Upload
others
View
4
Download
0
Embed Size (px)
Citation preview
Research ArticleIdentification of Molecular Characteristics and New PrognosticTargets for Thymoma by Multiomics Analysis
Dazhong Liu, Pengfei Zhang, Jiaying Zhao, Lei Yang, and Wei Wang
Department of Thoracic Surgery, Second Affiliated Hospital of Harbin Medical University, Harbin 150086, China
Correspondence should be addressed to Wei Wang; [email protected]
Received 16 January 2021; Revised 16 March 2021; Accepted 1 April 2021; Published 20 May 2021
Academic Editor: Tao Huang
Copyright © 2021 Dazhong Liu 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.
Background. Thymoma is a heterogeneous tumor originated from thymic epithelial cells. The molecular mechanism of thymomaremains unclear. Methods. The expression profile, methylation, and mutation data of thymoma were obtained from TCGAdatabase. The coexpression network was constructed using the variance of gene expression through WGCNA. Enrichmentanalysis using clusterProfiler R package and overall survival (OS) analysis by Kaplan-Meier method were carried out for theintersection of differential expression genes (DEGs) screened by limma R package and important module genes. PPI networkwas constructed based on STRING database for genes with significant impact on survival. The impact of key genes on theprognosis of thymoma was evaluated by ROC curve and Cox regression model. Finally, the immune cell infiltration,methylation modification, and gene mutation were calculated. Results. We obtained eleven coexpression modules, and three ofthem were higher positively correlated with thymoma. DEGs in these three modules mainly involved in MAPK cascade andPPAR pathway. LIPE, MYH6, ACTG2, KLF4, SULT4A1, and TF were identified as key genes through the PPI network. AUCvalues of LIPE were the highest. Cox regression analysis showed that low expression of LIPE was a prognostic risk factor forthymoma. In addition, there was a high correlation between LIPE and T cells. Importantly, the expression of LIPE was modifiedby methylation. Among all the mutated genes, GTF2I had the highest mutation frequency. Conclusion. These results suggestedthat the molecular mechanism of thymoma may be related to immune inflammation. LIPE may be the key genes affectingprognosis of thymoma. Our findings will help to elucidate the pathogenesis and therapeutic targets of thymoma.
1. Introduction
Thymoma is the most common anterior mediastinal com-partment tumor, originating from the thymic epithelial cellpopulation [1]. The incidence of thymomas is approximately2.5 cases per million people per year, with an age distributionranging from 10 to 80 years [2]. In addition, thymoma isoften associated with autoimmune diseases, especially myas-thenia gravis (MG) [3, 4]. However, the potential molecularoncogenesis of thymoma remains unknown. Generally, whena thymoma is diagnosed, the patient will receive surgicaltreatment. For stages III and IV patients, the 5-year survivalrates were 74% and <25%, respectively [5]. At present, nei-ther surgeon nor physician can predict the prognosis andmetastasis status of thymoma patients through X-ray exami-nation, nor can detailed treatment plan be formulated before
operation [6]. Obviously, the establishment of additional pre-dictors is very beneficial for the identification and treatmentof thymoma.
The pathogenesis of thymoma is various, and the rapiddevelopment of “genome” technology, including whole-genome expression analysis and next-generation sequencing(NGS), provides new means to explore the complexity andmap of genomic alterations in thymoma [7–9]. Epigeneticmodifications, including epigenetic alterations, are a featureof cancer because they play an important role in the processof carcinogenesis [10, 11]. In addition, the thymus provides aspecial microenvironment for the development and selectionof mature T cells. Recent evidence suggests that immuneresponses such as T cells are involved in the developmentof thymoma [12, 13]. However, the understanding of thepathogenesis of thymoma is still limited.
HindawiBioMed Research InternationalVolume 2021, Article ID 5587441, 15 pageshttps://doi.org/10.1155/2021/5587441
In recent years, with the development of molecular biol-ogy, more and more research projects have begun to exploremethods to accurately predict the prognosis of thymoma. Inthis study, we used multiomics datasets from the tumorgenome map (TCGA). The results may be helpful to under-stand the pathogenesis of thymoma and identify LIPE as apotential new therapeutic target through bioinformaticsanalysis. The novelty of this work is that we combined thevariance and difference of gene expression to screen the genesrelated to the prognosis of thymoma through coexpressionnetwork and PPI network. Then, the key genes were furtherscreened by methylation modification.
2. Materials and Methods
2.1. TCGA Dataset Processing and Coexpression Analysis.Thymoma mRNA-seq expression data, methylation data,mutation data, and clinical materials were obtained fromTCGA website (https://portal.gdc.cancer.gov/). The varianceof gene expression was calculated, and the top 1/4 genes wereintercepted for coexpression analysis through weighted genecoexpression network analysis (WGCNA).
2.2. Screening of Differentially Expressed Genes. The differen-tially expressed genes (DEGs) between thymoma and controlwere identified by limma R package. Set the filtering thresh-old P < 0:05.
2.3. Construction of PPI Network. The gene was mapped intothe STRING database (https://string-db.org) to obtain theprotein-protein interaction (PPI) network. A significant PPInetwork was obtained by comprehensive score ≥ 0:7, whichwas demonstrated by the Cytoscape software. The selectionof key genes was based on their association with other pro-teins: genes with higher connectivity were considered to playan important role in the PPI network [14, 15].
2.4. Enrichment Analysis. In order to analyze the biologicalfunctions and signaling pathways of differentially expressedgenes in thymoma-related modules, we performed enrich-ment analysis. Gene Ontology (GO) and the Kyoto Encyclo-paedia of Genes and Genomes (KEGG) were enriched byclusterProfiler R package. P < 0:05 was the threshold usedfor the significant terms. Gene set enrichment analysis (GSEA)was performed with the GSEA software for genes [16, 17].
