Germline DNA copy number aberrations identified as potential prognostic factors for breast cancer recurrence

Abstract

Breast cancer recurrence (BCR) is a common treatment outcome despite curative-intent primary treatment of non-metastatic breast cancer. Currently used prognostic and predictive factors utilize tumor-based markers, and are not optimal determinants of risk of BCR. Germline-based copy number aberrations (CNAs) have not been evaluated as determinants of predisposition to experience BCR. In this study, we accessed germline DNA from 369 female breast cancer subjects who received curative-intent primary treatment following diagnosis. Of these, 155 experienced BCR and 214 did not, after a median duration of follow up after breast cancer diagnosis of 6.35 years (range = 0.60-21.78) and 8.60 years (range = 3.08-13.57), respectively. Whole genome CNA genotyping was performed on the Affymetrix SNP array 6.0 platform. CNAs were identified using the SNP-Fast Adaptive States Segmentation Technique 2 algorithm implemented in Nexus Copy Number 6.0. Six samples were removed due to poor quality scores, leaving 363 samples for further analysis. We identified 18,561 CNAs with ≥1 kb as a predefined cut-off for observed aberrations. Univariate survival analyses (log-rank tests) identified seven CNAs (two copy number gains and five copy neutral-loss of heterozygosities, CN-LOHs) showing significant differences (P<2.01×10(-5)) in recurrence-free survival (RFS) probabilities with and without CNAs.We also observed three additional but distinct CN-LOHs showing significant differences in RFS probabilities (P<2.86×10(-5)) when analyses were restricted to stratified cases (luminal A, n = 208) only. After adjusting for tumor stage and grade in multivariate analyses (Cox proportional hazards models), all the CNAs remained strongly associated with the phenotype of BCR. Of these, we confirmed three CNAs at 17q11.2, 11q13.1 and 6q24.1 in representative samples using independent genotyping platforms. Our results suggest further investigations on the potential use of germline DNA variations as prognostic markers in cancer-associated phenotypes.

Figure 1. Absolute counts of CNAs stratified by overlap with germline CNVs in DGV and their length.

Shown in the histograms are total numbers of copy number losses (CN Loss), copy number gains (CN Gain) and copy neutral-loss of heterozygosities (CN-LOHs) identified in 363 samples stratified by their lengths (≥1 KB–10 KB, >10 KB–100 KB and >100 KB–5 MB) and their overlap with known germline CNVs in the Database of Genomic Variants (DGV), Toronto. A 100% overlap is shown as ‘Complete’, less than 100% but more than 0% is shown as ‘Partial’ and no overlap is shown as ‘Absent’.

Figure 2. Chromosome-wide distributions of 9,164 CNAs tested for association with BCR in unstratified samples.

Shown on the x-axis are middle points of chromosomal start and end positions (NCBI Build 37) of 9,161CNAs and on the y-axis are –log10 P values for their association with the phenotype of BCR in unstratified 363 samples. P values were obtained from log-rank tests with one d.f.

Figure 3. Chromosome-wide distributions of 7,218 CNAs tested for association with BCR in 208 luminal A cases.

Shown on the x-axis are middle points of chromosomal start and end positions (NCBI Build 37) of 7,218 CNAs and on the y-axis are –log10 P values for their association with the phenotype of BCR in 208 luminal A cases. P values were obtained from log-rank tests with one d.f.

Figure 4. Relationships between RFS and three CN-LOHs validated by RT-qPCR and Sequenom genotyping.

Using the data from 208 luminal A cases, Kaplan-Meir survival plots were generated to evaluate the predictive power of three CN-LOHs validated in independent platform for RFS. The x-axes in all three plots show recurrence time in days and the y-axes show RFS probabilities with and without CN-LOHs. Differences in RFS probabilities were assessed by log-rank tests with one d.f. HRs and 95% CIs were estimated by Cox proportional hazards model adjusted for tumor stage and grade.