Deep multi-region whole-genome sequencing reveals heterogeneity and gene-by-environment interactions in treatment-naive, metastatic lung cancer

Abstract

Our understanding of genomic heterogeneity in lung cancer is largely based on the analysis of early-stage surgical specimens. Here we used endoscopic sampling of paired primary and intrathoracic metastatic tumors from 11 lung cancer patients to map genomic heterogeneity inoperable lung cancer with deep whole-genome sequencing. Intra-patient heterogeneity in driver or targetable mutations was predominantly in the form of copy number gain. Private mutation signatures, including patterns consistent with defects in homologous recombination, were highly variable both within and between patients. Irrespective of histotype, we observed a smaller than expected number of private mutations, suggesting that ancestral clones accumulated large mutation burdens immediately prior to metastasis. Single-region whole-genome sequencing of from 20 patients showed that tumors in ever-smokers with the strongest tobacco signatures were associated with germline variants in genes implicated in the repair of cigarette-induced DNA damage. Our results suggest that lung cancer precursors in ever-smokers accumulate large numbers of mutations prior to the formation of frank malignancy followed by rapid metastatic spread. In advanced lung cancer, germline variants in DNA repair genes may interact with the airway environment to influence the pattern of founder mutations, whereas similar interactions with the tumor microenvironment may play a role in the acquisition of mutations following metastasis.

Molecular epidemiology and diagnostics of KRAS mutations in human cancer

“…Accordingly, it is expected that such subclonality variations may affect the status of metastases as well. A recent deep-sequencing study on a small cohort of four KRAS mutant patients found complete concordance between the primary tumor and its metastases [ 43 ]. However, there are previous studies with larger patient cohorts of metastatic lung adenocarcinomas where concordance was not observed.…”

Chiral DNA sequences as commutable controls for clinical genomics

“…To demonstrate the utility of sequin controls in a clinical context, we performed a case study analyzing retrospective tumor biopsy samples from three metastatic lung cancer patients (two lung adenocarcinomas and one small-cell lung cancer) presenting at St. Vincent’s Hospital, Sydney 42 (see Methods). Sequins were added to extracted patient DNA, with the synthetic tumor and normal mixtures added to patient tumor and matched normal samples, respectively.…”

“…To evaluate the diagnostic performance of variant detection in the presence and absence of sequins, we compared the filtered sets of patient variant candidates to calls that were generated independently by high depth whole-genome NGS 42 (>150× coverage; see Methods). In total, 12/16 variants identified using sequins were independently validated, compared to 12/49 variants identified in the absence of internal controls (Supplementary Data 1 ).…”

“…Next we assessed the utility of chiral standards representing cancer driver mutations for the analysis of patient tumor samples. We analyzed retrospective tumor biopsy samples (tumor cellularity ≥ 20%) from metastatic lung cancer patients (two lung adenocarcinoma and one small-cell lung cancer) that were collected for a separate study 42 , with approval of ethics committees at participating institutions (St Vincent’s Hospital and Garvan Institute). Samples were processed for DNA extraction from the frozen cell suspension using the Qiagen (Hilden, Germany) DNEasy kit as per the manufacturer’s instructions, and then checked for quality, purity, and integrity before adding chiral DNA standards.…”

“…These optimum filtering thresholds were determined and applied separately for each individual sample. After filtering, either by the standard or optimized approach, variant candidates were queried against raw somatic variant calls generated independently by deep (>150×) whole-genome NGS performed on the same sample pairs 42 , for orthogonal validation. Variants of interest were evaluated for pathogenicity using the Ensembl Variant Effect Predictor tool, via the web interface ( https://asia.ensembl.org/Tools/VEP ).…”