Unique Genomic Alterations and Microbial Profiles Identified in Gastric Cancer Patients with African, European and Asian Ancestry: A Novel Path for Precision Oncology?

Abstract

Introduction:

Gastric cancer (GC) is a heterogenous disease with clinicopathologic variations due to a complex interplay of environmental and biological factors, which may affect disparities in oncologic outcomes. Here, we characterize differences in the genetic and microbial profiles of GC in patients of African, European, and Asian ancestry.

Methods:

We identified 1,042 GC patients with next-generation sequencing (NGS) data from an institutional Integrated Mutation Profiling of Actionable Cancer Targets (IMPACT) assay and the Cancer Genomic Atlas (TCGA) group. Genetic ancestry was inferred from markers captured by the IMPACT and TCGA whole exome sequencing panels. Tumor microbial profiles were inferred from sequencing data using a validated microbiome bioinformatics pipeline. Genomic alterations and microbial profiles were compared between GC patients of different ancestries.

Results:

We assessed 8,023 genomic alterations. The most frequently altered genes were TP53, ARID1A, KRAS, ERBB2, and CDH1. Patients of African ancestry had a significantly higher rate of CCNE1 alterations and a lower rate of KRAS alterations (p<0.05), and patients of East Asian ancestry had a significantly lower rate of PI3K pathway alterations (p<0.05) compared to other ancestries. Microbial diversity and enrichment did not differ significantly across ancestry groups (p>0.05).

Discussion:

Distinct patterns of genomic alterations and variations in microbial profiles were identified in GC patients of African, European, and Asian ancestry. Our findings of variation in the prevalence of clinically actionable tumor alterations among ancestry groups suggest that precision medicine can mitigate oncologic disparities.

Mini-abstract for Abate et al., Ancestry and Gastric Cancer Genomics

Distinct patterns of genomic alterations and relative microbial variations are identified in gastric cancer patients of African, European, and East Asian ancestry. African ancestry patients had high rates of CCNE1 and low rates of KRAS alterations. East Asian ancestry patients had low rates of PI3K pathway alterations.

INTRODUCTION

Gastric cancer (GC) is the third leading cause of death worldwide, with over 1 million cases diagnosed annually.¹ GC is a heterogeneous disease with clinicopathologic variations based on anatomic site,² histopathology,³ immunogenicity,^{3, 4} and gene expression.⁵

Clinically, GC presentation varies among racial backgrounds.¹ In the U.S., the African American and Asian populations have been shown to have an earlier age of presentation and earlier peritoneal metastases when compared to non-Hispanic whites.⁶ African American patients have significantly worse overall⁷ and disease-specific survival for all stages of GC when compared to Caucasian and Asian groups.⁸ In contrast, Asian patients with GC have better overall and recurrence-free survival in both the U.S. and Asia.^{9, 10}

Racial background is also correlated with rates of lymph node metastases,¹¹ the most important predictor of overall survival and rate of recurrence in GC.¹² African American patients have a significantly higher rate of lymph node metastases when compared to other groups, independent of socioeconomic status.¹¹ In contrast, Asian patients have lower rates of lymph node metastasis when compared to non-Asian groups with similar histopathologic characteristics.^{8, 13, 14}

Differences in survival and prognosis among GC patients of varying racial demographics irrespective of socioeconomic status suggest that intratumoral biologic factors may contribute to these disparities. Recent studies have shown that the number of polymorphisms and total genes altered vary among GC patients of different racial groups.¹⁵ In addition, a prior study from our group found that African American patients have a significantly higher rate (89%) of TP53 alterations compared with Asian (56%) and Caucasian (40%) groups.¹⁶

Furthermore, another recent study from our group reported that patients of inferred African ancestry have lower rates of clinically actionable genomic alterations compared to those of other ancestral groups. Given the rise in targeted treatment modalities such as monoclonal antibodies and immune checkpoint blockade in cancer treatment, differences in targetable genomic alterations across ancestry groups may contribute to the oncologic disparities noted across various patient populations.¹⁷

In this study, we identify differences in the GC genome and microbiome of patients of inferred European, African, East Asian, South Asian, and Native American ancestry. We apply uniform sequencing and analytic methods on two independent cohorts obtained from Memorial Sloan Kettering Cancer Center and The Cancer Genome Atlas group.

