studies see the
Literature Review, section 2.5.3.
A further study by Nieser et al was of limited usefulness in understanding miRNA
expression in SBNET and their metastases since it only included primary tumour samples
(Nieser et al., 2016). The primary focus of this study was instead on gene expression
changes related to the loss of heterozygosity of Chr18 (+/-), a common genetic event in
SBNET (see section 2.6.2). Only the 15 miRNA present on Chr18 were quantified in this
study and only with respect to the presence or absence of the loss of heterozygosity of
Chr 18 in the primary tumour. No changes in miRNA expression were identified (Table
8.1).
MiR-133a was downregulated in both lymph node and liver metastases in the profiling
results and this miRNA was also identified as having significantly reduced expression
in these tissues in both the global miRNA profiling study by Li et al. (2013b) and the
cancer panel study by Ruebel et al. (2010) (see Table 8.1). MiR-133a was also confirmed
in a later study in serum samples from SBNET patients by the Li et al group as having
reduced expression in serum samples from SBNET patients with all different tumour
stages compared to serum samples from healthy donors (see Literature Review, section
2.5.3) (Li et al., 2015). MiR-133a was one of the 15 miRNA included in the Nieser et
al study but the expression of these miRNA did not vary with respect to the loss of
heterozygosity of Chr18 (Nieser et al., 2016).
340
It would have been interesting if the Nieser et al study had investigated the presence
or absence of Chr18 (+/-) in tissue from SBNET metastases with respect to miR-133a
expression in liver metastases. This could provide a possible mechanism for the consistent
reduction in the expression of miR-133a in SBNET metastases observed in the study
presented in this thesis and in the other two studies (Table 8.1). Alternatively changes in
miR-133a expression in SBNET metastases may be unrelated to the loss in heterozygosity
of Chr18, despite miR-133a being located on this chromosome.
This is backed up by the findings of the Nieser et al. (2016) study in the primary
tumour. The presence or absence of Chr18 (+/-) in the primary tumour was not found
to be associated with changes in the expression of the 6/7 tumour suppressor genes or the
15 miRNA located on Chr18. These findings suggest that alternative, possibly epigenetic
mechanisms may be behind the reduction in miR-133a expression in SBNET metastases
adding further weight to the hypothesis that epigenetic factors are key drivers of disease
pathology in SBNET (see Literature review, section 2.6.2) (Karpathakis et al., 2013;
Miller et al., 2015b).
The miRNA profiling study represented the most comprehensive investigation of miRNA
expression in SBNET metastases to date and included nearly double the number of liver
metastasis samples (n=15) and three times the number of lymph node metastasis samples
(n=24) than the second largest study by Li et al, Table 8.1. Of particular interest for the
development of future prognostic biomarkers were the novel findings presented in chap-
ter 6 that showed that the expression of miR-1, miR-143-3p, miR-145-5p, miR-139-3p,
miR-139-5p and miR-1233 was reduced in the lymph node metastasis samples and even
further reduced in liver metastasis samples.
These results suggest that these miRNA could represent particularly promising novel
prognostic biomarkers for the further stratification of patients with low grade SBNET to
identify patients with more aggressive tumours. If these results could be confirmed in
serum studies investigating circulating miRNA then these miRNA could be potentially
341
used for the early identification of liver micrometastases or treatment response monitoring
using a liquid biopsy approach. The efficacy of any potential novel biomarkers for clinical
use would need to be confirmed in future clinical trials.
These results are in contrast to those for miR-133a expression which was reduced with
respect to the primary tumour but was expressed at the same level in the lymph node
and liver metastases. This suggests that miR-133a may be of less use as a prognostic
biomarker for the early detection of liver metastases. The results presented in chapter
4 demonstrated that metastatic disease was present in 92 %, of patients with low grade
SBNET suggesting that a prognostic biomarker for the prediction or early identification
of liver metastases would be the most useful type of future biomarker for use in patients
with low grade SBNET. These findings for miR-133a are in keeping with the findings by
Li et al. (2013b) since they also found that expression levels of miR-133a were the same
in lymph node and liver metastases from SBNET patients despite being reduced in both
these tissues with respect to the primary tumour.
The other 7 miRNA validated by Li et al. (2013b), miR-96, miR-182, miR-183, miR-
196a, miR-31, miR-129-5p and miR-215, also showed the same pattern of expression as
miR-133a, with dysregulated expression in metastases compared to the primary tumour,
but with similar expression in the lymph node and liver metastases (see table 8.1). This
could explain why serum levels of these particular miRNA were not able to stratify
patients based on disease stage but could identify SBNET patients (of all disease stages)
from normal controls (see Literature review, section 2.5.3 for more details) (Li et al.,
2015).
7/8 of the miRNA validated by Li et al. (2013b) were identified as being significantly
dysregulated in the primary tumours compared to the “normal” small bowel samples in
the miRNA profiling experiments (see chapter 5, Tables 5.3 and 5.4 and chapter 6, Table
6.3). The results for miR-129-5p differed between the results presented here and the Li
et al study. MiR-129-5p had increased expression in both dataset 1 and 2 in the SBNET
342
compared to the “normal” small bowel samples, while the expression of this miRNA was
reduced in SBNET in the Li et al. (2013b) study (Table 5.3, Table 6.3).
These results suggest that while miR-133a, miR-96, miR-182, miR-183, miR-196a, miR-
31, miR-129-5p and miR-215 could be useful to aid diagnosis if there is already a high
suspicion of the presence of a SBNET, these miRNA would be of limited usefulness for
prognostic prediction. These miRNA may however have an important role in promoting
tumourigenesis in SBNET patients. This warrants further investigation and could lead to
a better understanding of the tumour biology of SBNET with the identification of novel
targets for therapeutic intervention as proposed by Li et al. (2013b).
The miRNA identified in chapters 5 and 6 of this thesis, miR-1, miR-143-3p, miR-145-
5p, miR-139-3p, miR-139-5p and miR-1233, represent particularly promising candidates
for future novel prognostic biomarkers for use in patients with low grade SBNET since
they had reduced expression with disease progression from lymph node metastases to
liver metastases. Further studies would be needed to determine if these results can
be confirmed in patients from additional centres and in serum from SBNET patients
to enable a non-invasive liquid biopsy approach for the stratification of patients into
clinically useful subgroups and to determine if they can be used for the prediction or
early detection of liver metastases.
The miRNA expression profile of SBNET and their metastases and the predicted inter-
actions with key oncogenes identified in this thesis represent a promising starting point
to direct future research aimed at a better understanding the tumour biology of SBNET.
This could lead to the identification of key drivers of tumourigenesis and novel drug
targets with the potential to improve patient survival and quality of life.
343
8.2. Further work
The results presented in this thesis represent the largest and most complete investigation
of miRNA expression in SBNET and in their lymph node and liver metastases in the
literature to date. Novel miRNA were identified and novel predicted miRNA-mRNA
interactions that had not been previously associated with SBNET. Further studies of
these miRNA are warranted both to investigate their function in the tumour biology of
SBNET and to develop and validate promising candidate miRNA as novel biomarkers for
use in SBNET.
8.2.1. Experimental validation of bioinformatics results
The bioinformatics results in chapter 7 showed that the downregulation of miR-1 and
miR-143-3p in SBNET could be an important event in the development of metastases
in SBNET patients due an absence in the silencing of key oncogenes which were shown
experimentally to have increased expression in SBNET metastases.
These results require in vitro validation, with gene silencing experiments in cell lines
to prove experimentally that the predicted miRNA-mRNA interactions do indeed occur.
Functional studies could then identify phenotype changes associated with the dysregula-
tion of these miRNA (see section 8.2.2).
Reduced expression in miR-1 and miR-143-3p was predicted to prevent gene silencing
of the expression of oncogenes including BCL-2, KRAS, FOSB and NUAK2 that had
been found to be overexpressed in SBNET metastases in gene expression experiments.
Growth factors growth factors HGF and VEGFA were also identified for miR-1.
Oncogene KRAS is rarely mutated in SBNET which suggests that post-translational
changes in miRNA expression with a reduction in gene silencing of KRAS by miR-1 and
miR-143-3p are likely to be contributing to the upregulation of this oncogene in SBNET
patients (chapter 7, dataset a) (Banck et al., 2013; Miller et al., 2016). The phenotypic
344
effects of this could be tested in further in vitro and in vivo functional studies.
Further work should include experimental studies to confirm that the predicted miR-
1/miR-143-3p gene targets of interest for SBNET disease progression including BCL-2,
KRAS, FOSB, NUAK2, HGF and VEGFA are indeed targeted by miR-1/miR-143-3p in
vitro (luciferase reporter gene assay) and to confirm that overexpression in these miRNA
triggers a reduction in expression of these genes at the mRNA (qPCR) and protein level
(western blot). These genes were associated with enriched gene ontology terms related to
apoptosis. It would therefore be of particular interest to carry out phenotyping studies
to determine if rates of apoptosis was affected in cell lines with reduced miR-1 and miR-
143-3p expression and increased expression of oncogenes such as KRAS.
8.2.2. Functional studies
Functional studies, in vitro models
Functional studies would be of interest in SBNET cell lines in order to better understand
the phenotype changes that might be triggered by the downregulation of miR-1, miR-
143-3p, miR-145-5p, miR-139-3p, miR-139-5p and miR-1233 in SBNET metastases and
the upregulation of miR-7-5p, miR-204-5p and miR-375 in primary tumours. This would
enable a better understanding of the disease pathology in SBNET as well as the better
characterisation of the phenotypic changes caused by the dysregulation of these miRNA
and their gene targets in SBNET.
Of particular interest would be in vitro studies in SBNET cell lines to determine if the
downregulation of miR-1 and miR-143-3p and the removal of their gene silencing action
on oncogenes such as BCL-2, KRAS, FOSB and NUAK2 led to phenotype changes in
metastatic SBNET cell lines and in primary cultures established from metastatic SBNET
tissue. The H-STS and L-STS cell lines developed from SBNET liver and lymph node
metastases respectively would therefore be particularly useful for the characterisation
of the phenotype changes caused by miR-1 and miR-143-3p downregulation in SBNET
345
metastases (Pfragner et al., 2009). The CNDT2 cell line would be a less optimal choice
for in vitro studies of metastatic growth in SBNET. The cell line was isolated from a
liver metastasis from a patient with a primary SBNET, however there have been difficul-
ties in identifying CgA and secretory granules in cells, therefore the authenticity of the
neuroendocrine background of this cell line has been questioned (Van Buren et al., 2007;
Grozinsky-Glasberg et al., 2012).
Experiments could be carried out to identify if these epigenetic changes in miR-1 and
miR-143-3p expression trigger changes in cellular pathways such as apoptosis, cellular
proliferation, motility, inflammation and cell adhesion. This would provide a better
understanding of the effects of miR-1 and miR-143-3p downregulation on SBNET tu-
mourigenesis and metastatic growth. It would be of particular interest to carry out these
functional studies in primary cells taken from lymph node and liver metastases of SBNET
patients were possible. This would require cells to be isolated from tumour tissue and
successfully grown in culture for functional studies. Primary cells are more challenging
to grow in culture and are of limited lifespan however they retain more of the charac-
teristics of the tumour due to acquiring far less mutations than immortalised cell lines
which have undergone many passages (Pan et al., 2009). There have been no functional
studies carried out on primary cultures from SBNET patients to date, this may be due
to the cell culture challenges presented by the low proliferation rates in these tumours
(Grozinsky-Glasberg et al., 2012; Pfragner et al., 2009).
Analysis of phenotypic changes related to the effects of increased expression of miR-
7-5p, miR-204-5p, miR-375 in SBNET would be of interest and could be carried out
in P-STS or KRJ-I cell lines (developed form SBNET primary tumours) (Pfragner et
al., 2009; Pfragner, 1996). These studies could investigate if these miRNA could be
contributing to SBNET tumourigenesis in the absence of mutations in TP53 and RB1
in patients with low grade SBNET, see Literature Review, sections 2.6.2 and 2.6.2, for
more details on the low mutation rate of SBNET.
346
There have been recent advances in the techniques involved in the development of 3D
cell cultures and organoids containing multiple cell lineages that form organ like structures
(Fatehullah et al., 2016). Organoids have the potential to enable far more complex
morphological and phenotyping experiments to be carried out since these experiments
can be done in 3D space, making organoids a far closer representation of the system being
studied than a 2D monolayer of cells. Patient derived organoids from SBNET patients
grown in 3D culture would be of interest for future functional studies in SBNET to better
interrogate disease pathology and the phenotypic effects of changes in the expression of
miRNA such as miR-1 and miR-143-3p, miR-7-5p, miR-204-5p and miR-375.
Such studies would lead to a better understanding of tumour biology of low grade
SBNET through the functional characterisation of the miRNA that are dysregulated
in these tumours and their metastases. This would enable the further development of
miRNA as novel biomarkers and identify novel therapeutic targets with the potential to
improve treatment and outcomes for SBNET patients.
Functional studies, in vivo models
There is currently no mouse model of SBNET, only PNET, which limits the scope of
studies attempting to better understand the disease pathology of SBNET at a whole
organism level. It would therefore be very beneficial to studies in this area if transgenic
organisms were developed that develop tumours in enterochromaffin cells and recapitu-
late SBNET tumourigenesis. This would enable in vivo studies to be carried to either
overexpress or knock down the expression of miRNA of interest such as miR-1 and miR-
143-3p. This would determine if their overexpression inhibited the growth/development
of SBNET lymph node and liver metastases by silencing oncogene expression.
Also of interest would be xenograft models in mice, these would be very helpful to
determine the wider implications of miR-1 and miR-143-3p downregulation in SBNET
metastases. There have been no studies of miRNA in xenograft models of SBNET to
347
date.
Downregulation of miR-143-3p expression is a common occurrence in many different
cancers including pancreatic cancer, colorectal cancer, gastric cancer and B-cell lym-
phomas (Yang et al., 2010). Interestingly a functional study was carried out investigating
miR-143-3p expression in a pancreatic cancer cell line and in a mouse xenograft model
(Hu et al., 2012). This showed that overexpression of miR-143-3p inhibited the invasion
and migration of Panc-1 cells (established from a pancreatic cancer liver metastasis) and
inhibited metastasis growth in the Panc-1 xenograft model.
MiR-143-3p overexpression was found to reduce KRAS, ARHGEF1 and ARHGEF2
expression at the gene and protein level by gene silencing in the in vitro and in vivo mod-
els of pancreatic cancer, leading to reduced GTPase activity and increased E-cadherin
expression (Hu et al., 2012). ARHGEF1 and ARHGEF2 encode proteins that activate
the Rho GTPases which are thought to be crucial for tumour cell migration and inva-
sion leading to a metastatic cellular phenotype by triggering reorganisation of the actin
cytoskeleton and triggering E-cadherin mediated cell adhesion (Hu et al., 2012).
Interestingly the bioinformatics results presented in chapter 7 suggest that a similar
process may be occurring in metastatic SBNET patients as genes in the same gene family
to those that were identified as being regulated by miR-143-3p expression in the studies
in pancreatic cancer were identified as being dysregulated in the SBNET metastases. The
bioinformatics study identified two other genes in the Rho guanine nucleotide exchange
factor family, ARHGEF7 and ARHGEF18, amongst the upregulated genes associated
with the enriched gene ontology terms related to cell death and apoptosis. ARHGEF7
was a predicted gene target of both miR-1 and miR-143-3p, while ARHGEF18 was a
predicted gene target of miR-1 but not miR-143-3p, see chapter 7, Tables 7.5 and 7.8).
This suggests that in addition to the loss of gene silencing of the oncogene KRAS in
SBNET, there could also be a loss in the gene silencing of ARHGEF7 and ARHGEF18 in
SBNET patients. The reduction in miR-1 and miR-143-3p expression could be causing an
348
aggressive metastatic phenotype in a subset of SBNET patients with reduced miR-1/miR-
143-3p expression through the activation of Rho GTPases by ARHGEF7/ARHGEF18,
in an analogous process to that functionally characterised in the studies of miR-143-3p
in the pancreatic cancer models. Similar in vitro and in vivo studies would need to be
carried out in cell and mouse xenograft models of metastatic SBNET to determine if this
was the case for SBNET patients.
It would be interesting to carry out a similar in vivo study in SBNET to determine
if miR-143-3p and/or miR-1 overexpression could also inhibit the growth of xenograft
tissue from a SBNET liver metastasis. If such studies were successful it would suggest
that miR-143-3p would be a promising therapeutic target for gene therapy approaches to
rescue miR-143-3p expression in SBNET patients in order to silence oncogene expression
and inhibit metastatic growth.
It would also be interesting to carry out in vivo functional studies on the other miRNA
that were identified as being downregulated with disease progression in SBNET in chap-
ter 6, miR-145-5p, miR-139-3p, miR-139-5p and miR-1233 to determine the effect on
metastatic growth in vivo of rescuing the expression of these miRNA. Functional studies
of in vivo models of the primary tumour would also be of interest for the candidate miRNA
that were upregulated in SBNET compared to “normal” small bowel tissue, miR-7-5p,
miR-204-5p and miR-375, identified in chapters 5 and 6. These studies could elucidate
the possible role of these miRNA in the tumourigenesis of low grade SBNET and would
be of particular interest since the lack of mutations in key oncogenes in SBNET points to
the involvement of post-translational, epigenetic changes in miRNA expression or DNA
methylation as key drivers of tumourigenesis.
Novel therapies identified as a result of these functional studies could include therapeu-
tic approaches based on miRNA, see Literature review, section 2.5.2 for more details. For
example clinical trials could be carried out to determine if the therapeutic reintroduction
of tumour suppressor miRNA, miR-1 and miR-143-3p, using gene therapy approaches,
349
was able to prevent metastatic growth and improve patient survival in patients with
low grade SBNET by silencing the expression of oncogenes BCL-2, KRAS, FOSB and
NUAK2.
These functional studies would enable a better systemic understanding of the role of
miRNA in promoting tumourigenesis and metastatic growth in SBNET and enable novel
treatments and biomarkers for SBNET to be further tested and developed in order to
improve the patient journey for SBNET patients and potentially increase survival.
8.2.3. Future biomarker development
The results presented in this thesis identified miR-1 and miR-143-3p as being the most
promising candidate miRNA biomarkers for future prognostic prediction in SBNET pa-
tients. Further work would be needed with further studies and clinical trials to determine
if miR-1 and miR-143-3p expression could indeed be used to stratify patients with low
grade SBNET into clinically useful subgroups. This could enable the identification of pa-
tients with a more aggressive metastatic phenotype so that these patients could receive
tailored treatment and more frequent follow up.
Of particular interest would be studies to determine if serum levels of miR-1 and miR-
143-3p decrease with advancing tumour stage in SBNET patients, reflecting the findings
in the tissue studies. If this was the case then a non-invasive liquid biopsy approach could
be used to monitor miR-1 and miR-143-3p levels in SBNET patients. This would enable
samples to be taken at multiple time points to monitor the patient journey in real time
instead of relying on a biopsy taken at a single location and time point. Studies could be
carried out to determine if monitoring serum levels of these biomarkers over time could
potentially lead to the early identification of disease progression or the development of
micrometastases in SBNET patients before these were visible on imaging allowing for
early intervention.
Further studies and clinical trials would need to be carried out to determine if miR-1
350
and miR-143-3p could successfully stratify patients with low grade SBNET into clinically
useful subgroups based on clinical and pathological behaviour. These studies could de-
termine for example if these miRNA could be used for the prediction/early detection of
liver metastases or to identify the presence of a more aggressive phenotype in a subset
of SBNET patients. Clinical trials would then need to be carried out to validate the use
of miR-1 and miR-143-3p as novel SBNET biomarkers and to ensure that they would
provide a benefit over the use of Ki-67 % alone for prognostic prediction.
351
Bibliography
Abdel-Rahman, O (2014). “Vascular endothelial growth factor (VEGF) pathway and
neuroendocrine neoplasms (NENs): prognostic and therapeutic considerations”. In: Tu-
mour Biol 35.11, pp. 10615–10625. issn: 1423-0380.
Adams, D. J., C. P. Arthur, and M. H. B. Stowell (2015). “Architecture of the Synapto-
physin/Synaptobrevin Complex: Structural Evidence for an Entropic Clustering Func-
tion at the Synapse.” In: Scientific reports 5.April, p. 13659. issn: 2045-2322.
Ahmed, A, G Turner, B King, L Jones, D Culliford, D McCance, J Ardill, B. T. Johnston,
G Poston, M Rees, M Buxton-Thomas, M Caplin, and J. K. Ramage (2009). “Midgut
neuroendocrine tumours with liver metastases: results of the UKINETS study.” In:
Endocrine-related cancer 16.3, pp. 885–94. issn: 1479-6821.
Akerblom, B., G. Zang, Z. W. Zhuang, G. Calounova, M. Simons, and M. Welsh (2012).
“Heterogeneity among RIP-Tag2 insulinomas allows vascular endothelial growth factor-
A independent tumor expansion as revealed by studies in Shb mutant mice: Implications
for tumor angiogenesis”. In: Molecular Oncology 6.3, pp. 333–346. issn: 15747891.
Aksu, A. U., O. Egritas Gurkan, S. Sarı, Z. Demirtas, C. Turkyılmaz, A. Poyraz, and B.
Dalgıc (2016). “Mutant neurogenin-3 in a Turkish boy with congenital malabsorptive
diarrhea”. In: Pediatrics International 58.5, pp. 379–382. issn: 1442200X.
353
Altman, D. G., L. M. McShane, W. Sauerbrei, and S. E. Taube (2012). “Reporting
Recommendations for Tumor Marker Prognostic Studies (REMARK): explanation and
elaboration.” In: Breast Cancer Research and Treatment 9.5, e1001216. issn: 01676806.
Alzugaraya, M. E., S. Hernandez-Martınez, and J. R. Ronderos (2016). “Somatostatin
signaling system as an ancestral mechanism: Myoregulatory activity of an Allatostatin-
C peptide in Hydra”. In: Peptides 82, pp. 67–75. issn: 18735169.
Ambros, V., B. Bartel, D. P. Bartel, C. B. Burge, J. C. Carrington, X. Chen, G. Dreyfuss,
S. R. Eddy, S. Griffiths-Jones, M. Marshall, M. Matzke, G. Ruvkun, and T. Tuschl
(2003). “A uniform system for microRNA annotation.” In: RNA (New York, N.Y.)
9.3, pp. 277–9. issn: 1355-8382.
Anderson, C. W. and J. J. Bennett (2016). “Clinical Presentation and Diagnosis of Pan-
creatic Neuroendocrine Tumors”. In: Surgical Oncology Clinics of North America 25.2,
pp. 363–374. issn: 15585042.
Andersson, E., C. Sward, G. Stenman, H. Ahlman, and O. Nilsson (2009). “High-resolution
genomic profiling reveals gain of chromosome 14 as a predictor of poor outcome in ileal
carcinoids.” In: Endocrine-related cancer 16.3, pp. 953–66. issn: 1479-6821.
Ardill, J. E. and B. Erikkson (2003). “The importance of the measurement of circulat-
ing markers in patients with neuroendocrine tumours of the pancreas and gut”. In:
Endocrine-Related Cancer 10.4, pp. 459–462. issn: 13510088.
Arimura, A. and J. B. Fishback (1981). “Somatostatin: Regulation of secretion”. In:
Neuroendocrinology 33.4, pp. 246–256. issn: 14230194.
Arnold, R., A. Wilke, A. Rinke, C. Mayer, P. H. Kann, K. J. Klose, A. Scherag, M.
Hahmann, H. H. Muller, and P. Barth (2008). “Plasma Chromogranin A as Marker for
354
Survival in Patients With Metastatic Endocrine Gastroenteropancreatic Tumors”. In:
Clinical Gastroenterology and Hepatology 6.7, pp. 820–827. issn: 15423565.
Aryani, A. and B. Denecke (2015). “In vitro application of ribonucleases: comparison of
the effects on mRNA and miRNA stability.” In: BMC research notes 8.1, p. 164. issn:
1756-0500.
Ashburner, M., C. A. Ball, J. A. Blake, D. Botstein, H. Butler, J. M. Cherry, A. P. Davis,
K. Dolinski, S. S. Dwight, J. T. Epping, M. A. Harris, D. P. Hill, L. Issel-Tarver, A.
Kasarskis, S. Lewis, J. C. Matese, J. E. Richardshon, M. Ringwald, G. M. Rubin, and
G. Sherlock (2000). “Gene Ontology: tool for the unification of biology”. In: Nature
Genetics 25.1, pp. 25–29. issn: 1061-4036. arXiv: 10614036.
Avesson, L., J. Reimegard, E. G. H. Wagner, and F. Soderbom (2012). “MicroRNAs
in Amoebozoa: Deep sequencing of the small RNA population in the social amoeba
Dictyostelium discoideum reveals developmentally regulated microRNAs”. In: RNA
18.10, pp. 1771–1782. issn: 1355-8382.
Bajetta, E, L Ferrari, A Martinetti, L Celio, and G Procopio (1999). “Chromogranin
A., neuron specific enolase, carcinoembryonic antigen, and hydroxyi ndolacetic acid
evaluation in patients with neuroendocrine tumors.” In: pp. 858–865.
Bak, R. O. and J. G. Mikkelsen (2010). “Regulation of cytokines by small RNAs during
skin inflammation.” In: Journal of Biomedical Science 17.1, p. 53. issn: 1423-0127.
Baldelli, R., A Barnabei, L Rizza, a. M. Isidori, F Rota, P Di Giacinto, A Paoloni, F
Torino, S. M. Corsello, A Lenzi, and M Appetecchia (2014). “Somatostatin Analogs
Therapy in Gastroenteropancreatic Neuroendocrine Tumors: Current Aspects and New
Perspectives.” In: Frontiers in endocrinology 5.7, pp. 1–10. issn: 1664-2392.
355
Banck, M. S. and A. S. Beutler (2014). “Advances in small bowel neuroendocrine neo-
plasia.” In: Current opinion in gastroenterology 30.2, pp. 163–7. issn: 1531-7056.
Banck, M., R Kanwar, A. Kulkarni, G. Boora, F Metge, B. Kipp, L Zhang, E. Thorland,
K. Minn, R Tentu, B. Eckloff, E. Wieben, Y Wu, J. Cunningham, D. Nagorney, J.
Gilbert, M. Ames, and A. Beutler (2013). “The genomic landscape of small intestine
neuroendocrine tumors”. In: J Clin Invest. 123.6, pp. 2502–8.
Barker, N., J. H. van Es, J. Kuipers, P. Kujala, M. van den Born, M. Cozijnsen, A.
