Pan-tumor analyses of kinase fusions detected in circulating tumor DNA (ctDNA) and concordance with paired tissue.

Abstract

3517 Background: Oncogenic kinase gene fusions are targetable with approved and investigational therapies and can also emerge as acquired resistance (AR) to targeted therapy. To understand the clinical validity of liquid biopsy comprehensive genomic profiling (CGP) to detect kinase fusions, we compared patient-matched plasma and tissue-based CGP. Methods: Hybrid capture-based CGP was performed on 28,743 plasma and 325,131 tumor tissue samples in the course of clinical care. Complete exonic regions of 13 kinases involved in oncogenic fusions plus select introns in ALK, EGFR, FGFR2/3, PDGFRA, RET, and ROS1 were sequenced to capture fusions with well characterized breakpoints. ctDNA fraction was estimated by maximum somatic allele frequency (MSAF). Results: 86% of cases had detectable ctDNA in plasma (MSAF > 0). Kinase fusions were detected in 2.1% of ctDNA cases (478/23,294) and were most prevalent in patients (pts) with bladder cancer (4.5%), non-small cell lung cancer (NSCLC) (4.3%), and cholangiocarcinoma (3.9%). The most commonly rearranged kinases were ALK (60%, 162/271) and RET (19%, 51/271) in NSCLC, FGFR2 (85%, 11/13) in cholangiocarcinoma, and FGFR3 (88%, 7/8) in bladder cancer. ALK fusions were detected in 26% (54/207) of fusion+ non-NSCLC cases. Paired tissue and ctDNA samples where ≥1 sample harbored a kinase fusion were available for 147 pts; median time between sample collection was 150 days (interquartile range: 444 days). Positive percent agreement (PPA) to tissue and liquid biopsies was 76% and 80%. Median MSAF in concordant and discordant ctDNA samples was 2.3% and 0.41% (p = 0.04) and median time between specimen collection for concordant and discordant pairs was 110 and 344 days (p = 0.04). PPA to tissue and liquid was 86% and 88% for pairs collected < 60 days apart (n = 53), versus 70% and 74% for pairs collected > 60 days apart. 6 pts with paired samples all collected > 196 days apart (median 593 days) had initial tissue samples with EGFR driver mutations and had an acquired kinase fusion (4 ALK, 1 FGFR2, 1 FGFR3) in the 2nd ctDNA sample, likely representing AR. Conclusions: Kinase gene fusions identified by tissue-based CGP were detected by liquid biopsy CGP in 85% of temporally-matched plasma samples. Kinase fusion detection by liquid biopsy CGP is feasible and had high PPA to tissue-based CGP. Subsequent sampling by liquid biopsy identified acquired fusions in EGFR driver positive cases consistent with known mechanisms of resistance to EGFR inhibitors suggesting utility of liquid biopsy at progression to identify targetable mechanisms of AR.