2.5. Differential Methylation and Mutation Analysis. Thequality of the original probe data obtained from the methyl-ated microarray was checked, including background correc-tion, probe type difference adjustment, and probe exclusion.According to these in sample standardized procedures,DNA methylation was scored as a β value. We used samr Rpackage for differential methylation analysis. For a CpG siteto be considered differentially methylated, the difference inthe median β value in thymoma and normal samples shouldbe at least 0.1 and the P value <0.05. The nonsilent mutation(gene-level) data were analyzed using Maftools R-package.
2.6. Statistical Analysis. Statistical analysis was performedusing the SPSS software, version 23.0 (SPSS Inc., Chicago,
USA). Kaplan-Meier method was used to estimate the overallsurvival (OS). Cox regression model and Cox proportionalhazards regression method were used to identify predictorsof OS [18]. P value <0.05 was considered statistically signifi-cant [19].
3. Results
3.1. Coexpression of Genes in Thymoma. According to thevariance results of thymoma gene expression, the top 1/4genes with larger variance were selected for coexpressionanalysis. A coexpression network consisting of 5758 geneswas obtained. Taken 0.9 as the threshold of correlation coef-ficient, select the soft threshold as 7 (Figure 1(a)). A total of11 coexpression modules were identified through WGCNAanalysis (Figure 1(b)). In addition, we calculated the correla-tion between module genes and thymoma. We found thatMEgreen, MEblue, and MEturquoise had the highest correla-tion with tumor samples (Figure 1(c)). Furthermore, 2559differentially expressed genes (DEGs) were screened betweenthymoma and control group (P < 0:05) (Figure 1(d)).
3.2. Enrichment of Differentially Expressed Genes in Modules.Further, 913 intersection genes between DEGs and the threemodules with the highest correlation were selected as theimportant genes for subsequent study and enrichment analy-sis. The results of GO enrichment showed that these geneswere involved in 1234 biological processes (BP), 151 cell com-ponents (CC), and 214 molecular functions (MF). It mainlyincluded cell growth, positive regulation of MAPK cascade,ERK1 and ERK2 cascade, response to transforming growthfactor beta, and Wnt signaling pathway (Figure 2(a)). KEGGenrichment results showed a total of 40 terms, mainly involv-ing cell adhesion molecules, ECM-receptor interaction, focaladhesion, and PPAR signaling pathway (Figure 2(b)). In addi-tion, the GSEA results showed some of the same results asKEGG, mainly including cGMP-PKG signaling pathway, cho-lesterol metadata, and PPAR signaling pathway (Figure 2(c)).These same signaling pathways cover a large number of differ-entially expressed genes (Figure 2(d)).
3.3. Identification of Key Prognostic Genes. The overall survival(OS) analysis of selected important genes identified 88 geneswith significant impact on prognosis (P < 0:05). Mapping thesegenes into the STRING database yielded a PPI network of 45genes, which was displayed by Cytoscape (Figure 3(a)). Thetop six genes with the highest connectivity were analyzed indepth as key genes. Among them, the expression of MYH6and SULT4A1 in osteosarcoma was higher than that in controlgroup, while the expression of LIPE, ACTG2, KLF4, and TFwas decreased (Figure 3(b)). In addition, high expression ofLIPE and MYH6 could improve the OS of patients, andACTG2, KLF4, SULT4A1, and TF decreased the OS of patients(Figure 3(c)). ROC curves showed that the AUC values of thesesix genes were all greater than 0.6, especially those of LIPE, andKLF4 and TF were greater than 0.9 (Figure 3(d)).
3.4. The Effect of Key Genes on Prognosis. Multivariate sur-vival analysis was performed by Cox regression model, andnomogram was generated by Cox regression coefficients.
2 BioMed Research International
The nomogram showed that low expression of LIPE was a riskfactor for predicting the overall survival time of thymoma at 5and 8 years (Figure 4(a)). Calibration plots showed that thenomograms performed well compared with an ideal model(Figure 4(b)). In addition, Cox risk ratio model suggested thatthe survival rate of the high-risk population for thymoma waspoor (Figure 4(c)). Among them, low expression of LIPE andMYH6 and high expression of ACTG2, KLF4, SULT4A1, andTF were important risk factors.
3.5. Changes of Immune Microenvironment in Thymoma. Bycomparing the immune cell infiltration between thymoma
and control, we found that dendritic cells (DC) decreasedmost significantly in thymoma (Figure 5(a)). These differen-tially infiltrated immune cells were clustered into four groups(Figure 5(b)). The strongest correlation was found between Tcells and CD8 T cells or Th17 cells in thymoma tissues(Figure 5(c)). In addition, we analyzed the correlation betweenkey genes and immune cells (Figure 5(d)). LIPE had the stron-gest positive correlation with T cells and Th2 cells, MYH6 hadthe strongest positive correlation with NK cells, TF, KLF4, andaDC had the strongest positive correlation, SULT4A1 andpDC had the strongest positive correlation, and ACTG2 andneutrophils had the strongest positive correlation.
5 10 15 20
−0.2
0.0
0.2
0.4
0.6
0.8
1.0
Scale independence
Soft threshold (power)
Scal
e fre
e top
olog
y m
odel
fit,
signe
d R
2
1
2
3
4
5 6 78 910 12 14 16 18 20
5 10 15 20
0
500
1000
1500
Mean connectivity
Soft threshold (power)
Mea
n co
nnec
tivity
1
2
3
45 6 7 8 910 12 14 16 18 20
(a)
0.0
0.2
0.4
0.6
0.8
1.0Cluster dendrogram
Fastcluster::hclust (⁎, “average”)as.dist(dissTom)
Hei
ght
Module colors
(b)
Module−trait relationships
−1
−0.5
0
0.5
1MEyellow
MEblackMEgreen
MEgreenyellowMEturquoise
MEblueMEbrown
MEredMEpink
MEmagentaMEpurple
MEgrey
−0.03(0.7)
0.03(0.7)
−0.0077(0.9)
0.0077(0.9)
0.15(0.09)
−0.15(0.09)
−0.14(0.1)
0.14(0.1)
0.019(0.8)
−0.019(0.8)
0.072(0.4)
−0.072(0.4)
−0.061(0.5)
0.061(0.5)
−0.052(0.6)
0.052(0.6)
−0.27(0.002)
0.27(0.002)
−0.19(0.03)
0.19(0.03)
−0.093(0.3)
0.093(0.3)
0.0071(0.9)
−0.0071(0.9)
Tumor Normal
(c)
BCORP1
GPAM
ITGA7
KCNIP2
NKX2−1
PRKAR2B
RPS4Y1
SFTA3
DownUp
0
10
20
30
40
−25 −10 −5 0 5 10 25Log2 (fold change)
−Lo
g 10 P
val
ue
(d)
Figure 1: Coexpression analysis of gene expression in thymoma. (a) Determination of soft threshold power in coexpression analysis. The leftimage shows the scale-free fit index (y-axis) as a function of the soft-thresholding power (x-axis). The right image shows the averageconnectivity (degree, y-axis) as a function of the soft-thresholding power (x-axis). (b) Module cluster tree of thymoma genes with largevariance. Branches with different colors correspond to different modules. (c) The correlation between module and clinical trait. Each rowcorresponds to a module, and each column corresponds to a feature. Each cell contains the corresponding correlation and P value. (d)The differentially expressed genes between thymoma and control. Red nodes were significantly upregulated genes, and green nodes weresignificantly downregulated genes.
