208
UNIVERSITE LIBRE DE BRUXELLES Faculty of Medicine Laboratory of Cancer Epigenetics Mining breast cancers with novel epigenetic modifications Evelyne COLLIGNON A thesis submitted for the degree of Doctor of Philosophy in Biomedical and Pharmaceutical Sciences Ph.D Thesis Director: Dr. François Fuks 2017

Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

  • Upload
    others

  • View
    2

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

UNIVERSITE LIBRE DE BRUXELLES

Faculty of Medicine

Laboratory of Cancer Epigenetics

Mining breast cancers with

novel epigenetic modifications

Evelyne COLLIGNON

A thesis submitted for the degree of Doctor of Philosophy

in Biomedical and Pharmaceutical Sciences

Ph.D Thesis Director: Dr. François Fuks

2017

Page 2: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

2

Page 3: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

3

“Nothing in life is to be feared, it is only to be understood. Now is the time to understand

more, so that we may fear less.”

Marie Curie

Page 4: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

4

Page 5: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

5

Résumé

Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications

des histones ont longtemps focalisé l’attention de la recherche biologique et médicale.

Toutefois, notre vision de l’épigénétique s’est fortement élargie avec la découverte de

« nouvelles » modifications épigénétiques de l’ADN et de l’ARN. Dès lors, au cours de

cette thèse, nous avons voulu explorer trois de ces modifications, et leurs enzymes, dans

le cadre des cancers du sein.

Premièrement, nous avons étudié l’enzyme TET1, responsable de

l’hydroxyméthylation de l’ADN (5hmC). Nous avons découvert que le niveau

d’expression de TET1 dans les tumeurs mammaires basal-like corrélait avec les

changements de 5hmC par rapport au tissu sain. Nous avons aussi établi un lien inédit

entre la répression de TET1 et l’infiltration immune dans ces cancers. Nous avons ensuite

démontré que cette répression était liée à l’activation canonique de NF-κB, un régulateur

majeur de l’immunité et l’inflammation. Enfin, nous avons étendu ce nouveau mode de

régulation de TET1 par l’immunité à d’autres types de cancers, y compris le mélanome,

le cancer du poumon et le cancer de la thyroïde.

Dans la deuxième partie de cette thèse, nous avons effectué la première étude

transcriptomique d’une nouvelle modification de l’ARN, l’hydroxyméthylation de l’ARN

(5hmrC). Chez la drosophile, nous avons révélé la distribution de cette marque le long du

transcriptome, associé une fonction régulatrice de la traduction protéique à la marque et

révélé le rôle central de 5hmrC et dTet, l’enzyme responsable de sa formation, dans le

développement du système nerveux central. A la suite de cette étude pionnière, nous

avons investigué 5hmrC en lignées mammaires et nous avons découvert plus de 700

ARNs différentiellement hydroxyméthylés dans les cellules cancéreuses par rapport aux

cellules mammaires normales. Globalement, nos résultats indiquent que 5hmrC constitue

un nouveau niveau de dérégulation de la fonction des gènes dans les cancers du sein.

Dans la dernière partie, nous avons examiné le rôle de la méthylation de l’ARN

(m6A) dans les cancers mammaires. Nous avons cartographié la distribution de m6A en

lignées et en tissus humains et identifié près de 2000 ARNs différentiellement méthylés

dans les cellules cancéreuses. Ensuite, nous avons découvert qu’une déméthylase de

l’ARN, FTO, était sous-exprimée dans les cancers du sein, et nous avons démontré que

cette dérégulation s’accompagnait d’une augmentation globale de m6A et d’un mauvais

pronostic de survie chez les patients. In vitro, la sous-expression de FTO cause un

phénotype plus agressif en termes de migration, invasion et caractère souche des cellules

cancéreuses mammaires. Ainsi, nos résultats semblent assigner une fonction suppressive

de tumeur à FTO dans la glande mammaire.

En conclusion, nos résultats démontrent l’importance biologique des

« nouvelles » modifications de l’ADN et l’ARN et de leur dérégulation dans le

développement des tumeurs mammaires. Nos données mettent au jour de nouveaux

mécanismes par lesquels l’épigénétique contribue au processus de cancérogénèse et

indiquent que l’étude de ces modifications pourrait avoir une utilité clinique.

Page 6: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

6

Page 7: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

7

Acknowledgements

First, I would like to thank my promoter, François Fuks, for giving me the

opportunity to pursue my PhD thesis in his team and work on exciting new themes of

research. Our many discussions, and sometimes disagreements, help me evolve

scientifically and develop a critical spirit. You also encouraged me and supported me in

my choices for the future, which I appreciated very much.

I would like also to thank all the members of the Laboratory of Cancer

Epigenetics, past and present, who shared this epic journey with me. First of all, Sarah,

who guided my first steps in the lab. Then, of course, Emilie, my “lab mom”, who was

always so helpful throughout the years (and who has the patience of a saint). Clémence,

my padawan, whom I could share my geekiness with (but not coffee). Jana, for all our

brainstorming and discussions on Schrödinger's Cat, among many topics. Martin, our dear

bioinformatician, who taught me some basic notions of R despite myself. Eric, our

resident star singer (and honorary member of the geek team). Pascale, whose knowledge

of good old classics awed us all. Princess Bouchra, who always stayed calm during the

storm. Iolanda, who listened to all of us and always supported us like a kindergarten

teacher. Rachel, who knew how to cheer us up with good music and cocktails. And all

the other members of the lab whom I had the chance to meet and share a good laugh with:

Elise, Laurence, Olivier, Christelle, Nick, Benjamin, Micha, Gordana, Jie, Matthieu,

Romy, Nitesh, Audrey, Thibaud, Andrea and Valentina.

I wish to thank all of our collaborators from the ULB and ULg, with a special

thought for Annalisa who worked by my side for four years, as well as the organisms

supporting our research: the FNRS, the l’Oréal Foundation, the Télévie, the IUAP and

the Walloon Region.

And finally, my family and my friends, who were ever present and supporting in

so many ways. Thank you.

Page 8: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

8

Page 9: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

9

Table of contents

Abbreviations ......................................................................................................... 13

List of figures .......................................................................................................... 19

General introduction .............................................................................................. 21

1. Epigenetics ...................................................................................................... 23

1.1 Introduction to epigenetics ................................................................................. 23

1.2 Chromatin .......................................................................................................... 26

1.3 Epigenetic modifications and their machineries ................................................... 30 1.3.1 Covalent DNA modifications ................................................................................................. 30

1.3.1.1 Methylation of cytosine ............................................................................................... 31 1.3.1.1.1 5mC distribution ...................................................................................................... 31 1.3.1.1.2 The DNA methyltransferases ................................................................................... 33 1.3.1.1.3 Gene silencing mediated by DNA methylation ........................................................ 34 1.3.1.1.4 Biological and pathological relevance of 5mC ......................................................... 35 1.3.1.1.5 Mapping the methylome ......................................................................................... 38 1.3.1.2 Hydroxymethylation of cytosine .................................................................................. 39 1.3.1.2.1 5hmC distribution .................................................................................................... 39 1.3.1.2.2 The TET enzymes ..................................................................................................... 40 1.3.1.2.3 Active demethylation and DNA hydroxymethylation .............................................. 41 1.3.1.2.4 Biological and pathological relevance of 5hmC and TETs........................................ 43 1.3.1.2.5 Mapping the hydroxymethylome ............................................................................ 46 1.3.1.3 Other covalent DNA modifications .............................................................................. 47 1.3.1.3.1 Oxidative derivatives of 5hmC ................................................................................. 47 1.3.1.3.2 Methylation of adenine ........................................................................................... 48

1.3.2 Histone modifications ........................................................................................................... 48 1.3.2.1 Histone acetylation ...................................................................................................... 49 1.3.2.2 Histone methylation .................................................................................................... 50 1.3.2.3 Other histone modifications ........................................................................................ 52

1.3.3 Epigenetic regulation of the chromatin ................................................................................ 53 1.3.3.1 Crosstalk between epigenetic modifications ............................................................... 53 1.3.3.2 Chromatin remodeling ................................................................................................. 55

1.3.4 RNA modifications ................................................................................................................ 58 1.3.4.1 Methylation of adenosine ............................................................................................ 59 1.3.4.1.1 m6A distribution ...................................................................................................... 59 1.3.4.1.2 The m6A machinery ................................................................................................. 60 1.3.4.1.3 Biological relevance of m6A .................................................................................... 61 1.3.4.2 Methylation of cytosine ............................................................................................... 63 1.3.4.2.1 5mrC in non-coding RNAs ........................................................................................ 64 1.3.4.2.2 5mrC in mRNAs ........................................................................................................ 64 1.3.4.2.3 Oxidation of 5mrC into 5hmrC ................................................................................ 65 1.3.4.3 Mapping RNA modifications ........................................................................................ 65

2. Breast cancers ................................................................................................. 67

2.1 Introduction ....................................................................................................... 67

Page 10: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

10

2.2 Diversity of breast cancers .................................................................................. 69 2.2.1 Histopathological classifications ........................................................................................... 69 2.2.2 Molecular classifications ....................................................................................................... 70

2.3 The immune system: a double-edged sword ........................................................ 72 2.3.1 Immune infiltration and its clinical relevance ...................................................................... 73 2.3.2 The NF-κB signaling pathway ................................................................................................ 74

2.4 Management and treatments .............................................................................. 76 2.4.1 Breast cancer diagnosis ........................................................................................................ 76 2.4.2 General therapeutic options................................................................................................. 76 2.4.3 Systemic therapies ................................................................................................................ 77

2.5 Breast cancer and epigenetics ............................................................................. 79 2.5.1 Epigenetics alterations in breast cancers ............................................................................. 79

2.5.1.1 Alterations of DNA modifications in BC ....................................................................... 79 2.5.1.2 Alterations of histone modifications in BC .................................................................. 82 2.5.1.3 Alterations of chromatin remodeling in BC ................................................................. 83 2.5.1.4 Alterations of RNA modifications in BC ....................................................................... 83

2.5.2 Clinical relevance of epigenetics........................................................................................... 84 2.5.2.1 Epigenetics modifications as biomarkers for cancer ................................................... 84 2.5.2.2 Epigenetic therapy of cancer ....................................................................................... 87

3. Aims of the project .......................................................................................... 93

Results ................................................................................................................... 95

1. Immune activation of NF-κB drives TET1 dysregulation in cancer ...................... 97

1.1. Introduction ............................................................................................................ 97

1.2. Results .................................................................................................................... 99 1.2.1. TET1 expression is associated with 5hmC dysregulation in BLBC .................................... 99 1.2.2. Link between TET1 expression and immunity in BLBC ................................................... 102 1.2.3. Activation of NF-κB drives TET1 repression ................................................................... 104 1.2.4. TET1 is repressed through binding of NF-κB to its promoter ......................................... 108 1.2.5. TET1 is downregulated by NF-κB in other cancer types ................................................. 110

1.3. Key findings .......................................................................................................... 113

2. RNA hydroxymethylation: a new player in the game ...................................... 115

2.1. Introduction .......................................................................................................... 115

2.2. Results .................................................................................................................. 117 2.2.1. RNA hydroxymethylation by dTet in Drosophila S2 cells ............................................... 117 2.2.2. Transcriptome-wide mapping of 5hmrC in S2 cells ........................................................ 119 2.2.3. Hydroxymethylation can favor mRNA translation ......................................................... 121 2.2.4. In vivo relevance of 5hmrC in Drosophila ...................................................................... 123 2.2.5. Alterations of 5hmrC in breast cancer ........................................................................... 125

2.3. Key findings .......................................................................................................... 129

3. Dysregulations of m6A and its machinery support breast cancer .................... 131

3.1. Introduction .......................................................................................................... 131

3.2. Results .................................................................................................................. 133 3.2.1. Transcriptome-wide m6A landscape in breast cancer ................................................... 133

Page 11: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

11

3.2.2. Deregulation of the m6A machinery in BC ..................................................................... 137 3.2.3. Phenotypic effect of FTO in BC ....................................................................................... 139 3.2.4. FTO regulates the Wnt/β-catenin pathway in BC .......................................................... 141

3.3. Key findings .......................................................................................................... 146

Discussion ............................................................................................................ 149

1. Immune activation of NF-κB drives TET1 dysregulation in cancer .................... 151

2. RNA hydroxymethylation: a new player in the game ...................................... 159

3. Dysregulations of m6A and its machinery support breast cancer .................... 165

4. Concluding remarks ....................................................................................... 171

References ........................................................................................................... 175

Appendix .............................................................................................................. 205

Page 12: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

12

Page 13: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

13

Abbreviations

2-OG 2-Oxoglutarate

5-Aza-CdR 5-Aza-2'-deoxycytidine

5-Aza-CR 5-Azacytidine

5caC DNA 5-carboxycytosine

5fC DNA 5-formylcytosine

5hmC DNA 5-hydroxymethylcytosine

5hmrC RNA 5-hydroxymethylcytosine

5hmU DNA 5-hydroxymethyluracile

5mC DNA 5-methylcytosine

5mrC RNA 5-methylcytosine

A Adenine

Ac Acetylation

ADAM19 A Disintegrin And Metalloproteinase Domain 19

ADP Adenosine Diphosphate

AID Activation-Induced Deaminase

ALKBH Alkylated DNA Repair Protein AlkB Homolog

APC Adenomatosis Polyposis Coli

APOBEC Apolipoprotein B mRNA editing enzyme catalytic polypeptide-like

ASB2 Ankyrin Repeat And SOCS Box Containing 2

ATP Adenosine Triphosphate

B2M Beta-2-Microglobulin

BC Breast Cancer

Bdnf Brain-derived neurotrophic factor (mouse)

BER Base Excision Repair

BLBC Basal-like Breast Cancer

BRCA1-2 Breast Cancer 1-2

BS Bisulfite

BS-Seq Bisulfite Sequencing

BTRC Beta-Transducin Repeat Containing E3 Ubiquitin Protein

C Cytosine

CBP Cyclic AMP response element-binding protein

CCL2 C-C Motif Chemokine Ligand 2

CD Cluster of Differentiation

CDO1 Cysteine Dioxygenase Type

CGI CpG island

ChIP-Seq Chromatin Immunoprecipitation Sequencing

CIBERSORT Cell type Identification By Estimating Relative Subsets Of known RNA

Transcripts

CLIP UV Cross-Linking and Immunoprecipitation

CM Conditioned Media

CMS Cytosine 5-Methylenesulphonate

COMPASS Complex Proteins Associated with Set1

CpG Cytosine-phosphate-Guanosine

CPM Counts Per Million

CTCF CCCTC-Binding Factor

CTL Control

CTLA-4 Cytotoxic T-Lymphocyte Associated Protein 4

Dam DNA adenine methylase

DAPK1 Death Associated Protein Kinase 1

Page 14: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

14

DAVID Database for Annotation, Visualization and Integrated Discovery

dhmR differentially hydroxymethylated regions

DMFS Distant Metastasis Free Survival

DNA Deoxyribonucleic Acid

DNMT DNA Methyl Transferase

DOX Doxocycline

DSBH Double-Stranded β-Helix

dTet drosophila Tet

eIF3 Eukaryotic Translation Initiation Factor 3

EMT Epithelial to Mesenchymal Transition

ER Estrogen Receptor

Erbb2 Erb-B2 Receptor Tyrosine Kinase 2

ES Embryonic Stem

ESMO European Society for Medical Oncology

EZH2 Enhancer Of Zeste 2 Polycomb Repressive Complex 2 Subunit

FBS Fetal Bovine Serum

FC Fold-Change

FDA Food and Drug Administration

FDR False Discovery Rate

Fgf1 Fibroblast Growth Factor 1

FK-228 Romidepsin

F-LUC Firefly Luciferase

FN1 Fibronectin 1

FOXM1 Forkhead Box M1

FTO Fat Mass And Obesity Associated

FZD Frizzled homolog Drosophila

G Guanine

GADD45 Growth Arrest And DNA Damage Inducible Alpha

GEO Gene Expression Omnibus

GFP Green Fluorescent Protein

GGI Genomic Grade Index

GlcNAcylation O-linked N-acetylglucosaminylation

GNAT GCN5-related N-acetyltransferases

GSK3 Glycogen Synthase Kinase 3

GSTP1 Glutathione S-Transferase Pi 1

GWAS Genome-Wide Association Studies

H1 Histone 1

H2A Histone 2A

H2B Histone 2B

H3 Histone 3

H3K27ac Acetylation of Histone 3 lysine 27

H3K27me3 Trimethylation of Histone 3 lysine 27

H3K4ac Acetylation of Histone 3 lysine 4

H3K4me3 Trimethylation of Histone 3 lysine 4

H3K9ac Acetylation of Histone 3 lysine 9

H3K9me3 Trimethylation of Histone 3 lysine 9

H4 Histone 4

H4K27ac Acetylation of Histone 4 lysine 27

HCF1 Host Cell Factor C1

HDAC Histone Deacetylase

HDM Histone Demethylases

HER2 Human Epidermal Growth factor Receptor 2

hMeDIP Hydroxymethylated DNA Immunoprecipitation

hMe-Seal 5hmC-selective chemical labeling

HMT Histone Methyl Transferase

Page 15: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

15

HNRNP Heterogeneous Nuclear Ribonucleoprotein

HOTAIR HOX Transcript Antisense RNA

HOX Homeobox

HP1 Heterochromatin Protein 1

HuR Hu Antigen R

ICM Inner Cell Mass

ICR1 Imprint Control Region 1

IDAX Inhibition Of The Dvl And Axin Complex

IDC-NOS Invasive Ductal Carcinoma Not Otherwise Specified

IDC-NST Invasive Ductal Carcinoma of No Special Type

IGF1R Insulin Like Growth Factor 1 Receptor

IGF2 Insulin Like Growth Factor 2

IgG Immunoglobulin G

IHC Immunohistochemistry

IKK IkB kinase

IL Interleukin

ILC Invasive Lobular Carcinoma

IP Immunoprecipitation

IPA Ingenuity Pathway Analysis

IκB Inhibitor of kappa B

JBP J-Binding Protein

LBH-589 Panobinostat

LINE-1 LINE retrotransposable element 1

LncRNAs Long non-coding RNA

LPS Lipopolysaccharide

LRP LDL Receptor Related Protein

LSD1 Lysine (K)-Specific Demethylase 1

LST1 Leukocyte Specific Transcript 1

LUAD LUng Adenocarcinoma

m6A N6-methyladenosine

MAGE Melanoma Antigen Family

MAPK Mitogen-Activated Protein Kinase

MASPIN Mammary Serine Protease Inhibitor

MBD Methyl-CpG-Binding domain

MDSCs Myeloid Derived Suppressor Cells

Me Methylation

MeCP2 Methyl-CpG Binding Protein 2

MeDIP-Seq Methylated DNA Immunoprecipitation Sequencing

MethylCap-Seq Methylated DNA Capture and Sequencing

MeTIL Methylation of TIL

METTL Methyltransferase Like

MGMT O-6-Methylguanine-DNA Methyltransferase

MHC Major Histocompatibility Complex

miCLIP m6A individual-nucleotide resolution using CLIP

miRNA microRNA

MLH1 MutL Homolog 1

MLL Mixed Lineage Leukemia

MMTV-PyMT Mouse Mammary Tumor Virus Promoter-polyoma Middle T-antigen

MRI Magnetic Resonance Imaging

mRNA Messenger RNA

MYST Moz, Ybf2/Sas3, Sas2, Tip60

NANOG Nanog homeobox

NCBI National Center for Biotechnology Information

NF-κB Nuclear factor kappa B

NK Natural Killer

Page 16: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

16

NSUN2 NOP2/Sun RNA Methyltransferase Family Member 2

OCT4 Octamer-binding Transcription factor 4

OGA O-GlcNAcase

O-GlcNAc O-linked β-N-acetylglucosamine

OGT O-GlcNAc transferase

OxBS-Seq Oxidative Bisulfite Sequencing

P Phosphorylation

p16 cyclin-dependent kinase inhibitor 2A

p38 P38 Mitogen Activated Protein Kinase

p50 Nuclear Factor Kappa B Subunit 1

p52 Nuclear Factor Kappa B Subunit 2

p65 RELA Proto-Oncogene, NF-KB Subunit

PAM50 Prediction Analysis of Microarray 50

PARP Poly(ADP-Ribose) Polymerase 1

PCDHGB7 Protocadherin Gamma Subfamily B, 7

PD-1/PD-L1 Programmed Cell Death 1 Ligand 1

PGC Primordial Germ Cells

pHEMA Poly(2-hydroxyethyl methacrylate)

PI3K Phosphatidylinositol-4,5-Bisphosphate 3-Kinase

PIK3CA Phosphatidylinositol-4,5-Bisphosphate 3-Kinase Catalytic Subunit Alpha

PITX2 Paired Like Homeodomain 2

Poly-A Polyadenylation

PPP2CB Protein Phosphatase 2 Catalytic Subunit Beta

PR Progesterone Receptor

PRMT Protein Arginine Methyltransferase

PTEN Phosphatase And Tensin Homolog

PXD-101 Belinostat

RAD51 RecA-Like Protein

RARA Retinoic Acid Receptor Alpha

RAR-b Retinoic Acid Receptor beta

RAS3 Ras-related protein R-Ras3

RASSF1A Ras Association Domain Family Member 1

Rb Retinoblastoma

Rel REL proto-oncogene

RFS Recurrence-Free Survival

RIP RNA Immunoprecipitaion

R-LUC Renilla luciferase

RNA Ribonucleic Acid

RNA-seq RNA sequencing

RPKM Reads Per Kilobase per Million mapped reads

R-Ras Related RAS Viral (R-Ras) Oncogene Homolog

rRNA Ribosomial RNA

RSEM RNA-Seq by Expectation Maximization

RTqPCR Reverse transcription polymerase chain reaction

RUNX1T1 RUNX1 Translocation Partner 1

SAHA Vorinostat

SAM S-adenosylmethionine

SATR-1 Satellite-like Repeat 1

SD Standard Deviation

SET1 SET Domain Containing 1A

SFN Stratifin

shRNA Short hairpin RNA

Sin3a SIN3 homolog A

siRNA Small Interfering RNA

SKCM SKin Cutaneous Melanoma (TCGA cohort)

Page 17: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

17

SMUG1 Single-Strand-Selective Monofunctional Uracil-DNA Glycosylase 1

snRNA small nuclear RNA

SOX Sex Determining Region Y-Box

SRSF2 Serine And Arginine Rich Splicing Factor 2

T Thymine

TAB-Seq Tet-Assisted Bisulfite Sequencing

TCGA The Cancer Genome Atlas

TDG Thymine DNA Glycosylase

TE Trophectoderm

TET Ten-Eleven Translocation

TF Transcription Factor

TFBIND Transcription Factor BINDing site

TH T Helper

THCA THyroid CArcinoma (TCGA cohort)

TIL Tumor-Infiltrating Lymphocyte

TIMP TIMP Metallopeptidase Inhibitor

TME Tumor Microenvironment

TN Triple Negative

TNF Tumor Necrosis Factor

TNM Tumor Node Metastasis

TP53 Tumor Protein P53

tRNA Transfer RNA

TSG Tumor Suppressor Gene

TSS Transcription Start Site

TYROBP TYRO Protein Tyrosine Kinase Binding Protein

Ub Ubiquitination

UCSC University of California Santa Cruz

UHRF1 Ubiquitin Like With PHD And Ring Finger Domains 1

UTR Untranslated regions

UV Ultraviolet

WHO World Health Organization

WNT Wingless-Type MMTV Integration Site Family

WTAP WT1 Associated Protein

XIST X Inactive Specific Transcript

YTHDC YTH Domain Containing

YTHDF YTH Domain Family

ZO1 Zona Occludens 1

Page 18: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

18

Page 19: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

19

List of figures

Fig. 1: Genome versus epigenome.

Fig. 2: The nucleosome.

Fig. 3: Chromatin organization and packing.

Fig. 4: Euchromatin and heterochromatin.

Fig. 5: Epigenetic modifications regulate the flow of genetic information.

Fig. 6: The methylome.

Fig. 7: DNMT-mediated DNA methylation.

Fig. 8: De novo and maintenance methylation.

Fig. 9: Transcription silencing by DNA methylation.

Fig. 10: Dynamicity of 5mC during development.

Fig. 11: Aberrant DNA methylation in cancer.

Fig. 12: 5hmC distribution in the cerebellum.

Fig. 13: Activity and structure of human TET proteins.

Fig. 14: Pathways of cytosine demethylation mediated by TET proteins.

Fig. 15: View of DNA demethylation in the zygote.

Fig. 16: 5hmC changes in cancer.

Fig. 17: The main modifications of the four core histones.

Fig. 18: Histone acetylation.

Fig. 19: Major methylation events of histones H3 and H4.

Fig. 20: Crosstalk between DNA and histone modifications.

Fig. 21: Cooperative regulation and gene silencing.

Fig. 22: Connecting TETs and OGT.

Fig. 23: SWI/SNF complexes.

Fig. 24: Chemical modifications in eukaryotic mRNA.

Fig. 25: Metagene profiles of m6A.

Fig. 26: The writer, eraser and reader proteins of m6A.

Fig. 27: Mechanisms and functions of m6A.

Fig. 28: Functions of 5mrC.

Fig. 29: Ten leading cancer types.

Fig. 30: Anatomy of the breast.

Fig. 31: TNM stage is a predictor of overall survival.

Fig. 32: Expression subtypes are associated with clinical features.

Fig. 33: Balance of the immune TME.

Fig. 34: Canonical NF-κB pathway.

Fig. 35: Standard treatments for BC.

Fig. 36: DNA methylation changes are associated with ER status in BC.

Fig. 37: Loss of 5hmC in BC, measured by IHC.

Fig. 38: TET1 act as a TSG in BC.

Fig. 39: Example of epigenetic prognostic marker.

Fig. 40: DNA methylation as a tool to quantify immune infiltration.

Fig. 41: Activation of TSG by epigenetic drugs.

Fig. 42: Epigenetic drugs in cancer therapy.

Fig. 43: TET1 expression in BC subtypes.

Fig. 44: TET1 regulation is associated with distinct 5hmC changes in BLBC.

Fig. 45: Link between 5hmC, 5mC and gene expression in BLBC.

Page 20: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

20

Fig. 46: TET1 is anticorrelated with genes linked to immune pathways.

Fig. 47: High TET1 expression discriminates BLBC tumors with low immune infiltration.

Fig. 48: Leukocyte-conditioned media represses TET1 expression.

Fig. 49: TET1 expression and NF-κB in breast cancer.

Fig. 50: TET1 expression is repressed by NF-κB activation in vitro.

Fig. 51: TET1 expression is repressed by NF-κB activation in vivo.

Fig. 52: Schematic view of TET1 gene promoter.

Fig. 53: Binding of p65 to TET1 promoter.

Fig. 54: TET1 expression and immune markers in cancer.

Fig. 55: TET1 expression and NF-κB signature.

Fig. 56: NF-κB represses TET1 expression in cancer.

Fig. 57: Validation of 5hmrC-detecting dot blot method.

Fig. 58: 5hmrC is enriched in poly A RNA in S2 Drosophila cells.

Fig. 59: dTet knockdown leads to reduced 5hmrC levels.

Fig. 60: Transcriptome-wide distribution of 5hmrC in Drosophila cells.

Fig. 61: dTet mediates transcriptome-wide RNA hydroxymethylation in Drosophila.

Fig. 62: 5hmrC and gene expression upon depletion of dTet in S2 cells.

Fig. 63: Highly translated mRNAs display high levels of 5hmrC.

Fig. 64: 5hmrC favors mRNA translation.

Fig. 65: Levels of dTet and 5hmrC in early embryogenesis.

Fig. 66: dTet and 5hmrC levels in the brain.

Fig. 67: dTet-deficient fruit flies show impaired brain development, accompanied by

decreased 5hmrC.

Fig. 68: RNA methylation and hydroxymethylation in mammary cells.

Fig. 69: Distribution of 5hmrC peaks in breast cells.

Fig. 70: Changes in 5hmrC in BC.

Fig. 71: Examples of 5hmrC tracks in breast cells.

Fig. 72: Transcriptome-wide mapping of m6A in SKBR3 cells.

Fig. 73: m6A changes in cultured BC cells.

Fig. 74: Transcriptome-wide mapping of m6A in human BC biopsies.

Fig. 75: Expression of m6A enzymes in BC.

Fig. 76: Quantification of m6A in BC by mass spectrometry.

Fig. 77: FTO expression and survival in BC.

Fig. 78: FTO-knockdown in BC.

Fig. 79: FTO depletion enhances the mobility of cancer cells.

Fig. 80: FTO depletion enhances in vitro tumorsphere formation.

Fig. 81: Canonical Wnt pathway.

Fig. 82: FTO regulates β-catenin in BC.

Fig. 83: Loss of FTO enhances β-catenin signaling.

Fig. 84: Loss of FTO affects response to Wnt inhibitor in BC.

Fig. 85: Proposed model illustrating the immune regulation of TET1 in cancer.

Fig. 86: FTO, breast cancer and adipocytes.

Fig. 87: Proposed model illustrating effects of FTO depletion in BC.

Fig. 88: The molecular portrait of tumors is a multidimensional picture.

Page 21: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

21

General introduction

General introduction

Page 22: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

22

Page 23: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

23

1. Epigenetics

1.1 Introduction to epigenetics

In 2001, the unravelling of the human genome generated considerable excitement

in the scientific community and provided an essential basis for the study of the inheritance

of phenotypical features, as described by Mendel over a century earlier (Lander et al.

2001). The genetic information, coded in our DNA, is written with 4 “letters” (A, T, C,

G) and inherited from our parents. The genes guide our development from a single cell to

a fully formed human being and outline our identity as a species. However, many

questions remain and research in the past decades indicates that we are “more than the

sums of our genes”. A striking illustration is the vast diversity in cell types within the

human body, from neurons to leukocytes or hepatocytes. All those cells share the same

genomic DNA, yet they display very different shapes and functions (Wade 2009).

Another example is the discrepancies in phenotypes observed between monozygotic

twins (Silva et al. 2011). To address those questions, one must look beyond the single

lecture of the DNA sequence to understand the fine regulation of gene expression.

The word “epigenetics”, which comes from the prefix epi- (in Greek επί: over,

above) and genetics, was broadly used to characterize events that could not be explained

by genetics alone, hence the need to go “beyond” genetics. It was defined in 1942 by

Conrad Waddington as “the branch of biology which studies the causal interactions

between genes and their products, which bring the phenotype into being” (Waddington

1942). Thus, in the original sense of the definition, any mechanism modulating the

apparition of a phenotype from a certain genotype could be considered as “epigenetic”.

However, the definition of epigenetics evolved over time, along with our knowledge of

genetics and gene regulation. It also narrowed over the years, focusing on “the study of

mitotically and/or meiotically heritable changes in gene function that cannot be explained

by changes in DNA sequence” (Russo et al. 1996; Riggs & Porter 1996; Bird 2007). The

field now comprises a range of biological mechanisms, including, among others, genomic

imprinting, X chromosome inactivation and retrotransposon regulation. Nevertheless, the

Page 24: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

24

exact definition of epigenetics remains open to debate. In particular, the heritability trait

has yet to be clearly demonstrated for many events generally accepted as “epigenetic”

(Deans et al. 2015).

The field of epigenetics has attracted growing attention over the past decades, as

our understanding of its molecular basis has increased. The study of how epigenetic

mechanisms can influence the structure of the chromatin (the complex of DNA and its

intimately associated protein histones) and gene expression has improved our knowledge

of physiological development and homeostasis, as well as their deregulation in

pathologies (Berner et al. 2010; Portela & Esteller 2010). In the strict sense, two known

molecular mechanisms currently satisfy the definition of epigenetics: the first are the

covalent modifications of DNA, and the second are the post-translational modifications

of histones. Next to the classical epigenetic modifications, which focus on the regulation

of the genome and the chromatin, a third class of modifications has recently gained the

attention of the scientific community: the RNA modifications. A new field of study,

commonly referred to as “RNA epigenetics” or “epitranscriptomics”, has emerged as a

new regulatory level of gene function (Liu & Pan 2017).

Taken together, the epigenetic modifications that either activate or inactivate the

genes form a code that controls their expression. The epigenetic state of a given cell is

referred to as the “epigenome”, by analogy to the genome, for which the genetic

information is encoded in the DNA. A striking characteristic of epigenetics, which makes

its study so compelling, is its outstanding versatility: while all the cells of a given

organism essentially contain the same genome stocking the genetic information, many

distinct epigenomes are present within the same organism and organize this information

differently (Fig. 1) (Wade 2009). This versatility is both spatial and temporal. The

epigenomes vary vastly between cell types and tissues, and they evolve during the

development from a single-cell zygote to an adult, as well as during the aging events

occurring beyond (Lokk et al. 2014; Booth & Brunet 2016). It can also be influenced by

external factors, such as the nutritional status, thus offering a bridge between nature and

nurture (Boyce & Kobor 2015). In that regard, the study of epigenetics is both very

complex and powerful. Interestingly, the recent technical progresses in biomedical

sciences, particularly in the domain of micro-array and sequencing technologies, have

eased the study of the epigenomes by making them more accessible, both in terms of cost

and workload.

Page 25: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

25

Fig. 1: Genome versus epigenome.

All the cells of the organism contain

the same genetic information in their

DNA. In contrast, the epigenetic

information contained in the

chromatin varies from one cell type

to another.

Because of this versatility, epigenetics has recently been integrated in the study of

many biological fields. It is now taken into consideration in research of early

development, reproduction, immunity, neurology, and evolution, among others (Portela

& Esteller 2010; Mirabella et al. 2016). This research is not restricted to humans and

mammals; it encompasses all types of species, from bacteria and yeast to worms, fish or

plants (Willbanks et al. 2016). In addition, epigenetic mechanisms and their deregulations

are also studied in the context of most pathologies, including all cancers,

neurodegenerative disorders (e.g. Parkinson, Huntington, and Alzheimer diseases),

inflammation, auto-immune diseases and cardio-vascular diseases (Portela & Esteller

2010; Brookes & Shi 2014). Thus, the scope of epigenetics has been broadening over the

years and is now interwoven with many other disciplines in biology.

In summary, epigenetics offers a complex and flexible reading of a cell’s state and

its study crosses-over disciplines such as physiology, medicine and evolution. The

epigenetic mechanisms are of vital importance for the proper regulation of many

biological processes. In the following sections, we will develop the main notions of

epigenetics. We will first present the concept of the chromatin, then we will detail the

molecular bases of the major epigenetic mechanisms.

DNA Chromatin

Page 26: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

26

1.2 Chromatin

The diploid human genome consists of approximately 6 billion base pairs of DNA

which, in a loose state, represent a length of about 2 meters and contain about 20,000

genes (Ezkurdia et al. 2014). The genome is organized in 23 chromosome pairs that are

confined in the nucleus of the cells that measures about 10µm of diameter. Therefore, the

DNA must be carefully packed and organized in order to both fit in the nucleus and allow

the proper expression of the genes. This high compaction (of over 100,000-fold in

magnitude) is accomplished through the wrapping of the DNA around a core of histone

proteins – much like thread around a bobbin – and the nucleoprotein complex thus formed

is called the chromatin (Chakravarthy et al. 2005). This enfolding of the DNA is repeated

along the genome, each unit of this particular polymer being called a “nucleosome”. The

structure of the chromatin is extremely flexible, allowing various state of compaction and

modulating gene expression.

The nucleosome is the fundamental unit of the chromatin (Fig. 2). It is made of

approximatively 147bp of DNA wrapped around an octamer of proteins containing two

copies of the four core histones (H2A, H2B, H3 and H4) (Mirabella et al. 2016). The

octamer is formed by the association of two H2A-H2B dimers with one H3-H4 tetramer.

An additional histone protein, H1, binds to the “linker DNA” located between

nucleosomes. It is involved in the packing and compaction of the DNA. All histones

consist of a globular hydrophobic core and a basic N-terminal tail that projects from the

surface of the protein (Chakravarthy et al. 2005). This tail can be significantly altered by

a range of post-translational modifications. The role of these modifications in the

regulation of chromatin will be detailed in a further chapter (see 1.3.2 Histones

modifications).

Page 27: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

27

Fig. 2: The nucleosome. Schematic

representation of DNA wrapped

around the histone octamer core.

Histone tails protrude from the core

and are subjected to chemical

modifications. (Adapted from

neuropsychotherapist.com)

The succession of nucleosomes forms a fiber of about 10nm diameter, commonly

referred to as the “beads on a string” structure (Fig. 3). It is the first level of compaction

of the DNA (Szerlong & Hansen 2011). The chromatin can be further compacted through

interactions between the nucleosomes, due to the incorporation of the H1 histone, leading

to a 30nm fiber. Further compaction of the chromatin fiber eventually produces the

formation of the metaphase chromosome structure (Fig. 3) (Mirabella et al. 2016). This

level constitutes the highest degree of packing of the chromatin and is clearly visible in

microscopy. The visualization of chromosomes through karyotyping methods has clinical

implications, as it allows the detection of dire chromosomal defects in fetus and newborn

children, such as the trisomy of chromosome 21 which causes Down syndrome (Ho &

Crabtree 2010).

Fig. 3: Chromatin organization and packing. Double-stranded DNA wraps around

histone proteins to form nucleosomes in the “beads on a string” structure. Further

compaction of the chromatin leads to 30-nm and 300-nm fibers of chromatin, and then to

the chromosome structure observed in metaphase. (Adapted from

www.geneticliteracyproject.org/?p=337515)

Page 28: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

28

Chromatin is commonly classified in two categories, based on the degree of

compaction and the transcriptional activity: euchromatin and heterochromatin (Fig. 4).

The terms originally refer to differences in staining intensity observed in the nucleus,

which corresponded to distinct levels of compaction of the chromatin (Heitz 1928). This

difference is also visible with electron microscopy imaging of the nucleus:

heterochromatin appears as dark regions while euchromatin constitutes much brighter

areas (Wolf & Sumner 1996).

Heterochromatin corresponds to a highly-condensed state of chromatin, which

hinders gene transcription (Fig.4). It can be subdivided in two categories. On one hand,

the constitutive heterochromatin refers to chromatin regions that are always compact. It

comprises mainly gene-poor, noncoding and repetitive sequences. Of note, the presence

of constitutive heterochromatin in the telomeres and centromeres is important for genome

integrity (Almouzni & Probst 2011; Postepska-Igielska et al. 2013). On the other hand,

the facultative heterochromatin corresponds to regions that can alternate between a

closed, condensed state and a more open, transcriptionally active state (Trojer & Reinberg

2007). The regulation of such chromatin is notably linked to morphogenesis and cell

differentiation. A striking example of facultative heterochromatin is the X chromosome

in female, for which one copy is silenced by tight compaction of the chromatin while the

other remains open and allows gene transcription (Wutz 2011).

In contrast to heterochromatin, euchromatin refers to open, loose regions of

chromatin (Fig. 4). It is often called “active chromatin”, as it consists mainly of coding

regions of the genomes that are accessible to the transcription machinery (Tamaru 2010).

However, euchromatin is not always transcriptionally active. While some genes of

euchromatin are ubiquitously expressed (and thus commonly named “housekeeping

genes”), other genes require the involvement of additional transcription factors to

promote their expression.

Fig. 4: Euchromatin and

heterochromatin. Chromatin can

be divided in euchromatin (loosely

packed and accessible to the

transcription machinery) and

heterochromatin (highly condensed

and not transcribed). (Adapted from

Sha & Boyer 2008)

Page 29: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

29

However, this view must be nuanced, as heterochromatin and euchromatin

represent two extreme situations on a wide spectrum that rarely reflect the reality of

chromatin state. For instance, the telomeric regions are considered as heterochromatic, as

previously mentioned, and yet they display some level of transcriptional activity.

Telomeric non-coding RNAs notably appear to regulate heterochromatin formation and

their dysregulation can lead to cellular senescence (Arnoult et al. 2012; Maicher et al.

2012). Likewise, centromeric regions contain large domains of heterochromatin,

nevertheless active transcription occurs and is a critical element to maintaining

centromere function (Hall et al. 2012). Thus, heterochromatin regions are not always

transcriptionally inert and their RNAs fulfill important functions.

In conclusion, DNA is not “naked” in the nucleus, but is associated with proteins

in a complex structure called the chromatin. The regulation of chromatin is key to the

establishment of a proper gene expression profile for any given cell. Several mechanisms

are involved in this regulation, including epigenetic modifications that are described

below.

Page 30: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

30

1.3 Epigenetic modifications and their machineries

In this chapter, we explore the main epigenetic modifications and their regulation

(Fig. 5). First, we focus on the best-characterized modifications, namely DNA and histone

modifications. Then, we expand on an emerging theme in epigenetics, the field of RNA

modifications.

Fig. 5: Epigenetic modifications regulate the flow of genetic information. In the

central dogma of biology, genetic information is passed from DNA to RNA and then to

protein. Epigenetic DNA modifications and histone modifications are known to have

important roles in regulating gene transcription. RNA modifications add an additional

layer of dynamic regulation of biological processes. (Adapted from Y. Fu et al. 2014)

1.3.1 Covalent DNA modifications

The fine regulation of gene expression is essential to the establishment of various

cell types and the completion of their functions. One of the central mechanisms

underlying this regulation is the modification of the DNA through the addition of

chemical moieties. The most studied modification is, by far, the methylation of cytosine.

For decades, it was the only epigenetic modification of the DNA extensively investigated

Page 31: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

31

in biomedical sciences. However, in recent years, light was shed on additional DNA

modifications, such as the hydroxymethylation of cytosine, thus emphasizing a system

more complex and dynamic than first thought. In this chapter, we will describe the main

covalent modifications of the DNA, their regulation and physiological relevance. We will

also describe the technologies available to profile their distributions.

1.3.1.1 Methylation of cytosine

The methylation of cytosine, commonly called “DNA methylation”, involves the

enzymatic transfer of a methyl group to the 5’ carbon of a cytosine residue to form the 5-

methylcytosine (5mC). This modified cytosine is sometimes referred to as the “fifth base”

of DNA. This modification is present on the genome of animals, plants and prokaryotes.

It is involved in a broad range of biological mechanisms, including bacterial defense

against bacteriophages, X chromosome inactivation, repression of transposons (Bergman

& Cedar 2013; Deans et al. 2015).

1.3.1.1.1 5mC distribution

In humans, cytosine methylation occurs nearly exclusively in a “CpG” context,

i.e. a cytosine-phosphate-guanine dinucleotide, and it is often associated with

transcriptional repression (Ziller et al. 2011). Importantly, CpG sites are not distributed

evenly in the genome, instead they tend to cluster together in regions called “CpG islands”

(CGIs) (Fan & Zhang 2009). These regions are short sequences of DNA with an

exceptionally high density of CpGs that are often located in the 5’ region of genes.

Interestingly, Alu sequences, the most abundant of the repetitive elements of the human

genome, are also CpG-rich regions and contain approximately one-third of all human

CpG dinucleotides (Y. Luo et al. 2014).

In mammals, almost 70% of all CpG sites are methylated (Lister et al. 2009), and

the patterns of methylation are not random. The highly-methylated regions include

repeated sequences, centromeric regions, satellite DNA and transposons (Almouzni &

Probst 2011; Postepska-Igielska et al. 2013). This methylation is essential for the

repression of endoparasitic elements, which could lead to genome instability through

translocation and gene disruption. In contrast, the CGIs located in the promoter of genes

Page 32: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

32

are mostly unmethylated. Nevertheless, methylation of these CGIs can still be acquired

during development in a tissue-specific manner (Fan & Zhang 2009).

Traditionally, 5mC is considered in a gene promoter context and has long been

associated with gene silencing. However, this dogma has been extended in recent years

with the advances of technologies (Dedeurwaerder et al. 2013; Moran et al. 2015).

Genome-wide profiling has shown that 5mC is in fact far from restricted to promoter

regions. Methylation across the gene (often called intragenic or gene body methylation)

is now thought to be involved in transcriptional regulation and efficiency. And, strikingly,

whereas promoter methylation is a repressive mark, intragenic methylation tends to show

a positive correlation with gene expression (Kulis et al. 2013). Intragenic methylation

also appears to inhibit the transcription from alternative promoters, thus controlling

tissue-specific isoforms (Maunakea et al. 2010; Neri et al. 2017). Similarly, increasing

evidence suggests that methylation at intergenic regions can also regulate gene expression

(Shore et al. 2010; Moran et al. 2015). The noncoding regions contain regulatory

elements, including enhancers, silencers and noncoding RNAs, which can be affected by

5mC. Therefore, it is essential to consider DNA methylation across the entire genome

(also called “methylome”, see Fig. 6) rather than focusing solely on promoter regions

(Jeschke et al. 2015).

Fig. 6: The methylome. CGI promoters are usually protected from methylation and are

prone to active transcription. CpG-poor regions and repetitive elements are often

methylated. Enhancers, promoters, and intragenic regions can also be differentially

methylated in a cell type-specific fashion. (Adapted from Carrio & Suelves 2015)

Page 33: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

33

1.3.1.1.2 The DNA methyltransferases

DNA methylation is mediated by a family of proteins called the DNA

methyltransferases (DNMTs). These enzymes transfer a methyl group (-CH3) from S-

adenosylmethionine (SAM) to the 5’ carbon of a cytosine (Fig. 7). There are five

members of the DNMT family in humans: DNMT1, DNMT2, DNMT3A, DNMT3B and

DNM3L. However, only three of them – DNMT1, 3A and 3B – are active DNA

methyltransferases.