Haegebarth, J. Korving, H. Begthel, P. J. Peters, and H. Clevers (2007). “Identification
of stem cells in small intestine and colon by marker gene Lgr5”. In: Nature 449.7165,
pp. 1003–1007. issn: 1476-4687.
Barry, M. (1998). “PSA screening for prostate cancer: the current controversy.” In: Annals
of oncology : official journal of the European Society for Medical Oncology / ESMO
9.12, pp. 1279–82. issn: 0923-7534 (Print).
Bassett, J. H., S. A. Forbes, A. A. Pannett, S. E. Lloyd, P. T. Christie, C Wooding, B
Harding, G. M. Besser, C. R. Edwards, J. P. Monson, J Sampson, J. A. Wass, M. H.
Wheeler, and R. V. Thakker (1998). “Characterization of mutations in patients with
multiple endocrine neoplasia type 1.” In: American journal of human genetics 62.2,
pp. 232–44. issn: 0002-9297.
Bauer, W., U. Briner, W. Doepfner, R. Haller, R. Huguenin, P. Marbach, T. J. Petcher,
and J. Pless (1982). “SMS 201-995: A very potent and selective octapeptide analogue
of somatostatin with prolonged action”. In: Life Sciences 31.11, pp. 1133–1140. issn:
00243205.
Bell, E. and M. A. Taylor (2017). “Functional Roles for Exosomal MicroRNAs in the
Tumour Microenvironment”. In: Computational and Structural Biotechnology Journal
15, pp. 8–13. issn: 20010370.
356
Ben, Q., J. Zhong, J. Fei, H. Chen, L. Yv, J. Tan, and Y. Yuan (2016). “Risk Factors
for Sporadic Pancreatic Neuroendocrine Tumors: A Case-Control Study.” In: Scientific
reports 6, p. 36073. issn: 2045-2322.
Benjamini, Y. and Y. Hochberg (1995). “Controlling the false discovery rate: apractical
and powerful approach to multiple testing”. In: Journal of the Royal Statistical Society
Series B 57.1, pp. 289–300.
Berge, T and F Linell (1976). “Carcinoid tumours. Frequency in a defined population
during a 12-year period.” In: Acta Pathologica Microbiologica Scandinavica Section A
Pathology 84A.4, pp. 322–330.
Berger, M., J. A. Gray, and B. L. Roth (2009). “The expanded biology of serotonin.” In:
Annual Review of Medicine 60.August 2016, pp. 355–366. issn: 0066-4219.
Best, J., H. F. Nijhout, and M. Reed (2010). “Serotonin synthesis, release and reuptake
in terminals: a mathematical model.” In: Theoretical biology & medical modelling 7,
p. 34. issn: 1742-4682.
Blakeley, J. O. and S. R. Plotkin (2016). “Therapeutic advances for the tumors associated
with neurofibromatosis type 1, type 2, and schwannomatosis”. In: Neuro-Oncology 18.5,
pp. 624–638. issn: 15235866.
Bleazard, T., J. A. Lamb, and S. Griffiths-Jones (2015). “Bias in microRNA functional
enrichment analysis”. In: Bioinformatics 31.10, pp. 1592–1598. issn: 14602059.
Boeri, M., C. Verri, D. Conte, L. Roz, P. Modena, F. Facchinetti, E. Calabro, C. M.
Croce, U. Pastorino, and G. Sozzi (2011). “MicroRNA signatures in tissues and plasma
predict development and prognosis of computed tomography detected lung cancer”. In:
Proceedings of the National Academy of Sciences 108.9, pp. 3713–3718. issn: 0027-8424.
arXiv: arXiv:1408.1149.
357
Bono, J. S. de, H. I. Scher, R. B. Montgomery, C. Parker, M. C. Miller, H. Tissing, G. V.
Doyle, L. W.W. M. Terstappen, K. J. Pienta, and D. Raghavan (2008). “Circulating
tumor cells predict survival benefit from treatment in metastatic castration-resistant
prostate cancer.” In: Clinical cancer research : an official journal of the American
Association for Cancer Research 14.19, pp. 6302–9. issn: 1078-0432.
Booth, D. G., M. Takagi, L. Sanchez-Pulido, E. Petfalski, G. Vargiu, K. Samejima, N.
Imamoto, C. P. Ponting, D. Tollervey, W. C. Earnshaw, and P. Vagnarelli (2014). “Ki-
67 is a PP1-interacting protein that organises the mitotic chromosome periphery”. In:
eLife 3, e01641. issn: 2050084X.
Bornstein, J. C. (2012). “Serotonin in the gut: What does it do?” In: Frontiers in Neu-
roscience 6.16, pp. 1–2. issn: 16624548.
Bossuyt, P. M., J. B. Reitsma, D. E. Bruns, C. A. Gatsonis, P. P. Glasziou, L. M. Irwig,
D. Moher, D. Rennie, H. C. W. de Vet, J. G. Lijmer, and Standards for Reporting
of Diagnostic Accuracy Group (2003). “The STARD statement for reporting studies
of diagnostic accuracy: explanation and elaboration. The Standards for Reporting of
Diagnostic Accuracy Group.” In: Croatian medical journal 44.5, pp. 639–50. issn: 0353-
9504.
Bottarelli, L., C. Azzoni, S. Pizzi, T. D’Adda, E. M. Silini, C. Bordi, and G. Rindi
(2013). “Adenomatous polyposis coli gene involvement in ileal enterochromaffin cell
neuroendocrine neoplasms.” In: Human pathology 44.12, pp. 2736–42. issn: 1532-8392.
Boucher, J., A. Kleinridders, and R. C. Kahn (2014). “Insulin Receptor Signaling in
Normal”. In: Cold Spring Harb Perspect Biol 2014 6.1, a009191. issn: 1943-0264.
Bowman, S. L., D. J. Shiwarski, and M. A. Puthenveedu (2016). “Distinct G protein-
coupled receptor recycling pathways allow spatial control of downstream G protein
signaling”. In: Journal of Cell Biology 214.7, pp. 1–10. issn: 15408140.
358
Braga, F., S. Ferraro, R. Mozzi, A. Dolci, and M. Panteghini (2013). “Biological variation
of neuroendocrine tumor markers chromogranin A and neuron-specific enolase”. In:
Clinical Biochemistry 46.1-2, pp. 148–151. issn: 00099120.
Brandi, M. L., R. F. Gagel, a Angeli, J. P. Bilezikian, P Beck-Peccoz, C Bordi, B Conte-
Devolx, a Falchetti, R. G. Gheri, a Libroia, C. J. Lips, G Lombardi, M Mannelli,
F Pacini, B. a. Ponder, F Raue, B Skogseid, G Tamburrano, R. V. Thakker, N. W.
Thompson, P Tomassetti, F Tonelli, S. a. Wells, and S. J. Marx (2001). “Guidelines
for diagnosis and therapy of MEN type 1 and type 2.” In: The Journal of clinical
endocrinology and metabolism 86.12, pp. 5658–5671. issn: 0021-972X.
Brazeau, Vale, Burgus, Ling, Butcher, Rivier, and Guillemin (1973). “Hypothalamic
polypeptide that inhibits the secretion of immunoreactive pituitary growth hormone”.
In: Science (New York, NY) 179.68, pp. 77–79. issn: 0036-8075.
Bremnes, R. M., T. Dønnem, S. Al-Saad, K. Al-Shibli, S. Andersen, R. Sirera, C. Camps,
I. Marinez, and L. T. Busund (2011). “The role of tumor stroma in cancer progression
and prognosis: Emphasis on carcinoma-associated fibroblasts and non-small cell lung
cancer”. In: Journal of Thoracic Oncology 6.1, pp. 209–217. issn: 15561380.
Briest, F. and P. Grabowski (2014). “PI3K-AKT-mTOR-Signaling and beyond: the Com-
plex Network in Gastroenteropancreatic Neuroendocrine Neoplasms.” In: Theranostics
4.4, pp. 336–365. issn: 1838-7640.
Bueno, M. J. and M. Malumbres (2011). “MicroRNAs and the cell cycle”. In: Biochimica
et Biophysica Acta - Molecular Basis of Disease 1812.5, pp. 592–601. issn: 09254439.
Bullwinkel, J., B. Baron-Luhr, A. Ludemann, C. Wohlenberg, J. Gerdes, and T. Scholzen
(2006). “Ki-67 protein is associated with ribosomal RNA transcription in quiescent and
proliferating cells.” In: Journal of cellular physiology 206.3, pp. 624–35. issn: 0021-9541.
359
Buscaglia, L. E. B. and Y. Li (2011). “Review Hallmarks of Cancer Apoptosis MiR-21
Targets”. In: Chinese journal of cancer 30.6, pp. 371–380.
Bussolati, G. and A. G. Pearse (1967). “Immunofluorescent localization of calcitonin in
the ’C’ cells of pig and dog thyroid.” In: Journal of Endocrinology 37.2, pp. 205–209.
issn: 00220795.
Caicedo, A. (2013). “PARACRINE AND AUTOCRINE INTERACTIONS IN THE HU-
MAN ISLET: MORE THAN MEETS THE EYE”. In: Seminars in cell & developmental
biology 24.1, pp. 11–21. issn: 15378276. arXiv: NIHMS150003.
Calin, G. A., C. D. Dumitru, M. Shimizu, R. Bichi, S. Zupo, E. Noch, H. Aldler, S.
Rattan, M. Keating, K. Rai, L. Rassenti, T. Kipps, M. Negrini, F. Bullrich, and C. M.
Croce (2002). “Frequent deletions and down-regulation of micro- RNA genes miR15
and miR16 at 13q14 in chronic lymphocytic leukemia”. In: Proceedings of the National
Academy of Sciences 99.24, pp. 13–18.
Cao, P., A. Abedini, and D. P. Raleigh (2013). “Aggregation of islet amyloid polypeptide:
From physical chemistry to cell biology”. In: Current Opinion in Structural Biology
23.1, pp. 82–89. issn: 0959440X. arXiv: NIHMS150003.
Caplin, M. E., M. Pavel, J. B. Cwik la, A. T. Phan, M. Raderer, E. Sedlackova, G. Ca-
diot, E. M. Wolin, J. Capdevila, L. Wall, G. Rindi, A. Langley, S. Martinez, J. Blum-
berg, P. Ruszniewski, and CLARINET Investigators (2014). “Lanreotide in metastatic
enteropancreatic neuroendocrine tumors.” In: The New England journal of medicine
371.3, pp. 224–33. issn: 1533-4406.
Caplin, M. E., E. Baudin, P. Ferolla, P. Filosso, M. Garcia-Yuste, E. Lim, K. Oberg,
G. Pelosi, A. Perren, R. E. Rossi, W. D. Travis, D. Bartsch, J. Capdevila, F. Costa,
J. Cwikla, W. de Herder, G. D. Fave, B. Eriksson, M. Falconi, D. Ferone, D. Gross, A.
Grossman, T. Ito, R. Jensen, G. Kaltsas, F. Kelestimur, R. Kianmanesh, U. Knigge, B.
360
Kos-Kudla, E. Krenning, E. Mitry, M. Nicolson, J. O’Connor, D. O’Toole, U. F. Pape,
M. Pavel, J. Ramage, E. Raymond, G. Rindi, A. Rockall, P. Ruszniewski, R. Salazar,
A. Scarpa, E. Sedlackova, A. Sundin, C. Toumpanakis, M. P. Vullierme, W. Weber,
B. Wiedenmann, and Z. Zheng-Pei (2015). “Pulmonary neuroendocrine (carcinoid)
tumors: European Neuroendocrine Tumor Society expert consensus and recommenda-
tions for best practice for typical and atypical pulmonary carcinoids”. In: Annals of
Oncology 26.8, pp. 1604–1620. issn: 15698041.
Capurso, G., S. Festa, R. Valente, M. Piciucchi, F. Panzuto, R. T. Jensen, and G. Delle
Fave (2012). “Molecular pathology and genetics of pancreatic endocrine tumours.” In:
Journal of molecular endocrinology 49.1, R37–50. issn: 1479-6813.
Catalanotto, C., C. Cogoni, and G. Zardo (2016). “MicroRNA in control of gene expres-
sion: An overview of nuclear functions”. In: International Journal of Molecular Sciences
17.10. issn: 14220067.
Chen, C., A. Neugut, and H Rotterdam (1994). “Risk factors for adeno- carcinomas and
malignant carcinoids of the small intestine: preliminary findings.” In: Cancer Epidemi-
ology, Biomarkers and Prevention 3.May, pp. 205–7.
Chen, L., J. Hou, L. Ye, Y. Chen, J. Cui, W. Tian, C. Li, and L. Liu (2014). “MicroRNA-
143 regulates adipogenesis by modulating the MAP2K5-ERK5 signaling.” In: Scientific
reports 4.3819. issn: 2045-2322.
Cheng, C. W., L. C. Chang, T. L. Tseng, C. C. Wu, Y. F. Lin, and J. S. Chen (2014).
“Phosphotriesterase-related protein sensed albuminuria and conferred renal tubular
cell activation in membranous nephropathy”. In: Journal of Biomedical Science 21.1,
pp. 1–8. issn: 14230127.
361
Cheng, H and C. P. Leblond (1974). “Origin, differentiation and renewal of 4 main ep-
ithelial cell types in mouse small intestine. 5 Unitarian theory of origin of 4 epithelial
cell types.” In: American Journal of Anatomy 141.4, pp. 537–561.
Cherenfant, J., S. J. Stocker, M. K. Gage, H. Du, T. a. Thurow, M. Odeleye, S. W.
Schimpke, K. L. Kaul, C. R. Hall, I. Lamzabi, P. Gattuso, D. J. Winchester, R. W.
Marsh, K. K. Roggin, D. J. Bentrem, M. S. Baker, R. a. Prinz, and M. S. Talamonti
(2013). “Predicting aggressive behavior in nonfunctioning pancreatic neuroendocrine
tumors.” In: Surgery 154.4, 785–91; discussion 791–3. issn: 1532-7361.
Chi, S. W., G. J. Hannon, and R. B. Darnell (2012). “An alternative mode of microRNA
target recognition.” In: Nature structural & molecular biology 19.3, pp. 321–7. issn:
1545-9985. arXiv: NIHMS150003.
Cho, M. Y., J. M. Kim, J. H. Sohn, M. J. Kim, K. M. Kim, W. H. Kim, H. Kim, M. C.
Kook, D. Y. Park, J. H. Lee, H. Chang, E. S. Jung, H. K. Kim, S. Y. Jin, J. H. Choi,
M. J. Gu, S. Kim, M. S. Kang, C. H. Cho, M. I. Park, Y. K. Kang, Y. W. Kim,
S. O. Yoon, H. I. Bae, M. Joo, W. S. Moon, D. Y. Kang, and S. J. Chang (2012).
“Current trends of the incidence and pathological diagnosis of gastroenteropancreatic
neuroendocrine tumors (GEP-NETs) in Korea 2000-2009: Multicenter study”. In: Can-
cer Research and Treatment 44.3, pp. 157–165. issn: 15982998.
Chojnacki, C., M. Wisniewska-Jarosinska, G. Kulig, I. Majsterek, R. J. Reiter, and J.
Chojnacki (2013). “Evaluation of enterochromaffin cells and melatonin secretion ex-
ponents in ulcerative colitis”. In: World Journal of Gastroenterology 19.23, pp. 3602–
3607. issn: 10079327.
Christopher, A., R. Kaur, G. Kaur, A. Kaur, V. Gupta, and P. Bansal (2016). “Mi-
croRNA therapeutics: Discovering novel targets and developing specific therapy”. In:
Perspectives in Clinical Research 7.2, p. 68. issn: 2229-3485.
362
Cives, M. and J. Strosberg (2015). “The expanding role of somatostatin analogs in gas-
troenteropancreatic and lung neuroendocrine tumors”. In: Drugs 75.8, pp. 847–858.
issn: 11791950.
Cives, M., H. P. Soares, and J. Strosberg (2016). “Will clinical heterogeneity of neuroen-
docrine tumors impact their management in the future? Lessons from recent trials.”
In: Current opinion in oncology 28.4, pp. 359–66. issn: 1531-703X.
Clift, A. K., P. Drymousis, A. Al-Nahhas, H. Wasan, J. Martin, S. Holm, and A. Frilling
(2015). “Incidence of Second Primary Malignancies in Patients with Neuroendocrine
Tumours”. In: Neuroendocrinology 102.1-2, pp. 26–32. issn: 0028-3835.
Clift, A. K., O. Faiz, A. Al-Nahhas, A. Bockisch, M. O. Liedke, E. Schloericke, H. Wasan,
J. Martin, P. Ziprin, K. Moorthy, and A. Frilling (2016). “Role of Staging in Pa-
tients with Small Intestinal Neuroendocrine Tumours”. In: Journal of Gastrointestinal
Surgery 20.1, pp. 180–188. issn: 1091-255X.
Cohen, S. J., C. J. A. Punt, N. Iannotti, B. H. Saidman, K. D. Sabbath, N. Y. Gabrail, J.
Picus, M. Morse, E. Mitchell, M. C. Miller, G. V. Doyle, H. Tissing, L. W.M. M. Ter-
stappen, and N. J. Meropol (2008a). “Relationship of circulating tumor cells to tumor
response, progression-free survival, and overall survival in patients with metastatic col-
orectal cancer”. In: Journal of Clinical Oncology 26.19, pp. 3213–3221. issn: 0732183X.
Cohen, S. J., C. J. A. Punt, N. Iannotti, B. H. Saidman, K. D. Sabbath, N. Y. Gabrail,
J. Picus, M. Morse, E. Mitchell, M. C. Miller, G. V. Doyle, and H. Tissing (2008b).
“Relationship of circulating tumor cells to tumor response, progression-free survival,
and overall survival in patients with metastatic colorectal cancer.” In: Journal of clinical
oncology : official journal of the American Society of Clinical Oncology 26.19, pp. 3213–
21. issn: 1527-7755.
363
Cortez, E., H. Gladh, S. Braun, M. Bocci, E. Cordero, N. K. Bjorkstrom, H. Miyazaki, I. P.
Michael, U. Eriksson, E. Folestad, and K. Pietras (2016). “Functional malignant cell
heterogeneity in pancreatic neuroendocrine tumors revealed by targeting of PDGF-
DD”. In: Proceedings of the National Academy of Sciences 113.7, E864–E873. issn:
0027-8424.
Costa, F., E. Domenichini, G. Garavito, R. Medrano, G. Mendez, J. O’Connor, W. Rojas,
S. Torres, R. N. Younes, G. Delle Fave, and K. Oberg (2008). “Management of neu-
roendocrine tumors: A meeting of experts from Latin America”. In: Neuroendocrinology
88.3, pp. 235–242. issn: 00283835.
Costedio, M. M., N. Hyman, and G. M. Mawe (2007). “Serotonin and its role in colonic
function and in gastrointestinal disorders”. In: Diseases of the Colon and Rectum 50.3,
pp. 376–388. issn: 00123706.
Courtney, M., E. Gjernes, N. Druelle, C. Ravaud, A. Vieira, N. Ben-Othman, A. Pfeifer, F.
Avolio, G. Leuckx, S. Lacas-Gervais, F. Burel-Vandenbos, D. Ambrosetti, J. Hecksher-
Sorensen, P. Ravassard, H. Heimberg, A. Mansouri, and P. Collombat (2013). “The
Inactivation of Arx in Pancreatic alpha-Cells Triggers Their Neogenesis and Conversion
into Functional beta-Like Cells”. In: PLoS Genetics 9.10, pp. 1–18. issn: 15537390.
Couvelard, A., L. Deschamps, P. Ravaud, G. Baron, A. Sauvanet, O. Hentic, N. Colnot,
V. Paradis, J. Belghiti, P. Bedossa, and P. Ruszniewski (2009). “Heterogeneity of tumor
prognostic markers: a reproducibility study applied to liver metastases of pancreatic
endocrine tumors”. In: Modern Pathology 22.2, pp. 273–281. issn: 0893-3952.
Cristofanilli, M., T. Budd, M. Ellis, A. Stopeck, J. Matera, M. C. Miller, J. M. Reuben,
G. V. Doyle, W. J. Allard, L. W.M. M. Terstappen, and D. F. Hayes (2004). “Circu-
lating Tumor Cells, Disease Progression, and Survival in Metastatic Breast Cancer”.
In: The New England Journal of Medicine 351.August, pp. 781–791. issn: 0028-4793.
364
Crona, J. and B. Skogseid (2016). “GEP- NETS UPDATE: Genetics of neuroendocrine
tumors”. In: European Journal of Endocrinology 174.6, R275–R290. issn: 0804-4643.
Cross, A. J., A. R. Hollenbeck, and Y. Park (2013). “A large prospective study of
risk factors for adenocarcinomas and malignant carcinoid tumors of the small intes-
tine.” In: Cancer causes & control : CCC 24.9, pp. 1737–46. issn: 1573-7225. arXiv:
NIHMS150003.
Crowley, E., F. Di Nicolantonio, F. Loupakis, and A. Bardelli (2013). “Liquid biopsy:
monitoring cancer-genetics in the blood.” In: Nature reviews. Clinical oncology 10.8,
pp. 472–84. issn: 1759-4782.
Cunningham, J., T Dıaz de Stahl, T Sjoblom, G Westin, J. Dumanski, and E. Janson
(2011). “Common pathogenetic mechanism involving human chromosome 18 in familial
and sporadic ileal carcinoid tumors”. In: Genes Chromosomes Cancer. 50.2, pp. 82–94.
Cutz, E (1982). “Neuroendocrine cells of the lung. An overview of morphologic charac-
teristics and development.” In: Experimental lung research 3.3-4, pp. 185–208. issn:
0190-2148.
Cuylen, S., C. Blaukopf, A. Z. Politi, T. Muller-Reichert, B. Neumann, I. Poser, J. Ellen-
berg, A. A. Hyman, and D. W. Gerlich (2016). “Ki-67 acts as a biological surfactant
to disperse mitotic chromosomes”. In: Nature 535.7611, pp. 308–312. issn: 0028-0836.
D’amico, M. A., B. Ghinassi, P. Izzicupo, L. Manzoli, and A Di Baldassarre (2014).
“Biological function and clinical relevance of chromogranin A and derived peptides.”
In: Endocrine connections 3.2, R45–54. issn: 2049-3614.
Davies, L. and M. O. Weickert (2016). “Gastroenteropancreatic neuroendocrine tumours:
an overview”. In: British Journal of Nursing 25.4, S12–15.
365
De Baere, T., F. Deschamps, L. Tselikas, M. Ducreux, D. Planchard, E. Pearson, A.
Berdelou, S. Leboulleux, D. Elias, and E. Baudin (2015). “GEP-NETs update: Inter-
ventional radiology: Role in the treatment of liver metastases from GEP-NETs”. In:
European Journal of Endocrinology 172.4, R151–R166. issn: 1479683X.
De Bono, J. S., H. I. Scher, R. B. Montgomery, C. Parker, M. C. Miller, H. Tissing, G. V.
Doyle, L. W.W. M. Terstappen, K. J. Pienta, and D. Raghavan (2008). “Circulating
tumor cells predict survival benefit from treatment in metastatic castration-resistant
prostate cancer”. In: Clinical Cancer Research 14.19, pp. 6302–6309. issn: 10780432.
Dejeux, E., J. A. Rønneberg, H. Solvang, I. Bukholm, S. Geisler, T. Aas, I. G. Gut, A.-L.
Børresen-Dale, P. E. Lønning, V. N. Kristensen, and J. Tost (2010). “DNA methylation
profiling in doxorubicin treated primary locally advanced breast tumours identifies
novel genes associated with survival and treatment response.” In: Molecular cancer
9.68. issn: 1476-4598.
Delle Fave, G., D. O’Toole, A. Sundin, B. Taal, P. Ferolla, J. K. Ramage, D. Ferone,
T. Ito, W. Weber, Z. Zheng-Pei, W. W. De Herder, A. Pascher, and P. Ruszniewski
(2016). “ENETS Consensus Guidelines Update for Gastroduodenal Neuroendocrine
Neoplasms”. In: Neuroendocrinology 103.2, pp. 119–124. issn: 14230194.
Desai, S., Z. Loomis, A. Pugh-Bernard, J. Schrunk, M. J. Doyle, A. Minic, E. McCoy, and
L. Sussel (2008). “Nkx2.2 regulates cell fate choice in the enteroendocrine cell lineages
of the intestine”. In: Developmental Biology 313.1, pp. 58–66. issn: 00121606.
Diaz, L. A. and A. Bardelli (2014). “Liquid biopsies: genotyping circulating tumor DNA.”
In: Journal of clinical oncology : official journal of the American Society of Clinical
Oncology 32.6, pp. 579–86. issn: 1527-7755.
366
Dıez, M., A. Teule, and R. Salazar (2013). “Gastroenteropancreatic neuroendocrine tu-
mors: diagnosis and treatment”. In: Annals of Gastroenterology : Quarterly Publication
of the Hellenic Society of Gastroenterology 26.1, pp. 29–36. issn: 1108-7471.
DiGruccio, M. R., A. M. Mawla, C. J. Donaldson, G. M. Noguchi, J. Vaughan, C. Cowing-
Zitron, T. van der Meulen, and M. O. Huising (2016). “Comprehensive alpha, beta
and delta cell transcriptomes reveal that ghrelin selectively activates delta cells and
promotes somatostatin release from pancreatic islets”. In: Molecular Metabolism 5.7,
pp. 449–458. issn: 22128778.
Dimitriadis, G. K., M. O. Weickert, H. S. Randeva, G. Kaltsas, and A. Grossman (2016).
“Medical management of secretory syndromes related to gastroenteropancreatic neu-
roendocrine tumours”. In: Endocrine-Related Cancer 23.9, R423–436. issn: 1479-6821.
Dromain, C., T. de Baere, J. Lumbroso, H. Caillet, A. Laplanche, V. Boige, M. Ducreux,
P. Duvillard, D. Elias, M. Schlumberger, R. Sigal, and E. Baudin (2005). “Detection
of liver metastases from endocrine tumors: A prospective comparison of somatostatin
receptor scintigraphy, computed tomography, and magnetic resonance imaging”. In:
Journal of Clinical Oncology 23.1, pp. 70–78. issn: 0732183X.
Druce, M. R., N. Bharwani, S. A. Akker, W. M. Drake, A. Rockall, and A. B. Grossman
(2010). “Intra-abdominal fibrosis in a recent cohort of patients with neuroendocrine
(’carcinoid’) tumours of the small bowel”. In: QJM 103.3, pp. 177–185. issn: 14602725.
