Genomics

Multi-omics Approach Reveals DistinctDifferences in Left- and Right-SidedColon CancerWangxiong Hu1, Yanmei Yang2, Xiaofen Li1,3, Minran Huang1, Fei Xu1,Weiting Ge1, Suzhan Zhang1,4, and Shu Zheng1,4

Abstract

Increasing evidence suggests that left-sided colon cancer(LCC) and right-sided colon cancer (RCC) are emerging as twodifferent colorectal cancer types with distinct clinical character-istics. However, the discrepancy in the underlying molecularevent between these types of cancer has not been thoroughlyelucidated to date and warrants comprehensive investigation.To this end, an integrated dataset from The Cancer GenomeAtlas was used to compare and contrast LCC and RCC, coveringmutation, DNA methylation, gene expression, and miRNA.Briefly, the signaling pathway cross-talk is more prevalent inRCC than LCC, such as RCC-specific PI3K pathway, which oftenexhibits cross-talk with the RAS and P53 pathways. Meanwhile,methylation signatures revealed that RCC was hypermethylatedrelative to LCC. In addition, differentially expressed genes(n ¼ 253) and differentially expressed miRNAs (n ¼ 16) were

determined between LCC and RCC. Especially for ProstateCancer Susceptibility Candidate 1 (PRAC1), a gene that wasclosely associated with hypermethylation, was the top signif-icantly downregulated gene in RCC. Multi-omics comparisonof LCC and RCC suggests that there are more aggressive markersin RCC and that tumor heterogeneity occurs within the loca-tion-based subtypes of colon cancer. These results clarify thedebate regarding the conflicting prognosis between LCC andRCC, as proposed by different studies.

Implications: The underlying molecular features present in LCCand RCC identified in this study are beneficial for adoptingreasonable therapeutic approaches to prolong overall survivaland progression-free survival in colorectal cancer patients. MolCancer Res; 16(3); 476–85. �2017 AACR.

IntroductionColorectal cancer is the third most commonly diagnosed

cancer in males and the second in females (1). Typically, thececum, ascending colon, hepatic flexure, and transverse colonare classified as the right colon, and the descending colon andsigmoid colon are classified as the left colon (2). It is knownthat the right and left colons have different embryologic origins(right colon arises from the embryonic midgut, and left colon

originates from the hindgut), microenvironments, and distinctblood supplies (3). In addition, left-sided colon cancer (LCC)and right-sided colon cancer (RCC) showed distinct differencesin epidemiology, biology, pathology, genetic mutations, andclinical outcomes (3–5).

It has been shown that RCC patients were older, had increasedtumor sizes and more advanced tumor stages, were more oftenfemale, and had poorly differentiated tumors (5, 6). In addition,increasing numbers of studies demonstrated a poor survival inRCC comparedwith LCC (7). Recently, Petrelli and colleagues (8)found that LCCwas associated with a significantly reduced risk ofdeath, which was independent of stage, race, adjuvant chemo-therapy, year of study, number of participants, and quality of theincluded studies through interrogating 1,437,846 patients frompublicly available data. However, the underlying molecularevents associated with huge difference between LCC and RCCare poorly understood. In this study, we comprehensively char-acterize the somatic mutations, genome-wide transcriptional(mRNA andmiRNA), and epigenetic (DNAmethylation) profilesof the LCC and RCC combinedwith the correlative analyses of theexpression, methylation, and clinical data from the CancerGenome Atlas (TCGA).

Materials and MethodsMutation data retrieval and processing

The colon adenocarcinoma (COAD) somatic mutation dataand clinical information were downloaded from the TCGA dataportal (March 2, 2015). Silent mutation and RNA mutation

1Cancer Institute (Key Laboratory of Cancer Prevention and Intervention, ChinaNational Ministry of Education), The Second Affiliated Hospital, ZhejiangUniversity School of Medicine, Hangzhou, Zhejiang, China. 2Key Laboratory ofReproductive and Genetics, Ministry of Education, Women's Hospital, ZhejiangUniversity, Hangzhou, Zhejiang, China. 3Department of Abdominal Oncology,West China Hospital, Sichuan University, Chengdu, Sichuan, China. 4ResearchCenter for Air Pollution and Health, School of Medicine, Zhejiang University,Hangzhou, Zhejiang, China.

Note: Supplementary data for this article are available at Molecular CancerResearch Online (http://mcr.aacrjournals.org/).

W. Hu and Y. Yang contributed equally to this article.

Corresponding Authors: W. Hu, Cancer Institute (Key Laboratory of CancerPrevention and Intervention, China National Ministry of Education), The SecondAffiliated Hospital, Zhejiang University School of Medicine, 88 Jiefang Rd,Hangzhou, Zhejiang 310009, China. Phone: 86-151-67166683; E-mail:wxhu@zju.edu.cn;W. Ge, geweiting@zju.edu.cn; S. Zhang, zuci@zju.edu.cn; andS. Zheng, zhengshu@zju.edu.cn

doi: 10.1158/1541-7786.MCR-17-0483

�2017 American Association for Cancer Research.

MolecularCancerResearch

Mol Cancer Res; 16(3) March 2018476

on August 18, 2020. © 2018 American Association for Cancer Research. mcr.aacrjournals.org Downloaded from

Published OnlineFirst November 29, 2017; DOI: 10.1158/1541-7786.MCR-17-0483

were discarded. Next, clinical information (i.e., neoplasm loca-tion) of each patient was added right after the mutational genesvia the unique patient ID. Frequent mutational gene sets(FMGS) with point mutations and small insertions/deletionsand hidden association rules (AR) were mined as in ourprevious work (9). Samples with ambiguous location annota-tion were excluded.

HM450k data retrieval and processCOAD level three DNA methylation data (HumanMethyla-

tion450) were downloaded from the TCGA data portal (March10, 2015). The methylation level of each probe was measuredas the beta value ranging from 0 to 1, which is calculated as theratio of the methylated signal to the sum of the methylated andunmethylated signal. Probes with an "NA" value in more than10% of the LCC or RCC samples were discarded. Next, bum-phunter Bioconductor package was used to seek the differen-tially methylated region (DMR) in the remaining probes (10).GRCh37 genome annotation was retrieved from Ensembl(http://grch37.ensembl.org/index.html) since the TCGA levelthree DNA methylation data (HumanMethylation450) usingGRCh37 for probe annotation. The circular diagram was plot-ted by the RCircos package (11).

Gene expression data processing and normalizationAll level three mRNA expression datasets (RNASeqV2) were

obtained from the TCGA (October 2015). Gene expressiondata analysis was performed similar to our previous work(12). Briefly, differentially expressed mRNA analysis betweenLCC and RCC was performed by the limma package forR/Bioconductor. Genes with an expression level < 1 (RSEM-normalized counts) in more than 50% of the samples wereremoved. Significantly differentially expressed mRNAs wereselected according to the FDR-adjusted P value < 0.05 andfold change > 2 condition. Validation mRNA expressiondataset was downloaded from the Gene Expression Omnibusdatabase under the accession numbers: GSE14333. The data-set was explored similarly to what was mentioned previously.Volcano plots were created with the ggplot2 package (http://ggplot2.org/).

