PD-L1 gene alterations identify a subset of diffuse large B-cell lymphoma harboring a T-cell-inflamed phenotype

Abstract

Programmed death-ligand 1 (PD-L1) expression on malignant cells is a dominant immune escape mechanism across a variety of human cancers. A unique genetic mechanism underlying PD-L1 upregulation has been uncovered in classical Hodgkin lymphoma (cHL), in which copy gains of the chromosomal region (9p24.1) containing the programmed death-1 (PD-1) ligands PD-L1 and PD-L2 are recurrently observed. While chromosome 9p24.1 copy-number alterations are ubiquitous in cHL, they also occur in diffuse large B-cell lymphoma (DLBCL), albeit with a lower incidence. Here, fluorescence in situ hybridization was used to identify DLBCLs harboring PD-L1 gene alterations, thereby enabling a characterization of the immunogenomic landscape of these lymphomas. Among 105 DLBCL cases analyzed, PD-L1 alterations were identified in 27%. PD-L1 alterations were highly enriched among non-germinal center DLBCLs and exhibited robust PD-L1 protein expression. These lymphomas were heavily infiltrated by clonally restricted T cells and frequently downregulated human leukocyte antigen expression. RNA sequencing of PD-L1-altered DLBCLs revealed upregulation of genes involved in negative T-cell regulation and NF-κB pathway activation, while whole-exome sequencing identified frequent mutations in genes involved in antigen presentation and T-cell costimulation. Many of these findings were validated in a large external data set. Interestingly, DLBCL patients with PD-L1 alterations had inferior progression-free survival following front-line chemoimmunotherapy; however, in the relapsed/refractory setting, PD-L1 alterations were associated with response to anti-PD-1 therapy. Collectively, our results indicate that PD-L1 alterations identify a unique biological subset of DLBCL in which an endogenous antilymphoma immune response has been activated, and that is associated with responsiveness to PD-1 blockade therapy.

Graphical abstract

Figure 1.

PD-L1 gene alterations are recurrently observed in DLBCL and are associated with inferior PFS to front-line chemoimmunotherapy. (A) Chromosome 9 structure. Enlarged below is the 9p24.1 region encoding PD-L1 and PD-L2. PD-L1, centromere 9, and translocation FISH probes are shown in orange, blue, and green respectively. (B) Incidence of PD-L1 alterations according to underlying mechanism. (C) Representative DLBCL cells with the indicated PD-L1 locus status as assessed by FISH. Arrows indicate signals for PD-L1 (orange) and translocation (green) probes. Split signals are seen between PD-L1 and translocation probes in the representative DLBCL with a PD-L1 translocation. (D) Waterfall plot demonstrating the frequency of lymphoma cells (30 analyzed/specimen) with the indicated PD-L1 locus status. The horizontal black line represents the minimum number of lymphoma cells required for a case to be considered PD-L1 altered (20% of cells). (E) Incidence of PD-L1 alterations by cell of origin (*P < .05, Fisher’s exact test). (F) PFS of DLBCL patients with or without PD-L1 gene alterations. (*P = .024, log-rank test). (G) PFS of DLBCL patients according to specific PD-L1 alteration (*P = .012, log-rank test). Clinical and genetic univariate (H) and multivariate (I) risk models for PFS in DLBCL.

Figure 2.

PD-L1 gene alterations are associated with increased PD-L1 protein expression, enhanced CD8⁺ T-cell infiltration, and decreased HLA class I expression in DLBCL. (A) Percentage of PD-L1–positive (H-score ≥30) DLBCLs according to PD-L1 locus status (**P < .01, Fisher’s exact test) (B) Violin plot demonstrating PD-L1 H-scores for DLBCLs with the indicated PD-L1 locus status (***P < .001 and ****P < .0001, Mann-Whitney U test). (C) Numbers of CD8⁺ T cells/hpf in DLBCLs with vs without PD-L1 gene alterations (*P < .05, Mann-Whitney U test). (D) TCR-β repertoire diversity among DLBCLs with and without PD-L1 alterations as measured by Shannon entropy (**P < .01, Mann-Whitney U test). (E) Overall contribution to the TCR repertoire by the top 10 TCR-β sequences in PD-L1 altered vs unaltered DLBCLs (P < .01, Mann-Whitney U test). (F) Representative IHC staining for HLA class I and II expression in DLBCL (40× magnification). Note in the HLA class I negative case the positive HLA I staining on infiltrating immune cells serving as an internal positive control. (G-H) Quantitative data showing percentage of DLBCLs with decreased/absent HLA class I (*P < .05, Fisher’s exact test) and II (P = .063) expression according to PD-L1 locus status.

