Clonal Structures of Regionally Synchronous Gastric ... · Seung-Hyun Jung1,2, Shin Young Kim2,3, Chang Hyeok An3, Sung Hak Lee4, Eun Sun Jung4, Hyeon-Chun Park2, Min Sung Kim1,5,Yeun-Jun

Precision Medicine and Imaging

Clonal Structures of Regionally SynchronousGastric Adenomas and CarcinomasSeung-Hyun Jung1,2, Shin Young Kim2,3, Chang Hyeok An3, Sung Hak Lee4,Eun Sun Jung4, Hyeon-Chun Park2, Min Sung Kim1,5, Yeun-Jun Chung1,2,6, andSug Hyung Lee1,5

Abstract

Purpose: Gastric adenoma (GA) is a premalignant lesionthat precedes intestinal-type gastric carcinoma (GC). How-ever, genetic progressionmechanisms fromGA toGChave notbeen clarified.

Experimental Design: We performed whole-exomesequencing–based mutational analyses for 15 synchronouspairs of attached GAs and GCs.

Results: There was no significant difference in the numberof driver mutations or copy-number alterations between GAsand GCs. Well-known mutations of TP53, APC, RNF43, andRPL22were recurrently detected in synchronous GA/GC pairs.In addition, we discovered novel KDM6A, PREX2, FAT1,KMT2C, GLI3, and RPL22 mutations and hypermutation inGAs, but did not identify recurrent drivers for GA-to-GCprogression. Clonal structure analyses revealed that mostGA/GC pairs exhibit parallel evolution with early divergencerather than stepwise evolution during GA-to-GC progression.

Of note, three cases were identified as clonally nonrelatedGA/GC pairs despite the lack of histologic differences. Wefound differences in dominant mutational signatures 1, 6, 15,and 17 in GA/GC trunks, GA branches, and GC branches.Compared with our previous work on synchronous colonadenoma/carcinoma genome structures, where most driverswere in the trunkwith parallel evolution, synchronous GA/GCgenomes showed a different model of parallel evolution, withmany drivers in the branches.

Conclusions: The preferred sequence of mutational eventsduring GA-to-GC progression might be more context-dependent than colon adenomaprogression.Our results showthat nonclonal synchronous GA/GC is common and that GAgenomes have already acquired distinct genomic alterations,suggesting caution in the diagnosis of synchronous GAand GC, especially in residual or recurrent cases. Clin CancerRes; 24(19); 4715–25. �2018 AACR.

IntroductionGastric carcinoma (GC) has a high incidence in Asia and is one

of the leading causes of cancer deaths worldwide (1). The twotypes of GCs according to Lauren's classification (intestinal anddiffuse types) are known to have different pathologic and geneticpathways (2). The intestinal type begins with atrophic gastritis/intestinalmetaplasia, dysplasia, adenoma, and carcinoma,where-as the diffuse type does not have these preceding steps (3). Gastric

adenoma (GA) is a neoplastic lesion that consists of intestinal-type GA (the most common type), foveolar-type GA, and inde-terminate GA (4). It is well known that GA is a direct precursor ofintestinal-type GCs (5, 6). Progression of GA to GC comprisesfrom 2.5% to 50% of GA cases (4), indicating the importance ofelucidating the progression mechanisms as well as the need fortherapeutic removal of GAs. Large-scale efforts such as the CancerGenome Atlas consortium (7, 8) have led to a previously unrec-ognized molecular understanding of a premalignant diseaseprogression to cancer, but the molecular mechanisms underlyingmalignant progression from GA to GC remain largely unknown.One way to identify these mechanisms of progression is toevaluate the premalignant lesions together with an adjacentmalignant lesion by genetic analyses. It is challenging to analyzepremalignant and malignant lesions in the stomach togetherdue to the loss of GA tissues during malignant progression, butresidualGAsoften continue to exist alongwithGC lesions. In suchsynchronousGAs andGCs, the temporal genetic evolution duringtheGA-to-GCprogression in a givenpatient canbe investigated byanalyzing genomic profiling of the synchronous GA and GClesions (9). In addition, the possibility of seemingly clonal butnonclonal concurrent GA and GC remains to be clarified.

Next-generation sequencing (NGS), which allows for the inter-rogation of thousands of variants from multiple genes within agiven tumor sample, has enabled cancer evolution studies.Although NGS-based whole-exome sequencing (WES) analyseshave reported mutational landscapes of GCs (7, 8, 10, 11), thoseof synchronous GAs and GCs remain unexplored. Given that theGA-to-GC transition is an evolutionary process and is encrypted

1Department of Cancer Evolution Research Center, College of Medicine, theCatholic University of Korea, Seoul, Republic of Korea. 2Department of Inte-grated Research Center for Genome Polymorphism, College of Medicine, theCatholic University of Korea, Seoul, Republic of Korea. 3Department of Surgery,College of Medicine, the Catholic University of Korea, Seoul, Republic of Korea.4Department of Hospital Pathology, College of Medicine, the Catholic Universityof Korea, Seoul, Republic of Korea. 5Department of Pathology, College ofMedicine, the Catholic University of Korea, Seoul, Republic of Korea. 6Depart-ment of Microbiology, College of Medicine, the Catholic University of Korea,Seoul, Republic of Korea.

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

Corresponding Authors: Sug Hyung Lee, College of Medicine, the CatholicUniversity of Korea, 505 Banpo-dong, Socho-gu, Seoul 137-701, Republic ofKorea. Phone: þ82-2-2258-7311; Fax: þ82-2-537-6586; E-mail:[email protected]; and Yeun-Jun Chung, Phone: þ82-2-2258-7343; Fax:þ82-2-537-0572; E-mail: [email protected]

doi: 10.1158/1078-0432.CCR-18-0345

�2018 American Association for Cancer Research.

ClinicalCancerResearch

www.aacrjournals.org 4715

on March 4, 2021. © 2018 American Association for Cancer Research. clincancerres.aacrjournals.org Downloaded from

Published OnlineFirst June 26, 2018; DOI: 10.1158/1078-0432.CCR-18-0345

http://crossmark.crossref.org/dialog/?doi=10.1158/1078-0432.CCR-18-0345&domain=pdf&date_stamp=2018-9-4

http://clincancerres.aacrjournals.org/

within the genome, evolutionary perspectives of genome-widemutational abundance and distribution may provide valuableinsights into GC development, with potential clinical benefits. Inthis study, we performed WES-based mutational analyses for 15synchronous pairs of GAs and GCs.

