Upload
others
View
10
Download
0
Embed Size (px)
Citation preview
Research ArticleCOL1A1 Is a Potential Prognostic Biomarker and Correlated withImmune Infiltration in Mesothelioma
Cangang Zhang ,1 Shanshan Liu,2 Xin Wang,1 Haiyan Liu,1 Xiaobo Zhou ,1
and Haibo Liu 3
1Department of Pathogenic Microbiology and Immunology, School of Basic Medical Sciences, Xi’an Jiaotong University, Xi’an,710061 Shaanxi, China2Health Science Center, Xi’an Jiaotong University, Xi’an, 710061 Shaanxi, China3Department of Hematology, The First Affiliated Hospital of Xi’an Jiaotong University, Xi’an, 710061 Shaanxi, China
Correspondence should be addressed to Xiaobo Zhou; [email protected] and Haibo Liu; [email protected]
Received 24 May 2020; Revised 15 November 2020; Accepted 23 December 2020; Published 5 January 2021
Academic Editor: Lucian Ilie
Copyright © 2021 Cangang Zhang et al. This is an open access article distributed under the Creative Commons Attribution License,which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Objective. Mesothelioma (MESO) is a rare tumor derived from mesothelium cells. The aim of this study was to explore keycandidate genes and potential molecular mechanisms for mesothelioma through bioinformatics analysis. Methods. The MESOexpression profiles came from the Gene Expression Omnibus (GEO) and The Cancer Genome Atlas (TCGA) databases. Thedifferences in the infiltration levels of immune cells between MESO and normal tissues were assessed using CIBERSORT.Differentially expressed genes (DEGs) were identified by comprehensive analysis of multiple datasets. A protein-protein interaction(PPI) network was constructed, and a hub gene COL1A1 was selected for MESO. The expression and mutation of COL1A1 inMESO were analyzed in the cBioPortal database. The correlation between COL1A1 expression and immune cell infiltration wasevaluated using the TIMER database. Gene Set Enrichment Analysis (GSEA) of COL1A1 was then performed. Finally, Kaplan-Meier survival analysis was presented to predict the survival times between high and low COL1A1 expression groups for MESOpatients. Results. There were distinct differences in the infiltration levels of immune cells between MESO and normal tissues. Atotal of 118 DEGs were identified by comprehensively analyzing three expression profile datasets. COL1A1, a hub gene, wasidentified to be highly expressed in MESO compared to normal tissues. COL1A1 genetic mutation occurred in 9% of MESOsamples, and amplification was the most common type of mutation. COL1A1 expression was significantly correlated to theinfiltration levels of CD4+ T cells, macrophages, and neutrophils. GSEA results indicated that COL1A1 could be involved in keybiological processes and pathways like extracellular matrix and PI3K-Akt pathway. Patients with high COL1A1 expression usuallyexperienced shorten overall survival time than those with its low expression. Conclusion. Our findings revealed that COL1A1 couldbecome a potential prognostic biomarker for MESO, which was significantly related to immune cell infiltration.
1. Introduction
MESO is a rare tumor that mainly originates from mesothe-lial cells [1]. The incidence of MESO is on the rise in recentyears due to asbestos exposure [2]. Malignant pleural meso-thelioma (MPM) exhibits the highest incidence (81%) andthe worst prognosis among all cases of MESO [1]. MESOcan also occur in membranous structures in other parts,including peritoneum (9%), pericardium, and testicularsheath [3, 4]. The histological subtypes of MESO are com-
prise of epithelial (the most common), sarcomatoid, andbiphasic (mixture of epithelial and sarcomatoid) [5, 6].Patients with epithelial type often experience better progno-sis than those with mixed or sarcomatous type [5, 6]. Mostpatients with malignant MESO are at an advanced stage atthe time of diagnosis, with a median overall survival of only1 year and the 5-year overall survival rate of about 10% [7–9].
At present, surgical resection and subsequent chemother-apy are the main treatment strategies for MESO patients. Sys-temic chemotherapy based on cisplatin and pemetrexed is
HindawiBioMed Research InternationalVolume 2021, Article ID 5320941, 13 pageshttps://doi.org/10.1155/2021/5320941
the main palliative treatment for most malignant pleuralMESO patients [10]. MPM patients’ prognosis is usually poor,with a median survival of 12 months for patients receivingpemetrexed and cisplatin chemotherapy [11]. Immunother-apy has achieved significant results in a variety of tumors[12], which is expected to be used to improve the prognosisof MESO. However, whether the interaction between MESOand the immune system could affect the prognosis of MESOpatients remains unclear. Under normal physiological condi-tions, the immune system strictly controls cell proliferationand apoptosis with the help of immune cells [13]. In addition,tumor-infiltrating lymphocytes (TILs) play an important rolein tumor-related immune responses [14]. Several studies haveshown that TILs can effectively improve the prognosis of var-ious tumors [15–18]. Therefore, it is of great clinical signifi-cance to explore the infiltration of immune cells in MESO.
The purpose of this study was to identify key gene relatedto MESO patients’ clinical outcomes and to explore the cor-relation between its expression and immune cell infiltration,which could help us determine the treatment decisions andpredict the prognosis for patients. A hub gene, COL1A1was identified to be upregulated in MESO tissues comparedto normal tissues and significantly related to the infiltrationlevels of various immune cells. For MESO patients, highCOL1A1 expression indicated a poorer prognosis than itslow expression. Hence, it could become a potential prognos-tic biomarker for MESO.
2. Materials and Methods
2.1. Microarray Data. The GEO (https://www.ncbi.nlm.nih.gov/geo) is a public functional genomics data repositoryof high-throughput gene expression data, chips, andmicroarrays [19]. Three gene expression datasets includingGSE51024 (Affymetrix Human Genome U133 Plus 2.0Array) [20], GSE42977 (Illumina HumanRef-6 v2.0expression bead chip) [21], and GSE2549 (AffymetrixHuman Genome U133A Array) [22] were downloadedfrom the GEO database. The GSE51024 dataset contained55 MESO samples and 41 normal samples. The GSE42977included 39 MESO samples and 7 normal samples, andthe GSE2549 included 40 MESO samples and 5 normalsamples. Normal tissues were obtained from the adjacentcancer of the same MESO patient during surgery. Theprobes were transformed into the corresponding genesymbol according to the annotation information. Figure 1depicted the workflow of this study. The clinical informa-tion of MESO patients in the GSE51024 dataset was listedin Supplementary table 1.
2.2. Identification of DEGs. The DEGs between MESO andnormal samples were screened using the GEO2R (https://www.ncbi.nlm.nih.gov/geo/geo2r). GEO2R is an interactiveweb tool that enables users to compare two or more data-sets in a GEO sequence through experimental conditions.
Analysis the immune cell profiles by CIBERSORT
Identify COL1A1 as hub gene by using 8 different methods of cytoHubba
GSE2549(patients=40, controls=5)GSE42977(patients=39, controls=7)
Identification of common118 DEGs in 3 GEO datasets
GSE 51024 (patients=55, controls=41)
Analysis the effect of COL1A1 expression on MESO immune cell
infiltration
Kaplan-Meier survival analysis of MESO patients using GEPIA and
TIMER
Anotation of the mRNA data
Normalize mRNA data with ‘limma’package of R
Include 82 paired samples(patients=41, controls=41)
Analysis of infiltration of 22 immune cells in tumor tissue
and normal tissue
GSEA based on the expression of COL1A1
using LinkedOmics
Analysis the mutation of COL1A1 in MESO using
cBioPortal
Figure 1: The flowchart in this study.