3BioMed Research International
GO:0016049
GO
:0043410
GO:0070371
GO:00
7155
9
GO:0016055
Decreasing Increasing
z−score
logFCDownregulated
Upregulated
IDGO:0016049
Description
GO:0043410GO:0070371GO:0071559GO:0016055
Cell growthPositive regulation of MAPK cascade
ERK1 and ERK2 cascadeResponse to transforming growth factor beta
Wnt signaling pathway
(a)
ECM−receptor interactionPPAR signaling pathway
Focal adhesionCell adhesion molecules
CD36COL2A1
COL6A6
COL9A2
COL9A3
FRAS1HSPG2
ITGA8
LAMA2LAMA4
LAMC3
NPNTRELN
SDC1
TNN
TNXBADIPOQ
ANGPTL4LPL
PCK1PLIN1
PLIN4PLIN5PLTPRXRGSCD
SORBS1BUB1B−PAK6
HRASMYLK2
PAK6
PDGFA
PDGFRA
PDGFRBPGF
PIK3R2
ALCAM
CDH1
CDH2CDH4
CLDN6
CLDN7
CLDN8
CNTNAP2IGSF11
L1CAMNECTIN1
NEO1NRCAM
SELP
CategoryCell adhesion moleculesECM−receptor interaction
Focal adhesionPPAR signaling pathway
−10
−5
0
5
Fold change
Size1215
1720
(b)
Figure 2: Continued.
4 BioMed Research International
3.6. Regulatory Factors Associated with Thymoma. By com-paring gene methylation modifications between thymomaand control, we obtained 943 differential methylation sites(Figure 6(a)). Among them, the methylation sites of chr1
accounted for the most, accounting for 13% (Figure 6(b)).Fourteen genes were identified as methylation factorsbecause they had opposite levels of methylation and expres-sion (Figure 6(c)). Among them, LIPE was significantly
p value0.0044
KEGG_CGMP−PKG SIGNALING PATHWAYp adjust
0.003
KEGG_CHOLESTEROL METABOLISM
0.0032
KEGG_PPAR SIGNALING PATHWAY 0.0490.04730.0473
−0.75
−0.50
−0.25
0.00
Runn
ing
enric
hmen
t sco
re
−10
0
10
20
5000 10000 15000 20000Rank in ordered dataset
Rank
ed li
st m
etric
(c)
AD
CY6
ADRA
2A
AGTR1
ATP1A2
CALML3
EDNRB
KCNMB2
MYH6
MYLK2
NPR1
PDE3B
PLN
ANGPTL4
APOBAPOC1
CD36LP
L
LRP1LR
P2
PLTPADIPOQ
ANGPTL4
CD36
LPL
PCK1
PLIN1
PLIN4
PLIN5
PLTP
RXRG
SCD
SORBS1
KEGG_CGMP−PKG SIGNALING PATHWAY
KEGG_CHOLESTEROL METABOLISM
KEGG_PPAR SIGNALING PATHWAY
(d)
Figure 2: Enrichment analysis of thymoma-related module genes. (a) Important genes were involved in biological processes. Red nodes areupregulated genes, and blue nodes are downregulated genes. (b) Important genes were involved in KEGG pathway. Different line colorsrepresent different signaling pathways which genes involved in. (c) KEGG pathway in GSEA for important genes. These pathways weresignificantly upregulated in thymoma. (d) The DEGs involved in the same KEGG pathway in the results of enrichment and GSEA.Different colors represent genes involved in different signaling pathways.
5BioMed Research International
DACT1
MST1WNT4
ABCC9 KLF4 ESPN
APOC1POU5F1BPNCK
TF
SULT4A1 SDSL
CFDLIPE
GPD1 DGAT2
AP3B2
OLFM1
SNCBITLN1FBLN5
SVEP1
FBLN2 ITIH5
HEPACAM2
CALCRL
ADCYAP1R1
LYVE1
OGDHL
SLCO1A2
MCIDAS
NOX4
DUOX1
CLCNKB
SLC16A2
MEGF11
ABCA10 FIGN
ACTG2SPAG6 NUDT14FGF17
KIF21B
MYH6
JPH2
(a)
ACTG2 KLF4 LIPE MYH6 SULT4A1 TF
Tumor Normal Tumor Normal Tumor Normal Tumor Normal Tumor Normal Tumor Normal
0
5
10
15
Groups
vst−
norm
aliz
ed co
unts
⁎⁎⁎⁎
⁎⁎⁎⁎ ⁎⁎⁎⁎⁎⁎⁎⁎
⁎⁎⁎⁎⁎⁎⁎⁎
(b)
Figure 3: Continued.
6 BioMed Research International
associated with OS in thymoma. In addition, GTF2I, the genewith the highest frequency of mutations in thymoma, wasmissense mutation in all samples (Figure 6(d)).