Materials and MethodsTumor specimens

Specimens from surgical gastrectomy for advanced gastriccancers (AGC; STC04, 05, 06, 07, 12, 13, and 14), early gastriccancer (EGC; STC03), and endoscopicmucosal resection for EGCs(STC01, 02, 08, 09, 10, 11, and 15) from 15 patients, all of themwith coexisting GA, were used for this study and came from thebiobanks of university-affiliated hospitals (Seoul St. Mary Hos-pital, St. Paul Hospital, and Bucheon St. Mary Hospital, Korea).All patients were Korean, and we only collected sporadic caseswithout any positive family history of familial adenomatosis coli or

GC. Approval for this study was obtained from the CatholicUniversity of Korea College of Medicine's Institutional ReviewBoard. The board permittedwaiver of informed consent under theregulation of Korean government because the study involvedanonymous materials and data that had been collected originallyfor nonresearch purposes. The studies were conducted in accor-dancewithHelsinki declaration. Clinicopathologic features of thecases are summarized in Table 1. The tissues used in this studyconsisted of 8 cases with formalin-fixed paraffin-embedded tis-sues, 6 with methacarn-fixed tissues, and 1 case with frozentissues. Examination of these stained tissue sections showed thatGA andGC in each casewere spatially close to each other andwerehistologically similar (Supplementary Figs. S1 and S2). Accordingto the Vienna classification (12), the GCs in our study werecategorized into category 5 (invasive neoplasia, either intramu-cosal carcinoma or submucosal carcinoma or beyond). Allof the GAs were categorized into either category 3 (noninvasivelow-gradeneoplasia; low-grade adenoma/dysplasia) or category4(noninvasive high-grade neoplasia; high-grade adenoma/dysplasia). The tissues with adenomas as well as associatedcarcinomas were serially cut and stained with hematoxylin andeosin. Adenoma cells and carcinoma cells were selectively pro-cured from hematoxylin-stained frozen sections (13). Adenomaand carcinoma cell purities of the microdissection were approx-imately 50% to 80%. For normal DNA, we used tissue blocks thatwere devoid of adenoma and carcinoma cells. For genomic DNAextraction, we used the DNeasy Blood & Tissue Kit (Qiagen)according to the manufacturer's recommendation.

WESWESwas performedwith the genomic DNA obtained fromGA,

GC, and matched normal tissues using the Agilent SureSelectHuman All Exome 50Mb kit (Agilent Technologies) according tothe manufacturer's instructions. One GA and one GC per eachpatient were analyzed in this study. For STC15, additional GA(STC15GA-2) was also analyzed. Acquisition and processing ofthe sequencing data were performed as previously described (9).

Translational Relevance

Gastric adenoma (GA) is a precursor of an intestinal-typegastric carcinoma (GC), but its genetic progression mechan-isms are largely unknown. In this study, we found that geno-mic structures of GA-to-GC progression had short trunks andlong branches, suggesting parallel evolution with an earlydivergence. Notably, clonally nonrelated GA/GC pairs (colli-sion of GA and GC) were common despite their lack ofhistologic differences; this could lead to difficulty in thepathologic diagnosis of gastric tumors, especially for residualor recurrent cases. In addition, our results revealed differencesin themutational signatures 1, 6, 15, and 17 in GA/GC trunks,GA branches, and GC branches, suggesting that dynamicmutational processes and diverse carcinogenic events haveconstantly shaped tumor genomes over time.

Table 1. Clinicopathologic parameters of 15 gastric cancer patients

MSI statusa

CaseAge/sex GC GA

Diameterof carcinomawith adenoma(cm)

Proximitybetweenadenoma andcarcinoma (mm)

Laurenclassificationof carcinoma(differentiation) T N M

TNM stage(Viennaclassificationb,surgical treatment)

Associated adenomas(Vienna classificationb)

STC01 51/F MSS MSS 1.5 <0.2 Intestinal (W) 1b 0 0 IA (category 5.2, M) Tubular (category 4.1)STC02 65/M MSS MSS 0.9 <0.2 Intestinal (W) 1b 0 0 IA (category 5.1, M) Tubular (category 4.1)STC03 87/F MSS MSS 6.5 <0.2 Intestinal (W) 2 0 0 IB (category 5.2, G) Tubulovillous (category 4.1)STC04 61/F MSS MSS 11.0 >1.5 Mixed (M) 2 2 0 IIB (category 5.2, G) Tubular (category 3)STC05 70/M MSS MSS 15.0 <0.2 Intestinal (M) 2 0 0 IB (category 5.2, G) Tubular (category 4.1)STC06 67/M MSS MSS 7.0 >1.5 Intestinal (M) 2 2 0 IIB (category 5.2, G) Tubular (category 4.1)STC07 66/M MSS MSI 5.0 0.2 Intestinal (W) 2 3 0 IIIA (category 5.2, G) Tubular (category 4.1)STC08 66/M MSI MSI 1.5 <0.2 Intestinal (M) 1a 0 0 IA (category 5.1, M) Villous (category 4.1)STC09 57/F MSI MSI 2.0 <0.2 Intestinal (W) 1a 0 0 IA (category 5.1, M) Tubular (category 4.1)STC10 54/F MSI MSI 0.7 <0.2 Intestinal (W) 1a 0 0 IA (category 5.2, M) Tubular (category 3)STC11 77/F MSI MSI 1.0 <0.2 Intestinal (M) 1a 0 0 IA (category 5.1, M) Tubular (category 3)STC12 64/M MSI MSI 12.0 <0.2 Intestinal (M) 3 1 0 IIB (category 5.2, G) Tubular (category 4.1)STC13 67/F MSI MSI 7.0 <0.2 Intestinal (M) 2a 0 0 IB (category 5.2, G) Tubular (category 3)STC14 75/F MSI MSI 6.0 <0.2 Intestinal (M) 2a 0 0 IB (category 5.2, G) Tubular (category 3)STC15 78/M MSI MSS (GA)/

MSI (GA-2)8.3 <0.2 (GA)/

<0.2 (GA-2)Intestinal (M) 1b 0 0 IA (category 5.2, M) Tubular (category 4.1, GA)/

Tubular (category 4.1, GA-2)

Abbreviations: F, female; M (sex), male; M (differentiation), moderate; W, well; G, gastrectomy; M (surgical treatment), mucosectomy.aThe MSI events were identified using WES data. All of them were consistent with MSI-PCR.bCategory 3: Noninvasive low-grade neoplasia (low-grade adenoma/dysplasia); Category 4.1: Noninvasive high-grade neoplasia (high-grade adenoma/dysplasia);Category 5.1: Intramucosal carcinoma with invasion to the lamina propria or muscularis mucosae; Category 5.2: submucosal carcinoma or beyond.