2 BioMed Research International
Rela
tive p
erce
nt
0%
20%
40%
60%
80%
100%
B cells naiveB cells memoryPlasma cellsT cells CD8T cells CD4 naiveT cells CD4 memory restingT cells CD4 memory activatedT cells follicular helper
T cells regulatory (Tregs)T cells gamma deltaNK cells restingNK cells activatedMonocytesMacrophages M0Macrophages M1Macrophages M2
Dendritic cells restingDendritic cells activatedMast cells restingMast cells activatedEosinophilsNeutrophils
(a)
−1
−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8
1
Neu
trop
hils
Den
driti
c cel
ls ac
tivat
ed
Eosin
ophi
ls
T ce
lls C
D4
mem
ory
resti
ng
Mon
ocyt
es
T ce
lls fo
llicu
lar h
elpe
r
Mac
roph
ages
M1
T ce
lls g
amm
a del
ta
NK
cells
activ
ated
Mac
roph
ages
M0
Mas
t cel
ls ac
tivat
ed
B ce
lls n
aive
T ce
lls C
D4
naiv
e
B ce
lls m
emor
y
Plas
ma c
ells
T ce
lls re
gula
tory
(Tre
gs)
Mac
roph
ages
M2
T ce
lls C
D8
Den
driti
c cel
ls re
sting
Mas
t cel
ls re
sting
T ce
lls C
D4
mem
ory
activ
ated
NK
cells
resti
ng
Neutrophils
Dendritic cells activated
Eosinophils
T cells CD4 memory resting
Monocytes
T cells follicular helper
Macrophages M1
T cells gamma delta
NK cells activated
Macrophages M0
Mast cells activated
B cells naive
T cells CD4 naive
B cells memory
Plasma cells
T cells regulatory (Tregs)
Macrophages M2
T cells CD8
Dendritic cells resting
Mast cells resting
T cells CD4 memory activated
NK cells resting
1
0.24
0.32
0.18
0.25
–0.3
–0.23
–0.15
–0.26
–0.11
–0.08
0.02
−0.17
−0.11
−0.08
−0.12
0
−0.22
0.16
0.09
0.12
0.06
0.24
1
0.49
0.27
0.26
–0.39
–0.46
–0.34
–0.39
–0.24
0.05
0.03
−0.06
−0.08
−0.19
−0.11
0.07
0.09
0.09
0.08
0.04
0.04
0.32
0.49
1
0.49
0.36
–0.36
–0.26
–0.2
–0.3
–0.16
0.08
−0.05
0
−0.13
−0.1
−0.39
−0.3
−0.2
0.18
−0.08
0.14
0.21
0.18
0.27
0.49
1
0.42
–0.33
–0.23
–0.15
–0.08
–0.1
0
−0.05
−0.11
−0.36
0.05
−0.41
−0.08
−0.53
−0.05
0.21
−0.06
0.08
0.25
0.26
0.36
0.42
1
–0.39
–0.32
–0.29
–0.21
0.01
0.13
−0.11
0.01
−0.24
−0.05
−0.25
−0.36
−0.25
0.07
0.09
0.03
0.32
–0.3
–0.39
–0.36
–0.33
–0.39
1
0.42
0.34
0.35
–0.15
–0.06
0.28
0.22
0.21
0.05
−0.01
0.1
−0.03
−0.28
−0.17
−0.26
−0.26
–0.23
–0.46
–0.26
–0.23
–0.32
0.42
1
0.41
0.37
–0.11
–0.22
−0.04
−0.09
−0.12
−0.03
−0.07
−0.11
0.15
−0.02
−0.2
0.12
−0.13
–0.15
–0.34
–0.2
–0.15
–0.29
0.34
0.41
1
0.38
0.06
–0.06
0.21
0.15
−0.17
−0.12
−0.29
−0.18
−0.32
−0.42
−0.1
0.12
−0.35
–0.26
–0.39
–0.3
–0.08
–0.21
0.35
0.37
0.38
1
–0.07
0.02
−0.04
−0.13
−0.22
−0.07
−0.03
−0.04
−0.11
−0.18
−0.11
−0.16
−0.19
–0.11
–0.24
–0.16
–0.1
0.01
–0.15
–0.11
0.06
–0.07
1
–0.01
−0.11
−0.03
−0.11
−0.13
−0.02
−0.21
−0.25
−0.19
−0.08
−0.13
0.02
–0.08
0.05
0.08
0
0.13
–0.06
–0.22
–0.06
0.02
–0.01
1
0.09
0.11
−0.03
−0.12
−0.04
−0.26
−0.24
−0.21
−0.38
−0.16
−0.15
0.02
0.03
–0.05
–0.05
–0.11
0.28
–0.04
0.21
–0.04
–0.11
0.09
1
0.46
−0.11
−0.23
−0.21
−0.11
−0.19
−0.11
−0.11
−0.21
−0.19
–0.17
–0.06
0
–0.11
0.01
0.22
–0.09
0.15
–0.13
–0.03
0.11
0.46
1
0.17
−0.17
−0.08
−0.28
−0.28
−0.12
−0.01
0.05
0.27
–0.11
–0.08
–0.13
v0.36
–0.24
0.21
–0.12
–0.17
–0.22
–0.11
–0.03
–0.11
0.17
1
0.3
0.21
0.09
0.23
−0.06
0.06
0.11
0.04
–0.08
–0.19
–0.1
0.05
–0.05
0.05
–0.03
–0.12
–0.07
–0.13
–0.12
–0.23
−0.17
0.3
1
0.2
0.1
0.11
−0.19
0.29
−0.09
0.03
–0.12
–0.11
–0.39
–0.41
–0.25
–0.01
–0.07
–0.29
–0.03
–0.02
–0.04
–0.21
−0.08
0.21
0.2
1
0.28
0.39
0.11
0.01
−0.13
0.03
0
0.07
–0.3
–0.08
–0.36
0.1
–0.11
–0.18
–0.04
–0.21
–0.26
−0.11
−0.28
0.09
0.1
0.28
1
0.19
−0.07
0.07
−0.21
−0.24
–0.22
0.09
–0.2
–0.53
–0.25
–0.03
0.15
–0.32
–0.11
–0.25
–0.24
−0.19
−0.28
0.23
0.11
0.39
0.19
1
0.43
0.07
0.06
−0.05
0.16
0.09
0.18
–0.05
0.07
–0.28
–0.02
–0.42
–0.18
–0.19
–0.21
−0.11
−0.12
−0.06
−0.19
0.11
−0.07
0.43
1
0.07
0.1
0.17
0.09
0.08
–0.08
0.21
0.09
–0.17
–0.2
–0.1
–0.11
–0.08
–0.38
–0.11
−0.01
0.06
0.29
0.01
0.07
0.07
0.07
1
0.07
0.07
0.12
0.04
0.14
–0.06
0.03
v0.26
0.12
0.12
–0.16
–0.13
–0.16
–0.21
0.05
0.11
−0.09
−0.13
−0.21
0.06
0.1
0.07
1
0.32
0.06
0.04
0.21
0.08
0.32
–0.26
–0.13
–0.35
–0.19
0.02
–0.15
–0.19
0.27
0.04
0.03
0.03
−0.24
−0.05
0.17
0.07
0.32
1
(b)
Figure 2: Landscape of immune cell infiltration inMESO and normal tissues by CIBERSORT. (a) The composition of 22 kinds of immune cellsin MESO and normal tissues. (b) Heat map visualizing the correlation between 22 kinds of immune cells across MESO and normal samples.