4. Discussion
Like other malignant tumors, the growth and proliferation ofthymoma have many biological factors. However, the exact
molecular basis of thymoma occurrence remains unclear. Inthis study, the possible molecular mechanism and regulatoryfactors of thymoma were explored through multiomics.
Early studies have shown that changes in certain genesseem to be associated with the development of thymic tumors[20, 21]. Our data suggest that there is a large difference ingene expression between thymoma and control. By identify-ing coexpression network constructed by genes with larger
++ ++++++++++++++++++++++++++++++++++++++++ ++
++
+
+ +
+ ++++++++++++++++++++++ ++++++++++++++++++++++++ +++ + +++ + + + ++
P = 0.043
0.00
0.25
0.50
0.75
1.00
0 1000 2000 3000 4000 5000Time (days)
Surv
ival
pro
babi
lity
0.00
0.25
0.50
0.75
1.00
0 1000 2000 3000 4000 5000Time (days)
Strata++
ACTG2=highACTG2=low
Survival analysis of ACTG2+++
+++++++ +++++++++++++++ ++++++++++++++++ ++ +
+ +
+
+ +
++++++++++++++++++ +++++++++++++++++ ++++++++++ ++ + + +++ + + + + +
P = 0.0061
Strata++
KLF4=highKLF4=low
++ +++++++++ +++++++++++++++++++++++++++++ +++++++ + + ++ ++ + + + ++ ++++++++++++++++++++++ +++++++++++++++++ +
++ +
+ +
+ + +
P = 0.023
Strata++
MYH6=highMYH6=low
+++++++++++++++++++++++++++++++++++
+ ++++++ +
+ +++
+ + +
+++++++++++++++++++++++++++ ++++++++++++++++++++ + ++
+ +++ + + + +
P = 0.0096
Strata++
SULT4A1=highSULT4A1=low
+++++++++ +++++++++++++++++++++++++++
+++++ ++ + + +
++ +
+
+ ++++++++++++++++++++++++++++++++++ ++++++++++
++ + ++ +++ + ++
P = 0.038
Strata++
TF=highTF=low
+ +++++++++++++++++++++++++++++++ +++++++++++ +
+++ + + ++ ++ + + +
++++++++++++++++++ +++++++++++++++++++++
++++ +++ + +
+ +
+ +
P = 0.045
Strata++
LIPE=highLIPE=low
Survival analysis of MYH6 Survival analysis of SULT4A1 Survival analysis of KLF4
Survival analysis of LIPE Survival analysis of TF
0.00
0.25
0.50
0.75
1.00
0 1000 2000 3000 4000 5000Time (days)
0.00
0.25
0.50
0.75
1.00
0 1000 2000 3000 4000 5000Time (days)
0.00
0.25
0.50
0.75
1.00
0 1000 2000 3000 4000 5000Time (days)
Surv
ival
pro
babi
lity
0.00
0.25
0.50
0.75
1.00
0 1000 2000 3000 4000 5000Time (days)
(c)
AUC
1 − specificity
Sens
itivi
ty
0.0 0.2 0.4 0.6 0.8 1.00.0
0.2
0.4
0.6
0.8
1.0
ACTG2, AUC = 0.89KLF4, AUC = 0.94LIPE, AUC = 0.99
MYH6, AUC = 0.64SULT4A1, AUC = 0.77TF, AUC = 0.9
(d)
Figure 3: Identification of key genes affecting overall survival of thymoma. (a) Cytoscape software shows the PPI network of important genesbased on the STRING database. (b) The expression of six genes with the highest connectivity in the PPI network. ∗∗∗P < 0:001. (c) The effectof six genes with the highest connectivity in the PPI network on the overall survival of thymoma (Kaplan-Meier plot). Red and green curvesare for high expression and low expression, respectively. (d) ROC curve of key genes. Different color curves represent different genes.
7BioMed Research International
Points0 10 20 30 40 50 60 70 80 90 100
ACTG2 0 800
KLF4 0 3000 7000 12000
LIPE 4500 4000 3500 3000 2500 2000 1500 1000 500 0
MYH6 7000 6000 5000 4000 3000 2000 1000 0
SULT4A1 3500
TF 0 1000 2500 4000 5500
Total points 0 20 40 60 80 100 120 140 160 180 200
Linear predictor −7 −5 −3 −1 0 1 2 3 4 5 6
5−year survival probability0.95 0.9 0.8
0.70.6
0.50.4
0.30.2
0.1
8−year survival probability0.95 0.9 0.8
0.70.6
0.50.4
0.30.2
0.1
(a)
0.0 0.2 0.4 0.6 0.8 1.0
0.0
0.2
0.4
0.6
0.8
1.0
Nomogram−prediced OS (%)
Obs
erve
d O
S (%
)
n = 118 d = 9 p = 6, 50 subjects per groupGray: ideal
X−resampling optimism added, B = 952Based on observed−predicted
5−year8−year
(b)
Figure 4: Continued.
8 BioMed Research International
variance, module genes with high correlation with thymomawere obtained. Intersection with differentially expressedgenes yielded 913 genes possibly associated with thymomadevelopment.
GO functional enrichment analysis is very powerful andwidely used to identify biological functions of gene expres-sion data [22]. In the GO functional enrichment results,MAPK cascade, ERK1 and ERK2 cascade, response to trans-forming growth factor beta (TGF-β), and Wnt signalingpathway were mainly involved. Mitogen-activated proteinkinase (MAPK) is a complex and interrelated signal cascade,which is closely related to the occurrence and progress oftumor, and plays an important regulatory role in cell prolifer-ation, differentiation, migration, and survival [23, 24]. ERK1/2 is also an effective target for anticancer [25]. Studies haveshown that MAPK signal and ERK 1/2 were significantlyactivated in thymoma [26]. TGF-β inhibited apoptosis andhad reduced expression of IFN-γ in effector cell, a key medi-ator of antitumor immunity [27]. Recently, it had been
proved that Wnt pathway was activated in human thymoma,which may be involved in the tumorigenesis [28]. These find-ings further confirmed that a variety of inflammatory pro-cesses and cytokines were involved in the pathogenesis ofthymoma.