Jung et al.

Clin Cancer Res; 24(19) October 1, 2018 Clinical Cancer Research4716




In brief, 101 paired-end sequence reads were generated by usingIllumina HiSeq2000 platform. A Burrows–Wheeler aligner wasused to align the sequencing reads onto the human referencegenome (UCSC hg19). The aligned sequencing reads were eval-uated using Qualimap (14), and the sequences were deposited inthe SRA database (Project ID: PRJNA429261).

Identification of somatic variants and microsatellite instabilityPoint mutations and indels were identified using MuTect (15)

and SomaticIndelDetector (16), respectively. The ANNOVARpackage (17) was used to select somatic variants located in theexonic sequences and to predict their functional consequences. Toget reliable and robust mutation calling, the following somaticvariantswere eliminated: (i) read depth fewer than 20 in either thetumor (GA or GC) or matched normal and (ii) polymorphismsreferenced in either 1000 Genomes Project or Exome AggregationConsortium with a minor allele frequency 1% or more in EastAsian. In addition, somatic variants detected in any of paired GAand GC were reexamined for the presence of sequencing readssupporting the corresponding variants. For this, number ofsequencing reads for GA- or GC-specific variants in each case wasextracted from paired GC or GA using bam-readcount (https://github.com/genome/bam-readcount). If the number of sequenc-ing reads supporting the corresponding variants was three ormore, the variant was manually rescued. Microsatellite instability(MSI) event in each specimenwas assessed directly fromWESdatausing mSINGS software (18). For each microsatellite locus, thedistribution of size lengths was compared with a TCGA baselinepopulation of normal controls provided by mSINGS. Loci wereconsidered unstable if the number of repeats is statistically greaterthan in the TCGA baseline population. A fraction of >0.1 (10%unstable loci) was considered as MSI. MSI status by WES wasconsistent with conventional capillary-based MSI-PCR.

DNA copy-number and LOH analysisDNA copy number and LOH were estimated using WES data.

The ngCGH module and SNPRank Segmentation statisticalalgorithm in NEXUS software v9.0 (Biodiscovery) were used todefine copy-number alterations (CNA) of each sample (9). Seg-ments were classified gains and losses when the log2 ratio wasgreater than 0.25 and less than –0.25, respectively.We inferred theLOH events using Sequenza (19). All of the identified CNAs andLOH events were manually curated in terms of depth ratio and Ballele frequency.

Genomic similarity and subclonal reconstructionFor genomic similarity analysis, all somatic mutations and

CNAs were classified into three categories: "common," "adenoma-specific," and "carcinoma-specific." "Common" represents thealterations identified in both GA and the paired GC, and"specific" represents those detected in either GA or GC. Con-cordance level of CNAs was estimated by calculating the ratio ofoverlapped CNA lengths between GA and GC against total CNAlengths in each case. In this analysis, copy-number gain and lossat same genomic location were considered independent events.Depending on the concordance of somatic alterations, eachspecimen was classified into three similarity patterns: highsimilarity (>10% concordance in both mutations and CNAs),high CNA/low mutation similarity (>10% concordance inCNAs and <10% concordance in mutations), and low similarity(<10% concordance in both mutations and CNAs). Subclone

analyses were performed using the SciClone algorithm (20).Only somatic mutations detected in copy-number neutralregions were used in these analyses, which allowed for thehighest confidence quantification of variant allele frequencies(VAF) and inference of clonality. The phylogenetic relationshipas well as bell plots and spheres of cells were inferred by thesomatic mutations using the ClonEvol algorithm (21).

Mutation signaturesTo obtain aggregated mutational signature profiles of GAs

and GCs for microsatellite-stable (MSS) and MSI, somaticsingle-nucleotide variants (SNV) of each sample were mergedappropriately. These data were then used to determine COSMICmutation signatures (http://cancer.sanger.ac.uk/cosmic/signatures)using deconstructSigs R package (22). To analyze a more gener-alized temporal dynamics of mutagenic processes during GA-to-GC transition, onehypermutated case (STC06) and three collisiontumor cases (STC04, 07, and15GA)were excluded in this analysis.To dissect the changing ofmutagenic processes, we also calculatedthe contributions of individual mutational signatures to eachtumor.

Targeted deep sequencingMultiple regions (three of GA and eight of GC) of the two

collision tumors (STC04 and STC07) were sequenced by targeteddeep-amplicon sequencing. For this, we randomly selected thecopy-neutral GA-specific (n¼ 25 for STC04 and n¼ 63 for STC07)andGC-specific (n¼ 51 for STC04 and n¼ 26 for STC07) somaticvariants detected by WES. Sequencing libraries were generatedusing AmpliSeq Library Kit 2.0 with customized target panel(Life Technologies) then sequenced using the Ion S5 system (LifeTechnologies) according to the manufacturer's instructions.Sequencing reads were aligned to UCSC hg19, and genomicvariants were called using the Torrent Suite 5.2.2.