Integrated analysis of miRNAs and mRNAsDifferentially expressed miRNAs between LCC and RCC were

explored similar to differentially expressed genes (DEG). Putativetarget genes of differentially expressed miRNAs were predictedusingmiRanda (13) and TargetScan (14). Intersection of putativetarget genes and DEGs was retained for further analysis. Wecalculated correlation between miRNA and its mRNA targets andretained negatively correlated pairs (Pearson correlation coeffi-cients < 0 and P < 0.01). MiRNA–mRNA interaction network wasconstructed by using Cytoscape (v3.2.1).

Network constructionMatched mRNA–miRNA samples were retained for coexpres-

sion analysis. The coexpression network was constructed byweighted correlation network analysis (WGCNA) package for R(15) and explored as in our previous work (12). The subnet-works constituted by the 50 top hub genes in the specificmodule were visualized by VisANT (16). Next, the top 50 hubgenes were subjected to Gene Ontology (GO) interrogation.

The P value was determined by the hypergeometric test withthe whole annotation as reference set and then adjusted formultiple testing using the Benjamini–Hochberg FDR correc-tion method. GO enrichment analysis was conducted by BiN-GO implemented in Cytoscape (17, 18). All statistical analysesand graphical representations were performed in the R pro-gramming language (� 64, version 3.0.2) unless otherwisespecified.

Immunohistochemistry stainingFifteen LCC and 32 RCC formalin-fixed paraffin-embedded

(FFPE) samples were collected from the Second AffiliatedHospital, Zhejiang University School of Medicine between2010 and 2016. Immunohistochemistry (IHC) staining wasperformed similar to our previous work (19) except for anti-body was replaced with anti-PRAC (1:500 dilution, Invitrogen,PA1-46237).

ResultsSparse somatic mutation pattern in RCC

Here, we used the Apriori algorithm to find the FMGSs withpoint mutations and small insertions/deletions from 84 LCCsand 145 RCCs as part of the TCGA Pan-Cancer effort. Totally,65 and 326 unique k-1 FMGSs were identified in LCC and RCC,respectively. The largest FMGS size identified in LCC and RCCwas three. Deeper analysis of the FMGSs showed that 8-foldmore one-itemsets (k ¼ 1) of FMGSs were found in RCC thanLCC (Supplementary Table S1). Though the top driver muta-tion genes were basically the same, the pattern of the mutationfrequencies differed greatly. In LCC, APC (support: 0.835) andTP53 (support: 0.647) were absolutely the dominant drivermutation genes. However, in RCC, a decreasing gene mutationpattern was observed. Despite that finding, APC and TP53remained the top mutation genes. The sparse mutation patternin RCC results in more co-occurred driver mutation genes.Except for TP53 and APC, other drivers (e.g., RNF43, FAT4,PIK3CA) also frequently co-occurred with other cancer-relatedgenes in RCC. For example, in RCC, RNF43 often co-occurredwith, for instance, SYNE1, OBSCN, BRAF, and LAMA5 (Sup-plementary Table S1).

It is well-known that microsatellite instability (MSI) patientshave an elevated frequency of single-nucleotide variantsbecause of inactivation of the DNA mismatch repair (MMR)system (20). MSI is observed in 15% of sporadic colorectaltumors diagnosed in the United States (21, 22). Therefore, wefurther divided the LCC and RCC into MSI-high (MSI-H) andmicrosatellite stable (MSS) subgroups according to the clinicalannotation by TCGA (23) and the findings by Hause andcolleagues (21). Intriguingly, we found that the proportion ofMSI-H in RCC was 20% (29 samples) and only approximately5% in LCC (4 samples). In this context, we speculated that thesparse somatic mutation pattern in RCC may be biased by theMSI-H samples since the MSI-H samples had an elevatedmutational load. As such, we independently calculated theFMGSs in MSS and MSI-H subgroups of RCC. However, similarresults held in the MSS of RCC, with 127 unique k-1 FMGSs and68 one-itemset (k ¼ 1) FMGSs, which is 2-fold and 3-fold morethan in LCC, respectively.

Regarding the RCC-MSI-H, a much higher mutational burdenwas observed, with 1,026 mutated genes being observed per

Left- and Right-Sided Colon Cancer Comparison

www.aacrjournals.org Mol Cancer Res; 16(3) March 2018 477

on August 18, 2020. © 2018 American Association for Cancer Research. mcr.aacrjournals.org Downloaded from

Published OnlineFirst November 29, 2017; DOI: 10.1158/1541-7786.MCR-17-0483

tumor. In addition, as many as 200 one-itemset (k ¼ 1) FMGSswith support > 0.3 were found in RCC-MSI-H. Especially forRNF43, the frequent and validated mutation target in MSI-Hcancers (21, 24), was the most significantly mutated gene inRCC-MSI-H (support: 0.76). Nevertheless, an exclusivity of theMSI-H status and canonical colorectal cancer driver mutationgenes, such asAPC (support: 0.24 vs. 0.73 inMSS),TP53 (support:0.24 vs. 0.53 in MSS), and KRAS (support: 0.28 vs. 0.45 in MSS),were observed. This finding may account for a better MSI-Hpatient survival than MSS patients in colorectal cancer. Collec-tively, our data revealed that LCC and RCC had distinct driver

patterns that should be considered in future targeted cancertherapy.

Hypermethylation in RCCChanges in the DNA methylation status are thought to be

involved in colorectal cancer initiation and development. There-fore,we attempted to identify theDMRs to decipher the epigeneticdifference between LCC and RCC. Notably, 552 DMRs (bumpssize � 2, cutoff ¼ 0.2, permutation test P < 0.01, SupplementaryTable S2) were determined using bumphunter. Among them, 499DMRs (92%)were locatedwithinCpG island, and 27DMRs (5%)