3BioMed Research International
Benjamini and Hochberg’s false discovery rates wereapplied to adjust p values [23]. The screening criteria ofDEGs were as follows: ∣log fold change ðFCÞ ∣ >1 andadjusted p value < 0.05. The common DEGs were over-lapped among the three datasets (GSE51024, GSE42977,and GSE2549).
2.3. Cell Type Identification by Estimating Relative Subsets ofRNA Transcripts (CIBERSORT). CIBERSORT can accuratelyquantify the percentage of various tumor-infiltrating
immune cells (TIICs) under the complex “gene signaturematrix” based on 547 genes [24]. In the current study, theimmune infiltration of 22 kinds of immune cells for eachsample was assessed by the LM22 signature file, with the pre-set signature matrix at 100 permutations. The distribution of22 subtypes of TIICs was then presented, followed by calcu-lation of correlation coefficient, p value, and root meansquared error (RMSE). The p value < 0.05 represents a statis-tical connotation of deconvolution outcomes for all cell sub-sets for each sample and has been useful for less precise
Macrophages M2Macrophages M0NK cells activatedT cells follicular helperMacrophages M1B cells naiveT cells CD4 naiveNK cells restingB cells memoryT cells regulatory (Tregs)Plasma cellsT cells CD4 memory activatedMast cells restingMonocytesDendritic cells activatedEosinophilsDendritic cells restingNeutrophilsT cells CD8T cells gamma deltaT cells CD4 memory restingMast cells activated
Type
TypeConTreat
0
0.1
0.2
0.3
0.4
0.5
(a)
NormalTumor
PCA1
PCA
2
GroupNormalTumor
–4
−2
0
2
4
6
–2.5 0.0 2.5
(b)
Figure 3: The differences in immune cell infiltration between MESO and normal tissues. (a) Heatmap showing the difference in infiltrationlevels of 22 kinds of immune cells between MESO and normal samples. (b) Principal component analysis depicting the distinction betweenMESO and normal tissues based on these immune cells. The blue dots indicate normal samples, and the red dots indicate tumor samples.
4 BioMed Research International
exclusion of outcomes. Finally, 41 MESO samples and 41paired control samples which met the required p value <0.05 were selected for further analysis.
2.4. Protein-Protein Interaction PPI Networks and HubGenes. A PPI network was constructed based on DEGs using
the STRING (version 11.0; http://string-db.org) database.STRING is an online database used to predict interactionsbetween proteins [25], which is essential for recognizing themechanisms of cell activities at the molecular levels in cancerprogression. The cut-off value was defined as an interactionscore (median confidence) of 0.4. The PPI network was
0.0
0.1
0.2
0.3
0.4
0.5
0.6
Frac
tion
p=0.721p=0.113
p=1
p=0.171
p=1
p<0.001
p=0.005
p<0.001 p<0.001
p=0.017
p=0.001p=0.009
p<0.001
p=0.29
p=0.011
p=0.001
p=0.002
p<0.001p=0.282
p=0.933
p<0.001
p=0.001
B ce
lls n
aive
B ce
lls m
emor
y
Plas
ma c
ells
T ce
lls C
D8
T ce
lls C
D4
naiv
e
T ce
lls C
D4
mem
ory
resti
ng
T ce
lls C
D4
mem
ory
activ
ated
T ce
lls fo
llicu
lar h
elpe
r
T ce
lls re
gula
tory
(Tre
gs)
T ce
lls g
amm
a del
ta
NK
cells
resti
ng
NK
cells
activ
ated
Mon
ocyt
es
Mac
roph
ages
M0
Mac
roph
ages
M1
Mac
roph
ages
M2
Den
driti
c cel
ls re
sting
Den
driti
c cel
ls ac
tivat
ed
Mas
t cel
ls re
sting
Mas
t cel
ls ac
tivat
ed
Eosin
ophi
ls
Neu
trop
hils
(a)
0.20
0.15
0.10
0.10
0.00Normal Tumor
Macrophages M1p value = 0.011
(b)
0.4
0.3
0.2
0.1
0.0
p value = 0.001Macrophages M2
Normal Tumor
(c)
0.15
0.10
0.05
0.00
T cells follicular helper
p value = 0.039e-07
Normal Tumor
(d)
0.35
0.300.25
0.200.15
0.10
0.05
0.10
T cells gamma delta
p value = 0.017
Normal Tumor
(e)
0.12
0.10
0.08
0.06
0.04
0.02
0.00
Dendritic cells activatedp value = 1.462e-07
Normal Tumor
(f)
0.08
0.06
0.04
0.02
0.00
Eosinophilsp value = 6.93e-12
Normal Tumor
(g)
0.20
0.15
0.10
0.05
0.00
Monocytesp value = 6.88e-10
Normal Tumor
(h)
0.20
0.15
0.10
0.05
0.00
Neutrophilsp value = 6.433e-04
Normal Tumor
(i)
Figure 4: The differences in immune cell infiltration between MESO and paired normal tissues. (a) Violin plots visualizing the distributionsof TIICs between MESO and paired normal tissues. Blue represents normal samples and red indicates MESO samples. The infiltration levelsof (b) M1 macrophages, (c) M2 macrophages, (d) T cell follicular helper, (e) T cell gamma delta, (f) activated dendritic cells, (g) eosinophils,(h) monocytes, and (i) neutrophils were different between the two groups.
5BioMed Research International
visualized by the Cytoscape software (version 3.7.2; http://www.cytoscape.org/). Hub genes were ranked by the cyto-Hubba plug-in [26].
