Intrapatient genomic divergence across multiple primary tumors in young Korean patients

Abstract

Purpose

Multiple primary tumors arising in the same individual pose challenges for precision oncology, particularly in the context of hereditary cancer syndromes such as Lynch syndrome. While these tumors may originate from a shared germline predisposition, it remains unclear whether they also share somatic alterations that could be therapeutically exploited. This study aimed to characterize the extent of somatic genomic overlap between synchronous or metachronous gastric and colorectal cancers within young Korean patients.

Methods

Nineteen patients diagnosed with both gastric and colorectal cancers before age 55 underwent whole exome sequencing of formalin-fixed paraffin-embedded tumor tissues. Microsatellite instability (MSI) status was determined, and germline mismatch repair (MMR) variants were assessed to identify Lynch syndrome cases. Somatic mutations, mutational signatures, and copy number alterations were analyzed to quantify intertumoral genomic similarity within individual patients.

Results

Among the 37 tumors analyzed, 36.8% of gastric and 44.4% of colorectal cancers were MSI-high. Germline pathogenic MMR variants were identified in seven patients. Despite shared hypermutated phenotypes, the proportion of overlapping somatic mutations between paired tumors was consistently low (<5%). Mutational signatures varied by MSI status and included SBS1/5 (aging) and SBS15 (MMR deficiency). Notably, one non-Lynch patient exhibited MYC amplification in both tumors, confirmed by fluorescence in situ hybridization.

Conclusion

Even in patients with shared germline predisposition, primary tumors arising in different organs demonstrate substantial genomic divergence. These findings suggest that organ-specific selective pressures drive independent tumor evolution, underscoring the need for individualized molecular profiling and therapeutic targeting in cases of multiple primary cancers.

INTRODUCTION

Tumor heterogeneity, a significant barrier to effective cancer treatment [1], is generally classified into intratumoral heterogeneity (diversity within a single tumor) and intertumoral heterogeneity (differences among separate tumors) [2,3]. Intertumoral heterogeneity encompasses differences between tumors in distinct patients [4], as well as between primary and metastatic tumors within the same individual [5,6], which often diverge through clonal evolution.

However, intertumoral heterogeneity among multiple primary tumors arising in different organs within the same patient remains poorly understood, particularly in hereditary cancer syndromes such as Lynch syndrome, where a common germline mismatch repair (MMR) alteration initiates tumorigenesis and leads to microsatellite instability-high (MSI-H) phenotypes [7]. In young patients with multiple primary cancers, these tumors may share molecular characteristics due to their inherited predisposition; yet, whether they also harbor substantial shared somatic alterations or instead diverge independently through organ-specific selective pressures remains unclear [8]. Addressing this question is critical to determining whether simultaneous therapeutic targeting of multiple primary tumors is feasible, or if distinct, organ-specific treatment strategies are required—highlighting an important gap in precision oncology. Given the increasing number of individuals who develop multiple primary cancers during their lifetime, a comprehensive understanding of the genomic and molecular characteristics of these tumors is becoming increasingly important [9].

In this study, we aimed to characterize the somatic mutational landscape of synchronous gastric and colorectal cancers arising in young Korean patients, many of whom harbor germline mutations in MMR genes [10]. Using whole exome sequencing (WES), we explored the degree of somatic variant overlap, mutational signatures, and actionable genomic alterations between paired tumors within individuals.

METHODS

Study cohort and tissue sample collection

This study included patients diagnosed with multiple primary cancers involving both the stomach and the colon before the age of 55. Eligible patients were selected from the surgical pathology archives of Severance Hospital, Yonsei University College of Medicine, spanning from January 2000 to December 2016. The inclusion criteria were as follows: (1) histologically confirmed primary adenocarcinoma of the stomach and colon; (2) surgical or endoscopic resection performed for both tumors; (3) availability of formalin-fixed paraffin-embedded (FFPE) tissue blocks for each tumor and matched normal tissue; and (4) adequate DNA quality for downstream genomic analyses. All histopathological diagnoses were confirmed by gastrointestinal pathologists. The clinical–pathological characteristics of the patients including age, sex, family history, location of tumors, number of tumors, histology and TNM stage, and MSI status were evaluated. This study was approved by the Institutional Review Board (IRB) of Severance Hospital of the Yonsei University Health System (4-2017-0434). The IRB of Severance Hospital of the Yonsei University Health System waived the requirement for patient informed consent as the study is retrospective by design. This cohort is a part of previously reported study, for evaluating germline risk of multiple primary cancers [10].

