CDH1 pathogenic variants and cancer risk in an unselected patient population

Abstract

Background:

CDH1 pathogenic variants confer a markedly elevated lifetime risk of developing diffuse gastric cancer (DGC) and lobular breast cancer (LBC). The aim of this study was to evaluate the prevalence and clinical impact of CDH1 pathogenic variants in the unselected and ancestrally diverse BioMe Biobank.

Methods:

We evaluated exome sequence data from 30,223 adult BioMe participants to identify CDH1 positive individuals, defined as those harboring a variant previously classified as pathogenic or likely pathogenic (P/LP) or a predicted loss-of-function variant in CDH1. We reviewed electronic health records and BioMe enrollment surveys for personal and family history of malignancy and evidence of prior clinical genetic testing.

Results:

Using a genomics-first approach, we identified 6 CDH1 positive individuals in BioMe (~1 in 5,000). CDH1 positive individuals had a median age of 42 years (range 35 to 62 years), all were non-European by self-report, and one was female. None had evidence of either a personal or family history of DGC or LBC.

Conclusion:

Our findings suggest a low risk of DGC and LBC in unselected patients harboring a pathogenic variant in CDH1. Knowledge of CDH1-related cancer risk in individuals with no personal or family history may better inform surveillance and prophylactic measures.

Introduction:

While most cases of gastric cancer are sporadic, an estimated 1–3% are the result of an inherited cancer syndrome [1]. Germline pathogenic variants in CDH1, encoding the tumor suppressor protein E-cadherin, increase the risk for diffuse gastric cancer (DGC). Approximately 10–18% of individuals meeting clinical diagnostic criteria for hereditary DGC (HDGC) harbor a pathogenic variant in CDH1 (i.e., are CDH1 positive) [2]. Previous studies have shown that CDH1 positive individuals have a lifetime risk of DGC of approximately 42–70% in men and 33–56% in women [1], with an average age at diagnosis of 38 years [3]. In addition, CDH1 confers an elevated lifetime risk of lobular breast cancer (LBC) of 39–55% in women, and cleft lip/palate has also been reported [1]. Recently, the International Gastric Cancer Linkage Consortium (IGCLC) issued updated guidelines for the management of CDH1 positive individuals. These recommend prophylactic total gastrectomy or specialized endoscopic surveillance with multiple biopsies every 6–12 months, and for women, annual breast magnetic resonance imaging (MRI) starting at age 30 to screen for LBC [1]. These guidelines are based on the high lifetime risk of DGC and LBC in CDH1 positive individuals; however, these risk estimates are mainly derived from studies of highly penetrant families, which may be subject to ascertainment bias.

Pathogenic germline variants in CDH1 are typically identified after genetic testing based on a significant personal or family history of DGC and/or LBC [1]. However, with the introduction of multigene panel testing and whole exome or genome sequencing, pathogenic variants in CDH1 are increasingly being discovered in individuals who do not meet criteria for HDGC testing [1, 4]. Accordingly, cancer risk estimates have shifted downward [5], and may continue to decrease as more such individuals are identified. This uncertainty around disease penetrance creates a management challenge for clinicians given the lack of efficacy for gastric cancer screening [1, 4], particularly in light of the morbidity of risk-reducing gastrectomy. Reflective of these challenges, CDH1 is not currently recommended by the American College of Medical Genetics and Genomics for disclosure as a secondary (or incidental) finding to patients undergoing clinical exome sequencing [4, 6]. The present study used a genomics-first approach to identify CDH1 positive individuals in an unselected clinical care cohort, in order to gain a better understanding of CDH1-related cancer risk.

Methods:

The BioMe Biobank is an ancestrally diverse, electronic health record (EHR)-linked biobank of over 55,000 patients enrolled from ambulatory care practices across the Mount Sinai Health System in New York City. The present study population consisted of adult BioMe participants with available research exome sequence data generated through a collaboration with the Regeneron Genetics center, as previously described [7]. This study was approved by Mount Sinai’s Institutional Review Board and all BioMe participants provided informed consent.

We identified expected pathogenic variants in CDH1 among 30,223 adult BioMe Biobank participants with research exome sequence data available using a bioinformatic pipeline that annotates sequence variants using Variant Effect Predictor (VEP) and ClinVar assertions [7]. Expected pathogenic variants included variants with any pathogenic or likely pathogenic (P/LP) assertion in ClinVar, an open access catalogue of human genetic variant interpretation by genetic testing laboratories and other expert groups. All CDH1 variants with P/LP or conflicting interpretations of pathogenicity were manually reviewed in ClinVar (accessed June 2020). Expected pathogenic variants included additional predicted loss-of-function (pLOF) variants either absent from or classified as uncertain significance in ClinVar, including frameshift, stop-gain, start-loss, stop-loss, or variants disrupting canonical splice acceptor or donor sites [7]. Each expected pathogenic variant was reviewed by a clinical geneticist (N.S.A.-H.) and a certified genetic counselor (E.R.S.). Consistent with current standards for the reporting of secondary findings [6], CDH1 variants identified from research exome sequence data were not clinically confirmed for disclosure to participants.