Du, Y., M. Ter-Minassian, L. Brais, N. Brooks, A. Waldron, J. A. Chan, X. Lin, P. Kraft,
D. C. Christiani, and M. H. Kulke (2016). “Genetic associations with neuroendocrine
tumor risk: results from a genome-wide association study”. In: Endocrine-Related Can-
cer 23.8, pp. 587–594. issn: 1351-0088.
Du Rieu, M. C., J. Torrisani, J. Selves, T. Al Saati, A. Souque, M. Dufresne, G. J.
Tsongalis, A. A. Suriawinata, N. Carrere, L. Buscail, and P. Cordelier (2010). “MicroRNA-
367
21 is induced early in pancreatic ductal adenocarcinoma precursor lesions”. In: Clinical
Chemistry 56.4, pp. 603–612. issn: 00099147.
Duffy, M. J. and J. Crown (2013). “Companion biomarkers: paving the pathway to per-
sonalized treatment for cancer.” In: Clinical chemistry 59.10, pp. 1447–1456. issn:
15308561.
Edfeldt, K., P. Bjorklund, G. Akerstrom, G. Westin, P. Hellman, and P. Stalberg (2011).
“Different gene expression profiles in metastasizing midgut carcinoid tumors.” In:
Endocrine-related cancer 18.4, pp. 479–89. issn: 1479-6821.
Ellis, L., M. J. Shale, and M. P. Coleman (2010). “Carcinoid tumors of the gastroin-
testinal tract: trends in incidence in England since 1971.” In: The American journal of
gastroenterology 105.12, pp. 2563–9. issn: 1572-0241.
Engelstoft, M. S., K. L. Egerod, M. L. Lund, and T. W. Schwartz (2013). “Enteroen-
docrine cell types revisited”. In: Current Opinion in Pharmacology 13.6, pp. 912–921.
issn: 14714892.
Eriksson, B., G uuml nter Kl ouml ppel, E. Krenning, H. Ahlman, U. Pl ouml ckinger,
B. Wiedenmann, R. Arnold, C. Auernhammer, M. K ouml rner, G. Rindi, and S.
Wildi (2008). “Consensus guidelines for the management of patients with digestive neu-
roendocrine tumors–well-differentiated jejunal-ileal tumor/carcinoma.” In: Neuroen-
docrinology 87.1, pp. 8–19. issn: 1423-0194.
Erspamer, V (1957). “Occurrence and distribution of 5-hydroxytryptamine (enteramine)
in the living organism.” In: Z Vitam Horm Fermentforsch. 9.1-2, pp. 74–96.
Esquela-Kerscher, A. and F. J. Slack (2006). “Oncomirs - microRNAs with a role in
cancer.” In: Nature reviews. Cancer 6.4, pp. 259–69. issn: 1474-175X.
368
Essand, M., K. Oberg, and J. Leja-Jarblad (2016). New cancer treatment to be tested,
2016-03-09.
Essen, M. van, A. Sundin, E. P. Krenning, and D. J. Kwekkeboom (2014). “Neuroen-
docrine tumours: the role of imaging for diagnosis and therapy.” In: Nature reviews.
Endocrinology 10.2, pp. 102–14. issn: 1759-5037.
Faggiano, A., P. Ferolla, F. Grimaldi, D. Campana, M. Manzoni, M. V. Davı, A. Bianchi,
R. Valcavi, E. Papini, D. Giuffrida, D. Ferone, G. Fanciulli, G. Arnaldi, G. M. Franchi,
G. Francia, G. Fasola, L. Crino, A. Pontecorvi, P. Tomassetti, and A. Colao (2012).
“Natural history of gastro-entero-pancreatic and thoracic neuroendocrine tumors. Data
froma large prospective and retrospective Italian Epidemiological study: The net man-
agement study”. In: Journal of Endocrinological Investigation 35.9, pp. 817–823. issn:
03914097.
Falconi, M., B. Eriksson, G. Kaltsas, D. K. Bartsch, J. Capdevila, M. Caplin, B. Kos-
Kudla, D. Kwekkeboom, G. Rindi, G. Kloppel, N. Reed, R. Kianmanesh, and R. T.
Jensen (2016). “ENETS Consensus Guidelines Update for the Management of Patients
with Functional Pancreatic Neuroendocrine Tumors and Non-Functional Pancreatic
Neuroendocrine Tumors.” In: Neuroendocrinology 103.2, pp. 153–171. issn: 14230194.
Falconi, M., D. K. Bartsch, B. Eriksson, G. Kloppel, J. M. Lopes, J. M. O’Connor, R.
Salazar, B. G. Taal, M. P. Vullierme, and D. O’Toole (2012). “ENETS Consensus
Guidelines for the Management of Patients with Digestive Neuroendocrine Neoplasms
of the Digestive System: Well-Differentiated Pancreatic Non-Functioning Tumors.” In:
Neuroendocrinology 95.2, pp. 120–134. issn: 1423-0194.
Fan, S. T., Y. P. Le Treut, V. Mazzaferro, A. K. Burroughs, M. Olausson, S. Breitenstein,
and A. Frilling (2015). “Liver transplantation for neuroendocrine tumour liver metas-
369
tases”. In: HPB : the official journal of the International Hepato Pancreato Biliary
Association 17.1, pp. 23–28. issn: 14772574.
Fatehullah, A., S. H. Tan, and N. Barker (2016). “Organoids as an in vitro model of
human development and disease”. In: Nature Cell Biology 18.3, pp. 246–254. issn:
14764679.
Fazio, N. and M. Milione (2016). “Heterogeneity of grade 3 gastroenteropancreatic neu-
roendocrine carcinomas: New insights and treatment implications”. In: Cancer Treat-
ment Reviews 50, pp. 61–67. issn: 15321967.
Foroutan, B. (2015). “Personalized Medicine: A Review with Regard to Biomarkers”. In:
Journal of Bioequivalence & Bioavailability 07.06, pp. 244–256. issn: 09750851.
Fotouhi, O., M. Fahmideh, Adel, M. Kjellman, L. Sulaiman, A. Hoog, J. Zedenius, J.
Hashemi, and C. Larsson (2014). “Global hypomethylation and promoter methylation
in small intestinal neuroendocrine tumors: An in vivo and in vitro study”. In: Epige-
netics 9.7, pp. 987–97.
Fraenkel, M., M. Kim, A. Faggiano, W. W. De Herder, and G. D. Valk (2014). “Inci-
dence of gastroenteropancreatic neuroendocrine tumours: A systematic review of the
literature”. In: Endocrine-Related Cancer 21.3, pp. 153–163. issn: 14796821.
Frampton, A. E., L. Castellano, T. Colombo, E. Giovannetti, J. Krell, J. Jacob, L. Pel-
legrino, L. Roca-Alonso, N. Funel, T. M. H. Gall, A. De Giorgio, F. G. Pinho, V.
Fulci, D. J. Britton, R. Ahmad, N. a. Habib, R. C. Coombes, V. Harding, T. Knosel,
J. Stebbing, and L. R. Jiao (2014). “MicroRNAs Cooperatively Inhibit a Network of
Tumor Suppressor Genes to Promote Pancreatic Tumor Growth and Progression.” In:
Gastroenterology 146.1, pp. 268–77. issn: 1528-0012.
370
Francis, J. M., A. Kiezun, A. H. Ramos, S. Serra, C. S. Pedamallu, Z. R. Qian, M. S.
Banck, R. Kanwar, A. a. Kulkarni, A. Karpathakis, V. Manzo, T. Contractor, J. Philips,
E. Nickerson, N. Pho, S. M. Hooshmand, L. K. Brais, M. S. Lawrence, T. Pugh, A.
McKenna, A. Sivachenko, K. Cibulskis, S. L. Carter, A. I. Ojesina, S. Freeman, R. T.
Jones, D. Voet, G. Saksena, D. Auclair, R. Onofrio, E. Shefler, C. Sougnez, J. Grimsby,
L. Green, N. Lennon, T. Meyer, M. Caplin, D. C. Chung, A. S. Beutler, S. Ogino, C.
Thirlwell, R. Shivdasani, S. L. Asa, C. R. Harris, G. Getz, M. Kulke, and M. Meyerson
(2013). “Somatic mutation of CDKN1B in small intestine neuroendocrine tumors.” In:
Nature genetics 45.12, pp. 1483–6. issn: 1546-1718.
Friedman, E. B., S. Shang, E. V.-S. de Miera, J. U. Fog, M. W. Teilum, M. W. Ma, R. S.
Berman, R. L. Shapiro, A. C. Pavlick, E. Hernando, A. Baker, Y. Shao, and I. Osman
(2012). “Serum microRNAs as biomarkers for recurrence in melanoma.” In: Journal of
translational medicine 10, p. 155. issn: 1479-5876.
Friedman, R. C., K. K. H. Farh, C. B. Burge, and D. P. Bartel (2009). “Most mammalian
mRNAs are conserved targets of microRNAs”. In: Genome Research 19, pp. 92–105.
issn: 10889051.
Friedman-Mazursky, O., R. Elkon, and S. Efrat (2016). “Redifferentiation of expanded
human islet β cells by inhibition of ARX.” In: Scientific reports 6.January, p. 20698.
issn: 2045-2322.
Frilling, A., J. Li, E. Malamutmann, K. W. Schmid, A. Bockisch, and C. E. Broelsch
(2009). “Treatment of liver metastases from neuroendocrine tumours in relation to
the extent of hepatic disease”. In: British Journal of Surgery 96.2, pp. 175–184. issn:
00071323.
Frilling, A. and A. K. Clift (2015). “Therapeutic strategies for neuroendocrine liver metas-
tases”. In: Cancer 121.8, pp. 1172–1186. issn: 0008543X.
371
Frilling, A., G. Akerstrom, M. Falconi, M. Pavel, J. Ramos, M. Kidd, and I. M. Modlin
(2012). “Neuroendocrine tumor disease: an evolving landscape.” In: Endocrine-related
cancer 19.5, R163–85. issn: 1479-6821.
Frilling, A., I. M. Modlin, M. Kidd, C. Russell, S. Breitenstein, R. Salem, D. Kwekkeboom,
W.-y. Lau, C. Klersy, V. Vilgrain, B. Davidson, M. Siegler, M. Caplin, E. Solcia, and R.
Schilsky (2014). “Recommendations for management of patients with neuroendocrine
liver metastases.” In: The lancet oncology 15.1, e8–21. issn: 1474-5488.
Fujimori, Y., T. Fujimori, J. Imura, T. Sugai, T. Yao, R. Wada, Y. Ajioka, and Y. Ohkura
(2012). “An assessment of the diagnostic criteria for sessile serrated adenoma/polyps:
SSA/Ps using image processing software analysis for Ki67 immunohistochemistry.” In:
Diagnostic pathology 7, p. 59. issn: 1746-1596.
Furness, J. B., C. Jones, K. Nurgali, and N. Clerc (2004). “Intrinsic primary afferent
neurons and nerve circuits within the intestine”. In: Progress in Neurobiology 72.2,
pp. 143–164. issn: 03010082.
Gabriel, M., C. Decristoforo, D. Kendler, G. Dobrozemsky, D. Heute, C. Uprimny, P.
Kovacs, E. Von Guggenberg, R. Bale, and I. J. Virgolini (2007). “68Ga-DOTA-Tyr3-
octreotide PET in neuroendocrine tumors: comparison with somatostatin receptor
scintigraphy and CT.” In: Journal of nuclear medicine : official publication, Society
of Nuclear Medicine 48.4, pp. 508–18. issn: 0161-5505.
Galligan, J. J. (2017). “5-HT secretion by enterochromaffin cells is a very touching story”.
In: The Journal of Physiology 595.1, p. 3. issn: 00223751.
Garcia-Carbonero, R., J. Capdevila, G. Crespo-Herrero, J. A. Dıaz-Perez, M. P. Martınez
del Prado, V. Alonso Orduna, I. Sevilla-Garcıa, C. Villabona-Artero, A. Beguiristain-
Gomez, M. Llanos-Munoz, M. Marazuela, C. Alvarez-Escola, D. Castellano, E. Vilar,
P. Jimenez-Fonseca, A. Teule, J. Sastre-Valera, M. Benavent-Vinuelas, A. Monleon,
372
and R. Salazar (2010). “Incidence, patterns of care and prognostic factors for outcome
of gastroenteropancreatic neuroendocrine tumors (GEP-NETs): Results from the Na-
tional Cancer Registry of Spain (RGETNE)”. In: Annals of Oncology 21.9, pp. 1794–
1803. issn: 09237534.
Garcia-Carbonero, R., H. Sorbye, E. Baudin, E. Raymond, B. Wiedenmann, B. Niederle,
E. Sedlackova, C. Toumpanakis, M. Anlauf, J. Cwikla, M. Caplin, D. O’Toole, and
A. Perren (2016). “ENETS Consensus Guidelines for High-Grade Gastroenteropancre-
atic Neuroendocrine Tumors and Neuroendocrine Carcinomas”. In: Neuroendocrinology
103.2, pp. 186–194. issn: 14230194.
Gedde-Dahl, M., E. Thiis-Evensen, A. M. Tjølsen, K. S. Mordal, M. Vatn, and D. S.
Bergestuen (2013). “Comparison of 24-h and overnight samples of urinary 5-hydroxyindoleacetic
acid in patients with intestinal neuroendocrine tumors.” In: Endocrine connections 2.1,
pp. 50–4. issn: 2049-3614.
Geiss, G. K., R. E. Bumgarner, B. Birditt, T. Dahl, N. Dowidar, D. L. Dunaway, H. P.
Fell, S. Ferree, R. D. George, T. Grogan, J. J. James, M. Maysuria, J. D. Mitton, P.
Oliveri, J. L. Osborn, T. Peng, A. L. Ratcliffe, P. J. Webster, E. H. Davidson, L. Hood,
and K. Dimitrov (2008). “Direct multiplexed measurement of gene expression with
color-coded probe pairs.” In: Nature biotechnology 26.3, pp. 317–25. issn: 1546-1696.
Genomics England (2017). The 100,000 Genomes Project.
Gerdes, J, U Schwab, H Lemke, and H Stein (1983). “Production of a mouse monoclonal
antibody reactive with a human nuclear antigen associated with cell proliferation”. In:
Int J Cancer 31.1, pp. 13–20.
Gerlinger, M, A. Rowan, S. Horswell, J. Larkin, D. Endesfelder, E. Gronroos, P. Martinez,
N Matthews, A Stewart, P Tarpey, I Varela, B Phillimore, S Begum, N. McDonald, A
Butler, D Jones, K Raine, C Latimer, C. Santos, M Nohadani, A. Eklund, B Spencer-
373
Dene, G Clark, L Pickering, G Stamp, M Gore, Z Szallasi, J Downward, P. Futreal,
and C. Swanton (2012). “Intratumor heterogeneity and branched evolution revealed by
multiregion sequencing”. In: N Engl J Med 366.10, pp. 883–92.
Gershon, M. D. (2004). “Review article: serotonin receptors and transporters – roles
in normal and abnormal gastrointestinal motility.” In: Alimentary pharmacology &
therapeutics 20 Suppl 7, pp. 3–14. issn: 0269-2813.
Giovinazzo, F., S. Schimmack, B. Svejda, D. Alaimo, R. Pfragner, I. Modlin, and M.
Kidd (2013). “Chromogranin A and its fragments as regulators of small intestinal
neuroendocrine neoplasm proliferation.” In: PloS one 8.11, e81111. issn: 1932-6203.
Gold, P and S. O. Freedman (1965). “Demonstration of Tumor-Specific Antigens in Hu-
man Colonic Carcinomata By Immunological Tolerance and Absorption Techniques.”
In: The Journal of experimental medicine 121, pp. 439–62. issn: 0022-1007.
Goldstein, L. J., R. Gray, S. Badve, B. H. Childs, C. Yoshizawa, S. Rowley, S. Shak, F. L.
Baehner, P. M. Ravdin, N. E. Davidson, G. W. Sledge, E. A. Perez, L. N. Shulman, S.
Martino, and J. A. Sparano (2008). “Prognostic utility of the 21-gene assay in hormone
receptor-positive operable breast cancer compared with classical clinicopathologic fea-
tures”. In: Journal of Clinical Oncology 26.25, pp. 4063–4071. issn: 0732183X.
Gosset, A. and P. Masson (1914). “Tumeurs endocrines de l’appendice.” In: Presse Med.
25, pp. 237–240.
Gradwohl, G, a Dierich, M LeMeur, and F Guillemot (2000). “Neurogenin3 Is Required for
the Development of the Four Endocrine Cell Lineages of the Pancreas.” In: Proceedings
of the National Academy of Sciences of the United States of America 97.4, pp. 1607–
1611. issn: 0027-8424.
374
Grajo, J. R., R. M. Paspulati, D. V. Sahani, and A. Kambadakone (2016). “Multiple
Endocrine Neoplasia Syndromes: A Comprehensive Imaging Review”. In: Radiologic
Clinics of North America 54.3, pp. 441–451. issn: 15578275.
Griniatsos, J. and O. Michail (2010). “Appendiceal neuroendocrine tumors: Recent in-
sights and clinical implications”. In: World J Gastrointest Oncol 2.4, pp. 192–196. issn:
1948-5204.
Gross, S., D. C. Garofalo, D. A. Balderes, T. L. Mastracci, J. M. Dias, T. Perlmann,
J. Ericson, and L. Sussel (2016). “The novel enterochromaffin marker Lmx1a regu-
lates serotonin biosynthesis in enteroendocrine cell lineages downstream of Nkx2.2”.
In: Development. issn: 0950-1991.
Grozinsky-Glasberg, S., I. Shimon, and H. Rubinfeld (2012). “The role of cell lines in
the study of neuroendocrine tumors”. In: Neuroendocrinology 96.3, pp. 173–187. issn:
00283835.
Grun, D., A. Lyubimova, L. Kester, K. Wiebrands, O. Basak, N. Sasaki, H. Clevers,
and A. van Oudenaarden (2015). “Single-cell messenger RNA sequencing reveals rare
intestinal cell types”. In: Nature 525.7568, pp. 251–5. issn: 0028-0836.
Gu, G., J. Dubauskaite, and D. A. Melton (2002). “Direct evidence for the pancreatic
lineage: NGN3+ cells are islet progenitors and are distinct from duct progenitors.” In:
Development (Cambridge, England) 129.10, pp. 2447–2457. issn: 0950-1991.
Guo, Z., M. Maki, R. Ding, Y. Yang, B. Zhang, and L. Xiong (2015). “Genome-wide
survey of tissue-specific microRNA and transcription factor regulatory networks in 12
tissues”. In: Scientific Reports 4.1, p. 5150. issn: 2045-2322.
Gut, P., A. Czarnywojtek, J. Fischbach, M. Baczyk, K. Ziemnicka, E. Wrotkowska, M.
Gryczynska, and M. Rucha la (2016). “Chromogranin A - unspecific neuroendocrine
375
marker. Clinical utility and potential diagnostic pitfalls.” In: Archives of medical science
: AMS 12.1, pp. 1–9. issn: 1734-1922.
Halfdanarson, T., W. Bamlet, R. McWilliams, T. Hobday, P. Burch, K. Rabe, and G.
Petersen (2014). “Risk Factors for Pancreatic Neuroendocrine Tumors (PNETs): A
Clinic-Based Case-Control study”. In: Pancreas 43.8, pp. 1219–22. issn: 1946-6242.
arXiv: NIHMS150003.
Hallet, J., C. H. L. Law, M. Cukier, R. Saskin, N. Liu, and S. Singh (2015). “Exploring
the rising incidence of neuroendocrine tumors: A population-based analysis of epidemi-
ology, metastatic presentation, and outcomes”. In: Cancer 121.4, pp. 589–597. issn:
10970142.
Hamam, R., D. Hamam, K. A. Alsaleh, M. Kassem, W. Zaher, M. Alfayez, A. Aldahmash,
and N. M. Alajez (2017). “Circulating microRNAs in breast cancer: novel diagnostic
and prognostic biomarkers”. In: Cell Death and Disease 8.9, e3045. issn: 2041-4889.
Hanahan, D. and R. A. Weinberg (2011). “Hallmarks of cancer: The next generation”.
In: Cell 144.5, pp. 646–674. issn: 00928674. arXiv: 0208024 [gr-qc].
Hanahan D. (1985). “Heritable formation of pancreatic beta-cell tumours in transgenic
mice expressing recombinant insulin/simian virus 40 oncogenes.” In: Nature 315.6015,
pp. 115–22. issn: 00280836.
Hansen, M. B. and A. B. Witte (2008). “The role of serotonin in intestinal luminal sensing
and secretion”. In: Acta Physiologica 193.4, pp. 311–323. issn: 17481708.
Harris, L., H. Fritsche, R. Mennel, L. Norton, P. Ravdin, S. Taube, M. R. Somerfield, D. F.
Hayes, and R. C. Bast (2007). “American society of clinical oncology 2007 update of
recommendations for the use of tumor markers in breast cancer”. In: Journal of Clinical
Oncology 25.33, pp. 5287–5312. issn: 0732183X.
376
Hashemi, J., O. Fotouhi, L. Sulaiman, M. Kjellman, A. Hoog, J. Zedenius, and C. Larsson
(2013). “Copy number alterations in small intestinal neuroendocrine tumors determined
by array comparative genomic hybridization.” In: BMC cancer 13, p. 505. issn: 1471-
2407.
Hassan, M. M., A Phan, D Li, C. G. Dagohoy, C Leary, and J. C. Yao (2008a). “Family
history of cancer and associated risk of developing neuroendocrine tumors: a case-
control study”. In: Cancer Epidemiol Biomarkers Prev 17.4, pp. 959–965. issn: 1055-
9965.
Hassan, M. M., A. Phan, D. Li, C. G. Dagohoy, C. Leary, and J. C. Yao (2008b). “Risk
factors associated with neuroendocrine tumors: A U.S.-based case-control study”. In:
International Journal of Cancer 123.4, pp. 867–873. issn: 00207136.
Hauge-Evans, A. C., A. J. King, D. Carmignac, C. C. Richardson, I. C.A. F. Robinson,
M. J. Low, M. R. Christie, S. J. Persaud, and P. M. Jones (2009). “Somatostatin se-
creted by islet delta-cells fulfills multiple roles as a paracrine regulator of islet function.”
In: Diabetes 58.2, pp. 403–11. issn: 1939-327X.
Haugvik, S.-P., R Valente, E Korsæth, D Siuka, P Hedenstrom, A Hayes, I. P. Gladhaug,
P Maisonneuve, B Lindkvist, and G Capurso (2015). “Smoking, alcohol consumption,
diabetes mellitus and family history of cancer as risk factors for sporadic pancreatic
neuroendocrine tumors: A systematic review”. In: Pancreatology 14.3, S80. issn: 1424-
3903.
Hauso, O., B. I. Gustafsson, M. Kidd, H. L. Waldum, I. Drozdov, A. K. C. Chan, and
I. M. Modlin (2008). “Neuroendocrine tumor epidemiology: contrasting Norway and
North America.” In: Cancer 113.10, pp. 2655–64. issn: 0008-543X.
Haynes, C. M., A. R. Sangoi, and R. K. Pai (2011). “PAX8 is expressed in pancreatic
well-differentiated neuroendocrine tumors and in extrapancreatic poorly differentiated
377
neuroendocrine carcinomas in fine-needle aspiration biopsy specimens.” In: Cancer cy-
topathology 119.3, pp. 193–201. issn: 1934-6638.
Heller, R. S., D. A. Stoffers, A. Liu, A. Schedl, E. B. Crenshaw, O. D. Madsen, and P.
Serup (2004). “The role of Brn4/Pou3f4 and Pax6 in forming the pancreatic glucagon
cell identity”. In: Developmental Biology 268.1, pp. 123–134. issn: 00121606.
Heller, R. S., M. Jenny, P. Collombat, A. Mansouri, C. Tomasetto, O. D. Madsen, G.
Mellitzer, G. Gradwohl, and P. Serup (2005). “Genetic determinants of pancreatic ε-cell
development”. In: Developmental Biology 286.1, pp. 217–224. issn: 00121606.
Helman, L. J., A. F. Gazdar, J. G. Park, P. S. Cohen, J. D. Cotelingam, and M. A. Israel
(1988). “Chromogranin A expression in normal and malignant human tissues”. In: J
Clin Invest 82.August, pp. 686–690.
Helwak, A., G. Kudla, T. Dudnakova, and D. Tollervey (2013). “Mapping the human
miRNA interactome by CLASH reveals frequent noncanonical binding”. In: Cell 153.3,
pp. 654–665. issn: 00928674.
Hemminki, K, X Li, A Yellin, D. A. Simansky, M Paley, and Y Refaely (2001). “Incidence
trends and risk factors of carcinoid tumors: a nationwide epidemiologic study from
Sweden.” In: Cancer 92.8, pp. 2204–10. issn: 0008-543X.
Henry, N. L. and D. F. Hayes (2012). “Cancer biomarkers”. In: Molecular Oncology 6.2,
pp. 140–146. issn: 15747891.
Hiripi, E., J. L. Bermejo, J. Sundquist, and K. Hemminki (2009). “Familial gastrointesti-
nal carcinoid tumours and associated cancers”. In: Annals of Oncology 20.5, pp. 950–
954. issn: 09237534.
Hobday, T. J., R. Qin, D. Reidy-Lagunes, M. J. Moore, J. Strosberg, A. Kaubisch, M.
Shah, H. L. Kindler, H. J. Lenz, H. Chen, and C. Erlichman (2015). “Multicenter
378
phase II trial of temsirolimus and bevacizumab in pancreatic neuroendocrine tumors”.
In: Journal of Clinical Oncology 33.14, pp. 1551–1556. issn: 15277755.
Hofmann, P. G., A. Baez Saldana, T. Fortoul Van Der Goes, M. Gonzalez del Pliego, and
G. Gutierrez Ospina (2013). “Neuroendocrine cells are present in the domestic fowl
ovary.” In: Journal of Anatomy 222.2, pp. 170–177. issn: 00218782.
Hsu, Y.-C. (2015). “Theory and Practice of Lineage Tracing”. In: Stem Cells 33.11,
pp. 3197–3204. issn: 15494918.
Hu, Y., Y. Ou, K. Wu, Y. Chen, and W. Sun (2012). “MiR-143 inhibits the metastasis
of pancreatic cancer and an associated signaling pathway.” In: Tumour Biology 33.6,
pp. 1863–1870. issn: 14230380.
Huang, D. W., B. T. Sherman, and R. A. Lempicki (2009a). “Bioinformatics enrichment
tools: Paths toward the comprehensive functional analysis of large gene lists”. In: Nu-
cleic Acids Research 37.1, pp. 1–13. issn: 03051048.
