Introduction PD-L1 and its receptor PD-1 are both expressed on multiple cell types, including antigen-presenting cells, T cells and tumor cells. In DLBCL, PD-L1 expression on non-malignant immune cells is common, whereas tumor-cell PD-L1 expression is restricted to a small percentage of pts. Because PD-L1/PD-1 are expressed on macrophages after in vitro activation, and activated macrophages contribute to antibody-dependent cellular phagocytosis (ADCP), high PD-L1 expression may identify tumors enriched with activated tumor-infiltrating macrophages primed for ADCP. Considering the role of ADCP in clinical responses to anti-CD20 (rituximab/obinutuzumab), we hypothesized that PD-L1 expression could reflect abundance of activated tumor-infiltrating macrophages and may be associated with prolonged PFS in 1L DLBCL. We leveraged 2 large Phase 3 trials and cell-line immune gene expression to better understand the biological and clinical relevance of PD-L1 expression in 1L DLBCL.
Methods We used clinical, genomic and pre-treatment immunohistochemistry (IHC) data from 1L DLBCL pts treated in the MAIN (NCT00486759) and GOYA (NCT01287741) trials. Pre-treatment tissue microarray cores from 201 MAIN pts were stained for PD-L1 (clone SP263), with 21 uninvolved lymph node FFPE sections used as controls. PD-L1+ samples were defined as IHC > 1+ (> 1% tissue area occupied by PD-L1+ cells). Tumor cells were identified based on morphological/cytological features, and PD-L1 tumor expression was scored as a percentage. Activated B-cell (ABC), germinal center B-cell (GCB) and unclassified DLBCL prognostic subtypes were determined by the NanoString Lymph2Cx cell-of-origin (COO) assay. We quantified mRNA expression of PD-L1, macrophage and B-cell genes using TruSeq (Illumina) RNAseq among 702 DLBCL pts (GOYA, n = 552; MAIN, n = 150), with 28 DLBCL cell lines and 15 purified normal B-cell samples as comparators. High/low expression of each gene signature was determined by median cutoffs. Median follow-up durations were 24 and 29 mo in MAIN and GOYA, respectively (≈ 30% of PFS events in both cohorts). We used Cox regression to examine associations between these markers and PFS, adjusting for IPI, age, COO, sex and treatment arm.
Results Among 201 DLBCL pts in MAIN, PD-L1 was primarily detected on inter-follicular cells with myeloid/dendritic morphology, similar to the PD-L1 staining pattern of normal lymph nodes. In samples with evaluable germinal centers (GCs), a subset of GC cells expressed PD-L1 with a punctate and granular pattern consistent with a myeloid origin. Overall, there were significantly more pts with high PD-L1 expression by IHC (IHC 3+) and mRNA among ABC vs GCB DLBCL pts. A minority of DLBCL pts (10%) had PD-L1+ tumor cells, consistent with low/undetectable CD274 (PD-L1) mRNA in the majority of DLBCL cell lines and purified normal B-cell samples tested. Additionally, among the 702 evaluable pts in both studies, CD274 mRNA inversely correlated with a B-cell gene signature (MAIN, r= −0.55; GOYA, r= −0.32). High PD-L1 expression (IHC 2+/3+) correlated with prolonged PFS (HR, 0.44 [0.24, 0.80]; P < 0.01) among the 201 evaluable pts in MAIN. Among 552 evaluable pts in GOYA, CD274 expression was higher in pts with a PET-CR (P= 0.02), particularly in pts with ABC DLBCL (P= 0.0001), with a trend toward prolonged PFS in the ABC subtype (HR, 0.80 [0.63, 1.0]; P= 0.054). CD274 mRNA positively correlated with macrophage gene expression in both cohorts (MAIN, r= 0.61; GOYA, r= 0.57); pts with high PD-L1 expression (IHC 2+/3+) also demonstrated high mRNA expression of the macrophage marker CD68. Consistent with PD-L1 reflecting abundance of activated tumor-infiltrating macrophages relevant to anti-CD20 response, high expression of a macrophage signature correlated with prolonged PFS (HR, 0.63 [0.45, 0.87]; P= 0.003) in GOYA and a trend toward prolonged PFS among DLBCL pts treated in MAIN (HR, 0.64 [0.35, 1.16]; P= 0.10), although statistical significance was not reached.
Conclusions In 1L DLBCL, most pts expressed PD-L1 on immune cells with myeloid/dendritic morphology, potentially reflecting activated macrophages primed for ADCP and supportive of the correlation between high PD-L1 expression and improved clinical outcomes among pts treated with immunochemotherapy. In the context of anti-CD20, the biology of PD-L1 expression on macrophages and the impact of PD-L1/PD-1 blockade on ADCP merit further study.
McCord: Genentech, Inc.: Employment. Bolen: Genentech: Employment, Equity Ownership. Koeppen: Genentech, Inc.: Employment. Kadel: Genentech, Inc.: Employment. Fingerle-Rowson: F. Hoffmann-La Roche Ltd: Employment, Equity Ownership. Oestergaard: F. Hoffmann-La Roche Ltd: Employment. Venstrom: Genentech, Inc.: Employment.
Asterisk with author names denotes non-ASH members.
This icon denotes a clinically relevant abstract