Targeted sequencing analysis of cell-free DNA from metastatic non-small-cell lung cancer patients: clinical and biological implications

Background: Sequencing artifacts, clonal hematopoietic mutations of indeterminate potential (CHIP) and tumor heterogeneity have been hypothesized to contribute to the low concordance between tissue and cell-free DNA (cfDNA) molecular profiling with targeted sequencing. Methods: We analyzed by targeted sequencing cfDNA from 30 healthy individuals, and cfDNA and matched tumor samples from 30 EGFR-mutant and 77 EGFR wild-type metastatic non-small-cell lung cancer (mNSCLC) patients. Discordant cases were solved by droplet digital PCR (ddPCR). Results: By testing cfDNA from healthy donors, we developed an algorithm to recognize sequencing artifacts. Applying this method to cfDNA from mNSCLC patients, EGFR mutations were detected with a good sensitivity (76.7%) and specificity (97.4%). In contrast, sensitivity and specificity for KRAS variants were 61.5% and 93.8%, respectively. All EGFR and KRAS variants detected in plasma but not in tissue were confirmed by ddPCR, thus excluding sequencing artifacts. In a fraction of cases, KRAS mutations found in plasma samples were confirmed in tumor tissue suggesting tumor heterogeneity. KRAS variants were found to be more likely sub-clonal as compared with EGFR mutations, and a correlation between clonal origin and frequency of detection in plasma was found. In a case with both EGFR and KRAS variants in cfDNA, we could demonstrate the presence of the KRAS variant in tumor tissue associated with lack of response to tyrosine kinase inhibitors (TKIs). Conclusions: Although sequencing artifacts can be identified in targeted sequencing of cfDNA, tumor heterogeneity and CHIP are likely to influence the concordance between plasma and tissue testing.