2.5. Immune Cell Infiltration. The mutations of COL1A1across 87 MESO samples were assessed in the cBioPortal
database (https://www.cbioportal.org/) [27]. In addition, thecorrelation between COL1A1 expression and the abundanceof 6 types of infiltrating immune cells (B cells, CD4+ T cells,CD8+ T cells, neutrophils, macrophages, and dendritic cells)was calculated among MESO samples via The TumorImmune Estimation Resource (TIMER) algorithm database
965213
1961
222
118
138
1066
GSE42977 GSE2549
GSE51024
(a)
KNTC1
HMMR
CENPF
KIAA0101
ASPMSEMA3G
CDH2
KIF20A
MCM2
CKS2
CDK6
ANGPT1
EPHB2
MCM4
ADAM10
SPARCL1GPC3
CLDN5
FAM107AHLF
TSPAN7
TGFBR3
GPM6A
PGM5ROR1
RAI2
GADD45B
SOCS2LDB2
TOP2A
KIF23
SKP2FEN1
CEP55
BUB1
MYH11
LPL
SPP1
FCN1RUNX2
BACE2
HBB S100A8 PDGFRL
ALDH1A1
LMO3
PPARGPTGER3
CD36
CAT
CFD
STX11
ADH1B
FABP4PDK4
LAMA1
CALD1
ANG
COL5A1
PLOD2
PECAM1
CDH5
SHMT2
LOXL1
COL1A1
THBS3
TGFBINMU
COL6A1
WISP1PDE3B
MAOA
AOC3CDO1 ACADL
GNG11GPX3TSPAN8
EDNRB
ADRB2
EFEMP1
JAM2
SORBS1
SERPINH1
EMCNC6
COL11A1
UCHL1
COL3A1
GART
CDH11
MFAP2
COL10A1
CKAP4
SFRP1
FHL2
ISYNA1
(b)
COL5A1COL1A1
COL10A1COL11A1
PLOD2SERPINH1COL3A1
COL6A1
(c)
Figure 5: Identification of DEGs and hub genes for MESO. (a) Venn diagram depicting common DEGs among 3 different GEO datasets withthe threshold of ∣logFC > 1 ∣ and adjusted p value < 0.05. (b) A PPI network based on DEGs using Cytoscape. (c) A key module based on thePPI network by the MCODE plugin. Light purple suggests upregulated genes, and light green suggests downregulated genes.
Table 1: The ranking of hub genes according to different methods using cytoHubba.
MNC Degree EPC Bottleneck Closeness Radiality Betweenness Stress
COL1A1 COL1A1 SPP1 SPP1 SPP1 SPP1 PPARG SPP1
MCM4 PPARG COL1A1 PPARG COL1A1 PPARG SPP1 COL1A1
MCM2 MCM4 RUNX2 CDH2 PPARG COL1A1 COL1A1 PPARG
BUB1 MCM2 MCM4 COL1A1 RUNX2 RUNX2 PECAM1 CDH2
ASPM SPP1 COL3A1 RUNX2 PECAM1 PECAM1 RUNX2 RUNX2
Abbreviation: MCC: maximal clique centrality; MNC: maximum neighborhood component; Degree: node connects degree; EPC: edge percolated component.
6 BioMed Research International
(https://cistrome.shinyapps.io/timer/) [28]. TIMER is a pow-erful online tool that can analyze the infiltration of immunecells in different tumors [28].
2.6. Gene Set Enrichment Analysis (GSEA). The LinkedOmicsdatabase (https://www.linkedomics.org/) contains multio-mics and clinical data across 32 cancer types and 11,158patients from The Cancer Genome Atlas (TCGA) project[29]. Furthermore, the database has numerous collated dataavailable for download and has a very powerful online analy-sis function. In this study, GSEA was presented to study thedifferences in the high and low expression of COL1A1 groupsfor MESO based on this powerful database, including GeneOntology (GO) and Kyoto Encyclopedia of Genes andGenomes (KEGG).
2.7. Survival Analysis. MESO patients were divided into thehigh- and low-expression groups based on the median valueof the COL1A1 expression. The differences in overall survivaland recurrence-free survival were analyzed between the twogroups and Kaplan–Meier curves were depicted via the
online tool Gene Expression Profiling Interactive Analysis(GEPIA) [30]. The impact of 6 types of immune cell infiltra-tion on the overall survival of MESO patients was also ana-lyzed using the TIMER.
2.8. Statistical Analysis. All statistical analysis was carried outusing R language (version 3.6.2) packages including “limma,”“CIBERSORT,” “pheatmap,” “corrplot,” “vioplot,” and“ggplot2.” p value < 0.05 was considered statisticallysignificant.
3. Results
3.1. The Distribution of TIICs in MESO and Matched NormalTissues.We investigated the differences in 22 subpopulationsof TIICs between MESO tissues and normal tissues using theCIBERSORT algorithm. Finally, 41 MESO tissues and thepaired control tissues met the screening criteria (p < 0:05).Figure 2(a) illustrated the distribution of 22 kinds of TIICsin 41 MESO tissues and control tissues. As shown inFigure 2(b), the fractions of different immune cells were
pRCC
(TCG
A P
anCA
n)
Chro
mop
hobe
RCC
(TCG
A P
anCa
n)
ccRC
C (T
CGA
Pan
Can)
Uve
al m
elan
oma (
TCG
A P
anCa
n 20
18)
Ute
rine C
S (T
CGA
Pan
Can
2018
)
Ute
rine (
TCG
A P
anCa
n 20
18)
�yr
oid
(TCG
A P
anCa
n 20
18)
�ym
oma (
TCG
A P
anCa
n 20
18)
Test
icul
ar g
erm
cell
(TCG
A P
anCa
n 20
18)
Stom
ach
(TCG
A P
anCa
n 20
18)
Pros
tate
(TCG
A P
anCa
n 20
18)
Panc
reas
(TCG
A P
anCa
n 20
18)
PCPG
(TCG
A P
anCa
n 20
18)
Ova
rian
(TCG
A P
anCa
n 20
18)
Mes
othe
liom
a (TC
GA
Pan
Can
2018
)
Mela
nom
a (TC
GA
Pan
Can
2018
)
Lung
squ
(TCG
A P
anCa
n)
Lung
aden
o (T
CGA
Pan
Can)
Live
r (TC
GA
Pan
Can)
Hea
d &
nec
k (T
CGA
Pan
Can)
GBM
(TCG
A P
anCa
n)
DLB
C (T
CGA
Pan
Can)
Col
orec
tal (
TCG
A P
anCa
n)
Chol
angi
ocar
cino
ma (
TCG
A P
anCa
n)
Cer
viva
l (TC
GA
Pan
Can)
AM
L (T
CGA
Pan
Can)
ACC
(TCG
A P
anCa
n)
COL1
A1
expr
essio
n - R
NA
seq
v2 (l
og2)
25
20
15
10
5
–5
–10
Fusion Truncating (VUS) Missense (VUS) Not mutatedNot profiled for mutations Amplification Gain DiploidShallow deletion Deep deletion Not profiled for CNA
0
(a)
Profiled in protein expression Z-scores (RPPA)
Profiled in protein expressionZ-scores (RPPA)
Yes No
COL1A1
Genetic alteraions Amplification mRNA high No alterations
9%
(b)
Figure 6: The COL1A1 expression and mutation in MESO tissues from the cbioportal database. (a) The expression of COL1A1 across 27kinds of cancers. (b) COL1A1 genetic alteration in MESO using reverse-phase protein arrays (RPPA).
7BioMed Research International
weakly to moderately correlated in MESO tissues. Amongthem, there was the strongest positive correlation in infiltra-tion mode between eosinophils and activated dendritic cells(coef = 0:49), alongside with correlation between CD4 mem-ory resting T cells and eosinophils (coef = 0:49). Further-more, there was the strongest negative correlation ininfiltration mode between CD8+ T cells and CD4 memoryresting T cells (coef = −0:53). On the whole, some immunecells were relatively abundant in tumor tissues, and somewere relatively abundant in normal tissues (Figure 3(a)).For example, the infiltration levels of monocytes, activateddendritic cells, and eosinophils in normal tissues were obvi-
ously higher than those in MESO tissues. We further probedinto whether the two tissue types could be differentiated bythese 22 immune cells. Based on the differences in infiltrationof 22 types of immune cells, tumor tissues were clearly distin-guished from normal tissues by principal component analysis(Figure 3(b)).
