PD-L1 Expression Identifies High Risk Diffuse Large B-Cell Lymphoma and Is Associated with Several Genomic Markers

Diffuse large B-cell lymphomas (DLBCLs) are aggressive tumors and the most common form of non-Hodgkin lymphoma (NHL). DLBCLs arise from germinal center (GC) or post-GC B cells and display marked heterogeneity in genetic features and clinical outcomes DLBCLs, similar to other NHL and to solid tumors, are known to evade host immune response by usurping immune checkpoint pathways such as CTLA4 and PD-1. The initial Phase I and Phase II trials of checkpoint inhibitors in NHL revealed promising results, but rational and reliable biomarkers for selecting patients that benefit from such therapies is missing. Even in the case of the better-studied PD1-PD-L1 pathway, the characterization of expression of those proteins in DLBCLs is not standardized, and their relationship with molecular and clinical features in DLBCLs are still largely unknown. We set out to study PD-L1 expression using three commonly used anti-PD-L1 antibodies in a well-characterized single-institutional cohort of 52 primary DLBCLs from Keck School of Medicine of USC treated with R-CHOP or R-CHOP-like therapy with clinical follow up. All cases were characterized by aCGH; Focus::Lymphoma NGS test (Cancer Genetics Inc), which includes 220 most commonly mutated genes in B NHL; FISH including bcl2 and c-myc; and immunohistochemical staining to define cell of origin (Hans classification) and with three anti-PD-L1 antibodies (Abcam EPR1161, Abcam 28-8 and Ventana SP263 clones).

We observed variable sensitivity, specificity and dynamic range of three anti-PD-L1 antibodies. EPR1161 and SP263 clones had highest correlation (Pearson correlation=0.69) and similar staining characteristics demonstrating robust linear membranous staining on neoplastic cells and on reactive infiltrating Immune Cells (IC). PD-1 expression on IC had inverse correlation with PD-L1 expression on neoplastic cells, suggesting immunosuppressive effect of PD-L1 positive neoplastic cells on tumor microenvironment. Non-GCB DLBCLs had higher PD-L1 positivity than GCB DLBCLs (p=0.0001, 0.01 and 0.06 respectively for EPR1161, DAKO 28-8 and Ventana SP263 Abs, t-test).

Using molecular characterization, we observed that PD-L1 expression marked high risk aggressive DLBCLs, which also had higher TP53 expression, TP53 mutation and/or del 17p13 (Pearson correlation= 0.176). Moreover, higher expression of PD-L1 had negative correlation with mutations associated with GCB-like DLBCLs including BCL2, FOXO1, KMT2D, EZH2 (p=0.003, 0.01, 0.06, 0.08; t-test) and positive correlation with mutations associated with ABC-like DLBCLs including FAT2 (p=0.04, t-test). Interestingly, while PD-L1 expression correlated with high risk molecular and cytogenetic groups of DLBCL, it did not show correlation with BCL2 or C-MYC expression nor with BCL2/C-MYC double expressors.

Most importantly, Kaplan Meyer analysis showed that even in the settings of chemotherapy, higher PD-L1 expression correlated with more aggressive disease, using all three antibodies ( p=0.03, 0.01, 0.04 respectively for Abcam EPR1161, Abcam 28-8 and Ventana SP263 clones).

We conclude that PD-L1 expression reflects the biology of aggressive subsets of DLBCLs, which are also characterized by high risk cytogenetic and molecular biomarkers. PD-L1 expression can be measured by several antibody clones, which have different performance characteristics but show positive correlation.