Fig. 7: DNMT-mediated DNA methylation. DNMTs catalyze the transfer of a methyl

group from S-adenosylmethionine (SAM) to the 5-carbon position of cytosine (Adapted

from Gibney & Nolan 2010).

DNMT1 is called the “maintenance methyltransferase” because it is responsible

for the methylation of the newly synthetized strand during DNA replication (Fig. 8). In a

methylated CpG, the cytosines on both strands of the DNA are normally methylated.

However, after the genome duplication in S phase, the neo-synthetized strand is initially

unmodified, while the remaining parent strand is still methylated, thus generating a

“hemi-methylated site”. DNMT1 displays a high affinity for such hemi-methylated CpGs

and is able to methylate the new strand. Hence, it ensures the maintenance of the

methylation patterns following the genome duplication. Without DNMT1, the methylome

would not be maintained during cell divisions and would be progressively lost as the

consequence of a dilution effect (Li et al. 1992).

DNMT3A and 3B are called the “de novo methyltransferases” because they are

responsible for the establishment of the methylation patterns during early development

(Fig. 8). Loss of one or both enzymes in mice leads to embryonic or post-natal mortality,

highlighting the crucial role of DNA methylation in development (Kato et al. 2007). The

DNMT3 enzymes are also important for methylation pattern in adults. For instance,

DNMT3A was shown to be required for the maintenance of DNA methylation and

synaptic function in adult forebrain neurons (Feng et al. 2010). However, DNMT1 and

Page 34: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

34

DNMT3 enzymes are not strictly restricted to their respective “maintenance” and “de

novo” methylation functions. Emerging evidence indicate that DNMT1 may also be

necessary for de novo methylation, while DNMT3 enzymes also contribute to

maintenance methylation (Fatemi et al. 2002; Walton et al. 2011).

Fig. 8: De novo and maintenance methylation. Unmethylated DNA is methylated de

novo by DNMT3A and 3B. Upon DNA replication, the newly synthesized DNA strand is

unmethylated at first. 5mC symmetry is then maintained by DNMT1. (Adapted from

http://atlasgeneticsoncology.org/Deep/DNAMethylationID20127.html)

In contrast, DNMT3L displays no catalytic activity. It is nevertheless required for

the establishment of genomic imprints, as the protein enhances the activity of DNMT3A

and 3B (Hata et al. 2002). The last member of the family, DNMT2, is active but the

enzyme but does not appear to methylate DNA in human cells. Instead, it is involved in

the methylation of the anticodon loop of aspartic acid transfer RNA (Tuorto et al. 2012).

1.3.1.1.3 Gene silencing mediated by DNA methylation

Repression of a gene by methylation is now a well-established event, particularly

in a promoter context. Two different models have been proposed for the underlying

mechanisms (Fig. 9). In the first model, the presence of 5mC prevents the binding of

transcription factors that are required to activate the expression of the gene. For instance,

at the well-known imprinted IGF2 gene cluster, binding of the CTCF factor to the imprint

control region 1 (ICR1) is prevented by methylation on the paternal allele (Hark et al.

2000). The second model relies on the recruitment of methyl-CpG-binding domain

(MBD) proteins which can specifically recognize and bind methylated CpGs. Those

Page 35: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

35

MBDs can then in turn recruit repressing factors. For instance, MeCP2, the founding

member of the MBD family, is able to recruit histone deacetylases (HDAC) and other

chromatin-remodeling factors (Jones et al. 1998). This last example highlights how, most

often, epigenetic mechanisms are not acting alone, but instead they are part of

multifaceted regulatory networks. The role of histone modifications will be explored in a

subsequent chapter (see 1.3.2 Histone modifications).

Fig. 9: Transcription silencing by DNA methylation. Methylation of CpGs can prevent

the binding of transcription factors (TF) required for gene expression, as well as recruit

repressive complexes through binding of MBDs (e.g. MeCP2). The complexes can

comprise histone modifiers (e.g. HDAC1, HMT) or other a transcriptional regulatory

protein (e.g. Sin3a). (Adapted from Feng & Nestler 2010).

1.3.1.1.4 Biological and pathological relevance of 5mC

It is important to note that DNA methylation patterns are vastly modified during

the life of an organism. This is a dynamic process: 5mC can be added by its writers, the

DNMTs, but it can also be removed. DNA demethylation occurs either as a passive or an

active mechanism. In the first case, 5mC is progressively lost as the result of the absence

of DNMT1 across successive cycles of DNA replication. In the second case, an enzymatic

reaction is involved, independently of the cell cycle. The latter will be explained in detail

in a subsequent chapter (see section 1.3.1.2 Hydroxymethylation of cytosines).

In mammals, there are two massive waves of DNA methylation and demethylation

during development (Fig. 10) (Lee et al. 2014). The first wave of demethylation occurs

in early embryogenesis, starting in the zygote just after fecundation and continuing in the

first few embryonic replication cycles of morula and blastula. DNA is then remethylated

in a cell-type and tissue-specific manner. The second wave of 5mC remodeling occurs

during gametogenesis: all DNA patterns are erased in the germline and the genome is

specifically remethylated in the gametes.

Page 36: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

36

Fig. 10: Dynamicity of 5mC during development. Two waves of DNA demethylation

and (re)methylation occur during development: in the zygote and early embryo; and in

the germline (male: blue line, female: red line). In somatic cells, DNA methylation varies

in a tissue-specific manner and remains globally stable. PGC: Primordial Germ Cells.

ICM: Inner Cell Mass. (Adapted from Lee et al. 2014)

DNA methylation patterns are crucial to proper development and cell

differentiation. In embryonic stem (ES) cells, maintenance of the pluripotency is

controlled by a set of transcription factors, including OCT4, NANOG, and SOX2, whose

expression is controlled by promoter methylation (Altun et al. 2010). In somatic adult

tissues, as well, emerging evidence has highlighted the role of 5mC in multipotency, cell-

fate commitment and reprogramming. For instance, similarly to ES cells, OCT4 appears

to be regulated by promoter methylation in adult stem cells (Lee et al. 2010). Additionally,

DNA methylation changes were observed during somatic differentiation, particularly in

enhancer and promoter regions. As such, in the intestinal tract, abnormal 5mC during

differentiation can lead to aberrant crypt foci and disrupt the absorptive barrier of the

intestinal epithelium (C.-Z. Huang et al. 2015).

Page 37: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

37

DNA methylation aberrations have been observed in a wide range of pathologies,

yet in none so extensively as in cancer. Malignant cells are characterized by a global loss

of 5mC, concomitant with a local hypermethylation (Fig. 11) (Dawson & Kouzarides

2012). On one hand, the global hypomethylation of the genome is mainly linked to a

decrease of 5mC in repetitive and gene-poor regions. This event affects genome stability

and allows the reactivation of transposable elements that can randomly integrate in the

genome, thus potentially disrupting genes. DNA hypomethylation can also lead to the

aberrant activation of genes involved in proliferation and tumor growth (e.g. R-Ras,

MASPIN and MAGE). On the other hand, the hypermethylation observed locally can also

cause the repression of control genes, particularly tumor suppressor genes (TSG, e.g. Rb,

p16, BRCA1 and MLH1). Interestingly, these aberrant 5mC patterns arise even in the

absence of any mutation in the DNMT genes (Brookes & Shi 2014; Mirabella et al. 2016).

Fig. 11: Aberrant DNA methylation in cancer. Local hypermethylation of tumor

suppressor genes (left) and global hypomethylation, including repetitive elements (right),

are hallmarks of cancer cells. (Adapted from Lao & Grady 2011)

Thus, DNA methylation changes are very broad in cancer, and cancerology has

taken an interest in epigenetics. First, 5mC displays an intriguing potential as a prognostic

and predictive tool. For instance, repression of MGMT (O6-methylguanine-DNA

methyltransferase) gene by promoter hypermethylation is associated with a better

response to chemotherapy in glioma patients (Esteller et al. 2000). DNA methylation,

which is very specific of the cell-type, also offers a snapshot of the tissue composition,

with implications for survival and response to treatment, which can notably be predicted

Page 38: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

38

by the immune infiltration of the tumor (Oble et al. 2009; Dedeurwaerder & Fuks 2012;

Melichar et al. 2014). Beyond the use of 5mC as a marker, DNA methylation also offers

opportunities in terms of therapeutics. Demethylating agents have shown promising

results in the treatment of myelodysplastic syndromes, demonstrating minimal side

effects of long-term treatment, and they can be used in combination with conventional

treatments, such as chemotherapy and immunotherapy (Dawson & Kouzarides 2012).

These topics are further detailed in the context of breast cancer in a subsequent chapter

(see 2.5.2 Clinical relevance of epigenetics).

1.3.1.1.5 Mapping the methylome

Recent advances in DNA methylation profiling technologies have widely

improved the coverage of the methylome queried. Such techniques have changed 5mC

profiling, from a candidate gene and promoter focus to genome-wide approaches

encompassing all genomic regions.

The gold standard for DNA methylation is the bisulfite sequencing (BS-Seq), a

method based on the chemical conversion of unmethylated cytosines into thymidines,

combined with high-throughput sequencing. The strength of bisulfite sequencing is the

single nucleotide resolution (Andrews et al. 2012). Another approach consists in the

enrichment of methylated DNA fragments, followed by sequencing. The enrichment can

be achieved by immunoprecipitation (MeDIP-Seq) or by binding to MBD domains

(MethylCap-Seq or MBD-Seq) (Nair et al. 2011). The coverage of the sequencing-based

methods is unrivaled. Yet, these technologies exhibit a number of technical issues,

including fragment size selection and sequence depth. Also, the costs remain relatively

high, often preventing the processing of large cohorts.

In contrast, the array-based Infinium technology (Illumina, San Diego, USA)

constitutes a compromise between coverage and costs (both in terms of labor and money),

a clear benefit when studying larger cohorts (Dedeurwaerder et al. 2013). Although the

first version of the array focused mainly on promoters, the latest version can investigate

over 850,000 CpGs targeting promoters, intragenic and intergenic regions (Moran et al.

2015). This improvement in the Infinium technology clearly reflects the growing demand

in research to assess DNA methylation across the entire genome.

Page 39: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

39

1.3.1.2 Hydroxymethylation of cytosine

While 5mC has long focalized the attention of researchers in terms of DNA

modifications, additional epigenetic marks have since come to light. In 2009, two groups

reported the presence of 5-hydroxymethylcytosine (5hmC) in mammalian genome

(Kriaucionis & Heintz 2009; Tahiliani et al. 2009). This discovery, along with the

involvement of 5hmC in DNA demethylation, completely changed the perception of

DNA modifications. It revealed a remarkably dynamic system and raised a series of

questions regarding the roles of 5hmC in transcriptional regulation, development and

pathologies.

1.3.1.2.1 5hmC distribution

DNA hydroxymethylation occurs through the oxidation of pre-existent 5mC, thus

the mark is predictably also found at CpGs in the mammalian genome. Yet, studies

showed that the distributions of the two marks are dissimilar to an extent. In both ES and

neuronal cells, 5hmC is found enriched in euchromatin, whereas 5mC accumulates rather

in the heterochromatin (Ficz et al. 2011; Szulwach et al. 2011; Chen et al. 2014). Overall,

the level of 5hmC across genes shows a clear drop in the promoter, around transcription

start sites (TSS), and increased levels in gene bodies (Fig. 12) (Song et al. 2011). And

while 5mC abundance is stable in most tissues, 5hmC levels are lower than 5mC and vary

highly between tissues: up to 40% of 5mC levels in Purkinje neurons versus only 7% in

mouse ES cells, and even less in other organs and cultured cell lines (Globisch et al. 2010;

Ito et al. 2010; Szwagierczak et al. 2010).

Globally, DNA hydroxymethylation is considered a mark of active gene

expression (Fig. 12) (Song et al. 2011), yet the correlation between 5hmC and expression

is not always straightforward. For instance, in ES cells, 5hmC is mostly absent of CGI

promoters, which are transcriptionally active. Yet, the mark is enriched at the “bivalent”

CGI promoters, which are marked by both repressive and active histone modifications.

Such genes, while repressed in ES cells, are “poised” for transcription upon

differentiation signals (Butler & Dent 2013).

DNA hydroxymethylation accumulation is frequently observed in the gene body,

especially in exons, and it positively correlates with gene expression: high intragenic

5hmC is observed in genes highly expressed and vice versa (Pastor et al. 2011; Song et

Page 40: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

40

al. 2011; Wu et al. 2011; Szulwach et al. 2011). Finally, 5hmC is also distributed at

intergenic cis-regulatory elements, such as active enhancers (Pastor et al. 2011; Szulwach

et al. 2011). The effect of 5hmC in such regions has yet to be elucidated.

Fig. 12: 5hmC distribution in the cerebellum.

Metagene profiles of 5hmC (divided into four

bins based on gene expression, see color

legend) and input genomic DNA in adult mouse

cerebellum. 5hmC drops at the promoter, is

enriched in the gene body, and correlates with

gene expression. (Adapted from Song et al.

2011)

1.3.1.2.2 The TET enzymes

Oxidation of 5mC is catalyzed by the Ten-Eleven Translocation (TET) enzymes

(Fig. 13). The founding member of the family, TET1, was originally described in myeloid

leukemia, in which the gene is often translocated from chromosome 10 to chromosome

11, hence its name (Ono et al. 2002). The TET enzymes are members of the TET/J-

binding protein (JBP) family of 2-oxoglutarate (2-OG) and iron (II)-dependent

dioxygenases. In both human and mice, all three TET proteins are active, and their

catalytic domain is composed of a cysteine-rich region and a double-stranded β-helix

(DSBH) domain. TET1 and TET3 both possess a CXXC DNA binding domain, while

TET2 does not. Instead, TET2 interacts with a separate CXXC protein, encoded by a gene

called IDAX or CXXC4.

Page 41: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

41

Fig. 13: Activity and structure of human TET proteins. TET1–3 can convert 5mC into

5hmC. The enzymes contain a cysteine (Cys)-rich region followed by the double-stranded

β-helix (DSBH) fold characteristic of the 2-oxoglutarate-Fe(II) dioxygenases and

required for catalytic activity. TET1 and TET3 also contain a CXXC domain. (Adapted

from Williams et al. 2012)

1.3.1.2.3 Active demethylation and DNA hydroxymethylation

In 2009, Tahiliani et al. showed in a seminal article that overexpression of

wildtype TET1, but not its catalytic mutant counterpart, led to a decrease of 5mC

(Tahiliani et al. 2009). This “demethylase” activity was soon extended to TET2 and TET3

(Ito et al. 2011). The intriguing possibility that TET-mediated DNA hydroxymethylation

might act as an intermediate in the demethylation process generated considerable interest

in the scientific community. While the existence of active demethylation had been long

postulated, the underlying mechanisms had remained mostly elusive. In this section, we

will explain what is known of TET-mediated demethylation.

Because the methyl group of 5mC is thermodynamically very stable, the direct

removal of the moieties from the cytosine is unlikely, as it would require huge energy

expenditure (Wu & Zhang 2010). Instead, one of the more realistic mechanisms for active

demethylation is the cleavage of the glycosyl bond between the ribose and the base by a

DNA glycosylase. This event results in an abasic site that can be removed and replaced

with an unmodified cytosine by the base excision repair (BER) machinery of the cell.

This mechanism would necessitate the selective targeting of glycosylases to the CpGs to

be demethylated. Studies have highlighted two main mechanisms by which 5hmC might

be targeted for such demethylation (Fig. 14). The first pathway involves the deamination

of 5hmC into 5-hydroxymethyluracile (5hmU) by AID/APOBEC proteins (Guo et al.

2011). The second possibility requires the iterative oxidation of 5hmC into 5-

formylcytosine (5fC) and 5-carboxycytosine (5caC), which is mediated by the TET

Page 42: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

42

enzymes (Shen et al. 2013). In all cases, removal of the 5hmC derivatives is achieved

through a glycosylase, such as TDG or SMUG1, followed by BER-mediated “repair”.

Fig. 14: Pathways of cytosine demethylation mediated by TET proteins. 5hmC might

facilitate passive and active demethylation pathway. (Adapted from Williams et al. 2012)

A key regulator of active demethylation events is the GADD45 family of proteins.

Despite a lack of any enzymatic activity, GADD45 acts as a link between demethylation

targets and the DNA repair machinery (Niehrs & Schäfer 2012). It was notably reported

that GADD45 directly binds to TET1 and increases its oxidation activity (Kienhöfer et

al. 2015). Furthermore, GADD45 enhances TDG-mediated removal of 5fC and 5caC.

Accordingly, knockout of both Gadd45a and Gadd45b from mouse ES cells leads to

hypermethylation of many genes that are targeted by TDG (Z. Li et al. 2015).

Other mechanisms for TET-mediated demethylation have also been proposed. Of

note, 5hmC might simply promote passive demethylation (Fig. 14), as DNMT1 appears

to methylate 5hmC‐containing DNA less efficiently than 5mC hemi‐methylated DNA

(Valinluck & Sowers 2007). Another intriguing hypothesis that has been suggested is the

oxidative demethylation, which consists in the decarboxylation of 5caC residues.

Carboxylases are widespread in protein signaling, which raises the question of a putative

“DNA carboxylase”. In support to this hypothesis, 5caC-decarboxylating activity was

reported in mouse ES cells and DNA methyltransferases were found to catalyze the direct

decarboxylation of 5caC in vitro (Schiesser et al. 2012; Liutkevičiutè et al. 2014).

However, the existence of an endogenous 5caC decarboxylase remains uncertain.

In summary, TET-mediated DNA demethylation is initiated by the oxidation of

5mC into 5hmC and can be achieved though several ways: passive demethylation,

Page 43: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

43

glycosylase/BER machinery, and possibly 5caC decarboxylation. These mechanisms are

not necessarily exclusive and might occur in a context-dependent manner.

1.3.1.2.4 Biological and pathological relevance of 5hmC and TETs

The role of 5hmC as an intermediate in DNA demethylation is critical, yet the

high abundance of the mark in some tissues suggests that 5hmC is more than a transient

residue (Kriaucionis & Heintz 2009). Interestingly, increasing evidence points out that

proteins can bind 5hmC with various affinities and specificities (compared to 5mC),

including some MBDs and UHRF1, and some of these proteins are known to be involved

in gene regulation (Jin et al. 2010; Mellén et al. 2012; Otani et al. 2013; Spruijt et al.

2013). Hence, 5hmC has extensive potential, both for its demethylating role and as an

epigenetic mark of its own. And, although the three human TET genes display a high

similarity in sequence, they are not fully redundant and, to an extent, exert different

functions. Accordingly, both TET expression and 5hmC patterns vary specifically

between tissues (Szwagierczak et al. 2010; Ponnaluri et al. 2017).

The brain and the nervous system constitute a particularly relevant setting to study

the function of TETs and 5hmC because (1) DNA hydroxymethylation levels are the

highest in those tissues, (2) 5mC plays an essential role in neurogenesis, and (3) gene

regulation by active DNA demethylation is particularly pertinent in post-mitotic neurons

(where passive demethylation cannot occur by lack of cell division). Interestingly, several

studies concluded that 5hmC patterns acquired during development are required for

normal neurodevelopment and neurological functions in the adult brain. Notably, Tet1-

mediated 5hmC causes promoter demethylation and activation of growth factor genes

Bdnf and Fgf1 in the adult mouse brain and provides protection against oxidative stress

and neuronal cell death (Guo et al. 2011; Xin et al. 2015; Hsieh et al. 2016). Accordingly,

alterations of 5hmC may contribute to neurodevelopmental and neurodegenerative

diseases, including Rett syndrome, schizophrenia, and Alzheimer’s disease, among others

(Cheng et al. 2015).

Another system in which 5hmC and TETs have been extensively studied is

embryonic stem cells, and initial reports indicated that TET1, in particular, might be

implicated in pluripotency maintenance. Both Tet1 and Tet2 are elevated in mouse ES

cells and in the inner cell mass of the blastocyte; and Tet1-knockdown led to spontaneous

Page 44: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

44

differentiation of ES cells (Ito et al. 2010). Another study reported that knockdown of

Tet1 and Tet2 during differentiation of ES cells was also associated with decreased 5hmC

levels at the promoters of pluripotency genes, leading to hypermethylation and gene

silencing (Ficz et al. 2011). However, this notion is in debate, as both Tet1-knockout and

Tet1/Tet2-double knockout mice are able to reproduce, despite some moderate perinatal

lethality and slightly reduced size and weight at birth. Thus, ES stemness could be

maintained in vivo without TET1 and TET2 (Dawlaty et al. 2011; Dawlaty et al. 2013).

As mentioned previously, a massive wave of demethylation of both maternal and

paternal genomes occurs in the zygote and during preimplantation development. Studies

have suggested that 5hmC and passive demethylation both contribute to paternal DNA

demethylation whereas the maternal genome appears to be demethylated mainly through

replication-dependent dilution effect (Fig. 15). Briefly, the paternal pronucleus appears

to be “demethylated” through maternal TET3-mediated oxidations of 5mC (into 5hmC,

5fC and 5caC), and those residues are then diluted in a replication-dependent manner

(Inoue & Zhang 2011). The maternal genome, however, is protected from demethylation

by the exclusion of TET3 by PGC7, a protein highly enriched in the maternal chromatin

(Kang et al. 2013). Interestingly, Tet3-knockout zygotes display a lack of 5mC

conversion into 5hmC in the paternal genome, which delays the subsequent activation of

paternal Oct4 and Nanog genes in early embryos. Loss of maternal TET3 led to reduced

fecundity and neonatal lethality of the offspring (Gu et al. 2011). Therefore, Tet3-

mediated epigenetic reprogramming of the zygotic paternal DNA is essential to early

development.

Fig. 15: View of DNA

demethylation in the

zygote. The maternal

genome is passively

demethylated while the

paternal pronucleus is

oxidized by TET3.

5hmC is then diluted

through successive

rounds of replications.

TE= trophectoderm,

ICM=inner cell mass.

(Adapted from Wu &

Zhang 2011)

Page 45: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

45

Similarly to 5mC, DNA hydroxymethylation is vastly affected in cancer. Global

loss of 5hmC (Fig. 16) and impaired TET function have become a new hallmark of

cancers (Haffner et al. 2011; Lian et al. 2012). In hematopoietic malignancies, decreased

5hmC levels have frequently been associated to genetic aberrations in the TET2 gene

(microdeletions, loss of heterozygosity or mutations targeting the catalytic domain). The

high frequency of mutations in leukemia (up to 27% of cases in myelodysplasia) is

thought to be linked to the essential role of TET2 in hematopoietic development and

transformation, which was characterized in several Tet2-deficient mice. In solid cancers,

loss of 5hmC is also observed, despite the lack of any frequent TET mutation. Instead,

impairment of TET activity may be achieved though promoter hypermethylation,

microRNA (miRNA) interference, post-translational regulation or alterations of cofactors

and interactors. These events result in low gene expression, altered protein stability and

localization, or reduced enzymatic activity (Jeschke et al. 2016).

Fig. 16: 5hmC changes in

cancer. The majority of cancers

display a global reduction in

5hmC compared to normal

tissue, which is reflected in

various regions of the genome.

(Adapted from Jeschke et al.

2016)

Genome-wide profiling of 5hmC in various tumor types revealed that loss of

5hmC affects all genomic regions: promoters, gene bodies, intergenic regulatory regions,

and even repetitive elements. Unlike DNA methylation, loss of 5hmC in cancer is not

focused on gene-poor regions and occurs across the entire genome. Accordingly, several

genes were reported to be hypo-hydroxymethylated, e.g. RAC3, IGF1R, and TIMP2 genes

in melanoma (Lian et al. 2012). However, despite the global loss, genes can be affected

by both gain and loss of 5hmC. Importantly, these changes correlate positively with gene

expression. In leukemia, intergenic regions, and, in particular, enhancers, were reported

to be extensively affected by 5hmC changes (Rampal et al. 2014; Rasmussen et al. 2015).

In conclusion, genome-wide studies of DNA hydroxymethylation remain rare, however

they highlight broad changes in the hydroxymethylome. Given that 5hmC is highly tissue-

specific in normal cells, further studies will be required to investigate the diversity of

5hmC changes among tumor types.

Page 46: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

46

1.3.1.2.5 Mapping the hydroxymethylome

Several techniques have been recently developed to map 5hmC patterns across the

genome. In this section, we will briefly explain the main existing methods.

First, it is important to note that the bisulfite sequencing (BS-Seq), which is the

gold standard to profile the methylome, cannot in fact distinguish 5mC from 5hmC: both

modifications protect DNA from C to T bisulfite-based conversion. Nevertheless, given

the much higher abundance of 5mC (about 100-fold compared to 5hmC in most tissues),

BS-Seq results are still relevant for the study of the DNA methylome.

Interestingly, 5hmC can be rendered sensitive to bisulfite conversion, by adding a

preceding step of selective chemical oxidation of 5hmC (into 5fC). Hence, in the

oxidative bisulfite sequencing (oxBS-Seq), unmodified C and 5hmC are both converted

to T, whereas in the classical BS-seq only unmodified C will be converted (Booth et al.

2013). Comparison of both profiles provides a quantitative measurement of 5mC and

5hmC in parallel and at single-base resolution. Another adaptation of the BS-Seq is the

Tet-assisted bisulfite sequencing (TAB-Seq). In this case, it is 5mC that is made sensitive

to bisulfite-conversion by TET-mediated oxidation (into 5caC), whereas 5hmC is

protected beforehand from oxidation by a specific glycosylation step (Yu et al. 2012).

Hence, in TAB-Seq, both unmodified C and 5mC are converted to T. Again, comparison

with classical BS-Seq (where only unmodified C are converted) provides 5hmC mapping

at single nucleotide resolution.

Other approaches rely on the enrichment of hyroxymethylated DNA fragments,

followed by high-throughput sequencing. The capture can be achieved by several

methods: (1) by direct 5hmC-targeting immunoprecipitation (hMeDIP), (2) by tagging

5hmC for a biotin pulldown through a specific glycosylation step, which is mediated by

the β-glucosyltransferase enzyme (hMe-Seal), (3) or by bisulfite conversion of 5hmC to

cytosine 5-methylenesulphonate (CMS), followed by immunoprecipitation of CMS by a

specific antibody (Song et al. 2011; Pastor et al. 2012; Nestor & Meehan 2014). While

the resolution of these affinity-based methods is lower than for oxBS and TAB-Seq, the

lower cost and availability of commercial, ready-to-use kits make them very attractive,

particularly for large cohorts of samples.

Page 47: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

47

1.3.1.3 Other covalent DNA modifications

Other covalent modifications of the DNA have also been investigated in recent

years. We already mentioned 5-formylcytosine (5fC) and 5-carboxycytosine (5caC), two

residues derived from the further oxidation of 5hmC. In addition, next to cytosine

modifications, methylation of adenine (m6A) has been recently described. In this chapter,

we will summarize emerging data on these “new” DNA modifications.

1.3.1.3.1 Oxidative derivatives of 5hmC

TET enzymes mediate the iterative oxidation of 5mC into 5hmC, 5fC and 5caC.

Although much less abundant than 5hmC, both 5fC and 5caC are detectable in ES cells,

and 5fC is also found in various tissues, including brain, spleen, or liver (Ito et al. 2011).

In mouse ES cells, 5fC is found enriched at CGI promoters and exons of

transcriptionally active genes. Loss of TDG leads to increased 5fC in CGIs and correlates

with increased 5mC in these regions during differentiation of ES cells (Raiber et al. 2012;

Shen et al. 2013). Therefore, 5fC appear to play a role in epigenetic reprogramming of

specific loci, specifically related to DNA methylation regulation. These results were

confirmed in vivo, as TDG also appears to shape 5fC distribution at CGI in mouse

embryos. The mark was also enriched at active enhancers and intragenic regions, with

highly tissue-specific patterns, which suggests a role in embryonic development (Iurlaro

et al. 2016).

Unlike other cytosine modifications, 5caC remains undetectable in most somatic

tissues, which suggests that its role might be predominantly that of a transient mark and

makes the study of this modification very difficult. It is however found in ES cells and

early mouse embryo. Like 5fC, 5caC accumulates upon loss of TDG at intra- and

intergenic regulatory elements (Shen et al. 2013). It was also reported that 5caC

transiently increases during lineage specification of neural stem cells and hepatic

differentiation, before dropping in differentiated cells (Wheldon et al. 2014; Lewis et al.

2017). Therefore, active DNA demethylation seem to occur extensively in the

mammalian genome during early development.

In conclusion, it is still very early day for the characterization of oxidative

derivatives of 5hmC and much effort has yet to be provided in order to fully comprehend

Page 48: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

48

their respective roles in the epigenome. Future studies will likely explore the distribution

of these epigenetic modifications in additional tissues. Also, identification of potential

specific binders might expand our understanding of their roles.

1.3.1.3.2 Methylation of adenine

In 2015, it was reported that adenosine can be methylated into N6-

methyladenosine (m6A) in mouse genomic DNA (Koziol et al. 2016). This modification

is extremely infrequent (0.00009% of A), although it is enriched in gene-poor regions. In

mouse ES cells, m6A enrichment correlates with epigenetic silencing of LINE-1

transposons (Wu et al. 2016). The Alkbh1 protein was identified as a demethylase for

m6A, although its writer enzyme is still unknown. In any case, early results indicate that

m6A might constitute an important component of the epigenetic regulation in mammalian

genomes.

Interestingly, m6A has been known to exist in bacterial DNA for a long time, but

its existence in mammalian genome is only a recent discovery. In bacteria, methylation

of adenine by the Dam methylases is well-characterized and is associated with protection

from bacteriophage restriction enzymes, DNA replication and bacterial virulence

regulation (Ratel et al. 2006). The roles of m6A in mammalians and bacteria might be

very different.

1.3.2 Histone modifications

The second category of epigenetic modifications is the post-translational

modification of histone tails (Fig. 17). It is a major type of epigenetic modification with

wide implications in terms of chromatin modulation and gene transcription. Over a

hundred histone modifications are known. Among them, the best-characterized are

acetylation, methylation, phosphorylation, and ubiquitination, all of which can affect

various residues of the histone tails with various consequences. Together, these

modifications form the “histone code” that regulates gene transcription (Bernstein et al.

2007). In this chapter, we will explain the main chemical modifications of histones, their

machineries, and their functions. Since the topic of histone modifications was not the

Page 49: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

49

focus of our research, we will restrict our description to the essential facts related to the

matter.

Fig. 17: The main modifications of the four core histones. Many residues of the histone

tails can be chemically modified. Ac, acetylation; Me, methylation; P, phosphorylation;

Ub, ubiquitination. (Adapted from Rodríguez-Paredes & Esteller 2011)

1.3.2.1 Histone acetylation

Histone acetylation corresponds to the transfer of an acetyl moiety from acetyl-

coenzyme A to a lysine residue of the N-terminal histone tail. This reaction is catalyzed

by a family of enzymes called histone acetyltransferases (HATs). The acetyl moiety can

also be removed by enzymes called histone deacetylases (HDACs), thus providing a

dynamic regulation of the mark (Dawson & Kouzarides 2012). HATs can be classified

into 2 main categories: the type A HATs (including the GNAT, MYST and CBP/p300

families) are localized in the nucleus and involved in chromatin regulation whereas the

type B HATs are localized in the cytoplasm and acetylate newly synthesized histones

before nucleosome formation. In contrast, there are 4 classes of HDACs, based on

sequence homology to the enzymes originally identified in the yeast and domain

organization (Witt et al. 2009).

Page 50: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

50

Acetylation is a histone mark associated with active transcription. Histone tails

are globally positively charged because of the presence of many lysine and arginine

residues and their positive amine group. These positive charges allow a proper interaction

with the DNA molecule, which is negatively charged due to the phosphate groups of the

molecule’s backbone. Acetylation neutralizes the positive amine charges of the lysine

residues, and thus decreases the binding affinity of histones to DNA, which in turn favor

chromatin opening (Fig. 18). Conversely, removal of the acetyl moiety by HDACs

increases binding of histones to the DNA, and thus promotes chromatin compaction and

silencing of transcription. In addition, acetylation of histone tails also allows the specific

recruitment of transcription factors through their bromodomains.

Fig. 18: Histone acetylation.

Acetylation neutralizes the

positive charge of the lysine

and promotes chromatin

opening. (Adapted from Korzus

2010)

Genome-wide mapping of histones marks is traditionally performed by

chromatin-immunoprecipitation, followed by high-throughput sequencing (ChIP-Seq).

Histone acetylation occurs at many residues, but the main sites are located on histones

H3 and H4 (Bannister & Kouzarides, 2011). Among them, acetylation of histone H3

lysine 4 (H3K4ac), lysine 9 (H3K9ac), and lysine 27 (H3K27ac) are enriched at

promoters of active genes. But histone acetylation is not restricted to promoters and is

also found at cis-regulatory regions. For instance, H3K27ac is also found at active

enhancers.

1.3.2.2 Histone methylation

Histone methylation consists in the addition of a methyl group to a lysine or an

arginine residue of an N-terminal histone tail. There are different levels of methylation,

Page 51: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

51

as the residues can be mono-, di-, or trimethylated. Methylation of arginine is catalyzed

by protein arginine methyltransferases (PRMTs), while methylation of lysine is catalyzed

by histone methyltransferases (HMTs). The methyl group can also be removed by histone

demethylases (HDMs), which once again allows a dynamic regulation of the mark

(Dawson & Kouzarides 2012).

The effect of the modification on gene expression depends on the specific residue

targeted and the level of methylation (Fig. 19). Unlike acetylation, methylation does not

neutralize the positive charge of these amino acids. Therefore, the effects of the

modifications are due to the recruitment of specific factors and vary widely. For instance,

trimethylation of histone H3 lysine 4 (H3K4me3) is typically associated with active

promoters, whereas trimethylation of histone H3 lysine 9 (H3K9me3) and lysine 27

(H3K27me3) are associated with gene repression (Mirabella et al. 2016). Other

methylation marks are involved in DNA replication, enhancer activity, transcription

elongation and DNA repair (Mosammaparast & Shi 2010).

Fig. 19: Major methylation events of histones H3 and H4. The effect of histone

methylation depends on the targeted amino acid and the level of methylation. The function

of each mono-, di-, and tri-methylation state is detailed by the color code, as explained

in the figure legend (Adapted from Mosammaparast & Shi 2010).

In embryonic stem cells, certain promoters of developmental regulatory genes,

referred to as “bivalent promoters”, bear both active H3K4 and repressive H3K27

trimethylation marks (Butler & Dent 2013). These promoters are not active as such, but

are poised for transcription. During differentiation, loss of H3K27 methylation marks can

Page 52: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

52

lead to a rapid activation of the gene. In contrast, bivalent genes that lose H3K4 methyl

marks will be silenced and targeted by heterochromatin.

1.3.2.3 Other histone modifications

Acetylation and methylation are the two major post-translational modifications of

histones, but other modifications have been characterized in recent years, broadening the

panel of histone marks. In this section, we will briefly explore additional histone marks.

Phosphorylation of histone is a highly dynamic mark that occurs on several amino

acids: serine, threonine, tyrosine, arginine, histidine and lysine. The addition of the

phosphate group is mediated by kinases, with ATP as a cofactor, while the removal of the

phosphate from the histone is mediated by phosphatases (Sawicka & Seiser 2012).

Similarly to acetylation, phosphorylation reduces the positive charge of histones and

influences chromatin structure. The modification can also be bound by specific factors

and has been implicated in a variety of functions, including gene activation, chromosome

condensation and segregation during mitosis and meiosis, DNA repair or apoptosis.

Mono-ubiquitination of histone is a large modification, with the addition of the

76-amino acid peptide that is ubiquitin. It occurs on histone H2A and H2B, where it is

associated with gene repression and transcription initiation, respectively (Bannister &

Kouzarides 2011).

Glycosylation has also been described in the context of histones. The OGT

enzyme can catalyze the addition of the O-linked N-acetylglucosamine (O-GlcNAc)

moiety on serine and threonine residues of histone H2B (Fujiki et al. 2011). This reaction,

referred to as “GlcNAcylation”, can antagonize histone phosphorylation because it occurs

on the same residues. The removal of O-GlcNAc is mediated by the O-GlcNAcase (also

called OGA) enzyme. This modification appears to act as a recruitment signal for the

H3K4me3 machinery, therefore it ultimately promotes gene expression. The O-GlcNAc

mark is closely related to the regulation of metabolic pathways and provides a link

between nutrition and epigenetics (Aquino-Gil et al. 2017).

Page 53: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

53

1.3.3 Epigenetic regulation of the chromatin

1.3.3.1 Crosstalk between epigenetic modifications

In previous chapters, we have described the main epigenetic modifications.

However, these modifications do not act as isolated marks, but rather as parts of a

complex system. In fact, various modifications decorate histones at the same time, and

together they form the “histone code”. In addition, the histone modifications interact

closely with DNA modifications. It is the integration of the various epigenetic signals that

ultimately leads to the regulation of chromatin (Fig. 20).

Fig. 20: Crosstalk between DNA and histone modifications. Shades of green indicate

active histone marks, whereas shades of pink represent repressive histone marks. The

orange and yellow colors mark regions of 5mC and 5hmC, respectively. (A) The

distribution of histone marks is illustrated across the promoter region, TSS

(transcriptional start site), and gene body of an active gene. (B) Histone H3 methylation

and DNA methylation are found in the promoter region and TSS in repressed genes. (C)

Bivalent chromatin domains display histone H3K4me3, H3K27me3, and 5hmC. (Adapted

from Butler & Dent 2013).

We have already mentioned the case of bivalent promoters in ES cells, which are

inactive, but poised, promoters of developmental genes that bear 5hmC, as well as active

H3K4me3 and repressive H3K27me3. We will now provide additional examples of how

Page 54: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

54

DNA modifications and the histone code can work hand in hand to regulate chromatin

and gene expression.

Histone marks have the capacity to recruit DNMTs and guide DNA methylation.

This occurs, for instance, during the formation of heterochromatin (Fig. 21). Regions of

repressed chromatin are typically marked by H3K9me3. The Heterochromatin Protein 1

(HP1) binds to H3K9me3 through its specific chromodomain and, in turn, recruits

DNMTs in order to methylate DNA. Therefore, post-translational modifications of

histones contribute to shaping the 5mC landscape. The presence of 5mC is not strictly

required for the formation of heterochromatin, however this double regulation allows to

lock tightly the heterochromatin. Conversely, DNA methylation can also influence

histone marks. For instance, 5mC found in heterochromatin can be bound by MBDs

which subsequently recruit HDACs. The deacetylation of histone is an additional

mechanism that represses gene expression, as the newly deacetylated H3K9 is then

available for trimethylation and further spreading of the heterochromatin.

Fig. 21: Cooperative regulation and

gene silencing. Gene repression in

heterochromatin occurs through a

multi-protein complex. The HP1 protein

recognizes H3K9me3. This eventually

leads to the recruitment of DNMTs for

DNA methylation, HDACs for histone

deacetylation and HMTs for the

spreading of heterochromatin through

additional H3K9 methylation. (Adapted

from Feinberg & Tycko 2004).

Another functional link has been established recently by our host laboratory, and

others, between histone glycosylation and DNA hydroxymethylation machineries in the

context of gene activation (Fig. 22) (Deplus et al. 2013; Chen et al. 2013; Vella et al.

2013). The enzyme responsible for histone GlcNAcylation, OGT, can interact with the

TET proteins, independently of their oxidation activity. Through this association, TETs

recruit OGT on CGI promoters and enhances its enzymatic activity. Subsequently, OGT

glycosylates its target histone H2B, as well as other target proteins. Among them is HCF1,

a key subunit of the H3K4 methyltransferase SET1/COMPASS complex. Glycosylation

of HCF1 stabilizes the complex, which favors H3K4me3 and gene activation. In

Page 55: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

55

conclusion, TETs act as scaffolding proteins by recruiting and activating OGT, which in

turn promotes the H3K4me3 machinery and transcription.

Fig. 22: Connecting TETs and

OGT. (1) TET–OGT interaction

promotes OGT GlcNAcylation on

numerous proteins, including HCF1.

(2) In a TET‐dependent manner, a

GlcNAcylated HCF1 stabilizes the

SET1/COMPASS complex. (3) Both

TET proteins and OGT activity favors

histone H3K4me3 and subsequent

transcriptional activation. (Adapted

from Deplus et al. 2013).

In conclusion, epigenetics brings together several machineries in order to regulate

gene expression. The relations between these different machineries give a more complex

picture, in which different modifications can either reinforce each other (e.g. H3K9me3

and 5mC in heterochromatin) or, on the contrary, balance each other (e.g. H3K4me3 and

H3K27me3 in bivalent promoters of ES cells). As of now, little is known about these

connections, yet they are key to understanding the intricacies of gene regulation. And

considering that new epigenetic modifications are found every year, there is still much

work to be done in that regard.

1.3.3.2 Chromatin remodeling

In addition to the previously described modifications of DNA and histone, a third

epigenetic mechanism can be involved in the regulation of the chromatin, which we will

briefly describe in this section.

Chromatin remodeling is an ATP-dependent mechanism mediated by the

SWI/SNF complexes. In human cells, these are large, multi-protein complexes containing

a single ATPase protein (either BRM or BRG1) and numerous core and accessory

subunits. The genes encoding these subunits belong mostly to the SMARC, ARID and

BCL families (Masliah-Planchon et al. 2015). SWI/SNF complexes are traditionally

divided in two categories, the “BAF complexes” and the “PBAF complexes”, based on

the presence of certain core subunits (Fig. 23, left panel). However, the composition of

Page 56: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

56

these complexes is highly variable, with multiple possible paralogues, and changes

widely depending on the cellular context. Thus, they should not be viewed as stable

complexes with a restricted set of well-defined subunits, but rather as a collection of

highly variable complexes (Wang et al. 1996; Mohrmann & Verrijzer 2005).

The chromatin remodeling complexes can affect histone-DNA interactions, using

the energy released by ATP consumption, and thus deeply change the organization of the

chromatin. This process involves nucleosome sliding, dissociation or replacement, and

these changes in the chromatin organization can either activate or repress transcription of

genes. Interestingly, SWI/SNF subunits contain domains interacting with both histones

and DNA, suggesting once again that different levels of epigenetic regulation can

influence one another. More broadly, members of the SWI/SNF complexes can recruit

transcription factors, as well as modulators of transcriptional activity (either coactivators

or repressors) (Helming et al. 2014). Hence, SWI/SNF complexes can influence

chromatin organization through several mechanisms.

Fig. 23: SWI/SNF complexes. SWI/SNF complexes are found in two major subtypes,

BAF and PBAF, and comprise multiple subunits (left). SWI/SNF complexes contribute to

transcription modulation by mobilizing nucleosomes and by interacting with

transcription factors, coactivators, and corepressors on DNA. Subunits found mutated in

cancer are denoted by a red star and are described in the table (right). (Adapted from

Helming et al. 2014)

Page 57: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

57

SWI/SNF complexes have been shown to play critical roles in physiology, notably

in the regulation of the balance between stemness and differentiation. For instance, there

is a shift in the composition of the SWI/SNF complex during the transition from neural

stem cells and progenitors to post-mitotic neurons, as BAF45a and BAF53a subunits are

replaced by BAF45b, BAF45c and BAF53b subunits (Lessard et al. 2007). SWI/SNF

subunits have also displayed critical roles in adipogenesis, osteogenesis and

hematopoiesis (Wilson & Roberts 2011).

Furthermore, SWI/SNF complexes are involved in the regulation of many genes

related to proliferation and/or cell motility, such as cell cycle regulator p16INK4A, oncogene

c-MYC, or genes of the Rho GTPase family. In direct link with this, alterations of

SWI/SNF chromatin remodeling have been implicated in multiple cancers (Fig. 23, right

panel). For instance, SMARCB1, which is now considered as a bona fide tumor suppressor

gene, is repressed in cancer, either by loss-of-function mutation (both somatic and

germline) or downregulation (Margol & Judkins 2014). In particular, repression of this

SWI/SNF subunit is involved in the development of the aggressive malignant rhabdoid

tumors. Numerous other SWI/SNF subunits are also known to be dysregulated in cancer,

including ARID1A, ARID2, SMARCA2, SMARCC1, and SMARCC2 (Masliah-Planchon

et al. 2015). The involvement of chromatin remodeling complexes in breast cancers will

be mentioned in a subsequent chapter (see section 2.5.1 Epigenetic alterations in breast

cancers).

Page 58: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

58

1.3.4 RNA modifications

Next to classical modifications of the chromatin, i.e. histone and DNA

modifications, a new field of research is emerging: RNA epigenetics. This domain, also

referred to as epitranscriptomics, investigates the post-transcriptional modifications of

RNA.

Over 100 distinct chemical modifications have been identified so far, and,

together, they constitute the “epitranscriptome” of our cells, by analogy to the epigenome

(Liu & Pan 2017). Already, the field of RNA epigenetics is changing our view on the

central dogma of biology, as it is clear that RNA transcripts are not merely transient

copies of the DNA. Instead, just like the chromatin, they constitute a true level of

regulation with the potential to fine-tune the functions of the genes. And these

modifications affect the various classes of RNA, from messenger RNAs (mRNAs) (Fig.

24), to ribosomal RNAs (rRNAs) or transfer RNAs (tRNAs). Since RNA epigenetics is

still in its early days, most of these modifications have not been well characterized yet.

Nevertheless, they represent an intriguing new field of research that will undoubtedly

expand in the years to come.

In this chapter, we will explore the main modifications of the mammalian

transcriptome, their machineries and biological relevance.

