PD-L1 expression in primary clear-cell renal cell carcinoma (ccRCC) increases the likelihood of response to anti-PD-1 inhibition, but fails to identify all responders. We hypothesized that PD-L1 levels assessed in randomly selected areas of the primary tumors may not accurately reflect expression levels in metastatic lesions, which are the target of systemic therapy. Therefore, we compared PD-L1 expression in a series of primary ccRCC and their metastases. Tissue blocks from 53 primary ccRCCs and 76 corresponding metastases were retrieved. Areas with predominant and highest nuclear grade were selected. Slides were immunostained with a validated anti-PD-L1 antibody (405.9A11). Membranous expression in tumor cells was quantified using H-score. Expression in tumor-infiltrating mononuclear cells (TIMC) was quantified using a combined score. Discordant tumor cell PD-L1 staining between primary tumors and metastases was observed in 11 of 53 cases (20.8%). Overall, tumor cell PD-L1 levels were not different in primary tumors and metastases (P = 0.51). Tumor cell PD-L1 positivity was associated with higher T stage (P = 0.03) and higher Fuhrman nuclear grade (P < 0.01). Within individual lesions, PD-L1 positivity was heterogeneous and almost exclusively detected in high nuclear grade areas (P < 0.001). No difference was found in PD-L1 levels in TIMCs between primary tumors and metastases (P = 0.82). The heterogeneity of PD-L1 expression in ccRCC suggests that its assessment as a predictive biomarker for PD-1 blockade may require analysis of metastatic lesions. Notably, because PD-L1 expression was mostly detected in high nuclear grade areas, to avoid false-negative results, these areas should be specifically selected for assessment.
©2015 American Association for Cancer Research.