The tumor immune microenvironment shapes response to pembrolizumab in microsatellite instability-high endometrial cancer

Abstract

Objectives: Despite FDA approval of pembrolizumab in microsatellite instability-high (MSI-H) / mismatch repair deficient solid tumors, approximately half of patients with MSI-H endometrial cancer are treatment-refractory. Our unpublished analysis of MSI-H endometrial cancer samples from the Cancer Genome Atlas (TCGA) suggests the possibility of immunologically ‘hot’ and ‘cold’ tumor microenvironments (TME); these differences may explain variable treatment responses. We sought to evaluate MSI-H endometrial tumor samples to examine differences in the TME and identify transcriptomic signatures associated with response/resistance to pembrolizumab. Methods: Archival tumor samples from MSI-H endometrial cancer patients treated at the University of Texas MD Anderson Cancer Center were obtained under an IRB-approved protocol. Tissue samples originating from patients who did not receive pembrolizumab treatment (‘untreated cohort’; n=11) were submitted for RNA sequencing analysis (RNA-seq) to evaluate TCGA-derived signatures in an independent sample set. Pre-treatment archival tumor samples (‘treated cohort’; n=23) from patients who were later treated with pembrolizumab were collected and submitted for RNA-seq to identify predictive immuno-genomic signatures associated with treatment response. Results: In the untreated cohort, there were observable differences in the transcriptomic profiles of the samples. Four samples were immunologically ‘hot’ as evidenced by an abundance of pro-inflammatory immune cell infiltrate (CD8+ T-cells, B-cells, monocytes, and dendritic cells). Three samples had a paucity of pro-inflammatory immune cell infiltrate (‘cold’) while 4 samples had intermediate amounts (‘warm’). In the treated cohort, there were 14 responders, 7 non-responders, and 2 patients with unknown response. The 14 responders consisted of samples with ‘hot’ (5/5; 100%), ‘cold’ (6/8; 75%), and ‘warm’ TMEs (3/8; 37.5%) while the 7 non-responders consisted of only ‘cold’ (2/8; 25%) and ‘warm’ (5/8; 62.5%) TME samples. We observed an enrichment of fibroblasts and endothelial cell transcriptomic signatures in the samples of the non-responders compared to responders (p=0.018). In particular, there was a significantly higher gene expression of PAMR1, HHIP, and MMRN1 in the non-responders. Subdividing samples into complete response, partial response, and no response to pembrolizumab, we observed a trend of increasing enrichment of fibroblast and endothelial cell transcriptomic signatures as response decreased. Specifically, there was increasing expression of TAGLN (fibroblast gene) and other endothelial cell genes (EMCN, KDR, MMRN1, MYCT1, PEAR1, PTPRB, and TEK). Conclusions: The TME composition appears to be heterogeneous among MSI-H endometrial cancer patients. Increased presence of fibroblasts and endothelial cells in the TME may contribute to innate resistance to pembrolizumab. Treatment aimed toward the reduction of these cellular subpopulations may improve sensitivity to PD-1 inhibitors. Future studies are needed to validate these findings.