Fig. 24: Chemical modifications in eukaryotic mRNA. (Adapted from X. Li et al.

2017).

Page 59: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

59

1.3.4.1 Methylation of adenosine

The methylation of adenosine is, by far, the most abundant modification of mRNA

to be identified. However, the low expression of mRNAs compared to rRNAs and tRNAs

limited for many years the possibilities of research on this modification. Recently, m6A

has regained the interest of the scientific community, due to advances in biochemical

approaches and sequencing technologies, as well as a better understanding of the players

involved in the regulation of m6A (X. Li et al. 2017).

1.3.4.1.1 m6A distribution

As previously mentioned, m6A is a frequent modification in mRNA, with about

1% to 2% of all adenosines being methylated and approximately 2 methylated sites per

transcript (Pan 2013; Yue et al. 2015). The exact abundance of the mark varies depending

on the biological context. In 2012, two independent studies reported the first mappings of

m6A in RNA by immunoprecipitation, followed by high-throughput sequencing (MeRIP-

Seq), and identified thousands of targets. Profiling of m6A across the mammalian

transcriptome revealed that the mark was preferentially found enriched near the

transcription start site (TSS), around the stop codons, and in the 3’ untranslated regions

(3’ UTR) (Fig. 25) (Dominissini et al. 2012; Meyer et al. 2012). Methylated sites

correlated with evolutionary conserved regions between human and mice, suggesting that

the function of m6A might be conserved between species.

Fig. 25: Metagene profiles of m6A. Metagene analysis of MeRIP–Seq data shows that

m6A is enriched around the TSS and the stop codon of mRNA transcripts. (Adapted from

Meyer & Jaffrey 2014)

How specific regions of RNAs are targeted for methylation is still unknown,

however consensus motifs of m6A have been identified, with the most predominant one

Page 60: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

60

being GGm6ACU. Nevertheless, the majority of the regions corresponding to the m6A

consensus motifs were in fact not methylated, suggesting that m6A mRNA modification

is regulated by more factors than simple sequence recognition.

1.3.4.1.2 The m6A machinery

The m6A methyltransferase machinery is a multi-protein complex, consisting of

two active methyltransferases, METTL3 and METTL14, in complex with the

catalytically inactive protein WTAP (Fig. 26) (Maity & Das 2016). METTL3 was the first

m6A methyltransferase identified and displays a SAM-binding domain. METTL14 is a

close homologue, with similar activity, that was shown to interact with METTL3.

Depletion of these enzymes leads to a decrease of m6A, both in vitro and in vivo (Y.

Wang et al. 2014). WTAP, a protein previously known to be involved in mRNA splicing,

was also reported to interact with the core complex of METTL3 and METTL14.

Strikingly, knockdown of WTAP reduced m6A levels more significantly than knockdown

of METTL3 or METTL14, despite the lack of enzymatic activity of the protein (Liu et al.

2014). This suggest that WTAP is required for the proper activity of the METTL3-

METTL14 complex.

Just like DNA and histone modifications, m6A is a dynamic mark that can be

removed from the RNA (Fig. 26). The first m6A demethylase to be identified was the fat

mass and obesity-associated protein (FTO) (Jia et al. 2011). Reports indicated that

knockdown of FTO increased m6A levels, while overexpression decreased them, both in

vitro and in vivo. Recent results indicate that FTO oxidizes m6A into 6-

hydroxymethyladenosine and 6-formyladenosine, suggesting an oxidative demethylation

process similar to TETs and 5mC in the DNA (Fu et al. 2013). A second m6A

demethylase was later identified: ALKBH5 (Zheng et al. 2013). This enzyme, unlike

FTO, catalyzes the direct removal of the methyl group from the adenosine.

Several proteins were suggested to preferentially bind (or be excluded from) m6A-

containing RNAs (Fig. 26). Of note, m6A appears to destabilize the potential binding to

the opposing U base within a hairpin RNA structure, thus forming a single-strand region.

The accessibility of RNA, in turn, allows a better binding of HNRNPC, an abundant

nuclear protein involved in pre-mRNA processing (Liu et al. 2015). In contrast, m6A-

dependent alteration of the RNA structure blocks the binding of HuR, a mediator of post-

transcriptional regulation (Dominissini et al. 2012; Y. Wang et al. 2014). Beyond

Page 61: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

61

structure alteration, several proteins were also identified as specific m6A binders,

including the YTHDF1-3 and YTHDC 1-2 proteins (Zhang et al. 2010). In particular,

YTHDF2 was shown to bind the methylated 3’UTRs and to be involved in mRNA

degradation regulation (X. Wang et al. 2014). In conclusion, m6A affects the interaction

of proteins to RNA, and this binding is, at least in part, responsible for the function of

m6A.

Fig. 26: The writer, eraser and

reader proteins of m6A.

Mammalian m6A writers function as

a protein complex with three

components: METTL3, METTL14,

and WTAP. Two m6A erasers have

been reported: FTO and ALKBH5.

The function of m6A is mediated

partly by reader proteins, which

have been identified in members of

the YTH domain-containing protein

and the heterogeneous nuclear

ribonucleoprotein (HNRNP) protein

families (Adapted from Zhao et al.

2016).

1.3.4.1.3 Biological relevance of m6A

A real breakthrough was achieved in recent years with the identification of key

functions of m6A in various levels of regulation of gene expression, and a link with

biological processes such as stemness and metabolism. This is carried out, at least in part,

by the interaction with specific m6A binder proteins (Fig. 27a).

Emerging evidence points towards a mechanistic relationship between m6A and

splicing regulation (Fig. 27c). All the m6A writers (METTL3, METTL14 and WTAP)

and erasers (ALKBH5 and FTO) have been localized, at least partially, in nuclear

speckles, which are important structures for pre-mRNA processing (Maity & Das 2016).

Knockdowns of METTL3, WTAP, ALKBH5 or FTO were all shown to affect splicing in

various biological contexts. This is achieved by m6A-mediated regulation of the binding

of various factors involved in splicing, including HNRNPC, YTHDC1 and SRSF2 (Zhao

et al. 2014; Liu et al. 2015; Xiao et al. 2016).

Page 62: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

62

Several recent studies also suggest that m6A might play a role in translation,

generally promoting its efficiency (Fig. 27d). Two m6A binding proteins were suggested

to enhance translation: YTHDF1, which is known to interact with translation initiation

factors; and eIF3, which is a key translation initiation factor (X. Wang et al. 2015; Meyer

et al. 2015). Identification of additional m6A binding proteins might improve our

understanding of the mark’s association with translation.

Furthermore, m6A has been described to regulate RNA stability (Fig. 27e).

Studies in cells depleted for METTL3, METTL14 and FTO have in turn associated m6A

with increased or decreased levels of transcripts. Enhanced stability was rather observed

for mRNAs with methylated introns, for which loss of m6A might lead to improper

splicing and subsequent degradation of the transcript. In contrast, m6A-mediated binding

of YTHDF2 (a promoter of RNA degradation) and blocking of HuR binding (a known

mRNA stabilizer), was associated with decreased half-life and mRNA decay (X. Wang

et al. 2014).

Fig. 27: Mechanisms

and functions of m6A.

Several mechanisms

have been attributed to

m6A in relation with

protein binding, RNA

base pairing, splicing,

translation, stability and

degradation. (Adapted

from Meyer & Jaffrey

2014)

Page 63: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

63

In mammalian ES cells, m6A appears to regulate stemness and cell fate transition

by promoting the transition from naïve pluripotency towards differentiation. Studies have

suggested that loss of m6A led to an increase in pluripotent markers and difficulties to

differentiate (Batista et al. 2014; Geula et al. 2015). Accordingly, Mettl3-knockout

blastocysts displayed an inability to repress pluripotent genes and ES cells formed poorly

differentiated teratomas in vivo (Batista et al. 2014).

Regulation of m6A has also been linked to obesity and adipogenesis. Variants of

FTO have been associated with childhood obesity in genome-wide association studies,

hence the name of the protein (fat mass and obesity-associated protein) (Farooqi 2011).

Also, FTO was suggested to promote obesity in mice by promoting food intake and

adiposity while decreasing energy expenditure (Fischer et al. 2009; Church et al. 2010).

In line with this, FTO-mediated demethylation of the RUNX1T1 mRNA was associated

with splicing regulation of this adipogenic regulatory factor (Merkestein et al. 2015).

The role of m6A in diseases remains unclear to date. Nevertheless, the mark has

recently been implicated in cancer. In MLL-rearranged leukemia, FTO appears to

demethylate a set of transcripts, including key regulators ASB2 and RARA, decreasing

their stability and thus promoting leukemogenesis (Z. Li et al. 2017). Similarly, in

glioblastoma, reduced m6A levels promoted tumorigenesis. Overexpression of METTL3

or chemical inhibition of FTO suppressed the progression of stem cell-mediated tumor

through the regulation of key transcripts, including ADAM19 and FOXM1 (Cui et al.

2017; Zhang et al. 2017). Thus, while it is still early days, m6A-mediated regulation

appears to play an important role in cancer. Given the complexity of the disease, and the

variety of tissues in which it occurs, additional studies are required to better understand

its function in malignancies.

1.3.4.2 Methylation of cytosine

Just like DNA, RNA can be methylated on the 5’ carbon of a cytosine residue to

form 5-methylcytosine (here abbreviated as 5mrC, to avoid confusion with the DNA

modification 5mC). This modification was first studied in the context of tRNAs, however

emerging evidence also points out towards a role in the regulation of mRNAs.

Page 64: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

64

1.3.4.2.1 5mrC in non-coding RNAs

DNMT2, a member of the DNA methyltransferase family, has been shown to

methylate several tRNAs, including those mediating the addition of aspartate, valine and

glycine (Fig. 28) (Schaefer et al. 2009). This methylation event has been linked to

increased tRNA stability and protein synthesis. Another enzyme, NSUN2, was also

implicated in the methylation of tRNAs. In line with this idea, loss of both Dnmt2 and

Nsun2 in mice led to a drastic decrease in 5mrC, as well as a global decrease in protein

synthesis (Tuorto et al. 2012).

Aside from tRNAs, methylation also occurs on other non-coding RNAs (Fig. 28).

Notably, loss of Nsun2-mediated 5mrC was reported to cause aberrant processing of

small RNA fragments that can function as miRNAs (Hussain, Sajini, et al. 2013). It was

also reported that two well-known long non-coding RNAs (lncRNAs), HOTAIR and

XIST, display 5mrC around regulatory regions mediating interaction with protein

complexes (Amort et al. 2013). And in rRNA, 5mrC could be involved in translation

regulation and tRNA recognition (Chow et al. 2007).

1.3.4.2.2 5mrC in mRNAs

The existence of 5mrC in mRNA was known for decades (Dubin & Taylor 1975),

yet the focus of research remained on tRNA methylation for many years. However, the

study of mRNA methylation recently gained attention with the first mappings of

transcriptome-wide 5mrC distribution (Squires et al. 2012; Khoddami & Cairns 2013;

Edelheit et al. 2013). The mark was found relatively enriched in both 5’ and 3’UTR, and

slightly depleted in coding regions. Of note, the enrichment in the 3’ UTR and near

Argonaute binding sites suggested that 5mrC in mRNA could be associated with miRNA

degradation pathway, although this remained to be demonstrated (Fig. 28). The exact

function of 5mrC in mRNA is not yet clarified.

The identification of the enzymes responsible for mRNA methylation remains in

debate, as there are at least six potential RNA cytosine methyltransferases in mammals,

in addition to Dnmt2 and Nsun2. Based on sequence homology in the catalytic domain,

Nsun1 and Nsun3-7 are all predicted to methylate RNA, although there might be some

substrate specificity. In addition, Nsun1, 2 and 5 have been identified as mRNA-binding

proteins (Hussain, Aleksic, et al. 2013).

Page 65: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

65

Fig. 28: Functions of 5mrC. Cytosine methylation on tRNAs and non-coding RNAs have

been associated with cleavage, processing and stability of the RNA. The function of

methylation in mRNAs is still unclear but could be linked to miRNA-mediated interference

(Adapted from Blanco & Frye 2014)

1.3.4.2.3 Oxidation of 5mrC into 5hmrC

It was recently discovered that, similar to what occurs in DNA, TET enzymes can

mediate the formation of 5-hydroxymethylcytosine in RNA (here abbreviated as 5hmrC,

to avoid confusion with the DNA modification 5hmC) (L. Fu et al. 2014; Huber et al.

2015). The modification was significantly enriched in poly-adenylated RNA (polyA

RNA), which contains mostly mRNAs and lncRNAs. In terms of abundance, 5hmrC was

about 1000-fold less abundant than 5mrC in total RNA, although this ratio dropped to 25-

fold in polyA RNA. The presence of the mark specifically in polyA RNA suggest that

5hmrC could play a role in the regulation of mRNAs and/or lncRNAs.

1.3.4.3 Mapping RNA modifications

Several techniques have been recently developed to map RNA modifications

across the transcriptome. Most of these methods are adapted from DNA or histone

modification mapping techniques. They are based on chemical modifications of the

nucleotides, immunoprecipitation, or a combination of both.

The principle of the bisulfite (BS) treatment, which is the gold standard to identify

5mC in DNA, can also be used to map 5mrC in RNA: unmethylated cytosines are

chemically converted into thymidines while 5mrC are unchanged. This technique

displays a single-nucleotide resolution; however, the treatment is rather aggressive to

RNA and can lead to degradation. It has nevertheless been used in previous studies, in

combination with high-throughput sequencing (Squires et al. 2012; Amort et al. 2017).

Page 66: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

66

Perhaps the most common technique used to study the distribution of RNA

modifications is the immunoprecipitation of RNA (RIP) followed by high-throughput

sequencing. It can be used to detect both RNA modifications and RNA-protein

interactions. In the case of methylation (e.g. 5mrC or m6A), this method is called MeRIP,

and has been successfully used in a number of studies (Dominissini et al. 2012; Meyer et

al. 2012; Amort et al. 2017). As all immunoprecipitation-based methods, RIP techniques

rely on the availability of a specific antibody targeting the modification and its resolution

is limited to about 100-200bp.

Another fundamental approach to map RNA modification was derived from the

RIP method. This technique relies on the use of ultraviolet (UV) light to irreversibly

crosslink RNA to binding proteins (similar to the use of formaldehyde in ChIP), hence its

name, crosslinking immunoprecipitation (CLIP). The main advantage of the crosslinking

step is that it allows more stringent washes during the IP, thus CLIP is theoretically more

specific than the traditional RIP. The CLIP principle was further expanded into several

adapted methods, one of which, called m6A individual-nucleotide resolution using CLIP

(miCLIP), was specifically designed to map m6A. The principle is to induce a mutational

signature with an anti-m6A antibody and UV-induced antibody-RNA crosslinking,

followed by retro-transcription. Analyzing the specific mutational signature subsequently

allows to map m6A at the single nucleotide resolution (Linder et al. 2015).

Page 67: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

67

2. Breast cancers

2.1 Introduction

Breast cancer (BC) is a malignant disease that originates in the mammary gland.

It is a major health burden worldwide with about 1.7 million diagnoses in 2012 (WHO

2012). In women, it is the most common cancer (Fig. 29), and the second deadliest, just

after lung cancer (Siegel et al. 2016). Most BCs are diagnosed in patients above 50 years

of age. Approximately 1 out of 3 women will develop a breast cancer throughout their

life and 1 out of 8 will die overall. Progress in the comprehension of the disease and in

treatment opportunities have improved the survival rate in the past decades: the 10-year

age-standardized net survival for BC in women has increased from 40% from 1971 to

78% in 2011. Nevertheless, with around 450 000 deaths every year, BC remains a major

issue worldwide, and the survival rate varies greatly from one country to another. In

Europe and North America, up to 80% of women survive their breast cancer, whereas the

numbers drop below 40% in low-income countries (WHO 2012).

Fig. 29: Ten leading cancer types. Estimated new cancer cases by sex in the United

States, 2016 (Adapted from Siegel et al, 2016)

The etiology of BC is complex and not fully understood, with many potential risk

factors (Dumalaon-Canaria et al. 2014). Familial antecedents, particularly in young, first-

Page 68: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

68

degree relatives, increase significantly the risk of developing BC. Such cases have been

associated with hereditary mutations in tumor suppressor genes, such as the DNA repair

genes BRCA1 and BRCA2. However, so-called familial BC only accounts for less than

10% of all BC, whereas sporadic (non-familial) cancers represent the vast majority of

BCs. Factors associated with hormonal status have also been linked to higher risk of BC

(e.g. late first pregnancy, no pregnancy, no breast-feeding, late menopause, hormone

replacement therapy). Other factors could be attributed to lifestyle (e.g. obesity, alcohol

consumption, diet) or environment (e.g. exposure to pesticides or heavy metal cadmium).

Importantly, BC constitutes a very heterogeneous disease that could be seen as a

set of several malignant conditions whose common feature is essentially the tissue of

origin. In adult women, the breast contains glandular tissue, fibrous stroma, and fat tissue

(Fig. 30, left panel) (Jesinger 2014). The mammary gland, which is the functional part of

the breast, is constituted of 10 to 20 lobules, responsible for milk production, and a series

of ducts that collect and drain the milk towards the nipple. Within breast lobules, the

epithelium is formed of a single layer of luminal cells, surrounded by an underlying layer

of basal myoepithelial cells (Fig. 30, right panel). During breastfeeding, prolactin induces

luminal cells to secrete milk in the lumen, and oxytocin elicits the contraction of the

myoepithelial cells to eject the milk from the lobule.

Fig. 30: Anatomy of the breast. Schematic representation of the lobular and ductal

systems. (Adapted from http://humanbiologylab.pbworks.com/w/page/104941359

/Histology%20of%20the%20Mammary%20Gland)

Page 69: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

69

2.2 Diversity of breast cancers

The heterogeneity of BC is a huge hurdle in terms of patient care-taking, as they

can display very different biological features, as well as different clinical evolutions.

Hence, clinicians and biologists have long sought to classify BC into different subtypes,

in order to improve our understanding of the disease and therapy management. In this

section, we will present the main classifications used to categorize breast cancers.

2.2.1 Histopathological classifications

The first classification of BC, into the so-called histological types, relies on the

detection by pathologists of specific morphological and cytological features of the

tumors. BCs are first divided based on the extent of the disease: malignancies that are

contained within the tissue of origin, without any breach of the basal membrane, are called

in situ carcinoma, whereas those that extend beyond are called infiltrating or invasive

carcinoma. BCs are also distinguished by their origin, i.e. ductal versus lobular.

Histopathological observations further refine the classification into many histological

types (up to 21, depending on the taxonomy) (Dieci et al. 2014). Based on these criteria,

the most common type of BC is the invasive ductal carcinoma “not otherwise specified”

(IDC-NOS) or of “no special type” (IDC-NST), with up to 75% of all BC. In brief, this

IDC type represents the adenocarcinoma that do not display any particular feature that

would allow to distinguish them. Taken together, all the other “special” histological types

represent the remaining 25% of BC. Among them, the most common is the invasive

lobular carcinoma (ILC), with approximately 10% of BC.

Another classification commonly used by pathologists is the histological grade,

which is an assessment of the aggressiveness of the tumor, based on three morphological

features: tubule formation, mitotic count, and nuclear pleomorphism (Filho et al. 2011).

Each one is graded, and the combined score divides the tumors from grade 1 (poorly

proliferating and highly differentiated) to grade 3 (highly proliferative and poorly

differentiated). Unfortunately, almost half of the patients are classified into the

intermediate grade 2, which is not very informative for clinicians as patients of this group

display mixed phenotypes and survivals. However, an improved version of the

Page 70: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

70

histological grade, called the genomic grade index (GGI) allows to further distinguish

grade 2 patients into low and high risk categories, with outcomes similar to grade 1 and

3, respectively (Sotiriou et al. 2006). The GGI classification is based on a 97-gene

signature, focused on proliferation and cell cycle markers.

In routine, invasive BCs are often sorted based on the “tumor node metastasis”

(TNM) classification. This stratification is based on three criteria: the size of the tumor

(T), the infiltration of tumor cells in the armpit lymph nodes (N), and the metastasis status

(M). Based on the combined score, BCs are sorted into tumor stages I to IV, in increasing

order of aggressiveness (Fig. 31).

Fig. 31: TNM stage is a predictor of overall

survival. Patients with stage III and stage IV

BC have significantly worse overall survival

than patients with stage I and stage II breast

carcinoma (Adapted from Orucevic et al.

2015).

2.2.2 Molecular classifications

Extensive research has been pursued in order to identify molecular biomarkers

that might appropriately reflect the intrinsic diversity of BC, as well as to predict the

response to certain therapies. Nowadays, 4 standard biomarkers are evaluated by

immunohistochemistry (IHC) in routine. Two of them, the estrogen receptor (ER) and the

progesterone receptor (PR), identify a group of patients that could benefit from endocrine

therapies. Approximately 65% of all BCs fall in this category, and they are considered of

better prognosis overall. The third biomarker, the HER2 growth factor receptor, defines

a group that can be treated with the recently-developed HER2-targeted therapies. The

Erbb2 gene, coding for HER2, is amplified in about 10-20% of all BC and leads to an

oncogenic activation of the phosphatidylinositol-3-kinase (PI3K) signaling pathway.

Tumors that are negative for ER, PR and HER2 are commonly referred to as “triple

Page 71: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

71

negative” and do not benefit, as of today, of any specific therapy. The fourth biomarker,

Ki67, is expressed in all proliferative cells and thus reflects the rate of proliferation within

a tumor. Taken together, these 4 biomarkers help, to some extent, the clinicians in the

therapy decision-making.

With the recent progresses in transcriptome-wide technologies, many research

groups have turned to transcriptome-wide expression analyses in order to find an unbiased

approach to the topic of BC diversity. According to the “intrinsic” classification described

by Perou et al., BC can be classified into at least four subtypes: luminal A and B, HER2-

like (or “HER2-enriched”), and basal-like (Perou et al. 2000). Similar results were found

by numerous studies in the following years (Sorlie et al. 2001; Sorlie et al. 2003; Sotiriou

et al. 2003; Hu et al. 2006). Interestingly, there is a striking correspondence between these

four subtypes and the biomarkers routinely used by pathologists (Fig. 32). Luminal A

cancers (approximately 40-50% of all BC) are mainly ER-positive and/or PR-positive, of

low grade and low Ki67. These correspond to the subtype with the best overall survival.

Luminal B cancers (about 20-25% of all BC) are also ER-positive and/or PR-positive, but

of high grade and high Ki67. They are more aggressive than their luminal A counterpart.

HER2-like cancers (10-15% of all BC) are predominantly HER2-positive, hence the name

of this group. They are also often of high grade. Basal-like cancers (10-20% of all BC)

are mostly “triple negative” and of high grade. These last two subtypes correspond to

aggressive subtypes. A signature, commonly referred to as

“PAM50”, was developed based on the expression of 50

genes in order to classify BC into these 4 subtypes (Parker

et al. 2009; Wallden et al. 2015).

Fig. 32: Expression subtypes are associated with clinical

features. Tumor samples are grouped by mRNA subtype:

luminal A (n=225), luminal B (n=126), HER2-enriched

(n=57) and basal-like (n=93). Clinical data are shown as

follows: ER/PR/HER2 IHC status (dark grey: positive;

white: negative; light grey: N/A or equivocal); “T” tumor

size (dark grey: T2–4; white: T1; light grey: N/A or

equivocal) and “N” node status; (dark grey: positive;

white: negative; light grey: N/A or equivocal). (Adapted

from Koboldt et al. 2012).

Page 72: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

72

Additional gene expression-based subtypes have been described, although they

are rare and less clearly defined than the main four subtypes. The normal-like subtype (3-

6% of all BC) is, as its name suggests, the closest to normal breast in terms of gene

expression patterns. This group is also characterized by intermediate or good survival

(Weigelt et al. 2010). The claudin-low subtype (about 5% of all BC) represents a very

heterogeneous group, enriched in both triple negative and ER-positive tumors and it is

characterized by intermediate survival rates (Herschkowitz et al. 2007; Sabatier et al.

2014). Based on gene expression, the claudin-low subtype constitutes the most

undifferentiated tumors.

Further research has been performed in recent years in order to refine the

molecular classification of BC (Dedeurwaerder et al. 2011; Lehmann et al. 2011;

Netanely et al. 2016). Worth mentioning, a study performed on an extensive cohort

(nearly 2000 BC patients) suggested the existence of up to 10 BC subtypes with distinct

clinical outcomes, based on the integrative analysis of copy number and gene expression

data (Curtis et al. 2012). Some of the identified clusters could be associated with the

known expression subtypes. However, many clusters were a mix of different subtypes,

which indicates an even higher degree of complexity in BC. The Cancer Genome Atlas

(TCGA) consortium went one step further by comparing multiple levels of regulations in

a pan-cancer study. In the context of BC, they highlighted the existence of four main

subtypes by combining data from microRNAs, DNA methylation, copy number, PAM50

mRNA expression, and protein expression in a cohort of over 500 patients (Koboldt et al.

2012). Each level of gene regulation showed significant heterogeneity. Interestingly,

protein expression highlighted two potential novel subtypes that might be related to the

microenvironment of the tumor, thus illustrating that BC displays heterogeneity at many

levels and that genetics and RNA expression alone might not be sufficient to fully

understand the complexity of the disease.

2.3 The immune system: a double-edged sword

Aside from malignant cells, breast and other tumors contain various cell types,

which, together, are referred to as the tumor microenvironment (TME). The TME is a

major contributor to BC heterogeneity and its composition varies between tumor types,

Page 73: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

73

and stages of tumor development (Becht et al. 2016). The TME is recruited by cancer

cell-derived factors and includes immune cells, fibroblasts, adipocytes and endothelial

cells. In turn, TME cells influence the behavior of cancer cells with different outcomes in

terms of survival and response to therapy.

In this section, we will focus on the role of immune cells of the TME and their

clinical relevance in BC. We will then focus on a major regulator of immunity and

inflammation, the NF-κB family.

2.3.1 Immune infiltration and its clinical relevance

The role of the immune system is dual in cancer. On one hand, proper immune

reactions, such as immune surveillance, are required both for preventing tumor

development and for efficient anti-tumor response during treatment. On the other hand,

inadequate immune responses, such as chronic inflammation and immune escape

mechanisms, support the development and progression of cancer.

Immune cells are one of the major cell types found in the TME and they carry

important diagnostic information in cancer. But the immune reaction in cancer is very

complex, and different subtypes of immune cells have been associated with differential

prognostic and predictive values. Overall, the abundance of tumor-infiltrating

lymphocytes (TILs) is associated with a good clinical outcome in breast and other cancers

and is predictive of good response to chemotherapy (Aaltomaa et al. 1992; Dieci et al.

2015; Melichar et al. 2014). Nevertheless, distinct subtypes of lymphocytes can display

very different roles in BC. For instance, cytotoxic T cells (or CD8+ lymphocytes), which

are effector cells that can recognize and destroy cancer cells, have been associated with a

good outcome and response to therapy in many studies in BC (Mahmoud et al. 2011; S.

Liu et al. 2012; Seo et al. 2013). In contrast, high frequency of regulatory T cells , which

suppress immune surveillance, correlates with tumor grade and reduced patient survival

(Lança & Silva-Santos 2012). T-helper cells (or CD4+ lymphocytes) can act as both anti-

or pro-cancer agents, depending on their cytokine profile (e.g. TH1, TH2; TH17 profiles)

and their clinical relevance vary accordingly (Becht et al. 2016). The role of B

lymphocytes, which are mediators of humoral immunity, is in debate, studies having

shown contrasting results and attributing to B lymphocytes both supportive and

suppressive effects on tumor progression (Nelson 2010). Macrophages, like CD4+

Page 74: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

74

lymphocytes, display distinct effect based on their polarization: classically activated M1

macrophages and alternatively activated M2 macrophages are thought be anti- and pro-

tumorigenic in nature, respectively (Biswas & Mantovani 2010).

With such contrasting influences, the overall prognostic effect of the immune

microenvironment seems to be the balance between the different immune cell subtypes

(Fig. 33). For example, breast cancer patients with tumors characterized by a high

macrophage, high CD4+, but low CD8+ T-cell signature had a significantly decreased

recurrence-free survival than patients with tumors that showed a low macrophage, low

CD4+ and high CD8+ T-cell signature (DeNardo et al. 2011).

Fig. 33: Balance of the immune TME. The elimination of tumor cells involves CD4+

T-helper (TH1, TH2 and TH17) cells, CD8+ cytotoxic T cells, γδT cells, natural killer

(NK) cells, natural killer T (NKT) cells; M1 macrophages and dendritic cells. Escape

from the immune response include myeloid derived suppressor cells (MDSCs), regulatory

T (TReg) cells, TH17 cells and M2 macrophages. (Adapted from Kansara et al. 2014)

2.3.2 The NF-κB signaling pathway

Beyond tumor infiltration by immune cells, activation of immune and

inflammatory signaling pathways are also common events in cancer cells. Among them,

the Nuclear factor kappa B (NF-κB) is a major regulator, originally identified as a critical

transcription factor for the development, survival, and activation of leukocytes, including

B and T lymphocytes, as well as macrophages (Gerondakis & Siebenlist 2010). In

mammalian, NF-κB family is composed of five members, p65 (RelA), RelB, c-Rel, p50

(NF-κB1), and p52 (NF-κB2), which form dimers that can act as transcriptional activators

or repressors, directly binding to their target DNA sequence. In the canonical pathway

(Fig. 34), the p50-p65 dimer is kept inactivated in the cytoplasm by an inhibitor called

IκB (inhibitor of kappa B). Following activation of the pathway (e.g. upon binding of

Page 75: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

75

various cytokines to their receptor), IκB is targeted for

proteasomal degradation by IκB kinase (IKK)-mediated

phosphorylation. Once IκB is degraded, the NF-κB dimer

is free to translocate to the nucleus where it can bind to

DNA and recruit effectors to act on gene expression. In

cancer cells, activation of the NF-κB can occur upon

mutations in the pathway genes or be mediated by cytokine

release from various cells of the TME, including immune

cells (Lakshmi Narendra et al. 2013).

Fig. 34: Canonical NF-κB pathway. The IKK complex

targets the inhibitor of NF-κB (IκB) for proteasomal

degradation by phosphorylation. IκB degradation allows

the translocation of the NF-κB dimer p50-p65 into the

nucleus where it can bind to DNA and act on gene

expression. (Adapted from Gerondakis et al. 2014)

The role of NF-κB in cancer is complex. Generally considered as a pro-

tumorigenic factor, it is involved in cell survival, invasion, angiogenesis, metastasis and

chemoresistance. These effects are widely associated with the pro-inflammatory role of

NF-κB. This notion is supported by the observation that patients with chronic

inflammation have higher risks to develop cancer, including BC (Bhatelia et al. 2014).

Furthermore, NF-κB induces the release of cytokines (e.g. TNF, IL1, IL6 and IL8) which

lead to the recruitment of leukocytes to the TME. The following immune response, which

includes events such as the release reactive oxygen species by neutrophils, might cause

DNA-damage and mutations as a side effects (Schumacker 2015). Additionally, NF-κB

signaling was shown to enhance epithelial to mesenchymal transition (EMT) as wells as

vascularization of tumors in BC (Shibata et al. 2002; Huber et al. 2004). In contrast, NF-

κB activation is part of the immune response targeting malignant cells, and, notably, full

activation of NF-κB is accompanied by a high cytotoxic activity of immune cells against

cancer cells (Hoesel & Schmid 2013). Also, several reports have indicated that NF-κB

could also oppose cancer development by promoting the survival of non-cancerous cells

(Zhang et al. 2005; Maeda et al. 2005; Shibata et al. 2010). Overall, the diverse effects of

the NF-κB pathway are determined both by the mechanisms sustaining tumor induction

and the type of immune response involved.

Page 76: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

76

2.4 Management and treatments

In this section, we will summarize the current strategies in terms of breast cancer

management, from diagnosis to treatment. We will mention the main therapeutic options

and explain how they relate to the biology of BC.

2.4.1 Breast cancer diagnosis

BC can be discovered in a patient through several circumstances: screening by

mammography, palpation of the breast, changes in the appearance of the breast, or

unexpected leaking from a single nipple. The formal diagnosis is usually based on three

procedures: (i) the clinical examination of the breast and the neighboring lymph nodes by

a doctor; (ii) the radiological examination, such as X-ray mammography, ultrasound

examination or magnetic resonance imaging (MRI) of the breast and neighboring lymph

nodes; and (iii) the histopathological examination following a biopsy of the tumor. The

latter, which is performed with a needle, provides the first clues concerning the

characteristics of the cancer and help guide the clinicians (ESMO 2016).

2.4.2 General therapeutic options

Deciding the therapeutic plan requires an inter-disciplinary team of medical

professionals. The final decision takes into account the size of the tumor, its location, the

lymph node status, the stage of the tumor and its molecular features.

The surgical removal of the tumor is chosen in most cases, for non-invasive as

well as invasive cancers. Depending on the progression of the disease, the surgery might

remain local (lumpectomy) or require the full removal of the breast (mastectomy). Of

note, breast-conserving surgery and breast-reconstruction are now a valid option for many

patients and should be discussed with their doctor (ESMO 2016).

Unfortunately, BC can lead to the development of metastases, even after the

removal of the tumor. At the point of diagnosis, micrometastatic sites may already exist,

and BC should in fact be viewed and treated as a systemic disease rather than a localized

pathology (Ignatiadis et al. 2008). Thus, clinicians will often advise the use of additional

Page 77: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

77

therapies to prevent the development of metastases (Fig. 35). Such treatments may occur

before the surgery (in which case they are referred to as “neo-adjuvant” or “pre-

operative”) and/or after the surgery (in which case they are called “adjuvant”). These

therapies, which have helped reduce the mortality of BC over the past decades, include

radiotherapy, chemotherapy, endocrine (or hormone) therapy and antibody-based

targeted therapies. Radiotherapy relies on the use of radiation in order to destroy rapidly-

proliferative cancer cells. Importantly, it is recommended as an adjuvant therapy for the

vast majority of invasive BC. The recommended radiation dose is 45-50 Grays, divided

in about 25 sessions in order to lower the damage to surrounding tissues and to better

control the tumor over a long period of time. Aside from the local treatments (i.e. surgery

and radiation), clinicians choose the most appropriate options in terms of systemic

treatment, based on the patient’s clinical features. These options are explained in the next

section.

Fig. 35: Standard treatments for BC. The most common combination for BC treatment

is the combination of surgery and radiation to remove and control locally the tumor.

Systemic treatments (hormone therapy, chemotherapy or targeted therapy) are often used

as additional (neo)adjuvant therapies to prevent the transition into a metastatic disease.

(Adapted from http://www.puhuahospital.com/treatments/cancer/breast)

2.4.3 Systemic therapies

Nowadays, chemotherapy generally consists in the combination of several anti-

cancer agents. Given the toxicity of the treatment, chemotherapy is given in 4-8 cycles,

with a resting period in between cycles. Several combinations of drugs can be used, but

generally at least one drug of the anthracycline family is advised (e.g. doxorubicin or

epirubicin). The anthracyclines prevent DNA replication and are among the most efficient

anti-cancer drugs available, however they can cause cardiotoxicity. Other drugs

commonly used for chemotherapy include cyclophosphamides (alkylant agents blocking

Page 78: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

78

DNA replication), taxanes (blockers of microtubule formation, which is essential for cell

division) and carboplatines (platinium-based blockers of DNA replication). Despite the

inevitable side effects, chemotherapy remains a relevant therapeutic option both as a

(neo)adjuvant therapy (when local surgery is performed) and in the case of advanced BC

(when the disease has already spread beyond the breast area).

Endocrine or hormone therapy is a first-line treatment for ER and/or PR-positive

tumors (den Hollander et al. 2013). A combination of several treatments can be used:

tamoxifen, aromatase inhibitors, gonadotropin-releasing hormone analogues, or the

removal of the ovaries (ovariectomy). The choice of treatment is notably based on the

menopausal status of the patient. Although endocrine therapy is often used in combination

with chemotherapy, patient at low risk of recurrence might benefit from endocrine therapy

alone, and thus avoid chemotherapy.

HER2-targeted therapy has become an efficient treatment for HER2-positive

tumors. The most common drug, trastuzumab (also called Herceptin), relies on an anti-

HER2 antibody to kill the tumor cells (Li & Li 2013). This adjuvant treatment is used in

combination with chemotherapy and has helped improve the survival of HER2-positive

patients, who previously suffered from poor survival. The effects of the treatment appear

to be due to the blockade of the HER2 signaling pathway, as well as an increase in HER2

antigen presentation, which promotes the recognition of tumor cells by cytotoxic T

lymphocytes.

Recently, we have witnessed the emergence of a major therapeutic innovation in

oncology: the immune therapies based on checkpoint blockade. Briefly, antibodies target

the immune checkpoints responsible for preventing the activation of the immune system,

including CTLA-4 or the PD-1/PD-L1 pathway (McArthur 2016). These strategies allow

a tumor-specific immune response to be elicited and, taking into account immune

memory, durable antitumor response and cure may be achieved. The checkpoint

inhibitors have shown exciting promising results in the context of metastatic melanoma

first, then other types of solid tumors (Yang et al. 2007; Hodi et al. 2010; Robert et al.

2011; Dany et al. 2016). Preliminary results from clinical trials indicated that some BC

might respond to immunotherapy, particularly triple negative and basal-like tumors which

are heavily infiltrated by immune cells (McArthur 2016). Several clinical trials are

Page 79: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

79

ongoing to evaluate how to best benefit from immune therapies, which have generated

tremendous hope in the field of oncology.

With many possible combinations, the choice of the best strategy often remains

difficult, even when taking into consideration all the clinical data available. Much

research on BC is currently focused on optimizing therapeutic strategies based on the

combination of various agents, the doses and/or the timing. And given the heterogeneity

of BC, the main issue in the management of the disease, beyond the development of new

drugs, might be identifying which patients will benefit the most from which combination

of treatments.

2.5 Breast cancer and epigenetics

As all cancers, BC notoriously involves numerous genetic alterations.

Interestingly, a link can be established with the intrinsic heterogeneity of the tumors: the

luminal A subtype is enriched in PIK3CA mutations (about 45% of cases) whereas the

basal-like subtype is highly associated with TP53 mutations (about 80% of cases)

(Koboldt et al. 2012). Nevertheless, cancer is not just a genetic disease, but also an

epigenetic disease. Many epigenetic machineries are altered during carcinogenesis and

these alterations actively contribute to the progression of the cancer. In this section, we

will review the main epigenetic alterations observed in BC, their implication in the

development of the disease and their clinical relevance.

2.5.1 Epigenetics alterations in breast cancers

The first part of this chapter describes the epigenetic alterations commonly

observed in BC, with a particular focus on DNA modifications.

2.5.1.1 Alterations of DNA modifications in BC

DNA methylation is a hallmark of all cancers, including BC. The cancer genome

displays global hypomethylation concomitantly to local hypermethylation (see also

section 1.3.1.1.4 Biological and pathological relevance of 5mC). The observed global

Page 80: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

80

hypomethylation is, at least in part, caused by a loss of methylation in pericentromeric

regions and repeated sequences. This leads to the activation of previously silenced

retrotransposon-like regions such as LINE-1 and the satellite DNA sequence SATR-1,

and consequently to genomic recombination and chromosomal instability (Costa et al.

2006; Van Hoesel et al. 2012). Loss of 5mC occurs mainly in gene-poor regions, and thus

only a limited number of genes have been reported as hypomethylated (Paredes et al.

2005; Pakneshan et al. 2005; Ito et al. 2008; Sharma et al. 2010). By contrast, many gene

promoters have been reported as hypermethylated in BC. These genes are classically

involved in the regulation of proliferation, apoptosis, DNA repair, metastasis or

angiogenesis, and their silencing through promoter methylation furthers tumorigenesis.

Examples include the cell-cycle regulator p16ink4A, the DNA repair genes MGMT and

MLH1, and the tumor suppressor genes (TSG) RAR-b, RASSF1A and PTEN (Murata et

al. 2005; Park et al. 2011; Fumagalli et al. 2012; Wang et al. 2012). Intra- and intergenic

regions have also been reported as aberrantly methylated in breast cancer, however, their

numbers are small because of the promoter-centric approach that was used for many

years, and they are expected to increase in future as genome-wide technology is ever more

utilized (Shann et al. 2008; Lu et al. 2012; Shetty et al. 2011). Intriguingly, these

alterations in 5mC occurs even without any changes of the DNMTs.

DNA methylation has been extensively studied on a genome-wide level in the

context of BC subtypes. Striking differences in 5mC profiles have been observed between

ER-positive and ER-negative tumors (Fig. 36) (Li et al. 2010). Beyond the ER status,

DNA methylation profiling seems to segregate the different molecular BC subtypes. In

particular, basal-like BC tend to display a differential methylation profile, in both

promoters and other genomic regions (Holm et al. 2010). These results suggest that

distinct subtypes of BC, as defined by gene expression, may display specific alterations

of the methylome. Nevertheless, the two levels of analysis do not overlap completely and

DNA methylation brings additional information to transcriptome data. For example,

genome-wide DNA methylation analyses of 802 breast tumors by the TCGA network

revealed five distinct subgroups, only two of which coincided with known expression

subtypes (luminal B and basal-like); the other three were unknown (Koboldt et al. 2012).

Importantly, the robustness of results obtained by various groups was recently evaluated

in a systematic review of 22 genome-wide methylation studies, and many genes and

Page 81: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

81

pathways were found to be commonly affected across all studies (Day & Bianco-Miotto

2013).

Fig. 36: DNA methylation changes are associated with ER status in BC. BC samples

were stratified through hierarchical cluster analysis. ER status is a main discriminator

of DNA methylation-based clusters in BC. (Adapted from Dedeurwaerder et al. 2011)

As previously mentioned (section 1.3.1.2.4 Biological and pathological relevance

of 5hmC), DNA hydroxymethylation is also widely affected in cancer. As most

malignancies, BC displays a global loss of 5hmC (Fig. 37) (Haffner et al. 2011). No

genome-wide 5hmC mapping has been reported in human biopsies so far, therefore the

local changes of 5hmC in BC are yet unknown. Given (i) the vast changes of 5mC in BC,

and (ii) the close link between 5mC and 5hmC, it is expected that many 5hmC alterations

could be found in BC as well. How those changes relate to anomalies in 5mC and gene

expression also remains to be explored. The question was partially answered in a study

in which 5hmC changes were mapped in the MMTV-PyMT transgenic mice (which

spontaneously develop breast tumors) with or without exogenous expression of sFlk1, a

protein that enhances tumor vessel pruning and hypoxia. The authors observed a global

loss of 5hmC in hypoxic tumors occurring mainly in gene-rich regions. Promoter

hypermethylation of a subset of tumor suppressor genes (TSG) was also found under

hypoxic conditions, suggesting a link between loss of 5hmC loss and gain of 5mC in this

model (Thienpont et al. 2016). Whether this relation stands true in human breast tissues

remains to be demonstrated.

Fig. 37: Loss of 5hmC

in BC, measured by

IHC. (Adapted from

Haffner et al. 2011)

Page 82: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

82

Unlike hematological malignancies, no recurrent mutation was observed for TET

genes in the breast. Nevertheless, all TETs display decreased expression, particularly

TET1. (L. Yang et al. 2015). Loss of TET1 has been linked to hypermethylation of its

promoter and has also been reported as a marker for metastasis in BC (Sang et al. 2015).

Worth mentioning, although TET enzymes are almost never found mutated in solid

cancers, including BC, 5hmC readers show higher mutational frequencies (Scourzic et al.

2015). Importantly, TET1 was mostly described as a tumor suppressor and its reduced

expression appears to promote migration, proliferation, tumor growth and metastasis.

Downstream targets of TET1 include tissue inhibitors of metalloproteinase (TIMP) and

HOX genes (Fig. 38) (Hsu et al. 2012; Sun et al. 2013).

Fig. 38: TET1 act as a TSG in

BC. Activation of Tet1 expression,

or its downstream target HOXA9,

by doxocycline (+DOX) in an

inducible breast cancer model

decreases tumor growth and

tumorigenesis. (Adapted from Sun

et al. 2013)

Worth mentioning, two studies have explored the effect of hypoxia, a condition

commonly associated with cancer, on TETs in the breast with similar results. In the first,

hypoxia was found to induce TET1 and TET3 upregulation, leading to activation of the

TNF-p38-MAPK signaling pathway and tumor promotion (Wu et al. 2015). In the second

study, a modest increase in TET expression was also observed under hypoxic conditions,

contrasting with a global loss of 5hmC in 2 out of 3 breast cell lines. The authors

demonstrated that decreased oxygen levels directly affected the oxidative activity of TET

enzymes and they suggested that increased TET expression might be an attempt of

compensation mechanism (Thienpont et al. 2016). Therefore, beyond TET expression,

loss of activity is another mechanism that can affect 5hmC levels in BC.

2.5.1.2 Alterations of histone modifications in BC

Aberrant histone modification patterns, as well as changes in their respective

machinery, have been linked to genomic instability and repression of tumor suppressor

genes in BC. For instance, H3K4 demethylase protein LSD1 is overexpressed in ER-

Page 83: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

83

negative tumors and is considered as a predictive marker for aggressive BC (Nagasawa

et al. 2015). Another example is the Polycomb2 complex member and H3K27

methyltransferase EZH2, which is overexpressed in HER2-like and basal-like tumors

(Holm et al. 2012). While the effects of aberrant histone modifications vary widely based

on the mark, it is now well-established that these changes contribute to BC development

and progression (Pitta & Constantinou 2015).