WES analysis

Genomic DNA was isolated from FFPE matched normal and tumor (gastric and colorectal) tissues using the DNeasy Tissue Kit (Qiagen) according to the manufacturer’s instructions. The integrity of extracted DNA was verified using 1% agarose gel electrophoresis, and DNA concentrations were quantified using the PicoGreen dsDNA Assay (Invitrogen).

WES libraries were prepared using the Agilent SureSelect All Exon V6 Kit (Agilent Technologies) and processed on a Bravo automated liquid handler. Library quality was assessed using capillary electrophoresis on a Bioanalyzer (Agilent). Following library quantification via real-time quantitative polymerase chain reaction (qPCR) with SYBR Green PCR Master Mix (Applied Biosystems), index-tagged libraries were pooled at equimolar concentrations. Cluster generation was performed on a cBot automated system (Illumina), and sequencing was carried out on an Illumina HiSeq 2500 sequencing platform (Illumina) using paired-end reads of 2×100 bp.

MSI and MMR status assessment

DNA extracted from matched normal and tumor tissues underwent PCR amplification. The MSI status was evaluated using two mononucleotide markers (BAT25 and BAT26) and three dinucleotide markers (D5S346, D2S123, and D17S250), according to National Cancer Institute consensus guidelines. Tumors were classified as MSI-H if instability was observed in two or more markers, MSI-low if instability was detected in only one marker, and microsatellite stable (MSS) if no instability was detected [11].

Immunohistochemistry (IHC) was performed using a Ventana XT automated staining system (Ventana Medical Systems). Primary antibodies included cytokeratin (AE1/AE3, 1:300; DAKO), MLH1 (clone M1; Roche), MSH2 (clone G219-1129; Roche), MSH6 (clone 44, 1:100; Cell Marque), and PMS2 (clone MRQ28, 1:40; Cell Marque). Tissue sections were deparaffinized with EZ Prep solution (Ventana).

Loss of MMR protein expression (MMR deficiency) was defined as a complete absence of nuclear staining in neoplastic epithelial cells, while normal expression was indicated by nuclear positivity irrespective of staining intensity or proportion. Adjacent non-neoplastic epithelium, stromal cells, and infiltrating lymphocytes served as internal positive controls. A tumor was classified as MMR-deficient (dMMR) if it exhibited loss of nuclear staining for any of the four MMR proteins. In this study, tumors were considered MSI-H if they demonstrated MSI-H status by PCR or dMMR status by IHC [11].

Data analysis and bioinformatics pipeline

Continuous variables were summarized as means±standard deviations and compared using the Mann-Whitney U-test. Categorical variables were summarized as frequencies and percentages and compared using chi-square tests or Fisher exact tests, as appropriate. Statistical significance was defined as a P-value of less than 0.05. Analyses were performed using SPSS software, version 23.0 (IBM Corp.).

For bioinformatics analyses, WES data were aligned to the reference genome using Burrows-Wheeler Aligner. Duplicate reads were removed using Picard tools. Indel realignment and base quality recalibration were performed using Genome Analysis Toolkit (GATK). Somatic variant calling and filtering were conducted using established pipelines, and variant annotation was performed with SnpEff software. Genomic similarity was estimated by Treeomics [12], copy number was estimated by CNVkit [13], and mutation signature analysis was evaluated by SigProfilerExtractor [14].

To select high-confidence variants for further analysis, we included only those that passed the GATK Best Practices pipeline, with a tumor log odds >12, mapping quality ≥60, and sequencing depth >10. Furthermore, to ensure the somatic origin of the variants, we retained only those with a variant allele frequency >1% in tumor samples and no supporting reads in the matched normal sample.

RESULTS

A total of 19 patients were included in the present study. All patients were diagnosed with both gastric and colorectal cancer before the age of 55 and underwent surgical treatment for both malignancies. The mean age at diagnosis was 46 years for gastric cancer and 48 years for colorectal cancer. Most patients had a family history of gastric cancer, colorectal cancer, or other types of malignancies. MSI-H status was observed in 36.8% (n=7) of gastric cancers and 44.4% (n=8) of colorectal cancers. Pathogenic germline variants consistent with Lynch syndrome were identified in seven patients, accounting for 36.8% of the study population. More detailed clinical and molecular characteristics are summarized in Table 1.

Table 1.