1

2

3

5

6

7

8

9

10

11

12

13

14

15

16

17

18

1920

21

22

X

MM

EL1

LIN

C01

141

RU

NX

3D

LGAP

3R

NF2

20TA

L1PO

DN

PRKA

A2D

NAJ

C6

IL12

RB2

LPH

N2

NTN

G1

ANKR

D34

ASH

ELR

RC71

CD1D

TMEM

178A

SIX3

AC09

3702

.1KC

NK12

MEI

S1−A

S3MEI

S1

AC00

7392

.3CYP26

B1EMX1

NOTO

FOXI3

RFX8

POU3F3

RP11−13J10.1

AC104655.3

ST6GAL2

CAND2

FBLN2

MYL3

PTH1R

SLC38A3

BOC

ADCY5

ESYT3

FOXL2

C3orf72

CLSTN2

RP11−372E1.6

MED12L

IDUA

ZFYVE28CRMP1

JAKMIP1RP11−586D19.1

NKX3−2KIAA1239

FIP1L1EPHA5

RP11−807H7.1ADAMTS3

TRIM2CTD−2231H16.1UBE2QL1ADCY2C5orf49CTNND2NPR3

CTD−2353F22.1GHRRNF180MAP1BPDE8BEDIL3HIST1H2BKHIST1H4IHCG17HLA−LTNXB

HLA−DPB2GRM4

KCNK17

HMGCLL1

COL19A1

B3GAT2

AC093627.12

FAM20C

DFNA5

ELMO1

RP11−797H7.5

ZNF273

RP5−942I16.1

ZNF804B

SGCE

DYNC

1I1

CTD

−2135J3.3

SOX7

XKR6

GFR

A2

RP11−175E9.1

RP11−213G

6.2PU

RG

RP11−1134I14.2

RG

S20

DM

RT3

CDL

GELA

VL2

TRP

M3

LIN

C00

092

SVEP

1PA

PPA

RP1

1−12

3K19

.1

NTN

G2

GAT

A3

PTF1

A

ZEB1

−AS1

ZEB1

CXC

L12

GDF

10

ZNF3

65

EGR2

GRI

D1IGF2IN

S−IG

F2

DKK3CAL

CB

MPPED

2

RP5−10

24C24

.1

WT1W

T1−AS

C11orf

96

PIANPRERGPTPRO

C12orf39

ST8SIA1PRICKLE1

NELL2COL2A1WASF3CDX2FREM2ENOX1PCDH20RNF219−AS1POU4F1

SLC22A17NOVA1−AS1RP11−964E11.2PAX9SIX1PAPLNZDHHC22

GOLGA8MRP11−578F21.6WHAMMP2GLDNPYGO1PRTG

LMF1HS3ST6

CDIP1

RP11−420N3.2

HS3ST4

STX1B

ATP1B2

EFNB3

MYH10

RP11−214O1.1

RP11−1113L8.1

AQP4−AS1

TCF4

AMH

JSRP1

LINGO3RSPO4

OXTHSPA12B

RASSF2AP000282.2

CBSCO

L6A2CECR2

SCARF2KLH

L22

KAL1

LINC

00632

107502500 107503000 107503500 107504000

0.0

0.2

0.4

0.6

0.8

1.0

genomic location Chr 2

Met

hyla

tion

LCCRCC

Figure 1.

Identification of theDMRsbetween LCC andRCC. Notably, 540 hypermethylated regionswere found in RCC, and only 12 hypermethylated regionswere found in LCC.The innermost circle corresponds to theRCCmethylation signature (depicted by red scatter plots), and the outer circle corresponds to the LCCmethylation signature(depicted by blue scatter plots). The scatter location in each track is corresponding to themethylation level (0–1). Selected DMR-related genes are present inside thehg19 human chromosome ideogram. An example of RCC hypermethylated regions relative to LCC were shown in the center of the circle.

Hu et al.

Mol Cancer Res; 16(3) March 2018 Molecular Cancer Research478

on August 18, 2020. © 2018 American Association for Cancer Research. mcr.aacrjournals.org Downloaded from

Published OnlineFirst November 29, 2017; DOI: 10.1158/1541-7786.MCR-17-0483

were in the shores. Notably, a hypermethylation profile(540DMRs) was observed in RCC (Fig. 1). Of the 12 hypermethy-lated DMRs (34 probes) in LCC, 3 DMRs (13 probes) wereassociated with HOX family members HOXB5, HOXB7, andCDX2. GO enrichment analysis showed that the DMR-associatedgenesweremainly enriched in organmorphogenesis (Benjamini–Hochberg FDR correction 3.1798E–10, 42 genes), cell differenti-ation (6.8791E–9, 71 genes), and embryonic morphogenesis(2.0480E–9, 29 genes).

Activation of oncogenes and silence of tumor-suppressorgenes in RCC

Using mRNAseq data compiled in TCGA, LCC versus RCCrevealed 253 DEGs (fold change > 2 and adjusted P value< 0.05) between them. In LCC, 105 genes were upregulated, and148 geneswere downregulated, comparedwith the RCC (Fig. 2A).Functional annotation of these DEGs uncovered that oncogenes(e.g., ZIC5) were overexpressed in RCC, whereas tumor-suppres-sor genes (e.g., HOXB13) were repressed. To further confirm theDEGs identified using TCGAdata, theGSE14333 dataset with 122LCCs and 126 RCCs was used for identifying DEGs. In total, 330DEGs were identified (Fig. 2B), with 36 of them overlapping withthe TCGA result (Fig. 2C). The Prostate Cancer Susceptibility

Candidate (PRAC)was themost significantly downregulated geneinRCC (Fig. 2D andE). To validate this result, 15 LCCand32RCCFFPE samples were used for IHC staining. Results showed thatboth the IHC staining score and positive rate in LCC weresignificantly higher than in RCCs (Supplementary Fig. S1). Thefunction of PRAC in COAD remains elusive; however, anotherhighly downregulated gene in RCC,HOXB13, approximately 2 kbdownstream of PRAC, has been shown to be a colorectal cancersuppressor (25, 26). AndHOXC6, which also belongs to homeo-box transcription factor (TF) family, was the most significantlyupregulated gene in RCC (Fig. 2F and G).

Further correlation analysis of the DNA methylation andexpression showed that 13 downregulated genes (CLSTN2,ECEL1, ELAVL2,GDF10, LMX1A, PRAC, PRKAA2, RSPO4, SNX32,ST6GAL2, TAC1, VENTX, and WASF3) in RCC were possiblyrepressed by hypermethylation of DMR. For example, oneDMR, situated in the CpG island shore, was strongly associatedwith PRAC repression in RCC. The hypermethylated DMR in RCCencompassed 8 probes (cg00960395, cg04574034, cg09866983,cg12374721, cg14230397, cg20094830, cg20945566, andcg27170782) that cover 388 bps in chromosome 17 (Fig. 3).DNAmethylation has been shown to play a role in the regulationof PRAC expression in prostate cancer (27).

0.0

2.5

5.0

7.5

10.0

12.5

Left Right

Val

ue

0

5

10

Left Right

Log 2

(Val

ue +

1)

TCGA PRAC Expression GSE14333 PRAC Expression

GNG4

HOXB13

ELAVL2

HOXC6

PRAC

DRD5

0

10

20

30

−5 50

Log2 fold change

−Log

10 P

ChangeDOWNNOTUP

TCGA LCC vs. RCCThe number of up gene is 135

The number of down gene is 118

PRAC

HOXC6

0

10

20

30

−4 −2 20

Log2 fold change

−Log

10 P

ChangeDOWNNOTUP

GSE14333 LCC vs. RCCThe number of up gene is 97

The number of down gene is 233

3

6

9

Left Right

Val

ue

GSE14333 HOXC6 Expression

0.0

2.5

5.0

7.5

10.0

Left Right

Log 2

(val

ue +

1)

TCGA HOXC6 Expression

217 20136

TCGAGSE14333

APCDD1 ELAVL2 B3GNT7 INSL5PRAC LY6G6D WIF1 ZIC5XPNPEP2 LYZ ERP27 PNPLA3SLC19A3 PCP4 HOXC6 GNG4PRKAA2 PNMA2 ZNF880 DUSP4SEMG1 TCN1 WASF3 FOXD1RAB27B CLDN8 TRNP1 ZNF813MUC12 ZIC2 CLDN18 ST6GAL2GPR126 PYY HOXB13 NR1H4

BA C

ED GF

Figure 2.