In addition, in KEGG enrichment results, ECM also regu-lated intercellular communication, cell connectivity plasticity,and cell adhesion molecules interacting with various cytoki-nes/chemokines or growth factors [29, 30]. There were 34%of the genes in the ECM-receptor interaction pathway mutatedrepeatedly in cancer [31]. Focal adhesion kinase (FAK) ishighly expressed in thymic epithelial tumors and can be usedas an independent prognostic biomarker [32]. PPARγ overex-pression more than doubled insulin-stimulated thymoma viralprotooncogene phosphorylation during low lipid availability[33]. GSEA results had the same terms as KEGG enrichmentresults, in which cholesterol accumulation was a common fea-ture of cancer tissues. Recent evidence showed that cholesterolplayed a crucial role in the progress of cancer including breast,
−4
0
4
0 30 60 90 120Patient ID (increasing risk score)
Risk
scor
e
Risk groupHighLow
0
1000
2000
3000
4000
0 30 60 90 120Patient ID (increasing risk score)
Surv
ival
tim
e (da
ys)
EventDeathAlive
LIPE
MYH6
ACTG2
KLF4
SULT4A1
TF−1
−0.5
0
0.5
1
(c)
Figure 4: The expression of key genes affects the prognosis of patients with thymoma. (a) Nomogram for predicting overall survival inpatients with thymoma. (b) Plots depict the calibration of each model in terms of agreement between predicted and observed 5-year and8-year outcomes. (c) Risk factor correlation diagram. The green dot was the survival thymoma patient, and the red dot was the deadthymoma patient. The dotted line was the median risk score, the left side was the low-risk group, and the right side was the high-risk group.
9BioMed Research International
−0.20
−0.15
−0.10
−0.05
0.00
Immunity.cells
DC
Mas
t cel
lsEo
sinop
hils
Neu
trop
hils
B ce
llsTR
egiD
CTh
1 ce
llsT
help
er ce
llsTc
mCD
8 T
cells
Mac
roph
ages
Th17
cells
T ce
llsN
K CD
56br
ight
cells
Th2
cells
aDC
TFH
Tgd
Tem
pDC
NK
cells
NK
CD56
dim
cells
Cyto
toxi
c cel
ls
logF
C
DEG with immunity.cells in TCGA
(a)
aDC
B cells
Cytotoxic cells
Eosinophils
iDCMacrophages
NK CD56bright cellsNK cells
pDC
TFH
Th1 cells
Th2 cells
CD8 T cells
T helper cells
Tcm
Tem
DCMast cellsNeutrophils
NK CD56dim cellsTgd
TReg
T cells
Th17 cells
Cell cluster−ACell cluster−B
Cell cluster−CCell cluster−D
Logrank test, P value
0.05
0.001
1e−05
1e−08
Positive correlation with P < 0.0001Negative correlation with P < 0.0001
(b)
Figure 5: Continued.
10 BioMed Research International
prostate, and colorectal cancer [34]. Activation of cGMP PKGsignal may promote the growth of cervical cancer cells [35].
By screening the DEGs that had a significant impact on theprognosis of thymoma, we identified LIPE, MYH6, ACTG2,KLF4, SULT4A1, and TF as key genes. LIPEwas also predictedas a new prognostic marker of thymoma in other studies [36].Consistent with our analysis, MYH6 was differentiallyexpressed in thymoma [37]. We found that MYH6 may be apotential target for thymoma. ACTG2, KLF4, SULT4A1, andTF were all involved in the occurrence or development of can-cer, but their biological significance in thymoma was not clear[38–41]. This needs further study and discussion of the follow-up experiments.
From the perspective of immunemicroenvironment, innateimmune cells such as DC and adaptive immune cells such as Tcells played an important role in thymoma [13, 42, 43]. Therewas a strong correlation between LIPE and immune cells, sug-gesting that LIPE may participate in the prognosis of thymomaby regulating the immunity system. Interestingly, we found thatLIPE was also a gene regulated by methylation. DNA and RNAmethylation genes are commonly studied as biomarkers [44,45], which also seems to be a way for LIPE to participate inthe development of thymoma [46]. On the other hand, geneticdifference in thymoma was also an effective way to screenpotential therapeutic targets [9]. GTF2I mutation occurs at highfrequency in thymoma and is a marker of good prognosis [47].
aDC
B.ce
llsCD
8.T.
cells
Cyto
toxi
c.cel
lsD
CEo
sinop
hils
iDC
Mac
roph
ages
Mas
t.cel
lsN
eutr
ophi
lsN
K.CD
56br
ight
.cells
NK.
CD56
dim
.cells
NK.
cells
pDC
T.ce
llsT.
help
er.ce
llsTc
mTe
mTF
HTg
dTh
1.ce
llsTh
17.ce
llsTh
2.ce
llsTR
eg
aDCB cellsCD8 T cellsCytotoxic cellsDCEosinophilsiDCMacrophagesMast cellsNeutrophilsNK CD56bright cellsNK CD56dim cellsNK cellspDCT cellsT helper cellsTcmTemTFHTgdTh1 cellsTh17 cellsTh2 cellsTReg −1
−0.5
0
0.5
1
(c)
⁎⁎
⁎⁎
⁎⁎
⁎⁎
⁎⁎
⁎⁎⁎⁎
⁎⁎
⁎⁎
⁎⁎
⁎⁎
⁎ ⁎ ⁎
⁎⁎⁎⁎ ⁎⁎
⁎⁎
⁎⁎⁎⁎ ⁎⁎ ⁎⁎
⁎⁎
⁎⁎
⁎⁎ ⁎⁎
⁎⁎
⁎⁎⁎⁎
⁎⁎
⁎⁎ ⁎⁎
⁎⁎
⁎⁎
⁎⁎⁎⁎⁎⁎⁎⁎
⁎⁎⁎⁎
⁎⁎
⁎ ⁎ ⁎
⁎
⁎
⁎
⁎
⁎ ⁎
⁎
⁎ ⁎
⁎ ⁎
⁎ ⁎ ⁎ ⁎
⁎ ⁎ ⁎
⁎ ⁎
⁎ ⁎ ⁎⁎
⁎⁎
ACTG2
aDC
B ce
llsCD
8 T
cells
Cyto
toxi
c cel
lsD
CEo
sinop
hils
iDC
Mac
roph
ages
Mas
t cel
lsN
eutr
ophi
lsN
K CD
56br
ight
cells
NK
CD56
dim
cells
NK
cells
pDC
T ce
llsT
help
er ce
llsTc
mTe
mTF
HTg
dTh
1 ce
llsTh
17 ce
llsTh
2 ce
llsTR
egKLF4
LIPE
MYH6
SULT4A1
TF
Immune_cells
−0.50
−0.25
0.00
0.25
0.50
⁎ P < 0.05⁎⁎ P < 0.01
CorrelationHub genes cor with immune cells in THYM
(d)
Figure 5: Immune cell infiltration in thymoma. (a) The difference of immune cell infiltration between thymoma and control. The blue linerepresents a significant difference. (b) Clustering of immunocytes with differential infiltration. The red line represents the positive correlationbetween immune cells, and the blue line represents the negative correlation. (c) Correlation between immune cells in thymoma. Redrepresents positive correlation between immune cells, and blue line represents negative correlation. The size of the node represents thesize of the correlation coefficient. (d) Correlation between key genes and immune cells. Red represents positive correlation betweenimmune cells, and blue line represents negative correlation. ∗P < 0:05 and ∗∗P < 0:01.