3.2. Infiltration Levels of Immune Cells Are Distinct in Normaland MESO Tissues. We further determined the difference inimmune cell infiltration between MESO and normal tissues.The average proportion of each immune cell type in MESOtissues and paired normal tissues was evaluated, as shown
MESO⁎⁎
⁎ ⁎
Copy numberArm-level deletion
High amplication
Infil
trat
ion
leve
l 0.6
0.4
0.2
0.0
B ce
ll
CD8+
T ce
ll
CD4+
T ce
ll
Mac
roph
age
Neu
troph
il
Den
driti
c cel
l
Arm-level gainDiploid/normal
⁎⁎
⁎ ⁎
(a)
0.25
6
9
12
15
0.50 0.75 1.00 0.1 0.2 0.3 0.0 0.2 0.4 0.1 0.2Infiltration level
COL1
A1
expr
essio
n le
vel (
log2
TPM
)
0.3 0.00 0.05 0.10 0.15 0.100 0.125 0.150 0.175 0.3 0.4 0.5 0.6 0.7
cor = –0.257p = 1.67e-02
partial.cor = 0.028p = 8.03e-01
partial.cor = –0.07p = 5.26e-01
partial.cor = 0.221p = 4.38e-02
partial.cor = –0.414p = 8.86e-05
partial.cor = –0.266p = 1.45e-02
partial.cor = 0.095p = 3.91e-01
MES
O
Purity B cell CD8+ T cell CD4+ T cell Machrophage Neutrophil Dendritic cell
(b)
–2.5 –2.0 –1.5 –1.0 –0.5 0.0Normalized enrichment score
Biological process
Extracellular structure organizationOssification
AngiogenesisMitochondrial transportAntiobiotic metabolic processDefense response to other organismRibonucleoprotein complex subunit organizationTranslational initiationncRNA processingMitochondrial gene expression
0.5 1.0 1.5 2.0 2.5
(c)
RibosomeMitochondrial inner membraneMitochondrial matrixSpliceosomal complex
Actin cytoskeletonChromosomal region
MicrotubuleCell leading edge
Cell-substrate junctionExtracellular matrix
Cellular component
–2.5 –2.0 –1.5 –1.0 –0.5 0.0Normalized enrichment score
0.5 1.0 1.5 2.0 2.5
(d)
FDR≤0.05FDR>0.05
Molecular function
Extracellular matrix structural constituentGrowth factor binding
Metallopeptidase activityCell adhesion molecular binding
Calmodulin bindingTubulin binding
Catalytic activity, acting on RNACarboxylic ester hydrolase activityOxidoreductase activity, acting on CH-OH group of donorsStructural constituent of ribosome
–2.5 –2.0–3.0 –1.5 –1.0 –0.5 0.0Normalized enrichment score
0.5 1.0 1.5 2.0 2.5
(e)
Herpes simplex infection
MicroRNAs in RNAPI3K-Akt signaling pathway
Regulation of actin cytoskeletonCell adhesion molecules (CAMs)
Parathyroid hormoe systhesis, secretion and action
Glycolysis/gluconeogenesisRibosome biogenesis in eukaryotesOxidative phosphorylationRibosome
KEGG
–2.5 –2.0–3.0 –1.5 –1.0 –0.5 0.0Normalized enrichment score
0.5 1.0 1.5 2.0
(f)
Figure 7: COL1A1 expression is associated with immune cell infiltration in MESO. (a) The correlation between somatic copy number alterationsand immune cell infiltration. (b) The correlation between COL1A1 expression and immune cell infiltration in MESO samples. Functionalenrichment analysis of COL1A1 including (c) GO biological process, (d) GO cellular components, (e) GO molecular function, and (f) KEGG.
8 BioMed Research International
in Figure 4(a). Our data revealed that the proportions of Tregcells (p < 0:001; Figure 4(a)), M1 macrophages (p = 0:011;Figure 4(b)), M2 macrophages (p = 0:001; Figure 4(c)), T cellfollicular helper (p < 0:001; Figure 4(d)), and T cell gammadelta (p = 0:017; Figure 4(e)) were all distinctly higher inMESO tissues compared to normal tissues. Additionally, weobserved that the infiltration levels of resting CD4+ memoryresting T cells (p < 0:001; Figure 4(a)), CD4+ memory acti-vated T cells (p = 0:005; Figure 4(a)), resting dendritic cells(p = 0:002; Figure 4(a)), activated dendritic cells (p < 0:001;Figure 4(f)), eosinophils (p < 0:001; Figure 4(g)), monocytes(p < 0:001; Figure 4(h)), and neutrophils (p < 0:001;Figure 4(i)) in normal tissues were all significantly higherthan MESO tissues.
3.3. Identification of DEGs and Hub Genes for MESO.GEO2Rwas used to identify DEGs between MESO and normal tis-sues. With the threshold of ∣logFC ∣ >1 and adjusted p value< 0.05, DEGs (1,983 in GSE51024, 1,470 in GSE42977, and2,851 in GSE2549) were identified for MESO tissues com-pared to normal tissues. The overlap among the three data-sets contained 118 DEGs, as shown in the Venn diagram(Figure 5(a)), composed of 66 downregulated genes and 52upregulated genes in MESO tissues compared to normal tis-sues. Then, these DEGs were analyzed in the STRING data-base. A PPI network was constructed by the Cytoscapesoftware, including 97 nodes and 299 edges (Figure 5(b)).The degree of each node was calculated. Among them,COL1A1 had the highest degree according to the eightranked methods using cytoHubba (Table 1). The pluginMCODE of the Cytoscape was used to establish the key mod-ule based on the PPI network (Figure 5(c)). In this module,COL1A1 was at the core position and was most connectedto other genes.
3.4. Expression and Mutation of COL1A1 in MESO. In thecBioPortal database, we analyzed the expression of COL1A1across 27 kinds of cancers. As shown in Figure 6(a), COL1A1exhibited a higher expression in MESO than normal tissues(other tumors). Subsequently, 87 MESO samples (TCGA,
Firehose Legacy) were selected to analyze the COL1A1genetic alteration. It was found that the COL1A1 geneticalteration occurred 9% across 87 MESO patients(Figure 6(b)). Among them, amplification was the most com-mon type of mutation. Furthermore, using the reverse-phaseprotein arrays (RPPA), we found that COL1A1 protein wasfrequently expressed in MESO tissues.
3.5. COL1A1 Expression Is Associated with Immune CellInfiltration in MESO.Using the TIMER, we analyzed the cor-relation between different somatic copy number alterationsand immune cell infiltration in MESO samples. As shownin Figure 7(a), our data indicated that somatic copy numberalterations were significantly correlated to the infiltration ofCD4+ T cells (p < 0:01), neutrophils (p < 0:05), and dendriticcells (p < 0:05). Samples with diploid/normal exhibited thehighest infiltration levels of CD4+ T cells and dendritic cells.The correlation between COL1A1 expression and immunecell infiltration was analyzed across patients with MESO(Figure 7(b)). COL1A1 expression was distinctly correlatedto tumor purity (cor = −0:257; p = 1:67e − 02), CD4+ T cells(cor = −0:221; p = 4:38e − 02), macrophages (cor = 0:414; p= 8:86e − 05), and neutrophils (cor = −0:266; p = 1:45e − 02). GSEA including GO and KEGG was performed forCOL1A1. Our results showed that COL1A1 was significantlyrelated with key biological processes such as extracellularstructure organization, ossification, and angiogenesis(Figure 7(c)). COL1A1 was correlated with several criticalcellular components like extracellular matrix and cell-substrate junction (Figure 7(d)). COL1A1 could be involvedin regulating molecular functions of extracellular matrixstructural constituent, growth factor binding, and metallo-peptidase activity (Figure 7(e)). As for KEGG pathways,COL1A1 was associated with microRNAs in cancer, PI3K-Akt signaling pathway, and regulation of actin cytoskeleton(Figure 7(f)).