Demographics of enrolled population

Variable		Patients (n=19)
Age at cancer diagnosis (yr)
GC		46.0±4.6
CRC		48.3±4.3
Family history (n=13)
GC	No	3 (23.1)
	Yes	10 (76.9)
CRC	No	8 (61.5)
	Yes	5 (38.5)
Any cancer	No	0
	Yes	13 (100)
Location of tumor
GC	UB	1 (5.3)
	M-LB	18 (94.7)
CRC	Right	7 (36.8)
	Left	12 (63.2)
Number of lesions
GC	1	18 (94.7)
	≥2	1 (5.3)
CRC	1	17 (89.5)
	≥2	2 (10.5)
Histology
GC	Differentiated	6 (31.6)
	Undifferentiated	13 (68.4)
CRCa)	W-MD	15 (83.3)
	PD	3 (16.7)
TNM stageb)
GC	I	10 (52.6)
	II–III	9 (47.4)
CRC	I	8 (42.1)
	II–III	11 (57.9)
MSI status
GC	MSS	12 (63.2)
	MSI-H	7 (36.8)
CRCa)	MSS	10 (55.6)
	MSI-H	8 (44.4)
Presence of P/LP variant	No	12 (63.2)
	Yes	7 (36.8)

Values are presented as mean±standard deviation or number (%).

GC, gastric cancer; CRC, colorectal cancer; UB, upper body; M-LB, mid-lower body; W-MD, well-moderate differentiated; PD, poorly differentiated; MSI, microsatellite instability; MSS, microsatellite stable; MSI-H, MSI-high; P, pathogenic; LP, likely-pathogenic.

^a)MSI status was not available in one case;

^b)8th edition of the AJCC (American Joint Committee on Cancer) Cancer Staging Manual.

Clinically, all 19 patients were diagnosed with at least one gastric and one colorectal cancer and underwent surgical resection. Among them, one patient had two distinct gastric cancers, and two patients each had three separate colorectal cancers. In total, WES was performed on 37 gastric and colorectal tumor tissues for which FFPE blocks were available. Detailed quality control metrics for the WES data are provided in Supplementary Table 1. Detailed information regarding the order and age at diagnosis of gastric and colorectal cancers in each patient, the synchronous or metachronous nature of the tumors, family history of cancer, MSI status of each tumor, and the presence or absence of Lynch syndrome is summarized in Supplementary Table 2.

Somatic mutations identified in gastric and colorectal cancers are illustrated in Fig. 1. The most frequently observed mutations were in TP53 and APC, with the majority of APC mutations being deleterious nonsense or frameshift (INDEL) variants. When comparing mutational signatures across individual tumor samples, the clock-like signatures single base substitution (SBS)1 and SBS5 were consistently present in most tumors (Fig. 2). The dMMR signature SBS15 was predominantly observed in MSI-H tumors. In the colorectal tumor from patient dou_019, mutational signatures SBS10a and SBS10b, which are characteristic of polymerase epsilon exonuclease domain mutations, were detected. Correspondingly, this tumor harbored a hotspot POLE mutation, p.Pro286Arg.

Fig. 1.

Oncoplot of somatic mutations in gastric and colorectal cancers from patients with multiple primary cancers diagnosed before the age of 55 years.

Fig. 2.

Mutational signature of both gastric and colorectal cancers from patients diagnosed with multiple primary malignancies before the age of 55 years. A mismatch repair-associated signature, primarily SBS15, was predominantly observed in MSI-high tumors from patients with Lynch syndrome. Notably, one colorectal tumor (019-CRC) exhibited a high contribution of SBS10a and SBS10b, corresponding to a hotspot POLE mutation (p.Pro286Arg). CRC, colorectal cancer; GC, gastric cancer; MSI, microsatellite instability; MSS, microsatellite stable.

Table 2 presents the genomic similarity between multiple tumors arising in the same patient. All tumors from patients with Lynch syndrome were of the MSI-H type, resulting in over 1,000 somatic variants being identified per patient. Among these, the number of shared variants between any two tumors within a patient ranged from 7 to 118, accounting for less than 5% of the total variants. In the non-Lynch group, one patient (dou_009) had both gastric and colorectal cancers classified as MSI-H. However, only 12 variants (0.9%) were shared between the two tumors. In patient dou_019, whose colorectal cancer harbored a hotspot POLE mutation (p.Pro286Arg), the tumor exhibited a hypermutated phenotype with more than 13,000 variants identified. Yet, only a single variant was found in common with the corresponding gastric tumor.

Table 2.