2.5.1.3 Alterations of chromatin remodeling in BC

As most cancer types, genes related to chromatin remodeling are found

dysregulated in BC. Loss-of-function mutations are associated with several members of

the SWI/SNF complexes, including PBRM1, ARID1A1, SMARCA2 and BRD7 (Helming

et al. 2014). In addition, post-translational modification of SWI/SNF subunits resulting

in mistargeting of the complex is another mechanism of chromatin remodeling

dysregulation in BC. In that regard it was demonstrated that BAF155 protein (encoded

by the SMARCC1 gene) can be aberrantly methylated by the arginine methyltransferases

CARM1. This event, which promotes the expression of genes involved in glycogen

metabolism, enhances breast tumor progression and metastasis (Stefansson & Esteller

2014).

2.5.1.4 Alterations of RNA modifications in BC

Research on RNA modifications is still in its early days, particularly in the context

of breast cancer. In 2016, a study reported that BC cells upregulate m6A demethylase

ALKBH5 in hypoxic conditions, subsequently leading to m6A-dependent regulation of

NANOG and an enrichment in cancer stem cells (Zhang et al. 2016). Of note, genetic

variants of FTO, another m6A demethylase, have been associated with increased risk of

BC in genome-wide association studies (GWAS) (Kaklamani et al. 2011). Given that

obesity is a known risk factor for BC and that FTO was originally identified for its role

in obesity, it was previously thought that the association between FTO variants and BC

was mostly due to obesity. However, the recent identification of FTO as an m6A

demethylase has challenged this notion (Jia et al. 2011). Importantly, no transcriptome-

wide study of any RNA modification in BC has yet been reported, but given the recent

Page 84: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

84

interest in this topic, it is expected that many groups will tackle this issue in the years to

come.

2.5.2 Clinical relevance of epigenetics

In this section, we will explore the relevance of epigenetics to clinical oncology

regarding BC, following two main topics: (i) the potential of epigenetic marks as new

diagnostic, prognostic and predictive biomarkers, and (ii) the development of drugs

targeting epigenetic modifications as new tools for cancer therapy.

2.5.2.1 Epigenetics modifications as biomarkers for cancer

As explained in previous chapters, epigenetic alterations are massively involved

in cancer development. Next to genetic mutations, which have long been used as

biomarkers in cancer, the so-called “epimutations” (i.e. alterations in genes involving

epigenetic mechanisms) are increasingly recognized as relevant markers as well. DNA

modifications, in particular, display several key features that make them useful as cancer

biomarkers, although other epigenetic modifications remain relevant. First, DNA,

whether unmodified or bearing epigenetic modifications, is a stable molecule, particularly

in comparison to RNA, allowing its detection in liquid biopsies (Warton et al. 2016).

Secondly, epigenetic alterations are common in cancer and are thought to occur early in

the disease (Dawson & Kouzarides 2012). Therefore, they could be used for the early

detection of malignancies. Finally, epigenetic biomarkers allow to detect, not only

changes in the cancer cells themselves, but also changes in the tissue composition

(Dedeurwaerder et al. 2011; Sehouli et al. 2011). This is of great interest, because cells

of the tumor microenvironment, such as immune cells, fibroblasts and adipocytes,

contribute to the cancer phenotype and bear a predictive and prognostic value (see further

below) (Oble et al. 2009; Dieci et al. 2015; Wolfson et al. 2015; Kuzet & Gaggioli 2016).

Detecting specific DNA methylation marks in liquid biopsies could in fact be

useful for cancer diagnostics, either for early detection or post-treatment monitoring. In

the context of BC, relevant target genes from blood biopsy include APC, GSTP1, RAR-β,

RASSF1A, SFN, P16, MLH1, HOXD13, PCDHGB7 and DAPK1 (Hoque et al. 2006;

Page 85: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

85

Ahmed et al. 2010; Shan et al. 2016). Overall, quantification of DNA methylation in

bodily fluids is a very attractive method of cancer diagnosis in clinics, given its

noninvasive nature.

Given the heterogeneity of BC, much attention has been directed to the

identification of prognostic markers, i.e. biomarkers that can categorize patients based on

their survival probability, in order to shepherd therapy strategy and reach the best clinical

risk-benefit ratio. Currently, risk evaluation is mainly based on immunohistological

features, mutation signatures and gene-expression analysis. Epigenetic biomarkers

represent an additional tool that can complement the classical methods used in routine.

DNA methylation signatures have been widely showed to refine our ability to predict

patient outcome, both in tumor and blood biopsy (Müller et al. 2003; Müller et al. 2004;

Dedeurwaerder et al. 2011; Severi et al. 2014). Overall, hypermethylation within

functional promoters was associated with an increased risk, whereas hypermethylation of

genomic regions outside promoters was associated with decreased risk (Severi et al.

2014). Methylated genes associated high-risk included APC and RASSF1A (Fig. 39)

(Müller et al. 2004). Likewise, changes in histone modifications could also serve as

prognostic biomarker. For instance, high global levels of histone acetylation and

methylation (H3K4me2 and H4K20me3) were associated with good outcome and were

found almost exclusively in luminal-like breast tumors.

Fig. 39: Example of epigenetic prognostic

marker. Methylation of the APC gene is

associated with poor survival in BC patients

(Adapted from Müller et al. 2003)

A major challenge of modern oncology lies in the selection of the optimal

treatment, as the effectiveness of a given therapy can vary greatly from one patient to

another. Hence, clinicians have started to rely on the use of predictive markers, i.e.

biomarkers that can predict a patient’s response to treatment. Once more, epigenetic

markers are of interest in this context. For instance, hypermethylation of the DNA-repair

genes BRCA1 and BRCA2 predicts sensitivity to adjuvant chemotherapy and/or

Page 86: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

86

poly(ADP-ribose) polymerase (PARP) inhibitors (Veeck et al. 2010; Xu et al. 2013).

Other hypermethylated genes identified as predictive markers for adjuvant therapies

include TP53, PITX2 an CDO1 (Harbeck et al. 2008; Dietrich et al. 2010; Foedermayr et

al. 2014).

Furthermore, DNA methylation is increasingly recognized as a tool to investigate

the tissue composition of a tumor. Above all, tumor-infiltrating lymphocytes (TILs) are

associated with better survival and better response to treatment in breast and other cancers

(Oble et al. 2009; Melichar et al. 2014; Dieci et al. 2015). Interestingly, several studies

have highlighted the ability of DNA methylation to discriminate and accurately quantify

the various immune populations (Fig. 40) (Sehouli et al. 2011; Accomando et al. 2014;

Dedeurwaerder & Fuks 2012). This is notably due to the more linear relation between

DNA molecules and the numbers of cells (i.e. 2 copies of DNA per cell in diploid

organisms) as compared to RNA, as well as the high tissue-specificity of 5mC profiles

(Hackl et al. 2016). Recent work from our host laboratory has demonstrated the

prognostic value of a DNA methylation-based signature (called “MeTIL”) by measuring

TIL infiltration in breast and other cancers. Moreover, the MeTIL signature also predicted

response to chemotherapy independently of other clinical and pathological variables.

Fig. 40: DNA methylation as a tool to quantify immune infiltration. The

immunophenotype, in particular the composition of TILs in the tumor tissue, can be

estimated based on DNA methylation profiles using computational tools, such as

establishing signature scores and/or deconvolution of signal. (Adapted from Hackl et al.

2016)

In conclusion, DNA methylation is by far the epigenetic modification that is the

closest to a “bench to bedside” transition, while histone modifications have also shown

some potential in terms of biomarkers. Recent studies have suggested that other

epigenetic modifications (e.g. DNA hydroxymethylation and RNA methylation) could

Page 87: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

87

also be detected in liquid biopsies and have potential as new biomarkers (Godderis et al.

2015; Huang et al. 2016; Gilat et al. 2017). Unlike DNA methylation, the use of such

epigenetic modifications in routine oncology practice is still limited by the greater amount

of material required for their analysis. However, it is expected that technologies will

improve in the future and make them accessible to biomarker investigation.

2.5.2.2 Epigenetic therapy of cancer

Beyond their role as biomarkers, epigenetic modifications also represent an

appealing target for cancer therapy because their alterations are involved in the

progression of the disease and because they are reversible. Thus, in the past two decades,

much effort has been aimed at the development of drugs targeting epigenetic enzymes.

These drugs can either inhibit broadly the establishment of an epigenetic mark and lead

to a global loss of the modification, or they can target a specific enzyme or isoform for

more selective effects.

Among epigenetic drugs, DNA methylation inhibitors were the first to be used in

oncology (Fig. 41). These inhibitors, e.g. 5-Azacytidine (5-Aza-CR; azacytidine) and 5-

Aza-2’-deoxycytidine (5-Aza-CdR; decitabine), consist in nucleoside analogues that are

incorporated in the DNA but cannot be methylated, thus leading to 5mC depletion during

DNA replication. At low dose, DNA methylation inhibitors block proliferation in tumor

cells by reactivating previously hypermethylated tumor suppressor genes (TSG) (Phan et

al. 2016). At higher doses, they also induce cytotoxicity, due to the covalent binding of

DNMTs to DNA (Oka et al. 2005). Although these drugs do not specifically target tumor

cells, they appear to affect more efficiently rapidly-growing cells, which might explain

their relatively low side effects to healthy tissues. Following promising results in the

treatment of leukemia (Silverman et al. 2002), 5-Aza-CR and 5-Aza-CdR were approved

by the U.S. Food and Drug Administration (FDA) in 2004 and 2006, respectively, for

treatment of myelodysplastic syndromes.

Page 88: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

88

Fig. 41: Activation of TSG by

epigenetic drugs. In cancer

cells, DNA hypermethylation

and histone hypoacetylation act

to promote silencing of TSGs.

Treatment with DNMT and

HDAC inhibitors (DNMTi and

HDACi, respectively) reverse

these epigenetic alterations,

open chromatin structure and

increase TSG gene expression.

Restoration of normal TSG

functions promotes anti-cancer

properties of epigenetic drugs.

(Adapted from Seidel et al.

2012)

The second main category of epigenetic drugs in oncology consists of HDAC

inhibitors (Fig. 41). Increasing histone acetylation with HDAC inhibitors has indeed

shown promising results by blocking cell growth and inducing apoptosis. As for DNA

methylation inhibitors, their action was shown to be linked to the reactivation of TSG.

Many HDAC inhibitors have been developed over the years, with the first-generation

inhibitors showing a broad specificity for HDACs and the second-generation inhibitors a

greater intrinsic selectivity for their molecular targets (Valdespino & Valdespino 2015).

Currently, four drugs have been approved by the FDA for cancer therapy: Vorinostat

(SAHA), Romidepsin (FK-228), Belinostat (PXD-101) and Panobinostat (LBH-589).

Aside from directly targeting enzymes, epigenetic drugs can also target readers of

specific modifications. For instance, the “BET inhibitors” (iBETs) bind the

bromodomains of BET proteins BRD2, BRD3, BRD4, and BRDT. In doing so they block

the interaction between BET proteins and acetylated histones, and thus affect the

recruitment transcription factors and other chromatin modulators (Shi & Vakoc 2014).

These drugs have shown promising results as anti-cancer therapies in preclinical models.

For instance, BRD4 was implicated in the development of epithelial cancers, which might

explain the therapeutic effects of iBETs observed in xenograft models of breast cancer

(Shi et al. 2014). Studies are currently being conducted in order to evaluate the clinical

benefit associated with such therapies. Worth mentioning, iBETs are also known as potent

anti-inflammatory agents, which has been linked to the critical role of Brd2, Brd3, and

Page 89: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

89

Brd4 in the induction of inflammatory gene transcription, notably in macrophages and T

lymphocytes (Biswas & Mantovani 2010; Bandukwala et al. 2012).

Until recently, SWI/SNF complexes were thought to be inadequate targets for

inhibition in cancer therapy, because they mostly display tumor suppressive functions.

However, new potential therapeutic strategies have been developed on the basis of

synthetic lethality (Schiaffino-Ortega et al. 2014). Briefly, this concept relies on the

observation that mutations in SWI/SNF subunits, while potentially promoting cancer,

tend to also render the cancer cells more sensitive to repression of additional SWI/SNF

subunits. For example, studies suggested that the inhibition of SMARCA2 might have a

therapeutic benefit in SMARCA4 mutant tumors (Oike et al. 2013; Hoffman et al. 2014).

Therefore, targeting SWI/SNF proteins could be a new field for anti-cancer drug

discovery. Worth mentioning, the SWI/SNF complexes include several proteins with

bromodomains, such as SMARCA4, SMARCA2, BRD9, and PBRM1, that are targetable

with specific iBETs (Schiaffino-Ortega et al. 2014).

In the context of BC, epigenetic drugs, especially demethylating agents and

HDAC inhibitors, are frequently used in combination with other drugs because they can

enhance the susceptibility to other anti-cancer agents such as chemotherapeutic drugs (S.

Y. Li et al. 2015; Phan et al. 2016; Manal et al. 2016). Synergistic interaction can be of

particular interest in cancer therapy, as it allows lowering of doses and thus reduction of

side effects. In BC cells, epigenetic drugs have been shown to increase the anti-tumor

effects of several chemotherapeutic agents (including paclitaxel, doxorubicin, and 5-

fluorouracil), as well as tamoxifen and trastuzumab (i.e. the most common hormone and

HER2-targeting agents, respectively) (Sharma et al. 2006; Mirza et al. 2010; Huang et al.

2011). Phase I-III trials are ongoing to evaluate the clinical benefit of epigenetic drugs

used in combination with conventional treatments (C. et al. 2015; Jones et al. 2016;

Connolly et al. 2017).

Another potential benefit of epigenetic drugs is their ability to reverse cancer

immune evasion (Fig. 42) through increased expression of tumor surface antigens and

major histocompatibility complex (MHC) molecules (Roulois et al. 2015; Chiappinelli et

al. 2015). In particular, 5-Azacytidine has been identified as an immunomodulator in

cancer. Although clinical benefit has yet to be proven, this effect could potentially be

exploited as strategy to enhance the efficiency of immune therapies in breast and other

Page 90: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

90

cancers. In addition, iBETs appear to reduce the inflammatory pro-tumor effects of the

TME by suppressing the production of nitric oxide and a variety of inflammatory

cytokines (Leal et al. 2017). It is likely that the immunomodulatory effects of epigenetic

drugs will progressively be integrated in cancer management, in combination of either

conventional or immune-targeted therapies.

Fig. 42: Epigenetic drugs in cancer therapy. Normal cells can acquire both genetic and

epigenetic alterations during the tumorigenesis development. The interplay between these

two processes gives rise to alterations in the chromatin. Further selection can take place

following host immune surveillance or chemotherapy and, once again, the availability of

both genetic and epigenetic pathways can rapidly speed up the emergence of resistance.

Epigenetic therapy has the potential to reverse epigenetic abnormalities, thus restoring

sensitivity to treatment (Adapted from Jones et al. 2016).

The extension of the epigenetic repertoire in recent years has also broadened the

possibilities in terms of epigenetic drugs. In particular, given that TET enzymes have been

described as TSG in many cancers, including BC, it has been suggested that reactivation

of TETs in cancer might also be beneficial in cancer therapy. As of today, there is no

specific activator of TETs/5hmC available, however vitamin C has been identified as a

positive cofactor of TET enzymes that is involved in DNA demethylation (Blaschke et

al. 2013; Minor et al. 2013; Sasidharan Nair et al. 2016). This is of particular interest,

because meta-analysis suggests that vitamin C supplementation and/or dietary intake

could be associated with better survival in BC (Harris et al. 2014). Vitamin C is a broad

factor that act on many enzymes besides TETs and has far-reaching effects. Nevertheless,

it seems likely that some of the beneficial effects could be mediated by increased 5hmC

levels in cancer cells.

In regard to RNA modifications, regulators of related enzymes are just starting to

appear on the market (Y. Huang et al. 2015; T. Wang et al. 2015). A recent study reported

that inhibition of the m6A demethylase FTO with meclofenamic acid suppressed glioma

Page 91: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

91

progression and prolonged lifespan of xenografted mice, opening new applications for

RNA epigenetics (Cui et al. 2017). Furthermore, RNA-based therapeutics have recently

been gaining attention, as utilizing siRNA and mRNA in a therapeutic context is now

possible. In that regard, understanding how chemical modifications of RNA can either

improve or hinder the potency of RNAs is essential (Kaczmarek et al. 2017). However,

the field of epitranscriptomics is still in its early days, and much further research will be

required to determinate the potential of RNA modification in terms of clinical oncology.

In conclusion, epigenetic drugs have shown promising results as anti-cancer

agents. In particular, DNMT and HDAC inhibitors have already displayed benefits for

cancer patients, notably when acting in synergy with other agents. This notion is

supported by ongoing clinical trials, and seems, at least in part, associated with their

immunomodulator effects. Likewise, iBETs have recently shown promising results in

cancer therapy, with potent anti-inflammatory properties. And given the identification of

so-called “new” epigenetic modifications, the field of cancer epigenetics will probably

increase further in the next few years.

Page 92: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

92

Page 93: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

93

3. Aims of the project

As described in the introduction, the improvement in sequencing technologies and

the characterization of new enzymes has deeply changed our view of epigenetics by

unraveling the importance of the growing epigenetic repertoire in many biological

processes, as well as their implication in diseases such as cancer. This scientific revolution

allows a better understanding of the molecular mechanisms regulating the cells of an

organism. In the long run, exploring the “new” epigenetic modifications should improve

cancer management by providing potential novel biomarkers and therapeutic targets. Yet,

the study of these modifications is still in its early days and has not yet reached its full

potential. Given the high incidence of breast cancer, we have decided to mainly focus on

this disease as a model to explore the role of such epigenetic modifications in human

pathologies.

More specifically, our project was divided in three parts, each centered around a

different epigenetic modification. First, we sought to explore the regulation of TET1, an

enzyme responsible for DNA hydroxymethylation (5hmC) and already known to be

downregulated in many cancers, including breast. Secondly, we wanted to provide the

first mapping of RNA hydroxymethylation (5hmrC), a modification just recently

identified. We aimed to conduct a first fundamental study in Drosophila, then to extend

our data by exploring potential 5hmrC dysregulations in breast cancer. Finally, we wanted

to investigate the potential implication of RNA methylation (m6A), the most frequent

modification of mRNAs, in breast cancer.

The general goal of this project was to improve our understanding of how

recently-discovered epigenetic modifications are altered in cancer and provide evidence

that they constitute an important and novel level of dysregulation implicated in

carcinogenesis. Overall, we anticipated that our results could contribute to a better

knowledge of the role of epigenetics in health and disease.

Page 94: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

94

Page 95: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

95

Results

Results

Page 96: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

96

Page 97: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

97

1. Immune activation of NF-κB drives TET1

dysregulation in cancer

1.1. Introduction

In the past decades, changes in the epigenome have been increasingly recognized

as intrinsically linked with tumorigenesis and cancer progression (Dawson & Kouzarides

2012). It is now well-accepted that both DNA methylation and histone modifications, the

two major epigenetic modifications, are vastly affected in cancer and involved in the

disease. Yet, the recent expansion of the epigenetic repertoire has brought new questions

and challenges to the field of cancer epigenetics. Alterations of TET enzymes and DNA

hydroxymethylation (5hmC) have been identified as a hallmark of cancers and have

gained much attention in recent years. Loss of TETs and 5hmC have been associated with

cancer progression and metastases in many cancers, including breast cancer (Haffner et

al. 2011; Hsu et al. 2012). Hence, understanding TET dysregulation represents a key

challenge for the future in cancer epigenetics.

In our quest for mechanisms of TET dysregulation in breast cancer, we came upon

an unprecedented link with the immune system and, more specifically, the NF-κB

pathway. This was of tremendous interest to us, because the immune system has emerged

as a major feature of cancer in recent years with a dual effect: on one hand, secretion of

pro-inflammatory factors by immune cells are involved in tumor progression and

resistance to treatment; on the other hand, the immune system is heavily involved in the

anti-tumor response. The latter notion is supported by the fact that tumor immune

response, and in particular tumor-infiltrating lymphocytes (TILs), are increasingly

recognized to be associated with better clinical outcome. In addition, the recent

emergence of immune checkpoint inhibitors as a promising method to treat cancer has

brought a new hope in terms of cancer treatment, as demonstrated by ongoing clinical

trials. In regard to that, our findings are of great interest because epigenetic drugs have

been shown to modulate the antitumor immune response and dissecting the epigenetic

mechanisms underlying the cross-talk between the immune system and cancer could help

optimize therapeutic strategies.

Page 98: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

98

In this study, we started by exploring epigenetic dysregulations in breast cancer

and we demonstrated for the first time that TET1 repression, and its related alterations of

5hmC, are associated with activation of immune pathways and immune infiltration of the

tumor. We show that TET1 is repressed in breast cancer, both in vitro and in mice,

following activation of the major immune regulator NF-κB, through binding to the TET1

promoter. Finally, we extend our findings to other cancer types, including melanoma,

lung and thyroid cancer, suggesting that immunity-driven repression of TET1 could be a

characteristic shared by many cancer types.

Personal contribution to this study includes:

- The general coordination of this project through interactions with the other

contributors;

- The design of experiments and interpretation of the data (along with Annalisa

Canale from the University of Liège);

- RNA-seq analyses, gene ontology analyses, cell culture and cell treatments,

RT-qPCR, Western blot, luciferase assays, agarose-streptavidin binding

assays, ChIP-qPCR, statistical analyses;

- Preparation of the figures;

- Preparation of the manuscript (along with Annalisa Canale).

This chapter summarizes the major points of our study. For further details, please see the

related manuscript provided in Appendix (Collignon et al., “Immune activation of NF-κB

drives TET1 dysregulation in cancer”, currently under submission).

Page 99: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

99

1.2. Results

1.2.1. TET1 expression is associated with 5hmC dysregulation in BLBC

TET1 has been described as a tumor suppressor in breast cancer (BC). Studies

have indicated that the gene was downregulated, which was associated with increased

tumor growth in mice, as well as poor survival in patients (Hsu et al. 2012; Sun et al.

2013; Yang et al. 2015). However, none of these studies took into account the

heterogeneity of BC. Hence, we decided to look more in depth at the regulation of TET1,

particularly in regard to the various BC subtypes.

To assess TET1 expression in breast tumors, we used publicly available RNA-seq

data from The Cancer Genome Atlas (TCGA) consortium (Koboldt et al. 2012). We

subdivided the samples into the four main BC subtypes and compared TET1 expression

to normal breast (see Fig. 43). As reported in publications for BC, TET1 expression was

found decreased in three of the molecular subtypes – luminal A (n=213), luminal B

(n=116) and HER2-like (n=54) tumors – compared to normal tissues (n=101). In sharp

contrast, BLBC (n=90), which composed the fourth subtype, displayed a different pattern

of expression. The range of expression was much wider – approximately 4-fold wider

than other BC subtypes – with some tumors displaying low expression and other high

expression of the TET1 gene.

Fig. 43: TET1 expression in BC subtypes. Expression was assessed in breast tissue by

meta-analysis of public RNA-Seq data from the TCGA cohort. Normal: normal breast

(n=101); BLBC: basal-like breast cancer (n=90); HER2: HER2-like breast cancers

(n=54); LumA: luminal A breast cancers (n=213); LumB: luminal B (n=116).

Page 100: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

100

Given the wide range of TET1 expression in BLBC tumors, we took advantage of

this subtype to investigate whether 5hmC dysregulations were associated with TET1

levels. We obtained DNA from 4 pairs of matched breast tissues (i.e. tumor and normal

samples coming from the same patients). We used the previously described hMe-seal

method to specifically select hydroxymethylated fragments, followed by deep sequencing

(referred here as “5hmC-seq”) (Delatte et al. 2015).

We first clustered the breast sample pairs based on TET1 expression in the

respective tumor (2 pairs with high expression and 2 pairs with low expression). In the

group characterized by low tumor TET1 expression, we identified 256 differentially

hydroxymethylated regions (dhmRs) in tumors compared to their matched normal tissues.

Most dhmRs were hypohydroxymethylated (58%). In contrast, in the group characterized

by high TET1 expression, we identified 160 dhmRs that were almost exclusively

hyperhydroxymethylated (98%). All dhmRs are displayed in a heat map in Fig. 44. The

overlap between the dhmRs of the 2 groups was extremely low with only 2 gene bodies

and 1 intergenic region in common. These results indicate that BLBC tumors with distinct

TET1 expression levels display different patterns of 5hmC alterations, with high TET1

expression being associated with 5hmC gain and vice versa.

Fig. 44: TET1 regulation is associated with distinct 5hmC changes in BLBC.

Sequencing of 5hmC was performed in 4 pairs of BLBC and matched normal breast.

Paired samples were clustered based on TET1 expression in the respective tumor: 2 pairs

with low TET1 expression (left), 2 pairs with high TET1 expression (right). Based on

selection criteria (log FC>3 and FDR<0.05), 256 and 160 differentially

hydroxymethylated regions (dhmRs) were identified in BLBC of each group, respectively.

The heat maps illustrate 5hmC levels (in CPM, counts per million) of the dhmRs identified

in BLBC with low TET1 expression (left) and high TET1 expression (right), compared to

matched normal breast.

Page 101: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

101

Next, we investigated the potential link between 5hmC, 5mC and gene expression

in each BLBC groups. Specifically, we looked at the differential regulation of all coding

gene-associated dhmRs between cancers and normal breast (Fig. 45). In the TET1-low

group, loss of 5hmC was mostly associated with a gain of 5mC. Consistently, in the TET1-

high group, gain of 5hmC was mostly associated with loss of 5mC. Hence, there was a

negative relationship between 5hmC and 5mC changes in BLBC. Furthermore, in both

groups, genes displaying 5hmC changes were also found deregulated at the expression

levels, albeit with no specific direction (either up- or downregulated, regardless of the

hyper- or hypohydroxymethylated status). These results suggest that there is a link

between DNA hydroxymethylation, DNA methylation, and gene expression in BLBC

tumors.

Fig. 45: Link between 5hmC, 5mC and gene expression in BLBC. Heat maps

illustrating 5hmC, 5mC and expression changes in BLBC with low TET1 expression (left)

and high TET1 expression (right), compared to normal breast. Only coding genes

associated with dhmRs are represented for each tumor group. 5mC changes were

measured using Illumina 450K Infinium in the same matched samples. The most variant

probe of the corresponding region (promoter or gene body) was represented. Expression

(mRNA) changes were obtained from TCGA by comparing RPKM values of the 25 BLBC

tumors with the lowest and highest TET1 expression, respectively, and normal breast.

Page 102: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

102

1.2.2. Link between TET1 expression and immunity in BLBC

In order to unravel the mechanisms responsible for TET1 dysregulation in breast

cancer, we next investigated the relationship between TET1 expression and signaling

pathways. Given that BLBC tumors display both high and low expression of TET1, we

focused on that subtype in particular.

From the TCGA RNA-Seq data, we selected all genes presenting a Pearson

correlation coefficient r superior to 0.25 (either positive or negative) with TET1

expression and we performed a gene ontology analysis with DAVID. Strikingly, the top

pathways enriched in the genes negatively correlated with TET1 were all related to

immunity and defense (Fig. 46A). To illustrate this result, we computed a heat map of the

top 20 genes from the “immune response” category. The score of this “immune response”

signature presented an r correlation coefficient of -0.49 with TET1 expression

(p<0.00001) (Fig. 46B). In contrast, the same signature of genes displayed a much weaker

correlation with TET1 in other BC subtypes, with r coefficients for the signature score of

-0.18 (p=0.009) for luminal A tumors, -0.13 (p=0.16) for luminal B tumors, and -0.15

(p=0.29) for HER2-like tumors. Hence, TET1 expression displayed a negative correlation

with expression of many immune markers, specifically in BLBC.

Fig. 46: TET1 is anticorrelated with genes linked to immune pathways. (A)

Functional enrichment analysis was performed with DAVID on all genes presenting a

correlation coefficient r > 0.25 or r < -0.25 with TET1, based on gene expression (RPKM)

of TCGA BLBC samples (n=90). The top 5 of immune and defense categories are

represented. (B) Heat map illustrates gene expression (RSEM z score) of the top 20 genes

from the “immune response” category from panel A, based on correlation coefficient r.

TCGA BLBC samples were ordered by TET1 expression.

Page 103: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

103

The immune and defense pathways identified included genes related to the

myeloid/macrophage compartment, such as TYROBP and CD14, as well as genes related

to the lymphoid compartment, such as CD3D, CD4, CD8A, and LST1. Importantly, genes

of key regulatory factors involved in defense pathways, such as the NF-κB member

RELA, the major histocompatibility complex (MHC) class I partner B2M and the

chemokine CCL2, were also found inversely correlated with TET1 expression. Examples

are provided in Fig. 2 and supplementary Fig. S1 of the manuscript presented in appendix

I.

Next, we further characterized the immune status of BLBC tumors by performing

immunohistochemistry (IHC) of classical immune markers to quantify tumor infiltration

by immune cells. This was performed in collaboration with the team of K. Willard-Gallo

(Bordet Institute). As shown in Fig. 47A, the CD45 antigen, commonly used to score

global leukocyte infiltration, was negatively correlated with TET1 expression. BLBC

tumors with high TET1 expression displayed significantly lower infiltration by leukocytes

by IHC than BLBC tumors with low TET1 expression (p=0.04). To further explore the

infiltration in respect to the various immune populations, we scored the infiltration of T

and B lymphocytes by staining CD3 and CD20 antigens, respectively. Consistent with

our findings that TET1 is negatively associated with global leukocyte infiltration, we

observed a negative relation between TET1 expression and the infiltration of the tumors

by T and B lymphocytes (p=0.029 and p=0.005, respectively).

In order to confirm and extend our IHC results, we used CIBERSORT, a method

for characterizing cell composition of complex tissues from their gene expression profiles

(Fig. 47B). Consistently with IHC results, tumors with high TET1 expression displayed

lower infiltration of several immune populations, including CD4+ and CD8+ T

lymphocytes (with p-values of 0.008 and 1.10-5, respectively) and M1 macrophages (with

p=5.10-5).

Taken together, our results indicated that the global immune state of tumors w

negatively correlated with TET1 expression. BLBC with a low expression of many

immunity-related genes, as well as a low infiltration by the major types of leukocytes,

were characterized by a higher expression of TET1 overall, and vice versa.

Page 104: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

104

Fig. 47: High TET1 expression discriminates BLBC tumors with low immune

infiltration. (A) High expression of TET1 was associated with low leukocyte infiltration

in BLBC tumors. Tumor infiltration was measured by IHC. Staining of CD45, CD3, and

CD20 was performed to quantify leukocytes, T lymphocytes, and B lymphocytes,

respectively (n=18). (B) Infiltration by major immune sub-populations was further

analyzed by using CIBERSORT, a method for characterizing cell composition of complex

tissues from their gene expression profiles in BLBC (n=90). Gene expression data

(RPKM) were obtained from TCGA.

1.2.3. Activation of NF-κB drives TET1 repression

Given the strong link between immune markers and TET1 repression in BLBC,

we wondered whether activation of immune pathways could be involved in the

dysregulation of TET1. Because both immune cells and immune markers were found

negatively correlated with TET1 in BLBC tissues, we hypothesized that cytokine release

from immune cells of the tumor microenvironment could lead to TET1 repression in

breast cancer cells.

In order to test this hypothesis, we first treated triple negative breast cancer MDA-

MB-231 cells with media conditioned by myeloid U937 cells (Fig. 48). This method,

previously established by Mohamed (2012), allowed us to test the effects of U937-derived

Page 105: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

105

secreted cytokines, chemokines and growth factors on immune cells. In these

experimental conditions, a significant decrease of TET1 expression (2.6-fold, p=0.02),

but not TET2 or TET3, was detected by RT-qPCR. Furthermore, the decrease of TET1

was confirmed at the protein level by Western blot. This result suggested that mediators

secreted by leukocytes could indeed lead to TET1 repression in breast cancer.

Fig. 48: Leukocyte-conditioned media

represses TET1 expression. MDA-MB-231

cells were treated with media previously

conditioned by myeloid U937 cells. TET

expression was measured by RT-qPCR (left

panel, n=3, data expressed as mean ± SD,

relative to control) and nuclear TET1

protein and p65 levels were assessed in

control (CTL) and conditioned media (CM)

conditions by Western blot (right panel,

representative of three independent

experiments).

Next, we aimed to unravel the specific mechanisms related to immune pathways

that could drive TET1 regulation. Because the genes previously identified as anti-

correlated with TET1 covered a panel of miscellaneous immune genes, encompassing

innate and adaptive immunity, as well as inflammatory markers, it seemed likely that a

central immune regulator might affect TET1 expression, rather than a highly specific

factor. In regard to that, several clues pointed towards a potential involvement of the NF-

κB family, which affects many immune and inflammatory functions, as hereafter

explained.

First, the NF-κB member RELA, which codes for the protein p65, was among the

genes identified as negatively correlated with TET1 in BLBC tumors of the TCGA cohort

(see Fig. 2 of the manuscript presented in appendix I). Secondly, when MDA-MB-231

cells were treated with U937-conditioned media, there was an increase in nuclear p65, as

observed by Western blot, which revealed the activation of canonical NF-κB pathway in

these experimental conditions (Fig. 48). Thirdly, based on RNA-seq data of the TCGA

cohort, we scored a NF-κB signature, and we observed that high TET1 expression was

associated with low signature score in BLBC tumors (Fig 49A) (p=2.10-5). Finally, in a

publicly available dataset (GSE52707), TET1 was reduced when NF-κB member p65 was

overexpressed in breast cancer cells (Fig. 49B). Taken together, these data suggested that

NF-κB activation could contribute to TET1 repression.

Page 106: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

106

Fig. 49: TET1 expression and NF-κB in

breast cancer. (A) High expression of

TET1 was associated with a low NF-κB

signature in BLBC tumors (n=90). Gene

expression data (RSEM z scores) were

obtained from TCGA. (B) Breast cancer

cells stably expressing an active form of

NF-κB p65 (RELA) have a lower expression

of TET1. Data were obtained from

GSE52707 on NCBI GEO database.

Given the association between TET1 expression and NF-κB activation, we next

aimed to test the causality of the relationship by activating NF-κB in vitro through

different approaches. For each condition, we verified that NF-κB was properly activated

by assessing nuclear translocation of p65 (Fig. 50) by Western blot, and the induction of

target genes IL6 and IL8 by RT-qPCR (see supplementary figure S4 of the manuscript

presented in appendix I).

First, p65 was overexpressed in MDA-MB-231 cells, and we observed a decrease

in TET1 expression (1.7-fold; p=0.04) (Fig. 50A). Of note, this result was consistent with

public data on p65 overexpression in BC cells (Fig. 49B). However, in cancer tissues,

elevated NF-κB activity is often achieved through cytokine release from cells of the tumor

microenvironment (Ben-Neriah & Karin 2011). Hence, in our next experiment, we treated

MDA-MB-231 cells with TNF or LPS, two soluble factors that are well-known to induce

the activation of the NF-κB pathway through their signaling cascade (Hellweg et al.

2006). This led to a 1.8-fold (p=0.01) and 2.6-fold (p=0.001) decrease of TET1

expression, respectively (Fig. 50B and 50C). Taken together, our results from these three

modes of NF-κB activation (p65 overexpression, LPS, TNF) all suggested that NF-κB

could specifically downregulate TET1, with no effect on TET2 expression and no effect

or an increase of TET3 expression. The effect of TNF on TET1 expression was also

confirmed in two other triple negative breast cell lines, i.e. Hs 578T and BT549 (see

supplementary figure S5 of the manuscript presented in appendix I). However, a careful

interpretation of the data was required because the cell signaling effects of cytokines, such

as TNF, are broad and not restricted to NF-κB activation. Therefore, in our next

experiment MDA-MB-231 cells were pretreated with MG-132, a known blocker of NF-

κB activation before TNF treatment. In these conditions, repression of TET1 was

compromised (1.2-fold reduction, p=0.15) (Fig. 50D), which suggested that TNF-

mediated repression of TET1 was actually due to NF-κB activation.

Page 107: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

107

Fig. 50: TET1 expression is repressed by NF-κB activation in vitro. (A) NF-κB member

p65 was overexpressed in MDA-MB-231 cells and TET expression was measured by RT-

qPCR (left panel, n=3, data expressed as mean ± SD, relative to control). Nuclear p65

levels were assessed by Western blot (right panel, representative of three independent

experiments). (B,C) NF-κB was activated by LPS and TNF treatment in MDA-MB-231

cells for 4h and TET expression was measured by RT-qPCR (left panel, n=3, data

expressed as mean ± SD relative to control). Nuclear p65 levels were assessed by Western

blot (right panel, representative of three independent experiments). (D) TNF-dependent

activation of NF-κB was blocked by pre-treatment of MDA-MB-231 with MG-132 and

TET expression was measured by RT-qPCR (left panel, n=3, data expressed as mean ±

SD relative to control). Nuclear p65 levels were assessed by Western blot (right panel,

representative of three independent experiments).

Finally, in order to test whether NF-κB-dependent dysregulation of TET1 was

conserved in vivo, we took advantage of a transgenic mice system called IKMV. In this

model, aberrant NF-κB activation leads to a pre-cancerous state in mammary epithelium

(Barham et al. 2015). The description of the transgenic model can be found in the study

by Barham et al. and in the Fig. 4 of the manuscript presented in appendix I. Briefly,

mammary glands were collected after 3 days with or without activation of NF-κB, and a

significant reduction of Tet1 expression was detected in those pre-cancerous epithelia,

both by RT-qPCR (4.5-fold; p=0.001) and Western blot (Fig. 51).

This last result further confirmed previous findings: in vitro and in vivo data both

reinforce the concept that NF-κB activation negatively regulates TET1 expression in BC.

Taken together with previous observations on immune markers and immune infiltration

in breast tissues, our data support a model in which cytokine release by cells of the micro-

environment and subsequent action of NF-κB mediate TET1 dysregulation in BLBC.

Page 108: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

108

Fig. 51: TET1 expression is repressed by NF-κB activation in vivo. Effect of NF-κB

activation was assessed in vivo in the breast using IKMV transgenic mouse model,

previously described (Barham et al., 2015). Tet expression was measured by RT-qPCR

(left panel, n=3, data expressed as mean ± SD relative to control) and Tet1 protein level

was assessed by Western blot (right panel, n=2 for controls and n=3 for IKMV samples).

1.2.4. TET1 is repressed through binding of NF-κB to its promoter

Next, we investigated the molecular mechanisms involved in TET1 regulation.

Given that NF-κB can bind DNA and act as a transcription factor, we started by searching

for the known consensus sequence of NF-κB binding sites in the vicinity of the TET1

promoter with in silico analyses. We used three different prediction algorithms (JASPAR,

AliBaba and TFBIND), and we identified two putative p65 binding sites, thereafter

named sites A and B. Both sites were close to the transcription start site (TSS) of the gene

(Fig. 52).

Fig. 52: Schematic view of TET1 gene promoter.

Two NF-κB binding sites, named “A” and “B” were

identified based on 3 prediction algorithms

(JASPAR, AliBaba and TFBIND). Binding site

locations are indicated in base pair (bp), relative to

TET1 transcription start site (TSS).

The presence of consensus p65 binding sequences raised the possibility that NF-

κB might bind to TET1 promoter in order to regulate its expression, therefore we next

sought to test this hypothesis. First, we investigated whether NF-κB-mediated regulation

of TET1 was linked to its promoter region. This was performed with luciferase reporter

assay. Briefly, MDA-MB-231 cells were transfected with a plasmid coding for the

luciferase enzyme under the control of the TET1 promoter and luciferase activity was

measured. Upon NF-κB activation (attained by overexpressing p65 or TNF treatment),

luciferase signal was decreased, indicating that the effect on TET1 expression was, at least

in part, promoter-dependent (Fig. 53A).

Page 109: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

109

Fig. 53: Binding of p65 to TET1 promoter. (A) TET1 promoter activity was assessed

by cotransfecting a Firefly luciferase vector under the control of TET1 promoter (TET1-

LUC) and a control Renilla luciferase vector (R-LUC) in MDA-MB-231 cells before TNF

treatment or overexpressing p65 (n=3, data expressed as mean ± SD, relative to control

condition). (B) ChIP was performed with an antibody targeting p65 or control IgG to

assess binding to TET1 promoter in MDA-MB-231 cells. TNF treatment (30 min) was

used to induce nuclear translocation of p65. Positive and negative controls (Pos1-2 and

Neg1-2) were chosen based on public p65 ChIP-seq data (GSM1055811). (C)

Streptavidin-agarose pulldown assays were performed with biotinylated DNA probes

corresponding to the predicted NF-κB binding sites A and B. To assess the specificity of

the binding, pulldowns were achieved with either the wildtype sites or a mutated version

in which the consensus NF-κB binding sequence was disrupted (wildtype probes: A, B;

mutated probes: A mut, B mut, representative of three independent experiments).

Next, we tested the putative binding of NF-κB to TET1 promoter with two

different methods. First, we performed ChIP-qPCR with a p65-targeting antibody. Upon

TNF treatment, binding of p65 to TET1 promoter was increased by 2.7-fold over IgG

(p=0.03) (Fig. 53B). While this result confirmed that TET1 promoter could be bound by

p65, the close proximity of the two binding sites (around 200bp) did not allow to distinct

them properly by ChIP-qPCR. Hence, we used a second approach: streptavidin-agarose

pulldown assays. This method, based on short DNA probes, allowed a better resolution

than ChIP analyses and had been used previously to study NF-κB binding sites (Deng et

al. 2003; Wu 2006). Briefly, proteins extracted from cells (treated or not with TNF) were

incubated with biotinylated probes representing the two potential binding sites, and the

binding of proteins to the probes was revealed by pulldown with streptavidin-agarose

beads followed by Western blot (Fig. 53C). The assay was performed with either the

Page 110: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

110

wildtype A and B site probes or mutated versions, in which the consensus NF-κB binding

sequences were disrupted (see supplementary figure S6 of the manuscript presented in

appendix I for more information on the probe sequences). Strikingly, the B probe showed

a strong p65 binding upon TNF treatment that was mostly lost with the disruption of the

consensus sequence. In contrast, the A probe showed a weak binding at the background

level of its mutated version.

The above results suggested that TET1 is repressed through binding of NF-κB to

its promoter. The B site was the most potent binding site responsible for NF-κB-mediated

regulation in breast.

1.2.5. TET1 is downregulated by NF-κB in other cancer types

In the last part of this study, we investigated whether NF-κB-dependent

dysregulation of TET1 occurred in other cancer types, beside basal-like breast cancers.

This question was raised by the knowledge that both TET1 downregulation and NF-κB

activation have been observed in many cancer types (Ben-Neriah & Karin 2011; Jeschke

et al. 2016).

Therefore, we screened the TCGA cohorts for all available cancer types with

RNA-Seq data. First, we scored the “20 immune gene signature” initially identified in

BLBC (Fig. 46) in each cancer cohort in order to evaluate the global immune state of the

tumors. Then, we calculated the correlation between this signature score and TET1

expression. The list of all TCGA cancer cohorts and their correlation coefficient is

provided in Table 1 of the manuscript presented in appendix I. Most cancer types

displayed a shift in global immune state that was significantly correlated with TET1

expression, including thyroid carcinoma (THCA), skin cutaneous melanoma (SKCM)

and lung adenocarcinoma (LUAD) (Fig. 54). The combined scores for the immune

signature presented a correlation coefficient of r=-0.33 (p=10-14), r=-0.33 (p=10-13) and

r=-0.28 (p=9.10-11) for THCA, SKCM and LUAD, respectively. Interestingly, these

cancers are all known to be infiltrated or surrounded by immune-reactive cells (Oble et

al. 2009; Imam et al. 2014; Kargl et al. 2017).

Page 111: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

111

Fig. 54: TET1 expression and immune markers in cancer. Heat map illustrating the

expression (RSEM z score) of the 20 “immune response” gene signature from Fig. 45 in

several cancer types (from left to right, THCA: thyroid carcinoma, SKCM: skin cutaneous

melanoma, and LUAD: lung adenocarcinoma). Data were taken from TCGA cohorts and

ordered by TET1 expression for each cancer type.

Hence, TET1 downregulation was associated with increased expression of many

immune markers in several cancers. Furthermore, we investigated the potential link

between NF-κB signaling and TET1 regulation in the same cohorts by scoring the NF-κB

signature previously used (Van Laere et al. 2006). Consistent with results in BLBC, high

expression of TET1 was associated with low NF-κB signature, and vice versa, in THCA,

SKCM and LUAD (Fig. 55). This result was the first piece of evidence supporting the

concept that NF-κB activation might be associated with TET1 dysregulation in cancers,

beyond BC.

Fig. 55: TET1 expression and NF-κB signature. High expression of TET1 was

associated with a low NF-κB signature score in THCA (n=509), SKCM (n=472) and

LUAD (n=510) tumors. Gene expression data (RSEM z scores) were obtained from

TCGA.

Page 112: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

112

We further explored the association between TET1 and NF-κB as we tested the

causality of the relationship. This was performed by activating NF-κB with TNF in vitro

in TPC1, A375 and A579 cell lines, derived from thyroid cancer, melanoma, and lung

cancer, respectively (Fig. 56A). Similarly to our results in breast cancer cells, decreased

TET1 expression was consistently observed upon TNF treatment in TPC1, A375 and

A579 cells (3.2-fold with p=0.0001; 1.9-fold with p=0.02 and 1.6-fold with p=0.03;

respectively). Therefore, NF-κB-dependent regulation of TET1 appears to not be

restricted to BLBC, and instead can also occur in other tumors. Finally, we confirmed

that NF-κB could bind to the TET1 promoter via its consensus target sequence.