Huang, D. W., B. T. Sherman, and R. a. Lempicki (2009b). “Systematic and integrative
analysis of large gene lists using DAVID bioinformatics resources.” In: Nature protocols
4.2, pp. 44–57. issn: 1754-2189.
Iorio, M. V. and C. M. Croce (2012). “MicroRNA dysregulation in cancer: Diagnostics,
monitoring and therapeutics. A comprehensive review”. In: EMBO Molecular Medicine
4.3, pp. 143–159. issn: 17574676.
Ito, T, L Lee, and R. Jensen (2016). “Treatment of symptomatic neuroendocrine tumor
syndromes: recent advances and”. In: Expert Opin Pharmacother. 17.16, pp. 2191–2205.
issn: 1465-6566.
Ito, T., H. Sasano, M. Tanaka, R. Y. Osamura, I. Sasaki, W. Kimura, K. Takano, T.
Obara, M. Ishibashi, K. Nakao, R. Doi, A. Shimatsu, T. Nishida, I. Komoto, Y. Hirata,
379
K. Nakamura, H. Igarashi, R. T. Jensen, B. Wiedenmann, and M. Imamura (2010).
“Epidemiological study of gastroenteropancreatic neuroendocrine tumors in Japan”.
In: Journal of Gastroenterology 45.2, pp. 234–243. issn: 09441174.
Ito, T., H. Igarashi, H. Uehara, M. Berna J, and R. Jensen T (2013). “Causes of Death
and Prognostic Factors in Multiple Endocrine Neoplasia Type 1: A Prospective Study:
Comparison of 106 MEN1/Zollinger-Ellison Syndrome Patients With 1613 Literature
MEN1 Patients With or Without Pancreatic Endocrine Tumors.” In: Medicine 92.3,
pp. 135–181. issn: 0025-7974.
Jamali, M. and R. Chetty (2008). “Predicting prognosis in gastroentero-pancreatic neu-
roendocrine tumors: An overview and the value of Ki-67 immunostaining”. In: En-
docrine Pathology 19.4, pp. 282–288. issn: 10463976.
James, C. R., J. E. Quinn, P. B. Mullan, P. G. Johnston, and D. P. Harkin (2007).
“BRCA1, a Potential Predictive Biomarker in the Treatment of Breast Cancer”. In:
The Oncologist 12.2, pp. 142–150. issn: 1083-7159.
Jann, H., S. Roll, A. Couvelard, O. Hentic, M. Pavel, J. Muller-Nordhorn, M. Koch,
C. Rocken, G. Rindi, P. Ruszniewski, B. Wiedenmann, and U.-F. Pape (2011). “Neu-
roendocrine tumors of midgut and hindgut origin: tumor-node-metastasis classification
determines clinical outcome.” In: Cancer 117.15, pp. 3332–41. issn: 1097-0142.
Jenny, M., C. Uhl, C. Roche, I. Duluc, V. Guillermin, F. Guillemot, J. Jensen, M.
Kedinger, and G. Gradwohl (2002). “Neurogenin3 is differentially required for endocrine
cell fate specification in the intestinal and gastric epithelium”. In: EMBO Journal 21.23,
pp. 6338–6347. issn: 02614189.
Jensen, J, E. E. Pedersen, P Galante, J Hald, R. S. Heller, M Ishibashi, R Kageyama,
F Guillemot, P Serup, and O. D. Madsen (2000). “Control of endodermal endocrine
development by Hes-1.” In: Nature genetics 24.1, pp. 36–44. issn: 1061-4036.
380
Jensen, R. T., M. J. Berna, D. B. Bingham, and J. A. Norton (2008). “Inherited pan-
creatic endocrine tumor syndromes: Advances in molecular pathogenesis, diagnosis,
management, and controversies”. In: Cancer 113.7, pp. 1807–1843. issn: 0008543X.
Jensen, R. T., G. Cadiot, M. L. Brandi, W. W. de Herder, G. Kaltsas, P. Komminoth,
J.-Y. Scoazec, R. Salazar, A. Sauvanet, and R. Kianmanesh (2012). “ENETS Consen-
sus Guidelines for the Management of Patients with Digestive Neuroendocrine Neo-
plasms: Functional Pancreatic Endocrine Tumor Syndromes.” In: Neuroendocrinology
95.2, pp. 98–119. issn: 1423-0194.
Jiao, Y., C. Shi, B. H. Edil, R. F. de Wilde, D. S. Klimstra, A. Maitra, R. D. Schulick,
L. H. Tang, C. L. Wolfgang, M. a. Choti, V. E. Velculescu, L. a. Diaz, B. Vogelstein,
K. W. Kinzler, R. H. Hruban, and N. Papadopoulos (2011). “DAXX/ATRX, MEN1,
and mTOR pathway genes are frequently altered in pancreatic neuroendocrine tumors.”
In: Science (New York, N.Y.) 331.6021, pp. 1199–203. issn: 1095-9203.
Jonas, S. and E. Izaurralde (2015). “Towards a molecular understanding of microRNA-
mediated gene silencing.” In: Nature reviews. Genetics 16.7, pp. 421–433. issn: 1471-
0064.
Jovanovic, J., J. A. Rønneberg, J. Tost, and V. Kristensen (2010). “The epigenetics of
breast cancer.” In: Molecular oncology 4.3, pp. 242–54. issn: 1878-0261.
Jung, M., A. Schaefer, I. Steiner, C. Kempkensteffen, C. Stephan, A. Erbersdobler, and K.
Jung (2010). “Robust MicroRNA stability in degraded RNA preparations from human
tissue and cell samples”. In: Clinical Chemistry 56.6, pp. 998–1006. issn: 00099147.
Jung, Y. S., K. E. Yun, Y Chang, S Ryu, J. H. Park, H. J. Kim, Y. K. Cho, C. I. Sohn,
W. K. Jeon, B. I. Kim, and D. I. Park (2014). “Risk factors associated with rectal
neuroendocrine tumors: a cross-sectional study”. In: Cancer Epidemiol Biomarkers
Prev 23.7, pp. 1406–1413. issn: 1055-9965.
381
Kaerlev, L., P. S. Teglbjaerg, S. Sabroe, H. A. Kolstad, W. Ahrens, M. Eriksson, P.
Guenel, L. Hardell, D. Cyr, T. Ballard, P. Zambon, M. M. Morales Suarez-Varela,
A. Stang, and J. Olsen (2002a). “Occupational risk factors for small bowel carcinoid
tumor: a European population-based case-control study.” In: Journal of Occupational
and Environmental Medicine 44.6, pp. 516–522. issn: 1076-2752 (Print).
Kaerlev, L., P. S. Teglbjaerg, S. Sabroe, H. A. Kolstad, W. Ahrens, M. Eriksson, P.
Guenel, G. Gorini, L. Hardell, S. C. Causes, N. Feb, L. Kaerlev, P. S. Teglbjaerg,
S. Sabroe, H. A. Kolstad, W. Ahrens, P. Guenel, G. Gorini, L. Hardell, D. Cyr, P.
Zambon, and A. Stanglo (2002b). “The Importance of Smoking and Medical History
for Development of Small Bowel Carcinoid Tumor : A European Population-Based
Case-Control Study”. In: Cancer Causes & Control 13.1, pp. 27–34.
Kanehisa, M. and S. Goto (2000). “KEGG: Kyoto encyclopedia of genes and genomes”. In:
Nucleic Acids Research 28.1, pp. 27–30. issn: 03051048. arXiv: arXiv:1011.1669v3.
Kanehisa, M., S. Goto, M. Furumichi, M. Tanabe, and M. Hirakawa (2010). “KEGG for
representation and analysis of molecular networks involving diseases and drugs”. In:
Nucleic Acids Research 38.SUPPL.1, pp. D355–D360. issn: 03051048.
Karpathakis, a, H Dibra, and C Thirlwell (2013). “Neuroendocrine tumours: cracking the
epigenetic code.” In: Endocrine-related cancer 20.3, R65–82. issn: 1479-6821.
Khan, M. S., T. V. Luong, J Watkins, C Toumpanakis, M. E. Caplin, and T Meyer
(2013a). “A comparison of Ki-67 and mitotic count as prognostic markers for metastatic
pancreatic and midgut neuroendocrine neoplasms”. In: British Journal of Cancer 108.9,
pp. 1838–1845. issn: 0007-0920.
Khan, M. S., T. Tsigani, M. Rashid, J. S. Rabouhans, D. Yu, T. V. Luong, M. Caplin, and
T. Meyer (2011). “Circulating tumor cells and EpCAM expression in neuroendocrine
382
tumors.” In: Clinical cancer research : an official journal of the American Association
for Cancer Research 17.2, pp. 337–45. issn: 1078-0432.
Khan, M. S., A. Kirkwood, T. Tsigani, J. Garcia-Hernandez, J. a. Hartley, M. E. Caplin,
and T. Meyer (2013b). “Circulating tumor cells as prognostic markers in neuroen-
docrine tumors.” In: Journal of clinical oncology : official journal of the American
Society of Clinical Oncology 31.3, pp. 365–72. issn: 1527-7755.
Kharazmi, E., E. Pukkala, K. Sundquist, and K. Hemminki (2013). “Familial risk of small
intestinal carcinoid and adenocarcinoma”. In: Clinical Gastroenterology and Hepatology
11.8, pp. 944–949. issn: 15423565.
Kidd, M., I. M. Modlin, and I. Drozdov (2014). “Gene network-based analysis identifies
two potential subtypes of small intestinal neuroendocrine tumors.” In: BMC genomics
15.1, p. 595. issn: 1471-2164.
Kim, D., Y. Nagano, and I. Choi (2008a). “Alterations in well-differentiated neuroen-
docrine tumors (carcinoid tumors) identified by genome-wide single nucleotide poly-
morphism analysis and comparison with Pancreatic Endocrine Tumors.” In: Genes
Chromosomes Cancer. 47.1, pp. 84–92.
Kim, D. H., Y. Nagano, I.-s. Choi, J. A. White, J. C. Yao, and A. Rashid (2008b).
“Allelic Alterations in Well Differentiated Neuroendocrine Tumors (Carcinoid Tumors)
Identified by Genome-Wide Single Nucleotide Polymorphism Analysis and Comparison
with Pancreatic Endocrine Tumors”. In: Genes Chromosomes Cancer. 47.1, pp. 84–92.
Kim, J. Y. and S. M. Hong (2016). “Recent updates on neuroendocrine tumors from the
gastrointestinal and pancreatobiliary tracts”. In: Archives of Pathology and Laboratory
Medicine 140.5, pp. 437–448. issn: 15432165.
383
Kim, S., Y. Lee, and J. S. Koo (2015a). “Differential expression of lipid metabolism-
related proteins in different breast cancer subtypes”. In: PLoS ONE 10.3, pp. 1–15.
issn: 19326203.
Kim, Y. H., H. L. Larsen, P. Rue, L. A. Lemaire, J. Ferrer, and A. Grapin-Botton (2015b).
“Cell Cycle–Dependent Differentiation Dynamics Balances Growth and Endocrine Dif-
ferentiation in the Pancreas”. In: PLoS Biology 13.3, pp. 1–26. issn: 15457885.
Kimura, W, A Kuroda, and Y Morioka (1991). “Clinical pathology of endocrine tumors
of the pancreas. Analysis of autopsy cases.” In: Digestive Diseases and Sciences 36.7,
pp. 933–42.
Kincaid, R. P. and C. S. Sullivan (2012). “Virus-Encoded microRNAs: An Overview and
a Look to the Future”. In: PLOS Pathogens 8.12, e1003018.
Kizilgul, M. and T. Delibasi (2015). “Gastroenteropancreatic neuroendocrine tumors
(GEP-NETs)”. In: Transl Gastrointest Cancer 4.1, pp. 39–56. issn: 0020-9554.
Klimstra, D. S., I. R. Modlin, D. Coppola, R. V. Lloyd, and S. Suster (2010). “The
pathologic classification of neuroendocrine tumors: a review of nomenclature, grading,
and staging systems.” In: Pancreas 39.6, pp. 707–12. issn: 1536-4828.
Knudson, A. G. (1971). “Mutation and Cancer: Statistical Study of Retinoblastoma”. In:
Proceedings of the National Academy of Sciences 68.4, pp. 820–823. issn: 0027-8424.
Knudson, A. G. (1974). “Heredity and human cancer.” In: The American journal of
pathology 77.1, pp. 77–84. issn: 0002-9440.
Kozomara, A. and S. Griffiths-Jones (2014). “MiRBase: Annotating high confidence mi-
croRNAs using deep sequencing data”. In: Nucleic Acids Research 42.D1, pp. 68–73.
issn: 03051048.
384
Krebs, M. G., J.-M. Hou, T. H. Ward, F. H. Blackhall, and C. Dive (2010). “Circulat-
ing tumour cells: their utility in cancer management and predicting outcomes”. In:
Therapeutic Advances in Medical Oncology 2.6, pp. 351–365. issn: 1758-8340.
Krebs, M. G., R. L. Metcalf, L. Carter, G. Brady, F. H. Blackhall, and C. Dive (2014).
“Molecular analysis of circulating tumour cells-biology and biomarkers.” In: Nature
reviews. Clinical oncology 11.3, pp. 129–44. issn: 1759-4782.
Kulkarni, M. M. (2011). “Digital multiplexed gene expression analysis using the nanos-
tring ncounter system”. In: Current Protocols in Molecular Biology April.Chapter 25,
doi: 10.1002/0471142727.mb25b10s94. issn: 19343639.
Kulke, M. H., E. Freed, D. Y. Chiang, J. Philips, D. Zahrieh, J. N. Glickman, and
R. A. Shivdasani (2008). “High-Resolution Analysis of Genetic Alterations in Small
Bowel Carcinoid Tumors Reveals Areas of Recurrent Amplification and Loss”. In: Genes
Chromosomes Cancer. Cancer. 47.7, pp. 591–603.
Kulke, M., D. Horsch, M. Caplin, L. Anthony, E. Bergsland, K. Oberg, S. Welin, R.
Warner, C. Lombard-Bohas, P. Kunz, E. Grande, J. Valle, D. Fleming, P. Lapuerta,
P. Banks, S. Jackson, D. Wheeler, B. Zambrowicz, A. Sands, and M. Pavel (2015).
“Telotristat etiprate is effective in treating patients with carcinoid syndrome that is in-
adequately controlled by somatostatin analog therapy (the phase 3 TELESTAR clinical
trial)”. In: European Journal of Cancer 51.(suppl 3), S728. issn: 09598049.
Kuninty, P. R., J. Schnittert, G. Storm, and J. Prakash (2016). “MicroRNA Targeting
to Modulate Tumor Microenvironment”. In: Frontiers in Oncology 6.January, pp. 1–8.
issn: 2234-943X.
Kurzynska-Kokorniak, A., N. Koralewska, M. Pokornowska, A. Urbanowicz, A. Tworak,
A. Mickiewicz, and M. Figlerowicz (2015). “The many faces of Dicer: The complexity
385
of the mechanisms regulating Dicer gene expression and enzyme activities”. In: Nucleic
Acids Research 43.9, pp. 4365–4380. issn: 13624962.
Kvols, L. K., K. E. Oberg, T. M. O’Dorisio, P. Mohideen, W. W. De Herder, R. Arnold,
K. Hu, Y. Zhang, G. Hughes, L. Anthony, and B. Wiedenmann (2012). “Pasireotide
(SOM230) shows efficacy and tolerability in the treatment of patients with advanced
neuroendocrine tumors refractory or resistant to octreotide LAR: Results from a phase
II study”. In: Endocrine-Related Cancer 19.5, pp. 657–666. issn: 13510088.
Kwekkeboom, D. J., E. P. Krenning, R. Lebtahi, P. Komminoth, B. Kos-Kud la, W. W.
De Herder, U. Plockinger, G. Akerstrom, B. Annibale, R. Arnold, E. Bajetta, J. Bark-
manova, Y. J. Chen, F. Costa, A. Couvelard, J. Davar, G. Delle Fave, B. Eriksson,
M. Falconi, D. Ferone, D. Gross, A. Grossman, B. Gustafsson, R. Hyrdel, D. Ivan,
G. Kaltsas, R. Kianmanesh, G. Kloppel, U. P. Knigge, V. Lewington, A. M. McNicol,
E. Mitry, O. Nilsson, K. Oberg, J. O’Connor, D. O’Toole, U. F. Pape, M. Papotti,
M. Pavel, A. Perren, M. Platania, G. Rindi, P. Ruszniewski, R. Salazar, A. Scarpa,
K. Scheidhauer, J. Y. Scoazec, A. Sundin, W. Szpak, B. Taal, P. Vitek, M. P. Vul-
lierme, and B. Wiedenmann (2009). “ENETS consensus guidelines for the standards of
care in neuroendocrine tumors: Peptide receptor radionuclide therapy with radiolabeled
somatostatin analogs”. In: Neuroendocrinology 90.2, pp. 220–226. issn: 00283835.
La Rosa, S, F Sessa, C Capella, C Riva, B. E. Leone, C Klersy, G Rindi, and E Solcia
(1996). “Prognostic criteria in nonfunctioning pancreatic endocrine tumours.” In: Vir-
chows Archiv : an international journal of pathology 429.6, pp. 323–33. issn: 0945-6317.
Lagos-Quintana, M, R Rauhut, W Lendeckel, and T Tuschl (2001). “Identification of
novel genes coding for small expressed RNAs.” In: Science (New York, N.Y.) 294.5543,
pp. 853–8. issn: 0036-8075.
386
Lam, K. Y. and C. Y. Lo (1997). “Pancreatic endocrine tumour: A 22-year clinico-
pathological experience with morphological, immunohistochemical observation and a
review of the literature”. In: European Journal of Surgical Oncology 23.1, pp. 36–42.
issn: 07487983.
Landerholm, K., N. Zar, R. E. Andersson, S. E. Falkmer, and J. Jarhult (2011). “Survival
and prognostic factors in patients with small bowel carcinoid tumour”. In: British
Journal of Surgery 98.11, pp. 1617–1624. issn: 00071323.
Landerholm, K., S. Falkmer, and J. Jarhult (2010). “Epidemiology of small bowel car-
cinoids in a defined population”. In: World Journal of Surgery 34.7, pp. 1500–1505.
issn: 03642313.
Landry, C. S., H. Y. Lin, A. Phan, C. Charnsangavej, E. K. Abdalla, T. Aloia, J. Nicolas
Vauthey, M. H. G. Katz, J. C. Yao, and J. B. Fleming (2013). “Resection of at-risk
mesenteric lymph nodes is associated with improved survival in patients with small
bowel neuroendocrine tumors”. In: World Journal of Surgery 37.7, pp. 1695–1700.
issn: 03642313.
Langhans, T (1867). “Ueber einen Drusenpolyp im Ileum”. In: Virchows Arch Pathol
Anat 38, pp. 550–560.
Larrea, E., C. Sole, L. Manterola, I. Goicoechea, M. Armesto, M. Arestin, M. M. Caffarel,
A. M. Araujo, M. Araiz, M. Fernandez-Mercado, and C. H. Lawrie (2016). “New con-
cepts in cancer biomarkers: Circulating miRNAs in liquid biopsies”. In: International
Journal of Molecular Sciences 17.5. issn: 14220067.
Lau, N. C., L. P. Lim, E. G. Weinstein, and D. P. Bartel (2001). “An abundant class
of tiny RNAs with probable regulatory roles in Caenorhabditis elegans.” In: Science
(New York, N.Y.) 294.5543, pp. 858–62. issn: 0036-8075.
387
Lawrence, B, B. Gustafsson, A Chan, B Svejda, M Kidd, and I. Modlin (2011). “The
epidemiology of gastroenteropancreatic neuroendocrine tumors.” In: Endocrinol Metab
Clin North Am 40.1, pp. 1–18.
Lee, A., D. L. Chan, M. H. Wong, B. T. Li, S. Lumba, S. Clarke, J. Samra, and N. Pavlakis
(2016). “Systematic Review on the Role of Targeted Therapy in Metastatic Neuroen-
docrine Tumor (NET)”. In: Neuroendocrinology Ci, [ahead of print]. issn: 0028-3835.
Lee, J and P. Pilch (1994). “The insulin receptor: structure, function, and signaling.” In:
Am J Physiol. 1994 Feb;266(2 Pt 1):C319-34. 266.2, pp. C319–34.
Lee, M. and N. S. Pellegata (2013). “Multiple endocrine neoplasia type 4”. In: Frontiers
of Hormone Research 41, pp. 63–78. issn: 03013073.
Lee, R. C. and V Ambros (2001). “An extensive class of small RNAs in Caenorhabditis
elegans.” In: Science (New York, N.Y.) 294.5543, pp. 862–4. issn: 0036-8075.
Lee, R. C., L. Feinbaum, Rhonda, and V. Ambros (1993). “The C . elegans Heterochronic
Gene lin-4 Encodes Small RNAs with Antisense Complementarity to & II-14”. In: Cell
75.5, pp. 843–854. issn: 00928674.
Leja, J, D Yu, B Nilsson, L Gedda, a Zieba, T Hakkarainen, G Akerstrom, K Oberg,
V Giandomenico, and M Essand (2011). “Oncolytic adenovirus modified with somato-
statin motifs for selective infection of neuroendocrine tumor cells”. In: Gene Therapy
18.11, pp. 1052–1062. issn: 0969-7128.
Leja, J., H. Dzojic, E. Gustafson, K. Oberg, V. Giandomenico, and M. Essand (2007).
“A novel chromogranin-A promoter-driven oncolytic adenovirus for midgut carcinoid
therapy”. In: Clinical Cancer Research 13.8, pp. 2455–2462. issn: 10780432.
Leja, J., A. Essaghir, M. Essand, K. Wester, K. Oberg, T. H. Totterman, R. Lloyd,
G. Vasmatzis, J.-B. Demoulin, and V. Giandomenico (2009). “Novel markers for ente-
388
rochromaffin cells and gastrointestinal neuroendocrine carcinomas.” In: Modern pathol-
ogy : an official journal of the United States and Canadian Academy of Pathology, Inc
22.2, pp. 261–272. issn: 0893-3952.
Leja, J., B. Nilsson, D. Yu, E. Gustafson, G. Akerstrom, K. Oberg, V. Giandomenico, and
M. Essand (2010). “Double-detargeted oncolytic adenovirus shows replication arrest in
liver cells and retains neuroendocrine cell killing ability”. In: PLoS ONE 5.1, e8916.
issn: 19326203.
Leoncini, E., G. Carioli, C. La Vecchia, S. Boccia, and G. Rindi (2016). “Risk factors
for neuroendocrine neoplasms: A systematic review and meta-analysis”. In: Annals of
Oncology 27.1, pp. 68–81. issn: 15698041.
Lepage, C, a. M. Bouvier, J. M. Phelip, C Hatem, C Vernet, and J Faivre (2004). “In-
cidence and management of malignant digestive endocrine tumours in a well defined
French population.” In: Gut 53.4, pp. 549–553. issn: 0017-5749.
Lepage, C, L Ciccolallo, R De Angelis, a. M. Bouvier, J Faivre, and G Gatta (2010).
“European disparities in malignant digestive endocrine tumours survival.” In: Inter-
national journal of cancer. Journal international du cancer 126.12, pp. 2928–34. issn:
1097-0215.
Lepage, C., A.-M. Bouvier, S. Manfredi, V. Dancourt, and J. Faivre (2006). “Incidence
and Management of Primary Malignant Small Bowel Cancers: A Well-defined French
Population Study”. In: The American Journal of Gastroenterology 101, pp. 2826–2832.
Lepage, C., B. Rachet, and M. P. Coleman (2007). “Survival from malignant digestive en-
docrine tumors in England and Wales: a population-based study.” In: Gastroenterology
132.3, pp. 899–904. issn: 0016-5085.
389
Lewis, B. P., C. B. Burge, and D. P. Bartel (2005). “Conserved seed pairing, often flanked
by adenosines, indicates that thousands of human genes are microRNA targets”. In:
Cell 120, pp. 15–20. issn: 00928674.
Li, A., J. Yu, H. Kim, C. L. Wolfgang, M. I. Canto, R. H. Hruban, and M. Goggins
(2013a). “MicroRNA array analysis finds elevated serum miR-1290 accurately distin-
guishes patients with low-stage pancreatic cancer from healthy and disease controls.”
In: Clinical cancer research : an official journal of the American Association for Cancer
Research 19.13, pp. 3600–10. issn: 1078-0432.
Li, H. J., B. Johnston, D. Aiello, D. R. Caffrey, M. Giel-Moloney, G. Rindi, and A. B.
Leiter (2014). “Enterochromaffin-Like Cells in the Gastric Corpus”. In: Gastroenterol-
ogy 146.3, pp. 754–764.
Li, H. J., S. K. Ray, N. K. Singh, B. Johnston, and A. B. Leiter (2011). “Basic helix loop
helix transcription factors and enteroendocrine cell differentiation”. In: Diabetes Obes
Metab. 13.1, pp. 5–12. issn: 15378276. arXiv: NIHMS150003.
Li, S.-C., A. Essaghir, C. Martijn, R. V. Lloyd, J.-B. Demoulin, K. Oberg, and V. Gian-
domenico (2013b). “Global microRNA profiling of well-differentiated small intestinal
neuroendocrine tumors.” In: Mod Pathol 26.5, pp. 685–96. issn: 1530-0285.
Li, S.-C., M. Khan, M. Caplin, T. Meyer, K. Oberg, and V. Giandomenico (2015). “So-
matostatin Analogs Treated Small Intestinal Neuroendocrine Tumor Patients Circu-
lating MicroRNAs.” In: PloS one 10.5, e0125553. issn: 1932-6203.
Li, Y. and K. V. Kowdley (2012). “MicroRNAs in common human diseases.” In: Ge-
nomics, proteomics & bioinformatics 10.5, pp. 246–53. issn: 2210-3244.
Lin, S. and R. I. Gregory (2015). “MicroRNA biogenesis pathways in cancer”. In: Nature
Review Cancer 15.6, pp. 321–333. issn: 1474-175X.
390
Linan-Rico, A., F. Ochoa-Cortes, A. Beyder, S. Soghomonyan, A. Zuleta-Alarcon, V.
Coppola, and F. L. Christofi (2016). “Mechanosensory Signaling in Enterochromaffin
Cells and 5-HT Release: Potential Implications for Gut Inflammation”. In: Frontiers
in Neuroscience 10.December, pp. 1–19. issn: 1662-453X.
