Chromosomal instability, leading to full or partial aneuploidy, is a common feature of solid tumors and hematological malignancies, with several recurrent aberrations recognized as important diagnostic or prognostic molecular markers. While genome-wide genotyping and genomic hybridization microarray analyses have been performed in clinical laboratories for several years the usage of next generation sequencing for detection of acquired copy number alterations (CNA) has not yet reached full clinical integration. Algorithmic strategies based on whole exome sequencing (WES) have been proposed, primarily focusing on read depth alterations affected by copy gain or loss.
We aimed to identify copy-neutral loss of heterozygosity (CN-LoH), which is not detectable by sequencing read depth correlation or the analogous microarray CGH, by means of simple statistical analyses for increased transparency and direct clinical implementation.
23 paired samples from different hematological disorders were selected from 14 individuals with MCL, CLL, CML, AML, T-ALL and monoclonal B cell lymphocytosis. Detection of CN-LoH by means of significant allele frequency (AF) shift using Fisher's exact test and χ2 was performed in combination with paired gene-wise read depth ratios correlations in order to resolve both copy altering and neutral chromosomal aberrations. Alignment (GRCh37) and variant calling was performed using BWA (Li and Durbin 2009) and GATK (McKenna et al. 2010) (v. 3.6/3.8). Variants and reads located outside targeted regions were excluded, and a base quality threshold of 25 and minimum allele read depth of 20 was applied. Gene-wise read depths were retrieved with BEDTools (Quinlan and Hall 2010) from RefSeq coordinates (UCSC, Karolchik et al. 2004). Statistics and plotting were performed in Mathematica and software developed for demonstration purposes (http://doi.org/10.7910/DVN/KFMGNY, Harvard Dataverse).
Sequencing yield spanned 48.5–188.6 million (median 84.3x106) paired-end reads. The 23 paired variant call sets comprised a median of 19,101 coding variants, matching the expected number (Frebourg 2014; Ng et al. 2009), with a median read depth of 96 and lower and upper quartiles at 71 and 136, respectively. The combination of significant AF shifts identified 69 altered chromosomes and offered mutual confirmation. Six evident CN-LoHs (>1% AF shifts, p < 0.01) were found in MCL, CML, AML and T-ALL with one additional suspected low frequency deletion (estimated 10–15% burden) in a case of CLL. Robust test of equal variances (Brown and Forsythe 1974) between the paired diagnosis and remission sample was needed to detect this low burden aberration (p < 10–10 in contrast to the other paired autosomes (pmedian = 0.019)). This also clearly indicates that low burden CN-LoH pose a challenge – as with other methods. Another patient (MCL) harbored both a balanced tetrasomy (not evident by AF alone, see figure) and a CN-LoH on chromosome 22 (not evident by DP-ratios alone).
The samples from 6 of the 14 patients with hematological malignancies/disorders were found to harbor 7 copy neutral deletions in total – all escaping previous clinical cytogenetic analyses. One was detected in a diagnostic sample from a patient otherwise diagnosed as cytogenically normal AML, which emphasize the importance of such methods. We conclude that this low-complexity method, juxtaposing DP and AF, is directly applicable for the detection of copy neutral chromosomal loss of single genes from WES with intermediate coverage and medium to high burden chromosomal aberrations.