METHODS

Patient Selection and Tumor Acquisition

After approval from the Memorial Sloan Kettering Cancer Center (MSK) Institutional Review Board (IRB 16-1525), MSK patients were identified from a prospectively maintained institutional database.^{18, 19} Patients with gastric adenocarcinoma who underwent gastric tumor evaluation by a targeted next-generation sequencing (NGS) at MSK between January 2010 and March 2022 were included. All patients provided written informed consent for targeted sequencing under trial protocol NCT01775072 in accordance with the ethical guidelines of the Declaration of Helsinki. Germline analysis was conducted only in patients who specifically consented to germline analysis.

Tissue for MSK-IMPACT sequencing was obtained via gastric resection or upper esophagogastroduodenoscopy (EGD) and diagnostic laparoscopy entailing tissue removal as part of treatment or diagnosis. Tissue was obtained from primary GC or GC metastasis if no primary tumor sample was available. Patients with gastroesophageal junction tumors were excluded. Sequencing was performed on formalin-fixed paraffin-embedded (FFPE) GC tissue samples. Full clinical and demographic information was collected from a prospectively maintained database.

The TCGA cohort included patients with stomach adenocarcinoma from the TCGA project²⁰ in which primary gastric adenocarcinoma tissue was sampled for next-generation sequencing (NGS) using a whole exome panel. Human genome alignment files from the TCGA project were obtained from archives following approval.²⁰

MSK-IMPACT

The MSK-IMPACT assay is a hybrid capture-based NGS platform that enables accurate and robust identification of genomic alterations in up to 505 validated genes that have been implicated in tumorigenesis using FFPE archival tissue.²¹ The assay uses tumor and matched blood sample DNA to identify clinically relevant somatic mutations, structural variants, and copy number alterations in all coding regions and select introns of cancer-related genes in patients with solid tumors. MSK-IMPACT is performed in Clinical Laboratory Improvement Amendments (CLIA)-accredited laboratories.²¹

Ancestry Inference

Genetic ancestry was inferred from MSK-IMPACT data as previously described.¹⁷ Briefly, for each patient, using data from the 1000 genomes project²² as reference, and extracted genotypes for select autosomal bi-allelic common single nucleotide polymorphism (SNP) markers from MSK-IMPACT data as input, we ran ADMIXTURE²³ v1.3 in supervised mode to estimate ancestral proportions of African (AFR), European (EUR), East Asian (EAS), Native American (NAM) and South Asian (SAS) populations. If the fraction of a patient’s SNP profile from any single ancestral population was ≥80%, the patient was assigned that population label. Remaining patients were labeled as admixed (ADM). The admixed group was used as internal validation for the MSK cohort. Within the admixed group, predominant ancestry was assigned if a patient’s SNP profiles had markers from a single population ≥50% and no other ancestry population >30%.

For TCGA data, we used the ancestry calls published by Carrot-Zhang et al., whereby samples with ≥80% chromosome-level consensus calls that matched the genome-wide call were assigned to an ancestral population.²⁴

Driver Gene and Pathway Alterations Analysis

Oncogenic drivers were identified and variants of unknown significance excluded using OncoKB, a precision oncology knowledge base.²⁵ Microsatellite instability (MSI) status was determined using the MSIsensor algorithm, which calculates the percentage of microsatellite loci covered by the MSK-IMPACT assay that are unstable in the tumor as compared with the patient’s matched normal sample. Samples with a score of ≥10 were classified as MSI-high.²⁶ Tumors were characterized as genomically stable (GS) if the tumor purity was ≥20% and the fraction of the autosomal genome affected by DNA copy number alterations of any kind (FGA) was ≤5%. We classified tumors that were neither Epstein Barr virus-positive (EBV+), MSI, or GS as the chromosomal instability (CIN) molecular subtype.²⁷ Tumor mutation burden and FGA were calculated as previously described.²⁸ Pathway analyses were conducted using 11 canonical cancer-related signaling pathways as defined by the TCGA PanCancer Atlas Project: p53, cell cycle, Hippo, Myc, Notch, NRF2, PI3K, receptor tyrosine kinase (RTK)/RAS/MAPK, TGFβ, Wnt, and DNA damage response (DDR).^{29, 30} A tumor was considered altered in a specific pathway if ≥1 gene belonging to that pathway was altered.