Streptavidin-agarose pulldown assays were performed with the probe corresponding to

the NF-κB B binding site, and p65 binding was observed upon TNF induction, in thyroid

TPC1 cells, melanoma A375 cells and lung A549 cells (Fig. 56B). As observed in the

breast cells, disruption of the NF-κB consensus sequence of the probe reduced binding.

Fig. 56: NF-κB represses TET1 expression in cancer. (A) NF-κB was activated by TNF

treatment for 4h in thyroid cancer TPC1 cells, melanoma A375 cells and lung cancer

A549 cells. TET expression was measured by RT-qPCR (n=3, data expressed as mean ±

SD relative to control). (B) Streptavidin-agarose pulldown assays were performed as

previously described to assess the binding of NF-κB member p65 to TET1 promoter in

TPC1, A375 and A549 cell lines (representative of three independent experiments).

Taken together, these results suggested that the previously identified mechanism,

in which immunity drives TET1 downregulation through NF-κB activation and binding

to its promoter, was not restricted to BLBC and could instead be mechanisms shared by

many cancer types.

Page 113: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

113

1.3. Key findings

This study aimed to explore dysregulation of TETs and 5hmC in BC, and some of

the uncovered concepts extend well beyond the epigenetics of breast tissues. First, TET1

dysregulation (either up- or downregulation) correlates with distinct profiles of DNA

hydroxymethylation in BC. These profiles are further associated with changes in 5mC

and gene expression. Secondly and importantly, TET1 downregulation is associated with

increased expression of immune markers in basal-like breast cancers. Noteworthy,

beyond the mere activation of immune genes, the tumor infiltration by several types of

leukocytes is also negatively correlated with TET1 expression. Thirdly, activation of the

NF-κB pathway can downregulate TET1. NF-κB member p65 notably binds to the TET1

promoter upon activation of the pathway, and TET1 repression in this context is promoter-

dependent. Finally, examination of the TCGA cohorts and in vitro results suggest that the

mechanism of TET1 downregulation is relevant to many cancer types.

In conclusion, our findings unravel for the first time a paradigm in which the

immune system can regulate cancer epigenetics via dysregulation of the TETs, opening

new avenues in terms of cross-talk between cancer cells and the micro-environment.

Page 114: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

114

Page 115: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

115

2. RNA hydroxymethylation: a new player in the

game

2.1. Introduction

In DNA, TET enzymes (TET1, TET2, and TET3) catalyze the oxidation of 5-

methylcytosine (5mC) to 5-hydroxymethylcytosine (5hmC). Since the discovery of their

implication in DNA demethylation, TETs and 5hmC have been involved in many

mechanisms linked to health and disease (Delatte & Fuks 2013). Yet, TET-mediated

regulation is not restricted to their activity on DNA. For instance, several studies have

already suggested an effect of TETs independent of their catalytic activity (Kaas et al.

2013; Deplus et al. 2013). Furthermore, and most importantly, the existence of TETs’

substrate, 5-methylcytosine, in RNA transcripts (here abbreviated as 5mrC) raised the

question of the potential existence of TET-mediated oxidation to form RNA 5-

hydroxymethylcytosine (5hmrC). Two independent studies have confirmed the existence

of 5hmrC in various species, although its functional role remains undetermined (Fu et al.

2014; Huber et al. 2015).

In this context, our host laboratory set the aim to unravel the first transcriptome-

wide distribution and function of RNA hydroxymethylcytosine in order to provide a better

understanding of 5hmrC. We started by investigating 5hmrC in Drosophila melanogaster

for the following reasons: (i) 5mC and 5hmC in Drosophila DNA are absent, and (ii) there

is only one Tet gene in Drosophila, making it an easier model to study than mammals.

We notably mapped the transcriptome-wide 5hmrC landscape, revealing

hydroxymethylcytosine in the transcripts of many genes, and particularly in coding

sequences. Moreover, we found that RNA hydroxymethylation can favor mRNA

translation. Finally, Tet and hydroxymethylated RNA are particularly abundant in the

brain, and loss of Tet and 5hmrC impaired brain development.

Following up on this pioneer study, we started to explore the involvement of RNA

hydroxymethylcytosine in other contexts related to human health and disease. Hence, we

decided to map 5hmrC in cancer, a disease in which TETs are known to be vastly

Page 116: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

116

dysregulated, more specifically in breast cancer. By doing so, we identified hundreds of

transcripts differentially hydroxymethylated, including some key cancer factors. This

finding raises the question of the functional impact of 5hmrC dysregulations on

tumorigenesis and highlights a new level of cellular alterations in malignant cells.

Personal contribution to this study includes:

- Interactions with the other contributors, including the teams of Véronique

Kruys and Ruth Steward;

- Cell culture, knockdown by RNA interference, RT-qPCR, in vitro

transcription, dot blot assays, interpretation of sequencing results, statistical

analyses;

- Preparation of the figures (along with Rachel Deplus)

This chapter summarizes the major points of our study. For further details, please see the

related manuscript provided in Appendix II (Delatte et al., “Transcriptome-wide

distribution and function of RNA hydroxymethylcytosine”, published in Science in 2016).

Of note, the breast cancer results are not included in the published manuscript and are

part of an ongoing study.

Page 117: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

117

2.2. Results

2.2.1. RNA hydroxymethylation by dTet in Drosophila S2 cells

To detect and measure 5-hydroxymethylcytosine RNA modification in

Drosophila, we set up a dot blot method based on an antibody raised against the modified

base (Kallin et al. 2012; Wu et al. 2011). First, to confirm that this method can detect

5hmrC, we performed dot blot experiments using in vitro transcribed templates

containing either unmodified, methylated, or hydroxymethylated cytosines (Fig. 57). As

expected, the antibody specifically recognized the transcripts containing 5-

hydroxymethylcytosine, and the binding was abolished after ribonuclease (RNase) A

treatment.

Fig. 57: Validation of 5hmrC-detecting

dot blot method. Antibody specificity and

sensitivity for 5-hydroxymethylcytosine-

containing RNA. Serially halved amounts

(starting at 50 ng) of transcribed template

containing either C, 5mrC or 5hmrC were

dot blotted and detected with an anti-5hmC

antibody. A representative experiment is

shown, which was successfully repeated 3

times.

After validating the dot blot method, we detected 5hmrC in dot blot experiments

on total RNA extracted from Drosophila S2 cells (Fig. 58A). Then, in order to determine

whether certain RNA fractions were more enriched in 5hmrC in Drosophila, isolation of

polyadenylated (poly A) RNA from S2 cells was performed, followed by

immunoblotting. S2 poly A RNA showed strong enrichment in 5hmrC signal as compared

with that of total cellular RNA (Fig. 58B). In contrast, in fractions enriched in small RNAs

or ribosomal RNAs, no 5hmrC signal was detected (Fig. 58C). These results indicated

that 5hmrC was most frequent in poly A RNA, which includes mostly messenger RNAs

and long non-coding RNAs.

Page 118: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

118

Fig. 58: 5hmrC is enriched in poly A RNA in S2 Drosophila cells. (A) Dot blotting on

total RNA from Drosophila S2 cells with antibody to 5hmC, treated or not with RNase A

(serially halved amounts of RNA, starting at 1 µg). Data are mean ± SD (n = 4

experiments run) with a representative blot shown. (B) Immunoblotting with anti-5hmC

antibody was performed on polyadenylated and total RNA from S2 cells. Data are mean

± SD (n = 3 experiment run). (C) 5hmrC content of total RNA as well as fractions

enriched in small RNA or rRNA was assessed by dot blotting. Data are mean ± SD (n =

3 experiments run). A vertical line indicates juxtaposition of lanes within the same blot,

exposed for the same time.

Next, we sought to assess whether Tet was responsible for RNA

hydroxymethylation. As previously mentioned, Drosophila possesses only one conserved

Tet ortholog, hereafter named dTet (Gowher et al. 2000; Zhang et al. 2015). Following

knockdown of dTet in S2 cells by using RNA interference (dTet KD), we observed a 44%

reduction in 5hmrC signal detected by dot blot (Fig. 59). This result suggests that dTet is

the enzyme responsible, at least in part, for the formation of RNA hydroxymethylation.

Fig. 59: dTet knockdown leads to reduced

5hmrC levels. (Left) Knockdown of dTet was

assessed by RT-qPCR analysis. (Right)

5hmrC levels were assessed by dot blotting.

Data are mean ± SD (n = 4 experiments run).

In conclusion, dot blot analyses indicate that 5hmrC exists in Drosophila RNA, is

enriched in the poly A fraction, and could be catalyzed by dTet.

Page 119: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

119

2.2.2. Transcriptome-wide mapping of 5hmrC in S2 cells

Next, we aimed to provide the first transcriptome-wide mapping of 5hmrC. This

was performed by hydroxymethylated RNA immunoprecipitation followed by

sequencing (hMeRIP-seq). This method involves immunoprecipitation of 5hmrC-

containing RNA with the antibody to 5hmC, followed by next-generation sequencing.

In S2 cells, hMeRIP-seq uncovered 3058 significantly enriched regions (“5hmrC

peaks,” p < 10−10). Among those regions, 1597 coding gene transcripts were found.

Examples of enrichment profiles are shown in Fig. 60A. The distribution of 5hmrC peaks

revealed by hMeRIP-seq analyses showed an enrichment in coding sequences (48%) and

introns (17%) (Fig. 60B). Further analyses also uncovered a motif commonly associated

with 5hmrC peaks (64% of identified target sites) that was highly UC-rich and containing

UCCUC repeats (see Fig. 2 of the manuscript presented in appendix II for more details).

Fig. 60: Transcriptome-wide distribution of 5hmrC in Drosophila cells. (A)

Representative UCSC Genome Browser plot from hMeRIP-seq data. The upper lane

represents the hMeRIP data, the lower lane the control (Ctrl, representative of the sample

input). (B) Distribution of 5hmrC peaks according to the type of structural element within

the transcript. *P < 10−5).

Given that the hMeRIP-seq was a newly established technique, we included

several controls in order to ensure the specificity of the signal. First, bioinformatic

analyses established that our hMeRIP-seq data did not merely select abundant RNA

fragments in non-specific manner (see Fig. 2B and Fig. S5 of the published manuscript).

Secondly, upon dTet depletion, up to 79.4% of peak sites showed a significant reduction

in 5hmrC signal, as compared with control S2 cells. Moreover, among reduced targets,

85.5% of peak sites displayed more than four-fold of reduction in 5hmrC levels (Fig. 61).

Page 120: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

120

Finally, we replicated our hMeRIP-seq results by performing the immunoprecipitation

with a different antibody targeting 5hmC. We obtained a strong agreement between

sequencing performed with the two 5hmC antibodies (77% of common peaks, see Fig.

S7 of the published manuscript for more details).

Fig. 61: dTet mediates transcriptome-wide RNA hydroxymethylation in Drosophila.

(A) hMeRIP-seq in cells depleted of dTet shows reduced 5hmrC levels at the majority of

target regions. (Left) Box plot of the normalized number of 5hmrC reads in dTet-depleted

cells versus control cells. (Center) Pie chart showing the percentage of reduced 5hmrC

peaks, with (right) a more than fourfold reduction at most targets. (B) Example of

enrichment profiles (UCSC tracks) of hMeRIP-seq done in control and dTet-depleted

cells.

Next, we aimed to better understand the role of dTet in gene regulation. Thus, we

performed RNA-seq in S2 cells, and we identified 574 differentially expressed mRNAs

upon depletion of dTet: 50.4% showed increased expression, 49.6% decreased expression

(Fig. 62). We then compared dTet-regulated mRNAs with the targets identified by

hMeRIP-seq, and we identified a small but very significant subset of common targets (p

< 10−43). Precisely, 26% of dTet-regulated mRNAs contained at least one 5hmrC peak

(Fig. 61). It is worth noting that dTet contains an N-terminal CXXC Zn-finger domain.

In mammalian cells, the presence of this DNA-binding domain is thought to explain, at

least in part, the ability of Tets to regulate gene expression independently of their catalytic

activity (Williams et al. 2011; Deplus et al. 2013). Thus, in Drosophila cells, dTet might

also affect gene expression via its CXXC domain independently of its

hydroxymethylation activity.

Interestingly, by gene ontology analysis of the 5hmrC targets, we observed an

enrichment for genes involved in cellular processes, including in the regulation of

embryogenesis, development and neurogenesis (see Fig. S8B of the published

manuscript). These findings suggest that dTet-mediated RNA hydroxymethylation might

influence the development of fruit flies.

Page 121: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

121

Fig. 62: 5hmrC and gene expression upon depletion

of dTet in S2 cells. (A) RNA-seq was performed on

dTet-knockdown and control S2 cells. (B) A partial

overlap was observed between hMeRIP-seq and RNA-

seq data sets.

In conclusion, 5hmrC was identified in over 3000 mRNA regions in Drosophila

S2 cells, and this RNA hydroxymethylation seems, at least in part, mediated by dTet.

Gene ontology analysis performed on 5hmrC targets suggest a potential role in

developmental regulation.

2.2.3. Hydroxymethylation can favor mRNA translation

Next, we sought to determine how cytosine hydroxymethylation might affect

mRNA functions. One aspect of RNA metabolism that can be affected by chemical

modifications is the efficiency of protein translation (Meyer et al. 2015).

Thus, we looked at the distribution of 5hmrC, in respect with the translational

status of mRNAs in Drosophila S2 cells. Specifically, we performed standard sucrose-

gradient fractionation, in order to separate free mRNAs from mRNA bound by ribosome

subunits (40S and 60S), single ribosomes (80S) or polysomes, and we measured 5hmrC

by dot blot in each fraction. Interestingly, we observed a positive correlation between

5hmrC abundance and active mRNA translation: fractions with low translation activity

(free mRNAs and mRNAs bound by 40S/60S/80S) displayed low levels of 5hmrC,

whereas highly translated mRNAs (bound by several ribosomes to form a polysome)

showed a high 5hmrC content (Fig. 63A). In contrast, dot blot quantification of

methylated RNA (5mrC) in the ribosomal fractions showed an enrichment in lowly-

translated mRNAs (bound by single ribosomes) and low levels in polysomes (Fig. 63B).

This was a first piece of evidence that 5hmrC, but not 5mrC, was associated with high

levels of translation.

Page 122: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

122

Fig. 63: Highly translated mRNAs

display high levels of 5hmrC. (A)

Sucrose gradient fractionation followed

by dot blot quantification shows that

active translation is associated with high

5hmrC contents. Data are represented as

mean ± SD (n = 3 experiments run). (B)

Dot blot quantification of 5mrC in the

same fractions shows that low translation

is associated with high 5mrC contents.

Data are represented as mean ± SD (n =

3 experiments run).

Based on the above findings, we next investigated whether mRNA

hydroxymethylation might affect mRNA translation. Therefore, we produced by in vitro

transcription RNA templates (encoding Firefly Luciferase) bearing unmodified,

methylated, or hydroxymethylated cytosines. RNA templates were then incubated in

rabbit reticulocyte lysate to allow protein translation to occur in vitro, and translation

efficiency was assessed by the incorporation of 35S-radiolabeled methionine and by

Western blot (targeting the Firefly Luciferase) (Fig. 64). With both methods, a decrease

in translation was observed for methylated RNAs, compared to unmodified RNAs. In

contrast, 5hmrC-modified templates restored translation levels to near-control levels,

suggesting that hydroxymethylation can counteract the effect of RNA methylation. These

experiments were performed in collaboration with the team of Véronique Kruys.

Fig. 64: 5hmrC favors mRNA translation. In vitro

translation of unmodified C-, 5mrC-, and 5hmrC-

containing RNAs (encoding for Firefly Luciferase, or

“F-luc”), was measured by incorporation of 35S-

radiolabeled methionine (top panel, n=3 with data as

mean ± SD) and Western blot (bottom panel,

representative of 3 experiments run).

Page 123: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

123

Taken together, our data suggest that hydroxymethylation is associated with

highly translated transcripts and that it can restore the translation efficiency of previously

methylated substrates. Overall, 5hmrC appears to favor translation.

2.2.4. In vivo relevance of 5hmrC in Drosophila

Next, we collaborated with the team of Ruth Steward (Rutgers University, New

Jersey, USA), who specializes in the study of fruit flies, in order to assess 5hmrC in vivo.

First, we looked at the expression of dTet and the levels of 5hmrC by RT-qPCR and dot

blot, respectively, during early embryogenesis of Drosophila melanogaster (Fig. 65). Both

showed a similar pattern: increasing levels up to 10h post-fertilization, followed by a

decrease. Thus, dTet expression and 5hmrC levels displayed a positive correlation during

embryogenesis.

Fig. 65: Levels of dTet and 5hmrC

in early embryogenesis. dTet

expression and 5hmrC levels were

measured by RT-qPCR (left) and dot

blot (right), respectively, during

Drosophila embryogenesis. Time is

indicated in hours post-fertilization.

Data are mean ± SD (n = 4

experiments run).

We also analyzed publicly available RNA-seq data from the modENCODE

database in order to assess dTet expression in organs at different stages of D. melanogaster

development. Strikingly, dTet expression was the highest in the central nervous system

of the third-instar larvae (see Fig. S13B of the published manuscript). This observation,

combined with the presence of genes related to neurogenesis among 5hmrC targets in S2

cells, raised the question of a potential role for dTet-mediated hydroxymethylation of

RNA in the fruit fly brain development. To test this hypothesis, we first confirmed the

high expression of dTet in the central nervous system compared to other tissues by two

different methods. Firstly, we used a transgenic fly model, in which endogenous dTet was

tagged with the green fluorescent protein (GFP). The GFP-dTet fusion protein was

detected throughout the larval brain, with the highest levels being detected in the optic

lobe and central brain (Fig. 66A). Secondly, RT-qPCR confirmed the high expression of

Page 124: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

124

dTet as compared to two other organs, the salivary glands and the ovaries (Fig. 66B).

These organs were chosen as controls because: (i) ovaries were the organs in which it was

previously shown that dTet possesses a DNA m6A demethylation activity (Zhang et al.

2015), and (ii) the salivary gland was an organ of sufficient size for us to extract enough

RNA to measure 5hmrC levels. Importantly, 5hmrC positively correlates with dTet

expression, with the highest levels being in the brain (Fig 66B). Hence, several

observations supported the possibility of a role for 5hmrC in this organ.

Fig. 66: dTet and 5hmrC levels in the brain. (A) (Left) Pattern of endogenous GFP-

tagged dTet in the larval brain was visualized by confocal microscopy analysis. OL and

CB indicate the optic lobe and the central brain, respectively. Scale bar, 50 mm. (Center)

Scheme of the larval brain. (B) (Top) dTet expression in the salivary gland, brain, and

ovary, measured by RT-qPCR. Data are mean ± SD (n = 3 experiments run). (Bottom)

Immunoblotting with 5hmC antibody in RNA from salivary gland, brain, and ovary.

Vertical line indicates juxtaposition of lanes within the same blot, exposed for the same

time.

To confirm the role of RNA hydroxymethylation in fruit flies brain development,

we took advantage of loss-of-function mutant of dTet developed by the team of Dr.

Steward. In agreement with recent published data (Zhang et al. 2015), dTet-deficient flies

grew up to the pupal stage, then displayed massive lethality. From over 5000 dTet-null

animals analyzed, no animal reached the adult stage. Moreover, morphological defects

were found at larval stages, with decreased brain size and abnormal neuroblast

organization in the central part. In particular, the width of the medulla region was reduced

in dTet-null flies (p < 2.10−9) (Fig. 67A). Importantly, in dot blot analyses, 5hmrC was

also reduced in the brains of dTet deficient larvae (Fig. 67B). It is worth noting that even

in RNA extracted from dTet-null animals, there are residual levels of 5hmrC, which

means that dTet-mediated oxidation might not be the only source of 5hmrC.

Page 125: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

125

Taken together, our data indicate that dTet and 5hmrC are particularly relevant to

brain development in Drososphila.

Fig. 67: dTet-deficient fruit flies show

impaired brain development, accompanied

by decreased 5hmrC. (A) Average width of the

medulla region containing the neuroblasts (left,

red arrow). Error bars represent 95%

confidence intervals (p < 2.10−9) for 20 brain

lobes. (B) Brains of dTet-deficient larvae show

a decrease in 5hmrC by dot blot analysis. Data

are mean ± SD (n = 3 experiments run).

In conclusion, the investigation of RNA hydroxymethylation in Drosophila

constituted the first study addressing the distribution, localization, and function of this

RNA modification. Our work provided a picture of the hydroxymethylated transcriptome,

identified an unrecognized function for 5hmrC in translation regulation, and identified a

major role of this modification in neurogenesis. Following on this pioneer study, we next

aimed to explore the role of 5hmrC in other biological contexts, and in particular, human

health and disease. In the next section, we will specifically focus on the study of 5hmrC

in the context of breast cancer.

2.2.5. Alterations of 5hmrC in breast cancer

Taking advantage of the methods developed for the study of 5hmrC in Drosophila,

we next aimed to investigate whether 5hmrC could be dysregulated in breast cancer (BC).

For this study, we had at our disposal a panel of breast cell lines: one non-cancerous breast

cell line (MCF12) and six BC cell line (MCF7, ZR75-1, T47D, MDA-MB-231, BT474,

SKBR3). We started by measuring 5hmrC and 5mrC by dot blot analysis in the different

fractions of RNA in all these cell lines (Fig. 68). First, we observed that 5hmrC and 5mrC

had different distributions overall. In agreement with our results in Drosophila (Fig. 58),

5hmrC was globally enriched in poly A RNA compared to total RNA (Fig. 68A).

However, we were also able to detect 5hmrC – albeit at low to medium levels – in small

RNA and rRNA, in discrepancy with results from Drosophila where 5hmrC was

Page 126: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

126

undetectable in the same RNA fractions. This might be an interspecies difference between

human and fly cells. In contrast, 5mrC showed a different distribution: the mark was

globally depleted in poly A RNA and increased in small RNA and rRNA (Fig. 68B).

Moreover, and importantly, BC cells displayed overall reduced levels of 5hmrC, both in

total and poly A RNA, compared to non cancerous breast cells. In contrast, 5mrC was

either unchanged or increased in BC cells compared to non-cancerous cells.

Fig. 68: RNA methylation and hydroxymethylation in mammary cells.

Immunoblotting for (A) 5hmrC, and (B) 5mrC, in RNA from a panel of breast cell lines.

The different RNA species (100ng of poly A RNA or 200ng of total RNA, small RNA and

rRNA) were blotted on the same membrane and exposed for the same time in order to

allow comparison.

Next, we aimed to provide the first transcriptome-wide map of 5hmrC alterations

in BC. Thus, we performed hMeRIP-seq in the non-cancerous breast MCF12 cells and in

breast cancer MDA-MB-231 cells. We identified 1605 and 946 5hmrC peaks, which were

associated with 1093 and 681 transcripts, in MCF12 and MDA-MB-231, respectively. In

both breast cell lines, 5hmrC peaks showed a non-random distribution, with a specific

enrichment in intronic regions (Fig. 69). This result is in contrast with previous results

showing a preferential enrichment in coding sequences in Drosophila cells, suggesting

again that there might be interspecies differences in the transcriptomic distribution of

5hmrC.

Page 127: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

127

Fig. 69: Distribution of 5hmrC peaks in breast cells. The region enrichment

corresponds to the ratio between the observed number of peaks and the number expected

by chance, in each region.

When comparing the two cell lines, we identified 998 differentially

hydroxymethylated regions, associated with 725 RNA transcripts (Fig. 70A). Overall, the

majority of 5hmrC changes were hypo-hydroxymethylated in the cancer MDA-MB-231

cells, which was in agreement with the global loss previously observed (Fig. 68 and 70B).

Furthermore, the majority of transcripts (nearly 70%) showed at least a 4-fold difference

(Fig. 70B). Thus, breast cancer cells display a massive redistribution of 5hmrC.

Fig. 70: Changes in 5hmrC in BC. (A) Based on 5hmrC-sequencing, we identified 725

differentially methylated transcripts between the non-cancerous breast cells (MCF12A)

and the BC cells (MDA-MB-231). These changes are displayed on the heatmap. (B) Pie-chart showing that changes in 5hmrC included more losses (hypo-hydroxymethylation, in

green) than gains (hyper-hydroxymethylation, in red). The majority of changes

corresponded at least to a 4-fold difference.

Page 128: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

128

Gene ontology analysis, performed with IPA®, indicated that genes associated

with 5hmrC changes between MCF12 and MDA-MB-231 cells were significantly linked

to pathways such as “estrogen receptor signaling”, “breast cancer” and “cytoskeleton”,

which are all relevant pathways in the context of mammary tumors. This finding

suggested that changes in 5hmrC are associated with known molecular alterations of BC.

Changes in 5hmrC notably affected transcripts of key cancer genes, such as BRCA2, p53,

CDKs, or RAD51. Examples of 5hmrC tracks are provided in Fig. 71.

Fig. 71: Examples of 5hmrC tracks in breast cells.

In conclusion, we have performed a first investigation of RNA

hydroxymethylation dysregulations in breast cancer. Although these are still early results,

our data suggest that 5hmrC is vastly altered in BC, both globally and locally, and notably

at mRNAs that are known to be involved in cancer development.

Page 129: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

129

2.3. Key findings

This study aimed to uncover for the first time the distribution and function of

5hmrC. Hydroxymethylcytosine, which is well established in DNA, was recently shown

to also occur in RNA. In Drosophila, we showed that 5hmrC preferentially marked

polyadenylated RNAs and was mediated by dTet. We mapped the transcriptome-wide

landscape of the mark, revealing 5hmrC in the transcripts of many genes, notably in

coding sequences. We also found that RNA hydroxymethylation can favor mRNA

translation. Importantly, in Drosophila, dTet and 5hmrC are found to be most abundant

in the brain, and Tet-deficient fruit flies suffered from embryonic lethality with impaired

brain development and decreased RNA hydroxymethylation. Thus, we identified a central

role for 5hmrC in Drosophila development.

In a follow-up study, we provided the first hints that 5hmrC might be widely

dysregulated in breast cancer. Comparison between non-cancerous and cancer cell lines

notably showed hundreds of differentially transcripts, including of key cancer genes. Our

findings highlight a new level of dysregulation in cancer, thus supporting the potential of

RNA modifications as new cancer biomarkers.

Page 130: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

130

Page 131: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

131

3. Dysregulations of m6A and its machinery

support breast cancer

3.1. Introduction

N6-Methyladenosine (m6A) is by far the most abundant modification in mRNA,

and it also exists at lower levels in tRNA, rRNA, small nuclear RNA (snRNA) as well as

several long non-coding RNA. It has been identified in bacteria and eukaryote cells,

including mammals, insects, plants and yeast. It is also the RNA modification with the

best characterized machinery: the mark is added by a METTL3/METTL14 complex, and

it removed by ALKBH5 and FTO, which offers a very dynamic system. Furthermore,

several m6A readers have been identified (e.g. the YTHDF and YTHDC families) and

they can carry out the effect of m6A on RNA stability, splicing, localization and

translation. Taking advantage of this understanding of the m6A dynamicity and functions,

recent attention was brought to the biological roles of the mark. It has notably been

involved in the regulation of pluripotency and cell differentiation during development

(Wang et al. 2014; Geula et al. 2015), thus suggesting that RNA modifications can have

crucial roles in health and diseases.

In that context, we aimed to study the role of m6A and its machinery in breast

cancer development and progression. Recent studies have indeed suggested that loss of

m6A on the transcripts of several key cancer factors could promote leukemia and

glioblastoma (Zhang et al. 2017; Cui et al. 2017; Li et al. 2017), indicating that the study

of m6A players could be valuable in cancer. In breast cancer, we notably mapped the

transcriptome-wide m6A landscape, revealing N6-methyladenosine in the transcripts of

many genes, and particularly in coding sequences, near the stop codon and the 3’UTR. In

cell lines, we identified many transcripts differentially methylated between non-

cancerous breast cells and cancer cells, particularly in major cancer pathways. In human

tumors, we uncovered that FTO, the m6A demethylase, is often downregulated, which is

associated with increased m6A levels and poor survival. Finally, we showed in vitro that

depletion of FTO in mammary cells led to a more aggressive phenotype of the cancer

cells.

Page 132: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

132

Taken together, our data on m6A and 5hmrC (see previous chapter) in breast

cancer both indicate that alterations of RNA modifications and their machinery constitute

a novel level of aberrant gene regulation promoting cancer. Thus, understanding how

RNA modifications can affect various cancers will be a key challenge for the future.

Personal contribution to this study includes:

- The general coordination of this project through interactions with the other

contributors;

- The design of experiments and interpretation of the data (along with Dr. Jana

Jeschke from our host laboratory);

- Cell culture and cell treatments, establishment of stable knockdown in cell

lines, RT-qPCR, dot blot analysis, Western blot, luciferase assays,

xCELLigence experiments (migration and invasion), statistical analyses;

- Preparation of the figures;

This chapter summarizes the major points of our study, which will be prepared for

publication in a near future.

Page 133: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

133

3.2.Results

3.2.1. Transcriptome-wide m6A landscape in breast cancer

Our first aim in this study was to determine whether changes in m6A occurred in

BC. To answer this question, we mapped m6A in several breast cell lines, including

MCF12, SKBR3, and MCF7. This was performed by methylated RNA

immunoprecipitation followed by sequencing (MeRIP-seq), based on a method described

in a published study (Dominissini et al. 2012).

Fig. 72: Transcriptome-wide mapping of m6A in SKBR3 cells. (A) Number of m6A

peaks and m6A-containing transcripts, based on MeRIP-seq in SKBR3 cells (top).

Examples of two m6A peaks are shown (bottom). (B) Region m6A enrichment, measured

by the ratio between the number of observed peaks and the number expected by chance

for each transcriptomic region (from left to right: 5’UTR, near start codon, coding

regions, introns, near stop codon and 3’UTR). (C) Motif analysis revealed that the GACU

motif, a known m6A motif, was commonly observed in m6A targets.

We began by assessing the quality of our m6A-sequencing experiments, and this

was done by looking the features of m6A distribution. In the SKBR3 BC cells, we

identified 5760 m6A peaks, associated with 3342 different transcripts. Examples of peaks

are shown in Fig. 72A. The m6A signal was specifically enriched in the coding region,

near the stop codon and in the 3’UTR (Fig. 72B). These were the same regions that were

shown to be enriched for m6A in other biological contexts (Luo et al. 2014; Geula et al.

2015). Furthermore, analyses also uncovered a motif commonly associated with m6A

target regions (approximately 1/3 of all peaks), i.e. the GACU motif (Fig. 72C). This

sequence corresponded exactly to one of the main known motifs for m6A, which has been

Page 134: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

134

previously detected in ES and other cells (Wu et al. 2016). Similar results were obtained

for the other breast cell lines (i.e. MCF7, BT20, T47D – data not shown). Hence, the

distribution of m6A was non-random and the “classical” m6A feature were conserved.

Next, we aimed to provide the first transcriptome-wide map of m6A alterations in

BC, and we performed this by comparing m6A sequencing results in SKBR3 breast

cancer cells and in MCF12 breast cells.

The latter is a non-tumorigenic epithelial cell line established from a reduction

mammoplasty (Paine et al. 1992). MCF12 cells are non-transformed but immortalized

and hyperplastic. They differ from true normal breast cells, notably through their

abnormal karyotype. Nevertheless, they constitute a non-cancerous breast model that has

been previously used to highlight epigenetic alterations in BC (Stolzenburg et al. 2012;

Park et al. 2015; Bagu et al. 2017). This cell line is particularly useful in studies that

require large amount of material, as was the case for our m6A MeRIP-seq experiments.

Regrettably, reduction mammoplasty only yields limited amount of material, as the

majority of the volume is taken by fatty tissues, and we were unable to acquire enough

RNA from normal breast tissue. Thus, we decided to compare our cancerous SKBR3 cells

to the non-cancerous MCF12 breast cells.

By comparing SKBR3 and MCF12 cells, we identified 1962 differentially

methylated regions (DMRs, defined by a minimum fold-change of 2), with a majority

(about 2/3) of hypo-methylation in cancer cells (Fig. 73A and 73B). Overall these changes

were vast, with nearly half of them displaying at least a 4-fold difference. A heat map

illustrating the top 100 DMRs and example of m6A tracks are shown on Fig. 73C and

73D, respectively. In conclusion, our data from MeRIP-seq suggested that there was a

vast redistribution of m6A in breast cancer cells.

Page 135: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

135

Fig. 73: m6A changes in cultured BC cells. (A) Changes observed in m6A MeRIP-seq

(IP), in respect with gene expression (input), between SKBR3 BC cells and MCF12 breast

cells. (B) Pie-chart representing the gains (hyper) and losses (hypo) of m6A in BC cells

(left panel). About half of these changes corresponded at least to a 4-fold difference (right

panel). (C) Heat map showing the m6A changes of the top 100 differentially methylated

regions (DMRs). (D) Examples of m6A tracks in SKBR3 and MCF12 cells.

Importantly, we were able to get access to, and profile m6A by MeRIP-seq in three

human breast cancer biopsies. This was the first mapping of m6A performed in any

human cancer tissue, all previous sequencing of m6A in cancer having been performed

on RNA extracted from cell lines (Cui et al. 2017; Li et al. 2017; Zhang et al. 2017). The

three samples yielded similar numbers of m6A peaks (4132, 4124 and 3541, respectively)

and associated transcripts (2681, 2721 and 2350) (Fig. 74A). Notably, profiling of m6A

in human biopsies showed a high level of consistency, with 86% of m6A-containing

targets being shared by at least two out of three samples, and 67% being shared by all

three. Even within transcripts, the position and shape of the peaks showed great similarity

(Fig. 74B). Furthermore, we looked at the transcriptomic distribution of m6A, and all

three human biopsies showed an enrichment of peaks in coding sequences, near the stop

codon, and in the 3’UTR (Fig. 74C), which was similar to our observations in breast cell

Page 136: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

136

lines (Fig. 72B) and in published results from other biological models (Luo et al. 2014;

Geula et al. 2015)

Fig. 74: Transcriptome-wide mapping of m6A

in human BC biopsies. (A) MeRIP-seq was

performed on RNA extracted from three distinct

human BC biopsies. The number of m6A peaks

and m6A-containing transcripts are indicated for

each sample. (B) Example of m6A tracks for a

peak shared by all three samples. (C) Region m6A

enrichment, measured by the ratio between the

number of observed peaks and the number

expected by chance for each transcriptomic

region (from left to right: 5’UTR, near the start

codon, coding regions, introns, near the stop

codon and 3’UTR). The data displayed are from

one sample and are representative of the three

biopsies.

Next, taking advantage of the m6A sequencing in human BC biopsies, we assessed

whether our data from BC cell lines, cultured in vitro, constituted a good model for the

study of m6A in BC. Thus, we compared the m6A peaks from breast cancer cells lines

and human biopsies. In terms of transcripts, there was a significant overlap between the

two types of samples. For instance, SKBR3 shared 67% of its m6A targets with at least

one human breast biopsy. In comparison, the transcripts obtained by the same MeRIP-

seq method in our host laboratory in an unrelated model (i.e. ES cells) shared only 37%

of its m6A targets with the breast biopsies. While both overlaps were significant (p <

2.10-16), the breast samples clearly shared more similarity between themselves than with

an unrelated control. In addition, gene ontology analyses performed with IPA® revealed

that the BC cell lines and the human biopsies were both enriched in signaling pathways

such as “molecular mechanisms of cancer”, “Wnt signaling”, “glioblastoma” and “stem

cells”. Hence, cell lines and human tissues displayed changes of m6A in the same key

cancer pathways, and BC cell lines represent a relevant model for the study of m6A.

Page 137: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

137

In conclusion, our transcriptome-wide data presented a picture in which RNA

from BC conserved the known features of m6A (e.g. the region distribution and the

motifs) but was deregulated in many transcripts associated with major cancer pathways.

3.2.2. Deregulation of the m6A machinery in BC

Given our that many m6A changes occurred in breast cancer, we investigated

whether the enzymes related with its cycle were also commonly deregulated in BC. To

assess expression in breast tumors, we used publicly available RNA-seq data from The

Cancer Genome Atlas (TCGA) consortium (Koboldt et al. 2012) (Fig. 75A). The m6A

methyltransferases METTL3 and METTL14, as well as the ALKBH5 demethylase, showed

little change in expression. In contrast, the FTO demethylase appeared to be globally

downregulated in BC. We confirmed the downregulation of FTO in an in-house cohort

by RT-qPCR (p=0.002; Fig. 75B).

Fig. 75: Expression of m6A enzymes in BC. (A) Expression was assessed in breast

tissue by meta-analysis of public RNA-Seq data from the TCGA cohort (101 normal

breast versus 812 BC samples). (B) Validation of FTO expression in an in-house cohort

(9 normal breast versus 47 BC samples)

Given the downregulation of FTO in BC, we next investigated whether the global

amount of m6A was also affected. Thus, we quantified m6A by mass spectrometry, which

is considered the gold standard method for m6A quantification. This was done in

collaboration with the team of Pr. Bi Feng Yuan (Wuhan University, China). First, we

were able to analyze RNA from 6 pairs of matched breast tissues (i.e. tumor and normal

samples coming from the same patients), and we observed an increase in global m6A

levels in BC (p=0.02; Fig. 76A). Secondly, in our in-house BC cohort, we observed a

Page 138: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

138

negative correlation between FTO expression and m6A levels: tumors with low FTO

expression had significantly more m6A (p=0.002; Fig. 76B). Thus, loss of FTO, the m6A

demethylase, was associated with a global gain of m6A in BC.

Fig. 76: Quantification of m6A in BC by

mass spectrometry. (A) Relative amount

of m6A in paired matched samples (normal

breast and BC from the same patients)

(n=6). Lines connect corresponding pairs.

Red dots indicate the mean of each group.

(B) Quantification of m6A in BC in respect

with FTO expression (25 tumors with low

expression versus 15 tumors with high

expression).

Finally, we investigated the clinical relevance of FTO expression in BC.

Interestingly, decreased FTO expression was significantly associated with poor survival

in Kaplan-Meier analysis, both in terms of recurrence-free survival (RFS) and distant

metastasis free survival (DMFS) (Fig. 77). Survival analyses were performed with the

online Kaplan-Meier plotter tool (Szász et al. 2016).

In conclusion, loss of FTO is a common event in breast cancer and it is associated

with a global gain in m6A and poor prognosis.

Fig. 77: FTO expression and survival in BC. (A) Recurrence-free survival (RFS) and

(B) distant metastasis free survival (DMFS) both displayed significant association with

FTO expression in BC (low FTO expression: black line; high FTO expression: red line).

Page 139: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

139

3.2.3. Phenotypic effect of FTO in BC

Taking the previous findings into account, we next investigated whether FTO

dysregulation had an impact on BC phenotype. To address this matter, we produced a

stable knockdown of FTO in SKBR3 cells, which was verified by RT-qPCR and by

Western blot (Fig. 78A and 78B). The subsequent gain in m6A was measured by dot blot

(Fig. 78C).

Fig. 78: FTO-knockdown in BC. (A) Knockdown of FTO was performed in SKBR3 cells.

The downregulation was verified (A) at the mRNA level by RT-qPCR (n=3) and (B) at the

protein level by Western blot (representative of 3 independent experiments). (C) Serially

halved amounts (starting at 100 ng) of poly A RNA from control (shScramble) or FTO-

knockdown (shFTO) cells were dot blotted and detected with an anti-m6A antibody

(representative of 3 independent experiments).

Then, we evaluated the effects of FTO depletion, according to three classical

features of cancer cells: migration, invasion and stemness.

First, we used the xCELLigence system to assess the migratory properties of BC

cells. This technology allows real-time measurement of the passage of cells from an upper

chamber to a lower chamber, guided by a chemoattractant (i.e. serum). With this system,

we observed an increase in migration upon depletion of FTO (Fig. 79A). In order to verify

that this effect was not specific to SKBR3 cells, we also used another BC cell line with

migrating capacities: the BT20 cells. A similar increase in migration was observed in this

cell line upon depletion of FTO (Fig. 79B). Furthermore, we mimicked the loss of FTO

function by treating cells meclofenamic acid (MA), a known inhibitor of its catalytic

activity (Huang et al. 2015), to ensure that the effect was merely not an off-target effect

of RNA interference. Consistently with previous results, inhibition of FTO activity led to

increased migrating properties in SKBR3 cells (Fig. 79C). This result also suggested that

Page 140: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

140

the observed phenotype was linked to the enzyme’s m6A-demethylating activity. Finally,

in order to assess the invasive properties of the cells (i.e. the capacity to travel through a

matrix or membrane), we added a layer of Matrigel between the two chambers, before

performing the xCELLigence assay. In these conditions, we observed an increase in

invasion upon depletion of FTO in SKBR3 cells (Fig. 79D). Taken together, these data

indicated that loss of FTO could enhance the mobility properties of cancer cells, which is

a feature commonly associated with EMT, and thus cancer progression.

Fig. 79: FTO depletion

enhances the mobility of

cancer cells. Knockdown of

FTO promoted cell migration

in (A) SKBR3 cells and (B)

BT20 cells. (C) Treatment of

SKBR3 with meclofenamic

acid (MA) increased

migration, as compared to

control (CTL) cells. (D)

Knockdown of FTO

promoted cell invasion.

All experiments were

performed in triplicate, one

representative result is

shown.

Next, we performed with FTO-depleted cells a tumorsphere (or mammosphere)

assay, in which breast cancer cell were isolated at the single cell level and grew in a media

that did not allow attachment. The capacity of the cells to grow in “colonies” (i.e.

tumorspheres) in such conditions is considered as a reflection of their stemness-like

properties, in particular the self-renewal capability. Using this assay, we observed an

increase in both the number (Fig. 80A) and the size (Fig. 80B) of tumorspheres upon loss

of FTO. This result indicated that reduced FTO expression could support the stemness of

cancer cells, a feature associated high tumorigenicity.

Page 141: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

141

Fig. 80: FTO depletion enhances in

vitro tumorsphere formation. (A)

Tumorsphere assays were performed

with control (shScramble) and FTO-

knockdown (shFTO) cells (n=3, data

as mean ± standard deviation). (B)

Representative pictures of the

tumorsphere are shown.

Taking into account the various phenotypic assays, loss of FTO appeared to

support an aggressive phenotype in terms of migration, invasion and stemness in breast

cancers.

3.2.4. FTO regulates the Wnt/β-catenin pathway in BC

Given the impact of FTO expression over the phenotype of BC cells, we wanted

to uncover the molecular mechanisms underlying this effect and whether they relied on

its demethylase activity.

Based on our sequencing data (see 3.2.1 Transcriptome-wide m6A landscape in

breast cancer), we noticed that the Wnt/β-catenin pathway was often targeted by m6A,

both in BC cell lines and in human biopsies. Thus, we wondered whether dysregulations

of m6A, through loss of FTO, could influence this signaling pathway, which is known to

play a major oncogenic role in breast and several other cancers. The canonical Wnt

pathway is summarized at Fig. 81. Briefly, on a basal level, β-catenin is a transcription

factor that is targeted for degradation by phosphorylation mediated by a multiprotein

“destruction complex” containing, among others, the GSK3 kinase. Upon binding of Wnt

ligands to their Frizzled receptors at the cellular membrane, a transduction signal leads to

the inactivation of the destruction complex, notably through inhibitory phosphorylation

of GSK3. Escaping degradation, β-catenin can then be accumulated, migrate to the

nucleus and bind to DNA to influence the expression of many genes. In BC, aberrant

activation of this pathway causes increased EMT and increased stem cell population, and

it is associated with recurrence and metastases (Lamb et al. 2013). The transcripts that

bore m6A peaks in BC cell lines included Wnt ligands (e.g. WNT3, WNT4, WNT6,

WNT8B), receptors and co-receptors (e.g. FZD1, FZD2, FZD4, FZD5, FZD6, FZD7,

Page 142: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

142

FZD8, FZD9, LRP5, LRP6), members of the destruction complex (GSK3B, PPP2CB,

BTRC), and even β-catenin itself. Thus, all levels of the signaling pathways displayed

m6A peaks.

Fig. 81: Canonical Wnt pathway.

(Left) GSK3-mediated phosphorylation

and proteolyse of β-catenin. (Right)

Binding of Wnt ligand to its receptor

leads to GSK3 phosphorylation and

inactivation of the destruction

complex. Subsequent binding of β-

catenin to DNA influences gene

expression and favors an aggressive

phenotype in breast cancer.

Therefore, we next assessed whether the canonical Wnt pathway was affected by

FTO dysregulation by measuring levels of β-catenin. First, in our FTO-knockdown cells,

we observed increased amounts of β-catenin by Western blot (Fig. 82A). This result

suggested that loss of FTO might promote the Wnt pathway. We further confirmed the

role of FTO in Wnt/ β -catenin regulation by producing a model of inducible

overexpression for FTO. This was performed in SKBR3 cells with the T-Rex™ system,

in which expression of the target is triggered by adding tetracycline in the culture medium.

We used either the wildtype form of FTO, or a catalytically inactive mutant. While

overexpression of WT FTO led to reduced β-catenin, overexpression of the mutant did

not change β-catenin levels (Fig. 82B). This finding suggested that the enzymatic activity

of FTO was required to sustain the regulation of the Wnt/β-catenin pathway.