Ling, H., M. Fabbri, and G. a. Calin (2013). “MicroRNAs and other non-coding RNAs
as targets for anticancer drug development.” In: Nature reviews. Drug discovery 12.11,
pp. 847–65. issn: 1474-1784.
Listing, H., W. A. Mardin, S. Wohlfromm, S. T. Mees, and J. Haier (2015). “MiR-23a/-
24-induced gene silencing results in mesothelial cell integration of pancreatic cancer”.
In: British Journal of Cancer 112.1, pp. 131–139. issn: 15321827.
Liu, L., R. R. Broaddus, J. C. Yao, S. Xie, J. a. White, T.-T. Wu, S. R. Hamilton, and A.
Rashid (2005). “Epigenetic alterations in neuroendocrine tumors: methylation of RAS-
association domain family 1, isoform A and p16 genes are associated with metastasis.”
In: Modern pathology : an official journal of the United States and Canadian Academy
of Pathology, Inc 18.12, pp. 1632–40. issn: 0893-3952.
Liu, Q., H. Fu, F. Sun, H. Zhang, Y. Tie, J. Zhu, R. Xing, Z. Sun, and X. Zheng (2008).
“miR-16 family induces cell cycle arrest by regulating multiple cell cycle genes”. In:
Nucleic Acids Research 36.16, pp. 5391–5404. issn: 03051048.
Logan, R. F. A., J. Patnick, C. Nickerson, L. Coleman, M. D. Rutter, and C. von Wagner
(2012). “Outcomes of the Bowel Cancer Screening Programme (BCSP) in England
after the first 1 million tests”. In: Gut 61.10, pp. 1439–1446. issn: 0017-5749.
Love, M. I., W. Huber, and S. Anders (2014). “Moderated estimation of fold change and
dispersion for RNA-seq data with DESeq2”. In: Genome Biology 15.12, p. 550. issn:
1474-760X.
391
Lu, J., G. Getz, E. A. Miska, E. Alvarez-Saavedra, J. Lamb, D. Peck, A. Sweet-Cordero,
B. L. Ebert, R. H. Mak, A. A. Ferrando, J. R. Downing, T. Jacks, H. R. Horvitz,
and T. R. Golub (2005). “MicroRNA expression profiles classify human cancers”. In:
Nature 435.7043, pp. 834–838. issn: 0028-0836. arXiv: 1512.00567.
Lubarsch, O. (1888). “Ueber den primaren Krebs des Ileum nebst Bemerkungen uber das
gleichzeitige Vorkommen von Krebs und Tuberculose.” In: Virchows Arch Pathol Anat
111, pp. 280–317.
Ludwig, N., P. Leidinger, K. Becker, C. Backes, T. Fehlmann, C. Pallasch, S. Rhein-
heimer, B. Meder, C. Stahler, E. Meese, and A. Keller (2016). “Distribution of miRNA
expression across human tissues”. In: Nucleic Acids Research 44.8, pp. 3865–3877. issn:
13624962.
Luzi, E., F. Marini, F. Giusti, G. Galli, L. Cavalli, and M. L. Brandi (2012). “The
negative feedback-loop between the oncomir Mir-24-1 and menin modulates the Men1
tumorigenesis by mimicking the ”Knudson’s second hit”.” In: PloS one 7.6, e39767.
issn: 1932-6203.
Magerl, C., J. Ellinger, T. Braunschweig, E. Kremmer, L. K. Koch, T. Holler, R. Buttner,
B. Luscher, and I. Gutgemann (2010). “H3K4 dimethylation in hepatocellular carci-
noma is rare compared with other hepatobiliary and gastrointestinal carcinomas and
correlates with expression of the methylase Ash2 and the demethylase LSD1.” In: Hu-
man pathology 41.2, pp. 181–9. issn: 1532-8392.
Mall, C., D. M. Rocke, B. Durbin-Johnson, and R. H. Weiss (2013). “Stability of miRNA
in human urine supports its biomarker potential.” In: Biomarkers in medicine 7.4,
pp. 623–31. issn: 1752-0371.
Marinoni, I., A. S. Kurrer, E. Vassella, M. Dettmer, T. Rudolph, V. Banz, F. Hunger,
S. Pasquinelli, E. J. Speel, and A. Perren (2014). “Loss of DAXX and ATRX are
392
associated with chromosome instability and reduced survival of patients with pancreatic
neuroendocrine tumors”. In: Gastroenterology 146.2, 453–460.e5. issn: 00165085.
Masson, M. H., C. Poisson, A. Guerardel, A. Mamin, J. Philippe, and Y. Gosmain (2014).
“Foxa1 and Foxa2 regulate alpha-cell differentiation, glucagon biosynthesis, and secre-
tion”. In: Endocrinology 155.10, pp. 3781–3792. issn: 19457170.
Mastracci, T., C. Lin, and L Sussel (2013). “Generation of mice encoding a conditional
allele of Nkx2.2”. In: Transgenic Research 22.5, pp. 965–72. issn: 09628819.
Matei, D., F. Fang, C. Shen, J. Schilder, A. Arnold, Y. Zeng, W. a. Berry, T. Huang,
and K. P. Nephew (2012). “Epigenetic resensitization to platinum in ovarian cancer.”
In: Cancer research 72.9, pp. 2197–205. issn: 1538-7445.
Maxwell, J. E. and J. R. Howe (2015). “Imaging in neuroendocrine tumors: an update
for the clinician”. In: International Journal of Endocrine Oncology 2.2, pp. 159–168.
issn: 2045-0869.
May, C. L. and K. H. Kaestner (2010). “Gut endocrine cell development”. In: Molecular
and Cellular Endocrinology 323.1, pp. 70–75. issn: 03037207.
McCarthy, D. J., Y. Chen, and G. K. Smyth (2012). “Differential expression analysis
of multifactor RNA-Seq experiments with respect to biological variation”. In: Nucleic
Acids Research 40.10, pp. 4288–4297. issn: 03051048.
McShane, L. M., D. G. Altman, W. Sauerbrei, S. E. Taube, M. Gion, and G. M. Clark
(2005). “REporting recommendations for tumor MARKer prognostic studies (RE-
MARK)”. In: J Natl Cancer Inst 97.16, pp. 1180–4. issn: 01676806.
Meijer, H. a., E. M. Smith, and M. Bushell (2014). “Regulation of miRNA strand selection:
follow the leader?” In: Biochemical Society transactions 42.4, pp. 1135–40. issn: 1470-
8752.
393
Merola, E., M. Rinzivillo, N. Cicchese, G. Capurso, F. Panzuto, and G. Delle Fave (2016).
“Digestive Neuroendocrine Neoplasms: a 2016 overview”. In: Digestive and Liver Dis-
ease 48.8, pp. 829–835. issn: 18783562.
Miki, M., T. Ito, M. Hijioka, L. Lee, K. Yasunaga, K. Ueda, T. Fujiyama, Y. Tachibana, K.
Kawabe, R. T. Jensen, and Y. Ogawa (2017). “Utility of chromogranin B compared with
chromogranin A as a biomarker in Japanese patients with pancreatic neuroendocrine
tumors”. In: Japanese Journal of Clinical Oncology, pp. 1–9. issn: 0368-2811.
Miller, C. P., R. E. McGehee, and J. F. Habener (1994). “IDX-1: a new homeodomain
transcription factor expressed in rat pancreatic islets and duodenum that transactivates
the somatostatin gene.” In: The EMBO journal 13.5, pp. 1145–1156. issn: 0261-4189.
Miller, H. C., P. Drymousis, R. Flora, R. Goldin, D. Spalding, and A. Frilling (2014).
“Role of Ki-67 proliferation index in the assessment of patients with neuroendocrine
neoplasias regarding the stage of disease.” In: World Journal of Surgery 38.6, pp. 1353–
1361. issn: 1432-2323.
Miller, H. C., M. Kidd, I. M. Modlin, P. Cohen, R. Dina, P. Drymousis, P. Vlavianos,
G. Kloppel, and A. Frilling (2015a). “Glucagon receptor gene mutations with hyper-
glucagonemia but without the glucagonoma syndrome”. In: World Journal of Gastroin-
testinal Surgery 7.4, pp. 60–66. issn: 1948-9366.
Miller, H. C., M. Kidd, L. Castellano, and A. Frilling (2015b). “Molecular genetic findings
in small bowel neuroendocrine neoplasms: a review of the literature”. In: International
Journal of Endocrine Oncology 2.2, pp. 143–150.
Miller, H. C., A. E. Frampton, A. Malczewska, S. Ottaviani, E. A. Stronach, R. Flora,
D. Kaemmerer, G. Schwach, R. Pfragner, O. Faiz, B. Kos-Kudta, G. B. Hanna, J.
Stebbing, L. Castellano, and A. Frilling (2016). “MicroRNAs associated with small
394
bowel neuroendocrine tumours and their metastases.” In: Endocrine-Related Cancer
23.9, pp. 711–726. issn: 14796821.
Mima, K., R. Nishihara, J. Yang, R. Dou, Y. Masugi, Y. Shi, A. D. Silva, Y. Cao,
M. Song, J. Nowak, M. Gu, W. Li, T. Morikawa, X. Zhang, K. Wu, H. Baba, E. L.
Giovannucci, J. A. Meyerhardt, A. T. Chan, C. S. Fuchs, Z. R. Qian, and S. Ogino
(2016). “MicroRNA MIR21 (miR-21) and PTGS2 expression in colorectal cancer and
patient survival”. In: Clinical Cancer Research 22.15, pp. 3841–3848. issn: 15573265.
Minnetti, M. and A. Grossman (2016). “Somatic and germline mutations in NETs: Impli-
cations for their diagnosis and management”. In: Best Practice and Research: Clinical
Endocrinology and Metabolism 30.1, pp. 115–127. issn: 15321908.
Missiaglia, E., I. Dalai, S. Barbi, S. Beghelli, M. Falconi, M. Della Peruta, L. Piemonti, G.
Capurso, A. Di Florio, G. Delle Fave, P. Pederzoli, C. M. Croce, and A. Scarpa (2010).
“Pancreatic endocrine tumors: Expression profiling evidences a role for AKT-mTOR
pathway”. In: Journal of Clinical Oncology 28.2, pp. 245–255. issn: 0732183X.
Mitchell, P. S., R. K. Parkin, E. M. Kroh, B. R. Fritz, S. K. Wyman, E. L. Pogosova-
Agadjanyan, A. Peterson, J. Noteboom, K. C. O’Briant, A. Allen, D. W. Lin, N. Urban,
C. W. Drescher, B. S. Knudsen, D. L. Stirewalt, R. Gentleman, R. L. Vessella, P. S.
Nelson, D. B. Martin, and M. Tewari (2008). “Circulating microRNAs as stable blood-
based markers for cancer detection”. In: Proceedings of the National Academy of Sci-
ences 105.30, pp. 10513–10518. issn: 0027-8424. arXiv: pnas.0804549105 [10.1073].
Modlin, I., K. Oberg, D. Chung, and R. Jensen (2008). “Gastroenteropancreatic neuroen-
docrine tumours”. In: The lancet oncology 9.January, pp. 61–72.
Modlin, I., S. Moss, K Oberg, R Padbury, R. Hicks, B. Gustafsson, N. Wright, and M
Kidd (2010). “Gastrointestinal neuroendocrine (carcinoid) tumours: current diagnosis
and management”. In: Medical Journal of Australia 193.1, pp. 46–52.
395
Modlin, I. M., K. D. Lye, and M. Kidd (2003). “A 5-decade analysis of 13,715 carcinoid
tumors.” In: Cancer 97.4, pp. 934–59. issn: 0008-543X.
Modlin, I. M., M. D. Shapiro, and M. Kidd (2004). “Siegfried oberndorfer: Origins and
perspectives of carcinoid tumors”. In: Human Pathology 35.12, pp. 1440–1451. issn:
00468177.
Modlin, I. M., I. Latich, M. Kidd, M. Zikusoka, and G. Eick (2006). “Therapeutic Op-
tions for Gastrointestinal Carcinoids”. In: Clinical Gastroenterology and Hepatology 4.5,
pp. 526–547. issn: 15423565.
Modlin, I. M., I. Drozdov, D. Alaimo, S. Callahan, N. Teixiera, L. Bodei, and M. Kidd
(2014). “A multianalyte PCR blood test outperforms single analyte ELISAs (chro-
mogranin A, pancreastatin, neurokinin A) for neuroendocrine tumor detection.” In:
Endocrine-related cancer 21.4, pp. 615–28. issn: 1479-6821.
Modlin, I. M., L. Bodei, and M. Kidd (2016). “Neuroendocrine tumor biomarkers: From
monoanalytes to transcripts and algorithms”. In: Best Practice and Research: Clinical
Endocrinology and Metabolism 30.1, pp. 59–77. issn: 15321908.
Murtaza, M., S.-J. Dawson, D. W. Y. Tsui, D. Gale, T. Forshew, A. M. Piskorz, C.
Parkinson, S.-F. Chin, Z. Kingsbury, A. S. C. Wong, F. Marass, S. Humphray, J.
Hadfield, D. Bentley, T. M. Chin, J. D. Brenton, C. Caldas, and N. Rosenfeld (2013).
“Non-invasive analysis of acquired resistance to cancer therapy by sequencing of plasma
DNA”. In: Nature 497.7447, pp. 108–12. issn: 0028-0836.
Navin, N., J. Kendall, J. Troge, P. Andrews, L. Rodgers, J. McIndoo, and M. Wigler
(2011). “Tumor Evolution Inferred by Single Cell Sequencing”. In: Nature 472.7341,
pp. 90–94. issn: 1476-4687.
396
Naya, F. J., H. P. Huang, Y. Qiu, H. Mutoh, F. J. DeMayo, A. B. Leiter, and M. J. Tsai
(1997). “Diabetes, defective pancreatic morphogenesis, and abnormal enteroendocrine
differentiation in BETA2/NeuroD-deficient mice”. In: Genes and Development 11.18,
pp. 2323–2334. issn: 08909369.
Nell, S., R. S. van Leeuwaarde, C. R. C. Pieterman, J. M. de Laat, A. R. Hermus, O. M.
Dekkers, W. W. de Herder, A. N. van der Horst-Schrivers, M. L. Drent, P. H. Bisschop,
B. Havekes, I. H. M. Borel Rinkes, M. R. Vriens, and G. D. Valk (2015). “No Association
of Blood Type O With Neuroendocrine Tumors in Multiple Endocrine Neoplasia Type
1.” In: The Journal of clinical endocrinology and metabolism 100.10, pp. 3850–5. issn:
1945-7197.
Niederle, B, U.-F. Pape, F Costa, D Gross, F Kelestimur, U Knigge, K Oberg, M Pavel, A
Perren, C Toumpanakis, J O’Connor, D O’Toole, E Krenning, N Reed, R Kianmanesh,
and all other Vienna Consensus Conference participants (2016). “ENETS Consensus
Guidelines Update for Neuroendocrine Neoplasms of the Jejunum and Ileum.” In: Neu-
roendocrinology 103.2, pp. 125–38. issn: 1423-0194.
Nieser, M., T. Henopp, J. Brix, L. Stoß, B. Sitek, W. Naboulsi, M. Anlauf, A. M. Schlit-
ter, G. Kloppel, T. Gress, R. Moll, D. K. Bartsch, A. E. Heverhagen, W. T. Knoefel, D.
Kaemmerer, J. Haybaeck, F. Fend, J. Sperveslage, and B. Sipos (2016). “Loss of Chro-
mosome 18 in Neuroendocrine Tumors of the Small Intestine: The Enigma Remains”.
In: Neuroendocrinology 104.3, pp. 302–312. issn: 14230194.
Nishi, T., Y. Kawabata, Y. Hari, H. Imaoka, N. Ishikawa, S. Yano, R. Maruyama, and
Y. Tajima (2012). “A case of pancreatic neuroendocrine tumor in a patient with
neurofibromatosis-1.” In: World journal of surgical oncology 10, p. 153. issn: 1477-
7819.
397
Norheim, I., K. Oberg, E. Theodorsson-Norheim, P. G. Lindgren, G. Lundqvist, A. Mag-
nusson, L. Wide, and E. Wilander (1987). “Malignant carcinoid tumors. An analysis
of 103 patients with regard to tumor localization, hormone production, and survival.”
In: Annals of Surgery 206.2, pp. 115–125.
Norlen, O., P. Stalberg, K. Oberg, J. Eriksson, J. Hedberg, O. Hessman, E. T. Janson, P.
Hellman, and G. Akerstrom (2012). “Long-term results of surgery for small intestinal
neuroendocrine tumors at a tertiary referral center”. In: World Journal of Surgery 36.6,
pp. 1419–1431. issn: 03642313.
Oberg, K. and E. Barbro (2005). “Endocrine tumours of the pancreas”. In: Best Practice
& Research Clinical Gastroenterology 19.5, pp. 753–781. issn: 03090167.
Oberg, K. and S. W. Lamberts (2016). “Somatostatin analogues in acromegaly and gas-
troenteropancreatic neuroendocrine tumours: past, present and future”. In: Endocrine-
related cancer 5920.October, pp. 1–36. issn: 1351-0088.
Oberg, K., I. M. Modlin, W. De Herder, M. Pavel, D. Klimstra, A. Frilling, D. C. Metz,
A. Heaney, D. Kwekkeboom, J. Strosberg, T. Meyer, S. F. Moss, K. Washington, E.
Wolin, E. Liu, and J. Goldenring (2015). “Consensus on biomarkers for neuroendocrine
tumour disease”. In: The Lancet Oncology 16.9, e435–e446. issn: 14745488.
Oberndorfer, S (1928). “Karzinoide handbuch der speziellen”. In: Henke F., Lubarsch
O. (eds): Handbuch der Speziellen Pathologischen Anatomie und Histologie. Berlin,
Germany: Verlag von Julius Springer, pp. 814–847.
Oberndorfer, S. (1907). “Karzinoide Tumoren des Dunndarms.” In: Frankf Z Pathol 1,
pp. 426–32.
O’Connor, J. M., F. Marmissolle, C. Bestani, V. Pesce, S. Belli, E. Dominichini, G.
Mendez, P. Price, N. Giacomi, A. Pairola, F. S. Loria, E. Huertas, C. Martin, K.
398
Patane, C. Poleri, M. Rosenberg, A. Cabanne, M. Kujaruk, A. Caino, V. Zamora, J.
Mariani, M. Dioca, P. Parma, G. Podesta, O. Andriani, G. Gondolesi, and E. Roca
(2014). “Observational study of patients with gastroenteropancreatic and bronchial
neuroendocrine tumors in Argentina: Results from the large database of a multidis-
ciplinary group clinical multicenter study.” In: Molecular and clinical oncology 2.5,
pp. 673–684. issn: 2049-9450.
O’Connor, T. M., J. O’Connell, D. I. O’Brien, T. Goode, C. P. Bredin, and F. Shanahan
(2004). “The role of substance P in inflammatory disease”. In: Journal of Cellular
Physiology 201.2, pp. 167–180. issn: 00219541.
Offield, M. F., T. L. Jetton, P. A. Labosky, M. Ray, R. W. Stein, M. A. Magnuson,
B. L. Hogan, and C. V. Wright (1996). “PDX-1 is required for pancreatic outgrowth
and differentiation of the rostral duodenum.” In: Development (Cambridge, England)
122.3, pp. 983–95. issn: 0950-1991.
Ohisson, H., K. Karisson, and T. Edlund (1993). “IPF1, a homeodomain-containing trans-
activator of the insulin gene.” In: EMBO Journal 12.11, pp. 4251–4259.
Olson, P., J. Lu, H. Zhang, A. Shai, M. G. Chun, Y. Wang, S. K. Libutti, E. K. Nakakura,
T. R. Golub, and D. Hanahan (2009). “MicroRNA dynamics in the stages of tumorige-
nesis correlate with hallmark capabilities of cancer.” In: Genes & development 23.18,
pp. 2152–65. issn: 1549-5477.
O’Mahony, S. M., G. Clarke, Y. E. Borre, T. G. Dinan, and J. F. Cryan (2015). “Serotonin,
tryptophan metabolism and the brain-gut-microbiome axis”. In: Behavioural Brain
Research 277, pp. 32–48. issn: 18727549.
O’Toole, D., R. Kianmanesh, and M. Caplin (2016). “ENETS 2016 Consensus Guidelines
for the Management of Patients with Digestive Neuroendocrine Tumours: An Update”.
In: Neuroendocrinology 103.2, pp. 117–118. issn: 14230194.
399
Ozturk, C., J. Fleer, H. J. Hoekstra, and J. E.H. M. Hoekstra-Weebers (2015). “Delay in
diagnosis of testicular cancer; A need for awareness programs”. In: PLoS ONE 10.11,
pp. 1–10. issn: 19326203.
Pan, C., C. Kumar, S. Bohl, U. Klingmueller, and M. Mann (2009). “Comparative Pro-
teomic Phenotyping of Cell Lines and Primary Cells to Assess Preservation of Cell
Type-specific Functions”. In: Molecular & Cellular Proteomics 8.3, pp. 443–450. issn:
1535-9476.
Pape, U.-F., U. Berndt, J. Muller-Nordhorn, M. Bohmig, S. Roll, M. Koch, S. N Willich,
and B. Wiedenmann (2008). “Prognostic factors of long-term outcome in gastroen-
teropancreatic neuroendocrine tumours”. In: Endocrine Related Cancer 15.4, pp. 1083–
1097. issn: 1351-0088.
Pape, U. F., B. Niederle, F. Costa, D. Gross, F. Kelestimur, R. Kianmanesh, U. Knigge, K.
Oberg, M. Pavel, A. Perren, C. Toumpanakis, J. O’Connor, E. Krenning, N. Reed, and
D. O’Toole (2016). “Consensus Guidelines for Neuroendocrine Neoplasms of the Ap-
pendix (Excluding Goblet Cell Carcinomas)”. In: Neuroendocrinology 103.2, pp. 144–
152. issn: 14230194.
Pardi, E., S. Mariotti, N. S. Pellegata, K. Benfini, S. Borsari, F. Saponaro, L. Torregrossa,
A. Cappai, C. Satta, M. Mastinu, C. Marcocci, and F. Cetani (2014). “Functional
characterization of a CDKN1B mutation in a Sardinian kindred with multiple endocrine
neoplasia type 4 (MEN4)”. In: Endocrine connections, pp. 1–8. issn: 2049-3614.
Park, T. Y., S. K. Lee, J. S. Park, D Oh, T. J. Song, H Park do, S. S. Lee, D. W. Seo, and
M. H. Kim (2015). “Clinical features of pancreatic involvement in von Hippel-Lindau
disease: a retrospective study of 55 cases in a single center”. In: Scand J Gastroenterol
50.3, pp. 360–367. issn: 0036-5521.
400
Patel, D., D. Chan, G. Cehic, N. Pavlakis, and T. J. Price (2016). “Systemic thera-
pies for advanced gastroenteropancreatic neuroendocrine tumors”. In: Expert Review
of Endocrinology & Metabolism 6651.June, pp. 1–17. issn: 1744-6651.
Pavel, M., D. O’Toole, F. Costa, J. Capdevila, D. Gross, R. Kianmanesh, E. Krenning, U.
Knigge, R. Salazar, U. F. Pape, and K. Oberg (2016). “Consensus Guidelines Update
for the Management of Distant Metastatic Disease of Intestinal, Pancreatic, Bronchial
Neuroendocrine Neoplasms (NEN) and NEN of Unknown Primary Site”. In: Neuroen-
docrinology 103.2, pp. 172–185. issn: 14230194.
Pavel, M., E. Baudin, A. Couvelard, E. Krenning, K. Oberg, T. Steinmuller, M. An-
lauf, B. Wiedenmann, and R. Salazar (2012). “ENETS consensus guidelines for the
management of patients with liver and other distant metastases from neuroendocrine
neoplasms of foregut, midgut, hindgut, and unknown primary”. In: Neuroendocrinology
95.2, pp. 157–176. issn: 00283835.
Pawa, N, K Clift, A, H Osmani, P Drymousis, A Cichocki, R Flora, N Goldiawa, K
Clift, A, R Flora, R Goldin, D Patsouras, A Baird, A Malczewska, J Kinross, O Faiz,
A Antoniou, H Wasan, A Kaltsas, G, A Darzi, B Cwikla, J, and A Frilling (2018).
“Surgical Management of Patients with Neuroendocrine Neoplasms of the Appendix:
Appendectomy or More?” In: Neuroendocrinology 106, pp. 242–251.
Pearse, A. (1966). “Common cytochemical properties of cells producing polypeptide hor-
mones, with particular reference to calcitonin and thyroid C-cells.” In: Vet. Rec. 79,
pp. 587–590.
Pellegata, N., L. Quintanilla-Martinez, H. Siggelkow, E. Samson, K. Bink, H. Hofler,
F. Fend, Jochen Graw, and M. J. Atkinson (2006). “Germ-line mutations in p27Kip1
cause a multiple endocrine neoplasia syndrome in rats and humans”. In: Proceedings
of the . . . 103.42, pp. 15558–63.
401
Pelosi, G, F Bresaola, and G Bagina (1996). “Endocrine Tumours of the Pancreas: Ki67
immunoreactivity on paraffin section is an indipendetn predictor for malignancy: a
comparative study with proliferating-cell nuclear antigen and progesterone receptor
protein immunostaging, mitotic index and other c”. In: Human PAthology, pp. 1124–
1134.
Pfeffer, S. R., C. H. Yang, and L. M. Pfeffer (2015). “The Role of MIR-21 in Cancer”.
In: Drug Development Research 76.6, pp. 270–277. issn: 10982299.
Pfragner, R (1996). “Establishment of a continuous cell line from a human carcinoid of
the small intestine (KRJ-I)”. In: Int J Oncol. 8.3, pp. 513–520.
Pfragner, R., A. Behmel, H. Hoger, A. Beham, E. Ingolic, I. Stelzer, B. Svejda, V. A.
Moser, A. C. Obenauf, V. Siegl, O. Haas, and B. Niederle (2009). “Establishment
and characterization of three novel cell lines - P-STS, L-STS, H-STS - derived from a
human metastatic midgut carcinoid.” In: Anticancer research 29.6, pp. 1951–61. issn:
0250-7005.
Phan, A. T., D. M. Halperin, J. A. Chan, D. R. Fogelman, K. R. Hess, P. Malinowski, E.
Regan, C. S. Ng, J. C. Yao, and M. H. Kulke (2015). “Pazopanib and depot octreotide
in advanced, well-differentiated neuroendocrine tumours: a multicentre, single-group,
phase 2 study”. In: Lancet Oncol 16.6, pp. 695–703. issn: 0036-8075. arXiv: 15334406.
