Abstract 2780: Loss of tumor HLA-A and B2M expression and association with biomarkers and clinical outcome

Loss of MHC-I and/or Beta 2 microglobulin (B2M) expression is regarded as a frequent resistance mechanism to immunotherapies. Indeed, downregulation of MHC-I expression and/or defects in the pathways of antigen presentation machinery are commonly reported in human cancers. Despite many reports published, available tools lack the robustness, specificity and sensitivity needed for systematic integration and analysis in clinical trials. In order to explore loss of antigen presentation as one of the major resistance mechanisms to checkpoint inhibitor (CPI) therapy treatment and to enable identification of patients with tumors harboring antigen presentation defects, we have developed immunohistochemistry (IHC)-based robust prototype assays to quantify HLA-A and B2M proteins in patient tissue samples. We have determined the prevalence of loss of HLA-A and B2M in banked primary tumor blocks as well as in 329 tumor metastatic biopsies collected during Roche-sponsored Phase 1 clinical trials investigating novel immunotherapies as single agents or in combination with CPI. Below are the key findings and preliminary interpretations:We confirm a significant prevalence of HLA-A and B2M loss (up to 65%) depending on indication, revealing heterogeneous expression patterns across patients and indications.We describe an increase of HLA-A and B2M loss from primary to metastatic lesions, which may reflect a selection of MHC-I-low/negative clones in spreading cancer cells (and maybe lines of therapies).We show that high on-treatment B2M expression is positively associated with clinical outcome (RECIST), while high baseline B2M expression only shows a trend (but no statistical significance) towards positive association with response. If the former (on-treatment expression) could be seen as a consequence of the drug effect (induction of inflammation, IFN-signaling and subsequent induction of MHC-I expression), the latter (baseline expression) would have been the foundation for enriching patient populations based on MHC-I expression. Based on these data, the absence of strong correlation between baseline expression and response does not support a cross-CIT enrichment approach at this stage.We see a treatment-induced increase of MHC-I expression in most of the patients. Preliminary conclusions could be that 1/MHC-I loss can be rescued in most of the cases we analyzed (few hard losses), 2/Since we only see few responders, MHC-I loss is not the dominant resistance mechanism or the d