Identifying Mechanisms of Resistance to Immunotherapy for Merkel Cell Carcinoma

Merkel cell carcinoma (MCC) is the most deadly skin cancer on a case-by-case basis. Until recently, little was known about its pathogenesis. Thus, there were no effective therapies. In a landmark manuscript, we identified the genomic landscape of Merkel cell carcinomas. 80% of MCCs harbor a clonally integrated tumor virus, Merkel cell polyomavirus. These cancers have few, if any, somatic mutations. In contrast, 20% are virus negative. These cancers harbor 100’s to 1000’s of somatic mutations. These data explain why both MCC subtypes are immunogenic. The virus-positive MCCs express foreign viral peptides but few tumor neoantigens. In contrast, the virus negative MCCs express more neoantigens than any other epithelial cancer. Our work supports the use of immunotherapy for both MCC subtypes. Indeed, PD1 inhibitors induce objective responses in 35-50% of patients. While this number appears high, at least half of patients do not respond. Moreover, even in those who initially respond to PD1 inhibitors, patients can develop acquired resistance. Unfortunately, we know little about the mechanisms of primary or acquired immunotherapy resistance. Thus, we do not have biomarkers to predict who is or is not going to respond. In addition, we do not have effective therapies for those who fail PD1 immunotherapy. To meet this critical unmet need, we have built a large multi-institutional clinically annotated biobank of patient samples. We are applying a number of orthogonal high dimensional assays to identify the landscape of immune cells in each tumor and the receptor/ligands which govern their behavior. These include DNA-seq to identify tumor neoantigens, RNA-Seq (bulk, single cell, and spatial) to identify MHC Class I molecules and immune cell infiltrates, high dimensional immunofluorescence to identify spatial relationships of immune cells, and functional assays. The goal is to immunophenotype each tumor and identify naturally occurring endotypes that predict responses to immunotherapy. To analyze this data, we have assembled an elite team of cancer geneticists, pathologists, biochemists, and machine learning experts on digital pathology. Based on a preliminary analysis of the first 42 samples, we have identified specific patterns. We have found at least three naturally occurring endotypes: immune deserts, immune excluded, and immune infiltrated tumors. Immune deserts do not respond to immunotherapy. Moreover, immune excluded and immune infiltrated tumors harbor many cell types including myeloid cells, ab, and gd T cells. We found that responders are enriched for specific MHC Class I molecules. These data suggest that specific tumor antigens (expressed by these MHC Class I molecules) may be preferentially immunogenic or lead to the induction of cytotoxic T cells that are poised to kill MCC cells. Orthogonally, spatial and bulk deconvolution RNA-Seq analysis also found statistically significant enrichment of gd cells in immunotherapy responders. In this proposal, we will validate our preliminary findings in our larger cohort and use the larger dataset to understand how and why these signals contribute to responsiveness to immunotherapies. The ultimate goal is to identify biomarkers that predict responses to PD1 immunotherapy and novel therapeutic regimens for those who fail immune check point blockade.