Markers of immune checkpoint inhibitor efficacy in the AACR Project GENIE database.

Abstract

e15091 Background: Although immune checkpoint inhibitors (ICIs) have been shown to be effective in many tumor types, some highly lethal cancers are not responsive to ICI. Potential biomarkers of ICI response include tumor mutational burden (TMB), which is thought to correlate with increased neoantigen production, and 9p24.1 copy number gain (CNG), which can result in over-expression of programmed death ligand 1 (PD-L1). 9p24.1 CNGs have been described in ICI-sensitive (ICI-S) hematologic malignancies (e.g., Hodgkin lymphoma), but are not well described in solid malignancy. We sought to investigate TMB and 9p24.1 CNG for ICI-S and ICI-resistant (ICI-R) tumor types in the publicly available AACR Project GENIE database, version 7.0. Methods: TMB was calculated by counting somatic mutations with tumor reference allele frequency ≥5% and sequencing depth ≥200X. Samples with < 0.5 MB sequenced were excluded. 9p24.1 CNG was extrapolated from gene-level data. Samples with two or more consecutive gene amplifications in the 9p24.1 region were determined to have 9p24.1 CNG. Samples whose sequencing assay did not include at least two genes in 9p24.1 were excluded. Using an overall response rate (ORR) of ≥10% to define ICI-S, we assessed three ICI-S cancers: hepatobiliary cancer (HBC), melanoma (MEL), non-small cell lung cancer (NSCLC); and two ICI-R types: metastatic breast cancer (MBC) & pancreatic ductal adenocarcinoma (PDAC). Groupwise TMB was compared using Wilcoxon rank-sum and 9p24.1 CNG was compared using Chi-squared. Results: MEL had the highest median TMB but a low 9p24.1 CNG rate; NSCLC had the highest rate of 9p24.1 CNG (Table). PDAC had both the lowest median TMB and 9p24.1 CNG rate. As a group, the ICI-S cancers had higher median TMB (p < .001) and 9p24.1 CNG rate (p < .001). Conclusions: Although rates of 9p24.1 CNG were low across the database as a whole, the NSCLC finding replicates findings described in early stage resected NSCLC (Inoue et al. 2016). Relatively high median TMB in MEL may explain ICI-sensitivity in this cancer type. The combination of low median TMB and low rates of 9p24.1 CNG in PDAC may explain the general lack of efficacy of ICIs in this disease. These findings demonstrate the utility of GENIE as a clinico-genomic database, and also highlight the need to identify better markers of responsiveness to these potentially effective but toxic therapies. [Table: see text]