Identification of rare germline copy number variations over-represented in five human cancer types

Abstract

Background: Copy number variations (CNVs) are increasingly recognized as significant disease susceptibility markers in many complex disorders including cancer. The availability of a large number of chromosomal copy number profiles in both malignant and normal tissues in cancer patients presents an opportunity to characterize not only somatic alterations but also germline CNVs, which may confer increased risk for cancer.

Results: We explored the germline CNVs in five cancer cohorts from the Cancer Genome Atlas (TCGA) consisting of 351 brain, 336 breast, 342 colorectal, 370 renal, and 314 ovarian cancers, genotyped on Affymetrix SNP6.0 arrays. Comparing these to ~3000 normal controls from another study, our case-control association study revealed 39 genomic loci (9 brain, 3 breast, 4 colorectal, 11 renal, and 12 ovarian cancers) as potential candidates of tumor susceptibility loci. Many of these loci are new and in some cases are associated with a substantial increase in disease risk. The majority of the observed loci do not overlap with coding sequences; however, several observed genomic loci overlap with known cancer genes including RET in brain cancers, ERBB2 in renal cell carcinomas, and DCC in ovarian cancers, all of which have not been previously associated with germline changes in cancer.

Conclusions: This large-scale genome-wide association study for CNVs across multiple cancer types identified several novel rare germline CNVs as cancer predisposing genomic loci. These loci can potentially serve as clinically useful markers conferring increased cancer risk.

Figure 1

Workflow for our CNV analysis. The numbers of samples at various steps are indicated.

Figure 2

Illustrative examples of rare germline CNVRs over-represented in specific cancers. In each panel, the first two tracks after the genomic coordinates show the RefSeq gene annotations and the positions of the probles on the Affymetrix SNP6.0 arrays. Below that, germline CNVs for cancers cases are indicated in blue (losses) and red (gains), followed by CNVs observed in control individuals in black. (A) 22 kb loss affecting NKX2-3 in 6 colorectal cancer cases (n = 342); none is present in the controls (n = 2956). (B) 56 kb loss affecting RET in 9 glioblastoma patients; 9 CNVs are also found in the controls but the sample size of the control set is almost 10-fold greater (351 vs 2956), making this statistically significant. (C) 173 kb gain affecting ERBB2 in 13 kidney cancer cases (n = 370); 19 are present in the controls. (D) 37 kb gain affecting DCC in five ovarian cancer cases (n = 380); none are present in the controls.