MHC Class I and PD-L1 Expression May Predict Treatment Response to Anti-PD-1/PD-L1 Therapy in Metastatic Triple-negative Breast Cancer Patients.

Although immune checkpoint inhibition has proven to be efficacious in a subset of breast cancer patients, there remains a heterogeneity in clinical responses. Major histocompatibility complex class I (MHC I) expression has been touted as a possible predictor of response to immune checkpoint inhibition given its role in the adaptive immune system. Loss of tumoral MHC I expression in breast cancer as an evasion of immune checkpoint inhibition has been described; however, outcome data are limited. We herein evaluate the relationship between PD-L1 and MHC I expression and response to PD-1/PD-L1 inhibitors in a cohort of treated metastatic triple-negative breast cancer (TNBC) patients. Best treatment response was assessed in 46 metastatic TNBC patients treated with anti-PD-1/PD-L1 immune checkpoint inhibition with or without chemotherapy. Primary tumors were evaluated by immunohistochemistry (IHC) for expression of PD-L1 and MHC I. PD-L1 was scored via the combined positive score (CPS). MHC I was evaluated for retention, partial loss (≥10%), or complete loss within the tumor cells. Models using logistic ordinal regression were constructed to determine the ability of PD-L1 and MHC I by IHC to predict treatment response. The best responses by RECIST 1.1 were 16 (34%) with complete response (CR), 15 (33%) with partial response (PR), and 15 (33%) with stable or progressive disease (no response; NR). Of the patients with CR, 13 (87%) had retained MHC, and of those, 12 (92%) were positive for PD-L1. Of the patients with PR, 7 (47%) had retained MHC, and of those, 4 (57%) were positive for PD-L1. Of the patients with NR, 8 had partial or complete loss of MHC (53%), and of those, all (100%) expressed PD-L1. We found that MHC I status is independently and significantly associated with treatment response (P=0.024), with intact MHC I more common in complete responders and partial or complete loss of MHC I more common in partial responders and nonresponders. We found no association between PD-L1 expression alone with treatment response in our cohort of TNBC patients. Loss of tumoral MHC I expression is more frequently seen in patients without treatment response, despite most of these tumors having some PD-L1 expression (CPS ≥1). Using a logistical ordinal regression model, we found that both MHC I and binary PD-L1 (bPD-L1) status play significant roles in the prediction of grouped treatment response (MHC I P=0.002 and bPD-L1 P=0.03). Additional studies evaluating MHC I loss as a potential barrier to successful immune checkpoint inhibition are needed to ascertain the potential use of MHC I as a predictive biomarker in combination with PD-L1.