Germline Analysis

For MSK patients who specifically consented, DNA from blood samples was evaluated for alterations in up to 90 hereditary cancer predisposition genes. Germline variants were called as previously described.^{31, 32} Identified variants were independently assessed and manually curated according to the American College of Medical Genetics and Genomics guidelines.³³

Site of Metastasis Analysis

Data were collected and processed for site of metastasis analysis as in MSK-MET.³⁴ Briefly, clinical data were retrieved from the institutional electronic health records (EHR) database. Metastatic events were extracted from a combination of the pathology report of the sequenced samples and International Classification of Diseases (ICD) billing codes. The anatomic location of the sequenced samples is described in sample pathology reports. Metastatic events from the sample pathology report and ICD billing codes from the EHR were systematically mapped to a curated list of 10 organs as previously described.³⁴ Metastatic burden was defined as the number of distinct organs (excluding lymph nodes) affected by metastases throughout a patient’s clinical course (ranging from 1 to 15 in the present study). Patients with more than 6 affected organ sites were grouped for analyses of metastatic burden.

Microbial Bioinformatics Pipeline

Microbial profiles were determined from tumor NGS of GC samples as previously described.³⁵ Briefly, off-target captured reads, usually 8–12% of the typical 8 million reads sequenced using IMPACT, were separated on a binary alignment (BAM) file using samtools (v1.9) and saved as fasta files. Off-target reads were analyzed using the basic local alignment search tool (BLAST) algorithm blastn (v2.5.0). BLAST output files were mapped to taxonomy IDs from the National Center for Biotechnology and Information (NCBI) database using ClassifyBlast from the KronaTools package.³⁶ Reads for each genus and species (positivity threshold of ≥2 reads) were quantified using an internally developed R script (executed using R v3.5.1). Bacteriophages were excluded from the analysis. The same method was used to analyze the microbiome of the TCGA cohort from downloadable BAM files.⁵

Microbial diversity was quantified using the inverse Simpson’s index^{37, 38} and abundance was calculated by determining the proportion of microbes present per sample, with a positivity threshold of ≥2 microbial reads.

Data Classification and Statistical Analysis

Categorical variables were compared using a two-sided Fisher’s exact test and continuous variables were compared using the Mann-Whitney U test. Testing for multiple hypotheses was corrected using the Benjamini-Hochberg method and all results with q <0.1 were considered statistically significant. All statistical analyses were performed using R software (version 3.6.3).

RESULTS

Patient Clinicopathologic Characteristics

A total of 1,042 patients for whom NGS data on their gastric adenocarcinoma specimen were included. The MSK cohort included 602 patients of which 341 (57%) were of European, 72 (12%) East Asian, 33 (5.5%) African, 20 (3.3%) South Asian, and 11 (1.8%) Native American inferred ancestry (Table 1, Supplemental Table 1). A total of 121 (20%) patients’ ancestry was categorized as an admixture. In the TCGA cohort, 397 of the 440 patients had ancestry inference available. Of these, 292 (74%) were of European, 90 (23%) East Asian, and 8 (2.0%) African ancestry (Table 1, Supplemental Table 2). The remaining 7 (1.8%) patients’ ancestry was categorized as an admixture.

Table 1.

Clinicopathologic characteristics of gastric cancer patients from the Memorial Sloan Kettering (MSK) and The Cancer Genome Atlas (TCGA) cohorts.