Pinney, S. E., J. Oliver-Krasinski, L. Ernst, N. Hughes, P. Patel, D. A. Stoffers, P. Russo,
and D. D. De Leon (2011). “Neonatal diabetes and congenital malabsorptive diarrhea
attributable to a novel mutation in the human neurogenin-3 gene coding sequence”. In:
Journal of Clinical Endocrinology and Metabolism 96.7, pp. 1960–1965. issn: 0021972X.
Ploeckinger, U., G. Kloeppel, B. Wiedenmann, and R. Lohmann (2009). “The German
NET-registry: An audit on the diagnosis and therapy of neuroendocrine tumors”. In:
Neuroendocrinology 90.4, pp. 349–363. issn: 00283835.
402
Pruitt, S. C., A. Freeland, M. E. Rusiniak, D. Kunnev, and G. K. Cady (2013). “Cdkn1b
overexpression in adult mice alters the balance between genome and tissue ageing”. In:
Nature Communications 4, pp. 1–12. issn: 20411723.
Pusceddu, S., D. Femia, G. Lo Russo, S. Ortolani, M. Milione, M. Maccauro, C. Vernieri,
N. Prinzi, L. Concas, L. Leuzzi, F. De Braud, and R. Buzzoni (2016). “Update on
medical treatment of small intestinal neuroendocrine tumors”. In: Expert Review of
Anticancer Therapy 16.9, pp. 969–976. issn: 1473-7140.
Pyo, J. H., S. N. Hong, B.-H. Min, J. H. Lee, D. K. Chang, P.-L. Rhee, J. J. Kim, S. K.
Choi, S.-H. Jung, H. J. Son, and Y.-H. Kim (2016). “Evaluation of the risk factors
associated with rectal neuroendocrine tumors: a big data analytic study from a health
screening center.” In: Journal of gastroenterology. issn: 1435-5922.
Quraish, R., N. Sudou, K. Nomura-komoike, F. Sato, and H. Fujieda (2016). “p27 KIP1
loss promotes proliferation and phagocytosis but prevents epithelial – mesenchymal
transition in RPE cells after photoreceptor damage”. In: Molecular Vision 22.October
2015, pp. 1103–1121.
Ramage, J., W. W. De Herder, G. F. Delle Fave, P. Ferolla, D. Ferone, T. Ito, P.
Ruszniewski, A. Sundin, W. Weber, Z. Zheng-Pei, B. Taal, and A. Pascher (2016).
“ENETS Consensus Guidelines Update for Colorectal Neuroendocrine Neoplasms”. In:
Neuroendocrinology 103.2, pp. 139–143. issn: 14230194.
Ramage, J., A. Davies, J Ardill, N Bax, M Caplin, A Grossman, R Hawkins, A. McNicol,
N Reed, R Sutton, R Thakker, S Aylwin, D Breen, K Britton, K Buchanan, P Corrie, A
Gillams, V Lewington, D McCance, K Meeran, and A Watkinson (2005). “Guidelines
for the management of gastroenteropancreatic neuroendocrine (including carcinoid)
tumours.” In: Gut 54.4, pp. iv1–16. issn: 1468-3288.
403
Randle, R. W., S. a. Northrup, S. J. Sirintrapun, D. S. Lyles, and J. H. Stewart (2013).
“Oncolytic vesicular stomatitis virus as a treatment for neuroendocrine tumors.” In:
Surgery 154.6, pp. 1323–30. issn: 1532-7361.
Ransom, W. B. (1890). “A case of primary carcinoma of the ileum.” In: Lancet 2,
pp. 1020–1023.
Rask-Andersen, M., M. S. Almen, and H. B. Schioth (2011). “Trends in the exploitation
of novel drug targets”. In: Nature Reviews: Drug discovery 10.8, pp. 579–590. issn:
1474-1784.
Raymond, E, L Dahan, J Raoul, Y Bang, I Borbath, C Lombard-Bohas, J Valle, P
Metrakos, D Smith, A Vinik, J Chen, D Horsch, P Hammel, B Wiedenmann, E Van
Cutsem, S Patyna, D. R. Lu, C Blanckmeister, R Chao, and P Ruszniewski (2011).
“Sunitinib Malate for the Treatment of Pancreatic Neuroendocrine Tumors”. In: New
England Journal 364.6, pp. 501–513.
Reboucas, E. d. L., J. J.d. N. Costa, M. J. Passos, J. R.d. S. Passos, R. van den Hurk,
and J. R. V. Silva (2013). “Real time PCR and importance of housekeepings genes for
normalization and quantification of mRNA expression in different tissues”. In: Brazilian
Archives of Biology and Technology 56.1, pp. 143–154. issn: 15168913.
Reinhardt, H. C. and B. Schumacher (2012). “The p53 network: Cellular and systemic
DNA damage responses in aging and cancer”. In: Trends in Genetics 28.3, pp. 128–136.
issn: 01689525.
Rindi, G, G Kloppel, H Alhman, M Caplin, A Couvelard, W. W. de Herder, B Erikssson,
A Falchetti, M Falconi, P Komminoth, M Korner, J. M. Lopes, A.-M. McNicol, O
Nilsson, A Perren, A Scarpa, J.-Y. Scoazec, and B Wiedenmann (2006). “TNM staging
of foregut (neuro)endocrine tumors: a consensus proposal including a grading system.”
404
In: Virchows Archiv : an international journal of pathology 449.4, pp. 395–401. issn:
0945-6317.
Rindi, G., G. Kloppel, A. Couvelard, P. Komminoth, M. Korner, J. M. Lopes, A.-M.
McNicol, O. Nilsson, A. Perren, A. Scarpa, J.-Y. Scoazec, and B. Wiedenmann (2007).
“TNM staging of midgut and hindgut (neuro) endocrine tumors: a consensus proposal
including a grading system”. In: Virchows Archiv 451.4, pp. 757–762. issn: 0945-6317.
Rindi, G, M Falconi, C Klersy, L Albarello, L Boninsegna, M. W. Buchler, C Capella, M
Caplin, A Couvelard, C Doglioni, G Delle Fave, L Fischer, G Fusai, W. W. de Herder,
H Jann, P Komminoth, R. R. de Krijger, S La Rosa, T. V. Luong, U Pape, A Perren,
P Ruszniewski, A Scarpa, A Schmitt, E Solcia, and B Wiedenmann (2012). “TNM
Staging of Neoplasms of the Endocrine Pancreas: Results From a Large International
Cohort Study.” In: Journal of the National Cancer Institute 104.10, pp. 764–777. issn:
1460-2105.
Rindi, G., C Bordi, S La Rosa, E Solcia, and G. Delle Fave (2011). “Gastroenteropancre-
atic (neuro)endocrine neoplasms: the histology report.” In: Digestive and liver disease
43 Suppl 4, S356–60. issn: 1878-3562.
Rinke, A., H. H. Muller, C. Schade-Brittinger, K. J. Klose, P. Barth, M. Wied, C. Mayer,
B. Aminossadati, U. F. Pape, M. Blaker, J. Harder, C. Arnold, T. Gress, and R.
Arnold (2009). “Placebo-controlled, double-blind, prospective, randomized study on
the effect of octreotide LAR in the control of tumor growth in patients with metastatic
neuroendocrine midgut tumors: A report from the PROMID study group”. In: Journal
of Clinical Oncology 27.28, pp. 4656–4663. issn: 0732183X.
Rinzivillo, M., G. Capurso, D. Campana, N. Fazio, F. Panzuto, F. Spada, N. Cicchese,
S. Partelli, P. Tomassetti, M. Falconi, and G. Delle Fave (2016). “Risk and Protec-
405
tive Factors for Small Intestine Neuroendocrine Tumours: A Prospectivecase-Control
Study.” In: Neuroendocrinology 103.5, pp. 531–537. issn: 1423-0194.
Rmostell (2011a). High Sensitivity Low Specificity.
– (2011b). Low Sensitivity High Specificity.
Robinson, M. D., D. J. McCarthy, and G. K. Smyth (2009). “edgeR: A Bioconductor
package for differential expression analysis of digital gene expression data”. In: Bioin-
formatics 26.1, pp. 139–140. issn: 13674803.
Roldo, C., E. Missiaglia, J. P. Hagan, M. Falconi, P. Capelli, S. Bersani, G. A. Calin,
S. Volinia, C.-G. Liu, A. Scarpa, and C. M. Croce (2006). “MicroRNA expression
abnormalities in pancreatic endocrine and acinar tumors are associated with distinctive
pathologic features and clinical behavior.” In: Journal of clinical oncology : official
journal of the American Society of Clinical Oncology 24.29, pp. 4677–84. issn: 1527-
7755.
Rosenfeld, N., R. Aharonov, E. Meiri, S. Rosenwald, Y. Spector, M. Zepeniuk, H. Ben-
jamin, N. Shabes, S. Tabak, A. Levy, D. Lebanony, Y. Goren, E. Silberschein, N.
Targan, A. Ben-Ari, S. Gilad, N. Sion-Vardy, A. Tobar, M. Feinmesser, O. Kharenko,
O. Nativ, D. Nass, M. Perelman, A. Yosepovich, B. Shalmon, S. Polak-Charcon, E.
Fridman, A. Avniel, I. Bentwich, Z. Bentwich, D. Cohen, A. Chajut, and I. Barshack
(2008). “MicroRNAs accurately identify cancer tissue origin”. In: Nature Biotechnology
26.4, pp. 462–469. issn: 1087-0156.
Ruebel, K., A. A. A. Leontovich, G. A. Stilling, S. Zhang, A. Righi, L. Jin, and R. V. Lloyd
(2010). “MicroRNA expression in ileal carcinoid tumors: downregulation of microRNA-
133a with tumor progression”. In: Modern Pathology 23.3, pp. 367–375. issn: 1530-0285.
406
Sadanandam, A., S. Wullschleger, C. A. Lyssiotis, C. Grotzinger, S. Barbi, S. Bersani, J.
Korner, I. Wafy, A. Mafficini, R. T. Lawlo, M. Simbolo, J. M. Asara, H. Blaker, L. C.
Cantley, B. Wiedenmann, A. Scarpa, and D. Hanahan (2015). “A cross-species analysis
in pancreatic neuroendocrine tumors reveals molecular subtypes with distinctive clin-
ical, metastatic, developmental, and metabolic characteristics”. In: Cancer Discovery
5.12, pp. 1296–1313. issn: 21598290. arXiv: 15334406.
Sakurai, A., S. Suzuki, S. Kosugi, T. Okamoto, S. Uchino, A. Miya, T. Imai, H. Kaji,
I. Komoto, D. Miura, M. Yamada, T. Uruno, K. Horiuchi, A. Miyauchi, M. Imamura,
T. Fukushima, K. Hanazaki, S. Hirakawa, T. Igarashi, T. Iwatani, M. Kammori, T.
Katabami, M. Katai, T. Kikumori, K. Kiribayashi, S. Koizumi, S. Midorikawa, R.
Miyabe, T. Munekage, A. Ozawa, K. Shimizu, I. Sugitani, H. Takeyama, and M. Ya-
mazaki (2012). “Multiple endocrine neoplasia type 1 in Japan: Establishment and anal-
ysis of a multicentre database”. In: Clinical Endocrinology 76.4, pp. 533–539. issn:
03000664.
Sandvik, O. M., K. Søreide, E. Gudlaugsson, J. T. Kvaløy, and J. A. Søreide (2016).
“Epidemiology and classification of gastroenteropancreatic neuroendocrine neoplasms
using current coding criteria”. In: British Journal of Surgery 103.3, pp. 226–232. issn:
13652168.
Sato, T., R. G. Vries, H. J. Snippert, M. van de Wetering, N. Barker, D. E. Stange,
J. H. van Es, A. Abo, P. Kujala, P. J. Peters, and H. Clevers (2009). “Single Lgr5 stem
cells build crypt-villus structures in vitro without a mesenchymal niche.” In: Nature
459.7244, pp. 262–5. issn: 1476-4687.
Sato-Kuwabara, Y., S. A. Melo, F. A. Soares, and G. A. Calin (2015). “The fusion of
two worlds: Non-coding RNAs and extracellular vesicles - Diagnostic and therapeutic
407
implications (Review)”. In: International Journal of Oncology 46.1, pp. 17–27. issn:
17912423.
Scacheri, P. C., S. Davis, D. T. Odom, G. E. Crawford, S. Perkins, M. J. Halawi, S. K.
Agarwal, S. J. Marx, A. M. Spiegel, P. S. Meltzer, and F. S. Collins (2006). “Genome-
wide analysis of menin binding provides insights into MEN1 tumorigenesis.” In: PLoS
genetics 2.4, e51. issn: 1553-7404.
Schaffer, A. E., B. L. Taylor, J. R. Benthuysen, J. Liu, F. Thorel, W. Yuan, Y. Jiao, K. H.
Kaestner, P. L. Herrera, M. A. Magnuson, C. L. May, and M. Sander (2013). “Nkx6.1
Controls a Gene Regulatory Network Required for Establishing and Maintaining Pan-
creatic Beta Cell Identity”. In: PLoS Genetics 9.1. issn: 15537390.
Schernthaner-Reiter, M. H., G. Trivellin, and C. A. Stratakis (2016). “MEN1, MEN4, and
Carney Complex: Pathology and Molecular Genetics”. In: Neuroendocrinology 103.1,
pp. 18–31. issn: 14230194.
Scherubl, H., R. T. Jensen, G. Cadiot, U. Stolzel, and G. Kloppel (2010). “Neuroendocrine
tumors of the small bowels are on the rise: Early aspects and management.” In: World
journal of gastrointestinal endoscopy 2.10, pp. 325–34. issn: 1948-5190.
Scherubl, H., B. Streller, R. Stabenow, H. Herbst, M. Hopfner, C. Schwertner, J. Stein-
berg, J. Eick, W. Ring, K. Tiwari, and S. M. Zappe (2013). “Clinically detected gas-
troenteropancreatic neuroendocrine tumors are on the rise: Epidemiological changes in
Germany”. In: World Journal of Gastroenterology 19.47, pp. 9012–9019. issn: 10079327.
Schimmack, S., B. Svejda, B. Lawrence, M. Kidd, and I. M. Modlin (2011). “The diversity
and commonalities of gastroenteropancreatic neuroendocrine tumors.” In: Langenbeck’s
archives of surgery / Deutsche Gesellschaft fur Chirurgie 396.3, pp. 273–98. issn: 1435-
2451.
408
Schokrpur, S., J. Hu, D. L. Moughon, P. Liu, L. C. Lin, K. Hermann, S. Mangul, W.
Guan, M. Pellegrini, H. Xu, and L. Wu (2016). “CRISPR-Mediated VHL Knockout
Generates an Improved Model for Metastatic Renal Cell Carcinoma.” In: Scientific
reports 6.February, p. 29032. issn: 2045-2322.
Scholzen, T. and J. Gerdes (2000). “The Ki-67 Protein : From the Known and the Un-
known”. In: Journal of Cellular Physiology 322.August 1999, pp. 311–322.
Schonhoff, S. E., M. Giel-Moloney, and A. B. Leiter (2004). “Minireview: Development
and differentiation of gut endocrine cells”. In: Endocrinology 145.6, pp. 2639–2644.
issn: 00137227.
Schreiter, N. F., W. Brenner, M. Nogami, R. Buchert, A. Huppertz, U. F. Pape, V.
Prasad, B. Hamm, and M. H. Maurer (2012). “Cost comparison of 111In-DTPA-
octreotide scintigraphy and 68Ga-DOTATOC PET/CT for staging enteropancreatic
neuroendocrine tumours”. In: European Journal of Nuclear Medicine and Molecular
Imaging 39.1, pp. 72–82. issn: 16197070.
Schunemann, V., K. Huntoon, and R. R. Lonser (2016). “Personalized Medicine for Ner-
vous System Manifestations of von Hippel-Lindau Disease.” In: Frontiers in surgery
3.June, p. 39. issn: 2296-875X.
Schwanhausser, B (2011). “Global quantification of mammalian gene expression control”.
In: Nature 473, pp. 337–342. issn: 0028-0836.
Sean, D. and P. S. Meltzer (2007). “GEOquery: A bridge between the Gene Expression
Omnibus (GEO) and BioConductor”. In: Bioinformatics 23.14, pp. 1846–1847. issn:
13674803.
409
Seok, H, J Ham, E. S. Jang, and S. W. Chi (2016). “MicroRNA Target Recognition:
Insights from Transcriptome-Wide Non-Canonical Interactions”. In: Mol Cells 39.5,
pp. 375–381. issn: 0219-1032.
Sharma, A., J. S. Khan, and P. J. Devereaux (2015). “Erratum:Is crowdfunding a viable
source of clinical trial research funding? (Lancet 2015; 386: 338)”. In: The Lancet
386.9991, p. 338. issn: 1474547X.
Shi, C., R. S. Gonzalez, Z. Zhao, T. Koyama, T. C. Cornish, K. R. Hande, R. Walker,
M. Sandler, J. Berlin, E. H. Liu, and E. H. Shi, C., Gonzalez, R. S., Zhao, Z., Koyama,
T., Cornish, T. C., Hande, K. R., . . . Liu (2015). “Liver metastases of small intes-
tine neuroendocrine tumors: Ki-67 heterogeneity and world health organization grade
discordance with primary tumors”. In: Am J Clin Pathol. 143.3, pp. 398–404. issn:
19437722.
Shroyer, N. F., D. Wallis, K. J. T. Venken, H. J. Bellen, and H. Y. Zoghbi (2005). “Gfi1
functions downstream of Math1 to control intestinal secretory cell subtype allocation
and differentiation”. In: Genes and Development 19.20, pp. 2412–2417. issn: 08909369.
Singh, S, J Hallet, C Rowsell, and C. H. L. Law (2014). “Variability of Ki67 labeling
index in multiple neuroendocrine tumors specimens over the course of the disease.” In:
European journal of surgical oncology : the journal of the European Society of Surgical
Oncology and the British Association of Surgical Oncology, pp. 1–6. issn: 1532-2157.
Sita-Lumsden, a, D. a. Dart, J Waxman, and C. L. Bevan (2013a). “Circulating microR-
NAs as potential new biomarkers for prostate cancer.” In: British journal of cancer
108.10, pp. 1925–30. issn: 1532-1827.
Sita-Lumsden, A., C. E. Fletcher, D. A. Dart, G. N. Brooke, J. Waxman, and C. L. Bevan
(2013b). “Circulating nucleic acids as biomarkers of prostate cancer”. In: Biomarkers
in Medicine 7.6, pp. 867–877. issn: 1752-0363.
410
Slater, E. P., K. Strauch, S. Rospleszcz, A. Ramaswamy, I. Esposito, G. Kloppel, E.
Matthai, K. Heeger, V. Fendrich, P. Langer, and D. K. Bartsch (2014). “MicroRNA-
196a and -196b as Potential Biomarkers for the Early Detection of Familial Pancreatic
Cancer.” In: Translational oncology xx.xx, pp. 1–8. issn: 1936-5233.
Song, J., Z. Bai, W. Han, J. Zhang, H. Meng, J. Bi, X. Ma, S. Han, and Z. Zhang (2012).
“Identification of suitable reference genes for qPCR analysis of serum microRNA in
gastric cancer patients”. In: Digestive Diseases and Sciences 57.4, pp. 897–904. issn:
01632116.
Sorbye, H., S. Welin, S. W. Langer, L. W. Vestermark, N. Holt, P. Osterlund, S. Dueland,
E. Hofsli, M. G. Guren, K. Ohrling, E. Birkemeyer, E. Thiis-Evensen, M. Biagini, H.
Gronbaek, L. M. Soveri, I. H. Olsen, B. Federspiel, J. Assmus, E. T. Janson, and U.
Knigge (2013). “Predictive and prognostic factors for treatment and survival in 305
patients with advanced gastrointestinal neuroendocrine carcinoma (WHO G3): The
NORDIC NEC study”. In: Annals of Oncology 24.1, pp. 152–160. issn: 09237534.
Srivastava, R., M. Kumar, S. Peineau, Z. Csaba, S. Mani, P. Gressens, and V. El Ghouzzi
(2013). “Conditional induction of Math1 specifies embryonic stem cells to cerebellar
granule neuron lineage and promotes differentiation into mature granule neurons”. In:
Stem Cells 31.4, pp. 652–665. issn: 10665099.
Steinhart, Z., Z. Pavlovic, M. Chandrashekhar, T. Hart, X. Wang, X. Zhang, M. Ro-
bitaille, K. R. Brown, S. Jaksani, R. Overmeer, S. F. Boj, J. Adams, J. Pan, H.
Clevers, S. Sidhu, J. Moffat, and S. Angers (2017). “Genome-wide CRISPR screens
reveal a Wnt-FZD5 signaling circuit as a druggable vulnerability of RNF43-mutant
pancreatic tumors”. In: Nature Medicine 23.1, pp. 60–68. issn: 1546170X.
411
Stewart, A. K., D. P. Winchester, and C. Y. Ko (2008). “Prognostic Score Predicting Sur-
vival After Resection of Pancreatic Neuroendocrine Tumors Analysis of 3851 Patients”.
In: Annals of Surgery 247.3, pp. 490–500. issn: 0003-4932.
Stridsberg, M., B. Eriksson, K. Oberg, and E. T. Janson (2003). “A comparison between
three commercial kits for chromogranin A measurements”. In: Journal of Endocrinology
177.2, pp. 337–341. issn: 00220795.
Strimbu, K. and J. a. Tavel (2011). “What are Biomarkers?” In: Curr Opin HIV AIDS
5.6, pp. 463–466. issn: 1746-6318.
Strosberg, J., E. Wolin, B Chasen, M. Kulke, D. Bushnell, M. Caplin, R. Baum, P. Kunz,
T. Hobday, and A. Hendifar (2016). “NETTER-1 phase III: Progression-free survival,
radiographic response, and preliminary overall survival results in patients with midgut
neuroendocrine tumors treated with 177-Lu-Dotatate.” In: Journal of Clinical Oncology
34.Supplement 4S, abstract 194.
Subramanian, A., P. Tamayo, and V. Mootha (2005). “GSEA : Gene set enrichment
analysis Gene set enrichment analysis : A knowledge-based approach for interpreting
genome-wide expression profiles”. In: PNAS 102.43, pp. 15545–15550.
Takeuchi, K., T. Endoh, S. Hayashi, and T. Aihara (2016). “Activation of Muscarinic
Acetylcholine Receptor Subtype 4 Is Essential for Cholinergic Stimulation of Gastric
Acid Secretion: Relation to D Cell/Somatostatin”. In: Frontiers in Pharmacology 7.Au-
gust, pp. 1–11. issn: 1663-9812.
Tamburrino, D., G. Spoletini, S. Partelli, F. Muffatti, O. Adamenko, S. Crippa, and M.
Falconi (2016). “Surgical management of neuroendocrine tumors.” In: Best practice &
research. Clinical endocrinology & metabolism 30.1, pp. 93–102. issn: 1878-1594.
412
Tan, E. H. (2011). “Imaging of gastroenteropancreatic neuroendocrine tumors”. In: World
Journal of Clinical Oncology 2.1, p. 28. issn: 2218-4333.
Thai, T. H., D. P. Calado, S Casola, K. M. Ansel, C Xiao, Y Xue, A Murphy, D Frendewey,
D Valenzuela, J. L. Kutok, M Schmidt-Supprian, N Rajewsky, G Yancopoulos, A Rao,
and K. Rajewsky (2007). “Regulation of the Germinal Center Response by MicroRNA-
155”. In: Science 316.5824, pp. 604–608.
Thakker, R. V. (2016). “Genetics of Parathyroid Tumours”. In: Journal of Internal
Medicine. issn: 1365-2796.
Thakker, R. V. (2014). “Multiple endocrine neoplasia type 1 (MEN1) and type 4 (MEN4)”.
In: Molecular and Cellular Endocrinology 386.1-2, pp. 2–15. issn: 03037207.
Thorns, C., C. Schurmann, N. Gebauer, H. Wallaschofski, C. Kumpers, V. Bernard, A. C.
Feller, T. Keck, J. K. Habermann, N. Begum, H. Lehnert, and G. Brabant (2014).
“Global MicroRNA profiling of pancreatic neuroendocrine Neoplasias”. In: Anticancer
Research 34.5, pp. 2249–2254. issn: 02507005.
Tischler, a. S. (1989). “The Dispersed Neuroendocrine Cells: The Structure, Function,
Regulation and Effects of Xenobiotics on this System”. In: Toxicologic Pathology 17.2,
pp. 307–316. issn: 0192-6233.
Tonelli, F., F. Giudici, F. Giusti, F. Marini, L. Cianferotti, G. Nesi, and M. L. Brandi
(2014). “A heterozygous frameshift mutation in exon 1 of cdkn1B gene in a patient
affected by MEN4 syndrome”. In: European Journal of Endocrinology 171.2. issn:
1479683X.
Ueno, K., H. Hirata, Y. Hinoda, and R. Dahiya (2013). “Frizzled homolog proteins,
microRNAs and Wnt signaling in cancer”. In: International Journal of Cancer 132.8,
pp. 1731–1740. issn: 00207136. arXiv: NIHMS150003.
413
Van Buren, G., A. Rashid, A. D. Yang, E. K. Abdalla, M. J. Gray, W. Liu, R. Somcio,
F. Fan, E. R. Camp, J. C. Yao, and L. M. Ellis (2007). “The development and char-
acterization of a human midgut carcinoid cell line.” In: Clinical cancer research : an
official journal of the American Association for Cancer Research 13.16, pp. 4704–12.
issn: 1078-0432.
Van Der Zwan, J. M., A. Trama, R. Otter, N. Larranaga, A. Tavilla, R. Marcos-Gragera,
A. P. Dei Tos, E. Baudin, G. Poston, and T. Links (2013). “Rare neuroendocrine
tumours: Results of the surveillance of rare cancers in Europe project”. In: European
Journal of Cancer 49.11, pp. 2565–78. issn: 1879-0852.
Van Peer, G., S. Lefever, J. Anckaert, A. Beckers, A. Rihani, A. Van Goethem, P. J.
Volders, F. Zeka, M. Ongenaert, P. Mestdagh, and J. Vandesompele (2014). “miRBase
Tracker: keeping track of microRNA annotation changes”. In: Database : the journal
of biological databases and curation 2014.bau080, pp. 1–8. issn: 17580463.
Vargas, A. J. and C. C. Harris (2016). “Biomarker development in the precision medicine
era: lung cancer as a case study”. In: Nature Reviews Cancer 16.8, pp. 525–537. issn:
1474-175X.
