Presence of immune infiltrates, increased expression of transposable elements, and viral response pathways in sarcoma associate with response to checkpoint inhibition.

11510 Background: Response to checkpoint inhibition (CPI) in sarcoma is overall low and varies between and within subtypes. Understanding tumor intrinsic determinants of this response may improve efficacy and patient selection. The de-repression of transposable elements (TEs), which are epigenetically silenced repetitive DNA elements of viral origin, is linked to anti-tumor immunity through an antiviral inflammatory response. We hypothesize that baseline expression of TEs and epigenetic regulators correlates with overall response rate (ORR) in sarcoma CPI clinical trials. Methods: This is a retrospective analysis of bulk RNA-sequencing data from pre-treatment biopsies of patients on CPI trials in sarcoma (pembrolizumab plus talimogene laherparepvec, nivolumab plus bempegaldesleukin, and pembrolizumab plus epacadostat). Sixty-seven samples from unique patients representing 12 subtypes were analyzed. The MCP counter deconvolution method and unsupervised clustering were used to group samples by immune phenotypes resulting in immune ‘hot’ and ‘cold’ clusters. ORR was defined by RECIST. To determine if baseline expression of TEs and epigenetic regulators significantly predicted immune types, we implemented a lasso penalized logistic regression. Results: Immune ‘hot’ tumors were characterized by increased immune infiltrates including CD8+ T-cells, B-cells, and NK cells vs ‘cold’ tumors. Patients with ‘hot’ vs ‘cold’ tumors had an ORR of 30.5% (11/36) vs. 3.2% (1/31) (p = 0.003; chi-squared). The best predictors of ‘hot vs ‘cold’ was the increased expression of multiple TE families including MER45A, MER57F, and LTR21B (respective lasso coefficients, 0.27, 0.07, and 0.07). Expression of IKZF1, a chromatin-interacting transcription factor, was also predictive (lasso coefficient, 0.35) and increased expression correlated with improved ORR (p = 0.003; unpaired t-test). TE and IKFZ1 expression was significantly correlated with CD8+ T-cell signaling and antiviral response pathways such as cGAS-STING (MER57F, r 2 = 0.43, padj = 1.75E-4; IKZF1, r 2 = 0.63, padj = 6.28E-9) and type II interferon (MER57F, r 2 = 0.67, padj = 2.51E-10; IKZF1, r 2 = 0.60, padj = 7.19E-8). Increased expression of cGAS-STING (p = 3.9E-4; unpaired t-test) and type II interferon pathways (p = 1.89E-10; unpaired t-test) was significant in ‘hot’ tumors. Conclusions: Immune ‘hot’ baseline immune profiles of sarcoma are associated with improved ORR to CPI and with increased expression of TEs and IKZF1. These differences in gene expression correlate with increased inflammatory signaling, which suggests a response to TE-encoded viral-like sequences that are typically epigenetically silenced. Induction of TE de-repression and IKZF1 expression through epigenetic targeting warrants pre-clinical investigation as a strategy to promote CPI response in sarcomas.