ResultsGeneral features of somatic alterations in synchronous GA andcarcinoma

We performed WES for synchronous GAs (n ¼ 16, 151x) andGCs (n¼ 15, 153x), as well as for matched normal tissue (n¼ 15,148x; Supplementary Table S1). A total of 26,044 somatic muta-tions (24,468 SNVs and 1,576 indels) were identified in thesynchronous GAs and GCs (Fig. 1A; Supplementary Table S2).TheMSI statuses of 14GAswere the same as those of synchronousGCs, but the other two samples (STC07 and STC15) had differentMSI statuses between GA and GC (Table 1), suggesting thatsome of the regionally coexisting GAs and GCs might not beclonally related. One MSS GA (STC06GA) was a hypermutatedtumor (mutation number ¼ 2,039) with POLE (p.D655Y)and MSH2 (p.Q645K) missense mutations (Fig. 1A), but thepaired GC (STC06GC) was not (mutation number¼ 215). Therewere no significant differences in the numbers of mutationsbetween GAs and GCs (Fig. 1B). Our results agreed with theCancer Genome Atlas data (225MSS GCs and 64MSI GCs; ref. 7)and our previous data (nine MSS GAs and six MSI GAs; ref. 23),confirming that there were no significant differences (Supplemen-tary Fig. S3).

We identified 532 CNAs (Fig. 1C; Supplementary Table S3),but there was no significant difference between GAs and GCs(Fig. 1D). MSS genomes harbored significantly higher numbers

Gastric Adenoma and Carcinoma Genomes

www.aacrjournals.org Clin Cancer Res; 24(19) October 1, 2018 4717



https://github.com/genome/bam-readcount

https://github.com/genome/bam-readcount

http://cancer.sanger.ac.uk/cosmic/signatures


of CNAs than MSI (P ¼ 3.1 � 10�5; Fig. 1D), which was inagreement with a previous report (24).

Genomic similarities of synchronous GA and carcinomaWe identified three types of similarity patterns between syn-

chronous GA and GC paired genomes [high similarity (n ¼ 11),high CNA/low mutation similarity (n ¼ 2), and low similarity(n¼ 3)]. Examples of the three types are illustrated in Fig. 2. Therewas no association of the grade of dysplasia with any genomicstructures as well as the abundance of the genomic alterations.The average concordance rates of mutations and CNAs betweenGA and GC were not correlated in MSS or MSI genomes (bothP > 0.05; Supplementary Figs. S4 and S5). The average concor-dance rate of CNAswas significantly higher than that ofmutations(60.4% vs. 24.9%, P ¼ 0.007; Fig. 2A). The 11 pairs (STC01-03,05, 08-13, and STC15GA-2) showed high similarities in bothsomatic mutations and CNA profiles (Fig. 2A). For example, 56%of mutations and 89% of CNAs overlapped between the GA andGC in STC02, and well-known putative driver genes (TP53 andAPC) were identified in both the synchronous GA and GC in thishigh-similarity case (Fig. 2B). Two cases (STC06 and 14) weregrouped intohighCNA/lowmutation similarity. For example, thehypermutated case (STC06GA) showed a relatively low concor-dance rate of mutations (7%) with its GC, but recurrent putativedriver alterations at variant levels (TGFBR2 andKDM6A) aswell asCNA profiles (48%), indicating that the STC06 GA and GC wereclonally related (Fig. 2B). By contrast, three low similarity cases(STC04, 07, and 15GA) had less than 10% genomic concordancein both somatic mutations and CNAs (Fig. 2). For the STC15, theinvasive GC region (STC15GC) and its overlying GA region(STC15GA-2) in a tissue block were negative for MLH1 immu-nostaining, whereas another GA region (STC15GA) in a different

tissue block was positive for MLH1 (Supplementary Fig. S2). Ofnote, all 3 cases (STC04, 07, and 15) were found in tumors largerthan 5.0 cm in diameter, whereas low similarities were not foundin smaller cases (<2.0 cm in diameter; Table 1). Two pathologistsexamined the histology of these GA/GC pairs, but did notidentify any histologic findings that support the idea that the GAand GC pair was clonally different in each case (SupplementaryFig. S1). These GA/GC pairs (STC14GA/GC, STC07GA/GC,and STC15GA/GC) did not harbor any shared putative drivermutations and exhibited largely different CNA profiles (Supple-mentary Figs. S4 and S5). In STC15, STC15GC shared all of theputative driver mutations with STC15GA-2, but not withSTC15GA. To address whether these GA/GCpairs originated fromdifferent clones (possible collision tumors), we further analyzedmultiple regions of STC04 (1 GA and 4 GC regions) and STC07(2 GA and 4 GC regions) by a targeted deep resequencing(average 3,717x; Supplementary Table S4). Interestingly, nomutations detected in the GA or GC regions were shared withmutations detected in its synchronous GA or GC (SupplementaryFig. S6), suggesting the existence of "collisions" of geneticallydistinct tumor.

Putative driver alterationsOf the nonsilent mutations detected in our study, 589 in 169

genes were cataloged in the COSMIC database (ref. 25; Supple-mentary Table S2). Of these, 22 genes overlapped with theCOSMIC top 20 genes for either GA or intestinal-type GC(Fig. 3A). Previously reported putative GC driver genes (7, 11)including GLI3, ERBB2, ERBB3, SMAD4, ARID2, KDM6A,and ATM were also identified (Fig. 3A). Putative drivermutation numbers between GA and GC were not different inMSS (P ¼ 0.413) or MSI (P ¼ 0.833). KDM6A, PREX2, FAT1,

Figure 1.

Abundance of somatic mutations and CNAs in GA and GC genomes. A, Numbers of somatic mutations in the synchronous GA and GC genomes are shown.B,Comparison of the number of somaticmutations in synchronousGAandGCgenomes; 14MSS vs. 17MSI, 7MSSGAvs. 7MSSGC, and9MSIGAvs. 8MSIGCgenomes.�� , P < 0.01 by Mann–Whitney U test. C, Numbers of CNAs in each synchronous GA and GC. D, Comparison of the number of CNAs in synchronous GAand GC genomes; 14 MSS vs. 17 MSI, 7 MSS GA vs. 7 MSS GC, and 9 MSI GA vs. 8 MSI GC genomes. �� , P < 0.01 by Mann–Whitney U test.

Jung et al.





KMT2C, GLI3, and RPL22 mutations found in our GAs are novelmutations that have not been reported in GAs (23, 26).