Characteristic	n	MSK, n = 602	TCGA, n = 440	p-value
Age	1,031	63 (52–72)	67 (58–73)	<0.001
Sex	1,040			0.016
Female		257 (43%)	156 (35%)
Male		343 (57%)	284 (65%)
Ancestry	995
Admixed		121 (20%)	7 (1.8%)
African		33 (5.5%)	8 (2.0%)
East Asian		72 (12%)	90 (23%)
European		341 (57%)	292 (74%)
Native American		11 (1.8%)	0 (0%)
South Asian		20 (3.3%)	0 (0%)
Procedure	1,006
Biopsy		376 (66%)	-
Resection		190 (34%)	440 (100%)
Sample type	1,042
Metastasis		171 (28%)	-
Primary		431 (72%)	440 (100%)
Neoadjuvant chemotherapy	970	137 (26%)	-
Subtype	897			<0.001
CIN		324 (62%)	223 (59%)
EBV		-	30 (8.0%)
GS		146 (28%)	50 (13%)
MSI		51 (9.8%)	73 (19%)
T stage	634			0.2
T <3		65 (32%)	115 (27%)
T ≥3		139 (68%)	315 (73%)
N stage≥1	624	137 (68%)	291 (69%)	0.8
Metastasis	809	202 (52%)	28 (6.7%)	<0.001

Data presented as n (%) for categorical variables and median (IQR) for continuous. P values derived by Pearson’s Chi-squared test and Wilcoxon rank-sum test, respectively.

CIN, chromosomal instability; EBV, Epstein Barr virus-associated; GS, genomic stability; MSI, microsatellite instability.

The MSK cohort included 376 (66%) gastric cancer samples obtained from EGD biopsy and 190 (34%) obtained from gastric resection. A total of 137 (26%) samples were obtained following neoadjuvant chemotherapy administration. A total of 431 (72%) samples were obtained from the primary site while 171 (28%) were obtained from a metastatic site (Table 1). The TCGA cohort only included patients who had not received prior neoadjuvant chemotherapy whose samples were obtained from resection of the primary tumor. Compared to the TCGA cohort, MSK patients were younger, had a higher rate of M1 disease, and included more female patients (Table 1).

Self-Identified Race and Ancestry Concordance

Inferred genetic ancestry correlated with available self-reported race and provided a more granular level of detail than the categories available for self-identified race, particularly for the race category “Black,” “Asian,” and for unreported race categories (Fig. 1A). Of the patients who self-reported as Asian, 73.1% were of East Asian, 16.1% South Asian, and 9.7% admixture inferred ancestries. Of the self-reported white individuals ,84.1% had European, 0% East Asian, and 14.1% admixture inferred ancestries. Of the self-reported Black or African American individuals, 58.3% had African, 0% South Asian, and 39.6% admixture inferred ancestries. Of the patients with no self-identified race selection, 23.4% had European, 5.2% East Asian, 5.2% South Asian, and 50.6% admixture inferred ancestries. Within the TCGA cohort, of the patients that were self-reported black, 48% were of African ancestry and 52% had admixed ancestry (Fig. 1A). All patients with self-reported Asian race had East Asian ancestry. Of the patients with self-reported white race, 99% were of European ancestry and 1% of African ancestry.

Figure 1. Ancestry concordance with self-identified race and comparison of molecular subtype among ancestry groups.

(A) Distribution of ancestry within each racial category in the MSK (left) and TCGA (right) cohorts. (B) Molecular subtype. CIN, chromosomal instability; GS, genomic stability, MSI, microsatellite instability; POLE, polymerase ε; EBV, Epstein Barr virus positive

Distribution of Gastric Cancer Molecular Classification

The MSK cohort included a larger proportion of GS gastric cancers compared with the TCGA cohort (28% vs. 13%, p<0.001) (Table 1). In the MSK cohort, there was a significantly higher proportion of genomically stable gastric cancer in the East Asian compared to the African ancestry group (35.0% vs. 12.0%, p=0.019) (Fig. 1B, Supplemental Table 1). While patients of East Asian and South Asian ancestry had a relatively lower proportion of MSI gastric cancer compared to the other groups in the MSK cohort, this difference was not significant (Fig. 1B, Supplemental Table 1). The distribution of molecular classifications did not differ among ancestry groups in the TCGA cohort (Fig. 1B, Supplemental Table 2).

Driver Gene and Pathway Alterations

Neither tumor mutational burden nor fraction of the genome altered differed significantly among ancestry groups in either the MSK or TCGA cohorts (Fig. 2A–B). The most frequently altered genes in the entire cohort were TP53 (53%), ARID1A (20%), ERBB2 (13%), KRAS (13%), and CDH1 (11%). The most common types of genomic alterations were mutations (n=5602), followed by copy-number alterations (n=1005), and then fusions (n=168).

