Prevalence of Germline Alterations on Targeted Tumor-Normal Sequencing of Esophagogastric Cancer

Key Points Question What are the prevalences of likely pathogenic or pathogenic (LP/P) germline variants in esophageal, gastroesophageal junction, and gastric cancers? Findings In this cross-sectional study of 515 patients, LP/P germline variants occurred in approximately one-fifth of patients with gastric cancer. Those with early-onset esophagogastic cancer (≤50 years of age at diagnosis) were significantly more likely to harbor an LP/P germline variant. Meaning The results of this study support germline testing in patients with gastric cancer and early-onset esophagogastric cancer.


Introduction
Esophagogastric cancer is one of the most common cancers globally, with more than 1 million cases diagnosed annually. 1 Although most esophagogastric cancers are sporadic, 2 gastric cancer can arise within the context of heritable cancer predisposition syndromes, specifically hereditary diffuse gastric cancer, hereditary breast and ovarian cancer, and Lynch syndrome. 3 However, the prevalence of pathogenic germline variants in patients with esophagogastric cancer remains poorly defined because only individuals who present with features of these heritable cancer syndromes are routinely referred for genetic testing. 4 We retrospectively reviewed patients with esophagogastric cancer who underwent targeted sequencing of tumor tissue and blood (to provide a germline comparison) using Memorial Sloan Kettering-Integrated Mutation Profiling of Actionable Cancer Targets (MSK-IMPACT), a nextgeneration sequencing platform. We investigated whether the prevalence of likely pathogenic/ pathogenic (LP/P) germline variants differed according to tumor location, age of diagnosis, and race/ ethnicity and evaluated the association between the presence of germline variants and response to platinum-based chemotherapy regimens.

Patients
In this cross-sectional study, we identified all patients with esophagogastric cancer whose tumors and matched germline DNA from blood were analyzed by MSK-IMPACT at the Memorial Sloan Kettering Cancer Center (MSKCC) from January 1, 2014, to December 31, 2019; in general, testing was offered to patients who were followed up at MSKCC and who would potentially be candidates for an experimental strategy based on the identification of an actionable somatic alteration. Starting in May 2015, patients were prospectively offered secondary germline analysis after providing written informed consent for tumor genetic analysis in the context of an MSKCC institutional review boardapproved protocol. The identities of those who provided consent only for somatic testing were anonymized before analysis.
Patients underwent anonymized or identified germline analysis, the latter in a prospective study starting in May 2015, with results returned to the patient. For the anonymized analysis, only patients who consented to the MSKCC institutional review board-approved prospective study of tumor genomic analysis via MSK-IMPACT were analyzed. A retrospective protocol for anonymized assessment of germline cancer susceptibility using normal DNA was approved by the MSKCC institutional review board. In this cohort, patient identities were anonymized after demographic data were summarized before other analyses. Institutional procedures for anonymization were followed to prevent inadvertent patient identification, with binning of clinical data into categorical variables of at least 3 individuals per bin. The identified germline analysis consisted of patients who consented to secondary germline analysis. For all patients, pathology records were reviewed and confirmed at MSKCC. Demographic and clinical data were manually extracted from electronic medical records by 2 clinical research coordinators (Y.T. and K.C.); data were extracted into standard fields in an Excel spreadsheet (designed by G.Y.K. and Y.K.). There was no a priori study hypothesis. Any ambiguous clinical data were adjudicated (G.Y.K.). Race/ethnicity was based on self-reporting by patients and was extracted because of potential differences in germline variants based on race/ethnicity.

Germline Analysis
Normal nontumor DNA from blood was analyzed using a 76-or 88-gene panel (eAppendix in the Supplement), including all cancer-predisposing genes identified by the American College of Medical Genetics and Genomics (ACMG) guidelines. 5 DNA was sequenced and variants reported as described previously. 6 Variant rediscovery was performed using standard germline variant-calling methods with stringent quality controls. Variant interpretation relied on an initial pass by the automated Overall survival (OS) and progression-free survival (PFS) were analyzed among 304 identified patients who were treated with first-line platinum therapy. Both OS and PFS were calculated from the date of first-line platinum therapy initiation until date of death for OS or first progression or death, whichever occurred first, for PFS and compared between subgroups using the log-rank test.

JAMA Network Open | Oncology
The end date for follow-up was July 31, 2020.
All statistical analyses were conducted using R, version 3.6.0 (R Foundation for Statistical Computing). P values were calculated using a 2-sided test with a cutoff of P < .05 to indicate statistical significance.

Germline Variant Prevalence
Among the 515 patients, the tumor was located in the esophagus in 161 patients (31.3%), the GEJ in 111 patients (21.6%), and the stomach in 243 patients (47.2%) ( locations, in line with the presumption that most of these variants were incidental and did not contribute to cancer development in these patients. We also assessed germline variant prevalence in patients with early-onset cancer, namely in patients diagnosed with esophagogastric cancer at 50 years or younger, a cutoff selected based on Surveillance Epidemiology and End Results data indicating that the age of 50 years is more than 1 SD below the mean age at esophagogastric cancer diagnosis. 9 In the overall cohort, 29 patients 50 years or younger (21.0%; 95% CI, 14.5%-28.8%) vs 52 patients older than 50 years (13.8%; 95% CI, 10.5%-17.7%) harbored a germline alteration (P = .046). After stratifying by cancer location, although      11 Only 4 (28.6%) were patients diagnosed with early-onset disease at 50 years or younger. Figure 2B is a representative pedigree from a patient with very early-onset gastric cancer and a BRCA2 alteration. Tumor sequencing revealed a somatic alteration in BRCA2, a presumed second hit, suggesting that the germline event was likely a driver of tumorigenesis.