For CNAs, 67 and 10 CNA regions were recurrently detected(�3 cases) in MSS and MSI, respectively (Fig. 3B; SupplementaryTable S5). Recurrent CNA lengths between GA and GC genomes

were not significantly different in MSS (P ¼ 0.842) or MSI (P ¼0.783). Among the recurrent CNAs, 56 including EGFR and APCwere common to GAs and GCs, whereas the other 21 includingPIK3CA (GC-specific) and ATM (GA-specific) were GA- or GC-specific (Fig. 3B; Supplementary Table S5). Twelve high-level

Figure 2.

Genomic similarities of synchronous GA and GC. A, Concordance rate of mutations and CNAs. B, Somatic mutations and genome-wide copy-number profilesare shown in the left and right plots, respectively. Per case, each lane represents the synchronous GA and GC. Copy-number gains and losses are shown as redand blue, respectively. Putative driver gene mutations and CNAs are marked with color and underlined (loss-of-function mutations; red, reported at COSMICdatabase; underline).






amplifications (>5 copies) and two homozygous deletions weredetected in the MSS genomes (Supplementary Table S3). Well-known oncogenes, ERBB2 and CCNE1, were included in theamplification and were successfully validated using quantitativereal-time PCR (Supplementary Fig. S7). STC07GC, but not its GA,displayed potential chromothripsis areas with complex recombi-nation events on chromosomes 4, 6, 11, and 19. STC15GC, butnot its GA, harbored copy-neutral LOH on chromosome 8 (Sup-plementary Figs. S8 and S9), further confirming the existence ofclonally nonrelated GA and GC collisions in the stomach.

Subclonal relationships of synchronous GA and carcinomaNext, we examined the subclonal architectures of synchronous

GA/GC genomes and identified two major patterns: clonallydifferent (collision; discordant pairs in STC04, 07, and 15GA)and clonally related (stepwise or parallel, the other 13 casesincluding STC15GA-2; Fig. 4). In parallel evolution, there shouldbe no distinct subclonal cluster in the GA that is simultaneouslyclonal in the GC (27). All samples except STC12 (stepwiseevolution) exhibited parallel evolution. With respect to branch-based trees, clonally related cases were further categorized into

Figure 3.

Cancer-related mutations and CNAs in the GA and GC genomes. A, Cancer-related mutations according to WES for 15 pairs of GA and GC genomes.Gene symbols in the COSMIC top 20 genes for GA or intestinal-type GC are underlined. The "X" symbol in the box represents the mutation reported in theCOSMIC database at the variant level. B, Frequencies (y-axis) of copy-number gains and losses across the whole genomes of MSS GA and GC genomes (left)and MSI GA and GC genomes (right). Red denotes copy-number gains, and blue denotes copy-number losses.

Jung et al.





two patterns: a simple pattern and a complex pattern (Fig. 4;Supplementary Fig. S10). The simple cluster pattern, found inmost cases (STC01, 02, 05, 06, 08, 11, 13, 14, and 15),exhibited one common trunk and one branch each for GA andGC. The complex pattern, found in another four cases (STC03,09, 10, and 12), exhibited one common trunk and threebranches (one for GA, one for GC, and the remaining one forGA or GC). We also measured the lengths of the trunks and thebranches in each case and categorized them into a short trunk(<average branch length) or a long trunk pattern (>averagebranch length). Five cases (STC02, 03, 09, 12, and 15) had longtrunks, whereas the other eight had short trunks. Together,these two analyses suggest that most GA-to-GC progressionfollows an early divergence with parallel evolution. In bothsimple and complex patterns, many putative GC driver genessuch as TP53, APC, RPL22, and RNF43 were located in thetrunk. However, nine cases harbored putative driver mutationsin either GA- or GC-specific branches.

Mutation signaturesWe analyzed the mutation signatures in paired GAs and GCs

according to the COSMIC database (22). Overall mutationalsignatures were highly enriched with signature 1 (associatedwith age) in the trunks of bothMSS andMSI genomes. However,the signatures of the branches, especially those of MSS, weresomewhat different from those of the trunks (Fig. 5A). In theMSS branches, we identified decreased signature 1 and increasedsignature 17. Of note, signature 15, which was associated withmismatch repair (MMR) deficiency, newly emerged in the MSSGC branch, but not in the GA branch (Fig. 5A). In the MSI,signature 6 increased in the branches. Next, we calculated thecontributions of individual mutational signatures to eachtumor and identified 20mutational signatures (Fig. 5B). Amongthem, seven (signatures 15, 17, 18, 20, 21, 26, and 28) havebeen found in stomach cancers (28). These stomach cancer–related signatures were significantly enriched in MSS GCs com-pared with MSI GCs (P ¼ 0.026). To the contrary, MMRdeficiency–related signature 6 was enriched in MSI GCs com-pared with MSS GC (P ¼ 0.026). The hypermutated GA showedsignature 9, which was associated with the activity of activation-induced cytidine deaminase during somatic hypermutation.Signature 9 has also been related to hypermutation in leukemiaand lymphoma, but we do not find any evidence of undiag-nosed hematologic malignancies in our case. Among the colli-sion cases, STC07 and STC15 showed different mutationalsignatures between GA and GC, whereas the other one (STC04)had similar signatures (Fig. 5C).

DiscussionIn this study,we investigated the genomic differences, clonality,

and evolution of synchronous GA and GC using WES-basedmutations and DNA copy-number profiling. To the best of ourknowledge, this is the first whole exome-wide study performed onregionally synchronous GA/GC pairs. Our data support threemajor conclusions: first, the numbers of putative drivermutationsand CNAs are not significantly different between GAs and GCs;second, most GAs exhibit parallel evolution with early divergencerather than a stepwise evolution during GA-to-GC progression;and third, in some cases, synchronous GA/GC pairsmay originatefrom different cancer-initiating cells (clonally nonrelated),

despite histologic similarities. These findings may provide newinsights into the phylogenetic principles of early gastric carcino-genesis and malignant progression.