Figure 3.

RNA-seq identifies DEGs in PD-L1–altered relative to PD-L1–unaltered DLBCL. Heatmap (A) and volcano plot (B) depicting 137 DEGs in DLBCLs with vs without PD-L1 gene alterations (false discovery rate-adjusted P < .05, fold change ≥1.5 or ≤−1.5). (C) GO analysis performed on the 137 DEGs. GO terms significantly enriched in DLBCLs with PD-L1 gene alterations are denoted by red dots. (D) IPA revealing predicted upstream regulators of gene expression that are activated (positive z score) and inhibited (negative z score) in DLBCLs with PD-L1 alterations relative to those without. Dashed vertical line indicates the position of P = .05 on the x-axis. (E) GSEA demonstrating enrichment of NF-κB–regulated genes in PD-L1–altered compared with PD-L1–unaltered DLBCLs. IFN-γ, interferon-γ; NES, normalized enrichment score; PTEN, phosphatase and tensin homolog.

Figure 4.

WES reveals PD-L1–altered DLBCLs are enriched for mutations in immune-related genes. (A) Profiles of recurrent nonsynonymous somatic mutations in DLBCLs with and without PD-L1 alterations. Each column represents a separate DLBCL case. Above each column is the mutational burden of each case as assessed by the total number of nonsynonymous somatic mutations. (B) Nonsynonymous somatic mutations (NSSMs) in genes involved in antigen presentation and T-cell costimulation in DLBCLs with vs without PD-L1 gene alterations (P = .02, Mann-Whitney U test). (C) Total mutational burden in DLBCLs with and without PD-L1 alterations (P = .34, Wilcoxon rank sum test with continuity correction).

Figure 5.

PD-L1 gene alterations are associated with objective response following pembrolizumab treatment in relapsed/refractory DLBCL patients. (A) Incidence and mechanism of PD-L1 gene alterations among 29 DLBCL patients enrolled on the KEYNOTE-013 study of pembrolizumab. (B) Swimmer’s plot demonstrating response characteristics and PFS to pembrolizumab according to PD-L1 locus status. Arrow depicts ongoing response. (C) Positron emission tomography (PET) scan images before and after pembrolizumab treatment in a patient with a PD-L1 gene–amplified DLBCL prior to and following 4 cycles of pembrolizumab treatment. PD, progressive disease; PR, partial response; SD, stable disease.

Figure 6.

Graphical and schematic summaries of immune landscapes in DLBCLs with and without PD-L1 gene alterations. (A) Summary radar plot (left), with 20% radial intervals depicting 7 parameters that characterize the immunogenomic landscape of PD-L1 gene–altered vs unaltered subsets. Indices were normalized from 0 to 1; bold connecting lines represent average scores for PD-L1–altered (red) vs PD-L1–unaltered (blue) subsets. Curves for individual patients are represented as gray lines (left) or in individual radar plots (right), colored by PD-L1–altered (red) or PD-L1–unaltered (blue) status. (B) Schematic representation of the immunogenomic landscapes of PD-L1 altered DLBCLs. PD-L1 gene–altered DLBCLs are characterized by frequent loss-of-function mutations in TNFAIP3 (A20) and increased NF-κB pathway activation. As a result, these lymphomas demonstrate evidence of enhanced immune surveillance, including robust infiltration by clonally restricted T cells. In addition to marked upregulation of PD-L1 protein on the lymphoma cell surface, PD-L1 gene–altered DLBCLs also commonly downregulate HLA I and II expression and harbor loss-of-function mutations in other genes critical for maintaining tumor immune surveillance.