Identification of the DEGs between LCC and RCC. A, Volcano plot revealed that 135 genes were overexpressed in RCC, and 118 genes were underexpressed inRCC using TCGA data. B, Volcano plot uncovered 97 genes were overexpressed in RCC, and 233 genes were underexpressed in RCC using publicly availableGSE14333 data. C, Venn diagram revealed that 36 DEGs were overlapped between TCGA and GSE14333. For D to G, boxplot distribution of PRAC andHOXC6 expression level by using TCGA (D and F) and GSE14333 (E and G) data.

Left- and Right-Sided Colon Cancer Comparison

www.aacrjournals.org Mol Cancer Res; 16(3) March 2018 479

on August 18, 2020. © 2018 American Association for Cancer Research. mcr.aacrjournals.org Downloaded from

Published OnlineFirst November 29, 2017; DOI: 10.1158/1541-7786.MCR-17-0483

miRNA expressionIn addition to mRNA, we also want to know if expression

discrepancies exist between LCC and RCC in noncoding RNAs.Interestingly, 15 cancer-related miRNAs were differentiallyexpressed (adjusted P value < 0.05 and fold change > 1.5)between LCC and RCC. Eight miRNAs, namely, miR-196b,miR-466, miR-296, miR-552, miR-592, miR-1247, miR-1275,and miR-3131, were downregulated in RCC, whereas sevenmiRNAs were upregulated, including miR-10b, miR-31, miR-146a, miR-155, miR-615, miR-625, and miR-1293 (Fig. 4A).The abundantly expressed mir-10b, an miRNA closely associ-ated with cancer metastasis, had significantly higher expressionlevel in RCC (Fig. 4B).

Further integrated analysis ofmiRNAandDEGs showed that 16DEGs were putatively regulated by seven miRNAs (Fig. 4C). TwomiRNAs, miR-625 and miR-146a, were highly expressed in RCC,and their predicted targets were all downregulated in RCC. As forthe other low expression of miRNAs in RCC, their predictedtargets were upregulated in RCC. For example, miR-466 wasdownregulated, and its predicted target, ZIC5, a transcriptionalrepressor that elevated expression contributed to cancer progres-sion, was upregulated in RCC.

RCCs tend to bemore aggressive throughorchestrating invasivegene modules

Considering the huge expression difference between LCC andRCC, we speculate that the regulatory network also differs greatly.To this end, we used WGCNA to explore the core gene regulatorymodules, which may shed light on the intrinsic expression var-iation between LCC and RCC. Interestingly, six and ninemodulesconsisting of at least 100 genes were found in LCC and RCC,respectively. Five modules related with mRNAmetabolic process,DNA ligation, regulation of ARF GTPase activity, translationalelongation, vasculature development, cell adhesion, complementactivation, T-cell activation, and lymphocyte activation wereshared by LCC and RCC (Fig. 5, linked by colored band). Ofnote, because certain modules such as brown- and blue modulesin LCC were rather large. Therefore, the top 50 hub genes wereextracted in eachmodule.We found that nomiRNAs ranked in thetop hubs in the module. This observation raises the possibilitythat miRNA, unlike TF, has a restricted role or at least does not actas a key orchestrator in COAD. The module peculiar to LCC was

linked to cellular nitrogen compound metabolic process (FDR-corrected P value 3.3739E–2, 13 genes). The rest of the four RCC-relevant modules were linked with cell-cycle phase (magentamodule, 5.4696E–28, 27 genes), mitosis (magenta module,2.9294E–29, 24 genes), generation of precursor metabolites andenergy (brown module, 1.0886E–4, nine genes), microtubule-based process (purple module, 3.1047E–2, five genes), responseto virus (7.5314E–8, nine genes), and innate immune response(salmon module, 4.4101E–2, four genes). As such, enhancedactivation of cell division, energy metabolism, and immunesystem process in RCC underscores the speculation of its moreaggressive state.

Pathway alteration difference between LCC and RCCIntegrated comparison of mutations and mRNA expression

alteration in 85 LCCs and 116 RCC-MSS will enable betterunderstanding of the intrinsic difference in the pathway dysre-gulation. Alteration frequency between LCC and RCC-MSS wasexamined in five well-known pathways (WNT, RAS, PI3K, TGF-beta, and P53) that are closely associated with colorectal cancer.Basically, the mutation rate in RCC-MSS is much more frequentthan LCC in TGF-beta, PI3K, and RAS signaling pathways, butalmost the same in WNT and TP53 pathways (Fig. 6). The WNTsignaling pathway was altered in 89% LCC and 85% RCC-MSS,respectively, which mainly contributed by inactivation of APC(78% and 73%), followed by the mutation of FAM123B,FBXW7, SOX9, CTNNB1, AXIN2, and LRP1B. Downregulationof LRP1B in colon cancer has recently been shown to beassociated with promoting growth and migration of cancercells (28). In the TGF-beta pathway, higher mutation rates(LCC vs. RCC-MSS, SMAD2, 0% vs. 3%; SMAD3, 1% vs. 4%;SMAD4, 12% vs. 16%) in RCC-MSS were observed in SMADfamily members. As for the PI3K pathway, PIK3CAwas mutatedin 11% LCC samples and 26% RCC-MSS samples. Resemblingthe case in PI3K, a much higher mutation rate of KRAS (30% vs.45%) and BRAF (7% vs. 11%) was found in RCC-MSS (Fig. 6).Activation of the PI3K and RAS pathways may be more prone toprogression for RCC-MSS.

To further explore the co-occurrence of the dysregulation ofdifferent pathways in LCC and RCC-MSS, we re-examined theFMGSs that were identified as mentioned earlier. The resultsshowed that five FMGSs (k > 1, support > 0.1) were in common

0.0

2.5

5.0

7.5

10.0

12.5

0.00 0.25 0.50 0.75 1.00

DNA Methylation (cg12374721)

PR

AC E

xpre

ssio

n

LeftRight

0.0

2.5

5.0

7.5

10.0

12.5

0.00 0.25 0.50 0.75 1.00

DNA Methylation (cg04574034)

PR

AC E

xpre

ssio

n

0.0

2.5

5.0

7.5

10.0

12.5

0.00 0.25 0.50 0.75 1.00

DNA Methylation (cg09866983)

PR

AC E

xpre

ssio

n

0.0

2.5

5.0

7.5

10.0

12.5

0.00 0.25 0.50 0.75 1.00

DNA Methylation (cg20094830)

PR

AC E

xpre

ssio

n

0.0

2.5

5.0

7.5

10.0

12.5

0.00 0.25 0.50 0.75 1.00

DNA Methylation (cg00960395)

PR

AC E

xpre

ssio

n

0.0

2.5

5.0

7.5

10.0

12.5

0.00 0.25 0.50 0.75 1.00

DNA Methylation (cg14230397)

PR

AC E

xpre

ssio

n

0.0

2.5

5.0

7.5

10.0

12.5

0.00 0.25 0.50 0.75 1.00

DNA Methylation (cg20945566)

PR

AC E

xpre

ssio

n

0.0

2.5

5.0

7.5

10.0

12.5

0.00 0.25 0.50 0.75 1.00

DNA Methylation (cg27170782)

PR

AC E

xpre

ssio

n

Figure 3.