Vijayvergia, N., P. M. Boland, E. Handorf, K. S. Gustafson, Y. Gong, H. S. Cooper, F.
Sheriff, I. Astsaturov, S. J. Cohen, and P. F. Engstrom (2016). “Molecular profiling
of neuroendocrine malignancies to identify prognostic and therapeutic markers: a Fox
Chase Cancer Center Pilot Study”. In: British Journal of Cancer 115.5, pp. 564–570.
issn: 0007-0920.
Volinia, S., G. a. Calin, C.-G. Liu, S. Ambs, A. Cimmino, F. Petrocca, R. Visone, M. Iorio,
C. Roldo, M. Ferracin, R. L. Prueitt, N. Yanaihara, G. Lanza, A. Scarpa, A. Vecchione,
M. Negrini, C. C. Harris, and C. M. Croce (2006). “A microRNA expression signature
of human solid tumors defines cancer gene targets.” In: Proceedings of the National
414
Academy of Sciences of the United States of America 103.7, pp. 2257–61. issn: 0027-
8424.
Wang, F., K. Knutson, C. Alcaino, D. R. Linden, S. J. Gibbons, P. Kashyap, M. Grover,
R. Oeckler, P. A. Gottlieb, H. J. Li, A. B. Leiter, G. Farrugia, and A. Beyder (2017).
“Mechanosensitive ion channel Piezo2 is important for enterochromaffin cell response
to mechanical forces”. In: The Journal of Physiology 595.1, pp. 79–91. issn: 00223751.
Wang, H., Y. Chen, C. Fernandez-Del Castillo, O. Yilmaz, and V. Deshpande (2013).
“Heterogeneity in signaling pathways of gastroenteropancreatic neuroendocrine tumors:
a critical look at notch signaling pathway.” In: Modern pathology : an official journal
of the United States and Canadian Academy of Pathology, Inc 26, pp. 139–47. issn:
1530-0285.
Wang, J., G. Cortina, S. V. Wu, R. Tran, J.-H. Cho, M.-J. Tsai, T. J. Bailey, M. Jamrich,
M. E. Ament, W. R. Treem, I. D. Hill, J. H. Vargas, G. Gershman, D. G. Farmer, L.
Reyen, and M. G. Martın (2006). “Mutant neurogenin-3 in congenital malabsorptive
diarrhea”. In: The New England Journal of Medicine 355.3, pp. 270–280. issn: 1533-
4406.
Wang, Q., Y. Zhou, P. Rychahou, T. W.-M. Fan, A. N. Lane, H. L. Weiss, and B. M.
Evers (2016). “Ketogenesis contributes to intestinal cell differentiation”. In: Cell Death
and Differentiation, pp. 1–11. issn: 1350-9047.
Wardlaw, R. and J. W. Smith (2008). “Gastric carcinoid tumors”. In: The Ochsner Jour-
nal 8, pp. 191–196. issn: 00224790.
Watson, S. a., A. M. Grabowska, M. El-Zaatari, and A. Takhar (2006). “Gastrin - ac-
tive participant or bystander in gastric carcinogenesis?” In: Nature reviews. Cancer 6,
pp. 936–946. issn: 1474-175X.
415
Wee, N. K., D. C. Weinstein, S. T. Fraser, and S. J. Assinder (2013). “The mammalian
copper transporters CTR1 and CTR2 and their roles in development and disease”.
In: International Journal of Biochemistry and Cell Biology 45.5, pp. 960–963. issn:
13572725.
Weigel, M. T. and M. Dowsett (2010). “Current and emerging biomarkers in breast cancer:
Prognosis and prediction”. In: Endocrine-Related Cancer 17.4. issn: 13510088.
Weisbrod, A. B., D. J. Liewehr, S. M. Steinberg, E. E. Patterson, S. K. Libutti, W. M.
Linehan, N. Nilubol, and E. Kebebew (2012). “Association of type O blood with pan-
creatic neuroendocrine tumors in Von Hippel-Lindau syndrome.” In: Annals of surgical
oncology 19.6, pp. 2054–9. issn: 1534-4681.
Weisbrod, A. B., N. Nilubol, L. S. Weinstein, W. F. Simonds, S. K. Libutti, R. T. Jensen,
S. J. Marx, and E. Kebebew (2013). “Association of Type-O Blood with Neuroen-
docrine Tumors in Multiple Endocrine Neoplasia Type 1”. In: Journal of Clinical En-
docrinology & Metabolism 98.1, E109–E114. issn: 0021-972X.
Weiss, M, D. Steiner, and L. Philipson (2000). Insulin Biosynthesis, Secretion, Structure,
and Structure-Activity Relationships. Ed. by L. De Groot, G Chrousos, and K Dungan.
2000th ed. South Dartmouth (MA): Endotext, MDText.com, Inc., https://www.ncbi.nlm.nih.gov/books/NBK279029/.
Welin, S., M. Stridsberg, J. Cunningham, D. Granberg, B. Skogseid, K. Oberg, B. Eriks-
son, and E. T. Janson (2009). “Elevated plasma chromogranin a is the first indication
of recurrence in radically operated midgut carcinoid tumors”. In: Neuroendocrinology
89.3, pp. 302–307. issn: 00283835.
Whitaker-Azmitia, P. M. (1999). “The Discovery of Serotonin and its Role in Neuro-
science”. In: Neuropsychopharmacology 21, 2S–8S.
416
Wiedenmann, B, W. W. Franke, C Kuhn, R Moll, and V. E. Gould (1986). “Synapto-
physin: a marker protein for neuroendocrine cells and neoplasms.” In: Proceedings of
the National Academy of Sciences of the United States of America 83.10, pp. 3500–
3504. issn: 0027-8424.
Williams, D. R., S. A. Mohammed, J. J. Leavell, and C. Collins (2012). “Race, socioeco-
nomic status, and health: complexities, ongoing challenges, and research opportunities”.
In: Ann N Y Acad Sci 1189, pp. 69–101.
Williams, E. and M. Sandler (1963). “The classification of carcinoid tumours.” In: Lancet
1.7275, pp. 238–239.
Witwer, K. W. (2015). “Circulating MicroRNA biomarker studies: Pitfalls and potential
solutions”. In: Clinical Chemistry 61.1, pp. 56–63. issn: 15308561.
Wolin, E., B Jarzab, B Eriksson, T Walter, C Toumpanakis, M. Morse, P Tomassetti,
M. Weber, D. Fogelman, J Ramage, D Poon, B Gadbaw, J Li, J. Pasieka, A Ma-
hamat, F Swahn, J Newell-Price, W Mansoor, and K Oberg (2015). “Phase III study
of pasireotide long-acting release in patients with metastatic neuroendocrine tumors
and carcinoid symptoms refractory to available somatostatin analogues”. In: Drug De-
sign, Development and Therapy 2015.9, pp. 5075–5086.
Wollam, J., S. Mahata, M. Riopel, A. Hernandez-Carretero, A. Biswas, G. K. Bandyopad-
hyay, N.-W. Chi, L. E. Eiden, N. R. Mahapatra, A. Corti, N. J. G. Webster, and S. K.
Mahata (2017). “Chromogranin A regulates vesicle storage and mitochondrial dynamics
to influence insulin secretion”. In: Cell and Tissue Research Published.doi:10.1007/s00441-
017-2580-5. issn: 0302-766X.
Yachida, S., E. Vakiani, C. M. White, Y. Zhong, T. Saunders, R. Morgan, R. F. de
Wilde, A. Maitra, J. Hicks, A. M. Demarzo, C. Shi, R. Sharma, D. Laheru, B. H. Edil,
C. L. Wolfgang, R. D. Schulick, R. H. Hruban, L. H. Tang, D. S. Klimstra, C. A.
417
Iacobuzio-Donahue, T. Saunders, R. Morgan, R. F. D. Wilde, A. Maitra, J. Hicks,
A. M. Demarzo, C. Shi, R. Sharma, D. Laheru, B. H. Edil, C. L. Wolfgang, R. D.
Schulick, R. H. Hruban, L. H. Tang, D. S. Klimstra, and C. A. Iacobuzio-Donahue
(2012). “Small Cell and Large Cell Neuroendocrine Carcinomas of the Pancreas are
Genetically Similar and Distinct From Well-differentiated Pancreatic Neuroendocrine
Tumors”. In: Am J Surg Pathol 36.2, pp. 173–184. issn: 0147-5185.
Yamaguchi, T., T. Fujimori, S. Tomita, K. Ichikawa, H. Mitomi, K. Ohno, Y. Shida, and
H. Kato (2013). “Clinical validation of the gastrointestinal NET grading system: Ki67
index criteria of the WHO 2010 classification is appropriate to predict metastasis or
recurrence.” In: Diagnostic pathology 8, p. 65. issn: 1746-1596.
Yang, Q, N. a. Bermingham, M. J. Finegold, and H. Y. Zoghbi (2001). “Requirement of
Math1 for secretory cell lineage commitment in the mouse intestine.” In: Science (New
York, N.Y.) 294.5549, pp. 2155–2158. issn: 00368075.
Yang, X., Y. Yang, Z. Li, C. Cheng, T. Yang, C. Wang, L. Liu, and S. Liu (2015). “Di-
agnostic value of circulating chromogranin a for neuroendocrine tumors: A systematic
review and meta-analysis”. In: PLoS ONE 10.4, pp. 1–14. issn: 19326203.
Yang, Y., R. Chaerkady, K. Kandasamy, T.-C. Huang, L. D. N. Selvan, S. B. Dwivedi,
O. A. Kent, J. T. Mendell, and A. Pandey (2010). “Identifying targets of miR-143
using a SILAC-based proteomic approach”. In: Molecular BioSystems 6.10, p. 1873.
issn: 1742-206X.
Yang, Z., L. H. Tang, and D. S. Klimstra (2011). “Effect of tumor heterogeneity on
the assessment of Ki67 labeling index in well-differentiated neuroendocrine tumors
metastatic to the liver: implications for prognostic stratification.” In: The American
journal of surgical pathology 35.6, pp. 853–60. issn: 1532-0979.
418
Yao, J. C., M. H. Shah, T Ito, C. L. Bohas, E. M. Wolin, E Van Cutsem, T. J. Hobday,
T Okusaka, J Capdevila, E. G. de Vries, P Tomassetti, M. E. Pavel, S Hoosen, T Haas,
J Lincy, D Lebwohl, K Oberg, and T. T.S. G. Rad001 in Advanced Neuroendocrine
Tumors (2011). “Everolimus for advanced pancreatic neuroendocrine tumors”. In: N
Engl J Med 364.6, pp. 514–523. issn: 1533-4406. arXiv: NIHMS150003.
Yao, J. C., M. Hassan, A. Phan, C. Dagohoy, C. Leary, J. E. Mares, E. K. Abdalla,
J. B. Fleming, J. N. Vauthey, A. Rashid, and D. B. Evans (2008). “One hundred years
after ”carcinoid”: Epidemiology of and prognostic factors for neuroendocrine tumors in
35,825 cases in the United States”. In: Journal of Clinical Oncology 26.18, pp. 3063–
3072. issn: 0732183X.
Yao, J. C., N. Fazio, S. Singh, R. Buzzoni, C. Carnaghi, E. Wolin, J. Tomasek, M.
Raderer, H. Lahner, M. Voi, L. B. Pacaud, N. Rouyrre, C. Sachs, J. W. Valle, G. D.
Fave, E. Van Cutsem, M. Tesselaar, Y. Shimada, D. Y. Oh, J. Strosberg, M. H. Kulke,
and M. E. Pavel (2016). “Everolimus for the treatment of advanced, non-functional
neuroendocrine tumours of the lung or gastrointestinal tract (RADIANT-4): A ran-
domised, placebo-controlled, phase 3 study”. In: The Lancet 387.10022, pp. 968–977.
issn: 1474547X.
Yates, C. J., P. J. Newey, and R. V. Thakker (2015). “Challenges and controversies in
management of pancreatic neuroendocrine tumours in patients with MEN1.” In: The
lancet. Diabetes & endocrinology 3.11, pp. 895–905. issn: 2213-8595.
Ye, D. Z. and K. H. Kaestner (2009). “Foxa1 and Foxa2 Control the Differentiation
of Goblet and Enteroendocrine L- and D-Cells in Mice”. In: Gastroenterology 137.6,
pp. 2052–2062. issn: 00165085. arXiv: NIHMS150003.
419
Young, G. P., E. L. Symonds, J. E. Allison, S. R. Cole, C. G. Fraser, S. P. Halloran, E. J.
Kuipers, and H. E. Seaman (2015). “Advances in Fecal Occult Blood Tests: The FIT
Revolution”. In: Digestive Diseases and Sciences 60.3, pp. 609–622. issn: 15732568.
Yu, D, C Jin, J Leja, N Majdalani, B Nilsson, F Eriksson, and M Essand (2011). “Ade-
novirus with hexon Tat-protein transduction domain modification exhibits increased
therapeutic effect in experimental neuroblastoma and neuroendocrine tumors”. In: J
Virol 85.24, pp. 13114–13123. issn: 0022-538X.
Zatelli, M. C., G. Fanciulli, P. Malandrino, V. Ramundo, A. Faggiano, and A. Colao
(2016). “Predictive factors of response to mTOR inhibitors in neuroendocrine tu-
mours”. In: Endocrine-Related Cancer 23.3, R173–R183. issn: 1351-0088.
Zhan, H. X., L. Cong, Y. P. Zhao, T. P. Zhang, and G. Chen (2013). “Risk factors for
the occurrence of insulinoma: A Case-control study”. In: Hepatobiliary and Pancreatic
Diseases International 12.3, pp. 324–328. issn: 14993872.
Zhang, H.-Y., K. M. Rumilla, L. Jin, N. Nakamura, G. a. Stilling, K. H. Ruebel, T. J.
Hobday, C. Erlichman, L. a. Erickson, and R. V. Lloyd (2006). “Association of DNA
methylation and epigenetic inactivation of RASSF1A and beta-catenin with metastasis
in small bowel carcinoid tumors.” In: Endocrine 30.3, pp. 299–306. issn: 1355-008X.
Zhang, J., Y. Song, C. Zhang, X. Zhi, H. Fu, Y. Ma, Y. Chen, F. Pan, K. Wang, J. Ni, W.
Jin, X. He, H. Su, and D. Cui (2015). “Circulating MiR-16-5p and MiR-19b-3p as two
novel potential biomarkers to indicate progression of gastric cancer”. In: Theranostics
5.7, pp. 733–745. issn: 18387640.
Zhang, S., H. Liu, C. L. Chuang, X. Li, M. Au, L. Zhang, A. R. J. Phillips, D. W. Scott,
and G. J. S. Cooper (2014). “The pathogenic mechanism of diabetes varies with the
degree of overexpression and oligomerization of human amylin in the pancreatic islet
beta cells”. In: FASEB Journal 28.12, pp. 5083–5096. issn: 15306860.
420
Zhou, G., J. Sinnett-Smith, S. H. Liu, J. Yu, J. Wu, R. Sanchez, S. J. Pandol, R. Abrol, J.
Nemunaitis, E. Rozengurt, and F. C. Brunicardi (2014). “Down-regulation of pancreatic
and duodenal homeobox-1 by somatostatin receptor subtype 5: A novel mechanism for
inhibition of cellular proliferation and insulin secretion by somatostatin”. In: Frontiers
in Physiology 5 JUN.June, pp. 1–7. issn: 1664042X.
Zwan, W. van der, L Bodei, J Mueller-Brand, W. de Herder, L. Kvols, and D. Kwekke-
boom (2015). “GEPNETs update: Radionuclide therapy in neuroendocrine tumors”.
In: European journal of endocrinology 172.1, R1–R8. issn: 1479683X.
421
A. Sample ID dataset 1
The sample ID numbers for the samples included in Dataset 1 are shown in Table A.1
below. Clinical details for Dataset 1 are shown in Table A.2.
Table A.1.: Sample ID of FFPE tissue available for miRNA analysis (Dataset 1)
Patient No. Sample IDSB-NET
SBadjacentnormal
LNmetastasis
LNnormal
Livermetastasis
Liveradjacentnormal
S1 1.1 1.2 1.3 1.4S2 1.5 1.6 1.7 1.8 1.9 2.0S3 2.1 2.2S4 2.3 2.4 2.5 2.6*S5 2.7 2.8 6.1S6 2.9S7 3.0 3.1 3.2 3.3S8 3.4 3.5 3.6 3.7S9 3.8 3.9 4.0 4.1 4.2 4.3S10 4.4 4.5 4.6S11 4.7S12 4.8S13 5.2 5.3 5.4 5.5S14 5.6 5.7S15 5.9 6.0
*: excluded from qPCR validation, insufficient RNA SB: Small Bowel LN: Lymph Node
422
Table A.2.: Clinical details miRNA Dataset 1Pa-tientno.
Gen-der
Age Ki-67 % Grade Tumour stage Functioning Multifocalprimary
Angiolym-phaticinvasion
Perineuralinvasion
Livermetastases
Syn-chronous/metachronous
Additionalmetastases?
Patientdied?
S1 F 76 3 G2 T3N1M1 no no yes yes yes synchronous no noS2 M 81 1 G1 T4N1M1 Carcinoid no yes not
mentionedyes synchronous no yes
S3 F 75 1-2 G1 T3N1M1 no no yes yes yes synchronous no noS4 M 38 < 2 G1 T2N1M0 no no yes not
mentionedno N/A no no
S5 F 59 < 1 G1 T2N1M1 Carcinoid no no no yes synchronous no noS6 F 69 < 1 G1 T3N1M0 no no yes no no N/A no noS7 F 57 < 0.5 G1 T4N1M1 no no yes yes yes synchronous uterus,
omentum,ovaries
no
S8 F 84 < 1 G1 T4N1M0 no yes yes yes no N/A no noS9 M 83 2 G1 T3N1M1 no no yes yes yes synchronous no yesS10 M 69 < 2 G1 T3N1M0 no no no no no N/A no noS11 M 75 < 2 G1 T3N0M0 no no no no no N/A no noS12 M 59 < 2 G1 T4N1M0 Carcinoid no yes yes no N/A no noS13 M 77 4-5 G2 T4N1M1 Carcinoid yes yes yes no N/A peritoneum yesS14 M 60 2-3 G2 T3N1M1 no yes yes not
mentionedyes synchronous no no
S15 F 61 1 G1 T1N0M1 no yes notmentioned
notmentioned
yes synchronous no no
423
B. Primers for qPCR
The primers for the qPCR experiments were all Taqman® primers (Life Technologies
Ltd.). Table B.1 shows the full name and miRBase accession number for each miRNA and
the catalogue number/assay ID for the Taqman® primers used in the reverse transcription
and qPCR experiments.
Table B.1.: miRNA primers for qPCR
Target Database, accession number Catalogue, assay IDhsa-miR-215-5p miRBase, MIMAT0000272 4427975, 000518hsa-miR-378i miRBase, MIMAT0019074 4427975, 464668 mathsa-miR-378a-3p miRBase, MIMAT0000732 4427975, 001314hsa-miR-451a miRBase, MIMAT0001631 4427975, 001141hsa-miR-7-5p miRBase, MIMAT0000252 4427975, 000268hsa-miR-204-5p miRBase, MIMAT0000265 4427975, 000508hsa-miR-375 miRBase, MIMAT0000728 4427975, 000564hsa-miR-1-3p(hsa-miR-1)
miRBase, MIMAT0000416 4427975, 002222
hsa-miR-143-3p miRBase, MIMAT0000435 4427975, 002249hsa-miR-1233-3p(hsa-miR-1233)
miRBase, MIMAT0005588 4427975, 002768
Small nuclear RNA,RNU6-1 (U6)
NCBI, NR 004394 4427975, 001973
Small nucleolar RNA,SNORD44 (RNU44)
NCBI, NR 002750 4427975, 001094
424
C. RNA extractions
The tables below, Table C.1, Table C.2, and Table C.3 show the quantity and quality of
RNA obtained from the RNA extractions for Datasets 1 and 2.
Table C.1.: Dataset 1, RNA extractions for quantification NanoString
Sample ID Concentration ng/µL 260/280 260/2301.1 84.1 1.74 1.381.2 161.6 1.97 2.151.3 232.3 1.99 2.161.4 237.6 1.93 2.211.5 234.3 1.93 1.871.6 74.5 1.97 1.991.7 351.4 1.91 1.781.8 323.4 1.93 2.151.9 422.2 1.91 1.722 550.2 1.97 2.122.1 217.4 1.91 1.872.2 132.1 1.93 2.072.3 215 1.92 2.22.4 76.3 1.93 2.032.5 235.8 1.93 2.32.6 146.8 1.94 2.32.7 200.3 1.99 2.082.8 100 1.95 2.222.9 130 1.95 2.093 117.5 1.9 1.813.1 122.6 1.93 2.163.2 400.2 1.88 2.093.3 74.7 1.88 2.223.4 459.7 1.88 1.863.5 143.9 2 2.22
425
Continuation of Table C.1
Sample ID Concentration ng/µL 260/280 260/2303.6 199.2 1.92 2.183.7 394.3 1.91 2.253.8 139.3 1.95 2.033.9 126.9 1.99 2.084 223.5 1.93 2.34.1 93.1 1.98 2.144.2 410 1.92 1.964.3 356.5 1.99 2.084.4 380.5 1.96 1.984.5 87.1 1.95 2.154.6 185.1 1.81 1.44.7 150.8 1.94 24.8 95.2 1.86 1.995.2 182.8 1.97 2.055.3 76.8 1.93 2.175.4 191.1 1.94 2.235.5 552.4 1.97 2.255.6 157.2 2.02 1.925.7 116.7 1.96 2.135.8 119.3 1.96 2.215.9 106.2 1.92 2.196 89.5 1.94 2.166.1 269.5 2 2.04
Table C.2.: Dataset 1, RNA extractions for quantification qPCR
Sample ID Concentration ng/µL 260/280 260/2301.1 295.7 2 2.041.2 86.2 1.99 2.151.3 261.5 1.92 2.31.4 306.8 1.92 2.281.5 94 1.91 2.031.6 69.4 1.97 2.061.7 314.3 1.88 2.111.8 212.9 1.92 2.251.9 160.1 1.99 2.12 260.9 2.02 2.172.1 186.2 1.89 2.112.2 135.6 1.98 2.22.3 107.7 1.87 2.16
426
Continuation of Table C.2
Sample ID Concentration ng/µL 260/280 260/2302.4 62.1 1.9 1.992.5 316.2 1.89 2.282.7 194.2 1.94 2.232.8 111.3 1.96 2.22.9 105.1 1.91 2.213 74.4 1.85 1.753.1 79.7 1.93 2.143.2 85.3 1.88 2.133.3 121 1.89 2.233.4 209.2 1.96 2.023.5 218.7 1.96 2.183.6 242.5 1.93 2.193.7 399.3 1.91 2.263.8 170.5 1.98 2.123.9 71.6 1.95 2.094 148.9 1.93 2.174.1 289.9 1.93 2.294.2 158 1.95 2.114.3 401.9 1.97 2.124.4 224.6 1.88 2.074.5 143.9 1.94 2.224.6 257.5 1.92 2.114.7 157.7 1.89 2.134.8 119.5 1.85 2.065.2 142.6 1.91 2.125.3 67.5 1.94 2.075.4 115.2 1.88 2.185.5 235.6 1.92 2.215.6 88.5 1.95 2.085.7 184.6 2 2.255.9 103.1 1.96 2.216 81.7 1.94 2.126.1 132.1 1.96 2.22
Table C.3.: Dataset 2, RNA extractions for quantification NanoString
Sample ID Concentration ng/µL 260/280 260/2300.1 385.8 2.01 2.080.2 1227.6 2.05 2.247.7 1159.7 2.06 2.2
427
Continuation of Table C.3
Sample ID Concentration ng/µL 260/280 260/2307.8 1582.8 2.1 2.38.0 352.2 2.05 2.168.1 944.7 2.09 2.168.3 1743.3 2.11 2.318.6 1470.1 2.1 2.298.7 1663.4 2.1 2.298.8 463.1 2.02 2.188.9 1399 2.09 2.249.0 249.6 2.03 2.129.1 1259.5 2.06 2.199.2 106.3 2.07 2.229.3 1666.7 2.08 2.269.4 1966.7 2.07 2.249.5 1857.5 2.1 2.279.6 4434.1 1.68 1.879.7 562.5 2.09 2.059.8 894.7 2.08 1.999.9 652.6 2.11 2.1810.1 729 2.1 2.1510.2 722 2.09 2.1810.4 1173.8 2.11 2.2510.6 1493 2.1 2.2710.7 1629 2.08 2.2110.8 567.9 2.11 2.1210.9 662.5 2.08 2.2211.0 835 2.1 2.2611.2 1245 2.08 2.2411.3 161.5 2.07 2.2111.4 1287.4 2.07 2.2211.5 938 2.01 1.7311.6 850.5 2.09 2.2611.7 1903.2 2.09 2.311.8 36.1 2.02 2.0912.0 2221.3 2.06 2.2312.1 1444.3 2.07 2.2212.2 947.1 2.09 2.2312.3 1075.4 2.12 2.2912.4 410.4 2.06 2.2512.5 1437.9 2.07 2.2412.6 903.3 2.08 2.19
428
D. Dysregulated miRNA
Table D.1 shows significantly dysregulated miRNA in lymph node metastases versus
normal tissue from dataset 1 (FDR: < 0.05) . Table D.2, shows the 106 miRNA that
were significantly dysregulated in the SBNET relative “normal” small bowel tissue for
dataset 2 (FDR < 0.05).