3.6. COL1A1 Could Be a Potential Prognostic Marker forMESO Patients. Cox proportional hazard model was con-structed to assess the prognostic value of COL1A1 expressionin the survival of MESO patients using the TIMER, as listedin Table 2. Among other clinicopathological factors,COL1A1 had a significant association with prognosis ofMESO patients (p<0.0001). To further explore the associa-tion between COL1A1 expression and patients’ survival,MESO patients were divided into the high- and low-expression groups based on the median value of COL1A1expression using the online tool GEPIA. The data showedthat there was no significant difference in disease-free sur-vival between the high- and low-expression groups ofCOL1A1 for MESO patients (Figure 8(a)). However, patientswith high COL1A1 expression indicated poorer overall sur-vival time compared to those with its low expression(p = 0:0014; Figure 8(b)). We also evaluated the correlationbetween immune cell infiltration and MESO patients’ prog-nosis. As shown in Figure 8(c), high levels of neutrophil infil-tration (p = 0:001) and low levels of macrophage infiltration(p = 0:047) distinctly predicted better clinical outcomes.
Table 2: Cox proportional hazard model of MESO among 84patients.
Parameter Coef95%CI_lower
95%CI_upper
HR p value
Age· 0.029 0.998 1.062 1.029 0.07
Gender(male)
-0.25 0.415 1.464 0.779 0.438
Race (Black) 0.705 0.101 40.397 2.024 0.644
Race (White) -0.116 0.114 6.935 0.89 0.912
Stage I -0.151 0.345 2.14 0.86 0.745
Stage II -0.142 0.385 1.954 0.868 0.732
Stage III -0.483 0.24 1.588 0.617 0.317
Purity -0.297 0.248 2.223 0.743 0.595
COL1A1 0.28 1.134 1.544 1.323 <0.0001Abbreviation: coef: coefficient; CI: confidence interval; HR: hazard ratio.
9BioMed Research International
4. Discussion
Based on the MESO dataset from the GEO database, we foundthe differences in immune cell infiltration between MESO tis-sues and matched normal pleural tissues. By combining twoother GEO datasets, we identified 118 DEGs for MESO. Ahub gene, COL1A1, was identified based on the PPI network.Through the cBioPortal database, COL1A1 had a high geneticalteration (9%) across MESO samples, and amplification wasthe most common type of mutation. COL1A1 expressionwas markedly associated with MESO patients’ clinical out-comes, which could become a potential prognostic marker.
High expression of COL1A1 has been shown to be closelylinked to the progression of various cancers [31–34]. Consis-tently, in this study, the expression of COL1A1 in tumor tis-sues was distinctly higher compared with normal tissues.Furthermore, the results showed that the mutation ofCOL1A1 had a distinct correlation with the infiltration ofneutrophils, CD4+ T cells, and dendritic cells. Meanwhile,COL1A1 expression exhibited a significant association withtumor purity, CD4 T cells, macrophages, and neutrophils.Thus, COL1A1 expression might be related to tumorimmune microenvironment.
In GSEA results, we found that extracellular matrix(ECM) and cell-substrate junction were significantlyenriched in the high COL1A1 expression group, indicatingthat COL1A1 might affect the local infiltration of immunecells by affecting ECM, which was consistent with the results
of Homan et al. [35]. For KEGG pathway enrichment analysisresults, microRNAs in cancer, PI3K-Akt signaling pathway,and regulation of actin cytoskeleton were related to highCOL1A1 expression. A previous study has shown that thePI3K-Akt pathway is frequently activated inMPM [36]. Theseresults indicated that high expression of COL1A1 could pro-mote the progression of MESO. Cox proportional hazardmodel analysis results demonstrated that COL1A1 expressionwas in relationship with MESO patients’ prognosis. HighCOL1A1 expression usually predicted a poorer prognosis forMESO patients. Thus, COL1A1 expression could be a promis-ing immune-related prognostic marker for MESO.
Due to the neutrophils’ short life span, it is difficult to iso-late and study neutrophils associated with tumors, especiallyhuman tumor-associated neutrophils (TANs). Up to now, itremains unclear whether neutrophils could promote tumorgrowth [37, 38] or not [39–41]. In this study, we found thatfor patients with MESO, high levels of neutrophil infiltrationpredicted longer overall survival time. Moreover, the expres-sion of COL1A1 was negatively correlated with neutrophilinfiltration. Hence, it is of importance to further explore theassociation between neutrophil infiltration and MESOpatients’ prognosis. Consistent with previous studies, wefound that high levels of macrophage infiltration predicteda worse clinical outcome for MESO patients [42–44]. More-over, patients with high expression of COL1A1 indicated ahigher level of macrophage infiltration. Taken together, thesefindings indicated that COL1A1 might affect the survival of
0 20 40 60 800.0
0.2
0.4
0.6
0.8
1.0Disease free survival
Months
Perc
ent s
urvi
val
Logrank p=0.16HR(high)=1.5p(HR)=0.16n(high)=41n(low)=41
(a)
0 20 40 60 800.0
0.2
0.4
0.6
0.8
1.0Overall survival
Months
Perc
ent s
urvi
val
Low COL1A1 TPMHigh COL1A1 TPM
Logrank p=0.0014HR(high)=2.2p(HR)=0.0016
n(high)=41n(low)=41
(b)
Low (bottom 25%)High (top 25%)
B cellLog-rank p = 0.176 Log-rank p = 0.07 Log-rank p = 0.105 Log-rank p = 0.047 Log-rank p = 0.001 Log-rank p = 0.097
CD8+ T cell CD4+ T cell Macrophage Neutrophil Dendritic cell
LevelTime to follow-up (months)
0
1.00
0.75
0.50
0.25
0.00
Cum
ulat
ive s
urvi
val
20 40 60 80 0 20 40 60 80 0 20 40 60 80 0 20 40 60 80 0 25 50 75 0 20 40 60
MES
O
Log-rank agLL k p = 0.17660 17 Log-rank kL k p = 0.07700 0 Log-rank gLL k p = 0.10550 10 Log-rank n-LL k pp = 0.04700 04 Log-rank n-LL pp = 0.001000 00 Log-rank kroL k p = 0.09700 09
(c)
Figure 8: COL1A1 expression is associated with MESO patients’ prognosis. (a) Disease-free survival and (b) overall survival analyses ofCOL1A1 expression for MESO patients using the GEPIA online platform. (c) The correlation between the infiltrations of 6 immune cellsand MESO patients’ prognosis.