Genomic similarity between paired tumors in patients with multiple primary cancers diagnosed before the age of 55 years

	Patient ida)	MSI status	Concurrencyb)	No. of intersect_variants (%)	Total no. of_assesed_variants	Similarity coefficientb)	Median coverage
							GC	CRC
Non-Lynch	dou_001	All MSS	Metachronous	0	110	0.01	192.5	99.5
	dou_008	All MSS	Metachronous	3 (2.0)	152	0.03	133.5	110.5
	dou_009	All MSI-H	Metachronous	12 (0.9)	1,379	0.01	138	126
	dou_010	All MSS	Metachronous	0	137	0.01	82	129
	dou_012	All MSS	Synchronous	1 (1.1)	89	0.02	108	88
	dou_015	All MSS	Synchronous	0	71	0.01	168	161
	dou_018	All MSS	Metachronous	3 (2.7)	110	0.04	131.5	219
	dou_019c)	All MSS	Synchronous	1 (0)	13,463	0	128	164
Lynch	dou_003	All MSI-H	Synchronous (GC:metachronous)	72 (0.9)	7,734	0.03/0.04/0.02d)	123	116/116e)
	dou_005	All MSI-H	Metachronous	7 (0.2)	4,239	0.01	133	123
	dou_006	All MSI-H	Synchronous	104 (3.5)	2,963	0.04	114	112
	dou_011	All MSI-H	Synchronous	118 (5.0)	2,338	0.05	104	130
	dou_016	All MSI-H	Synchronous	62 (1.3)	4,614	0.03/0.03/0.03d)	130	148/106.5e)
	dou_017	All MSI-H	Synchronous	54 (2.9)	1,872	0.03	109	161.5

GC, gastric cancer; CRC, colorectal cancer; MSS, microsatellite stable; MSI-H, microsatellite instability high.

^a)In total, mutational similarity could not be evaluated in five patients. In three cases (dou_002, dou_007, dou_014), whole exome sequencing data were available for only one of the two tumors—either gastric or CRC—rendering comparative analysis unfeasible. In the remaining two cases (dou_004, dou_013), tumor samples failed to pass the quality control criteria of Treeomics due to low tumor purity, and thus similarity coefficients could not be estimated;

^b)Estimated by Treeomics. 1 represents 100% of similarity;

^c)The colorectal tumor in this patient harbored a hotspot POLE mutation (p.Pro286Arg), resulting in a hypermutated phenotype despite being microsatellite stable;

^d)Pairwise comparison order among three tumors: GC–CRC1, CRC1–CRC2, and GC–CRC2;

^e)CRC1/CRC2.

Overall, while tumors with hypermutation due to MSI-H or POLE mutations tended to harbor a higher number of variants, the proportion of shared variants between tumors from the same patient remained very low—typically below 5%. Consistently, the estimated genomic similarity coefficients across tumor pairs ranged from 0 to 0.05. These findings indicate that even in hereditary cancer syndromes such as Lynch syndrome, where tumors arise from a shared germline mutation, genomic similarity between primary tumors in different organs is minimal.

Copy number analysis revealed MYC copy number gain in both the gastric and colorectal tumors of a single patient (dou_008) (Fig. 3A and B). To validate this finding, fluorescence in situ hybridization (FISH) was performed, confirming MYC amplification in both tumors (Fig. 3C and D). This patient was classified as a non-Lynch case, and each tumor was considered sporadic. Although the tumors arose in a metachronous pattern, both exhibited MYC amplification. In a previous study [10], no pathogenic germline variant of clinical significance was identified in this patient. Therefore, it remains unclear whether the presence of MYC amplification in both tumors was merely coincidental or driven by an as-yet unidentified common underlying mechanism.

Fig. 3.

Copy number analysis and fluorescence in situ hybridization (FISH) of gastric and colon cancer from one patient (dou_008). (A, B) MYC amplification was observed in whole exome sequencing copy number analysis (by cnvkit). (C, D) MYC amplification was confirmed by FISH. This observation raises the possibility that MYC could be a viable therapeutic target in both tumors within a single patient; however, additional preclinical validation is necessary. GC, gastric cancer; CRC, colorectal cancer.

DISCUSSION

Tumorigenesis in hereditary cancer syndromes, such as Lynch syndrome, is initiated by germline mutations that compromise DNA repair fidelity, typically through disruption of the MMR pathway [7]. Despite this shared oncogenic priming, these findings reveal that primary tumors arising in distinct organs within the same individual—specifically the stomach and colorectum—exhibit minimal overlap in their somatic mutational landscapes. In the present cohort, the number of shared somatic variants between any two tumors from the same patient consistently remained low, accounting for less than 5% of the total variants. This observation suggests that, even in the context of a common germline predisposition, subsequent tumor evolution is largely governed by organ-specific selective pressures and the local tissue microenvironment, rather than by expansion from a common ancestral clone.