11BioMed Research International
Hypermethylated DMPs Hypomethylated DMPs
cg25270788
cg27544288
cg03634592 cg02235894
cg06405765
cg20384325cg23138608cg18262591cg07846737
cg04510420
0
10
15
20
30
−0.25 0.00 0.25 0.50Methylation difference (beta−value)
Feature1st Exon3’UTR5’UTRBodyIsland
OpenseaShelfShoreTSS1500TSS200
51.22 % 48.78 %
6.2%
3.2%
9.3%
39.2%
3.4%
7.3%
1.4%
1.2%
12.1%
16.8%
1stExon
3’UTR
5’UTR
Body
Island
Opensea
Shelf
Shore
TSS1500
TSS200
−lo
g10
(P v
alue
)
(a)
chr19: 7%
chr4: 5%
chr7: 6%
chr11: 5%chr16: 3%chr15: 3%
chr8: 5%
chr12: 7%
chr14: 3%chr18: 2%
chr10: 5%
chr1: 13%
chr22: 1%
chr6: 7%chr2: 7% chr3: 5%
chr20: 2%chr21: 1%
chr17: 6%
chr5: 5%
chr9: 2%
chr13: 3%
(b)
Figure 6: Continued.
12 BioMed Research International
TGFBR2
STX11
LIPE
LAMA4
KRT18
KLC3
HRAS
GPIHBP1
FZD4
FERMT2
ERBB4
EPHA1
CD36
ART4
Exp Methy
Sym
bol
TypeExpMethy
logFC123
45
StatusUPDOWN
Heatmap with Hub_Genes Exp in TCGAGroup
GroupTumorNormal
−2
0
2
(c)
0
688
NA
NA
SMARCA4
RYR2
NF1
DNAH3
DNAH17
CYLD
CUBN
CSMD2
CLIP2
BDP1
BCOR
TP53
OSGIN1
HDAC4
DMD
AL022578.1
MUC16
TTN
HRAS
GTF2I
2%
2%
2%
2%
2%
2%
2%
2%
2%
2%
2%
3%
3%
3%
3%
3%
7%
7%
8%
50%
0 62
Masaoka_stageVital_statusRaceGender
Missense_MutationSplice_SiteNonsense_Mutation
In_Frame_InsFrame_Shift_DelMulti_Hit
GenderMaleFemaleNA
RaceAsianWhiteNA
Not reportedBlack or African American
Vital_statusAliveDeadNA
Masaoka_stageStage IStage IIIStage IIb
Stage IIaNAStage IVb
Altered in 82 (66.67%) of 123 samples.
(d)
Figure 6: Methylation and mutation in thymoma. (a) Differential methylation sites between thymoma and control. (b) The proportion ofmethylation sites in different chromosomes. (c) The expression and methylation of methylation factors. Red node represents upregulation,and blue node represents downregulation. Yellow represents positive gene expression, while blue represents negative gene expression. (d)The top 20 genes with the highest mutation frequency in thymoma. Each cell represents a sample.
13BioMed Research International
However, this study also had some limitations. Firstly,conclusions may be limited by small samples, especially con-trol samples. Secondly, the results of this study had not beenverified by molecular experiments, so the interpretation ofthe results may be cautious. In this study, the possible molec-ular changes and pathogenesis of thymoma were investigatedusing the multiomics data from TCGA database. This studyidentified key genes related to the prognosis of thymoma,including LIPE, MYH6, ACTG2, KLF4, SULT4A1, and TF.The expression of these genes in thymoma may be a promis-ing biomarker, which needs further study.
5. Conclusion
In this study, potential targets associated with thymoma wereidentified by combining thymoma-related gene expression,methylation, and mutation data. Using a variety of bioinfor-matics analysis methods, we found that important genesrelated to thymoma were associated with immune inflamma-tory response. LIPE, MYH6, ACTG2, KLF4, SULT4A1, andTF were the key genes affecting the prognosis of thymoma.Among them, LIPE was also modified by methylation.
Data Availability
Thymoma mRNA-seq expression data, methylation data,mutation data, and clinical materials were obtained fromTCGA website (https://portal.gdc.cancer.gov/).
Conflicts of Interest
The authors declare that the research was conducted in theabsence of any commercial or financial relationships thatcould be construed as a potential conflict of interest.
Authors’ Contributions
Dazhong Liu and Pengfei Zhang contributed equally to thisarticle as co-first authors.
Acknowledgments
This study was supported by the Project of HeilongjiangHealth and Family Planning Commission (2016-066) andScience and Technology Innovation Talent Project of HarbinScience and Technology Bureau (2016RAQXJ149).