Inheritance of Variants in Frequently Altered Genes
LP/P variants in CDH1 were identified in 6 patients with gastric cancer. Figure 2C shows the pedigree of a patient with gastric signet ring cell adenocarcinoma who did not meet clinical criteria for CDH1 germline analysis but nonetheless harbored a pathogenic CDH1 alteration, presumably inherited from the paternal lineage, without a family history suggestive of hereditary diffuse gastric cancer.
Given potential clinical implications for treatment response, we also evaluated the prevalence of LP/P germline variants in genes associated with DNA damage repair (DDR) and found that 27 patients overall (5.2%) and 16 patients (6.6%) with gastric cancer, 5 (4.5%) with GEJ cancer, and 6 (3.7%) with esophageal cancer harbored variants in these genes.

Association With Treatment Outcomes
We evaluated whether patients with variants in homologous recombination deficiency (HRD) and CI, 43%-55%). A, ATM pathogenic alteration in a 29-year-old woman with stage IV gastric signet ring cell adenocarcinoma treated with folinic acid, fluorouracil, and oxaliplatin (FOLFOX) with treatment discontinued after 6 months because of toxic effects. Overall survival was 9.5 months. The identification of an ATM pathogenic alteration represented an incidental finding in a relative with familial gastric cancer. B, BRCA2 alteration in a 35-year-old woman with stage IV gastric cardia cancer to the lymph nodes with complete response to epirubicin, oxaliplatin, and capecitabine chemotherapy that lasted for 16 months. Overall survival was 53.5 months. In addition to the germline BRCA2 alteration, a somatic BRCA2 alteration was also identified, presumed to be the second hit in the context of a cancer caused by the germline mutation. C, Germline CDH1 alteration in a 45-year-old man with stage IV gastric signet ring cell carcinoma treated with FOLFOX for 9 months before progression. He remained alive 10.4 months into treatment. A germline CDH1 alteration, presumably inherited from the paternal lineage without familial gastric cancer, was identified. Because he did not meet genetic testing criteria for CDH1, it is considered an incidental finding. Square indicates male; circle, female; bracket around a circle, adopted individual; slash, deceased; black upper left portion of a square or circle, canceraffected individual; arrowhead, a family member who underwent matched tumor and germline sequencing; black asterisk, germline-positive status; blue asterisk, presence of a second somatic alteration; pink asterisk, obligate carrier; and gold asterisk, confirmed germline-negative status.

JAMA Network Open | Oncology
We also compared PFS and OS between patients with or without BRCA1/2 germline alterations who received platinum-based treatment and found no difference. However, our study may have been underpowered to detect such differences.

Discussion
This cross-sectional study found that pathogenic germline variants were significantly more common in patients with gastric compared with esophageal cancer, with carcinogenesis in the latter more closely linked to exogenous or environmental risk factors. Specifically, the study found not only a higher prevalence of actionable germline variants but also a 5-fold greater prevalence of highpenetrance variants in patients with gastric cancer compared with esophageal cancer (10% vs 2%).
The prevalence of germline LP/P variants in GEJ tumors was intermediate between those in gastric and esophageal cancers. The Cancer Genome Atlas analysis of esophageal cancers found a gradual transition of molecular subtypes from the distal esophagus through the GEJ into the proximal stomach. 12 In addition, GEJ cancers are heterogeneous because ascertaining the true epicenter of a tumor at the esophagogastric junction can be impossible, even in patients who undergo endoscopy. In contrast, the prevalence of a high-penetrance germline variant in the esophageal cancer cohort was less than 2%, which, on the basis of these data, does not justify routine testing in such patients.
Because the prevalence of high-penetrance variants was between these two values in the GEJ population (4.5%) and given the challenge in accurately classifying such tumors as belonging exclusively to the esophagus or the stomach, germline testing for patients with GEJ tumors can also be considered.
The survival analysis of patients with germline variants in HRD and DDR pathways and in particular the BRCA1/2 genes did not support enhanced responsiveness to treatment with platinumcontaining chemotherapy regimens, whose mechanism of action involves damaging DNA. This finding contrasts with studies in pancreatic cancers 25 and likely results from lack of power to detect

Limitations
This study has limitations, including its retrospective nature, a median age of diagnosis 10 years younger than the general US population, and a higher-than-average representation of patients with metastatic disease at the time of diagnosis (approximately two-thirds vs half). 2 This high prevalence is related to the fact that next-generation sequencing somatic and germline alteration testing have, up to now, been preferentially performed in patients with metastatic disease, in part to identify actionable alterations for targeted therapies. The patient population treated at MSKCC in the New York area also includes higher-than-average proportions of patients with Ashkenazi Jewish ancestry or Asian ethnicity, so these results may not be applicable to other patient populations. In addition, CTNNA1 (OMIM 116805), a recently implicated gene associated with inherited diffuse gastric cancer, was not included on our germline panel, possibly leading to underestimation of overall germline variant prevalence.

Conclusions
This large retrospective analysis suggests that pathogenic germline variants may be significantly more common in gastric than in esophageal cancer, with GEJ tumors having an intermediate prevalence, as well as in patients with esophagogastric cancer diagnosed at 50 years or younger, although this observation was of borderline statistical significance and needs to be validated.
Because many of the cancer predisposition syndromes identified have significant implications for cancer surveillance and risk reduction in at-risk relatives and could influence therapy selection in patients affected by cancer, germline testing should be considered in patients with GEJ or gastric cancer and any patient with esophagogastric cancer diagnosed at 50 years or younger. Identifying germline HRD and DDR pathway alterations also permits prioritization of standard platinum-based therapy and experimental PARP inhibitors for these patients. Finally, a large number of LP/P variants in ATM were identified in this cohort, especially in patients with early-onset disease, and this observation warrants confirmation in other data sets.