Divergent immunobiological correlates of FDA-/EMA-approved PD-L1 assays and scoring algorithms in muscle-invasive bladder cancer

Abstract

Immune checkpoint inhibition gained significant impact in therapeutic handling of advanced bladder cancer. PD-L1 testing is currently required for first-line immune therapy in platinum-ineligible patients and bases on complex scoring algorithms. We aimed to decipher the composition of distinct scoring algorithms and their associations with immune biological determinants. 193 consecutively collected MIBC patients were assessed with four approved PD-L1 assays for all currently established scoring algorithms (IC-Score [companion drug: atezolizumab], CPS [companion drug: pembrolizumab], ICarea/TC-score [companion drug: Durvalumab]). Results were correlated with a combined immune infiltration and immune checkpoint signature consisting of multiple immune cell populations (CD3, CD8, FOXP3, GZMB, CD56, CD68) and other immune checkpoint proteins (LAG3, CTLA4, PD-1). Findings were correlated with clinico-pathological data to elucidate the biological implications of different PD-L1 scoring algorithms. Of the three scoring algorithms exclusively the IC-score identified patient subsets with different survival rates between PD-L1 negative and positive subgroups (improved survival for PD-L1 positive subset; HR=0.44, P=0.0004) which predominantly detects patients with high PD-L1 expression on immune cells, while the CPS predominantly covers PD-L1 tumor cell positive cases. The ICarea/TC score equally identified both populations. Deeper analysis of the underlying immune biology revelead that high aberrant PD-L1 tumor cell expression associated with worse survival while PD-L1 immune cell expression is highly associated with improved survival. Compared to predominantly PD-L1 tumor cell positive tumors, general immune infiltration is higher in cases with strong PD-L1 expression on immune cells (P=0.0037). PD-L1 scoring algorithms identify different patient populations with either predominant immune cell (IC-Score [atezolizumab]) or tumor cell expression (CPS [pembrolizumab]), while the ICarea/TC-score (durvalumab) sufficiently covers both populations. High expression of PD-L1 on tumor cells associated with worse survival indicating high predictive potential for PD-L1 inhibition while high expression on immune cells is a baseline characteristic for improved patient survival (positive prognostic value).