Figure 2. Comparison of major genomic features among ancestry groups.

(A) Tumor mutation burden. (B) Fraction of the genome altered.

In the MSK cohort, CCNE1 alterations were more frequent among patients of African vs. European inferred ancestry in all patients (15% vs. 4%, p=0.02, q=0.09) and those with non-MSI tumors (17% vs. 4%, p=0.02, q=0.09) (Fig. 3A, Fig. 4A). Compared with the European ancestry group, patients of African ancestry had a lower frequency of KRAS alterations in all patients (3% vs. 16%, p=0.04, q=0.13) and those with non-MSI tumors (0% vs. 15%, p=0.02, q=0.09) (Fig. 3A, Fig. 4A). Compared to the European cohort, East Asian patients had a lower prevalence of PI3K pathway alterations (6% vs 19.4%, p=0.01, q=0.07) (Fig. 4A). While a higher proportion of TP53 alterations were identified in the African ancestry group (67%) compared to European (53%), East Asian (51%), and South Asian (50%) groups, this difference was not statistically significant (Fig. 3A, Fig. 4A).

Figure 3. Genomic alterations according to ancestry.

Oncoprints illustrating each individual’s alterations, grouped by ancestry. (A) MSK, (B) TCGA.

Figure 4. Frequency of oncogenic alterations according to ancestry and internal validation.

Comparison of frequency of alterations in (A) individual genes and (B) oncogenic pathways among ancestry groups. (C) Comparison of frequency of alterations among patients of admixed ancestry, grouped by predominant (>50%) ancestry group.

In the TCGA cohort, SMAD4 alterations were more frequent among patients of East Asian vs. European inferred ancestry (20% vs. 10%, p=0.04, q=0.39), as were APC genomic alterations among patients of African vs. East Asian ancestry (40% vs. 5%, p=0.04, q=0.68) (Fig. 3B, Fig. 4B). While compared to the European cohort, patients of African ancestry had higher proportions of CCNE1 (20% vs. 11%) and TP53 (60% vs. 49%) alterations, and a lower proportion of KRAS alterations (0% vs. 13%), these differences were not statistically significant (Fig. 3B, Fig. 4B). Similarly, while patients of East Asian ancestry had a lower prevalence of PI3K pathway alterations compared with European patients, this difference did not reach significance (Fig. 4B).

We next sought to determine whether the genomic patterns we observed extended to comparisons among patients of admixed ancestry. Among the 121 MSK admixed patients, 36 were grouped as predominantly (≥50%) European, 12 as predominantly African, 1 as predominantly East Asian and 2 as predominantly South Asian (Fig. 4C). Within this cohort, CCNE1 alterations were more frequent among patients of majority African vs. European ancestry was significant (50% vs. 6%, p=0.002, q=0.016). While patients in the African admixture group had low rates of KRAS alteration (8%) and high rates of TP53 alterations (67%), rates of these alterations did not differ significantly among ancestry groups (Fig. 4C).

Germline Alterations

The MSK cohort included data on germline variants for 376 patients, of whom 216 were of European, 38 East Asian, 18 African, and 13 South Asian inferred ancestry (Fig. 5A–B). Germline alterations associated with cancer predisposition were observed at similar rates among patients of European (25.9%) and East Asian ancestry (23.7%), while no alterations in known hereditary cancer predisposition genes were identified in any of the 18 patients of African ancestry (Fig. 5A–B).

Figure 5. Inheritance of gastric cancer-associated alterations and association of metastasis with genomic alterations according to ancestry.

(A) Percentage of patients with germline analysis. (B) Percentage of patients with germline alterations in known hereditary cancer predisposition genes. (C) Comparison of frequency of metastasis among ancestry groups. (D) Frequency of metastasis to major organ sites compared among ancestry groups. (E) Frequency of oncogenic alterations in individual genes among patients with metastasis to each site according to ancestry.