Page 143: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

143

Fig. 82: FTO regulates β-catenin in BC. (A) Knockdown of FTO in SKBR3 cells

increased β-catenin levels, as measured by Western blot (representative of 3 independent

experiments). (B). Overexpression of wildtype FTO, but not its catalytic mutant,

decreased β-catenin, as measured by Western blot. SKBR3 cells were stably transfected

with FTO (wildtype or mutant) in Trex™ vector and overexpression was induced by

tetracycline (+Tet) (representative of 3 independent experiments).

We further examined the upregulation of the Wnt/β-catenin pathway upon loss of

FTO in SKBR3 cells. First, β-catenin downstream targets c-MYC and SOX9 were found

increased by Western blot (Fig. 83A). Secondly, Flash reporter assay showed increased

sensibility to Wnt3a treatment (Fig. 83B). Briefly, this method consists in a luciferase

reporter measurement of β-catenin-mediated transcriptional activation. Moreover, we

also observed by Western blot reduced ZO-1 levels and increased FN1 levels, two

indicators of increased epithelial–mesenchymal transition (EMT), upon loss of FTO (Fig.

83C). This result was particularly interesting given that (i) the EMT process is known to

be supported by Wnt signaling, and (ii) this event is key in the formation of metastasis

and could be linked with the increased migration and invasion we previously observed

upon loss of FTO (Fig. 79).

Page 144: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

144

Fig. 83: Loss of FTO enhances β-catenin signaling. (A) Knockdown of FTO in SKBR3

cells increased β-catenin downstream targets c-MYC and SOX9, as measured by Western

blot (representative of 3 independent experiments). (B) SKBR3 cells were treated by

Wnt3a (100ng/ml) for 6h and increased activation of β-catenin was measured by the

Flash reporter assay upon depletion of FTO (n=2). (C) Loss of ZO-1 and gain of FN1,

two signs of EMT, were observed by Western blot upon depletion of FTO (representative

of 2 independent experiments).

Finally, we asked what the clinical relevance of FTO-mediated regulation of the

Wnt pathway could be. Interestingly, a range of inhibitors targeting specifically the Wnt

pathway have been developed in recent years and are currently being investigated as anti-

cancer agents in clinical trials, notably for the treatment of BC (Lu et al. 2016). Thus, we

used iCRT3, an inhibitor that can cause BC cell death by blocking Wnt signaling at the

level of the β-catenin transcriptional complex (Bilir et al. 2013). Interestingly, SKBR3

cells appeared less sensitive to the inhibitor upon loss of FTO (Fig. 84) as increased

concentrations of iCRT3 were required in order to kill cancer cells. Most likely, high

levels of the drug were needed to override the enhanced activation of the pathway in FTO-

knockdown cells. This result suggested that expression of FTO might influence the

patient response to Wnt inhibitors and that this gene could be a potential predictive

marker.

Fig. 84: Loss of FTO affects

response to Wnt inhibitor in BC.

Control (shScramble) and FTO-

knockdown (shFTO) SKBR3 cells

were treated with increasing

concentrations of Wnt inhibitor iCRT3

for 24h. Cell viability was measured as

the ratio between the number of living

cells in the treated and untreated

conditions (n=3, data are represented

as mean ± SD).

Page 145: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

145

Taken together, our data indicated that depletion of FTO in BC led to enhance

Wnt/β-catenin pathway, which was notably associated with enhanced EMT and resistance

to Wnt inhibitors. Thus, as suggested by our previous sequencing data, dysregulations of

m6A enzymes could affect key cancer pathways.

Page 146: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

146

3.3. Key findings

This study aimed to investigate the role of m6A and its machinery in breast cancer,

and we performed our research both in cultured cells and in human breast biopsy.

Importantly, we mapped for the first time the transcriptome-wide m6A profile in BC cells

lines as well as human biopsies. Although classical features of m6A distribution are

conserved in BC, we identified nearly 2000 regions differentially methylated between

non-cancerous and cancer breast cells. Then we highlighted that one m6A enzyme in

particular, FTO, was dysregulated in BC. Downregulation of this demethylase was

associated with higher global m6A levels and poor survival in patients. Therefore, we

explored the role of FTO in vitro and we observed that loss of FTO supported an

aggressive phenotype in SKBR3 BC cells, with enhanced migration and invasion

properties, as well as increased stemness. We also unraveled dysregulations of the Wnt

pathway – a major signaling pathway involved in cancer – occurred upon depletion of

FTO. This was associated with increased EMT (a feature associated notably with

migration and invasion) and reduced sensitivity to Wnt inhibitors.

Overall, and in line with recent studies, our findings brought further evidence of

the role of m6A in cancer regulation (Cui et al. 2017; Li et al. 2017; Zhang et al. 2017).

Hence, RNA modifications and their machinery appear to be a new level of dysregulation

that can support cancer development and progression.

Page 147: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

147

Page 148: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

148

Page 149: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

149

Discussion

Discussion

Page 150: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

150

Page 151: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

151

1. Immune activation of NF-κB drives TET1

dysregulation in cancer

Malignant transformation and progression are complex processes in which several

layers of regulation are disrupted. In addition to genetic aberrations, epigenetic

mechanisms play a key role in tumorigenesis. Cancer cells undergo widespread changes

in histone and cytosine modification patterns, and DNA methylation aberrations are a

well-established hallmark of cancer. Global loss of 5mC, mainly affecting large, gene-

poor regions and repeated sequences, has been linked to DNA recombination and

genomic instability, while a gain in 5mC at gene promoters causes transcriptional

inactivation of tumor suppressor genes (TSG). For a long time, 5mC was considered a

relatively permanent mark, but this view changed abruptly. The discovery of TET-

mediated DNA hydroxymethylation as a mechanism of DNA demethylation, along with

the observation of disrupted hydroxymethylation patterns in cancer, sparked high hopes

of better understanding malignant processes.

This work of research addressed the essential question of TET and 5hmC

dysregulation in cancer, and more specifically in BC. Several previous studies have

provided insight into cancer-related 5hmC changes in various genomic regions. First

mappings uncovered wide redistribution of 5hmC in melanoma and pancreatic and liver

cancer (Lian et al. 2012; Bhattacharyya et al. 2013; Thomson et al. 2013). The existence

of vast 5hmC alterations was further confirmed in colorectal tumors, which show reduced

5hmC levels at repetitive elements and within genes, despite 5hmC distributions globally

similar to those of normal tissues (Uribe-Lewis et al. 2015). Furthermore, as shown in

leukemia, intergenic regions and enhancers in particular are also widely affected by

changes in 5hmC (Rampal et al. 2014; Rasmussen et al. 2015). In regards to previous

5hmC profiling data, our results confirmed and extended our understanding of cancer-

related 5hmC changes. Firstly, 5hmC had not been mapped in breast cancer tissues before.

In the light of the finding that 5hmC is highly tissue-specific in normal cells (Nestor et al.

2012), it is necessary to examine the diversity of 5hmC changes among tumor types.

Secondly, we observed a global anti-correlation between 5hmC and 5mC in breast cancer.

Page 152: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

152

Despite the established role of 5hmC in DNA demethylation, such a relationship was not

always straightforward in published studies. Research groups have in turn highlighted a

negative correlation (Figueroa et al. 2010; Lian et al. 2012; Yamazaki et al. 2015) or

failed to show this relation, particularly when a concomitant loss of 5mC and 5hmC was

observed at many loci (Ko et al. 2010; Uribe-Lewis et al. 2015). Extended maps may

further allow to better understand the relationship between 5hmC and 5mC in cancer.

Thirdly, we established a link between TET1 expression and 5hmC patterns in BLBC.

Our data are partly in contrast with published results, because all previous cancer-related

mappings of 5hmC were performed in a context of TET loss of function (either by

downregulation or mutation) and, overall, unraveled mainly 5hmC loss. However, we

observed that basal-like tumors with high levels of TET1 displayed almost exclusively

gains of 5hmC compared to normal breast tissues, whereas tumors with low levels of

TET1 displayed a majority of loss of 5hmC. Therefore, we provide evidence that 5hmC

dysregulations might be more diverse than formerly thought.

Our observations in regard to the 5hmC landscape in BC should however be

considered with caution, as only a limited number of samples were sequenced (4 tumors

and 4 matched normal tissues). This limits the strength of our study for two reasons.

Firstly, although the use of matched samples and pairwise analysis allows to control for

some interindividual variability, the low number of samples reduced the statistical power

of the analysis. Secondly, we could not measure the actual range of dhmRs in BC, as

observed in a heterogeneous population, with only two pairs of samples per group.

Therefore, mapping of 5hmC should be performed in a bigger cohort in order to reflect

the changes of 5hmC in BC, both extensively and accurately. In addition, we specifically

focused on the BLBC subtype. Mapping of 5hmC in other subtypes could provide

additional clues to the diversity of BC.

In regard to TET dysregulation, BLBC stand apart from most cancer types.

Decreased abundance of TET and/or 5hmC has been observed, both in solid and liquid

tumors, and multiple mechanisms interfering either directly or indirectly with TET

expression have been identified, notably involving the transcriptional repressor HMGA2

(Sun et al. 2013), various miRNAs (Song et al. 2013; Cheng et al. 2013; Chuang et al.

2015), or methylation of the TET1 or TET2 gene (Chim et al. 2010; Kim et al. 2011;

Cimmino et al. 2015). However, a study by Huang et al. has demonstrated that, in MLL-

rearranged leukemia, upregulation of TET1 expression by various MLL fusion proteins

Page 153: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

153

led to increased expression of oncogenic target genes (Huang et al. 2013). As this finding

contrasts starkly with the many reports of impaired TET expression in several

malignancies, it would seem that TET proteins can play diverse roles in tumorigenesis

and thus be differently dysregulated depending on the context. The wide range of TET1

expression observed in BLBC, including both increased and decreased TET1 levels,

brings evidence supporting the idea that TET dysregulation in cancer is more complex

than previously thought. It is worth noting that the BLBC subtype displays a particular

chromatin environment with lower 5mC levels at many loci than the other BC subtypes,

including TET1 promoter. This lack of methylation could support the higher expression

of TET1 in a subset of BLBC, however some tumors display low TET1 expression despite

the hypomethylated status of its promoter, as observed in the TCGA cohort (Koboldt et

al. 2012). Therefore, we took advantage of this particular subtype to investigate new

mechanisms of TET regulation.

As TET downregulation is observed in nearly all cancer types, we speculated that

alterations in signaling pathways frequently associated with tumors could play a role in

this regulation. Therefore, investigating the potential association between TET1

expression and major signaling pathways in BLBC, we established an unprecedented

association with immune pathways. In the BLBC subtype, high TET1 expression was

exclusively observed in tumors with low expression of immune genes and low infiltration

by major immune populations, including B lymphocytes, CD4+ and CD8+ T

lymphocytes and macrophages. This is of most interest, because the immune system has

emerged in recent years as a prominent feature of cancer. Many cancer types, including

BLBC, display a striking infiltration by immune populations. Cancer infiltration by

immune cells is, at least in part, due to secretion of recruiting factors by the cancer cells

themselves, and has major impacts in terms of disease progression and response to

treatment (Mantovani 2010). The immune system displays a dual action in cancer

(Lakshmi Narendra et al. 2013). On one hand, secretion of pro-inflammatory factors has

been shown to enhance cancer progression and to favor resistance to treatment. On the

other hand, tumor immune response, and in particular tumor-infiltrating lymphocytes

(TILs), are increasingly recognized to be associated with better clinical outcome in many

cancers (Galon et al. 2006; Oble et al. 2009; Melichar et al. 2014). In addition, the recent

emergence of immunotherapy (e.g. PD-L1 and PD-1 inhibitors) as promising tools to

prevent cancer from escaping destruction by the immune system has brought a new hope

Page 154: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

154

in terms of cancer treatment (Dany et al. 2016; McArthur 2016). Thus, the immune system

is a major component of the TME with important clinical relevance, in breast and other

cancers. Accordingly, we extended our findings to other cancer types, including

melanoma, thyroid and lung cancers, supporting the idea that immune regulation of TET1

occurs in most cancer types.

In recent years, epigenetics and immunity have been increasingly interwoven in

the context of cancer. DNA methylation, in particular, has been shown to be altered in

relation with the immune infiltration of the tumor (Dedeurwaerder & Fuks 2012). In this

study, however, we highlight a new dimension in the epigenetics-immunity connection,

i.e. the immunity-driven repression of TET1 in cancer cells. Thus far, TET enzymes have

only been involved in the regulation of immune cells themselves. In T regulatory (Treg)

cells, for instance, TETs promote FOXP3 expression and Treg-cell-associated immune

homeostasis (Yang et al. 2015). In myeloid cells, Tet2 has been shown to repress the

proinflammatory cytokine IL6, thus controlling inflammation (Zhang et al. 2015). Tet2

has also been found to promote the activation of cytokine genes in CD4+ T cells

(Ichiyama et al. 2015). Additionally, TET1 was reported as an epigenetic regulator

involved in Th2 differentiation (Yang et al. 2016). In contrast to previously published

literature, we demonstrate that the link between immune pathways and TETs extends

beyond the immune system itself. We highlight a new paradigm in which the immune

system can influence cancer cell epigenetics, specifically by NF-κB-dependent repression

of TET1. This is of great importance as epigenetic drugs have been shown to modulate

the anti-tumor immune response (Roulois et al. 2015; Chiappinelli et al. 2015) and

dissecting the epigenetic mechanisms underlying the cross-talk between the immune

system and cancer could help optimize therapeutic strategies (Jeschke et al. 2017).

In this study, we specifically linked TET1 repression with the immune regulator

NF-κB. In order to establish the involvement of this transcription factor in TET1

repression, we activated NF-κB through three different methods: overexpression of the

p65 subunit, LPS treatment and TNF treatment. We also blocked the activation of the

pathway with an inhibitor of the proteasome, MG-132, before TNF treatment to prevent

TET1 repression. However, it is worth mentioning that LPS and TNF influence several

pathways and that not all of their effects can be attributed to NF-κB activation. Likewise,

MG-132 is not a specific inhibitor of NF-κB, as it acts on the entire proteasome function.

A stronger evidence of the implication of NF-κB could be provided by using specific

Page 155: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

155

inhibitors. For instance, a dominant-negative mutant of IKK2 has been engineered and

can block the activation of the pathway (Mercurio 1997). This would also provide the

reverse experiment of the phenotype observed in IKMV mice, where NF-κB activation

was in fact performed through expression of a constitutively active mutant of IKK2 in the

mammary gland. Likewise, it is also possible to take advantage of non-degradable forms

of IκB, bearing mutations on the phosphorylation and ubiquitylation sites targeting the

protein to the proteasome, in order to prevent NF-κB activation (Gupta et al. 2010). Using

these controls could provide stronger evidence of the role of NF-κB in TET1 regulation.

NF-κB is a major transcription factor, known to be activated in many cancer types

(Ben-Neriah & Karin 2011). It is mostly considered as pro-tumorigenic, particularly when

its activation is associated with an inflammatory context, however NF-κB signaling can

also be a marker of the immune response targeting malignant cells (Chen et al. 2008;

Hoesel & Schmid 2013). In the context of TET1 regulation, we highlighted the role of

NF-κB member p65. In general, p65-p50 dimer is widely recognized to promote the

expression of many cytokines and chemokines, while the repressive activity is more often

attributed to p50-p50 homodimer. Nevertheless, a repressive effect of p65-p50

heterodimer has been previously reported as well. For instance, p65 can recruit DNMT1

leading to repression of the tumor suppressor gene BRMS1 by promoter hypermethylation

(Y. Liu et al. 2012). In another study, histone deacetylase HDAC1 and HDAC2, both

known as co-repressors of gene expression, were shown to interact with NF-κB p65 and

could repress NF-κB-regulated genes (Ashburner et al. 2001). Alternatively, p65-

mediated repression can also occur as a consequence of promoter occupancy and

subsequent steric hindrance. The exact molecular mechanisms involved in TET1

regulation remains to be uncovered.

Interestingly, gene regulation by NF-κB is another example of bidirectional cross-

talk between epigenetics and immune signaling (Vanden Berghe et al. 2006). Firstly, and

as mentioned above, epigenetic enzymes can act as co-effectors in the NF-κB

transcriptional complex. And while NF-κB-responsive elements in promoters are mostly

responsible for the recruitment of the NF-κB complex, the response patterns can vary in

terms of kinetics and quantity, depending on the context. Modulation of the NF-κB

response is, at least in part, achieved through epigenetic alteration of the chromatin. Both

DNA methylation and histone modifications have been widely involved in NF-κB

signaling (Zhou et al. 2013; De Andrés et al. 2013). Furthermore, epigenetic factors have

Page 156: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

156

been linked to the regulation of NF-κB transcriptional activity through post-translational

modifications of its subunits. For instance, HDAC-mediated deacetylation of p65 is

critical for orchestrating the transcriptional program of NF-κB and can play a critical role

in cell differentiation (Chen et al. 2011). Thus, epigenetic factors are both downstream

effectors and upstream modulators of the NF-κB signaling pathway. Our study uncovered

a new dimension of this relationship. Our findings notably raise the question of the

potential role of TETs and 5hmC regulation in the modulation of NF-κB-mediated gene

regulation.

Fig. 85: Proposed model illustrating the

immune regulation of TET1 in cancer. (1) In

tumors with high immune infiltration and

expression of immune mediators, the canonical

NF-κB signaling pathway is activated in cancer

cells. (2) Nuclear translocation of p65 allows

p65-p50 NF-κB dimers to bind to TET1 gene

promoter. (3) Binding of NF-κB dimers to TET1

promoter leads to reduced expression of the gene.

Our work also opens new avenues in terms of perspectives. Firstly, the molecular

mechanisms involved in this regulation should be further examined. Notably, the

potential recruitment of co-repressor(s) by p65 could be demonstrated through

streptavidin-agarose assay followed by protein mass spectrometry. The modulation(s) of

the chromatin state of TET1 promoter upon NF-κB activation could also be studied by

ChIP-qPCR (e.g. targeting the main histone modifications or other features, depending

on the previously identified co-repressors). We should also keep in mind that different

mechanisms might be involved in different cancer types. Secondly, and importantly, in

light of the link between TET1 repression and immunity, as well as the prominent role of

immune infiltration in predicting clinical outcome, the relevance of TET1 expression in

cancer should be reconsidered. So far, loss of TET function has almost exclusively been

considered as a marker of bad prognosis, linked with their described tumor suppressive

role. However, our data demonstrate that this loss can also be associated with high

Page 157: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

157

immune infiltration, including of TILs, which is recognized as a marker of better clinical

outcome. Hence, further investigation should be completed in order to clarify this

apparent discrepancy. In that regard, it is important to recall that the role of TETs in

cancer might not be as one-dimensional as first thought, and that several studies have

brought an intriguing duality to light by displaying TET enzymes and 5hmC as potential

promoters of cancer (Huang et al. 2013; Ahsan et al. 2014; Navarro et al. 2014). It would

be particularly relevant and interesting to evaluate the effects of TET modulation (e.g. by

exogenous overexpression or by knockdown/knockout) in cancer cells in a context of

immune activation. For instance, would loss of TET function exacerbate the pro-

tumorigenic effect of inflammatory signaling or influence the recruitment of immune

cells to the TME? The potential impact of TET1 regulation on immune signaling remains

unknown.

In conclusion, our data assign a novel function to NF-κB, which is known to

orchestrate immune and inflammatory responses, as well as oncogenesis (Oeckinghaus &

Ghosh 2009; Ben-Neriah & Karin 2011). Although identified in BLBC, NF-κB-mediated

repression of TET1 appears to be a mechanism shared by many cancer types. Our findings

set the stage for future studies on TET dysregulation in cancer, in placing the immune

system under the spotlight and unraveling a new facet in the relationship between

immunity and cancer epigenetics.

Page 158: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

158

Page 159: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

159

2. RNA hydroxymethylation: a new player in the

game

A new area of epigenetic research is on the rise: RNA epigenetics. This field has

increasingly intrigued scientists due to the numerous existing modifications identified in

RNA, as well as their involvement in the regulation of major biological processes, such

as stem cell pluripotency and differentiation (Geula et al. 2015; Zhang et al. 2016).

Chemical modifications play an important role in modifying and regulating the function

of RNA. Until recently, attention had been focused on tRNA modifications and their

impact on protein translation (Torres et al. 2014). However, other classes of RNA can

also be modified, and these modifications impact many aspects of RNA metabolism, aside

from translation (e.g. stability, splicing, or localization). Of particular interest, mRNA is

now acknowledged to also bear several chemical modifications, and this is important

since they constitute a regulatory layer positioned between DNA and proteins. Thus,

although our understanding of the post-transcriptional modifications that decorate RNA

is still in its infancy, RNA epigenetics shows much promise for the future.

In this context, we have decided to tackle the challenge of yet uncharacterized

RNA modifications. Specifically, we have conducted a pioneer study addressing cytosine

hydroxymethylation in RNA, using Drosophila melanogaster as a model. During the

preparation of our manuscript, two studies demonstrated the existence of 5hmrC in RNA

(L. Fu et al. 2014; Huber et al. 2015). The first report indicated that 5hmrC occurred in

mammalian RNA, though at much lower frequency than 5mrC (approximately in a

1:5000 ratio), according to mass spectrometry. Through in vitro assays and

overexpression of Tets in HEK293T cells, they also demonstrated that Tet enzymes could

oxidize 5mrC into 5hmrC. The second report described the existence of 5hmrC across all

three domains of life (i.e. archaea, bacteria and eukaryotes). They also showed that 5hmrC

was enriched in mammalian poly A RNA, with up to 10 times more 5hmrC than in total

RNA. Therefore, these two studies set the bases for the investigation of a new RNA

modification by demonstrating the existence of 5hmrC and suggesting that it might be

part of a dynamic regulatory mechanism. Yet, these findings relied solely on global

Page 160: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

160

quantification and did not address the distribution and function of 5hmrC. Therefore, our

work confirmed and extended prior observations by (i) providing the first transcriptome-

wide picture of 5hmrC distribution, (ii) assigning a novel function for 5hmrC in

translation regulation, and (iii) suggesting a central role for this RNA modification and

dTet in the Drosophila brain.

An important issue in the study of any new epigenetic modification is the

identification of its enzymatic machinery. Results from our study and from Fu et al.

indicated that Tet enzymes are responsible for RNA hydroxymethylation. However,

several pieces of evidence also suggested that other mechanisms might be implicated in

5hmrC formation. Firstly, Fu et al. observed that while 5hmrC levels are reduced in Tet-

null ES cells, appreciable levels of the modification remained detectable by mass

spectrometry (approximately 50% of wildtype ES cells). Similarly, in the Drosophila

brain, approximately 40% of 5hmrC remained detectable upon knockout of dTet (Fig.

67). Finally, Huber et al. found 5hmrC in the RNA of species that lack both DNA

hydroxymethylation and TET homologues in their genomes, such as C. elegans and A.

thaliana. Taken together, these results suggest that the formation of 5hmrC in RNA can

also occur through a non-TET mechanism. In this context, two hypotheses could be

considered: (i) oxidation of 5mrC into 5hmrC might be induced by cellular reactive

oxygen species in a non-specific manner, or (ii) other enzyme(s), yet unidentified, might

have the capacity to oxidize 5mrC, in a reaction similar to TET-mediated RNA

hydroxymethylation. While we cannot formally exclude the first option, it seems unlikely

that the remaining 5hmrC signal derives merely from random oxidation, since we were

able to detect hundreds of regions significantly enriched for 5hmrC after depletion of dTet

in S2 cells. This result suggests that 5hmrC distribution was not random even in dTet-

depleted cells. In regard to the second option, a search for a Tet-like enzyme in Drosophila

by BLAST (Basic Local Alignment Search Tool) analysis did not provide any obvious

candidate, thus the identity of the putative enzyme(s) remains unknown. However, we

also cannot entirely reject the possibility that another type of oxidase enzyme might be

involved.

Another issue raised by our results is the tissue-specificity of 5hmrC. In our study,

we observed that both 5hmrC and dTet were enriched in the central nervous system, as

compared to other tissues, such as salivary glands and ovaries (Fig 65B). Similarly, Fu et

al. measured 5hmrC in several mouse tissues and observed the highest levels in the heart

Page 161: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

161

and the lowest in the pancreas. Interestingly, and in contrast with results from the

Drosophila, mouse brain only displayed intermediary levels of 5hmrC. These results

suggest not only that the amount of 5hmrC differs between tissues, but also that the tissue

distribution might be species-dependent. Beyond global quantification of 5hmrC, we can

expect that this tissue-specificity would also extend to local enrichments. In line with this

idea, several other epigenetic modifications have been shown to display highly tissue-

specific distributions, including DNA methylation and hydroxymethylation (5mC,

5hmC), as well as RNA methylation (m6A) (Meyer et al. 2012; Jeschke et al. 2015;

Ponnaluri et al. 2017). Therefore, the diversity of 5hmrC distribution across different

tissues should be examined by transcriptome-wide sequencing in the years to come. In

that regard, a first clue was provided by the sequencing of 5hmrC we performed in breast

cell lines as a follow-up of our first study in Drosophila. While the majority of 5hmrC

was found in coding regions in S2 cells, the mark showed a preferential enrichment in

intronic regions in the human breast cells, thereby displaying some discrepancies in the

general distribution of 5hmrC between the two types of sample. This example illustrate

how profiling 5hmrC in different tissues and species would help us understand whether

5hmrC peaks are conserved or display evolutionary divergence.

While our research in Drosophila uncovered several interesting features of 5hmrC,

this work constituted the first fundamental study on this “new” mark, which raised even

more questions and possibilities in terms of perspectives. These are discussed in the next

paragraphs.

Firstly, we have discovered that 5hmrC can favor protein translation in mRNA.

However, we cannot exclude that this modification has other regulatory effects, and

further transcriptome-wide analysis should be performed to test this hypothesis. For

instance, coupling 5hmrC-mapping with paired-end RNA sequencing would allow to

investigate a potential role of the modification in RNA splicing regulation (Rossell et al.

2014). It is worth noting that the intronic enrichment of 5hmrC in breast cells raised the

possibility of a potential role in splicing regulation in these cells. Moreover, to assess the

effect of 5hmrC on RNA stability, the half-life of transcripts could be measured by

treating the cells with actinomycin D (which blocks transcription) for several hours before

performing RNA sequencing (Ayupe & Reis 2017). Another way to get an insight into

the functions of 5hmrC would be the identification of potential readers. This might be

performed by pulldown of protein extracts with biotinylated RNAs (in vitro transcribed

Page 162: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

162

and containing 5hmrC), followed by proteomic mass spectrometry. A similar method

(relying on DNA probes) was used for the identification of 5mC and 5hmC readers

(Spruijt et al. 2013). And, as for tissue distribution, one should keep in mind that distinct

readers might be identified in various tissues or species. Overall, these different

approaches would extend our understanding of the role of 5hmrC in RNA metabolism.

Secondly, the discovery of a new epigenetic modification naturally raises the

question of its potential involvement in mechanisms related to human health and disease.

Two biological contexts stand out in particular: neurogenesis and cancer.

The first arises from our observation that dTet-null fruit flies displayed impaired

brain development, as well as the implication of mammalian Tets in neuronal functions

(Guo et al. 2011; Xin et al. 2015; Hsieh et al. 2016). In mammals, the role of Tets in

neuronal development and homeostasis has so far been attributed to 5hmC and DNA

demethylation. However, in light of the discovery that RNA can also be

hydroxymethylated by Tets, this notion should be reevaluated. In Drosophila, many RNA

transcripts related to the nervous system were found hydroxymethylated. Should this

feature be conserved in mammals, it would suggest that 5hmrC might also be involved in

Tet-mediated regulation of neurons.

Likewise, the choice of cancer as a potential model to study 5hmrC comes from

the fact that Tets are widely dysregulated in cancers in mammals, as was discussed in a

previous chapter. This idea is also backed up by our 5hmrC sequencing data. In cell lines,

we have shown that 5hmrC displays a wide redistribution between non-cancerous breast

cells (MCF12) and BC cells (MDA-MB-231), with over 700 differentially

hydroxymethylated RNAs. These transcripts were associated with key BC genes, such as

BRCA2 and p53. Further analyses are required to confirm and extend these results;

however, our observations suggest a major dysregulation of 5hmrC in breast cancer.

However, these results are very preliminary data that should only be considered as a pilot

study, performed to evaluate the feasibility of studying 5hmrC in breast tissues. Although

we were able to measure and map 5hmrC in breast cells, the amount of RNA required for

such analysis (about 1mg) far exceeds what can be reasonably obtained from human

breast biopsies, and especially from normal tissues which only yields very limited

amounts of material. This constraint forced us to rely on breast cell lines at this stage, for

which sufficient amplification was possible. Furthermore, MCF12 cells are, admittedly,

Page 163: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

163

not a perfect control as the cells are immortalized and differ from actual normal breast

cells, notably in terms of karyotype and proliferation rate. However, they are non-

tumorigenic and were the best available control. In order to truly measure 5hmrC in BC

and normal breast cells, the technology of hMeRIP must be improved to allow the

detection of 5hmrC from smaller amounts of RNA. Nevertheless, should our results be

confirmed, gaining insight into the effects of 5hmrC on RNA metabolism (e.g. in terms

of splicing, stability, etc.) would also help understand better the consequences of such

dysregulations in regard to tumorigenesis. Hence, mapping 5hmrC could provide a new

dimension of cell regulation in contexts related to human health and disease.

A major challenge in the study of TET enzymes will be to discriminate the effects

linked to DNA and RNA hydroxymethylation. In terms of phenotype, it is likely that both

levels of regulation contribute to TET-mediated effects in a variety of cellular contexts.

There is no known specific modulator, and thus it would be almost impossible to

distinguish the effects of the marks in vivo. In contrast, in vitro experiments can at least

allow to study the functions of these marks separately. Briefly, 5hmC mostly acts at the

transcriptional level and its dysregulation is expected to affect the expression of RNA

transcripts. However, 5hmrC constitutes a post-transcriptional level of regulation. Our

results indicate that dysregulation of this mark could affect translation, and we also

suspect that it might affect RNA splicing and/or stability. We mentioned in a previous

paragraph the techniques that would allow the study of these potential functions. Hence,

while we cannot, as of today, isolate the roles of 5hmC and 5hmrC on a given phenotype,

we can at least explore the effects of these modifications through their functions and,

accordingly, speculate on related mechanisms. Furthermore, it is worth mentioning that

an RNA binding domain was recently identified in TET2 (He et al. 2016). Similar

domains can also be found in TET1 and TET3. It has not yet been shown that these

domains are required for RNA hydroxymethylation. But, should it be the case, deletion

of these domains could prevent RNA hydroxymethylation and thus provide a tool to study

5hmrC specifically. This concept is, admittedly, at the level of speculation, but it

constitutes an intriguing possibility.

Page 164: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

164

In conclusion, our work in Drosophila has yielded the following key findings: (i)

the first mapping of the hydroxymethylated transcriptome, (ii) a newly identified function

for 5hmrC, and (iii) a central role for this RNA modification and dTet in the Drosophila

brain. All in all, we expect this fundamental study to change the way we think about the

roles played by cytosine hydroxymethylation and the Tet proteins. Our findings in breast

cells also open new research prospects for the study of health and disease by drawing

attention to an emerging realm of biological regulation: epitranscriptomics.

Page 165: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

165

3. Dysregulations of m6A and its machinery

support breast cancer

Discovered in the 1970s (Desrosiers et al. 1974), m6A is the most abundant

modification of mRNAs, detected in a wide range of species, ranging from bacteria to

eukaryotes (Deng et al. 2015; Wu et al. 2016). Nevertheless, the study of m6A

modifications had long been stalled, owing to the lack of understanding of its machinery

and the short half-life of most RNAs. The situation changed recently, however, with the

identification of m6A methyltransferases and demethylases, as well as the development

of new technologies, including high-throughput sequencing of the methylated

transcriptome (Dominissini et al. 2012; Meyer et al. 2012; Maity & Das 2016). In

particular, the recognition of m6A as being part of a dynamic process has reignited the

interest of researcher in RNA epigenetics and has shaken many concepts of cytogenetics

related to RNA metabolism.

The present work of research addressed the topic of dysregulations of m6A and

its machinery in cancer, and more specifically in BC. In that regard, three recent studies

have provided insight into the landscape of cancer-related m6A changes and the impact

of associated enzymes on tumorigenesis. In 2016, Li et al. reported that FTO, an m6A

demethylase, was highly expressed in MLL-rearranged acute myeloid leukemia (AML),

and that the enzyme functioned as an oncogene promoting leukemogenesis. This pro-

tumor function was notably mediated by m6A-dependent regulation of ASB2 and RARA

transcripts, two key regulators of hematopoiesis (Li et al. 2016). Likewise, two groups

recently assigned oncogenic functions to the m6A demethylases, FTO and ALKBH5, in

glioblastoma. Both studies showed that m6A regulated the self-renewal, proliferation and

tumorigenesis of glioblastoma stem cells by controlling the expression of key cancer

factors, such as ADAM19 and FOXM1 (Cui et al. 2017; Zhang et al. 2017). In contrast

with previous studies, our results assigned a tumor suppressive function to FTO, as we

demonstrated that FTO was depleted in BC, which was associated with poor survival,

increased global m6A levels and an aggressive phenotype in vitro. Hence, our findings

bring a novel duality to light, as m6A demethylases appears to display both oncogenic

Page 166: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

166

and anti-tumor activity, depending on the cancer type and the cellular context. Such

dichotomy is not unheard of for epigenetic enzymes. For instance, EZH2, a member of

the Polycomb repressive complex 2 promoting H3K27 methylation, has in turn showed

pro- and anti-tumor functions in various cancer contexts (Herviou et al. 2016). We also

discussed TET enzymes in a previous chapter, whose tumor suppressive role has been put

into question by several studies (Huang et al. 2013; Ahsan et al. 2014; Navarro et al.

2014). These examples, as well as our findings about FTO in mammary tumors, illustrate

that the effects we attribute to enzymes should always be considered within a biological

context.

It is worth mentioning that, on a global level, changes in m6A upon modulation

of its related enzymes can be modest in terms of fold-change, despite substantial

phenotypical effects. For instance, m6A levels increased approximately by 30% upon

ALKBH5-knockdown in glioblastoma cells (Zhang et al. 2017). In breast cells, we

observed a similar increase of m6A upon FTO-knockdown by dot blot analysis. In line

with these findings, studies have previously reported in various cell types that modulation

(either knockdown or overexpression) of m6A methyltransferases or demethylases could

lead to changes in global m6A levels ranging from 10% to 40% approximately, depending

on the targeted enzyme and the cell type (Jia et al. 2011; Liu et al. 2014). In comparison,

overexpression of TET1 alone increased 5hmC levels by 10-fold in DNA (Tahiliani et al.

2009). Thus, changes of m6A in RNA, following modulation of related enzymes, can be

relatively subtle compared to other epigenetic marks; nevertheless they are associated

with major phenotypic changes in cancer. It is possible that m6A dysregulation in few

key transcripts would be sufficient to carry out the observed effects.

In regard to the transcriptome-wide landscape of m6A, the three previously

mentioned studies (Li et al. 2016; Cui et al. 2017; Zhang et al. 2017) found the classical

known features of m6A to be conserved in cancer, such as enrichment near the stop codon

and the 3’UTR and expected binding motifs (G. Luo et al. 2014; Geula et al. 2015). We

observed similar features in breast cell lines, suggesting that the general topography of

m6A distribution is robust in cancer. Yet, in comparison to these other studies, our

profiling of m6A stands out for two reasons. Firstly, we started our investigation by

comparing cancer cells to their non-malignant counterpart. This is the first time such a

comparison is drawn, since all previous differential m6A analyses in cancer instead

involved the direct modulation of the m6A machinery in a cancerous cell line. In contrast,

Page 167: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

167

we took advantage of the MCF12 cell line, which is a model for non-cancerous breast

(Paine et al. 1992), in order to draw a picture of transcriptome-wide changes of m6A in

BC. As previously mentioned, the MCF12 cells are immortalized and differ from true

normal breast cells. Similarly to our study of 5hmrC, the decision of using this cell line

was justified by the amount of RNA material currently required for MeRIP-seq (about

150µg), which exceeds what can be reasonably obtained from reduction mammoplasty.

Hence, our analyses should be considered with caution, and mostly as a first approach of

possible m6A dysregulations in breast cancer. Mapping of m6A in genuine normal breast

cells will require an improvement of the technology allowing lesser amount of starting

material.

The second characteristic that sets our study apart is our sequencing of m6A in

human breast tissues, i.e. obtained from patient biopsies rather than cultured cells. As

mentioned in the previous paragraph, sequencing of human tissues is not trivial due to the

amount of material required for m6A mapping. In fact, the vast majority of published

m6A profiles in mammals are from cultured cells (e.g. ES cells or cancer cell lines such

as HeLa or HEK293T), and the rare sequencing experiments performed on RNA extracted

from tissues are all from mice (Meyer et al. 2012; Dominissini et al. 2012; Batista et al.

2014; Zhao et al. 2014; Geula et al. 2015). Hence, our sequencing of breast tumors was,

to our knowledge, the first mapping of m6A in any human tissue. These data allowed us

to assess the relevance of cultured cells as a model for the study of m6A, and we found

that BC cell lines appeared to conserve the main features of human breast tumors, both in

terms of m6A-targeted transcripts and signaling pathways. For all the above-mentioned

reasons, our study presents strong evidence to support the notion that m6A represents a

new layer of dysregulation in BC. Nevertheless, the true validation of our hypothesis

should come from the comparison between BC tissues and normal breast tissue. As

mentioned previously, we could not provide the latter for technical reasons. We expect

however that the MeRIP-seq technologies will improve in the coming years and render

possible the mapping of m6A from human biopsies.

It is important to note that identifying cancer-related m6A dysregulations can have

clinical implications. For instance, Cui et al. demonstrated that in glioblastoma – where

m6A demethylases displayed oncogenic functions – an FTO inhibitor suppressed cancer

progression and prolonged lifespan of xenografted animals, suggesting that targeting the

m6A machinery is a promising therapeutic tool (Cui et al. 2017). The use of such inhibitor

Page 168: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

168

would most likely not be beneficial in BC however, as FTO appeared to carry out anti-

tumor activity instead. Nevertheless, we linked in vitro downregulation of FTO with

reduced sensitivity to iCRT3, a Wnt inhibitor, in BC cells. Furthermore, several key

cancer pathways showed differential RNA methylation between non-cancerous and

cancer cell lines, including Wnt and stemness signaling. Therefore, while FTO might not

be a relevant therapeutic target in breast, dysregulations of FTO and m6A might bear

great potential as future prognostic and/or predictive biomarkers.

Interestingly, FTO had been suspected to play a role in BC long before its

demethylase activity was discovered, as genetic variants had been associated with

increased risk of BC (Kaklamani et al. 2011). Importantly, this observation was originally

thought to be linked with obesity. Indeed, in genome wide association studies (GWAS),

single nucleotide polymorphisms (SNPs) in intron 1 of FTO (namely, the Fat Mass and

Obesity-associated protein) had been connected to obesity and type 2 diabetes (Grant et

al. 2008; Thorleifsson et al. 2009; Cho et al. 2009). Studies in mice have since suggested

that FTO was involved in adipogenesis and weight gain regulation (Church et al. 2010;

Farooqi 2011; Merkestein et al. 2015). This notion was confirmed and extended in a study

by Zhao et al., who demonstrated that m6A-mediated regulation of mRNAs by FTO was

involved in adipogenesis (Zhao et al. 2014). Therefore, researchers assumed that the role

of FTO in BC was mostly owed to obesity and dysregulations in adipocyte functions, as

it is known that obesity causes adipocyte hypertrophy in breast, which is in turn associated

with increased release of pro-tumor inflammatory cytokines and adipokines (Monteiro &

Azevedo 2010). Yet, our data partly contradict this hypothesis. Notably, depletion of FTO

in in vitro cultured breast cells enhanced an aggressive cancer phenotype (i.e. migration,

invasion and stemness of BC cells), which supports the idea of an intrinsic, or cell-

autonomous, mechanism – that is, independently of effects mediated by adipocyte

hypertrophy in breast. Thus, although the influence of FTO on adipocyte homeostasis is

undeniable, downregulation of the enzyme in breast cells is sufficient to promote BC.

Whether FTO-mediated changes in adipocytes has an additional causal effect on BC or is

merely a confounding factor remains to be determined.

Page 169: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

169

Fig. 86: FTO, breast cancer and

adipocytes. Our data indicate that

FTO dysregulations intrinsically

affects the cancer phenotype of

breast cells. It is also well-

established that FTO regulates

adipocytes. And while it has been

suspected that the latter might also

influence BC development, this

hypothesis has not been proven yet.

While our research uncovered a new level of dysregulation in breast cancer

through m6A and assigned a tumor suppressive role to FTO, several major points have

yet to be addressed as perspectives.

Firstly, we have demonstrated that loss of FTO affects BC cells in vitro, but these

findings should be extended to in vivo experimentation. As mentioned previously, in

tissues, tumor cells do not develop alone – they constantly interact with their tumor

environment (TME), and the effects of these interactions cannot be properly characterized

through in vitro models. Therefore, we plan to assess the effects of FTO downregulation

in vivo in a xenograft model. Similar experiments were performed, for instance, by Cui

et al. to demonstrate the oncogenic role of FTO in glioblastoma (Cui et al. 2017). To

further investigate the role of FTO in BC, patient-derived xenograft (PDX) experiments

could also be conducted. Interestingly, PDX-derived tumors have been shown to display

a high degree of concordance with the original human tumor, because both the supporting

stroma and the cancer cells are maintained, contrary to cultured cell lines (Isella et al.

2015).

Secondly, and importantly, we have yet to fully characterize the molecular

mechanisms supporting the enhanced aggressiveness observed upon knockdown of FTO

in BC. So far, we have indications that this phenotype could be related to Wnt/β-catenin

pathway dysregulations and that it might be m6A-dependent. In order to validate our

hypothesis, the exact transcripts that are regulated by FTO through m6A changes must be

identified. Afterwards, we must also determine how they are related to the phenotype in

terms of mechanisms. Sequencing of m6A in FTO-knockdown cells would determine

which transcripts are altered. Changes in m6A should also be confirmed by m6A-qPCR,

as sequencing methods can present biases in terms of quantification. Further sequencing

Page 170: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

170

experiments can be performed in parallel to investigate the effects of m6A modulation on

RNA metabolism at the global level (e.g. stability, splicing, or translation – see

perspectives related to the study of 5hmrC for further explanation). Finally, once key

m6A targets are identified, their implication in the phenotype could be demonstrated

through rescue by overexpression or downregulation of said targets. For instance, if

transcripts of the “β-catenin destruction complex” (such as GSK3B or Axin2) display

changes in m6A, we would first explore the consequences for these targets in term of their

stability, splicing, and translation. Then we would mimic m6A-mediated regulation by

either downregulating or overexpressing the targets (depending on the effects observed)

and assess whether the β-catenin levels and the cancer phenotype (in terms of migration,

invasion, and tumorsphere formation) are affected as in the FTO-knockdown cells. These

experiments are needed to demonstrate that m6A regulation is truly responsible for the

observed phenotype, as well as to identify the underlying mechanisms. They would also

allow us to refine our model of the effects of FTO downregulation in breast cancer.

Fig. 87: Proposed model illustrating effects of FTO depletion in BC. Reduced FTO

expression in breast cancer cells is associated with a global increase in m6A levels and

changes in RNA regulation. These changes lead to enhanced activation of the Wnt/β-

catenin pathway, which supports an aggressive phenotype in BC, as observed in vitro for

migration, invasion and tumorsphere formation.

In conclusion, our study of m6A dysregulations in breast cancer has yielded the

following key results: (i) a transcriptome-wide portrait of cancer-related changes of m6A

in breast cells; (ii) the discovery of a novel anti-tumor function for FTO; and (iii) a

connection between FTO and a major cancer pathway, i.e. the Wnt/β-catenin signaling

pathway. Our findings support the notion that RNA epigenetics constitutes a new level of

molecular dysregulation in cancer, as suggested by recent studies (Li et al. 2016; Cui et

al. 2017; Zhang et al. 2017), while also bringing to light a duality in the role of the m6A

machinery.

Page 171: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

171

4. Concluding remarks

Throughout this PhD thesis, we had the opportunity to explore three epigenetic

modifications – and their related enzymes – that have been either identified or

characterized within the past decade: DNA hydroxymethylation (5hmC), RNA

hydroxymethylation (5hmrC) and RNA methylation (m6A). The field of epigenetics is a

domain in constant evolution, and, taking into account the growth of the known epigenetic

repertoire, the scientific community can no longer be satisfied by simply looking at the

“classical” epigenetic modifications (i.e. DNA methylation and histone modifications) as

a reflection of the epigenome.

Our research work was mainly focused on breast cancer, a disease that remains

the most frequent malignancy in women. A legitimate question would be the clinical

relevance of investigating all these “new” epigenetic modifications in cancer.

Considering the coming of age of the next-generation sequencing methods, both the

genome and the transcriptome of a tumor can now be profiled in a very limited amount

of time. One could thus doubt the interest of studying such epigenetic modifications, when

genetic and transcriptomic aberrations can be easily examined. In the last chapter of our

discussion, we wish to take a step back from our own research and consider the global

picture of current oncology field. Because the answer to the question of the clinical

relevance of epigenetics is both simple and extraordinary complex: profiling the genome

and the transcriptome of malignant cells, while certainly beneficial, is not a magic bullet

for cancer management.