Downregulation of PRAC correlates with high DNA methylation in RCC. Black dots and red dots correspond to LCC and RCC, respectively.

Hu et al.

Mol Cancer Res; 16(3) March 2018 Molecular Cancer Research480

on August 18, 2020. © 2018 American Association for Cancer Research. mcr.aacrjournals.org Downloaded from

Published OnlineFirst November 29, 2017; DOI: 10.1158/1541-7786.MCR-17-0483

between LCC and RCC-MSS. Inactivation of WNT and P53, butactivation of the RAS pathway, occurred simultaneously in morethan 15% of the COAD samples (both LCC and RCC-MSS).However, we observed nine more examples of RCC-MSS–specificpathway cross-talk that have >10% support (Fig. 6, right bottomplot). In addition to the interaction with the WNT pathway, thePI3K pathway often crosstalks with the RAS and P53 pathways inRCC-MSS. In addition, in RCC-MSS, the TGF-beta pathway alsooften interacted with the RAS andWNT pathways via SMAD4 andKRAS or APC, respectively. Of note, we also observed singlepathway members that were inactivated simultaneously, repre-sented byAPCandSOX9 (support¼0.121) or FBXW7 (support¼0.103) in the WNT pathway. Collectively, we found numerousmolecular signatures that were associated with tumor aggres-siveness in RCC-MSS, which suggests a poor prognosis forRCC-MSS patients.

DiscussionThe large molecular difference between LCC and RCC iden-

tified in this study consolidates the different embryonic originsof LCC and RCC. And the difference is independent of sex(X2 test, P ¼ 1). Actually, DNA methylation alteration is anearly event that occurs in cancer. Koestler and colleagues (29)determined that 168 probes enriched in homeobox genes thatwere differentially methylated between the right- and left-colonadenomas using the Illumina HumanMethylation450 Bead-Chip. Consistent with their observation of PRAC hypermethy-lation in right-colon adenomas and hypermethylation ofCDX2 in left-colon adenomas, the difference also held true inthe carcinoma.

As a TF, HOXC6 promotes prostate cancer metastasisvia inducing aberrant metastatic miRNA expression (30). It

LCC RCC

050

,000

100,

000

150,

000

200,

000

hsa−miR−10b

Rea

ds p

er m

illio

n

A B

hsa-miR-10b

hsa-miR-1247

hsa-miR-146a

hsa-miR-155

hsa-miR-196b hsa-miR-625hsa-miR-552

hsa-miR-592

hsa-miR-31hsa-miR-296

hsa-miR-3131

hsa-miR-615

hsa-miR-1275

hsa-miR-466

hsa-miR-1293

0

3

6

9

12

−1 0 1 2 3

Log2 Fold change

−Log

10 P

ChangeDOWNNOTUP

Cutoff for log2FC is 0.585The number of up miRNA is 7

The number of down miRNA is 8

SCN5A

ACSL6

FOXD1

MIA2

KRT23

hsa-miR-625

AMT

CRTAM

CLSTN2

hsa-miR-552

PLA2G3 PAX9

hsa-miR-146a

ZIC5

WASF3

hsa-miR-466hsa-miR-592

DAPK1

hsa-miR-196b

GPR126

HOXC8

hsa-miR-1275

C

Figure 4.

Identification of differentially expressed miRNAs between LCC and RCC. A, Volcano plot revealed that 11 miRNAs were overexpressed in RCC, and 5 miRNAswere underexpressed in RCCusing TCGAdata.B,Comparison of themiR-10b expression level between LCC andRCCbyboxplot.C, Integrated analysis ofmiRNA andDEGs revealed that some DEGs were putatively regulated by miRNAs. Yellow elliptical nodes represent miRNAs, and magenta rectangle nodes representmiRNA targets.

Left- and Right-Sided Colon Cancer Comparison

www.aacrjournals.org Mol Cancer Res; 16(3) March 2018 481

on August 18, 2020. © 2018 American Association for Cancer Research. mcr.aacrjournals.org Downloaded from

Published OnlineFirst November 29, 2017; DOI: 10.1158/1541-7786.MCR-17-0483

is tempting to believe that this result held in RCC due to theelevated expression of HOXC6. Previous studies demonstratethat the overexpression mir-10b can promote tumor invasion(31, 32) and confers chemoresistance in colorectal cancer cellsto 5-fluorouracil (33). In this context, 5-fluorouracil may not besuitable for the treatment of RCC due to a much higherexpression level of miR-10b in RCC than LCC. In addition,high miR-10b expression in primary colorectal cancer tissue canindependently predict distant metastasis (34), which suggeststhat primary RCC is more vulnerable to metastasis. Other onco-miRNAs (miR-31, miR-155, miR-625) that were overexpressedin RCC were all closely associated with colorectal carcinomacell proliferation, migration, and invasion (35–37). In contrast,two tumor-suppressor miRNAs in colorectal cancer, miR-296and miR-592, were underexpressed in RCC. MiR-296 inhibitsthe metastasis and epithelial–mesenchymal transition by tar-geting S100A4 and miR-592, which inhibits cell proliferationby suppressing of CCND3 expression (38, 39).

It must be noted that a more aggressive status was observedin RCC through determining the modules consisting of tightlycoregulated genes. The large brown module unique to RCCharbors 1,353 genes that were mainly involved in electrontransport chain. In addition, three other RCC-specific moduleswere enriched in cell-cycle regulating, mitosis/nuclear division,microtubule-based process, and innate immune response,which were all linked to tumor aggressiveness. In parallel, asdescribed earlier, miRNAs tend not to be hub genes with anextremely high connectivity. This finding can be ascribed tothe intrinsic target recognition discrepancy between plant andanimal miRNAs (40). In plants, almost all miRNA targets havea single miRNA-responsive element and are regulated by justone miRNA or one miRNA family through nearly perfectlycomplementary miRNA-target pairs. Thus, the scope of miRNAtargets appears to be only a handful of key targets, most of

which are TFs. In animals, miRNAs target transcripts viaimperfect base-pairing to multiple sites in the 30 untranslatedregions by the seed sequence (50 end of miRNAs that com-prises nucleotides 2–7). For this reason, over 90% transcriptswere putatively targeted by miRNAs, and one transcript wasoften under combinatorial control of multiple miRNAs (41,42). In this context, animal miRNAs generally act as fine-tuners that exert wide subtle regulatory effects across thetranscriptome and tend not to be hub genes in the regulatorynetwork.

PRAC, a small protein with a molecular weight of only 6 kDa,has been shown to be specifically expressed in the human pros-tate, rectum, and colon (27, 43). In prostate cancer, the expressionlevel of PRAC is significantly downregulated compared withbenign prostatic hyperplasia. Here, we demonstrated that itsexpression level is significantly downregulated in RCC, which isin line with the findings by Bauer and colleagues (44). Althoughthe concrete function of PRAC remains elusive in COAD, apossible cotranscribed gene HOXB13 has been shown to beinvolved in suppressing colorectal cancer progression (25, 45).In RCC, HOXB13 was significantly downregulated, as well. Thus,PRAC may serve as a tumor suppressor through interacting withHOXB13 in COAD.