These findings align with prior evidence highlighting the distinct carcinogenic pathways, immune landscapes, and environmental exposures encountered by gastric and colonic epithelia—for example, Helicobacter pylori in the stomach [15,16] versus microbial metabolites and bile acids in the colorectum [17]. Such divergent contexts may impose differential constraints on clonal selection, contributing to lineage-specific trajectories of somatic evolution. In support of this, we observed a predominance of TP53 and APC mutations, typically restricted to gastric and colorectal tumors, respectively. These mutually exclusive patterns further underscore the influence of tissue-specific vulnerabilities and developmental contexts in shaping early driver events. Consequently, the overall genomic similarity between synchronous or metachronous tumors within the same individual—despite shared germline predisposition—appears to be strikingly limited.

Notably, tumors harboring hypermutated phenotypes, such as those with MSI-H [4,18] or exonuclease domain mutations in POLE, demonstrated exceptionally high mutational burdens [19], yet the proportion of shared variants remained similarly low. This paradox can be attributed to the intrinsic nature of hypermutation: both MMR deficiency and POLE mutations promote replication-associated mutagenesis by impairing DNA repair and proofreading functions. MMR deficiency allows base-base mismatches and insertion-deletion loops to persist, while POLE mutations disable the 3′ to 5′ exonuclease activity of DNA polymerase ɛ, leading to the accumulation of widespread single-nucleotide substitutions. Although these mechanisms result in hypermutation across the genome, the mutational events are largely random and predominantly passenger in nature, reflecting stochastic replication errors rather than convergent selection. As such, even tumors arising contemporaneously within the same host accrue highly individualized mutational profiles. Importantly, the presence of a hypermutator phenotype does not imply increased genomic similarity, nor does it enhance the likelihood of identifying shared actionable targets. These findings underscore the importance of organ-specific molecular profiling and personalized therapeutic planning—even in patients with shared genetic predispositions or systemic mutational processes.

A particularly notable case in this study involved a patient without pathogenic germline mutations, in whom MYC amplification was identified in both gastric and colorectal tumors. This finding, confirmed by WES and validated by FISH, raises two possibilities: it may represent a coincidental convergence on a common oncogenic driver, or alternatively, may reflect the influence of an unrecognized shared predisposition, potentially involving non-coding regulatory variants or epigenetic dysregulation. MYC amplification is a well-established oncogenic event implicated in multiple malignancies and is associated with increased cellular proliferation and resistance to apoptosis [20,21]. Although rare in the present dataset, this case suggests that, in selected instances, multiple primary tumors may converge on shared molecular vulnerabilities, thereby offering potential for dual-organ targeted therapeutic strategies. However, such interpretations must be approached cautiously, given the isolated nature of this finding and the absence of corroborative germline or functional data.

This study has several limitations. First, the sample size was modest, comprising only 19 patients, with confirmed germline MMR mutations identified in a subset. The relatively small number of tumor pairs limits the generalizability and statistical power of the conclusions. Second, the present analysis relied on WES, which captures only coding regions and omits potentially significant alterations in non-coding regulatory elements, structural variants, and epigenetic modifications. Third, the temporal relationship between tumors was heterogeneous, with both synchronous and metachronous cases included, and longitudinal sampling was not available to assess clonal evolution over time. Lastly, while we validated MYC amplification via FISH, further multi-omic approaches—including transcriptomic, epigenomic, and single-cell analyses—would be necessary to elucidate the functional relevance and potential mechanisms of convergent evolution.

In summary, the present results demonstrate that primary tumors arising in different organs within the same individual—particularly in the context of Lynch syndrome or hypermutator phenotypes—exhibit largely distinct somatic genomic profiles, with minimal overlap despite a shared germline origin. These findings emphasize that tumor evolution is primarily shaped by organ-specific factors and stochastic mutational processes, rather than by clonal convergence. As such, comprehensive molecular profiling of each tumor site remains essential to guide effective, personalized therapeutic strategies. Although rare, instances of convergent alterations such as MYC amplification highlight the potential for shared therapeutic opportunities in selected cases. Nevertheless, the overarching paradigm in the management of multiple primary tumors should remain rooted in individualized, site-specific precision oncology.

Supplementary Materials

Supplementary materials are available at the Korean Journal of Clinical Oncology website (http://www.kjco.org/).