Let us begin with an optimistic note: cancer treatment, including for breast cancer,

has improved tremendously over the past 40 years, and survival rates have never been as

high as they are nowadays (WHO 2012). This progress results mostly from the

combination of two occurrences: (i) the growth and improvement of the therapeutic

arsenal, and (ii) the use of relevant biomarkers to guide the choice of therapy. With the

side by side evolution of these two aspects of cancer management, we are gradually

coming to an age of personal medicine where each patient and each tumor is considered

as unique. Hence, in order to continue to improve patient survival, the major challenge of

Page 172: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

172

modern oncology lies in finding the optimal treatment for each patient, as response to

treatment varies greatly between patients with the same cancer.

Variation in cancer response can be explained by differences in the molecular

profile of tumors. Already, profiling DNA mutations and the RNA transcriptome are

becoming accepted in oncology as a means to help guide clinicians and sequencing

protocols are being adapted to standard clinical practice (Gagan & Van Allen 2015).

However, cells are widely dysregulated in tumors, and multiple levels of regulation have

been implicated: genetic mutations and genomic defects, changes in gene expression, loss

of heterochromatin, abnormal DNA methylation, changes in histone modifications,

altered RNA interference, etc. To complicate matters even more, malignant cells can alter

their TME and these dysregulations can evolve over time. Thus, cancer is an extremely

complex disease, which explains the difficulty of finding a cure: whenever a treatment

targets a certain alteration, another level of dysregulation can take over and present a path

towards resistance. Accordingly, combination of therapies targeting different levels of

dysregulation can improve the efficiency of cancer treatment and reduce the risk of

relapse. For instance, demethylating agents of the DNA and HDAC inhibitors are known

to enhance the effects of chemotherapy (Dawson & Kouzarides 2012). In this context, it

appears obvious that restricting ourselves to profiling only the “classical” epigenetic

modifications limits our own understanding of the tumor biology.

Ideally, each tumor should be profiled at multiple molecular levels, and the

integration of all this information should determine the choice of the optimal therapeutic

strategy. We are not at this point yet, both because of our incomplete understanding of

tumor complexity, and because of technologic and economic limitations. Biomedical

sciences are making constant progresses though, and the clinical field would undeniably

benefit from integrating concepts developed by fundamental and translational research.

Page 173: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

173

Fig. 88: The molecular portrait of

tumors is a multidimensional picture.

Theoretically, molecular profiling should

take into account all levels of

dysregulation in a tumor. In that regard,

the recently discovered epigenetic

modifications offers a growing platform

for the study of cancer-related

mechanisms.

In light of this, we believe that studying epigenetic modifications is highly

relevant to the improvement of our understanding of a multidimensional cancer biology.

Of particular interest, the growing epigenetic repertoire offers a wide variety of

modifications with various features and benefits that can be complementary in terms of

clinical use. For instance, DNA modifications such as 5mC and 5hmC are extremely

specific of cell types and their distribution can be used as a potential clinical tool for TME

characterization and cell typing within tumors (Jeschke et al. 2015). One advantage of

DNA modifications, compared to transcriptomics or proteomics, is the linearity of the

relationship between the number of cells and the mark (given the two DNA copies that

can be modified per cell). Recently, our host laboratory demonstrated the value of such

approach with a 5mC-based immune response signature which improved patient

diagnosis and predicted the response to chemotherapy in multiple cancers (Jeschke et al.

2017). Furthermore, DNA marks are stable and are not degraded when released in the

blood stream by cancer cells, which makes them useable as biomarkers in non-invasive

liquid biopsies. By contrast, RNA modifications present the advantage of being a level of

dysregulation which is more downstream than genetics and transcriptomics, thus they can

capture complementary information. Additionally, the recent development of RNA-based

therapeutics has highlighted the clinical potential of RNA modifications. Chemical

modifications of injected RNA can, for instance, impart resistance to degradation or block

translation. The degree to which RNA modifications affect the potency of the transcript

depends both on the type of RNA and its mechanism of action (Kaczmarek et al. 2017).

Indeed, heavy chemical modification is now a standard for in vivo siRNA use, whereas

Page 174: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

174

mRNAs, which have to be translated by ribosomes, are more sensitive and studies have

suggested that they should only carry naturally-occurring modifications, such as 5mrC or

m6A (Behlke 2008; Kauffman et al. 2016). In line with this, the first RNA-based drugs

have recently gained FDA approval and are being tested in clinical trials, illustrating that

this research is beginning to bear fruit.

In conclusion, cancer management is, without a doubt, a multidisciplinary task.

Currently, the therapeutic decision is based on several criteria that include both clinical

features and the molecular profiling of the tumor. It has become clear in recent years,

however, that molecular profiling has not yet reached its full potential. In particular,

epigenetics is an important contributor of the molecular profile, and its expanding

repertoire bears a great potential in terms of cancer therapies and biomarkers that will

help chart a course towards future personal medicine.

Page 175: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

175

References

References

Page 176: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

176

Aaltomaa, S. et al., 1992. Lymphocyte infiltrates as a prognostic variable in female

breast cancer. European Journal of Cancer, 28(4–5), pp.859–864.

Accomando, W.P. et al., 2014. Quantitative reconstruction of leukocyte subsets using

DNA methylation. Genome Biol, 15(3), p.R50. Available at:

http://genomebiology.com/2014/15/3/R50%5Cnpapers3://publication/doi/10.1186/

gb-2014-15-3-r50.

Ahmed, I.A. et al., 2010. Epigenetic alterations by methylation of RASSF1A and

DAPK1 promoter sequences in mammary carcinoma detected in extracellular

tumor DNA. Cancer Genetics and Cytogenetics, 199(2), pp.96–100.

Ahsan, S. et al., 2014. Increased 5-hydroxymethylcytosine and decreased 5-

methylcytosine are indicators of global epigenetic dysregulation in diffuse intrinsic

pontine glioma. Acta neuropathologica communications, 2(1), p.59. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/24894482.

Almouzni, G. & Probst, A. V., 2011. Heterochromatin maintenance and establishment:

Lessons from the mouse pericentromere. Nucleus, 2(5), pp.332–338. Available at:

http://www.tandfonline.com/doi/abs/10.4161/nucl.2.5.17707.

Altun, G., Loring, J.F. & Laurent, L.C., 2010. DNA methylation in embryonic stem

cells. Journal of Cellular Biochemistry, 109(1), pp.1–6.

Amort, T. et al., 2017. Distinct 5-methylcytosine profiles in poly(A) RNA from mouse

embryonic stem cells and brain. Genome Biology, 18(1), p.1. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/28077169%0Ahttp://genomebiology.biomed

central.com/articles/10.1186/s13059-016-1139-1.

Amort, T. et al., 2013. Long non-coding RNAs as targets for cytosine methylation. RNA

biology, 10(6), pp.1003–1008. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/23595112.

De Andrés, M.C. et al., 2013. Loss of methylation in CpG sites in the NF-κB enhancer

elements of inducible nitric oxide synthase is responsible for gene induction in

human articular chondrocytes. Arthritis and Rheumatism, 65(3), pp.732–742.

Andrews, S.R. et al., 2012. BS-Seq Data Analysis. Babraham Bioinformatics, 9(July),

pp.145–151. Available at: www.bioinformatics.babraham.ac.uk.

Aquino-Gil, M. et al., 2017. OGT: a short overview of an enzyme standing out from

usual glycosyltransferases. Biochemical Society Transactions, 45(2), pp.365–370.

Available at: http://biochemsoctrans.org/lookup/doi/10.1042/BST20160404.

Arnoult, N., Van Beneden, A. & Decottignies, A., 2012. Telomere length regulates

TERRA levels through increased trimethylation of telomeric H3K9 and HP1α.

Nature Structural & Molecular Biology, 19(9), pp.948–956. Available at:

http://www.nature.com/doifinder/10.1038/nsmb.2364.

Ashburner, B.P., Westerheide, S.D. & Baldwin, A.S., 2001. The p65 (RelA) subunit of

NF-kappaB interacts with the histone deacetylase (HDAC) corepressors HDAC1

and HDAC2 to negatively regulate gene expression. Molecular and cellular

biology, 21(20), pp.7065–77. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/11564889%5Cnhttp://www.pubmedcentral.n

ih.gov/articlerender.fcgi?artid=PMC99882.

Page 177: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

177

Ayupe, A.C. & Reis, E.M., 2017. Evaluating the stability of mRNAs and noncoding

RNAs. Methods in Molecular Biology, 1468, pp.139–153.

Bagu, E.T. et al., 2017. Repression of Fyn-related kinase in breast cancer cells is

associated with promoter site-specific CpG methylation. Oncotarget. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/28077797%5Cnhttp://www.oncotarget.com/

abstract/14546.

Bandukwala, H.S. et al., 2012. Selective inhibition of CD4+ T-cell cytokine production

and autoimmunity by BET protein and c-Myc inhibitors. Proceedings of the

National Academy of Sciences, 109(36), pp.14532–14537. Available at:

http://www.pnas.org/cgi/doi/10.1073/pnas.1212264109.

Bannister, A.J. & Kouzarides, T., 2011. Regulation of chromatin by histone

modifications. Cell research, 21(3), pp.381–395. Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3193420&tool=pmcent

rez&rendertype=abstract.

Batista, P.J. et al., 2014. M6A RNA modification controls cell fate transition in

mammalian embryonic stem cells. Cell Stem Cell, 15(6), pp.707–719.

Becht, E. et al., 2016. Cancer immune contexture and immunotherapy. Current Opinion

in Immunology, 39, pp.7–13.

Behlke, M. a, 2008. Chemical modification of siRNAs for in vivo use.

Oligonucleotides, 18(4), pp.305–19. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/19025401.

Ben-Neriah, Y. & Karin, M., 2011. Inflammation meets cancer, with NF-κB as the

matchmaker. Nature Immunology, 12(8), pp.715–723. Available at:

http://www.nature.com/doifinder/10.1038/ni.2060.

Vanden Berghe, W. et al., 2006. Keeping up NF-kB appearances: Epigenetic control of

immunity or inflammation-triggered epigenetics. Biochemical Pharmacology, 72(9

SPEC. ISS.), pp.1114–1131.

Bergman, Y. & Cedar, H., 2013. DNA methylation dynamics in health and disease. Nat

Struct Mol Biol, 20(3), pp.274–281. Available at:

http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt

=Citation&list_uids=23463312.

Berner, C., Engle, S. & Haslberger, A.G., 2010. Epigenetics and human health. Journal

fur Ernahrungsmedizin, 12(1), pp.26–27.

Bernstein, B.E., Meissner, A. & Lander, E.S., 2007. The Mammalian Epigenome. Cell,

128(4), pp.669–681.

Bhatelia, K., Singh, K. & Singh, R., 2014. TLRs: Linking inflammation and breast

cancer. Cellular Signalling, 26(11), pp.2350–2357.

Bhattacharyya, S. et al., 2013. Genome-wide hydroxymethylation tested using the

HELP-GT assay shows redistribution in cancer. Nucleic Acids Research, 41(16).

Bird, A., 2007. Perceptions of epigenetics. Nature, 447(7143), pp.396–8. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/17522671.

Biswas, S.K. & Mantovani, A., 2010. Macrophage plasticity and interaction with

Page 178: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

178

lymphocyte subsets : cancer as a paradigm. Nature immunology, 11(10), pp.889–

896. Available at:

http://dx.doi.org/10.1038/ni.1937%5Cnhttp://www.ncbi.nlm.nih.gov/pubmed/2085

6220.

Blanco, S. & Frye, M., 2014. Role of RNA methyltransferases in tissue renewal and

pathology. Current Opinion in Cell Biology, 30(1), pp.1–7.

Blaschke, K. et al., 2013. Vitamin C induces Tet-dependent DNA demethylation and a

blastocyst-like state in ES cells. Nature, 500(7461), pp.222–226. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/23812591%5Cnhttp://www.nature.com/natu

re/journal/v500/n7461/pdf/nature12362.pdf.

Booth, L.N. & Brunet, A., 2016. The Aging Epigenome. Molecular Cell, 62(5), pp.728–

744. Available at: http://dx.doi.org/10.1016/j.molcel.2016.05.013.

Booth, M.J. et al., 2013. Oxidative bisulfite sequencing of 5-methylcytosine and 5-

hydroxymethylcytosine. Nature protocols, 8(10), pp.1841–1851. Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3919000%7B&%7Dto

ol=pmcentrez%7B&%7Drendertype=abstract$%5C$nhttp://www.ncbi.nlm.nih.gov

/pubmed/24008380.

Boyce, W.T. & Kobor, M.S., 2015. Development and the epigenome: The “synapse” of

gene-environment interplay. Developmental Science, 18(1), pp.1–23.

Brookes, E. & Shi, Y., 2014. Diverse epigenetic mechanisms of human disease.,

Available at: http://www.annualreviews.org/doi/abs/10.1146/annurev-genet-

120213-092518.

Butler, J.S. & Dent, S.Y.R., 2013. The role of chromatin modifiers in normal and

malignant hematopoiesis. Blood, 121(16), pp.3076–3084.

C., N., E., D.M. & Codacci-Pisanelli G. AO - Codacci-Pisanelli, G.O. http://orcid.

org/000.-0003-4662-7944, 2015. Epigenetic treatment of solid tumours: A review

of clinical trials. Clinical Epigenetics, 7(1), p.no pagination. Available at:

http://www.springer.com/biomed/human+genetics/journal/13148%5Cnhttp://ovids

p.ovid.com/ovidweb.cgi?T=JS&PAGE=reference&D=emed18b&NEWS=N&AN=

607277767.

Carrio, E. & Suelves, M., 2015. DNA methylation dynamics in muscle development

and disease. Front Aging Neurosci, 7, p.19. Available at:

http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt

=Citation&list_uids=25798107.

Chakravarthy, S. et al., 2005. Structure and dynamic properties of nucleosome core

particles. FEBS Letters, 579(4), pp.895–898. Available at:

http://doi.wiley.com/10.1016/j.febslet.2004.11.030.

Chen, F., Beezhold, K. & Castranova, V., 2008. Tumor Promoting or Tumor

Suppressing of NF-κ B, a Matter of Cell Context Dependency. International

Reviews of Immunology, 27(4), pp.183–204.

Chen, Q. et al., 2013. TET2 promotes histone O-GlcNAcylation during gene

transcription. Nature, 493(7433), pp.561–4. Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3684361&tool=pmcent

Page 179: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

179

rez&rendertype=abstract.

Chen, Y. et al., 2014. Diversity of two forms of DNA methylation in the brain.

Frontiers in Genetics, 5(46).

Chen, Y. et al., 2011. HDAC-mediated deacetylation of NF-κB is critical for Schwann

cell myelination. Nature neuroscience, 14(4), pp.437–441. Available at:

http://dx.doi.org/10.1038/nn.2780.

Cheng, J. et al., 2013. An Extensive Network of TET2-Targeting MicroRNAs Regulates

Malignant Hematopoiesis. Cell Reports, 5(2), pp.471–481.

Cheng, Y. et al., 2015. 5-Hydroxymethylcytosine: A new player in brain disorders?

Experimental Neurology, 268, pp.3–9.

Chiappinelli, K.B. et al., 2015. Inhibiting DNA Methylation Causes an Interferon

Response in Cancer via dsRNA Including Endogenous Retroviruses. Cell, 162(5),

pp.974–986.

Chim, C.S. et al., 2010. Methylation of TET2, CBL and CEBPA in Ph-negative

myeloproliferative neoplasms. Journal of clinical pathology, 63(10), pp.942–6.

Available at: http://www.ncbi.nlm.nih.gov/pubmed/20671051.

Cho, Y.M. et al., 2009. Type 2 diabetes-associated genetic variants discovered in the

recent genome-wide association studies are related to gestational diabetes mellitus

in the Korean population. Diabetologia, 52(2), pp.253–261.

Chow, C.S., Lamichhane, T.N. & Mahto, S.K., 2007. Expanding the nucleotide

repertoire of the ribosome with post-transcriptional modifications. ACS Chemical

Biology, 2(9), pp.610–619.

Chuang, K.H. et al., 2015. MicroRNA-494 is a master epigenetic regulator of multiple

invasion-suppressor microRNAs by targeting ten eleven translocation 1 in invasive

human hepatocellular carcinoma tumors. Hepatology, 62(2), pp.466–480.

Church, C. et al., 2010. Overexpression of Fto leads to increased food intake and results

in obesity. Nature genetics, 42(12), pp.1086–92. Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3018646&tool=pmcent

rez&rendertype=abstract.

Cimmino, L. et al., 2015. TET1 is a tumor suppressor of hematopoietic malignancy.

Nature immunology, 16(6), pp.653–62.

Connolly, R.M. et al., 2017. Combination epigenetic therapy in advanced breast cancer

with 5-azacitidine and entinostat: A phase II national cancer institute/stand up to

cancer study. Clinical Cancer Research, 23(11), pp.2691–2701.

Costa, F.F. et al., 2006. SATR-1 hypomethylation is a common and early event in breast

cancer. Cancer Genetics and Cytogenetics, 165(2), pp.135–143.

Cui, Q. et al., 2017. m6A RNA Methylation Regulates the Self-Renewal and

Tumorigenesis of Glioblastoma Stem Cells. Cell Reports, 18(11), pp.2622–2634.

Available at:

http://www.sciencedirect.com/science/article/pii/S2211124717302747 [Accessed

April 3, 2017].

Curtis, C. et al., 2012. The genomic and transcriptomic architecture of 2 , 000 breast

Page 180: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

180

tumours reveals novel subgroups. Nature, 486(7403), pp.346–352. Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3440846&tool=pmcent

rez&rendertype=abstract.

Dany, M. et al., 2016. Advances in immunotherapy for melanoma management. Human

Vaccines and Immunotherapeutics, 12(10), pp.2501–2511.

Dawlaty, M.M. et al., 2013. Combined Deficiency of Tet1 and Tet2 Causes Epigenetic

Abnormalities but Is Compatible with Postnatal Development. Developmental

Cell, 24(3), pp.310–323.

Dawlaty, M.M. et al., 2011. Tet1 is dispensable for maintaining pluripotency and its

loss is compatible with embryonic and postnatal development. Cell Stem Cell, 9(2),

pp.166–175.

Dawson, M.A. & Kouzarides, T., 2012. Cancer epigenetics: From mechanism to

therapy. Cell, 150(1), pp.12–27.

Day, T.K. & Bianco-Miotto, T., 2013. Common gene pathways and families altered by

DNA methylation in breast and prostate cancers. Endocrine-Related Cancer, 20(5).

Deans, C. et al., 2015. What Do You Mean, “Epigenetic”? Genetics, 199(4), pp.775–

792.

Dedeurwaerder, S. et al., 2013. A comprehensive overview of Infinium

HumanMethylation450 data processing. Briefings in bioinformatics, 15(6),

pp.929–41. Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=4239800&tool=pmcent

rez&rendertype=abstract [Accessed October 20, 2014].

Dedeurwaerder, S. et al., 2011. DNA methylation profiling reveals a predominant

immune component in breast cancers. EMBO Molecular Medicine, 3(12), pp.726–

741.

Dedeurwaerder, S. & Fuks, F., 2012. DNA methylation markers for breast cancer

prognosis: Unmasking the immune component. Oncoimmunology, 1(6), pp.962–

964. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/23162772%5Cnhttp://www.pubmedcentral.n

ih.gov/articlerender.fcgi?artid=PMC3489760.

DeNardo, D.G. et al., 2011. Leukocyte complexity predicts breast cancer survival and

functionally regulates response to chemotherapy. Cancer Discovery, 1(1), pp.54–

67.

Deng, X. et al., 2015. Widespread occurrence of N6-methyladenosine in bacterial

mRNA. Nucleic acids research, 43(13), pp.6557–67. Available at:

http://nar.oxfordjournals.org/content/early/2015/06/11/nar.gkv596.full.

Deplus, R. et al., 2013. TET2 and TET3 regulate GlcNAcylation and H3K4 methylation

through OGT and SET1/COMPASS. The EMBO Journal, 32(5), pp.645–655.

Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3590984&tool=pmcent

rez&rendertype=abstract.

Desrosiers, R., Friderici, K. & Rottman, F., 1974. Identification of methylated

nucleosides in messenger RNA from Novikoff hepatoma cells. Proceedings of the

Page 181: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

181

National Academy of Sciences of the United States of America, 71(10), pp.3971–5.

Available at:

http://www.ncbi.nlm.nih.gov/pubmed/4372599%5Cnhttp://www.pubmedcentral.ni

h.gov/articlerender.fcgi?artid=PMC434308.

Dieci, M. V et al., 2015. Prognostic and predictive value of tumor-infiltrating

lymphocytes in two phase III randomized adjuvant breast cancer trials. Annals of

oncology : official journal of the European Society for Medical Oncology / ESMO,

26(8), pp.1698–704. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/25995301%5Cnhttp://www.pubmedcentral.n

ih.gov/articlerender.fcgi?artid=PMC4511223.

Dieci, M.V. et al., 2014. Rare Breast Cancer Subtypes: Histological, Molecular, and

Clinical Peculiarities. The oncologist, 19(8), pp.805–813.

Dietrich, D. et al., 2010. CDO1 promoter methylation is a biomarker for outcome

prediction of anthracycline treated, estrogen receptor-positive, lymph node-positive

breast cancer patients. BMC cancer, 10, p.247. Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2893112&tool=pmcent

rez&rendertype=abstract.

Dominissini, D. et al., 2012. Topology of the human and mouse m6A RNA methylomes

revealed by m6A-seq. Nature, 485(7397), pp.201–206. Available at:

http://www.nature.com/doifinder/10.1038/nature11112%5Cnhttp://dx.doi.org/10.1

038/nature11112.

Dubin, D.T. & Taylor, R.H., 1975. The methylation state of poly A-containing-

messenger RNA from cultured hamster cells. Nucleic Acids Research, 2(10),

pp.1653–1668.

Dumalaon-Canaria, J.A. et al., 2014. What causes breast cancer? A systematic review of

causal attributions among breast cancer survivors and how these compare to

expert-endorsed risk factors. Cancer Causes and Control, 25(7), pp.771–785.

Edelheit, S. et al., 2013. Transcriptome-Wide Mapping of 5-methylcytidine RNA

Modifications in Bacteria, Archaea, and Yeast Reveals m5C within Archaeal

mRNAs. PLoS Genetics, 9(6).

ESMO, 2016. Breast Cancer: A Guide for Patients. Available at:

http://www.esmo.org/content/download/6593/114959/file/EN-Breast-Cancer-

Guide-for-Patients.pdf.

Esteller, M. et al., 2000. Inactivation of the DNA-repair gene MGMT and the clinical

response of gliomas to alkylating agents. New England Journal of Medicine,

343(19), pp.1350–1354. Available at:

http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt

=Citation&list_uids=11070098%5Cnhttp://www.nejm.org/doi/pdf/10.1056/NEJM

200011093431901.

Ezkurdia, I. et al., 2014. Multiple evidence strands suggest that theremay be as few as

19 000 human protein-coding genes. Human Molecular Genetics, 23(22), pp.5866–

5878.

Fan, S. & Zhang, X., 2009. CpG island methylation pattern in different human tissues

and its correlation with gene expression. Biochemical and Biophysical Research

Page 182: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

182

Communications, 383(4), pp.421–425.

Farooqi, I.S., 2011. FTO and Obesity: The missing link. Cell Metabolism, 13(1), pp.7–

8.

Fatemi, M. et al., 2002. Dnmt3a and Dnmt1 functionally cooperate during de novo

methylation of DNA. European Journal of Biochemistry, 269(20), pp.4981–4984.

Feinberg, A.P. & Tycko, B., 2004. The history of cancer epigenetics. Nature reviews.

Cancer, 4(2), pp.143–153.

Feng, J. et al., 2010. Dnmt1 and Dnmt3a maintain DNA methylation and regulate

synaptic function in adult forebrain neurons. Nature neuroscience, 13(4), pp.423–

30. Available at: http://www.ncbi.nlm.nih.gov/pubmed/20228804.

Feng, J. & Nestler, E.J., 2010. MeCP2 and drug addiction. Nat. neurosci, 13(9),

pp.1039–41. Available at: http://www.ncbi.nlm.nih.gov/pubmed/20740030.

Ficz, G. et al., 2011. Dynamic regulation of 5-hydroxymethylcytosine in mouse ES cells

and during differentiation. Nature, 473(7347), pp.398–402. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/21460836.

Figueroa, M.E. et al., 2010. Leukemic IDH1 and IDH2 Mutations Result in a

Hypermethylation Phenotype, Disrupt TET2 Function, and Impair Hematopoietic

Differentiation. Cancer Cell, 18(6), pp.553–567.

Filho, O.M., Ignatiadis, M. & Sotiriou, C., 2011. Genomic Grade Index: An important

tool for assessing breast cancer tumor grade and prognosis. Critical Reviews in

Oncology/Hematology, 77(1), pp.20–29.

Fischer, J. et al., 2009. Inactivation of the Fto gene protects from obesity. Nature,

458(7240), pp.894–8. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/19234441.

Foedermayr, M. et al., 2014. BRCA-1 methylation and TP53 mutation in triple-negative

breast cancer patients without pathological complete response to taxane-based

neoadjuvant chemotherapy. Cancer Chemotherapy and Pharmacology, 73(4),

pp.771–778.

Fu, L. et al., 2014. Tet-mediated formation of 5-hydroxymethylcytosine in RNA.

Journal of the American Chemical Society, 136(33), pp.11582–11585.

Fu, Y. et al., 2013. FTO-mediated formation of N6-hydroxymethyladenosine and N6-

formyladenosine in mammalian RNA. Nature communications, 4, p.1798.

Available at:

http://www.ncbi.nlm.nih.gov/pubmed/23653210%5Cnhttp://www.pubmedcentral.n

ih.gov/articlerender.fcgi?artid=PMC3658177.

Fu, Y. et al., 2014. Gene expression regulation mediated through reversible m(6)A RNA

methylation. Nature reviews. Genetics, 15, pp.293–306. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/24662220.

Fujiki, R. et al., 2011. GlcNAcylation of histone H2B facilitates its monoubiquitination.

Nature, 480(7378), pp.557–60. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/22121020.

Fumagalli, C. et al., 2012. Methylation of O 6-methylguanine-DNA methyltransferase

Page 183: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

183

(MGMT) promoter gene in triple-negative breast cancer patients. Breast Cancer

Research and Treatment, 134(1), pp.131–137.

Gagan, J. & Van Allen, E.M., 2015. Next-generation sequencing to guide cancer

therapy. Genome medicine, 7(1), p.80. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/26221189%5Cnhttp://www.pubmedcentral.n

ih.gov/articlerender.fcgi?artid=PMC4517547.

Galon, J. et al., 2006. Type, density, and location of immune cells within human

colorectal tumors predict clinical outcome. Science (New York, N.Y.), 313(5795),

pp.1960–4. Available at: http://www.ncbi.nlm.nih.gov/pubmed/17008531.

Gerondakis, S. et al., 2014. NF-κB control of T cell development. Nature immunology,

15(1), pp.15–25. Available at: http://www.ncbi.nlm.nih.gov/pubmed/24352326.

Gerondakis, S. & Siebenlist, U., 2010. Roles of the NF-κB Pathway in Lymphocyte

Development and Function. Cold Spring Harbor Perspective in Biology, 2(5),

p.29.

Geula, S. et al., 2015. Stem cells. m6A mRNA methylation facilitates resolution of

naïve pluripotency toward differentiation. Science, 347(6225), pp.1002–6.

Available at: http://www.ncbi.nlm.nih.gov/pubmed/25569111.

Gibney, E.R. & Nolan, C.M., 2010. Epigenetics and gene expression. Heredity, 105(1),

pp.4–13. Available at: http://dx.doi.org/10.1038/hdy.2010.54.

Gilat, N. et al., 2017. Single-molecule quantification of 5-hydroxymethylcytosine for

diagnosis of blood and colon cancers. Clinical Epigenetics, 9(1), p.70. Available

at: http://clinicalepigeneticsjournal.biomedcentral.com/articles/10.1186/s13148-

017-0368-9.

Globisch, D. et al., 2010. Tissue distribution of 5-hydroxymethylcytosine and search for

active demethylation intermediates. PLoS ONE, 5(12).

Godderis, L. et al., 2015. Global methylation and hydroxymethylation in DNA from

blood and saliva in healthy volunteers. BioMed Research International, 2015.

Grant, S.F.A. et al., 2008. Association analysis of the FTO gene with obesity in children

of Caucasian and African ancestry reveals a common tagging SNP. PLoS ONE,

3(3).

Gu, T.P. et al., 2011. The role of Tet3 DNA dioxygenase in epigenetic reprogramming

by oocytes. Nature, 477(7366), pp.606–610. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/21892189.

Guo, J.U. et al., 2011. Hydroxylation of 5-Methylcytosine by TET1 Promotes Active

DNA Demethylation in the Adult Brain. Cell, 145(3), pp.423–434.

Gupta, S. et al., 2010. Inhibiting NF-κB activation by small molecules as a therapeutic

strategy. Biochim Biophys Acta, 1799(10–12), pp.775–787. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/20493977%0Ahttp://www.ncbi.nlm.nih.gov/

pubmed/20493977%0Ahttp://www.ncbi.nlm.nih.gov/pubmed/20493977.

Hackl, H. et al., 2016. Computational genomics tools for dissecting tumour-immune cell

interactions. Nature reviews. Genetics, 17(8), pp.441–58. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/27376489.

Page 184: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

184

Haffner, M.C. et al., 2011. Global 5-hydroxymethylcytosine content is significantly

reduced in tissue stem/progenitor cell compartments and in human cancers.

Oncotarget, 2(8), pp.627–37.

Hall, L.E., Mitchell, S.E. & O’Neill, R.J., 2012. Pericentric and centromeric

transcription: A perfect balance required. Chromosome Research, 20(5), pp.535–

546.

Harbeck, N. et al., 2008. Multicenter study using paraffin-embedded tumor tissue

testing PITX2 DNA methylation as a marker for outcome prediction in tamoxifen-

treated, node-negative breast cancer patients. Journal of Clinical Oncology, 26(31),

pp.5036–5042.

Hark, A.T. et al., 2000. CTCF mediates methylation-sensitive enhancer-blocking

activity at the H19/Igf2 locus. Nature, 405(6785), pp.486–489.

Harris, H.R., Orsini, N. & Wolk, A., 2014. Vitamin C and survival among women with

breast cancer: A Meta-analysis. European Journal of Cancer, 50(7), pp.1223–

1231.

Hata, K. et al., 2002. Dnmt3L cooperates with the Dnmt3 family of de novo DNA

methyltransferases to establish maternal imprints in mice. Development

(Cambridge, England), 129, pp.1983–1993.

He, C. et al., 2016. High-Resolution Mapping of RNA-Binding Regions in the Nuclear

Proteome of Embryonic Stem Cells. Molecular Cell, 64(2), pp.416–430.

Heitz, E., 1928. Das Heterochromatin der Moose. Jahrbücher für Wissenschaftliche

Botanik, 69, pp.762–818.

Helming, K.C., Wang, X. & Roberts, C.W.M., 2014. Vulnerabilities of mutant

SWI/SNF complexes in cancer. Cancer Cell, 26(3), pp.309–317.

Herschkowitz, J.I. et al., 2007. Identification of conserved gene expression features

between murine mammary carcinoma models and human breast tumors. Genome

biology, 8(5), p.R76. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/17493263%5Cnhttp://www.pubmedcentral.n

ih.gov/articlerender.fcgi?artid=PMC1929138.

Herviou, L. et al., 2016. EZH2 in normal hematopoiesis and hematological

malignancies. Oncotarget, 7(3), pp.2284–2296. Available at:

http://www.impactjournals.com/oncotarget/index.php?journal=oncotarget&page=a

rticle&op=download&path[]=6198&path[]=15543.

Ho, L. & Crabtree, G.R., 2010. Chromatin remodelling during development. Nature,

463(7280), pp.474–84. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/20110991%5Cnhttp://www.pubmedcentral.n

ih.gov/articlerender.fcgi?artid=PMC3060774.

Hodi, F.S. et al., 2010. Improved survival with ipilimumab in patients with metastatic

melanoma. The New England journal of medicine, 363(8), pp.711–23. Available

at:

http://www.ncbi.nlm.nih.gov/pubmed/20525992%5Cnhttp://www.pubmedcentral.n

ih.gov/articlerender.fcgi?artid=PMC3549297.

Van Hoesel, A.Q. et al., 2012. Hypomethylation of LINE-1 in primary tumor has poor

Page 185: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

185

prognosis in young breast cancer patients: A retrospective cohort study. Breast

Cancer Research and Treatment, 134(3), pp.1103–1114.

Hoesel, B. & Schmid, J. a, 2013. The complexity of NF-κB signaling in inflammation

and cancer. Molecular cancer, 12(1), p.86. Available at: Molecular Cancer.

Hoffman, G.R. et al., 2014. Functional epigenetics approach identifies

BRM/SMARCA2 as a critical synthetic lethal target in BRG1-deficient cancers.

Proceedings of the National Academy of Sciences, 111(8), pp.3128–3133.

Available at: http://www.pnas.org/cgi/doi/10.1073/pnas.1316793111.

den Hollander, P., Savage, M.I. & Brown, P.H., 2013. Targeted therapy for breast

cancer prevention. Frontiers in oncology, 3(September), p.250. Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3780469&tool=pmcent

rez&rendertype=abstract.

Holm, K. et al., 2012. Global H3K27 trimethylation and EZH2 abundance in breast

tumor subtypes. Molecular Oncology, 6(5), pp.494–506.

Holm, K. et al., 2010. Molecular subtypes of breast cancer are associated with

characteristic DNA methylation patterns. Breast cancer research : BCR, 12(3),

p.R36.

Hoque, M.O. et al., 2006. Detection of aberrant methylation of four genes in plasma

DNA for the detection of breast cancer. Journal of clinical oncology : official

journal of the American Society of Clinical Oncology, 24(26), pp.4262–4269.

Available at:

http://jco.ascopubs.org/cgi/doi/10.1200/JCO.2005.01.3516%5Cnhttp://www.ncbi.n

lm.nih.gov/pubmed/16908936.

Hsieh, M.-C. et al., 2016. Tet1-dependent epigenetic modification of BDNF expression

in dorsal horn neurons mediates neuropathic pain in rats. Scientific reports, 6,

p.37411. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/27857218%5Cnhttp://www.pubmedcentral.n

ih.gov/articlerender.fcgi?artid=PMC5114645.

Hsu, C.-H. et al., 2012. TET1 suppresses cancer invasion by activating the tissue

inhibitors of metalloproteinases. Cell reports, 2(3), pp.568–79. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/22999938 [Accessed February 18, 2015].

Hu, Z. et al., 2006. The molecular portraits of breast tumors are conserved across

microarray platforms. BMC Genomics, 7, p.96. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/16643655%5Cnhttp://www.biomedcentral.c

om/1471-

2164/7/96%5Cnhttp://www.ncbi.nlm.nih.gov/pmc/articles/PMC1468408/pdf/1471

-2164-7-96.pdf.

Huang, C.-Z., Yu, T. & Chen, Q.-K., 2015. DNA Methylation Dynamics During

Differentiation, Proliferation, and Tumorigenesis in the Intestinal Tract. Stem Cells

and Development, 24(23), pp.2733–2739.

Huang, H. et al., 2013. TET1 plays an essential oncogenic role in MLL-rearranged

leukemia. Proceedings of the National Academy of Sciences of the United States of

America, 110(29), pp.11994–9. Available at:

http://www.pnas.org/content/110/29/11994.short.

Page 186: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

186

Huang, W.-H. et al., 2016. Determination of DNA and RNA Methylation in Circulating

Tumor Cells by Mass Spectrometry. Analytical Chemistry, 88(2), pp.1378–1384.

Available at:

http://pubs.acs.org/doi/10.1021/acs.analchem.5b03962%5Cnhttp://pubs.acs.org/doi

/abs/10.1021/acs.analchem.5b03962.

Huang, X. et al., 2011. HDAC inhibitor SNDX-275 enhances efficacy of trastuzumab in

erbB2-overexpressing breast cancer cells and exhibits potential to overcome

trastuzumab resistance. Cancer Letters, 307(1), pp.72–79.

Huang, Y. et al., 2015. Meclofenamic acid selectively inhibits FTO demethylation of

m6A over ALKBH5. Nucleic Acids Research, 43(1), pp.373–384.

Huber, M.A. et al., 2004. NF-kappaB is essential for epithelial-mesenchymal transition

and metastasis in a model of breast cancer progression. The Journal of clinical

investigation, 114(4), pp.569–81. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/15314694%5Cnhttp://www.pubmedcentral.n

ih.gov/articlerender.fcgi?artid=PMC503772.

Huber, S.M. et al., 2015. Formation and abundance of 5-hydroxymethylcytosine in

RNA. ChemBioChem, 16(5), pp.752–755.

Hussain, S., Aleksic, J., et al., 2013. Characterizing 5-methylcytosine in the mammalian

epitranscriptome. Genome biology, 14(11), p.215. Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=4053770&tool=pmcent

rez&rendertype=abstract [Accessed January 28, 2015].

Hussain, S., Sajini, A.A., et al., 2013. NSun2-mediated cytosine-5 methylation of vault

noncoding RNA determines its processing into regulatory small RNAs. Cell

Reports, 4(2), pp.255–261.

Ichiyama, K. et al., 2015. The Methylcytosine Dioxygenase Tet2 Promotes DNA

Demethylation and Activation of Cytokine Gene Expression in T Cells. Immunity,

42(4), pp.613–626.

Ignatiadis, M., Georgoulias, V. & Mavroudis, D., 2008. Micrometastatic disease in

breast cancer: Clinical implications. European Journal of Cancer, 44(18),

pp.2726–2736.

Inoue, A. & Zhang, Y., 2011. Replication-dependent loss of 5-hydroxymethylcytosine

in mouse preimplantation embryos. Science, 334(6053), p.194. Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3799877&tool=pmcent

rez&rendertype=abstract [Accessed February 17, 2015].

Isella, C. et al., 2015. Stromal contribution to the colorectal cancer transcriptome. Nat

Genet, 47(4), pp.312–319. Available at:

http://www.nature.com/ng/journal/v47/n4/pdf/ng.3224.pdf.

Ito, S. et al., 2010. Role of Tet proteins in 5mC to 5hmC conversion, ES-cell self-

renewal and inner cell mass specification. Nature, 466(7310), pp.1129–1133.

Ito, S. et al., 2011. Tet proteins can convert 5-methylcytosine to 5-formylcytosine and

5-carboxylcytosine. Science, 333(6047), pp.1300–3. Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3495246&tool=pmcent

rez&rendertype=abstract [Accessed January 13, 2015].

Page 187: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

187

Ito, Y. et al., 2008. Somatically acquired hypomethylation of IGF2 in breast and

colorectal cancer. Human Molecular Genetics, 17(17), pp.2633–2643.

Iurlaro, M. et al., 2016. In vivo genome-wide profiling reveals a tissue-specific role for

5-formylcytosine. Genome biology, 17(1), p.141. Available at:

http://dx.doi.org/10.1186/s13059-016-1001-

5%5Cnhttp://www.ncbi.nlm.nih.gov/pubmed/27356509%5Cnhttp://www.pubmedc

entral.nih.gov/articlerender.fcgi?artid=PMC4928330.

Jeschke, J. et al., 2017. DNA methylation–based immune response signature improves

patient diagnosis in multiple cancers. J Clin Invest., 18(94), p. Available at:

https://www.jci.org/articles/view/91095.

Jeschke, J., Collignon, E. & Fuks, F., 2015. DNA methylome profiling beyond

promoters - Taking an epigenetic snapshot of the breast tumor microenvironment.

FEBS Journal, 282(9), pp.1801–1814.

Jeschke, J., Collignon, E. & Fuks, F., 2016. Portraits of TET-mediated DNA

hydroxymethylation in cancer. Current Opinion in Genetics and Development, 36,

pp.16–26.

Jesinger, R.A., 2014. Breast anatomy for the interventionalist. Techniques in Vascular

and Interventional Radiology, 17(1), pp.3–9.

Jia, G. et al., 2011. N6-methyladenosine in nuclear RNA is a major substrate of the

obesity-associated FTO. Nature chemical biology, 7(12), pp.885–7. Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3218240&tool=pmcent

rez&rendertype=abstract.

Jin, S.G., Kadam, S. & Pfeifer, G.P., 2010. Examination of the specificity of DNA

methylation profiling techniques towards 5-methylcytosine and 5-

hydroxymethylcytosine. Nucleic Acids Research, 38(11).

Jones, P.A., Issa, J.-P.J. & Baylin, S., 2016. Targeting the cancer epigenome for

therapy. Nature reviews. Genetics, 17(10), pp.630–41. Available at:

http://www.nature.com/doifinder/10.1038/nrg.2016.93%5Cnhttp://www.ncbi.nlm.

nih.gov/pubmed/27629931.

Jones, P.L. et al., 1998. Methylated DNA and MeCP2 recruit histone deacetylase to

repress transcription. Nature genetics, 19(june), pp.187–191. Available at:

http://www.nature.com/doifinder/10.1038/561.

Kaczmarek, J.C., Kowalski, P.S. & Anderson, D.G., 2017. Advances in the delivery of

RNA therapeutics: from concept to clinical reality. Genome Medicine, 9(1), p.60.

Available at: http://genomemedicine.biomedcentral.com/articles/10.1186/s13073-

017-0450-0.

Kaklamani, V. et al., 2011. The role of the fat mass and obesity associated gene (FTO)

in breast cancer risk. BMC Medical Genetics, 12(1), p.52. Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3089782&tool=pmcent

rez&rendertype=abstract.

Kang, J., Kalantry, S. & Rao, A., 2013. PGC7, H3K9me2 and Tet3: regulators of DNA

methylation in zygotes. Cell Res, 23(1), pp.6–9. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/22868271.

Page 188: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

188

Kansara, M. et al., 2014. Translational biology of osteosarcoma. Nat Rev Cancer,

14(11), pp.722–735. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/25319867.

Kato, Y. et al., 2007. Role of the Dnmt3 family in de novo methylation of imprinted and

repetitive sequences during male germ cell development in the mouse. Human

Molecular Genetics, 16(19), pp.2272–2280.

Kauffman, K.J. et al., 2016. Efficacy and immunogenicity of unmodified and

pseudouridine-modified mRNA delivered systemically with lipid nanoparticles

in vivo. Biomaterials, 109, pp.78–87.

Khoddami, V. & Cairns, B.R., 2013. Identification of direct targets and modified bases

of RNA cytosine methyltransferases. Nature biotechnology, 31(5), pp.458–64.

Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3791587&tool=pmcent

rez&rendertype=abstract [Accessed February 17, 2015].

Kienhöfer, S. et al., 2015. GADD45a physically and functionally interacts with TET1.

Differentiation, 90(1–3), pp.59–68.

Kim, Y.H. et al., 2011. TET2 promoter methylation in low-grade diffuse gliomas

lacking IDH1/2 mutations. J Clin Pathol, 64(10), pp.850–852. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/21690245%5Cnhttp://jcp.bmj.com/content/6

4/10/850.full.pdf.

Ko, M. et al., 2010. Impaired hydroxylation of 5-methylcytosine in myeloid cancers

with mutant TET2. Nature, 468(7325), pp.839–43. Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3003755&tool=pmcent

rez&rendertype=abstract [Accessed July 16, 2014].

Koboldt, D.C. et al., 2012. Comprehensive molecular portraits of human breast

tumours. Nature, 490(7418), pp.61–70.

Koziol, M.J. et al., 2016. Identification of methylated deoxyadenosines in vertebrates

reveals diversity in DNA modifications. Nature structural & molecular biology,

23(1), pp.24–30. Available at:

http://www.nature.com/nsmb/journal/v23/n1/full/nsmb.3145.html%5Cnhttp://www

.nature.com/nsmb/journal/v23/n1/pdf/nsmb.3145.pdf%5Cnhttp://www.ncbi.nlm.ni

h.gov/pubmed/26689968%5Cnhttp://www.pubmedcentral.nih.gov/articlerender.fcg

i?artid=PMC4941928.

Kriaucionis, S. & Heintz, N., 2009. The nuclear DNA base 5-hydroxymethylcytosine is

present in Purkinje neurons and the brain. Science, 324(5929), pp.929–930.

Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3263819&tool=pmcent

rez&rendertype=abstract.

Kulis, M. et al., 2013. Intragenic DNA methylation in transcriptional regulation, normal

differentiation and cancer. Biochimica et Biophysica Acta - Gene Regulatory

Mechanisms, 1829(11), pp.1161–1174.

Kuzet, S.E. & Gaggioli, C., 2016. Fibroblast activation in cancer: when seed fertilizes

soil. Cell and Tissue Research, 365(3), pp.607–619.

Page 189: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

189

Lakshmi Narendra, B. et al., 2013. Immune system: A double-edged sword in cancer.

Inflammation Research, 62(9), pp.823–834.

Lança, T. & Silva-Santos, B., 2012. The split nature of tumor-infiltrating leukocytes:

Implications for cancer surveillance and immunotherapy. Oncoimmunology, 1(5),

pp.717–725. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/22934263%5Cnhttp://www.pubmedcentral.n

ih.gov/articlerender.fcgi?artid=PMC3429575.

Lander, E.S. et al., 2001. Initial sequencing and analysis of the human genome. Nature,

409(6822), pp.860–921. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/11237011%5Cnhttp://www.nature.com/natu

re/journal/v409/n6822/pdf/409860a0.pdf.