The first two conclusions are notable because the parallelmodelmay be inconsistent with the commonly accepted stepwiseprogression model (5). In fact, two different reports (one for theparallel model and the other for the stepwise model) exist withrespect to the genomic analysis of GA-to-GC progression (23, 26).However, because these studies analyzed independent cohorts forGA andGC (nonpaired GA andGC in an individual), theremightbe intertumoral bias that is not ideal for investigating a precisespatiotemporal genetic evolution. Our results adopting synchro-nous GA/GC pairs indicate that the divergence of GA/GC pairs orthe emergence of histologically distinct subclones occurred earlyin gastric tumorigenesis. Our previous study on synchronouscolon adenoma and carcinoma also identified such early diver-gence in colon tumorigenesis (27). The hypermutated phenotypein STC06GA, but not in STC06GC,may represent a good exampleof early divergence with parallel evolution in GAs. According tothe Big-Bang model recently suggested in colon cancer, a singleexpansion in the early stage may give rise to numerous subclonesthat shape subclonal architectures and intratumoral heterogeneityof a colon cancer (29). Together, our studies suggest that geneticprogression of adenoma to carcinoma in gastric and colon cancersmay share similar mechanisms.

Of note, the average concordance rates of mutations and CNAswere not correlated in MSS or MSI genomes, suggesting thepossibility that mutations and CNAs evolve independently. Inter-estingly, the average concordance rate of CNAs was significantlyhigher than that of mutations, further suggesting that acquisitionof CNAs during gastric carcinogenesis might be an earlierevent than the acquisition of somatic mutations. These resultssupport a previous notion that the majority of CNAs develop in apunctuated burst of evolution, while mutations are acquiredgradually (30).

In three (STC04, 07, and 15) of 15 cases (20%), we observedvery lowgenomic concordance betweenGAs andGCswith respectto both somatic mutations and CNAs, indicating the existence of"collisions" of genetically distinct tumors. For instance, differentTP53 mutations were identified in STC04GA (p.R116Q) andSTC04GC (p.R205C) and in STC07GA (p.P33fs) and STC07GC(p.R141C). The "collisions" were also identified by differentCNA profiles. Interestingly, the invasive GC region of STC15(STC15GC) and its overlying GA region (STC15GA-2) wereclonally related, whereas another GA region (STC15GA) in adifferent tissue block was clonally nonrelated to STC15GC, sug-gesting that it could lead to pathologic or genomic misdiagnosisof GA or GC. In this case, the STC15GA genome harbored APCp.E1536fs, whereas STC15GC and STC15GA-2 genomes har-bored APC p.R1432X. Our multiple regional sequencing of thesethree collision cases (STC04, 07, and 15) excluded the possibilityof undersampling or missampling. Synchronous multiple gastrictumors in the stomach, either attached or detached, have beenreported in 10% to 20% of patients (31–33), but no reports haveaddressed their differences and similarities at the genome level.Our results revealed that, even among attached GAs and GCs, aconsiderable fraction (3/15, 20%) of synchronous GAs and GCsproved to be collision tumors. Our data suggest that GA and GCcollisions tend to be found in large tumors; however, consideringthe small number of cases, further observationswill be required toconfirm this possibility.






Consistent with the previous reports (7, 11, 24, 26), the well-known somatic mutations TP53, APC, PREX2, SMAD4, KRAS,RNF43, RPL22, and GLI3 were recurrently detected in both GAs

and GCs in our study, suggesting that these mutations may beputative drivers of early gastric tumorigenesis. Comparing themutations detected in our synchronous GAs with GCs with those

Figure 4.

Subclonal clustering and predicted evolutionary trees of 15 stomach cancer genomes. The distribution of VAF in synchronous GA (x-axis) and GC (y-axis) isillustrated for 15 cases. Colors indicate assigned clusters by the SciClone algorithm (20). Only mutations in copy-number neutral regions are used forsubclonal clustering. Clonal evolution trees are also illustrated on the right of clustering plots. Each evolutionary treewas estimated using the ClonEvol algorithm (21)with a branch length scaled to the number of variants of subclones. Putative driver genes are labeled in the trees.

Jung et al.





Figure 5.

Mutational signatures in the 15 synchronous GA and GC genomes. A, Aggregated mutational signature profiles of GAs and GCs for MSS (left) and MSI (right)in this study. Signature composition of each tree was inferred using deconstructSigs (22) according to the 30 COSMIC mutation signatures. One hypermutatedcase (STC06) and three collision tumor cases (STC04, STC07, and STC15GA) were excluded from this analysis. B, Heatmap showing the proportion ofmutational signatures in each sample. Asterisks represent the samples that harbor <50 mutations. Bold underlines represent known mutational signatures instomach cancer. †, GA and z, GA-2. C, Colored bar charts illustrate mutational signatures in the collision tumors (STC04, STC07, and STC15) according to the96-trinucleotide mutation types.






in nonsynchronous GAs (23, 26), we found that TP53, KRAS,RNF43, and APC mutations were identified in nonsynchronousGAs as well. Of note, we identified previously unreported recur-rent mutations (KDM6A, PREX2, FAT1, KMT2C, GLI3, andRPL22) in our GAs. Interestingly, RNF43-truncating mutations,previously reported only in GA with high-grade dysplasia (26),were detected in the GA with low-grade dysplasia in our study.We did not observe any recurrent determinants for GA-to-GCprogression because most (8/12) clonally related cases showedearly divergence. In STC12, which had a stepwise progression,APC, RPL22, and KMT2C mutations were considered putativedrivers of progression.

Compared with our previous work on synchronous colonadenoma/carcinoma genome structures, where almost all driverswere located on the trunks with parallel evolution (27), synchro-nous GA/GC genomes showed a different pattern of parallelevolution, with many putative drivers located in the branches.For example, TP53 mutations were located in either the trunk(STC01, 02, and 05), the GA branch (STC13), or the GC branch(STC03, 08, and 10). These findings suggest that putative drivermutations in the GA-to-GC progression are not fixed but arecontext-dependent, which is different from thepreferred sequenceof mutational events in colon carcinogenesis.