10 BioMed Research International
patients partly by mediating the local infiltration of these twoimmune cells, which was consistent with the results fromGEPIA. In the GEO database, the data indicated that therewas a significant difference in immune cell infiltrationbetween MESO and normal tissues, which made our resultsmore reliable. The concept of immunophenotypic heteroge-neity of MESO is worthy of further exploration to find newtreatment strategies to improve the clinical outcomes ofpatients with this disease.
Our study has several limitations that should be acknowl-edged. Firstly, our study only analyzed the infiltration ofimmune cells between normal and MESO tissues, withoutpaying attention to the differences between subtypes. Thus,immune cell infiltration based on histological subtypes ofMESO needs further exploration. Secondly, our conclusionswere based on bioinformatics methods. Therefore, furthermolecular biological experiments should be presented toconfirm the biological functions of COL1A1 in MESO. Lastbut not least, in public databases, datasets on MESO are rel-atively scarce. More relevant clinical samples should be col-lected for MESO research. The research needs to be verifiedin a larger cohort of MESO.
5. Conclusion
With the help of bioinformatics analysis, we found that therewas a distinct difference in the infiltration of immune cellsbetween MESO and normal tissues. The differences inimmune infiltration were significantly associated with MESOpatients’ prognosis. A PPI network was constructed forMESO based on DEGs. A hub gene, COL1A1, was identified,which was highly expressed in MESO tissues than normal tis-sues. COL1A1 expression was in association with theimmune cell infiltration across MESO samples. Patients withhigh COL1A1 expression usually indicated a poorer progno-sis than those with its low expression. Thus, COL1A1 couldbecome a potential immune-related prognostic marker forMESO, which requires to be validated in a larger MESOcohort.
Abbreviations
MESO: MesotheliomaDEGs: Differentially expressed genesPPI: Protein-protein interactionGSEA: Gene Set Enrichment AnalysisMPM: Malignant pleural mesotheliomaTILs: Tumor-infiltrating lymphocytesGEO: Gene expression omnibusTCGA: The Cancer Genome AtlasFC: Fold changeCIBERSORT: Cell type identification by estimating relative
subsets of RNA transcriptsPPI: Protein-protein interaction.
Data Availability
The data used to support the findings of this study areincluded within the supplementary information files.
Conflicts of Interest
The authors declare no conflicts of interest.
Acknowledgments
This study was supported by the Natural Science Foundationof Shaanxi Province of China (2020JM-065).
Supplementary Materials
Supplementary Table 1: the clinical information of MESOpatients in the GSE51024 dataset. (Supplementary Materials)
References
[1] A. S. Tsao, I. Wistuba, J. A. Roth, and H. L. Kindler, “Malig-nant pleural mesothelioma,” Journal of Clinical Oncology,vol. 27, no. 12, pp. 2081–2090, 2009.
[2] R. A. Lemen, “Mesothelioma from asbestos exposures: epide-miologic patterns and impact in the United States,” Journalof Toxicology and Environmental Health. Part B, CriticalReviews, vol. 19, no. 5-6, pp. 250–265, 2016.
[3] P. Mirarabshahii, K. Pillai, T. C. Chua, M. H. Pourgholami,and D. L. Morris, “Diffuse malignant peritoneal mesothelioma- an update on treatment,” Cancer Treatment Reviews, vol. 38,no. 6, pp. 605–612, 2012.
[4] G. Carteni, C. Manegold, G. M. Garcia et al., “Malignant peri-toneal mesothelioma–results from the international expandedaccess Program using pemetrexed alone or in combinationwith a platinum agent,” Lung Cancer, vol. 64, no. 2, pp. 211–218, 2009.
[5] D. W. Henderson, G. Reid, S. C. Kao, N. van Zandwijk, andS. Klebe, “Challenges and controversies in the diagnosis ofmalignant mesothelioma: part 2. Malignant mesotheliomasubtypes, pleural synovial sarcoma, molecular and prognosticaspects of mesothelioma, BAP1, aquaporin-1 and microRNA,”Journal of Clinical Pathology, vol. 66, no. 10, pp. 854–861,2013.
[6] F. Galateau-Salle, A. Churg, V. Roggli, W. D. Travis, andWorld Health Organization Committee for Tumors of thePleura, “The 2015 World Health Organization classificationof tumors of the pleura: advances since the 2004 classification,”Journal of Thoracic Oncology, vol. 11, no. 2, pp. 142–154, 2016.
[7] J. M. Mazurek, G. Syamlal, J. M. Wood, S. A. Hendricks, andA. Weston, “Malignant mesothelioma mortality - UnitedStates, 1999-2015,” MMWR. Morbidity and Mortality WeeklyReport, vol. 66, no. 8, pp. 214–218, 2017.
[8] R. R. Meyerhoff, C.-F. J. Yang, P. J. Speicher et al., “Impact ofmesothelioma histologic subtype on outcomes in the Surveil-lance, Epidemiology, and End Results database,” The Journalof Surgical Research, vol. 196, no. 1, pp. 23–32, 2015.
[9] A.W. Musk, N. Olsen, H. Alfonso et al., “Predicting survival inmalignant mesothelioma,” The European Respiratory Journal,vol. 38, no. 6, pp. 1420–1424, 2011.
[10] B. Tolani, L. A. Acevedo, N. T. Hoang, and B. He, “Heteroge-neous contributing factors in MPM disease development andprogression: biological advances and clinical implications,”International Journal of Molecular Sciences, vol. 19, 2018.
[11] N. J. Vogelzang, J. J. Rusthoven, J. Symanowski et al., “Phase IIIstudy of pemetrexed in combination with cisplatin versus cis-platin alone in patients with malignant pleural mesothelioma,”
11BioMed Research International
Journal of Clinical Oncology, vol. 21, no. 14, pp. 2636–2644,2003.
[12] D. S. Chen and I. Mellman, “Elements of cancer immunity andthe cancer-immune set point,” Nature, vol. 541, no. 7637,pp. 321–330, 2017.
[13] J. A. Joyce and J. W. Pollard, “Microenvironmental regulationof metastasis,” Nature Reviews. Cancer, vol. 9, no. 4, pp. 239–252, 2009.
[14] S. E. Stanton and M. L. Disis, “Clinical significance of tumor-infiltrating lymphocytes in breast cancer,” Journal for Immu-notherapy of Cancer, vol. 4, no. 1, p. 59, 2016.
[15] R. M. Bremnes, L.-T. Busund, T. L. Kilvær et al., “The role oftumor-infiltrating lymphocytes in development, progression,and prognosis of non-small cell lung cancer,” Journal of Tho-racic Oncology, vol. 11, no. 6, pp. 789–800, 2016.
[16] M. Morris, C. Platell, and B. Iacopetta, “Tumor-infiltratinglymphocytes and perforation in colon cancer predict positiveresponse to 5-fluorouracil chemotherapy,” Clinical CancerResearch, vol. 14, no. 5, pp. 1413–1417, 2008.
[17] C. Truntzer, N. Isambert, L. Arnould, S. Ladoire, andF. Ghiringhelli, “Prognostic value of transcriptomic determi-nation of tumour-infiltrating lymphocytes in localised breastcancer,” European Journal of Cancer, vol. 120, pp. 97–106,2019.