Frequency of Metastasis Sites

Evaluation of metastatic frequency revealed higher rates of lung (46.2% vs. 25.6%, p=0.04), and CNS or brain (26.9% vs. 10.2%, p=0.04) metastasis in patients of African ancestry compared to European and East Asian ancestry groups, respectively (Fig. 5D). TP53 alterations were associated with increased rates of metastasis across all sites, as were CCNE1 amplifications with peritoneal metastasis in the European and African ancestry groups (p=0.001, q=0.01). Patients of African ancestry had high CCNE1 enrichment and depletion of KRAS alterations across all metastatic sites (Fig. 5E).

Microbial Diversity and Enrichment

Evaluation of differences in microbial diversity and composition across ancestry groups revealed a non-significantly lower alpha diversity in patients of African ancestry and non-significantly greater diversity in patients of East Asian ancestry compared with other ancestral groups (Fig. 6A), We also found that fewer patients of East Asian and South Asian ancestry had samples positive for Selenomonas and Fusobacterium, and more patients of African ancestry had samples positive for Helicobacter, Lactobacillus, Prevotella Corynebacterium, Sphingomonas and Pseudomonas compared with those of European ancestry (Fig. 6B, Supplemental Table 3).

Figure 6. Variation in microbial profiles among ancestry groups.

(A) Microbial diversity. (B) Proportion of samples in which the indicated bacterial genera were detected by targeted tumor sequencing.

DISCUSSION

In this study, we compared the genomic and microbial landscape of patients of European, East Asian, African, and South Asian inferred ancestry in the largest gastric cancer patient population to date. We found significantly more frequent genomic alterations in CCNE1 and less frequent alterations in KRAS in patients of African ancestry, and less frequent PI3K pathway alteration in patients of East Asian ancestry compared with patients of European ancestry. These findings were validated in an independent TCGA cohort and an internal group of patients of admixed ancestry where similar patterns were identified. We also identified distinct correlations between the frequency of site-specific metastasis and oncogenic alterations across ancestry groups. Given the rise in targeted therapies in oncologic treatment, and the noted lower rates of clinically actionable alterations in minority groups,¹⁷ our findings of distinct patterns of genomic alterations and microbial profiles across ancestry groups provides novel insight into how future trials and treatment strategies may help mitigate disparities in outcomes among ancestral groups.

We found higher rates of CCNE1 alterations in patients of African ancestry in two independent cohorts, supported by our internal validation. CCNE1 encodes the cyclin E1 protein, which, in complex with CDK2 kinase, enables S phase promotion.³⁹ CCNE1 alterations are associated with genomic instability, whole genome doubling, and copy number losses that promote oncogenesis and contribute to resistance to cytotoxic and targeted therapies.^40–42 Currently, there are no FDA-approved drugs targeting CCNE1.²⁵ However, PKMYT1 inhibition using RP-6306, an orally bioavailable and selective inhibitor that shows single-agent activity and induces durable tumor regression when combined with gemcitabine, has been identified in cellular models as a promising therapeutic strategy for CCNE1-amplified cancers.⁴⁰ As a result, RP-6306 is currently in phase 1 clinical studies as monotherapy (NCT04855656) and in combination with gemcitabine (NCT05147272) or FOLFIRI (NCT05147350) for advanced solid tumors. As patients of African ancestry have fewer clinically actionable targets for most cancers, our findings suggest that further investigation of CCNE1-targeted agents and enrollment of gastric cancer patients of African ancestry into these clinical trials may be of great benefit for this population.

In our gastric cancer cohort, we found that patients of African ancestry had significantly fewer KRAS alterations compared with those of European decent in both the MSK and TCGA cohorts. KRAS is one of 3 closely related RAS genes that encode 4 proteins that regulate proliferation and survival.⁴³ KRAS mutations are well-established biomarkers of resistance to anti-EGFR therapies such as cetuximab and ganitumab.^{44, 45} Our findings of lower KRAS alteration rates in patients of African ancestry groups suggests that these patients may be less likely to display resistance to anti-EGFR-directed therapies. In contrast to our finding, a recent study evaluating patients with colorectal cancer found KRAS alterations to be more common in patients of African ancestry compared with those of European ancestry.⁴⁶ This distinction in patterns of genomic alterations within the same ancestry group but different tumors suggests that ancestry-specific biomarkers may vary between different cancer types, and that investigations of the relationships between tumor features and ancestry should be cancer-specific.