In this study, we found evidence for shifts of dominant muta-tional signatures during gastric tumorigenesis. Most cases dis-played a prominent decrease in the relative proportion of signa-ture 1, the most common signature known in gastric cancer (34),with increases in the relative proportion of signatures 6, 15, and17 in the branches compared with trunk mutations. Of note,signature 15 (associated with MMR deficiency) newly emerged inonly the carcinoma branch, suggesting the contribution of DNArepair deficiency to gastric cancer progression. It is interesting thatMMR deficiency signature is present in MMR-proficient samples.When we assessed the mutational signatures of the 191 MSS GCsin TCGA with race category information (61 Asian and 130Caucasian), the individual mutational signature pattern waslargely consistent with our MSS data: a strong signal for signature1 and a weak signal of stomach cancer–related signatures(signatures 15, 17, 18, 20, 21, 26, and 28). The signature 15 wasdetected in 26% of Asian and 25% of Caucasian patients, indi-cating that the signature 15 may not be ethnicity-specific. Giventhe low amount of MMR deficiency signature, the signature inMMR-proficient samples in both our cohort and the TCGA is notclaiming that the samples are deficient rather just that the sampleshave mutations characteristic of MMR deficiency, of which the

exact source is not entirely clear. Signature 6, another MMRdeficiency–associated signature, has been reported inMSI tumorsincluding colorectal and uterine cancers, but not in gastric cancer(35, 36). It was striking that signature 6 was not detected in aprevious study (28). In our study, signature 6was the secondmostcommon signature and was significantly enriched in MSI com-pared with MSS tumors (P ¼ 0.026), suggesting that signature 6might be a footprint of the MSI phenotype.

In summary, we report the genomic profiles and structures ofpaired GAs/GCs, which may provide useful information forunderstanding GA-to-GC progression. Our results showed thatclonally nonrelated GA/GC pairs were common, and that GAgenomes have already acquired distinct genomic alterations.Practically, clinicians should be aware that collisionGA/GC couldlead to pathologic or genomic misdiagnosis of GA or GC.Although the present study has analyzed the largest number ofsynchronous GAs/GCs to date, it has somewhat preliminarynature due to the small number of cases, and a larger collaborativeproject should be needed for the definitive determination.

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

Authors' ContributionsConception and design: Y.-J. Chung, S.H. LeeDevelopment of methodology: S.-H. Jung, Y.-J. ChungAcquisition of data (provided animals, acquired and managed patients,provided facilities, etc.): C.H. An, S.H. Lee, E.S. Jung, M.S. KimAnalysis and interpretation of data (e.g., statistical analysis, biostatistics,computational analysis): S.-H. Jung, S.Y. Kim, H.-C. Park, M.S. Kim,Y.-J. Chung, S.H. LeeWriting, review, and/or revision of the manuscript: S.-H. Jung, Y.-J. Chung,S.H. LeeAdministrative, technical, or material support (i.e., reporting or organizingdata, constructing databases): S.-H. Jung, S.Y. Kim, H.-C. Park, Y.-J. ChungStudy supervision: Y.-J. Chung

AcknowledgmentsThis work was supported by grants from National Research Foundation of

Korea (2012R1A5A2047939, 2017R1A2B2002314, and 2017R1E1A1A01074913).

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 January 29, 2018; revised May 7, 2018; accepted June 22, 2018;published first June 26, 2018.

References1. Ferlay J, Soerjomataram I, Dikshit R, Eser S, Mathers C, Rebelo M, et al.

Cancer incidence and mortality worldwide: sources, methods and majorpatterns in GLOBOCAN 2012. Int J Cancer 2015;136:E359–86.

2. Lauren P. The two histological main types of gastric carcinoma: diffuse andso-called intestinal-type carcinoma. An attempt at a histo-clinical classifi-cation. Acta Pathol Microbiol Scand 1965;64:31–49.

3. Correa P. Human gastric carcinogenesis: a multistep and multifactorialprocess–first american cancer society award lecture on cancer epidemiologyand prevention. Cancer Res 1992;52:6735–40.

4. Abraham SC, Park SJ, Lee JH, Mugartegui L, Wu TT. Genetic alterations ingastric adenomas of intestinal and foveolar phenotypes. Mod Pathol2003;16:786–95.

5. Kolodziejczyk P, Yao T, Oya M, Nakamura S, Utsunomiya T, Ishikawa T,et al. Long-term follow-up study of patients with gastric adenomas with

malignant transformation. An immunohistochemical and histochemicalanalysis. Cancer 1994;74:2896–907.

6. Orlowska J, Jarosz D, Pachlewski J, Butruk E. Malignant transformation ofbenign epithelial gastric polyps. Am J Gastroenterol 1995;90:2152–9.

7. Cancer Genome Atlas Research N. Comprehensive molecular characteri-zation of gastric adenocarcinoma. Nature 2014;513:202–9.

8. Cancer Genome Atlas Research N, Analysis Working Group: Asan U,Agency BCC, Brigham, Women's H, Broad I, et al. Integrated genomiccharacterization of oesophageal carcinoma. Nature 2017;541:169–75.

9. Jung SH, Shin S, KimMS, Baek IP, Lee JY, Lee SH, et al. Genetic progressionof high grade prostatic intraepithelial neoplasia to prostate cancer. EurUrol2016;69:823–30.

10. Zang ZJ, Cutcutache I, Poon SL, Zhang SL, McPherson JR, Tao J, et al.Exome sequencing of gastric adenocarcinoma identifies recurrent somatic

Jung et al.





mutations in cell adhesion and chromatin remodeling genes. Nat Genet2012;44:570–4.

11. Wang K, Yuen ST, Xu J, Lee SP, Yan HH, Shi ST, et al. Whole-genomesequencing and comprehensive molecular profiling identify new drivermutations in gastric cancer. Nat Genet 2014;46:573–82.

12. SchlemperRJ, Riddell RH,KatoY, Borchard F,CooperHS,Dawsey SM, et al.The Vienna classification of gastrointestinal epithelial neoplasia. Gut2000;47:251–5.

13. Lee SH, ShinMS, ParkWS, Kim SY, KimHS, Han JY, et al. Alterations of Fas(Apo-1/CD95) gene in non-small cell lung cancer. Oncogene 1999;18:3754–60.

14. Garcia-Alcalde F, Okonechnikov K, Carbonell J, Cruz LM, Gotz S, TarazonaS, et al. Qualimap: evaluating next-generation sequencing alignment data.Bioinformatics 2012;28:2678–9.

15. Cibulskis K, LawrenceMS,Carter SL, SivachenkoA, JaffeD, SougnezC, et al.Sensitive detection of somatic point mutations in impure and heteroge-neous cancer samples. Nat Biotechnol 2013;31:213–9.

