Abstract
Endometrial cancer is the most common gynaecological cancer, with incidence increasing globally. A woman’s risk of developing the disease doubles if she has a first degree relative with endometrial cancer. Although, familial risk of endometrial cancer may be impacted by environmental or lifestyle risk factors, heritability of this disease is estimated between 27% and 52%. Known genetic risk factors, including risk-associated single nucleotide polymorphisms (SNPs), and rare pathogenic variants in endometrial cancer susceptibility genes (e.g., MLH1, MSH2, MSH6, PMS2 and PTEN) explain only a small proportion of suspected hereditary cases, leaving much of the genetic risk unexplained. Rare variants, particularly copy number variants (CNVs), may account for a significant portion of this ’missing heritability’. CNVs are structural changes that alter the diploid state of the genome via DNA gains or losses and have been previously linked to cancer risk, though their role in endometrial cancer is largely unexplored. This study aims to comprehensively investigate the role rare germline CNVs play in inherited endometrial cancer risk.
Using a relatively novel CNV detection pipeline (CamCNV), rare putative CNVs were identified in 4,115 endometrial cancer cases and 17,818 country-matched controls. Characterisation of the CNVs impacting the genomes of the study cohort were similar to what has previously been reported, with an average of 2.9 deletions and 2.2 duplications located throughout the genomes of endometrial cancer cases and controls. The majority (80%) of CNVs were unique events. CNVs were annotated by assessing genomic overlap with various functional features to enable global and targeted gene burden tests, and an association study.
Global CNV burden analysis identified a 1.22-fold greater number of rare CNVs in the genomes of those who developed endometrial cancer compared to controls, increasing to 1.35-fold and 1.86-fold when considering deletions overlapping exons or completely overlapping genes intolerant to loss of function variants (pLI > 0.9), respectively. Permutation testing was employed to increase confidence that cancer status drives this difference. Additionally, individuals with CNVs overlapping mismatch repair genes or DNA repair genes had more rare CNVs across the genome on average than non-carriers. A systematic review and meta-analysis of ten studies showed no evidence of a pan-cancer risk association with CNV burden. However, there was significant heterogeneity between studies that limited this analysis.
Under three models of CNV impact (deletion, duplication and putative loss of function) a genome-wide association analysis identified 141 candidate endometrial cancer risk genes (p < 0.01). Pathway analysis of the candidate genes revealed a significant enrichment of genes involved in 16p11.2 proximal deletion syndrome. Candidate gene expression was assessed in both normal and tumour endometrial tissue, and the relationship between gene copy and mRNA in tumour tissue identified 59 candidate endometrial cancer risk genes with evidence supporting dosage sensitivity.
Putative CNVs in endometrial cancer samples were validated using NanoString. In total, nine CNVs at a range of allele frequencies (0.0046% [singleton across entire dataset] - 1.49%) including six deletions overlapping five risk loci from CNV-GWAS (16p11.2, MSH2, NPL, PMS2 and SKAP1). The expression of three selected candidate endometrial cancer risk genes (NPL, MUTYH and SALL3) was determined in two endometrial cell lines (AN3CA and EM-E6/E7/hTERT) and NPL and MUTYH siRNA knockdown models were established in the EM-E6/E7/hTERT cell lines. The knockdown of both genes did not result in cell viability changes.
To investigate the evidence supporting a role of 34 cancer predisposition genes in endometrial cancer risk, the frequency of rare CNVs overlapping the genes was compared between endometrial cancer cases and controls. The proportion of endometrial cancer cases carrying a CNV overlapping one of these genes was significantly greater than controls and CNVs overlapping MSH2, PMS2, MSH6, STK11 and EPCAM were significantly enriched in endometrial cancer cases. Additionally, rare CNVs overlapping BARD1, MLH1, EPCAM, PIK3CA and PALB2 were over four-times more frequent in endometrial cancer cases compared to controls, however this did not reach significance. Furthermore, to evaluate the clinical significance of rare CNVs, gene-specific ACMG/AMP criteria were applied to 214 CNVs overlapping four known (MLH1, MSH2, MSH6, and PMS2) and six candidate (ATM, BARD1, BRCA1, BRCA2, CDH1, and PALB2) endometrial cancer risk genes. Of the 214 CNVs, 109 were assigned codes supporting at least a strong level of evidence towards pathogenicity (assigned PVS1 or PVS1_Strong). A secondary gene burden test demonstrated that restricting CNVs to only include those with strong evidence for pathogenicity resulted in the a higher magnitude cohort difference, and a more significant enrichment for CNVs overlapping MSH2 and MSH6.
Overall, this study provided strong evidence for the involvement of rare CNVs in endometrial cancer risk and a greater understanding of the genetics underlying endometrial cancer susceptibility. The findings from this study will facilitate future research to improve population stratification of individuals with increased risk of endometrial cancer and identify candidate causal endometrial cancer genes for the development of chemopreventive drugs.