[18] A. W. Zhang, A. McPherson, K. Milne et al., “Interfaces ofmalignant and immunologic clonal dynamics in ovariancancer,” Cell, vol. 173, no. 7, pp. 1755–1769.e22, 2018,e1722.
[19] R. Edgar, M. Domrachev, and A. E. Lash, “Gene ExpressionOmnibus: NCBI gene expression and hybridization array datarepository,”Nucleic Acids Research, vol. 30, no. 1, pp. 207–210,2002.
[20] M. B. Suraokar, M. I. Nunez, L. Diao et al., “Expression profil-ing stratifies mesothelioma tumors and signifies deregulationof spindle checkpoint pathway and microtubule network withtherapeutic implications,” Annals of Oncology, vol. 25, no. 6,pp. 1184–1192, 2014.
[21] A. De Rienzo, W. G. Richards, B. Y. Yeap et al., “Sequentialbinary gene ratio tests define a novel molecular diagnosticstrategy for malignant pleural mesothelioma,” Clinical CancerResearch, vol. 19, no. 9, pp. 2493–2502, 2013.
[22] G. J. Gordon, G. N. Rockwell, R. V. Jensen et al., “Identificationof novel candidate oncogenes and tumor suppressors in malig-nant pleural mesothelioma using large-scale transcriptionalprofiling,” The American Journal of Pathology, vol. 166,no. 6, pp. 1827–1840, 2005.
[23] Y. D. Tan and H. Xu, “A general method for accurate estima-tion of false discovery rates in identification of differentiallyexpressed genes,” Bioinformatics, vol. 30, no. 14, pp. 2018–2025, 2014.
[24] A. M. Newman, C. L. Liu, M. R. Green et al., “Robust enumer-ation of cell subsets from tissue expression profiles,” NatureMethods, vol. 12, no. 5, pp. 453–457, 2015.
[25] A. Franceschini, D. Szklarczyk, S. Frankild et al., “STRINGv9.1: protein-protein interaction networks, with increased cov-erage and integration,” Nucleic Acids Research, vol. 41,pp. D808–D815, 2013.
[26] C.-H. Chin, S.-H. Chen, H.-H. Wu, C.-W. Ho, M.-T. Ko, andC.-Y. Lin, “cytoHubba: identifying hub objects and sub-networks from complex interactome,” BMC Systems Biology,vol. 8, Supplement 4, p. S11, 2014.
[27] E. Cerami, J. Gao, U. Dogrusoz et al., “The cBio cancer geno-mics portal: an open platform for exploring multidimensionalcancer genomics data,” Cancer Discovery, vol. 2, no. 5, pp. 401–404, 2012.
[28] T. Li, J. Fan, B. Wang et al., “TIMER: a web server for compre-hensive analysis of tumor-infiltrating immune cells,” CancerResearch, vol. 77, no. 21, pp. e108–e110, 2017.
[29] S. V. Vasaikar, P. Straub, J. Wang, and B. Zhang, “LinkedO-mics: analyzing multi-omics data within and across 32 cancertypes,” Nucleic Acids Research, vol. 46, no. D1, pp. D956–D963, 2018.
[30] Z. Tang, C. Li, B. Kang, G. Gao, C. Li, and Z. Zhang, “GEPIA: aweb server for cancer and normal gene expression profilingand interactive analyses,” Nucleic Acids Research, vol. 45,no. W1, pp. W98–W102, 2017.
[31] B. Li, E. Severson, J. C. Pignon et al., “Comprehensive analysesof tumor immunity: implications for cancer immunotherapy,”Genome Biology, vol. 17, no. 1, p. 174, 2016.
[32] J. Liu, J. X. Shen, H. T. Wu et al., “Collagen 1A1 (COL1A1)promotes metastasis of breast cancer and is a potential thera-peutic target,” Discovery Medicine, vol. 25, no. 139, pp. 211–223, 2018.
[33] U. Oleksiewicz, T. Liloglou, K. M. Tasopoulou et al.,“COL1A1, PRPF40A, and UCP2 correlate with hypoxiamarkers in non-small cell lung cancer,” Journal of CancerResearch and Clinical Oncology, vol. 143, no. 7, pp. 1133–1141, 2017.
[34] S. Sun, Y. Wang, Y. Wu et al., “Identification of COL1A1 as aninvasion‑related gene in malignant astrocytoma,” Interna-tional Journal of Oncology, vol. 53, no. 6, pp. 2542–2554, 2018.
[35] H. Mohan, M. Krumbholz, R. Sharma et al., “Extracellularmatrix in multiple sclerosis lesions: Fibrillar collagens, bigly-can and decorin are upregulated and associated with infiltrat-ing immune cells,” Brain Pathology, vol. 20, no. 5, pp. 966–975,2010.
[36] S. Cedrés, S. Ponce-Aix, N. Pardo-Aranda et al., “Analysis ofexpression of PTEN/PI3 K pathway and programmed celldeath ligand 1 (PD-L1) in malignant pleural mesothelioma(MPM),” Lung Cancer, vol. 96, pp. 1–6, 2016.
[37] S. B. Coffelt, K. Kersten, C. W. Doornebal et al., “IL-17-pro-ducing γδ T cells and neutrophils conspire to promote breastcancer metastasis,” Nature, vol. 522, no. 7556, pp. 345–348,2015.
[38] S. L. Zhou, Z. Dai, Z. J. Zhou et al., “CXCL5 contributes totumor metastasis and recurrence of intrahepatic cholangiocar-cinoma by recruiting infiltrative intratumoral neutrophils,”Carcinogenesis, vol. 35, no. 3, pp. 597–605, 2014.
[39] Y. Koga, A. Matsuzaki, A. Suminoe, H. Hattori, and T. Hara,“Neutrophil-derived TNF-related apoptosis-inducing ligand(TRAIL): a novel mechanism of antitumor effect by neutro-phils,” Cancer Research, vol. 64, no. 3, pp. 1037–1043, 2004.
[40] P. C. Kousis, B. W. Henderson, P. G. Maier, and S. O. Gollnick,“Photodynamic therapy enhancement of antitumor immunityis regulated by neutrophils,” Cancer Research, vol. 67, no. 21,pp. 10501–10510, 2007.
[41] L. Andzinski, N. Kasnitz, S. Stahnke et al., “Type I IFNs induceanti-tumor polarization of tumor associated neutrophils inmice and human,” International Journal of Cancer, vol. 138,no. 8, pp. 1982–1993, 2016.
[42] F. Dammeijer, L. A. Lievense, M. E. Kaijen-Lambers et al.,“Depletion of tumor-associated macrophages with a CSF-1R
12 BioMed Research International
kinase inhibitor enhances antitumor immunity and survivalinduced by DC immunotherapy,” Cancer ImmunologyResearch, vol. 5, no. 7, pp. 535–546, 2017.
[43] A. G. Pappas, S. Magkouta, I. S. Pateras et al., “Versican mod-ulates tumor-associated macrophage properties to stimulatemesothelioma growth,” OncoImmunology, vol. 8, articlee1537427, 2018.
[44] S. H. Chew and S. Toyokuni, “Malignant mesothelioma as anoxidative stress-induced cancer: an update,” Free Radical Biol-ogy & Medicine, vol. 86, pp. 166–178, 2015.
13BioMed Research International