Allelic selection of amplicons in glioblastoma revealed by combining somatic and germline analysis

Abstract

Cancer is a disease driven by a combination of inherited risk alleles coupled with the acquisition of somatic mutations, including amplification and deletion of genomic DNA. Potential relationships between the inherited and somatic aspects of the disease have only rarely been examined on a genome-wide level. Applying a novel integrative analysis of SNP and copy number measurements, we queried the tumor and normal-tissue genomes of 178 glioblastoma patients from the Cancer Genome Atlas project for preferentially amplified alleles, under the hypothesis that oncogenic germline variants will be selectively amplified in the tumor environment. Selected alleles are revealed by allelic imbalance in amplification across samples. This general approach is based on genetic principles and provides a method for identifying important tumor-related alleles. We find that SNP alleles that are most significantly overrepresented in amplicons tend to occur in genes involved with regulation of kinase and transferase activity, and many of these genes are known contributors to gliomagenesis. The analysis also implicates variants in synapse genes. By incorporating gene expression data, we demonstrate synergy between preferential allelic amplification and expression in DOCK4 and EGFR. Our results support the notion that combining germline and tumor genetic data can identify regions relevant to cancer biology.

Figure 1. Biological rationale for selected allelic amplification in the tumor.

(A) The individual inherits an oncogenic variant from the maternal chromosome M (top). This variant is not directly typed, but is captured via linkage disequilibrium by surrounding array SNPs (labeled with nucleotide residue). The paternal chromosome P harbors the wild-type allele. In the tumor environment (bottom panel), the oncogenic allele is activated via amplification, which confers a selective growth advantage to the cell. The amplified haplotype is detected from the SNP array data, and its preferentially amplified state is revealed through analysis of data from hundreds of patients. (B) In SNP array data, the underlying biological phenomenon will manifest itself in an abundance of amplicons harboring alleles that tag the inherited variant that provides a selective advantage when amplified. The ADT tests for over-transmission of a particular allele from heterozygous “parent” cells to the “affected” (amplified) homolog in the tumor cell and examines deviations from the null hypothesis of a 1∶1 transmission ratio.

Figure 2. Genome wide ADT analysis of 178 TCGA glioblastoma samples.

Manhattan-style plot (A) of amplification distortion P-value (y-axis, log₁₀ scale) along the genome (x-axis). Save for the strongest hits, the ADT statistic follows the distribution expected under the null hypothesis, as demonstrated by the quantile-quantile plot (B) of P-values (log₁₀ scale). Only SNPs with nine or more amplified heterozygous samples are presented, to avoid effects of discrete probabilities in a small sample.

Figure 3. Heatmap of correlation, as measured by odds ratio estimates, between amplification status among six kinase/transferase activity genes showing signs of somatic allelic selection.

Values above one indicate amplification correlation, below one anti-correlation. Fisher's exact P-values are given in each heatmap pixel.

Figure 4. Selective allelic amplification and expression of SNPs in DOCK4 and EGFR.

The SNP rs6959338 in DOCK4 shows preferential amplification of the T allele (upper left), as well as higher expression levels in samples amplifying T instead of C (upper right). Similarly, rs13222385 shows preferential allelic amplification (lower left) and expression (lower right).