BioMe participants harboring expected pathogenic variants in CDH1 were termed CDH1 positive individuals. Through manual review of each CDH1 positive individual’s EHR, we collected data on demographics, any evidence of personal and/or family history of malignancy up to third-degree relatives, personal history of cleft lip/palate, and prior clinical genetic counseling and testing for CDH1. EHRs were also reviewed for evidence of surveillance upper endoscopy, total gastrectomy, and for female participants, breast MRI. These data were supplemented by information from participant surveys administered on enrollment into BioMe, which included personal and family history of breast (but not gastric) cancer. Self-reported race/ethnicity of participants was also derived from these surveys, as previously described [7].

Results:

We evaluated CDH1 variants among 30,223 adult participants of the BioMe Biobank with available exome sequence data. This BioMe population has been described previously; the majority of participants are female (59%) and self-report as non-European (74%), including a large proportion of African American/African (23%) and Hispanic/Latinx (35%) participants [7].

Within the BioMe Biobank, we identified three CDH1 variants classified as P/LP in the ClinVar database (Table 1). Another variant (CDH1 c.1057G>A) had conflicting interpretations of pathogenicity in ClinVar (2 likely pathogenic and 2 uncertain significance interpretations), and has been reported previously in individuals or families with HDGC (https://www.ncbi.nlm.nih.gov/clinvar/variation/234571/). We identified two additional pLOF variants in CDH1, both of which were classified as uncertain significance in ClinVar (Table 1). In all, there were 6 (0.02%) BioMe participants harboring one of these variants in CDH1, resulting in an estimated prevalence of ~1 in 5,000.

Table 1.

CDH1 variants and characteristics of CDH1 positive individuals identified in BioMe

cDNA Positiona	Protein Position	Variant Type	ClinVar Classification	Age	Sex	Self-Reported Race/Ethnicity	Personal or Family History of Cancer (Affected Individual)
c.1137+1G>A	n/a	Splice Donor	Likely pathogenic	42	F	AA	Prostate cancer (FDR); Ovarian cancer (TDR)
c.2095C>T	p.Gln699Ter	Stop Gained	Pathogenic	54	M	Other	GE junction spindle cell sarcoma (self)
c.2195G>A	p.Arg732Gln	Missense	Likely pathogenic	35	M	ESA	None
c.1057G>A	p.Glu353Lys	Missense	Conflicting interpretations of pathogenicity (Likely pathogenic and uncertain significance)	41	M	HL	None
c.2594G>A	p.Trp865Ter	Stop Gained	Uncertain significance	37	M	AA	None
c.2647T>C	p.Ter883Gln	Stop Lost	Uncertain significance	62	M	AA	None

Abbreviations: AA, African American/African; ESA, East/Southeast Asian; HL, Hispanic/Latinx; pLOF, predicted loss-of-function; FDR, first-degree relative; TDR, third-degree relative; GE, gastroesophageal; M, male; F, Female

^acDNA and protein position provided for CDH1 ENST00000261769 (NM_004360.4); Human reference genome build 38 (GRCh38).

We evaluated the demographics and clinical characteristics of the 6 CDH1 positive individuals. Their ages ranged from 35 to 62 years, with a median age of 42 years. Five were men, and self-reported race/ethnicities were African American/African (N=3), East/Southeast Asian (N=1), Hispanic/Latinx (N=1), and other (N=1). Among the 6 CDH1 positive individuals, none had a documented diagnosis of DGC, LBC, or cleft lip/palate. Two individuals had evidence of a personal or family history of other cancer types. One individual harboring CDH1 c.2095C>T had metastatic spindle cell sarcoma of the gastroesophageal junction, histologically confirmed and diagnosed at age 50 years. Another individual harboring CDH1 c.1137+1G>A had a third-degree relative (first cousin) with ovarian cancer noted in the EHR, and a first-degree relative (father) with prostate cancer noted in their BioMe enrollment questionnaire. None of the 6 CDH1 positive individuals had evidence of prior clinical genetic counseling or testing for CDH1, and none had evidence of surveillance upper endoscopy, breast MRI, or total gastrectomy.