In addition, RCC-MSS hadmore mutated BRAF (11% vs. 7%)and KRAS (45% vs. 30%), and similar results were observed inour Zhejiang University Cancer Institute–collected samplesthrough sequencing 338 pairs of matched fresh frozen colo-rectal cancer tissue and adjacent normal tissue samples (unpub-lished data). BRAF mutation has been shown to be associatedwith a worse prognosis (46) and a poorer response to cetux-imab therapy (3, 47). In addition, colon cancer patients withwild-type KRAS treated with cetuximab have significantlyimproved overall survival (OS; ref. 47). Thus, a worse prognosisin RCC can be partly ascribed to a higher mutation rate of KRAS

0.00.40.8

LCC

Height

Dyn

amic

tree

cut

0.0 0.4 0.8

RCC

Height

Dyn

amic

tree

cut

Cellular nitrogen compound Metabolic process

mRNA Metabolic process

Regulation of ARF GTPase activityTranslational elongation

DNA Ligation

Cell adhesion

Vasculature developmentComplement activationT-cell activationLymphocyte activation

Generation of precursor

Metabolites and energy

Cell-cycle phase

Mitosis

Microtubule-based process

Response to virusInnate immune response

Cell adhesionVasculature development

Complement activation

T-Ce

ll act

ivatin

g

Lym

phoc

yte

activ

atin

g

Figure 5.

Comparison of the regulatory network between LCC and RCC using WGCNA. Hierarchical cluster analysis dendrogram used to detect coexpression clustersfrom the dataset with 230 LCCs and 220 RCCs along with corresponding color assignments. Genes that were not coexpressed were assigned tothe gray group. Each vertical line corresponds to a gene, and branches are expression modules of the highly interconnected groups of genes. In total, six andnine modules ranging from 117 to 4,118 genes in size were identified, respectively. Functional annotation of each module is based on the GO enrichment analysisof the top 50 hub genes. Modules with similar functional annotation in LCC and RCC were linked by colored bands. In addition, two modules were selectedfor visualization of the subnetwork connections among the most connected genes. The plots show network connections whose topological overlap isabove the threshold of 0.10.

Hu et al.

Mol Cancer Res; 16(3) March 2018 Molecular Cancer Research482

on August 18, 2020. © 2018 American Association for Cancer Research. mcr.aacrjournals.org Downloaded from

Published OnlineFirst November 29, 2017; DOI: 10.1158/1541-7786.MCR-17-0483

and BRAF (Fig. 6). More recently, Ptashkin and colleagues (48)find that chromosome 20q amplification is associated with abetter OS in LCC. In addition, because cetuximab is mainlyused in advanced stages of colorectal cancer patients, FMGSmining is further conducted in four pathologic stages accordingto the American Joint Committee on Cancer standards. Nota-bly, we found KRAS harbored great stage heterogeneity betweenLCC (stage I: 25%, II: 53%, III: 9%, and IV: 20%) and RCC(53%, 46%, 31%, and 62%). Nevertheless, we should bear inmind that current targeted drugs used in treating colon cancerpatients are rather limited and that the prognosis may bereversed if the targeted drug is based on mutated KRAS orBRAF. Meanwhile, OS and progression-free survival (PFS) inRCC can benefit from combination therapy since frequentpathway cross-talk occurred in RCC.

Recently, Le and colleagues (49) demonstrated that PD-1blockade works well in MMR-deficient colorectal cancer, andprolonged PFS was associated with high somatic mutationalloads. In this manner, RCC patients that are deficient in MMR,approximately 20% of RCC, can benefit from PD1 therapy (23,50). In short, the underlyingmolecular feature present in LCC andRCC identified in this study is beneficial for medical decision-making and prolonged OS and PFS. Our results help to elucidatethe molecular basis of treating LCC and RCC as two differentdiseases in future cancer treatment.

Disclosure of Potential Conflicts of InterestNo potential conflicts of interest were disclosed.

Authors' ContributionsConception and design: W. Hu, W. Ge, S. Zhang, S. ZhengAnalysis and interpretation of data (e.g., statistical analysis, biostatistics,computational analysis): W. Hu, Y. Yang, X. Li, M. HuangWriting, review, and/or revision of the manuscript: W. Hu, X. Li, F. XuAdministrative, technical, or material support (i.e., reporting or organizingdata, constructing databases): M. Huang, W. Ge, S. Zhang, S. ZhengStudy supervision: W. Ge, S. Zhang, S. Zheng

AcknowledgmentsThis work was supported by the National High Technology Research and

Development Program of China (863 Program, 2012AA02A204) to W. Ge.This work was also partially supported by China Postdoctoral ScienceFoundation (2016M590532) and Postdoctoral Foundation of ZhejiangProvince to W. Hu (BSH1502129).

The authors thank Dr. Lizhen Zhu for her help in IHC staining.

The costs of publication of this article were defrayed in part by thepayment of page charges. This article must therefore be hereby markedadvertisement in accordance with 18 U.S.C. Section 1734 solely to indicatethis fact.

Received August 31, 2017; revised October 26, 2017; accepted November 27,2017; published OnlineFirst November 29, 2017.

Proliferation, stem/

progenitor phenotype

Nucleus

APC78% 73%

FZD10

1% <1%

FBXW712% 13%

FAM123B5% 17%

TCF7L29% 6%

ARID1A6% 12%

AXIN21% 3%

SOX98% 13%

CTNNB14% 9%

CTNNB14% 9% MYC

LRP5

WNT

WNT Signaling

Activated InactivatedGene mutation

frequency

RCC-MSS

Upregulated

TGFβ Signaling

TGFβ Activin

TGFBR1

2% 4%

TGFBR2

1% 3%

ACVR2A

1% 6%

ACVR1B

2% 2%

SMAD20% 3%

SMAD31% 4%

SMAD412% 16%

Proliferation, cell survival

IGF1R

IRS21% 4%

ERBB2

5% 3%

ERBB3

2% 3%

PIK3CA11% 26%PIK3R1

5% 3%

PTEN6% 6%

NRAS4% 5%

KRAS30% 45%

BRAF7% 11%

Proliferation

Cell survival

ATM6% 12%

TP5366% %53%

PI3K Signaling

P53 Signaling

Activation

Inhibition

1% <1%

DKK1-41% 4%

89% 85% 20% 36% 20% 32%

70% 64%

RAS Signaling 40% 61%LC

C

RC

C-M

SS

LC

C

RC

C-M

SS

LC

C

RC

C-M

SS

LC

C

RC

C-M

SS

LC

C

RC

C-M

SS

SOS1/24%4%

DVL20% <1%

LRP1B5% 19%

RNF43

5% 6%

LCCDownregulated

> 20%

10–20%

5–10%

1–5%

< 1%

TP53 24%

APC 24%

KRAS 27%

RCC-MSI-H

APC, KRAS 0.271 TP53, KRAS PIK3CA, APC APC, TP53 APC, TP53, KRAS

PIK3CA, TP53

APC, KRAS

SMAD4, KRAS

SMAD4, APC

APC, PIK3CA

APC, SOX9

PIK3CA, KRAS

TP53, KRAS

APC, TP53

APC, FBXW7

APC, TP53, KRAS

PIK3CA, APC, KRAS PIK3CA, TP53, APC PIK3CA, TP53, KRAS

0.200

0.5290.129

0.165

0.3880.2330.2070.3790.1980.1210.1030.1810.1380.1030.1120.1470.1120.103

LCC RCC-MSS

Figure 6.

