Immune checkpoint inhibitor (ICPI) efficacy and resistance detected by comprehensive genomic profiling (CGP) in non-small cell lung cancer (NSCLC)

Abstract

Background: The prediction of outcome to ICPI in advanced NSCLC is of great clinical interest. We considered CGP, PD-L1 IHC, and real world data to investigate potential biomarkers for ICPI response. Methods: CGP and IHC was performed on 1,619 FFPE NSCLC samples in the FoundationCORE database (FMI). The SP142 antibody was used to capture PD-L1 tumor expression (PD-L1 TE) for these 1,619 samples. NSCLC patients (n=2139) in the Flatiron Health Analytic Database with FoundationOne testing CGP results and real world IHC results for PD-L1 TE were analyzed separately (FMI-FIH). CGP used >=50 ng of DNA and a hybrid-capture, adaptor ligation-based assay (median coverage depth>600X). TMB (mut/Mb) was determined on 1.1Mb of sequenced DNA. Results: PD-L1 IHC TE correlated weakly with TMB (FMI samples) (Spearman's q 0.085, p=6.16e-4); mean TMB was 10.9 mut/Mb, median 8.1 mut/Mb and 14.5% had high TMB (>=20 mut/Mb). From FMI-FIH, high TMB but not PD-L1 status predicted longer mean duration on therapy (DOT) (p=0.001). Analysis of the FMI and FMIFIH datasets revealed relationships between GA, PD-L1 TE, TMB, and mean DOT. Inactivating STK11 GA were seen in 12.1% of FMI-FIH and 15.1% of FMI samples, most often adenocarcinomas (aCa). STK11 GA correlated with high TMB/low PD-L1 (FMI; p=0.0014) and preliminary analyses suggest correlation with negative ICPI treatment outcome. Several genes were commonly co-altered with STK11 (FMI): KRAS (54.5%), TP53 (43%), CDKN2A (27.5%), CDKN2B (20.1%), KEAP1 (18.9%), and MYC (13.5%). BRAF GA, most often short variants (SV) in aCa, were associated with prolonged DOT on ICPI regardless of TMB score (FMI-FIH; p=0.0073). MET SV also predicted prolonged DOT on ICPI, but insufficient events prevented calculation of statistical significance (FMI-FIH). Analysis of the TCGA lung aCa dataset revealed MET SV (2.8%) linked with immune activation gene expression profiles (p<0.05) and STK11 mutations (14.2%) with immune evasion profiles (p<0.05). Conclusions: Although TMB powerfully predicts ICPI outcome independent of tumor cell PD-L1 expression, considering GA in STK11, BRAF or MET may significantly increase the precision and improve outcomes when using genomics with IHC to guide to ICPI selection.