References
[1] L. Yu, J. Ke, X. du, Z. Yu, and D. Gao, “Genetic characteriza-tion of thymoma,” Scientific Reports, vol. 9, no. 1, p. 2369,2019.
[2] M. A. den Bakker, A. C. Roden, A. Marx, andM.Marino, “His-tologic classification of thymoma: a practical guide for routinecases,” Journal of Thoracic Oncology, vol. 9, no. 9, pp. S125–S130, 2014.
[3] B. Aydemir, “The effect of myasthenia gravis as a prognosticfactor in thymoma treatment,” Northern Clinics of Istanbul,vol. 3, no. 3, pp. 194–200, 2016.
[4] Y. Sun, C. Gu, J. Shi et al., “Reconstruction of mediastinal ves-sels for invasive thymoma: a retrospective analysis of 25 cases,”Journal of Thoracic Disease, vol. 9, no. 3, pp. 725–733, 2017.
[5] D. J. Kim, W. I. Yang, S. S. Choi, K. D. Kim, and K. Y. Chung,“Prognostic and clinical relevance of the World Health Orga-nization schema for the classification of thymic epithelialtumors: a clinicopathologic study of 108 patients and literaturereview,” Chest, vol. 127, no. 3, pp. 755–761, 2005.
[6] D. Bian, F. Zhou, W. Yang et al., “Thymoma size significantlyaffects the survival, metastasis and effectiveness of adjuvanttherapies: a population based study,” Oncotarget, vol. 9,no. 15, pp. 12273–12283, 2018.
[7] M. Radovich, C. R. Pickering, I. Felau et al., “The integratedgenomic landscape of thymic epithelial tumors,” Cancer Cell,vol. 33, no. 2, pp. 244–258.e10, 2018, e10.
[8] T. Sakane, T. Murase, K. Okuda et al., “A mutation analysis ofthe EGFR pathway genes, RAS, EGFR, PIK3CA, AKT1 andBRAF, and TP53 gene in thymic carcinoma and thymoma typeA/B3,” Histopathology, vol. 75, no. 5, pp. 755–766, 2019.
[9] F. Enkner, B. Pichlhöfer, A. T. Zaharie et al., “Molecular profil-ing of thymoma and thymic carcinoma: genetic differencesand potential novel therapeutic targets,” Pathology OncologyResearch, vol. 23, no. 3, pp. 551–564, 2017.
[10] S. Li, Y. Yuan, H. Xiao et al., “Discovery and validation of DNAmethylation markers for overall survival prognosis in patientswith thymic epithelial tumors,” Clinical Epigenetics, vol. 11,no. 1, p. 38, 2019.
[11] M. Venza, M. Visalli, C. Beninati, C. Biondo, D. Teti, andI. Venza, “Role of genetics and epigenetics in mucosal, uveal,and cutaneous melanomagenesis,” Anti-Cancer Agents inMedicinal Chemistry, vol. 16, no. 5, pp. 528–538, 2016.
[12] P. Christopoulos and P. Fisch, “Acquired T-cell immunodefi-ciency in thymoma patients,” Critical Reviews in Immunology,vol. 36, no. 4, pp. 315–327, 2016.
[13] M. Omatsu, T. Kunimura, T. Mikogami et al., “Difference indistribution profiles between CD163+ tumor-associated mac-rophages and S100+ dendritic cells in thymic epithelialtumors,” Diagnostic Pathology, vol. 9, no. 1, p. 215, 2014.
[14] C. Gu, X. Shi, Z. Huang et al., “A comprehensive study of con-struction and analysis of competitive endogenous RNA net-works in lung adenocarcinoma,” Biochimica et BiophysicaActa (BBA) - Proteins and Proteomics, vol. 1868, no. 8,p. 140444, 2020.
[15] X. Shi, T. Huang, J. Wang et al., “Next-generation sequencingidentifies novel genes with rare variants in total anomalouspulmonary venous connection,” eBioMedicine, vol. 38,pp. 217–227, 2018.
[16] C. Gu, Z. Huang, X. Chen et al., “TEAD4 promotes tumordevelopment in patients with lung adenocarcinoma via ERKsignaling pathway,” Biochimica et Biophysica Acta - MolecularBasis of Disease, vol. 1866, no. 12, article 165921, 2020.
[17] C. Gu, X. Shi, X. Dang et al., “Identification of common genesand pathways in eight fibrosis diseases,” Frontiers in Genetics,vol. 11, p. 627396, 2020.
[18] C. Gu, R. Wang, X. Pan et al., “Comprehensive study of prog-nostic risk factors of patients underwent pneumonectomy,”Journal of Cancer, vol. 8, no. 11, pp. 2097–2103, 2017.
[19] C. Gu, X. Pan, R.Wang et al., “Analysis of mutational and clin-icopathologic characteristics of lung adenocarcinoma withclear cell component,” Oncotarget, vol. 7, no. 17, pp. 24596–24603, 2016.
14 BioMed Research International
[20] X. D. Wang, P. Lin, Y. X. Li et al., “Identification of potentialagents for thymoma by integrated analyses of differentiallyexpressed tumour-associated genes and molecular dockingexperiments,” Experimental and Therapeutic Medicine,vol. 18, no. 3, pp. 2001–2014, 2019.
[21] F. J. Meng, S. Wang, J. Zhang et al., “Alteration in gene expres-sion profiles of thymoma: genetic differences and potentialnovel targets,” Thoracic Cancer, vol. 10, no. 5, pp. 1129–1135, 2019.
[22] K. Rue-Albrecht, P. A. McGettigan, B. Hernández et al.,“GOexpress: an R/bioconductor package for the identificationand visualisation of robust gene ontology signatures throughsupervised learning of gene expression data,” BMC Bioinfor-matics, vol. 17, no. 1, p. 126, 2016.
[23] C. Braicu, M. Buse, C. Busuioc et al., “A comprehensive reviewon MAPK: a promising therapeutic target in cancer,” Cancers,vol. 11, no. 10, p. 1618, 2019.