We compared the estimated prevalence of CDH1 positive individuals in BioMe (~1 in 5,000) to those in gnomAD (v2.1), a public database of exome and genome sequencing data from a variety of large-scale sequencing projects [8]. Among 125,748 individuals with exome sequence data in gnomAD, there were 16 individuals harboring either a ClinVar P/LP variant (N = 8) or an additional pLOF variant (N = 8) in CDH1, resulting in an estimated prevalence of ~1 in 8,000. Restricting to known ClinVar P/LP variants in both datasets, a more conservative estimate of prevalence would be ~1 in 10,000 in BioMe and ~1 in 16,000 in gnomAD.

Discussion:

We used a large-scale, population-based exome sequencing database in a diverse healthcare setting to characterize the prevalence and clinical impact of CDH1. We found that up to 1 in 5,000 individuals may harbor a pathogenic variant in CDH1. None of the six CDH1 positive individuals in our study had evidence of a personal or family history of CDH1-related cancers. One had metastatic spindle cell sarcoma of the gastroesophageal junction, which has not been described previously in association with CDH1. Consistent with recent studies from cohorts unselected for features of HDGC, these findings may suggest a lower risk of DGC and LBC in CDH1 positive individuals than previously reported [5].

CDH1 positive individuals may be identified incidentally via multigene panel testing or validated direct-to-consumer tests. Risk management recommendations have been proposed for these situations, which include annual gastric surveillance and consideration for prophylactic gastrectomy; females are recommended annual breast surveillance with consideration for prophylactic mastectomies [1]. These guidelines provide some direction to a previously unaddressed population. However, the recommendations reflect the unexpected and difficult decision these individuals face when weighing prophylactic gastrectomy, which poses concerns relating to morbidity, perioperative risks and prolonged recovery, against endoscopic surveillance, whose effectiveness at detecting early DGC is uncertain [1]. Recent studies in experienced centers using specialized endoscopic surveillance techniques show improved detection of early gastric lesions [1, 9]. However, it is not yet clear whether this approach leads to a survival benefit, and data on its efficacy in individuals with no relevant history of disease is particularly lacking [1, 10]. Findings from our study and others, which suggest a reduced penetrance in such individuals, highlight the important role of pre- and post-test genetic counseling to facilitate carefully considered patient decision-making regarding both the choice to pursue CDH1 genetic testing, and how to manage risk once a pathogenic variant is identified [4].

The prevalence of CDH1 pathogenic variants in the general population has not been well defined. Our genomics-first assessment of CDH1 in an ancestrally diverse clinical care cohort suggests a prevalence of 1 in 5,000 to 1 in 10,000 (if only known pathogenic variants are included). Previous studies of CDH1 have focused on distinct populations highly selected for DGC, with some research showing a significant presence of CDH1 variants in the Maori population of New Zealand and in a subset of people in Newfoundland [1, 2]. Our study included a relatively unselected clinical care cohort, which avoids such ascertainment bias. We also evaluated the prevalence of CDH1 expected pathogenic variants in gnomAD, a larger dataset including various disease-specific and population genetic studies. Further large-scale studies are needed to define the prevalence of CDH1 pathogenic variants in the general population.

There are limitations to this study. We retrospectively evaluated EHRs to assess for CDH1-related cancer risk. EHRs may be incomplete or inaccurate, particularly for family history information; however, we supplemented these with survey data on family history of breast cancer. Only 6 CDH1 positive individuals were identified, two of whom were under the age of 40. Given that younger individuals may not have had time to develop a CDH1-related malignancy, this could result in underestimating disease penetrance. Large copy number variants, such as multi-exon deletions and duplications, were not detectable using exome sequencing and therefore would have been missed, potentially leading to an underestimate of pathogenic variants in our cohort. On the other hand, we included some variants currently classified as conflicting or uncertain significance in ClinVar, which may in the future be reclassified as benign.

In conclusion, we identified 6 CDH1 positive individuals in an unselected EHR-linked biobank, none of whom had evidence of personal or family histories of DGC or LBC. Unselected individuals with pathogenic variants in CDH1 may reflect a population with low penetrance of HDGC in absence of a family history. If so, the benefit of prophylactic gastrectomy for many of these individuals may be limited. Considerations for risk management, including gastrectomy or surveillance via upper endoscopy, will need to balance the estimated lifetime risk of cancer with the nature of the intervention and its clinical utility. Prospective, longitudinal studies of CDH1 positive individuals with no personal or family cancer history are needed to better inform the clinical consequences associated with these variants and optimal risk management.