Our finding that PI3K pathway alterations were significantly less frequent in patients of East Asian ancestry in two independent gastric cancer cohorts has implications for current trials. Alterations in the PI3K pathway are one of the most common aberrations that drive tumorigenesis, and therapies targeting this pathway are currently being investigated in multiple tumors.⁴⁷ In gastric cancer, two trials evaluating the combination of capecitabine and the mTOR (upstream of PI3K) inhibitor everolimus have reached phase II.^{48, 49} Our findings of low rates of PI3K pathway alterations in the East Asian group suggests that baseline patient characteristics including patient race or ancestry could be a confounder in the evaluation of response to these therapies. In addition, because these therapies pose toxicity risks including hematological toxicity, lymphopenia, pulmonary toxicity and even death,^{48, 49} if our results are validated and reflected as lower likelihood of response to PI3K-targeted therapies among patients of East Asian ancestry, these patients could be spared associated toxicity through ancestry-specific precision oncology.

Among patients with metastasis, CCNE1 amplification was significantly more frequent among patients of African ancestry across all metastatic sites (Fig. 5E). This observation is in agreement with a prior study reporting an association of CCNE1 amplification with liver metastasis in gastric cancer⁵⁰ and suggests that the association may extend beyond the liver. Furthermore, the lower frequency of KRAS alterations noted in patients of African ancestry across all metastatic sites compared with other ancestral groups suggests that KRAS alterations may not drive metastasis in this patient population.

Germline analysis revealed high rates of inheritance of the hereditary cancer predisposition genes among patients of European ancestry and progressively lower rates among those of East Asian and South Asian ancestry, while no patients of African descent had risk-associated germline variants. Screening for inheritance of genomic alterations associated with risk for gastric cancer, such as CDH1, has allowed early identification of patients most likely to develop aggressive cancer.⁵¹ Prophylactic gastric resection in this setting has survival benefit for these patients.^{51, 52} As black patients have a high incidence of most cancers, the lack of known risk-associated germline variants in patients of African ancestry suggests that further investigation into inherited risk-associated alterations in patients of non-European ancestry necessary to understand germline variants and promote equitable cancer genetic counseling and prevention in gastric cancer.

Finally, we found no siginficant differences in microbial diversity (Fig. 6A) or frequency of microbes across ancestry groups (Fig. 6B). Interestingly, East Asian and South Asian patients in our cohort, who had relatively low rates of MSI gastric cancer, also had relatively low proportions of Selenomonas and Fusobacterium, which are known to be elevated in both MSI-high gastric³⁵ and colon cancer.⁵³ The lack of significant differences in microbial profiles across ancestry groups in a patient population from the United States suggests that intra-tumoral microbial differences are driven by environmental and geographic factors. Further studies using a larger sample size are necessary to corroborate our findings.

While we believe this is the most significant study of genomic alterations among ancestry groups in gastric cancer to date, we acknowledge its limitations. The study is retrospective and thus our data may reflect systemic and selection bias. However, all clinical variables were acquired prospectively without knowledge or anticipation of this study, and no variables were altered when doing the analysis. As the two cohorts differ in terms of clinical features (i.e., the MSK cohort is obtained from patient samples utilized in clinical practice, including those with prior therapy, and samples obtained from biopsy and metastatic sites, while the TCGA cohort included only treatment-naïve primary tumors), our analysis demonstrates reproducibility across very different patient populations.

CONCLUSION

In conclusion, we identified distinct patterns of genomic alterations and tumor microbial composition across ancestry groups. We found that alterations in CCNE1 were more frequent and alterations in KRAS were less frequent in patients of African ancestry, while PI3K pathway alterations were less frequent in patients of East Asian ancestry compared with patients of European ancestry. Given the rise in targeted therapies in cancer treatment, and the noted lower frequency of clinically actionable genomic alterations in minority groups, our findings of differences in rates of specific genomic alterations among ancestry groups provide novel insight into how future trials and treatment strategies can be used to mitigate oncologic disparities through the use of precision oncology.