BET-Bromodomain Inhibitors Engage the Host Immune System and Regulate Expression of the Immune Checkpoint Ligand PD-L1

Abstract

BET inhibitors (BETi) target bromodomain-containing proteins and are currently being evaluated as anti-cancer agents. We find that maximal therapeutic effects of BETi in a Myc-driven B cell lymphoma model required an intact host immune system. Genome-wide analysis of the BETi-induced transcriptional response identified the immune checkpoint ligand Cd274 (Pd-l1) as a Myc-independent, BETi target-gene. BETi directly repressed constitutively expressed and interferon-gamma (IFN-γ) induced CD274 expression across different human and mouse tumor cell lines and primary patient samples. Mechanistically, BETi decreased Brd4 occupancy at the Cd274 locus without any change in Myc occupancy, resulting in transcriptional pausing and rapid loss of Cd274 mRNA production. Finally, targeted inhibition of the PD-1/PD-L1 axis by combining anti-PD-1 antibodies and the BETi JQ1 caused synergistic responses in mice bearing Myc-driven lymphomas. Our data uncover an interaction between BETi and the PD-1/PD-L1 immune-checkpoint and provide mechanistic insight into the transcriptional regulation of CD274.

Keywords: BRD4; PD-L1; bromodomain inhibitor; immune checkpoint.

Graphical abstract

Figure 1

An Intact Host Immune System Is Required to Elicit Maximal Therapeutic Responses to BET Inhibitors (A–D) Cohorts of C57BL/6 mice (n = 10 per treatment group) were injected intravenously (i.v.) with 1 × 10⁵ Eμ-Myc lymphoma cells 3 days prior to commencement of daily dosing with JQ1 (50 mg/kg; 5 d/w), or DMSO vehicle, via i.p. injection. Kaplan-Meier survival curves representing cohorts of wild-type C57BL/6 (blue line) mice and immune compromised strains (red line) (A) C57BL/6.Rag1^−/− or (B) C57BL/6.Rag2cγ^−/− inoculated with Eμ-Myc lymphoma ^#1 and treated with JQ1 (solid line) or DMSO vehicle (dashed line). (C) Kaplan-Meier survival curves representing cohorts of wild-type C57BL/6 (blue line) and immune compromised C57BL/6.Rag2cγ^−/− mice (red line) inoculated with Eμ-Myc lymphoma ^#2 and treated with JQ1 (solid line) or DMSO vehicle (dashed line). In all therapy experiments, JQ1 conveyed a significant survival advantage to both immune-competent and immune-deficient mice bearing established Eμ-Myc lymphoma. However, immune-deficient mice (C57BL/6.Rag2cγ^−/− and C57BL/6.Rag1^−/−) succumbed to disease significantly earlier than tumor-bearing wild-type mice despite JQ1 treatment (^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001, Log-rank). (D) Fold-change in median survival across all immune-competent versus immune-deficient experiments (error bars represent SEM; ^∗∗∗p < 0.001, Student’s t test). (E) Gene Set Enrichment Analysis (GSEA) enrichment score plots showing significant correlation between genes downregulated by JQ1 and genes associated with inflammatory and interferon gamma response signatures using RNA-seq analysis of Eμ−Myc lymphoma ^#1 following 2 hr treatment with JQ1. (F) Overlap of inflammatory and interferon gamma response signatures identified in 1E with genes downregulated by JQ1 (Eμ−Myc lymphoma ^#1.2; adjusted (adj.) false discovery rate [FDR] < 0.05, −1 > log FC > 1) and genes involved in T cell costimulation.

Figure 2

PD-L1 Is a Direct Target of BET Inhibition In Vitro and In Vivo (A) Eμ-Myc lymphomas ^#1 and ^#2 transduced with retrovirus to overexpress Bcl-2, and Eμ-Myc^#3 were incubated in vitro with 0.5 μM of JQ1 or DMSO for 24 hr and cell surface Pd-l1 was assessed by flow cytometry. (B) Eμ-Myc lymphoma ^#3 was incubated in vitro with 1 μM of BET inhibitor (10 μM for RVX-208) or DMSO for 24 hr and cell surface Pd-l1 was assessed by flow cytometry. For (A) and (B) representative data from at least three independent experiments is presented as mean MFI of cells cultured and analyzed in triplicate ± SEM (^∗∗∗p < 0.001, Student’s t test). (C) Eμ-Myc lymphoma ^#3 was incubated in vitro with 0.5 μM JQ1 or DMSO control for indicated time points prior to flow cytometry analysis of Pd-l1 expression. Grey shaded line, isotype control antibody; black line, DMSO-treated; red line, JQ1-treated. (D) C57BL/6 bearing established Eμ-Myc lymphoma ^#1 were treated with DMSO or JQ1 (50 mg/kg) i.p. and peripheral lymph node cells were harvested 16 hr later for cell surface Pd-l1 expression by flow cytometry. Graphs show the MFI of Pd-l1 expression gated on live GFP-positive tumor cells. Data presented as mean MFI from three individual mice per group ± SEM. ^∗∗p < 0.01, Student’s t test). (E) C57BL/6 bearing established Eμ-Myc^#1/Bcl-2 lymphoma were treated daily with JQ1 (50 mg/kg; 5 d/w), or DMSO vehicle (n = 5 per treatment group). At day 18, peripheral blood was obtained and tumor cells were assessed by flow cytometry for Pd-l1 and MHC Class I (H-2K^b) expression on tumor cells. Data presented as mean MFI ± SEM (^∗∗∗p < 0.001, Student’s t test). (F) qPCR analysis of Cd274 mRNA levels in Eμ−Myc^#1/Bcl-2, Eμ-Myc^#2/Bcl-2 and Eμ-Myc^#3 following treatment with 1 μM JQ1 or DMSO for indicated time points. Transcript levels are normalized to Gapdh and presented as fold change compared to DMSO. (G) L540 cells were cultured in the presence and absence of JQ1 (1 μM) for 2 hr and MYC and CD274 mRNA levels were determined by qPCR (normalized to DMSO). Data presented as mean fold-change from three separate experiments ± SEM (^∗∗p < 0.01, Student’s t test).