16. DePristo MA, Banks E, Poplin R, Garimella KV, Maguire JR, Hartl C, et al.A framework for variation discovery and genotyping using next-generationDNA sequencing data. Nat Genet 2011;43:491–8.

17. WangK, LiM,HakonarsonH.ANNOVAR: functional annotationof geneticvariants from high-throughput sequencing data. Nucleic Acids Res 2010;38:e164.

18. Salipante SJ, Scroggins SM, Hampel HL, Turner EH, Pritchard CC. Micro-satellite instability detection by next generation sequencing. Clin Chem2014;60:1192–9.

19. Favero F, Joshi T, Marquard AM, Birkbak NJ, Krzystanek M, Li Q, et al.Sequenza: allele-specific copy number and mutation profiles from tumorsequencing data. Ann Oncol 2015;26:64–70.

20. Miller CA, White BS, Dees ND, Griffith M, Welch JS, Griffith OL, et al.SciClone: inferring clonal architecture and tracking the spatial and tem-poral patterns of tumor evolution. PLoS Comput Biol 2014;10:e1003665.

21. Dang HX, White BS, Foltz SM, Miller CA, Luo J, Fields RC, et al. ClonEvol:clonal ordering and visualization in cancer sequencing. Ann Oncol2017;28:3076–82.

22. Rosenthal R, McGranahan N, Herrero J, Taylor BS, Swanton C.DeconstructSigs: delineating mutational processes in single tumorsdistinguishes DNA repair deficiencies and patterns of carcinomaevolution. Genome Biol 2016;17:31.

23. Lim CH, Cho YK, Kim SW, Choi MG, Rhee JK, Chung YJ, et al.The chronological sequence of somatic mutations in early gastric

carcinogenesis inferred from multiregion sequencing of gastric adeno-mas. Oncotarget 2016;7:39758–67.

24. Kim TM, Jung SH, KimMS, Baek IP, Park SW, Lee SH, et al. The mutationalburdens and evolutionary ages of early gastric cancers are comparable tothose of advanced gastric cancers. J Pathol 2014;234:365–74.

25. Futreal PA, Coin L, Marshall M, Down T, Hubbard T, Wooster R, et al.A census of human cancer genes. Nat Rev Cancer 2004;4:177–83.

26. Min BH, Hwang J, KimNK, Park G, Kang SY, Ahn S, et al. DysregulatedWntsignalling and recurrent mutations of the tumour suppressor RNF43 inearly gastric carcinogenesis. J Pathol 2016;240:304–14.

27. KimTM, AnCH, Rhee JK, Jung SH, Lee SH, Baek IP, et al. Clonal origins andparallel evolution of regionally synchronous colorectal adenoma andcarcinoma. Oncotarget 2015;6:27725–35.

28. Alexandrov LB, Nik-Zainal S, Siu HC, Leung SY, StrattonMR. Amutationalsignature in gastric cancer suggests therapeutic strategies. Nat Commun2015;6:8683.

29. Sottoriva A, Kang H, Ma Z, Graham TA, Salomon MP, Zhao J, et al. A BigBang model of human colorectal tumor growth. Nat Genet 2015;47:209–16.

30. Gao R, Davis A, McDonald TO, Sei E, Shi X, Wang Y, et al. Punctuated copynumber evolution and clonal stasis in triple-negative breast cancer.Nat Genet 2016;48:1119–30.

31. Isomoto H, Shikuwa S, Yamaguchi N, Fukuda E, Ikeda K, Nishiyama H,et al. Endoscopic submucosal dissection for early gastric cancer: alarge-scale feasibility study. Gut 2009;58:331–6.

32. Goto O, Fujishiro M, Kodashima S, Ono S, Omata M. Outcomes ofendoscopic submucosal dissection for early gastric cancer with specialreference to validation for curability criteria. Endoscopy 2009;41:118–22.

33. Lee HH, Park JM, Lim CH, Kim JS, Cho YK, Choi MG. The impact of pre-resection endoscopic examination time on the rate of synchronous gastricneoplasms missed during endoscopic treatment. Surg Endosc 2017;31:3952–60.

34. Alexandrov LB, Jones PH, Wedge DC, Sale JE, Campbell PJ, Nik-Zainal S,et al. Clock-like mutational processes in human somatic cells. Nat Genet2015;47:1402–7.

35. AlexandrovLB,Nik-Zainal S,WedgeDC,Aparicio SA, Behjati S, BiankinAV,et al. Signatures of mutational processes in human cancer. Nature2013;500:415–21.

36. Alexandrov LB, Stratton MR. Mutational signatures: the patterns ofsomatic mutations hidden in cancer genomes. Curr Opin Genet Dev2014;24:52–60.






2018;24:4715-4725. Published OnlineFirst June 26, 2018.Clin Cancer Res Seung-Hyun Jung, Shin Young Kim, Chang Hyeok An, et al. and CarcinomasClonal Structures of Regionally Synchronous Gastric Adenomas

Updated version

10.1158/1078-0432.CCR-18-0345doi:

Access the most recent version of this article at:

Material

Supplementary

http://clincancerres.aacrjournals.org/content/suppl/2018/06/26/1078-0432.CCR-18-0345.DC1

Access the most recent supplemental material at:

Cited articles

http://clincancerres.aacrjournals.org/content/24/19/4715.full#ref-list-1

This article cites 36 articles, 4 of which you can access for free at:

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

Subscriptions

Reprints and

[email protected]

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://clincancerres.aacrjournals.org/content/24/19/4715To request permission to re-use all or part of this article, use this link



http://clincancerres.aacrjournals.org/lookup/doi/10.1158/1078-0432.CCR-18-0345

http://clincancerres.aacrjournals.org/content/suppl/2018/06/26/1078-0432.CCR-18-0345.DC1

http://clincancerres.aacrjournals.org/content/24/19/4715.full#ref-list-1

http://clincancerres.aacrjournals.org/cgi/alerts

mailto:[email protected]

http://clincancerres.aacrjournals.org/content/24/19/4715


Documents

Clonal Structures of Regionally Synchronous Gastric ... · Seung-Hyun Jung1,2, Shin Young Kim2,3, Chang Hyeok An3, Sung Hak Lee4, Eun Sun Jung4, Hyeon-Chun Park2, Min Sung Kim1,5,Yeun-Jun