Table D.1.: Significantly dysregulated miRNA in lymph node metastases versus normaltissue
miRNA log2FC FDRmiR-518b -0.8 0.026317893miR-1247-5p -0.7 0.003362739miR-572 -0.7 0.013975255miR-212-3p -0.4 0.049578071miR-1260b 0.5 0.043968467miR-362-3p 0.8 0.046142383miR-542-5p 0.8 0.029007409miR-210 0.8 0.026317893miR-335-5p 0.8 0.040282004miR-340-5p 0.8 0.035571455miR-30c-5p 0.9 0.046299278miR-139-3p 0.9 0.027076363miR-186-5p 0.9 0.049578071miR-1468 0.9 0.006880079miR-505-3p 0.9 0.025304494miR-409-3p 0.9 0.034325281miR-106b-5p 0.9 0.046299278miR-425-5p 1.0 0.04908112miR-125b-5p 1.0 0.025304494miR-500a-5p+501-5p 1.0 0.008350497miR-199b-5p 1.0 0.021213173
429
Continuation of Table D.1
miRNA log2FC FDRmiR-28-5p 1.0 0.005264096miR-100-5p 1.0 0.013987454miR-215 1.0 0.02194981miR-342-3p 1.0 0.038084306miR-532-3p 1.0 0.043787092miR-130b-3p 1.1 0.003276445let-7b-5p 1.1 0.03768921miR-365a-3p 1.1 0.012838063miR-874 1.1 0.014442499miR-374a-5p 1.1 0.023752601miR-423-3p 1.1 0.006198659miR-744-5p 1.1 0.014268138miR-374b-5p 1.2 0.013363771miR-135a-5p 1.2 0.002211389miR-132-3p 1.2 0.002423989miR-93-5p 1.2 0.009080377miR-328 1.2 0.002423989miR-23a-3p 1.3 0.011433857miR-15a-5p 1.3 0.013967733miR-191-5p 1.3 0.005794072miR-26b-5p 1.3 0.00664561miR-320a 1.3 0.00314078miR-423-5p 1.3 0.00580949miR-582-5p 1.3 0.000417834miR-127-3p 1.3 0.010196984miR-338-3p 1.4 0.004251624miR-26a-5p 1.4 0.006554137let-7g-5p 1.4 0.005928357let-7a-5p 1.4 0.005917467miR-421 1.4 0.000541669miR-151a-5p 1.4 0.001120897miR-1206 1.4 0.000761827miR-29c-3p 1.5 0.004597285miR-34a-5p 1.5 0.003394165miR-30d-5p 1.5 0.003177246miR-660-5p 1.5 0.001893788miR-22-3p 1.5 0.002423989miR-615-3p 1.5 6.13176E-05miR-29b-3p 1.5 0.003768205let-7f-5p 1.6 0.001173095miR-181b-5p+181d 1.6 0.001893788miR-128 1.6 0.000528183
430
Continuation of Table D.1
miRNA log2FC FDRmiR-454-3p 1.6 0.000128909miR-652-3p 1.7 0.000221524let-7i-5p 1.7 0.000514845miR-486-3p 1.7 0.000248478miR-24-3p 1.8 0.00036816miR-361-5p 1.8 0.000101852miR-642a-5p 1.9 3.06955E-05let-7d-5p 1.9 0.000161867miR-148b-3p 1.9 4.86516E-05miR-551b-3p 1.9 0.000128909miR-331-3p 2.0 2.50784E-05miR-330-3p 2.0 3.87887E-06miR-125a-5p 2.1 6.97908E-06miR-107 2.1 1.59442E-05miR-129-5p 2.1 5.21896E-06miR-181c-5p 2.1 1.97512E-05miR-532-5p 2.2 6.54536E-07miR-98 2.2 3.87887E-06miR-196a-5p 2.3 6.10305E-06let-7e-5p 2.3 1.59833E-06miR-27b-3p 2.4 4.49538E-07miR-301a-3p 2.4 2.08717E-07miR-324-5p 2.4 1.51052E-06miR-129-2-3p 2.5 4.40272E-06miR-99b-5p 2.5 7.2765E-08miR-96-5p 2.5 4.6921E-07miR-489 2.5 1.59833E-06miR-23b-3p 2.5 2.38457E-07miR-204-5p 2.6 4.54034E-07miR-95 2.6 5.31588E-08miR-1180 2.7 1.22316E-08miR-183-5p 2.8 3.72669E-08miR-137 3.0 5.86976E-10miR-182-5p 3.0 5.77696E-10miR-429 3.2 2.49704E-11miR-192-5p 3.3 1.24899E-11miR-141-3p 3.4 4.69356E-13miR-194-5p 3.5 2.68307E-13miR-7-5p 3.7 1.0035E-13miR-200b-3p 3.8 3.38215E-14miR-200a-3p 3.9 9.91134E-15miR-375 4.2 6.27468E-17
431
Continuation of Table D.1
miRNA log2FC FDRmiR-200c-3p 4.2 3.14223E-18
MiRNA had a FDR < 0.05.
Table D.2.: miRNA that were significantly dysregulated in SBNET relative to “normal”small bowel tissue
Upregulated miRNA (dataset 2) Downregulated miRNA (dataset 2)miRNA log2FC FDR miRNA log2FC FDRmiR-186-5p 0.994 0.048992697 miR-3180 -3.224 2.42985E-08miR-328 1.031 0.021326394 miR-31-5p -1.542 0.000622114miR-423-5p 1.046 0.048992697 miR-548aa -1.394 0.003173223miR-423-3p 1.077 0.029414165 miR-1299 -1.292 0.001453093miR-3200-3p 1.099 0.040816526 miR-548h-5p -1.258 0.005813017miR-299-3p 1.132 0.048992697 miR-329 -1.204 0.005407391miR-653 1.144 0.00068917 miR-770-5p -1.075 0.011899449miR-361-3p 1.147 0.015050766 miR-638 -1.045 0.042201986miR-371a-5p 1.149 0.036157756 miR-663a -1.035 0.009462954miR-197-3p 1.217 0.032588111 miR-581 -1.008 0.011139264miR-221-3p 1.219 0.047787816 miR-892b -1.005 0.017439505miR-652-3p 1.238 0.004989686 miR-1302 -1.002 0.015050766miR-151a-5p 1.258 0.007446208 miR-1256 -0.984 0.039768879miR-330-5p 1.326 0.004989686 miR-302d-3p -0.966 0.012158394miR-628-5p 1.328 0.011899449 miR-548ai -0.965 0.012327891miR-365a-3p 1.343 0.019746543 miR-320b -0.951 0.010521253miR-542-5p 1.374 0.019746543 miR-934 -0.944 0.01004802miR-331-3p 1.383 0.0237593 miR-519b-3p -0.919 0.011606903miR-330-3p 1.394 0.001756363 miR-1197 -0.909 0.017184072miR-22-3p 1.446 0.015547491 miR-761 -0.905 0.034671491miR-335-5p 1.448 0.00494862 miR-512-3p -0.900 0.038155946miR-615-3p 1.508 0.009462954 miR-624-3p -0.845 0.035864247let-7c 1.519 0.041537624 miR-762 -0.816 0.019171432miR-1468 1.521 0.021277504 miR-550a-5p -0.806 0.011945207miR-23b-3p 1.555 0.035290968 miR-
526a+520c-5p+518d-5p
-0.799 0.033167663
miR-181b-5p+181d
1.566 0.003106112 miR-586 -0.776 0.017184072
miR-16-5p 1.595 0.039458926 miR-297 -0.736 0.035864247miR-361-5p 1.599 0.003106112 miR-595 -0.723 0.032588111
432
Continuation of Table D.2
Upregulated miRNA (dataset 2) Downregulated miRNA (dataset 2)miRNA log2FC FDR miRNA log2FC FDRmiR-1180 1.600 0.019746543 miR-548ak -0.721 0.035864247miR-664-3p 1.603 0.00305366 miR-1538 -0.715 0.036814546miR-27b-3p 1.674 0.021083682miR-24-3p 1.741 0.011899449miR-132-3p 1.750 0.0123323miR-135a-5p 1.782 0.039768879miR-196a-5p 1.790 0.038155946miR-30b-5p 1.800 0.021083682let-7i-5p 1.801 0.015664337miR-421 1.805 0.00068917miR-128 1.851 0.003173223miR-342-3p 1.852 0.013795817miR-374b-5p 1.853 0.003903987miR-324-5p 1.856 0.004989686miR-505-3p 1.856 0.003135006miR-30c-5p 1.868 0.005875075miR-29c-3p 1.868 0.009462954miR-99b-5p 1.934 0.004989686miR-4284 1.938 0.003710895let-7f-5p 1.940 0.012158394miR-362-3p 1.942 0.00649524miR-1206 1.949 0.017439505miR-129-5p 1.951 0.012327891miR-660-5p 1.956 0.005681993miR-582-5p 1.965 0.003173223miR-551b-3p 1.968 0.017439505miR-96-5p 1.992 0.019746543miR-182-5p 2.033 0.017184072miR-340-5p 2.042 0.003173223miR-200a-3p 2.045 0.011945207miR-148b-3p 2.048 0.003106112miR-34a-5p 2.098 0.003566272miR-454-3p 2.107 0.001168798miR-98 2.115 0.003106112miR-429 2.115 0.008470798miR-183-5p 2.163 0.011899449miR-107 2.185 0.002905876miR-26a-5p 2.194 0.003106112miR-181c-5p 2.259 0.000712133miR-129-2-3p 2.393 0.003903987miR-204-5p 2.478 0.003106112
433
Continuation of Table D.2
Upregulated miRNA (dataset 2) Downregulated miRNA (dataset 2)miRNA log2FC FDR miRNA log2FC FDRmiR-642a-5p 2.541 0.00068917miR-301a-3p 2.565 0.0002353miR-7-5p 2.730 0.001217902miR-95 2.937 5.55994E-05miR-375 3.065 0.000154715miR-489 3.256 2.6165E-05miR-137 3.324 2.23998E-05
434
E. Bioinformatics
E.1. Genes list lymph node metastases
There were 278 genes with increased expression in lymph node metastases (versus SB-
NET) that were also predicted gene targets of both miR-1 and miR-143-3p:
ADRBK2 MON2 FKBP5 NGFR ACTN2 DIEXF INTS6 GPD2 CACNG8
SYNRG THEM4 AP2B1 POFUT1 RIN3 NPTX1 SP100 ZNF217 TCEANC2
QKI MRPL30 TRDMT1 SLC9A5 MTUS2 RBM27 COPA B4GALT1 CRAMP1L
POLH ZDHHC21 SYNCRIP FLRT2 PGR TEX35 PDE7A ARHGEF7 ADAMTS16
CUL3 RNF138 DNAJB1 RBBP9 ALX4 ZNF33A SCN11A G2E3 MYO1E
MAGT1 PTCD3 CAPRIN1 TIGIT FN1 SLC35E2 DIDO1 KIAA2022 SF3A1
DCP2 VSIG1 C5orf22 GNL3L TICRR RNGTT PALM2 MOXD1 ASXL2
CHM ZNF770 PAX9 NUS1 EPPIN PRTG PDS5A KIAA1549 GNPNAT1
ZFP91 RALGAPB IL2RA SSH1 TMOD3 COL5A2 NCBP1 ADAL THSD7A
POLH ZDHHC21 EMP1 BCL2 TGIF2 SLC1A3 PGR TEX35 XK TFAP2B
ALDH3A2 STK16 RBM33 TUBB ORAI2 TMPRSS13 SRD5A3 PRKCE CDC42SE1
CD244 DNAJC11 SEC22C EGFR RAB11FIP4 CDKL3 KANK4 WWC1 CNT-
NAP2 METTL21A RIC1 ZNF275 TCF20 TMEM245 SLC1A2 TMEM260
ZCCHC8 GTPBP10 MARCKS BMP7 TOR3A RNF165 TCAF2 MTPN CCPG1
KDELC2 TRA2B SRGAP1 TTC19 TKT GABRA4 SNX29 PTGIR ZMYND11
TMEM237 PPP1R9B PLEKHA2 WSCD2 NETO1 KRAS CNGB1 TBC1D20
435
ATG12 ABCC4 KLF12 SRSF1 KIF3C HNMT GPX5 ITGA11 RICTOR CSNK1G3
STRN MED1 IL17RA FXR1 FMNL2 KNCN AGO1 NAV3 FAM122C TPM4
SLC4A8 NQO1 COL4A4 FGF1 HUS1 RFWD3 SERPINB8 IGFBP5 SCAI
WISP1 SNX30 ADAMTSL1 ADORA3 SLC7A11 RAPGEF6 PTPRE ZBTB46
KCNJ13 APPL1 STAMBP MFAP3L PTPN14 SAMD8 ADCYAP1R1 KLF8
PLEKHA5 HOGA1 KIAA0226L SLC12A3 IL6ST SH3TC2 FAM91A1 RS1
FUT9 CNGA2 LAMP2 ERBB4 NPR3 SLC17A7 MTR EYA4 IGF2BP1 PLEKHG2
SEMA5A JOSD1 SLC41A1 TMEM97 BACH2 TRIM35 SLC35E1 ASPH FMNL3
PIK3R5 RASSF5 RAP2B SNX13 EDEM1 BAAT AKAP11 ADAM22 CHST11
PTPN2 PCSK5 SPTBN1 UQCC1 FOSB ADAM10 TNFRSF10D LGALS8
C6orf25 NT5C2 DLGAP2 TMED5 CGNL1 SFXN5 TVP23C SF3B3 IGF1
COX18 HECA INTS2 IKZF2 MED14 ENPP1 UBE2R2 HHLA2 NRAS IFI44L
CREM MAPK1 CBL C1orf109 DENND1B NUAK2 CA12 SLC7A6 EFR3B
PAX7 SLC24A1 ATP6V1A AGMAT TBC1D28 CD28 SLC38A2 CYFIP2
EBF1 ZNF236 PDCD2 CFLAR ANKRD29 RRAGD TPM3 ANKRD6 ARCN1
NKTR QRSL1
E.2. Enriched gene ontology terms SBNET
The top 10 enriched gene ontology terms from the bioinformatics analysis (DAVID) of the
downregulated predicted gene targets of the upregulated candidate miRNA in SBNET
are shown in Table E.1. For the full DAVID results for enriched gene ontology terms
see Supplementary Table 3 (Miller et al., 2016). None of the the enriched gene ontology
terms identified for miR-7-5p, miR-204-5p and miR-375 reached statistical significance
using a FDR of < 0.05.
436
Table E.1.: Top 10 enriched gene ontology terms for the predicted gene targets of miR-7-5p, miR-204-5p and miR-375Down-
regu-
lated
Gene ontology term Genes Count % List Pop Pop Fold
gene
targets
total hits total enrichment P Value Bonferroni Benjamini FDR
miR-7 GO:0006814 sodium ion
transport
SGK1, SLC12A2, SLC5A1,
SLC22A4, SLC4A7, SLC4A4,
SCNN1A, SLC5A12
8 5.096 129 130 13528 6.5 2.32E-04 0.26612 0.26612 0.37945
miR-7 GO:0006811 ion transport SLC36A1, SGK1, SLC39A11,
SLC12A2, SLC5A1, SLC22A23,
SLC26A2, ATP5G3, TMEM37,
CLIC5, SLC22A4, SLC30A4,
SLC4A7, PLLP, SLC4A4,
SLC31A2, SLC1A1, SCNN1A,
SLC5A12
19 12.102 129 768 13528 2.6 3.22E-04 0.34871 0.19297 0.52546
miR-7 GO:0006812 cation
transport
SLC36A1, SGK1, SLC12A2,
SLC39A11, SLC5A1, ATP5G3,
TMEM37, SLC22A4, SLC30A4,
SLC4A7, SLC4A4, SLC31A2,
SCNN1A, SLC5A12
14 8.917 129 553 13528 2.7 0.00222 0.94834 0.62757 3.57515
miR-7 GO:0010604 positive
regulation of
macromolecule metabolic
process
CDK1, CDX1, MYO6, PRKAG2,
PPARG, NDFIP2, EHF, IL6R,
MECOM, PPARGC1A, HMGA1,
PPARGC1B, GATA5, NEDD4,
BCL11B, BCL3, DYRK2, KLF4
18 11.465 129 857 13528 2.2 0.00292 0.97983 0.62313 4.68278
miR-7 GO:0015672 monovalent
inorganic cation transport
SLC36A1, SGK1, SLC12A2,
SLC5A1, SLC22A4, SLC4A7,
SLC4A4, SCNN1A, ATP5G3,
SLC5A12
10 6.369 129 318 13528 3.3 0.00323 0.98665 0.57822 5.16518
miR-7 GO:0030217 T cell
differentiation
CD8A, BCL11B, RELB, BCL3,
HLA-DMA
5 3.185 129 65 13528 8.1 0.00332 0.98806 0.52192 5.29503
miR-7 GO:0030001 metal ion
transport
TMEM37, SGK1, SLC12A2,
SLC39A11, SLC5A1, SLC22A4,
SLC30A4, SLC4A7, SLC31A2,
SLC4A4, SCNN1A, SLC5A12
12 7.643 129 465 13528 2.7 0.00462 0.99790 0.58560 7.29657
miR-7 GO:0002250 adaptive
immune response
CD8A, RELB, FCER1G, BCL3,
HLA-DMA
5 3.185 129 77 13528 6.8 0.00608 0.99970 0.63787 9.50199437
Continuation of Table E.1
Down-
regu-
lated
Gene ontology term Genes Count % List Pop Pop Fold
gene
targets
total hits total enrichment P Value Bonferroni Benjamini FDR
miR-7 GO:0002460 adaptive
immune response based
on somatic recombination
of immune receptors built
from immunoglobulin
superfamily domains
CD8A, RELB, FCER1G, BCL3,
HLA-DMA
5 3.185 129 77 13528 6.8 0.00608 0.99970 0.63787 9.50199
miR-7 GO:0046632 alpha-beta T
cell differentiation
BCL11B, RELB, BCL3 3 1.911 129 14 13528 22.5 0.00750 0.99996 0.67228 11.60567
miR-204 GO:0055114 oxidation
reduction
XDH, ACOX1, HSD17B2,
CYP2C18, FDX1, PTGS1, ADH5,
DECR1, PPARGC1A, ALDH3A2,
ACOX3, HSDL2, RRM2, SDHD,
CYBRD1, OXNAD1, NQO1,
RETSAT
18 11.613 129 639 13528 3.0 1.08E-04 0.10686 0.10686 0.17083
miR-204 GO:0019395 fatty acid
oxidation
ACOX1, DECR1, PPARGC1A,
ACOX3
4 2.581 129 39 13528 10.8 0.00590 0.99801 0.95534 8.97709
miR-204 GO:0034440 lipid
oxidation
ACOX1, DECR1, PPARGC1A,
ACOX3
4 2.581 129 39 13528 10.8 0.00590 0.99801 0.95534 8.97709
miR-204 GO:0016042 lipid
catabolic process
ACOX1, PLCB3, PAFAH2,
PLA2G12B, LIPH, DECR1,
ACOX3
7 4.516 129 173 13528 4.2 0.00602 0.99825 0.87958 9.16002
miR-204 GO:0051384 response to
glucocorticoid stimulus
SDC1, ALDOB, PTGS1, IL6R,
PPARGC1B
5 3.226 129 78 13528 6.7 0.00636 0.99878 0.81296 9.64700
miR-204 GO:0009725 response to
hormone stimulus
SDC1, EIF4EBP2, ALDOB,
PTGS1, PDGFRA, IL6R, GNG12,
CCNA2, PPARGC1B, PCK1
10 6.452 129 367 13528 2.9 0.00811 0.99981 0.81960 12.15031
miR-204 GO:0031960 response to
corticosteroid stimulus
SDC1, ALDOB, PTGS1, IL6R,
PPARGC1B
5 3.226 129 85 13528 6.2 0.00858 0.99988 0.77910 12.80886
miR-204 GO:0006091 generation of
precursor metabolites and
energy
ACOX1, SUCLG2, FDX1,
ALDOB, SDHD, CYBRD1,
PPARGC1A, ATP5G3, PCK1
9 5.806 129 313 13528 3.0 0.00978 0.99997 0.77139 14.46805
438
Continuation of Table E.1
Down-
regu-
lated
Gene ontology term Genes Count % List Pop Pop Fold
gene
targets
total hits total enrichment P Value Bonferroni Benjamini FDR
miR-204 GO:0048545 response to
steroid hormone stimulus
SDC1, ALDOB, PTGS1,
PDGFRA, IL6R, CCNA2,
PPARGC1B
7 4.516 129 192 13528 3.8 0.00983 0.99997 0.72689 14.53610
miR-204 GO:0006631 fatty acid
metabolic process
ACOX1, PTGS1, DECR1,
SLC27A2, PPARGC1A, ACOX3,
ACSL5
7 4.516 129 198 13528 3.7 0.01133 0.99999 0.73556 16.56508
miR-375 GO:0008610 lipid
biosynthetic process
DGAT1, SGMS2, B3GNT5,
LPGAT1, SEMA6D, PRKAG2,
PTGS1, SGMS1, CDS1
9 9.278 79 323 13528 4.8 5.31E-04 0.37365 0.37365 0.82287
miR-375 GO:0007423 sensory organ
development
MAF, DFNA5, MYO6, BCL11B,
ERBB2, PDGFRA, KLF4
7 7.216 79 229 13528 5.2 0.00205 0.83584 0.59483 3.14087
miR-375 GO:0008654 phospholipid
biosynthetic process
SGMS2, LPGAT1, SEMA6D,
SGMS1, CDS1
5 5.155 79 102 13528 8.4 0.00283 0.91776 0.56513 4.31620
miR-375 GO:0006631 fatty acid
metabolic process
ACOX1, PRKAG2, PTGS1,
BDH2, CPT1A, ACSL5
6 6.186 79 198 13528 5.2 0.00569 0.99338 0.71472 8.47985
miR-375 GO:0044242 cellular lipid
catabolic process
ACOX1, APOB, BDH2, CPT1A 4 4.124 79 76 13528 9.0 0.00959 0.99979 0.81650 13.90590
miR-375 GO:0006635 fatty acid
beta-oxidation
ACOX1, BDH2, CPT1A 3 3.093 79 28 13528 18.3 0.01126 0.99995 0.80998 16.13627
miR-375 GO:0046467 membrane
lipid biosynthetic process
SGMS2, B3GNT5, SGMS1 3 3.093 79 34 13528 15.1 0.01634 1.00000 0.87399 22.59228
miR-375 GO:0006686 sphin-
gomyelin biosynthetic
process
SGMS2, SGMS1 2 2.062 79 3 13528 114.2 0.01720 1.00000 0.85168 23.63443
miR-375 GO:0009062 fatty acid
catabolic process
ACOX1, BDH2, CPT1A 3 3.093 79 36 13528 14.3 0.01822 1.00000 0.83430 24.85354
miR-375 GO:0019395 fatty acid
oxidation
ACOX1, BDH2, CPT1A 3 3.093 79 39 13528 13.2 0.02119 1.00000 0.84813 28.31366
439
E.3. Enriched gene ontology terms lymph node
metastases
The 26 genes in common that are associated with the significantly enriched gene ontology
terms identified in the bioinformatics analysis (DAVID) of the predicted gene targets of
both miR-1 and miR-143 are shown below (FDR: < 0.05). For the full DAVID results
see Supplementary Table 4 (Miller et al., 2016).
CUL3, ZFP91, G2E3, PAX7, CHST11, NQO1, ALX4, EGFR, ARHGEF7,
ACTN2, PRKCE, MAPK1, TRIM35, TNFRSF10D, NGFR, PLEKHG2, TUBB,
KRAS, BCL2, STAMBP, B4GALT1, CFLAR, IL2RA, IGF1, NRAS, BMP7
440
F. Permission for reprints
The permissions obtained for the reproduction of figures and tables that appear in this
thesis are shown in Table F.1. The permission to reproduce documentation appears in
Figures: F.1, F.2, F.3, F.4, F.5 and F.6.
The results presented in this thesis were published in several research papers, for further
details see section F.1.
441
Table F.1.: Permissions for reprintsIdentifier Type of Description Source Copyright holder Permission given to reproduce Documenta-
tionwork
Table 2.1 table Ki-67 index Virchows Archiv (Rindi et al.,2006)
©2006, Springer-Verlag Creative Commons AttributionNonCommercial (CC BY-NC)
Figure: F.1
Virchows Archiv (Rindi et al.,2007)
©2007, Springer-Verlag Creative Commons AttributionNonCommercial (CC BY-NC)
Figure: F.2
Table 2.2 table staging SBNET Virchows Archiv (Rindi et al.,2007)
©2007, Springer-Verlag Creative Commons AttributionNonCommercial (CC BY-NC)
Figure: F.2
Table 2.4 table staging PNET Virchows Archiv (Rindi et al.,2006)
©2006, Springer-Verlag Creative Commons AttributionNonCommercial (CC BY-NC)
Figure: F.1
Figure 2.2 figure miRNAbiogenesis
Journal of Biomedical Science(Bak and Mikkelsen, 2010)
©2010, Bak and Mikkelsen;licensee BioMed Central Ltd. 2010
Creative Commons Attribution 2.0 Generic(CC BY 2.0)
Figure: F.3
Figure 2.3 figure miRNA cancer Nature Reviews Cancer(Esquela-Kerscher and Slack, 2006)
©2006, Macmillan Publishers Ltd yes Figure: F.4
Figure 2.4 figure sensitivity/speci-ficity
Rmostell (Rmostell, 2011a) ©2011, Rmostell Creative Commons Universal PublicDomain (CC0 1.0)
Figure: F.5
Rmostell (Rmostell, 2011b) ©2011, Rmostell Creative Commons Universal PublicDomain (CC0 1.0)
Figure: F.6
442
F.1. Published papers
The Ki-67 % results presented in this thesis were published in The World Journal of
Surgery in 2014 under the title “Role of Ki-67 proliferation index in the assessment of
patients with neuroendocrine neoplasias regarding the stage of disease.” (Miller et al.,
2014). For permissions see Figure F.7.
The miRNA results presented in this thesis were published in the journal Endocrine-
Related Cancer in 2016 under the title “MicroRNAs associated with small bowel neu-
roendocrine tumours and their metastases.” (Miller et al., 2016). For permissions see
Figure F.8.
Further publications were “Glucagon receptor gene mutations with hyperglucagonemia
but without the glucagonoma syndrome” published in The World Journal of Gastroin-
testinal Surgery (Miller et al., 2015a) and “Molecular genetic findings in small bowel
neuroendocrine neoplasms: a review of the literature” published in The International
Journal of Endocrine Oncology (Miller et al., 2015b).
443
Figure F.1.: CC BY-NC (Creative Commons Attribution NonCommercial) for Table 2.1and Table 2.4, for details see Table F.1
444
Figure F.2.: CC BY-NC (Creative Commons Attribution NonCommercial) for Table 2.1and Table 2.2, for details see Table F.1
Figure F.3.: CC BY 2.0 (Creative Commons Attribution 2.0 Generic) for Table 2.2, fordetails see Table F.1
445
Figure F.4.: Permissions for Figure 2.3, for details see Table F.1
446
Figure F.5.: Permissions for Figure 2.4, for details see Table F.1
Figure F.6.: Permissions for Figure 2.4, for details see Table F.1
447
Figure F.7.: Reprint permission from Springer Nature, [World Journal of Surgery], (Milleret al., 2014).
448
Figure F.8.: Reprint permission from BioScientifica Ltd., [Endocrine-Related Cancer],(Miller et al., 2016).
449