Lao, V.V. & Grady, W.M., 2011. Epigenetics and colorectal cancer. Nature reviews.

Gastroenterology & hepatology, 8(12), pp.686–700. Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3391545&tool=pmcent

rez&rendertype=abstract.

Leal, A.S. et al., 2017. Bromodomain inhibitors, JQ1 and I-BET 762, as potential

therapies for pancreatic cancer. Cancer Letters, 394, pp.76–87.

Lee, H.J., Hore, T.A. & Reik, W., 2014. Reprogramming the methylome: Erasing

memory and creating diversity. Cell Stem Cell, 14(6), pp.710–719.

Lee, S.H. et al., 2010. Dynamic methylation and expression of Oct4 in early neural stem

cells. Journal of Anatomy, 217(3), pp.203–213.

Lehmann, B.D. et al., 2011. Identification of human triple-negative breast cancer

subtypes and preclinical models for selection of targeted therapies. The Journal of

clinical investigation, 121(7), pp.2750–67. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/21633166%5Cnhttp://www.pubmedcentral.n

ih.gov/articlerender.fcgi?artid=PMC3127435.

Lessard, J. et al., 2007. An Essential Switch in Subunit Composition of a Chromatin

Remodeling Complex during Neural Development. Neuron, 55(2), pp.201–215.

Lewis, L.C. et al., 2017. Dynamics of 5-carboxylcytosine during hepatic differentiation:

Potential general role for active demethylation by DNA repair in lineage

specification. Epigenetics, 12(4), pp.277–286.

Li, E., Bestor, T.H. & Jaenisch, R., 1992. Targeted mutation of the DNA

methyltransferase gene results in embryonic lethality. Cell, 69(6), pp.915–926.

Li, L. et al., 2010. Estrogen and progesterone receptor status affect genome-wide DNA

methylation profile in breast cancer. Human molecular genetics, 19(21), pp.4273–

7. Available at: http://www.ncbi.nlm.nih.gov/pubmed/20724461 [Accessed

February 25, 2015].

Li, S.G. & Li, L., 2013. Targeted therapy in HER2-positive breast cancer. Biomedical

reports, 1(4), pp.499–505. Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3917005&tool=pmcent

rez&rendertype=abstract.

Li, S.Y. et al., 2015. Combination therapy with epigenetic-targeted and

chemotherapeutic drugs delivered by nanoparticles to enhance the chemotherapy

Page 190: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

190

response and overcome resistance by breast cancer stem cells. Journal of

Controlled Release, 205, pp.7–14.

Li, X., Xiong, X. & Yi, C., 2017. Epitranscriptome sequencing technologies: decoding

RNA modifications. Nature Methods, 14(1), p.23. Available at:

http://www.nature.com/doifinder/10.1038/nmeth.4110.

Li, Z. et al., 2017. FTO Plays an Oncogenic Role in Acute Myeloid Leukemia as a N6-

Methyladenosine RNA Demethylase. Cancer Cell, 31(1), pp.127–141. Available

at: http://www.sciencedirect.com/science/article/pii/S1535610816305608

[Accessed April 3, 2017].

Li, Z. et al., 2016. FTO Plays an Oncogenic Role in Acute Myeloid Leukemia as a N6-

Methyladenosine RNA Demethylase. Cancer Cell, pp.1–15. Available at:

http://linkinghub.elsevier.com/retrieve/pii/S1535610816305608.

Li, Z. et al., 2015. Gadd45a promotes DNA demethylation through TDG. Nucleic Acids

Research, 43(8), pp.3986–3997.

Lian, C.G. et al., 2012. Loss of 5-hydroxymethylcytosine is an epigenetic hallmark of

melanoma. Cell, 150(6), pp.1135–46. Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3770275&tool=pmcent

rez&rendertype=abstract [Accessed July 15, 2014].

Linder, B. et al., 2015. Single-nucleotide-resolution mapping of m6A and m6Am

throughout the transcriptome. Nat Methods, 12(8), pp.767–772. Available at:

http://dx.doi.org/10.1038/nmeth.3453.

Lister, R. et al., 2009. Human DNA methylomes at base resolution show widespread

epigenomic differences. Nature, 462(7271), pp.315–22. Available at:

http://dx.doi.org/10.1038/nature08514.

Liu, J. et al., 2014. A METTL3-METTL14 complex mediates mammalian nuclear RNA

N6-adenosine methylation. Nature chemical biology, 10(2), pp.93–5. Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3911877&tool=pmcent

rez&rendertype=abstract.

Liu, N. et al., 2015. N(6)-methyladenosine-dependent RNA structural switches regulate

RNA-protein interactions. Nature, 518(7540), pp.560–564. Available at:

http://eutils.ncbi.nlm.nih.gov/entrez/eutils/elink.fcgi?dbfrom=pubmed&id=257196

71&retmode=ref&cmd=prlinks%5Cnpapers2://publication/doi/10.1038/nature1423

4.

Liu, N. & Pan, T., 2017. RNA Epigenetics (Epitranscriptomics). In Translating

Epigenetics to the Clinic. pp. 19–35.

Liu, S. et al., 2012. CD8+ lymphocyte infiltration is an independent favorable

prognostic indicator in basal-like breast cancer. Breast Cancer Res, 14(2), p.R48.

Available at:

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3446382/pdf/bcr3148.pdf.

Liu, Y. et al., 2012. Phosphorylation of RelA/p65 promotes DNMT-1 recruitment to

chromatin and represses transcription of the tumor metastasis suppressor gene

BRMS1. Oncogene, 31(9), pp.1143–1154.

Liutkevičiutè, Z. et al., 2014. Direct decarboxylation of 5-Carboxylcytosine by DNA

Page 191: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

191

C5- Methyltransferases. Journal of the American Chemical Society, 136(16),

pp.5884–5887.

Lokk, K. et al., 2014. DNA methylome profiling of human tissues identifies global and

tissue-specific methylation patterns. Genome biology, 15(4), p.r54. Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=4053947&tool=pmcent

rez&rendertype=abstract.

Lu, L. et al., 2012. Association of large noncoding RNA HOTAIR expression and its

downstream intergenic CpG island methylation with survival in breast cancer.

Breast Cancer Research and Treatment, 136(3), pp.875–883.

Luo, G. et al., 2014. Unique features of the m 6 A methylome in Arabidopsis thaliana.

Nature Communications, 5, pp.1–8. Available at:

http://dx.doi.org/10.1038/ncomms6630.

Luo, Y., Lu, X. & Xie, H., 2014. Dynamic Alu Methylation during Normal

Development, Aging, and Tumorigenesis. BioMed Research International, 2014.

Maeda, S. et al., 2005. IKKβ couples hepatocyte death to cytokine-driven compensatory

proliferation that promotes chemical hepatocarcinogenesis. Cell, 121(7), pp.977–

990.

Mahmoud, S.M.A. et al., 2011. Tumor-infiltrating CD8+ lymphocytes predict clinical

outcome in breast cancer. Journal of Clinical Oncology, 29(15), pp.1949–1955.

Maicher, A. et al., 2012. Deregulated telomere transcription causes replication-

dependent telomere shortening and promotes cellular senescence. Nucleic Acids

Research, 40(14), pp.6649–6659.

Maity, A. & Das, B., 2016. N6-methyladenosine modification in mRNA: Machinery,

function and implications for health and diseases. FEBS Journal, 283(9), pp.1607–

1630.

Manal, M. et al., 2016. Inhibitors of histone deacetylase as antitumor agents: A critical

review. Bioorganic Chemistry, 67, pp.18–42.

Mantovani, A., 2010. Molecular pathways linking inflammation and cancer. Current

molecular medicine, 10(4), pp.369–73.

Margol, A.S. & Judkins, A.R., 2014. Pathology and diagnosis of SMARCB1-deficient

tumors. Cancer Genetics, 207(9), pp.358–364.

Masliah-Planchon, J. et al., 2015. SWI/SNF Chromatin Remodeling and Human

Malignancies. Annual Review of Pathology: Mechanisms of Disease, 10(1),

pp.145–171. Available at: http://www.annualreviews.org/doi/10.1146/annurev-

pathol-012414-040445.

Maunakea, A.K. et al., 2010. Conserved role of intragenic DNA methylation in

regulating alternative promoters. Nature, 466(7303), pp.253–7. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/20613842%5Cnhttp://www.pubmedcentral.n

ih.gov/articlerender.fcgi?artid=PMC3998662.

McArthur, H.L., 2016. Checkpoint inhibitors in breast cancer: Hype or promise?

Clinical Advances in Hematology and Oncology, 14(6), pp.392–395.

Melichar, B. et al., 2014. Predictive and prognostic significance of tumor-infiltrating

Page 192: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

192

lymphocytes in patients with breast cancer treated with neoadjuvant systemic

therapy. Anticancer research, 34(3), pp.1115–25. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/24596349 [Accessed February 24, 2015].

Mellén, M. et al., 2012. MeCP2 binds to 5hmC enriched within active genes and

accessible chromatin in the nervous system. Cell, 151(7), pp.1417–30. Available

at: http://www.ncbi.nlm.nih.gov/pubmed/23260135 [Accessed July 12, 2014].

Mercurio, F., 1997. IKK-1 and IKK-2: Cytokine-Activated IB Kinases Essential for NF-

B Activation. Science, 278(5339), pp.860–866. Available at:

http://www.sciencemag.org/cgi/doi/10.1126/science.278.5339.860.

Merkestein, M. et al., 2015. FTO influences adipogenesis by regulating mitotic clonal

expansion. Nature Communications, 6, p.6792. Available at:

http://www.nature.com/doifinder/10.1038/ncomms7792.

Meyer, K.D. et al., 2015. 5′ UTR m6A Promotes Cap-Independent Translation. Cell,

163(4), pp.999–1010.

Meyer, K.D. et al., 2012. Comprehensive analysis of mRNA methylation reveals

enrichment in 3′ UTRs and near stop codons. Cell, 149(7), pp.1635–1646.

Meyer, K.D. & Jaffrey, S.R., 2014. The dynamic epitranscriptome: N6-methyladenosine

and gene expression control. Nature reviews. Molecular cell biology, 15, pp.313–

26. Available at: http://www.ncbi.nlm.nih.gov/pubmed/24713629.

Minor, E.A. et al., 2013. Ascorbate induces ten-eleven translocation (Tet)

methylcytosine dioxygenase-mediated generation of 5-hydroxymethylcytosine.

Journal of Biological Chemistry, 288(19), pp.13669–13674.

Mirabella, A.C., Foster, B.M. & Bartke, T., 2016. Chromatin deregulation in disease.

Chromosoma, 125(1), pp.75–93.

Mirza, S. et al., 2010. Demethylating agent 5-aza-2-deoxycytidine enhances

susceptibility of breast cancer cells to anticancer agents. Molecular and Cellular

Biochemistry, 342(1–2), pp.101–109.

Mohrmann, L. & Verrijzer, C.P., 2005. Composition and functional specificity of

SWI2/SNF2 class chromatin remodeling complexes. Biochimica et Biophysica

Acta - Gene Structure and Expression, 1681(2–3), pp.59–73.

Monteiro, R. & Azevedo, I., 2010. Chronic inflammation in obesity and the metabolic

syndrome. Mediators of Inflammation, 2010.

Moran, S., Arribas, C. & Esteller, M., 2015. Validation of a DNA methylation

microarray for 850,000 CpG sites of the human genome enriched in enhancer

sequences. Epigenomics, 6(July 2015), p.epi.15.114. Available at:

http://www.futuremedicine.com/doi/10.2217/epi.15.114.

Mosammaparast, N. & Shi, Y., 2010. Reversal of histone methylation: biochemical and

molecular mechanisms of histone demethylases. Annual review of biochemistry,

79, pp.155–179.

Müller, H.M. et al., 2003. DNA Methylation in Serum of Breast Cancer Patients: An

Independent Prognostic Marker. Cancer Research, 63(22), pp.7641–7645.

Müller, H.M. et al., 2004. Prognostic DNA methylation marker in serum of cancer

Page 193: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

193

patients. In Annals of the New York Academy of Sciences. pp. 44–49.

Murata, H. et al., 2005. Roles of mismatch repair proteins hMSH2 and hMLH1 in the

development of sporadic breast cancer. Cancer Letters, 223(1), pp.143–150.

Nagasawa, S. et al., 2015. LSD1 overexpression is associated with poor prognosis in

basal-like breast cancer, and sensitivity to PARP inhibition. PLoS ONE, 10(2).

Nair, S.S. et al., 2011. Comparison of methyl-DNA immunoprecipitation (MeDIP) and

methyl-CpG binding domain (MBD) protein capture for genome-wide DNA

methylation analysis reveal CpG sequence coverage bias. Epigenetics, 6(1), pp.34–

44.

Navarro, A. et al., 2014. 5-hydroxymethylcytosine promotes proliferation of human

uterine leiomyoma: A biological link to a new epigenetic modification in benign

tumors. Journal of Clinical Endocrinology and Metabolism, 99(11), pp.E2437–

E2445.

Nelson, B.H., 2010. CD20+ B cells: the other tumor-infiltrating lymphocytes. J

Immunol, 185(9), pp.4977–4982. Available at:

http://www.jimmunol.org/content/185/9/4977.full.pdf.

Neri, F. et al., 2017. Intragenic DNA methylation prevents spurious transcription

initiation. Nature, 543(7643), pp.72–77. Available at:

http://www.nature.com/doifinder/10.1038/nature21373.

Nestor, C.E. et al., 2012. Tissue type is a major modifier of the 5-

hydroxymethylcytosine content of human genes. Genome Research, 22(3),

pp.467–477.

Nestor, C.E. & Meehan, R.R., 2014. Hydroxymethylated DNA immunoprecipitation

(hmeDIP). Methods in Molecular Biology, 1094, pp.259–267.

Netanely, D. et al., 2016. Expression and methylation patterns partition luminal-A

breast tumors into distinct prognostic subgroups. Breast Cancer Research, 18(1),

p.74. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/27386846%5Cnhttp://www.pubmedcentral.n

ih.gov/articlerender.fcgi?artid=PMC4936004%5Cnhttp://breast-cancer-

research.biomedcentral.com/articles/10.1186/s13058-016-0724-2.

Niehrs, C. & Schäfer, A., 2012. Active DNA demethylation by Gadd45 and DNA

repair. Trends in Cell Biology, 22(4), pp.220–227.

Oble, D.A. et al., 2009. Focus on TILs: Prognostic significance of tumor infiltrating

lymphocytes in human melanoma. Cancer Immunity, 9.

Oeckinghaus, A. & Ghosh, S., 2009. The NF-kappaB family of transcription factors and

its regulation. Cold Spring Harbor perspectives in biology, 1(4), p.a000034.

Oike, T. et al., 2013. A synthetic lethality-based strategy to treat cancers harboring a

genetic deficiency in the chromatin remodeling factor BRG1. Cancer Research,

73(17), pp.5508–5518.

Oka, M. et al., 2005. De novo DNA methyltransferases Dnmt3a and Dnmt3b primarily

mediate the cytotoxic effect of 5-aza-2’-deoxycytidine. Oncogene, 24, pp.3091–

3099.

Page 194: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

194

Ono, R. et al., 2002. LCX, leukemia-associated protein with a CXXC domain, is fused

to MLL in acute myeloid leukemia with trilineage dysplasia having

t(10;11)(q22;q23). Cancer Research, 62(14), pp.4075–4080.

Orucevic, A. et al., 2015. Breast Cancer in Elderly Caucasian Women-An Institution-

Based Study of Correlation between Breast Cancer Prognostic Markers, TNM

Stage, and Overall Survival. Cancers, 7(3), pp.1472–83. Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=4586779&tool=pmcent

rez&rendertype=abstract.

Otani, J. et al., 2013. Structural basis of the versatile DNA recognition ability of the

methyl-CpG binding domain of methyl-CpG binding domain protein 4. The

Journal of biological chemistry, 288(9), pp.6351–6362. Available at:

http://eutils.ncbi.nlm.nih.gov/entrez/eutils/elink.fcgi?dbfrom=pubmed&id=233160

48&retmode=ref&cmd=prlinks%5Cnpapers2://publication/doi/10.1074/jbc.M112.

431098.

Paine, T.M. et al., 1992. Characterization of epithelial phenotypes in mortal and

immortal human breast cells. International Journal of Cancer, 50(3), pp.463–473.

Pakneshan, P., Szyf, M. & Rabbani, S.A., 2005. Hypomethylation of urokinase (uPA)

promoter in breast and prostate cancer: prognostic and therapeutic implications.

Current cancer drug targets, 5(7), pp.471–88. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/16305345.

Pan, T., 2013. N6-methyl-adenosine modification in messenger and long non-coding

RNA. Trends in Biochemical Sciences, 38(4), pp.204–209.

Paredes, J. et al., 2005. P-cadherin overexpression is an indicator of clinical outcome in

invasive breast carcinomas and is associated with CDH3 promoter

hypomethylation. Clinical Cancer Research, 11(16), pp.5869–5877.

Park, S.-B. et al., 2015. Differential Epigenetic Effects of Atmospheric Cold Plasma on

MCF-7 and MDA-MB-231 Breast Cancer Cells. PLOS ONE, 10(6), p.e0129931.

Available at: http://dx.plos.org/10.1371/journal.pone.0129931.

Park, S.Y. et al., 2011. Promoter CpG island hypermethylation during breast cancer

progression. Virchows Archiv, 458(1), pp.73–84.

Parker, J.S. et al., 2009. Supervised risk predictor of breast cancer based on intrinsic

subtypes. Journal of Clinical Oncology, 27(8), pp.1160–1167.

Pastor, W.A. et al., 2011. Genome-wide mapping of 5-hydroxymethylcytosine in

embryonic stem cells. Nature, 473(7347), pp.394–7. Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3124347&tool=pmcent

rez&rendertype=abstract.

Pastor, W. a et al., 2012. The GLIB technique for genome-wide mapping of 5-

hydroxymethylcytosine. Nature Protocols, 7(10), pp.1909–1917.

Perou, C.M. et al., 2000. Molecular portraits of human breast tumours. Nature,

406(6797), pp.747–752.

Phan, N.L.-C., Trinh, N. Van & Pham, P. Van, 2016. Low concentrations of 5-aza-2’-

deoxycytidine induce breast cancer stem cell differentiation by triggering tumor

suppressor gene expression. OncoTargets and therapy, 9, pp.49–59. Available at:

Page 195: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

195

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=4694670&tool=pmcent

rez&rendertype=abstract.

Pitta, C.A. & Constantinou, A.I., 2015. Breast Cancer Epigenetics. In Epigenetic

Cancer Therapy. pp. 215–232.

Ponnaluri, V.K.C. et al., 2017. Association of 5-hydroxymethylation and 5-methylation

of DNA cytosine with tissue-specific gene expression. Epigenetics, 12(2), pp.123–

138.

Portela, A. & Esteller, M., 2010. Epigenetic modifications and human disease. Nature

biotechnology, 28(10), pp.1057–1068. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/20944598%5Cnhttp://dx.doi.org/10.1038/nb

t.1685.

Postepska-Igielska, A. et al., 2013. The chromatin remodelling complex NoRC

safeguards genome stability by heterochromatin formation at telomeres and

centromeres. EMBO reports, 14(8), pp.704–710. Available at:

http://embor.embopress.org/cgi/doi/10.1038/embor.2013.87.

Raiber, E.-A. et al., 2012. Genome-wide distribution of 5-formylcytosine in embryonic

stem cells is associated with transcription and depends on thymine DNA

glycosylase. Genome Biology, 13(8), p.R69. Available at:

http://genomebiology.com/2012/13/8/R69.

Rampal, R. et al., 2014. DNA Hydroxymethylation Profiling Reveals that WT1

Mutations Result in Loss of TET2 Function in Acute Myeloid Leukemia. Cell

Reports, 9(5), pp.1841–1856.

Rasmussen, K.D. et al., 2015. Loss of TET2 in hematopoietic cells leads to DNA

hypermethylation of active enhancers and induction of leukemogenesis. Genes and

Development, 29(9), pp.910–922.

Ratel, D. et al., 2006. N6-methyladenine: The other methylated base of DNA.

BioEssays, 28(3), pp.309–315.

Riggs, A.D. & Porter, T.N., 1996. Overview of Epigenetic Mechanisms. In Epigenetic

Mechanisms of Gene Regulation. pp. 29–45.

Robert, C. et al., 2011. Ipilimumab plus dacarbazine for previously untreated metastatic

melanoma. The New England journal of medicine, 364(26), pp.2517–2526.

Rodríguez-Paredes, M. & Esteller, M., 2011. Cancer epigenetics reaches mainstream

oncology. Nature medicine, 17(3), pp.330–339. Available at:

http://dx.doi.org/10.1038/nm.2305.

Rossell, D. et al., 2014. Quantifying alternative splicing from paired-end RNA-

sequencing data. Annals of Applied Statistics, 8(1), pp.309–330.

Roulois, D. et al., 2015. DNA-Demethylating Agents Target Colorectal Cancer Cells by

Inducing Viral Mimicry by Endogenous Transcripts. Cell, 162(5), pp.961–973.

Russo, V.E.A., Martienssen, R.A. & Riggs, A.D., 1996. Epigenetic mechanisms of gene

regulation. Cold Spring Harbor Laboratory Press, p.692.

Sabatier, R. et al., 2014. Claudin-low breast cancers: clinical, pathological, molecular

and prognostic characterization. Molecular cancer, 13(1), p.228. Available at:

Page 196: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

196

http://molecular-cancer.biomedcentral.com/articles/10.1186/1476-4598-13-228.

Sasidharan Nair, V., Song, M.H. & Oh, K.I., 2016. Vitamin C Facilitates Demethylation

of the Foxp3 Enhancer in a Tet-Dependent Manner. The Journal of Immunology,

196(5), pp.2119–2131. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/26826239%5Cnhttp://www.jimmunol.org/lo

okup/doi/10.4049/jimmunol.1502352.

Sawicka, A. & Seiser, C., 2012. Histone H3 phosphorylation - A versatile chromatin

modification for different occasions. Biochimie, 94(11), pp.2193–2201.

Schaefer, M. et al., 2009. RNA cytosine methylation analysis by bisulfite sequencing.

Nucleic acids research, 37(2), p.e12. Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2632927&tool=pmcent

rez&rendertype=abstract [Accessed January 27, 2015].

Schiaffino-Ortega, S. et al., 2014. SWI/SNF proteins as targets in cancer therapy.

Journal of hematology & oncology, 7(1), p.81. Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=4236461&tool=pmcent

rez&rendertype=abstract.

Schiesser, S. et al., 2012. Mechanism and stem-cell activity of 5-carboxycytosine

decarboxylation determined by isotope tracing. Angewandte Chemie -

International Edition, 51(26), pp.6516–6520.

Schumacker, P., 2015. Reactive Oxygen Species in Cancer: A Dance with the Devil.

Cancer Cell, 27(2), pp.156–157.

Scourzic, L., Mouly, E. & Bernard, O.A., 2015. TET proteins and the control of

cytosine demethylation in cancer. Genome medicine, 7(1), p.9. Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=4308928&tool=pmcent

rez&rendertype=abstract.

Sehouli, J. et al., 2011. Epigenetic quantification of tumor-infiltrating T-lymphocytes.

Epigenetics, 6(2), pp.236–246.

Seidel, C. et al., 2012. Chromatin-modifying agents in anti-cancer therapy. Biochimie,

94(11), pp.2264–2279.

Seo, A.N. et al., 2013. Tumour-infiltrating CD8+ lymphocytes as an independent

predictive factor for pathological complete response to primary systemic therapy in

breast cancer. British journal of cancer, 109(10), pp.2705–13. Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3833219&tool=pmcent

rez&rendertype=abstract.

Severi, G. et al., 2014. Epigenome-wide methylation in DNA from peripheral blood as a

marker of risk for breast cancer. Breast Cancer Research and Treatment, 148(3),

pp.665–673.

Sha, K. & Boyer, L.A., 2008. The chromatin signature of pluripotent cells. StemBook,

pp.1–21. Available at: http://www.ncbi.nlm.nih.gov/pubmed/20614601.

Shan, M. et al., 2016. Detection of aberrant methylation of a six-gene panel in serum

DNA for diagnosis of breast cancer. Oncotarget, 7(14), pp.18485–18494.

Shann, Y.-J. et al., 2008. Genome-wide mapping and characterization of

Page 197: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

197

hypomethylated sites in human tissues and breast cancer cell lines. Genome

research, 18(5), pp.791–801.

Sharma, D. et al., 2006. Restoration of tamoxifen sensitivity in estrogen receptor-

negative breast cancer cells: Tamoxifen-bound reactivated ER recruits distinctve

corepressor complexes. Cancer Research, 66(12), pp.6370–6378.

Sharma, G. et al., 2010. CpG hypomethylation of MDR1 gene in tumor and serum of

invasive ductal breast carcinoma patients. Clinical Biochemistry, 43(4–5), pp.373–

379.

Shen, L. et al., 2013. Genome-wide analysis reveals TET- and TDG-dependent 5-

methylcytosine oxidation dynamics. Cell, 153(3), pp.692–706.

Shetty, P.J. et al., 2011. Regulation of IGF2 transcript and protein expression by altered

methylation in breast cancer. Journal of Cancer Research and Clinical Oncology,

137(2), pp.339–345.

Shi, J. et al., 2014. Disrupting the Interaction of BRD4 with Diacetylated Twist

Suppresses Tumorigenesis in Basal-like Breast Cancer. Cancer Cell, 25(2),

pp.210–225.

Shi, J. & Vakoc, C.R., 2014. The Mechanisms behind the Therapeutic Activity of BET

Bromodomain Inhibition. Molecular Cell, 54(5), pp.72–736.

Shibata, A. et al., 2002. Inhibition of NF-kappaB activity decreases the VEGF mRNA

expression in MDA-MB-231 breast cancer cells. Breast cancer research and

treatment, 73(3), pp.237–43. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/12160329.

Shibata, W. et al., 2010. Conditional Deletion of IKKB-Kinase-beta Accelerates

Helicobacter-Dependent Gastric Apoptosis, Proliferation, and Preneoplasia.

Gastroenterology, 138(3).

Shore, A. et al., 2010. Role of Ucp1 enhancer methylation and chromatin remodelling in

the control of Ucp1 expression in murine adipose tissue. Diabetologia, 53(6),

pp.1164–1173.

Siegel, R.L., Miller, K.D. & Jemal, A., 2016. Cancer statistics, 2016. CA Cancer J Clin,

66(1), pp.7–30. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/26742998%5Cnhttp://onlinelibrary.wiley.co

m/store/10.3322/caac.21332/asset/caac21332.pdf?v=1&t=io59driw&s=f5d356a70

462e7927a65b0ec8811c011b92d7469.

Silva, S. et al., 2011. Why are monozygotic twins different? Journal of Perinatal

Medicine, 39(2), pp.195–202.

Silverman, L.R. et al., 2002. Randomized controlled trial of azacitidine in patients with

the myelodysplastic syndrome: a study of the cancer and leukemia group B.

Journal of clinical oncology : official journal of the American Society of Clinical

Oncology, 20(10), pp.2429–2440. Available at:

http://www.jco.org/cgi/doi/10.1200/JCO.2002.04.117%5Cnhttp://www.ncbi.nlm.ni

h.gov/pubmed/12011120.

Song, C.-X. et al., 2011. Selective chemical labeling reveals the genome-wide

distribution of 5-hydroxymethylcytosine. Nature biotechnology, 29(1), pp.68–72.

Page 198: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

198

Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3107705&tool=pmcent

rez&rendertype=abstract [Accessed February 3, 2015].

Song, S.J. et al., 2013. MicroRNA-antagonism regulates breast cancer stemness and

metastasis via TET-family-dependent chromatin remodeling. Cell, 154(2), pp.311–

24. Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3767157&tool=pmcent

rez&rendertype=abstract [Accessed January 31, 2015].

Sorlie, T. et al., 2001. Gene expression patterns of breast carcinomas distinguish tumor

subclasses with clinical implications. Proceedings of the National Academy of

Sciences of the United States of America, 98(19), pp.10869–10874. Available at:

http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt

=Citation&list_uids=11553815%5Cnhttp://www.ncbi.nlm.nih.gov/pmc/articles/P

MC58566/pdf/pq010869.pdf.

Sorlie, T. et al., 2003. Repeated observation of breast tumor subtypes in independent

gene expression data sets. Proceedings of the National Academy of Sciences of the

United States of America, 100(14), pp.8418–23. Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=166244&tool=pmcentr

ez&rendertype=abstract.

Sotiriou, C. et al., 2003. Breast cancer classification and prognosis based on gene

expression profiles from a population-based study. Proceedings of the National

Academy of Sciences, 100(18), pp.10393–10398. Available at:

http://dx.doi.org/10.1073/pnas.1732912100.

Sotiriou, C. et al., 2006. Gene expression profiling in breast cancer: Understanding the

molecular basis of histologic grade to improve prognosis. Journal of the National

Cancer Institute, 98(4), pp.262–272.

Spruijt, C.G. et al., 2013. Dynamic readers for 5-(Hydroxy)methylcytosine and its

oxidized derivatives. Cell, 152(5), pp.1146–1159.

Squires, J.E. et al., 2012. Widespread occurrence of 5-methylcytosine in human coding

and non-coding RNA. Nucleic acids research, 40 VN-r(11), pp.5023–5033.

Available at: /Users/yurikoharigaya/Documents/ReadCube Media/preiss RNA

methylation.pdf%5Cnhttp://dx.doi.org/10.1093/nar/gks144.

Stefansson, O.A. & Esteller, M., 2014. CARM1 and BAF155: an example of how

chromatin remodeling factors can be relocalized and contribute to cancer. Breast

Cancer Research, 16(3), p.307. Available at: http://breast-cancer-

research.biomedcentral.com/articles/10.1186/bcr3657.

Stolzenburg, S. et al., 2012. Targeted silencing of the oncogenic transcription factor

SOX2 in breast cancer. Nucleic Acids Research, 40(14), pp.6725–6740.

Sun, M. et al., 2013. HMGA2/TET1/HOXA9 signaling pathway regulates breast cancer

growth and metastasis. Proceedings of the National Academy of Sciences of the

United States of America, 110(24), pp.9920–5. Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3683728&tool=pmcent

rez&rendertype=abstract [Accessed February 25, 2015].

Szerlong, H.J. & Hansen, J.C., 2011. Nucleosome distribution and linker DNA:

Page 199: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

199

connecting nuclear function to dynamic chromatin structure. Biochemistry and cell

biology, 89(1), pp.24–34. Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3125042&tool=pmcent

rez&rendertype=abstract.

Szulwach, K.E. et al., 2011. 5-hmC-mediated epigenetic dynamics during postnatal

neurodevelopment and aging. Nat Neurosci, 14(12), pp.1607–1616. Available at:

http://dx.doi.org/10.1038/nn.2959.

Szwagierczak, A. et al., 2010. Sensitive enzymatic quantification of 5-

hydroxymethylcytosine in genomic DNA. Nucleic Acids Research, 38(19).

Tahiliani, M. et al., 2009. Conversion of 5-methylcytosine to 5-hydroxymethylcytosine

in mammalian DNA by MLL partner TET1. Science, 324(5929), pp.930–5.

Available at:

http://www.ncbi.nlm.nih.gov/pubmed/19372391%5Cnhttp://www.pubmedcentral.n

ih.gov/articlerender.fcgi?artid=PMC2715015.

Tamaru, H., 2010. Confining euchromatin/heterochromatin territory: Jumonji crosses

the line. Genes and Development, 24(14), pp.1465–1478.

Thienpont, B. et al., 2016. Tumour hypoxia causes DNA hypermethylation by reducing

TET activity. Nature, 537(7618), pp.63–68. Available at:

http://www.nature.com/doifinder/10.1038/nature19081.

Thomson, J.P. et al., 2013. Dynamic changes in 5-hydroxymethylation signatures

underpin early and late events in drug exposed liver. Nucleic Acids Research,

41(11), pp.5639–5654.

Thorleifsson, G. et al., 2009. Genome-wide association yields new sequence variants at

seven loci that associate with measures of obesity. Nature genetics, 41(1), pp.18–

24.

Torres, A.G., Batlle, E. & Ribas de Pouplana, L., 2014. Role of tRNA modifications in

human diseases. Trends in Molecular Medicine, 20(6), pp.306–314.

Trojer, P. & Reinberg, D., 2007. Facultative Heterochromatin: Is There a Distinctive

Molecular Signature? Molecular Cell, 28(1), pp.1–13.

Tuorto, F. et al., 2012. RNA cytosine methylation by Dnmt2 and NSun2 promotes

tRNA stability and protein synthesis. Nature Structural & Molecular Biology,

19(9), pp.900–905. Available at: http://dx.doi.org/10.1038/nsmb.2357.

Uribe-Lewis, S. et al., 2015. 5-hydroxymethylcytosine marks promoters in colon that

resist DNA hypermethylation in cancer. Genome biology, 16(1), p.69. Available at:

http://genomebiology.com/2015/16/1/69.

Valdespino, V. & Valdespino, P.M., 2015. Potential of epigenetic therapies in the

management of solid tumors. Cancer Management and Research, 7, pp.241–251.

Valinluck, V. & Sowers, L.C., 2007. Endogenous cytosine damage products alter the

site selectivity of human DNA maintenance methyltransferase DNMT1. Cancer

Research, 67(3), pp.946–950.

Veeck, J. et al., 2010. BRCA1 CpG island hypermethylation predicts sensitivity to

poly(adenosine diphosphate)-ribose polymerase inhibitors. Journal of Clinical

Page 200: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

200

Oncology, 28(29).

Vella, P. et al., 2013. Tet Proteins Connect the O-Linked N-acetylglucosamine

Transferase Ogt to Chromatin in Embryonic Stem Cells. Molecular Cell, 49(4),

pp.645–656.

Waddington, C.H., 1942. The Epigenotype. Endeavour, pp.18–20. Available at:

http://www.ije.oxfordjournals.org/lookup/doi/10.1093/ije/dyr184.

Wade, N., 2009. From One Genome, Many Types of Cells. But How ? New York Times,

p.D4. Available at:

http://www.nytimes.com/2009/02/24/science/24chromatin.html?pagewanted=all&

_r=0.

Wallden, B. et al., 2015. Development and verification of the PAM50-based Prosigna

breast cancer gene signature assay. BMC medical genomics, 8, p.54. Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=4546262&tool=pmcent

rez&rendertype=abstract.

Walton, E.L., Francastel, C. & Velasco, G., 2011. Maintenance of DNA methylation:

Dnmt3b joins the dance. Epigenetics, 6(11), pp.1373–1377.

Wang, L., Tang, L. & Xie, R., 2012. p16 promoter hypermethylation is associated with

increased breast cancer risk. Molecular medicine …, 6(4), pp.904–8. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/22824969%5Cnhttp://www.spandidos-

publications.com/mmr/6/4/904?text=fulltext.

Wang, T. et al., 2015. Fluorescein Derivatives as Bifunctional Molecules for the

Simultaneous Inhibiting and Labeling of FTO Protein. Journal of the American

Chemical Society, 137(43), pp.13736–13739.

Wang, W. et al., 1996. Purification and biochemical heterogeneity of the mammalian

SWI-SNF complex. The EMBO journal, 15(19), pp.5370–5382.

Wang, X. et al., 2014. N6-methyladenosine-dependent regulation of messenger RNA

stability. Nature, 505(7481), pp.117–20. Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3877715&tool=pmcent

rez&rendertype=abstract.

Wang, X. et al., 2015. N6-methyladenosine modulates messenger RNA translation

efficiency. Cell, 161(6), pp.1388–1399.

Wang, Y. et al., 2014. N6-methyladenosine modification destabilizes developmental

regulators in embryonic stem cells. Nature cell biology, 16(2), pp.191–8. Available

at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=4640932&tool=pmcent

rez&rendertype=abstract.

Warton, K., Mahon, K.L. & Samimi, G., 2016. Methylated circulating tumor DNA in

blood: Power in cancer prognosis and response. Endocrine-Related Cancer, 23(3),

pp.R157–R171.

Weigelt, B. et al., 2010. Breast cancer molecular profiling with single sample

predictors: A retrospective analysis. The Lancet Oncology, 11(4), pp.339–349.

Wheldon, L.M. et al., 2014. Transient accumulation of 5-carboxylcytosine indicates

Page 201: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

201

involvement of active demethylation in lineage specification of neural stem cells.

Cell Reports, 7(5), pp.1353–1361.

WHO, 2012. Globocan 2012 - Home. Globocan 2012. Available at:

http://globocan.iarc.fr/Default.aspx.

Willbanks, A. et al., 2016. The Evolution of Epigenetics: From Prokaryotes to Humans

and Its Biological Consequences. Genetics & epigenetics, 8, pp.25–36. Available

at: http://www.ncbi.nlm.nih.gov/pubmed/27512339.

Williams, K., Christensen, J. & Helin, K., 2012. DNA methylation: TET proteins-

guardians of CpG islands? EMBO reports, 13(1), pp.28–35. Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3246258&tool=pmcent

rez&rendertype=abstract%5Cnhttp://dx.doi.org/10.1038/embor.2011.233%5Cnhttp

://www.ncbi.nlm.nih.gov/pubmed/22157888%5Cnhttp://www.pubmedcentral.nih.g

ov/articlerender.fcgi?artid=PM.

Wilson, B.G. & Roberts, C.W.M., 2011. SWI/SNF nucleosome remodellers and cancer.

Nature Reviews Cancer, 11(7), pp.481–492. Available at:

http://www.nature.com/doifinder/10.1038/nrc3068.

Witt, O. et al., 2009. HDAC family: What are the cancer relevant targets? Cancer

Letters, 277(1), pp.8–21.

Wolf, K.W. & Sumner, A.T., 1996. Scanning electron microscopy of heterochromatin

in chromosome spreads of male germ cells in Schistocerca gregaria (Acrididae,

Orthoptera) after trypsinization. Biotech Histochem, 71(5), pp.237–244. Available

at:

http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt

=Citation&list_uids=8896797.

Wolfson, B., Eades, G. & Zhou, Q., 2015. Adipocyte activation of cancer stem cell

signaling in breast cancer. World journal of biological chemistry, 6(2), pp.39–47.

Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=4436905&tool=pmcent

rez&rendertype=abstract.

Wu, H. et al., 2011. Genome-wide analysis of 5-hydroxymethylcytosine distribution

reveals its dual function in transcriptional regulation in mouse embryonic stem

cells. Genes & development, 25(7), pp.679–84. Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3070931&tool=pmcent

rez&rendertype=abstract [Accessed January 8, 2015].

Wu, H. & Zhang, Y., 2011. Mechanisms and functions of Tet protein- mediated 5-

methylcytosine oxidation. Genes & Development, 25, pp.2436–2452.

Wu, M.Z. et al., 2015. Hypoxia drives breast tumor malignancy through a TET-TNFα-

p38-MAPK signaling axis. Cancer Research, 75(18), pp.3912–3924.

Wu, R. et al., 2016. N6-Methyladenosine (m6A) Methylation in mRNA with A

Dynamic and Reversible Epigenetic Modification. Molecular Biotechnology,

58(7), pp.450–459.

Wu, S.C. & Zhang, Y., 2010. Active DNA demethylation: many roads lead to Rome.

Nat Rev Mol Cell Biol, 11(9), pp.607–620. Available at:

Page 202: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

202

http://www.ncbi.nlm.nih.gov/pubmed/20683471.

Wutz, A., 2011. Gene silencing in X-chromosome inactivation: advances in

understanding facultative heterochromatin formation. Nature Reviews Genetics,

12(8), pp.542–553. Available at:

http://www.nature.com/doifinder/10.1038/nrg3035.

Xiao, W. et al., 2016. A Reader YTHDC1 Regulates mRNA Splicing. Molecular Cell,

61, pp.1–13. Available at: http://dx.doi.org/10.1016/j.molcel.2016.01.012.

Xin, Y.-J. et al., 2015. Tet1-mediated DNA demethylation regulates neuronal cell death

induced by oxidative stress. Scientific reports, 5, p.7645. Available at:

http://www.ncbi.nlm.nih.gov/pubmed/25561289.

Xu, Y. et al., 2013. Promoter methylation of BRCA1 in triple-negative breast cancer

predicts sensitivity to adjuvant chemotherapy. Annals of Oncology, 24(6),

pp.1498–1505.

Yamazaki, J. et al., 2015. TET2 mutations affect Non-CpG island DNA methylation at

enhancers and transcription factor-binding sites in chronic myelomonocytic

Leukemia. Cancer Research, 75(14), pp.2833–2843.

Yang, C. et al., 2016. TET1 and TET3 are essential in induction of Th2-type immunity

partly through regulation of IL-4/13A expression in zebrafish model. Gene, 591(1),

pp.201–208.

Yang, J.C. et al., 2007. Ipilimumab (Anti-CTLA4 Antibody) Causes Regression of

Metastatic Renal Cell Cancer Associated With Enteritis and Hypophysitis. Journal

of Immunotherapy, 30(8), pp.825–830. Available at:

http://content.wkhealth.com/linkback/openurl?sid=WKPTLP:landingpage&an=000

02371-200711000-00005.

Yang, R. et al., 2015. Hydrogen Sulfide Promotes Tet1- and Tet2-Mediated Foxp3

Demethylation to Drive Regulatory T Cell Differentiation and Maintain Immune

Homeostasis. Immunity, 43(2), pp.251–263.

Yu, M. et al., 2012. Tet-assisted bisulfite sequencing of 5-hydroxymethylcytosine.

Nature protocols, 7(12), pp.2159–70. Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3641661&tool=pmcent

rez&rendertype=abstract.

Yue, Y., Liu, J. & He, C., 2015. RNA N6-methyladenosine methylation in post-

transcriptional gene expression regulation. Genes and Development, 29(13),

pp.1343–1355.

Zhang, C. et al., 2016. Hypoxia induces the breast cancer stem cell phenotype by HIF-

dependent and ALKBH5-mediated m6A-demethylation of NANOG mRNA.

Proceedings of the National Academy of Sciences of the United States of America,

113(14), pp.E2047-56. Available at:

http://www.pnas.org/content/113/14/E2047.short%5Cnhttp://www.ncbi.nlm.nih.go

v/pubmed/27001847%5Cnhttp://www.pubmedcentral.nih.gov/articlerender.fcgi?ar

tid=PMC4833258.

Zhang, J.Y. et al., 2005. CDK4 regulation by TNFR1 and JNK is required for NF-κB-

mediated epidermal growth control. Journal of Cell Biology, 168(4), pp.561–566.

Page 203: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

203

Zhang, Q. et al., 2015. Tet2 is required to resolve inflammation by recruiting Hdac2 to

specifically repress IL-6. Nature, 525(7569), pp.389–393. Available at:

http://www.nature.com/doifinder/10.1038/nature15252.

Zhang, S. et al., 2017. m6A Demethylase ALKBH5 Maintains Tumorigenicity of

Glioblastoma Stem-like Cells by Sustaining FOXM1 Expression and Cell

Proliferation Program. Cancer Cell, 31(4), p.591–606.e6.

Zhang, Z. et al., 2010. The YTH domain is a novel RNA binding domain. The Journal

of biological chemistry, 285(19), pp.14701–14710. Available at:

http://www.jbc.org/cgi/content/abstract/285/19/14701.

Zhao, B.S., Roundtree, I.A. & He, C., 2016. Post-transcriptional gene regulation by

mRNA modifications. Nature Reviews Molecular Cell Biology, 18(1), pp.31–42.

Available at: http://www.nature.com/doifinder/10.1038/nrm.2016.132.

Zhao, X. et al., 2014. FTO-dependent demethylation of N6-methyladenosine regulates

mRNA splicing and is required for adipogenesis. Cell research, 24(12), pp.1403–

19. Available at:

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=4260349&tool=pmcent

rez&rendertype=abstract%5Cnhttp://dx.doi.org/10.1038/cr.2014.151.

Zheng, G. et al., 2013. ALKBH5 Is a Mammalian RNA Demethylase that Impacts RNA

Metabolism and Mouse Fertility. Molecular Cell, 49(1), pp.18–29.

Zhou, R. et al., 2013. Histone Deacetylases and NF-kB Signaling Coordinate

Expression of CX3CL1 in Epithelial Cells in Response to Microbial Challenge by

Suppressing miR-424 and miR-503. PLoS ONE, 8(5).

Ziller, M.J. et al., 2011. Genomic distribution and Inter-Sample variation of Non-CpG

methylation across human cell types. PLoS Genetics, 7(12).

Page 204: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

204

Page 205: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

205

Appendix

Appendix

Page 206: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

206

Page 207: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

207

The following documents are provided in appendix:

• Manuscript I: Immunity drives TET1 regulation in cancer through NF-κB

(research article submitted for publication).

• Manuscript II: Transcriptome-wide distribution and function of RNA

hydroxymethylcytosine (research article published in Science in 2016).

• Manuscript III: DNA methylome profiling beyond promoters – taking an

epigenetic snapshot of the breast tumor microenvironment (review published in

the FEBS Journal in 2015).

• Manuscript IV: Portraits of TET-mediated DNA hydroxymethylation in cancer

(review published in Current Opinion in Genetics & Development in 2016)

• Additional methods (related to unpublished results presented in this thesis).

Page 208: Apache Tomcat · 2017-12-22 · 5 Résumé Dans le domaine de l’épigénétique, la méthylation de l’ADN et les modifications des histones ont longtemps focalisé l’attention

208