[24] M. Cargnello and P. P. Roux, “Activation and function of theMAPKs and their substrates, the MAPK-activated proteinkinases,” Microbiology and Molecular Biology Reviews,vol. 75, no. 1, pp. 50–83, 2011.
[25] A. Plotnikov, K. Flores, G. Maik-Rachline et al., “The nucleartranslocation of ERK1/2 as an anticancer target,” Nature Com-munications, vol. 6, no. 1, p. 6685, 2015.
[26] Z. Yang, S. Liu, Y. Wang et al., “High expression of KITLG is anew hallmark activating the MAPK pathway in type A and ABthymoma,” Thoracic Cancer, vol. 11, no. 7, pp. 1944–1954,2020.
[27] M. Ibrahim, D. Scozzi, K. A. Toth et al., “Naive CD4+T cellscarrying a TLR2 agonist overcome TGF-β–mediated tumorimmune evasion,” Journal of Immunology, vol. 200, no. 2,pp. 847–856, 2018.
[28] P. Vodicka, L. Krskova, I. Odintsov et al., “Expression of mol-ecules of the Wnt pathway and of E-cadherin in the etiopatho-genesis of human thymomas,” Oncology Letters, vol. 19, no. 3,pp. 2413–2421, 2020.
[29] H. E. Barker, J. T. E. Paget, A. A. Khan, and K. J. Harrington,“The tumour microenvironment after radiotherapy: mecha-nisms of resistance and recurrence,” Nature Reviews. Cancer,vol. 15, no. 7, pp. 409–425, 2015.
[30] C. Ionescu, C. Braicu, R. Chiorean et al., “TIMP-1 expressionin human colorectal cancer is associated with SMAD3 geneexpression levels: a pilot study,” Journal of Gastrointestinaland Liver Diseases, vol. 23, no. 4, pp. 413–418, 2020.
[31] B. Liu, F. F. Hu, Q. Zhang et al., “Genomic landscape andmutational impacts of recurrently mutated genes in cancers,”Molecular Genetics & Genomic Medicine, vol. 6, no. 6,pp. 910–923, 2018.
[32] M. Li, F. Hou, J. Zhao et al., “Focal adhesion kinase is overex-pressed in thymic epithelial tumors and may serve as an inde-pendent prognostic biomarker,” Oncology Letters, vol. 15,no. 3, pp. 3001–3007, 2018.
[33] S. Hu, J. Yao, A. A. Howe et al., “Peroxisome proliferator-activated receptor γ decouples fatty acid uptake from lipidinhibition of insulin signaling in skeletal muscle,” MolecularEndocrinology, vol. 26, no. 6, pp. 977–988, 2012.
[34] T. Murai, “Cholesterol lowering: role in cancer prevention andtreatment,” Biological Chemistry, vol. 396, no. 1, pp. 1–11,2015.
[35] L. Gong, Y. Lei, X. Tan et al., “Propranolol selectively inhibitscervical cancer cell growth by suppressing the cGMP/PKG
pathway,” Biomedicine & Pharmacotherapy, vol. 111,pp. 1243–1248, 2019.
[36] Q. Li, Y. L. Su, and W. X. Shen, “A novel prognostic signatureof seven genes for the prediction in patients with thymoma,”Journal of Cancer Research and Clinical Oncology, vol. 145,no. 1, pp. 109–116, 2019.
[37] J. Xi, L. Wang, C. Yan et al., “The Cancer Genome Atlasdataset-based analysis of aberrantly expressed genes by Gen-eAnalytics in thymoma associated myasthenia gravis: focusingon T cells,” Journal of Thoracic Disease, vol. 11, no. 6,pp. 2315–2323, 2019.
[38] A. Simiczyjew, A. J. Mazur, E. Dratkiewicz, and D. Nowak,“Involvement of β- and γ-actin isoforms in actin cytoskeletonorganization and migration abilities of bleb-forming humancolon cancer cells,” PLoS One, vol. 12, no. 3, articlee0173709, 2017.
[39] M. Moral, C. Segrelles, A. B. Martínez-Cruz et al., “Transgenicmice expressing constitutively active Akt in oral epitheliumvalidate KLFA as a potential biomarker of head and neck squa-mous cell carcinoma,” In Vivo, vol. 23, no. 5, pp. 653–660,2009.
[40] J. Dai, Z. Bing, Y. Zhang et al., “Integrated mRNAseq andmicroRNAseq data analysis for grade III gliomas,” MolecularMedicine Reports, vol. 16, no. 5, pp. 7468–7478, 2017.
[41] D. Ferreira, R. Ponraj, A. Yeung, and J. de Malmanche, “Purered cell aplasia associated with thymolipoma: complete anae-mia resolution following thymectomy,” Case Reports in Hema-tology, vol. 2018, Article ID 8627145, 4 pages, 2018.
[42] L. Wang, O. E. Branson, K. Shilo, C. L. Hitchcock, and M. A.Freitas, “Proteomic signatures of thymomas,” PLoS One,vol. 11, no. 11, article e0166494, 2016.
[43] S. Shelly, N. Agmon-Levin, A. Altman, and Y. Shoenfeld,“Thymoma and autoimmunity,” Cellular & Molecular Immu-nology, vol. 8, no. 3, pp. 199–202, 2011.
[44] C. Gu, X. Shi, C. Dai et al., “RNAm6Amodification in cancers:molecular mechanisms and potential clinical applications,”The Innovation, vol. 1, no. 3, p. 100066, 2020.
[45] C. Gu and C. Chen, “Methylation in lung cancer: a briefreview,” Methods in Molecular Biology, vol. 2204, pp. 91–97,2020.
[46] Y. Bi, Y. Meng, Y. Niu et al., “Genome‑wide DNAmethylationprofile of thymomas and potential epigenetic regulation of thy-moma subtypes,” Oncology Reports, vol. 41, no. 5, pp. 2762–2774, 2019.
[47] I. Petrini, P. S. Meltzer, I. K. Kim et al., “A specific missensemutation in GTF2I occurs at high frequency in thymic epithe-lial tumors,” Nature Genetics, vol. 46, no. 8, pp. 844–849, 2014.
15BioMed Research International