MA17.06 Landscape of Somatic Mutations Involving Lung Cancer Associated Genes in Non-Small Cell Lung Cancer (NSCLC) Patient-Derived Xenografts

Abstract

Patient-derived tumor xenografts (PDXs) have high fidelity to their histological origins, and maintain the molecular heterogeneity and genetic aberrations of the donor patient tumors more faithfully than established in non-small cell lung cancer (NSCLC) cell lines. This study evaluated whether our panel of PDX models recapitulate known cancer-related gene mutations. Whole-exome sequencing was completed on 103 NSCLC PDX models, 47 adenocarcinoma (AdC) and 56 squamous (SqCC), with a mean coverage of 84x. After filtering for contaminating mouse reads, the exome data were aligned using the Burrows-Wheeler Aligner, processed using the standard GATK pipeline, and mutations were identified using MuTect. Additional filtering using dbSNP, ExAC and ESP was performed for cases without corresponding normal adjacent lung exome data (n = 80). The identified mutations were compared to 1260 frequently mutated cancer-related genes, which were compiled from a panel of cancer-related mutated genes (555) and a panel of lung cancer-specific mutated genes (1082). High rates of somatic mutations were observed in both AdC (mean of 12.4 mutations/megabase) and SqCC (mean of 11.7 mutations/megabase) PDX models. Compared to the rates observed in primary lung cancers in The Cancer Genome Atlas studies (mean of 8.9 mutations/megabase in AdC; 8.1 mutations/megabase in SqCC), these values appear higher, but may be inflated due to the lack of data from corresponding normal tissues. AdC models had a total of 953 mutated genes (median: 57 genes/model; range: 5-307), while SqCC models were characterized by 1007 mutated genes (median: 55 genes/model; range: 21-354). Specific mutation frequencies were compared to those determined in a recent study involving genomic alterations in human primary lung AdC and SqCC (Nature Genetics 2016; 48; 607–616). This comparison, based on mutated genes common in both studies, demonstrated significant correlation of the frequencies in 791 genes in AdC (ρ=0.78; p<2.2×10-16), as well as in 799 genes in SqCC (ρ=0.73; p<2.2×10-16). Three genes that were reported as significantly mutated in both AdC and SqCC primaries, and had higher mutation frequencies in SqCC, were also observed to be higher in our SqCC PDX models (TP53: 48.9% in AdC vs. 55.4% in SqCC; CDKN2A: 4.3% vs. 7.1% and PIK3CA: 2.1% vs. 23.2%); however, the statistical significance of these differences needs to be tested. Mutation landscapes in cancer genes are recapitulated in AdC and SqCC PDX models. The fidelity of these landscapes in matched patient primary tumour samples is being investigated.