Assessment of genetic susceptibility to multiple primary cancers through whole-exome sequencing in two large multi-ancestry studies

Abstract

Background: Up to one of every six individuals diagnosed with one cancer will be diagnosed with a second primary cancer in their lifetime. Genetic factors contributing to the development of multiple primary cancers, beyond known cancer syndromes, have been underexplored.

Methods: To characterize genetic susceptibility to multiple cancers, we conducted a pan-cancer, whole-exome sequencing study of individuals drawn from two large multi-ancestry populations (6429 cases, 165,853 controls). We created two groupings of individuals diagnosed with multiple primary cancers: (1) an overall combined set with at least two cancers across any of 36 organ sites and (2) cancer-specific sets defined by an index cancer at one of 16 organ sites with at least 50 cases from each study population. We then investigated whether variants identified from exome sequencing were associated with these sets of multiple cancer cases in comparison to individuals with one and, separately, no cancers.

Results: We identified 22 variant-phenotype associations, 10 of which have not been previously discovered and were significantly overrepresented among individuals with multiple cancers, compared to those with a single cancer.

Conclusions: Overall, we describe variants and genes that may play a fundamental role in the development of multiple primary cancers and improve our understanding of shared mechanisms underlying carcinogenesis.

Keywords: Germline genetics; Multiple primary cancers; Pleiotropy; Whole-exome sequencing.

Fig. 1

Cancer diagnosis pairs present in the combined study populations. Circos plot describing the pairs of first and second cancer diagnoses with at least 25 cases present in Kaiser Permanente Research Bank and the UK Biobank study populations combined. Each connection reflects the number of cases with both of the linked primary cancers, where the color of the line shows the first cancer site diagnosed

Fig. 2

Germline single-variant association results for multiple primary cancers combined or grouped by organ site. Suggestive (p < 5 × 10⁻⁶) germline variant associations with multiple cancer phenotypes versus cancer-free controls (n = 165,853) following a fixed-effects meta-analysis of Kaiser Permanente Research Bank and UK Biobank WES data. Associations were detected for any 2+ primary cancers (n = 6429) and with groups of cases defined by a shared index cancer, at any time point, plus any other cancer diagnosis: melanoma + (n = 1443), prostate + (n = 1977), breast + (n = 1874), head and neck + (n = 283), thyroid + (n = 198), urinary bladder + (n = 829), colorectal + (n = 1324), and lymphoid neoplasms + (n = 728). Variants that have been previously associated in single cancer studies have superscript (a). The heatmap reflects the number of carriers with the risk-increasing allele for each associated variant with the index (y-axis) and additional (x-axis) cancer over the total number of carriers, restricting to cancer cases. When the index and additional cancer are the same, the heatmap value represents all carriers with the specified cancer diagnosis divided by the total number of carriers. Abbreviations: SNP, single nucleotide polymorphism; EA, effect allele; OR, odds ratio

Fig. 3

Germline gene-based association results for multiple primary cancers combined or grouped by organ site. Burden tests were performed combining variants defined as pLOF with or without deleterious missense variants, defining deleteriousness by at least one (1/5) or all five (5/5) prediction algorithms used (Methods), at a MAF < 0.5%. Following a fixed-effects meta-analysis of Kaiser Permanente Research Bank and UK Biobank data, Bonferroni significant associations (p < 2.65 × 10⁻⁶ = 0.05/18,842) corrected for the number of genes tested were found for comparisons of cancer-free controls (n = 165,853) with all cases with any 2+ primary cancers (n = 6429) and with groups of cases defined by an index cancer for the following phenotypes: prostate + (n = 1977), breast + (n = 1874), and ovary + (n = 239). For each gene, the variant grouping with the smallest p-value was selected. The heatmap reflects the number of carriers of each associated variant, with the index (y-axis) and additional (x-axis) cancer over the total number of carriers, where the carrier is defined as having at least one alternate allele across all variants in a given gene, restricting to cancer cases. When the index and additional cancer are the same, the heatmap value represents all carriers with the specified cancer diagnosis divided by the total number of carriers. Abbreviations: OR, odds ratio; pLOF, predicted loss of function