Germline variation in cancer-susceptibility genes in a healthy, ancestrally diverse cohort: implications for individual genome sequencing

Abstract

Technological advances coupled with decreasing costs are bringing whole genome and whole exome sequencing closer to routine clinical use. One of the hurdles to clinical implementation is the high number of variants of unknown significance. For cancer-susceptibility genes, the difficulty in interpreting the clinical relevance of the genomic variants is compounded by the fact that most of what is known about these variants comes from the study of highly selected populations, such as cancer patients or individuals with a family history of cancer. The genetic variation in known cancer-susceptibility genes in the general population has not been well characterized to date. To address this gap, we profiled the nonsynonymous genomic variation in 158 genes causally implicated in carcinogenesis using high-quality whole genome sequences from an ancestrally diverse cohort of 681 healthy individuals. We found that all individuals carry multiple variants that may impact cancer susceptibility, with an average of 68 variants per individual. Of the 2,688 allelic variants identified within the cohort, most are very rare, with 75% found in only 1 or 2 individuals in our population. Allele frequencies vary between ancestral groups, and there are 21 variants for which the minor allele in one population is the major allele in another. Detailed analysis of a selected subset of 5 clinically important cancer genes, BRCA1, BRCA2, KRAS, TP53, and PTEN, highlights differences between germline variants and reported somatic mutations. The dataset can serve a resource of genetic variation in cancer-susceptibility genes in 6 ancestry groups, an important foundation for the interpretation of cancer risk from personal genome sequences.

Figure 1. Profile of the variability per individual.

(A) Boxplot of the total number of variants, the number of variants listed in HGMD, the number of likely deleterious variants, and the number of variants of unknown significance per individual for cancer-associated genes. (B) Distribution of the number of cancer genes with at least one nonsynonymous variant per individual.

Figure 2. Admixture coefficients for the subpopulations.

The admixture proportions of the 6 ancestral populations (colors) are displayed for all individuals in each of the 7 groups defined in the cohort (panels). (A) European (B) Central Asian (C) East Asian (D) African (E) African-European (F) Hispanic (G) Other. Red: European, Blue: Central Asian, Cyan: East Asian, Yellow: African, Green: Native American, Magenta: Oceania.

Figure 3. Number of cancer-gene variants per individual by ancestry.

The distribution of the number of nonsynonymous genes per subject for each of the 6 ancestry-based subpopulations.

Figure 4. Variation prevalence per gene.

Distribution of the number of individuals with a variant per gene for (A) all variants (B) rare variants.

Figure 5. Correlation between the number of variants and coding length.

The number of nonsynonymous variants vs. total number of coding bases for each of the 158 cancer-susceptibility genes.

Figure 6. p53 DNA-binding domain variants.

The DNA-binding domain of the p53 protein (black) bound to DNA (purple) . Common somatic mutations (yellow) contact the DNA or stabilize the structure. Variants in our cohort (red) occur at residues distal to the DNA binding site except for Arg 283 (green).