Comparison of the genetic change frequency leading to signaling pathway alteration between LCC and RCC. Alteration frequencies are defined as the percentage ofLCC (n ¼ 85) or RCC (n ¼ 116) samples. Mutation frequency of TP53, APC, and KRAS in MSI samples (n ¼ 29) of RCC are shown separately in the right plot.Red denotes the predicted activated genes, and blue represents the inactivated genes. Right bottom plot demonstrates the pathway cross-talk difference betweenLCC and RCC-MSS.

Left- and Right-Sided Colon Cancer Comparison

www.aacrjournals.org Mol Cancer Res; 16(3) March 2018 483

on August 18, 2020. © 2018 American Association for Cancer Research. mcr.aacrjournals.org Downloaded from

Published OnlineFirst November 29, 2017; DOI: 10.1158/1541-7786.MCR-17-0483

References1. Torre LA, Bray F, Siegel RL, Ferlay J, Lortet-Tieulent J, Jemal A. Global cancer

statistics, 2012. CA Cancer J Clin 2015;65:87–108.2. Nitsche U, Stogbauer F, Spath C, Haller B, Wilhelm D, Friess H, et al. Right

sided colon cancer as a distinct histopathological subtype with reducedprognosis. Dig Surg 2016;33:157–63.

3. Brule SY, JonkerDJ, Karapetis CS,O'CallaghanCJ,MooreMJ,Wong R, et al.Location of colon cancer (right-sided versus left-sided) as a prognosticfactor and a predictor of benefit from cetuximab in NCIC CO.17. Eur JCancer 2015;51:1405–14.

4. Hansen IO, Jess P. Possible better long-term survival in left versusright-sided colon cancer - a systematic review. Dan Med J 2012;59:A4444.

5. Benedix F, Kube R,Meyer F, SchmidtU, Gastinger I, Lippert H. Comparisonof 17,641 patients with right- and left-sided colon cancer: differences inepidemiology, perioperative course, histology, and survival. Dis ColonRectum 2010;53:57–64.

6. Weiss JM, Pfau PR, O'Connor ES, King J, LoConte N, Kennedy G, et al.Mortality by stage for right- versus left-sided colon cancer: analysis ofsurveillance, epidemiology, and end results–Medicare data. J Clin Oncol2011;29:4401–9.

7. Meguid RA, Slidell MB, Wolfgang CL, Chang DC, Ahuja N. Is there adifference in survival between right- versus left-sided colon cancers? AnnSurg Oncol 2008;15:2388–94.

8. Petrelli F, Tomasello G, Borgonovo K, Ghidini M, Turati L, Dallera P, et al.Prognostic survival associated with left-sided vs right-sided colon cancer: asystematic review and meta-analysis. JAMA Oncol 2016 Oct 27. [Epubahead of print].

9. Hu W, Li X, Wang T, Zheng S. Association mining of mutated cancer genesin different clinical stages across 11 cancer types. Oncotarget 2016;7:68270–7.

10. Ji Y, Ying H, Tran J, Dews P, Massanari RM. Integrating unified medicallanguage system and association mining techniques into relevancefeedback for biomedical literature search. BMC Bioinformatics 2016;17:264.

11. Zhang H, Meltzer P, Davis S. RCircos: an R package for Circos 2D trackplots. BMC Bioinformatics 2013;14:244.

12. Hu W, Yang Y, Li X, Zheng S. Pan-organ transcriptome variation across 21cancer types. Oncotarget 2017;8:6809–18.

13. Betel D, Wilson M, Gabow A, Marks DS, Sander C. The microRNA.org resource: targets and expression. Nucleic Acids Res 2008;36:D149–53.

14. Agarwal V, Bell GW, Nam JW, Bartel DP. Predicting effective microRNAtarget sites in mammalian mRNAs. Elife 2015;4.

15. Langfelder P, Horvath S. WGCNA: an R package for weighted correlationnetwork analysis. BMC Bioinformatics 2008;9:559.

16. Hu Z, Chang YC, Wang Y, Huang CL, Liu Y, Tian F, et al. VisANT 4.0:Integrative network platform to connect genes, drugs, diseases and ther-apies. Nucleic Acids Res 2013;41:W225–31.

17. Maere S, Heymans K, Kuiper M. BiNGO: a Cytoscape plugin to assessoverrepresentation of gene ontology categories in biological networks.Bioinformatics 2005;21:3448–9.

18. Shannon P, Markiel A, Ozier O, Baliga NS, Wang JT, Ramage D, et al.Cytoscape: a software environment for integrated models of biomolecularinteraction networks. Genome Res 2003;13:2498–504.

19. Li X, Hu W, Zhou J, Huang Y, Peng J, Yuan Y, et al. CLCA1 suppressescolorectal cancer aggressiveness via inhibition of the Wnt/beta-cateninsignaling pathway. Cell Commun Signal 2017;15:38.

20. Cortes-Ciriano I, Lee S, Park WY, Kim TM, Park PJ. A molecular portrait ofmicrosatellite instability across multiple cancers. Nat Commun 2017;8:15180.

21. Hause RJ, Pritchard CC, Shendure J, Salipante SJ. Classification andcharacterization of microsatellite instability across 18 cancer types. NatMed 2016;22:1342–50.

22. Vilar E, Gruber SB. Microsatellite instability in colorectal cancer-the stableevidence. Nat Rev Clin Oncol 2010;7:153–62.

23. Comprehensive molecular characterization of human colon and rectalcancer. Nature 2012;487:330–7.

24. Giannakis M, Hodis E, JasmineMu X, Yamauchi M, Rosenbluh J, CibulskisK, et al. RNF43 is frequentlymutated in colorectal and endometrial cancers.Nat Genet 2014;46:1264–6.

25. Jung C, Kim RS, Zhang H, Lee SJ, Sheng H, Loehrer PJ, et al. HOXB13 isdownregulated in colorectal cancer to confer TCF4-mediated transactiva-tion. Br J Cancer 2005;92:2233–9.

26. AkbariMR, Anderson LN, BuchananDD,ClendenningM, JenkinsMA,WinAK, et al. Germline HOXB13 p.Gly84Glu mutation and risk of colorectalcancer. Cancer Epidemiol 2013;37:424–7.

27. Lenka G, Weng WH, Chuang CK, Ng KF, Pang ST. Aberrantexpression of the PRAC gene in prostate cancer. Int J Oncol 2013;43:1960–6.

28. Wang Z, SunP,GaoC,Chen J, Li J, ChenZ, et al.Down-regulation of LRP1Bin colon cancer promoted the growth and migration of cancer cells. ExpCell Res 2017;357:1–8.

29. Koestler DC, Li J, Baron JA, Tsongalis GJ, Butterly LF, GoodrichM, et al. Distinct patterns of DNA methylation in conventionaladenomas involving the right and left colon. Mod Pathol 2014;27:145–55.

30. Xue M, Liu H, Zhang L, Chang H, Liu Y, Du S, et al. Computationalidentification of mutually exclusive transcriptional drivers dysregulat-ing metastatic microRNAs in prostate cancer. Nat Commun 2017;8:14917.

31. Ma L. Role of miR-10b in breast cancer metastasis. Breast Cancer Res2010;12:210.

32. Jiang H, Liu J, Chen Y, Ma C, Li B, Hao T. Up-regulation of mir-10bpredicate advanced clinicopathological features and liver metastasis incolorectal cancer. Cancer Med 2016;5:2932–41.

33. Nishida N, Yamashita S, Mimori K, Sudo T, Tanaka F, Shibata K, et al.MicroRNA-10b is a prognostic indicator in colorectal cancer and confersresistance to the chemotherapeutic agent 5-fluorouracil in colorectal cancercells. Ann Surg Oncol 2012;19:3065–71.

34. Hur K, Toiyama Y, Schetter AJ, Okugawa Y, Harris CC, Boland CR, et al.Identification of a metastasis-specific MicroRNA signature in humancolorectal cancer. J Natl Cancer Inst 2015;107.

35. Zhang GJ, Xiao HX, Tian HP, Liu ZL, Xia SS, Zhou T. Upregulation ofmicroRNA-155 promotes the migration and invasion of colorectal cancercells through the regulation of claudin-1 expression. Int J Mol Med2013;31:1375–80.

36. ZhengH,MaR,WangQ, ZhangP, LiD,Wang J, et al.MiR-625-3ppromotescell migration and invasion via inhibition of SCAI in colorectal carcinomacells. Oncotarget 2015;6:27805–15.

37. Li T, LuoW, LiuK, Lv X, Xi T.miR-31promotes proliferationof colon cancercells by targeting E2F2. Biotechnol Lett 2015;37:523–32.

38. Liu Z, Wu R, Li G, Sun P, Xu Q. MiR-592 inhibited cell proliferation ofhuman colorectal cancer cells by suppressing of CCND3 expression. Int JClin Exp Med 2015;8:3490–7.

39. He Z, Yu L, Luo S, Li M, Li J, Li Q, et al. miR-296 inhibits themetastasis andepithelial-mesenchymal transition of colorectal cancer by targetingS100A4. BMC Cancer 2017;17:140.

40. Brodersen P, Voinnet O. Revisiting the principles of microRNAtarget recognition and mode of action. Nat Rev Mol Cell Biol2009;10:141–8.

41. Hu W, Wang T, Yang Y, Zheng S. CPuORF correlates with miRNA respon-sive elements on protein evolutionary rates. Biochem Biophys ResCommun 2014;452:66–71.

42. Axtell MJ, Westholm JO, Lai EC. Vive la difference: biogenesis and evolu-tion of microRNAs in plants and animals. Genome Biol 2011;12:221.

43. Liu XF, Olsson P, Wolfgang CD, Bera TK, Duray P, Lee B, et al. PRAC: Anovel small nuclear protein that is specifically expressed in human prostateand colon. Prostate 2001;47:125–31.

44. Bauer KM, Hummon AB, Buechler S. Right-side and left-side colon cancerfollow different pathways to relapse. Mol Carcinog 2012;51:411–21.

45. Ghoshal K, Motiwala T, Claus R, Yan P, Kutay H, Datta J, et al. HOXB13, atarget ofDNMT3B, ismethylated at an upstreamCpG island, and functionsas a tumor suppressor in primary colorectal tumors. PLoS One 2010;5:e10338.

46. Kalady MF, Dejulius KL, Sanchez JA, Jarrar A, Liu X, Manilich E, et al.BRAF mutations in colorectal cancer are associated with distinctclinical characteristics and worse prognosis. Dis Colon Rectum 2012;55:128–33.

47. De Roock W, Claes B, Bernasconi D, De Schutter J, Biesmans B, FountzilasG, et al. Effects of KRAS, BRAF, NRAS, and PIK3CA mutations on the

Mol Cancer Res; 16(3) March 2018 Molecular Cancer Research484

Hu et al.

on August 18, 2020. © 2018 American Association for Cancer Research. mcr.aacrjournals.org Downloaded from

Published OnlineFirst November 29, 2017; DOI: 10.1158/1541-7786.MCR-17-0483

efficacy of cetuximab plus chemotherapy in chemotherapy-refractory met-astatic colorectal cancer: a retrospective consortium analysis. Lancet Oncol2010;11:753–62.

48. Ptashkin RN, Pagan C, Yaeger R, Middha S, Shia J, O'Rourke KP, et al.Chromosome 20q amplification defines a subtype of microsatellite stable,left-sided colon cancers withwild-type RAS/RAF andbetter overall survival.Mol Cancer Res 2017;15:708–13.

49. Le DT, Uram JN,WangH, Bartlett BR, Kemberling H, Eyring AD, et al. PD-1blockade in tumors with mismatch-repair deficiency. N Engl J Med2015;372:2509–20.

50. Overman MJ, McDermott R, Leach JL, Lonardi S, Lenz HJ, Morse MA, et al.Nivolumab in patients with metastatic DNA mismatch repair-deficient ormicrosatellite instability-high colorectal cancer (CheckMate 142): anopen-label, multicentre, phase 2 study. Lancet Oncol 2017;18:1182–91.

www.aacrjournals.org Mol Cancer Res; 16(3) March 2018 485

Left- and Right-Sided Colon Cancer Comparison

on August 18, 2020. © 2018 American Association for Cancer Research. mcr.aacrjournals.org Downloaded from

Published OnlineFirst November 29, 2017; DOI: 10.1158/1541-7786.MCR-17-0483

2018;16:476-485. Published OnlineFirst November 29, 2017.Mol Cancer Res   Wangxiong Hu, Yanmei Yang, Xiaofen Li, et al.   Right-Sided Colon CancerMulti-omics Approach Reveals Distinct Differences in Left- and

  Updated version

  10.1158/1541-7786.MCR-17-0483doi:

Access the most recent version of this article at:

  Material

Supplementary

  http://mcr.aacrjournals.org/content/suppl/2017/11/29/1541-7786.MCR-17-0483.DC1

Access the most recent supplemental material at:

   

   

  Cited articles

  http://mcr.aacrjournals.org/content/16/3/476.full#ref-list-1

This article cites 47 articles, 3 of which you can access for free at:

  Citing articles

  http://mcr.aacrjournals.org/content/16/3/476.full#related-urls

This article has been cited by 5 HighWire-hosted articles. Access the articles at:

   

  E-mail alerts related to this article or journal.Sign up to receive free email-alerts

  Subscriptions

Reprints and

  .pubs@aacr.org

To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at

  Permissions

  Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://mcr.aacrjournals.org/content/16/3/476To request permission to re-use all or part of this article, use this link

on August 18, 2020. © 2018 American Association for Cancer Research. mcr.aacrjournals.org Downloaded from

Published OnlineFirst November 29, 2017; DOI: 10.1158/1541-7786.MCR-17-0483