Figure 3

PD-L1 Expression Is Not Modulated by Changes in Expression of Myc (A) Eμ-Myc^#1 were treated in vitro for 30 min with 1 μM JQ1 or DMSO, prior to 30 min labeling of nascent RNA with 4-thiouridine (4sU). 4sU-labeled RNA was then biotinylated and purified prior to qPCR analysis of actin, endogenous or transgenic Myc and Cd274 nascent RNA levels (normalized to DMSO). (B) Eμ-Myc^#1 were transduced with TRMPV.shRNAs targeting Myc (1891 and 2105) or Renilla (REN) and pRetroX-tet-on and selected for venus expression and neomycin resistance. Cells were cultured in the presence of absence of DOX for 18 hr and then stained for Pd-l1 expression by flow cytometry. Data are presented as mean MFI of cells normalized to untreated (-DOX) and analyzed in triplicate ± SEM (^∗p < 0.05, ^∗∗p < 0.01, Student’s t test). (C) Eμ-Myc lymphoma ^#1 was transduced with retrovirus expressing human MYC (MYC) or empty vector control (MSCV). Total MYC and Cd274 mRNA levels were determined by qPCR (normalized to Eμ-Myc/MSCV). Eμ-Myc/MSCV and Eμ-Myc/MYC cells were cultured for 2 hr with 1 μM JQ1 or DMSO, and total MYC and Cd274 mRNA levels were determined by qPCR (normalized to the DMSO-treated condition). Data presented as mean fold-change from three separate experiments ± SEM (^∗∗p < 0.01, ^∗∗∗p < 0.001, Student’s t test). Eμ-Myc^#1 cells were cultured in the presence and absence of JQ1 (1 μM) for 2 hr, and ChIP-seq assays were performed using Abs against Brd4, Myc, and RNA Pol II. (D) Average profile of genome-wide Myc peaks in the presence or absence of JQ1. (E) Occupancy heatmap of genome-wide Myc peaks in the presence or absence of JQ1. (F) ChIP-seq reads for RNA pol II and Brd4 mapped to the Myc locus, alongside Myc RNA-seq reads (both transgenic and endogenous) in Eμ-Myc #1 cells treated for 2 hr with 0.25 μM JQ1 or DMSO. (G) Correlation of MYC and CD274 expression was measured in publically available TCGA datasets using Pearson’s correlation coefficient.

Figure 4

Pharmacological Inhibition or Genetic Depletion of Brd4 Is Associated with Loss of PD-L1 Expression (A) Eμ-Myc lymphoma ^#1 was transduced with TRMPVIR.shScrambled, TRMPVIR.shBRD4.498, and TRMPVIR.shBRD4.500 and exposed to doxycycline (DOX) to induce the DsRed-shRNA gene cassette. Brd4 and Myc mRNA levels were determined by qPCR in the presence or absence of 1 μM DOX for 18 hr. Venus⁺DsRed⁺ cells were sorted by flow cytometry. Transcript levels are normalized to Gapdh and presented as fold change compared to -DOX. Data presented as fold-change from three separate experiments ± SEM (^∗p < 0.05, ^∗∗p < 0.01, Student’s t test). (B) Pd-l1 expression following Brd4 knockdown was determined by flow cytometry following exposure to 1 μM DOX for 18 hr. Representative data are presented as mean MFI of cells cultured and analyzed in triplicate ± SEM (^∗p < 0.05, ^∗∗p < 0.01, Student’s t test). (C) Average profile of genome-wide Brd4 peaks in the presence or absence of JQ1. (D) Occupancy heatmap of genome-wide Brd4 peaks in the presence or absence of JQ1. (E) Metagene analysis of RNA polymerase II occupancy across genes repressed by JQ1 as determined by RNA-seq analysis. (F) Magnification of the RNA polymerase II metagene analysis in the gene body. (G) Brd4, RNA Pol II, and Myc binding at the Cd274 locus was assessed.

Figure 5

JQ1 Inhibits IFN-γ Induced PD-L1 Upregulation (A) AT3 cells were treated with either single agent or combination of 1 ng/mL IFN-γ and 1 μM JQ1 or DMSO for 2 hr prior to 3′-mRNA-seq. Heatmap depicts the top 16 IFN-γ induced genes from our IFN-γ gene signature (Figure S4F). (B) AT3 cells were treated as in (A) prior to qPCR analysis of Cd274 mRNA levels. Data represent mean of three independent experiments ± SEM (^∗∗∗p < 0.001 2-way ANOVA). (C) Representative flow cytometry histograms showing cell surface Pd-l1 expression on AT3 cells were treated as in (A) for 24 hr prior to FACs analysis. (D) AT3 cells were cultured as in (A) for 2 hr and ChIP-seq assays were performed using Abs against Brd4, H3K27Ac, Irf1, RNA Pol II, and Stat1. Binding of these factors at the Cd274 locus was then visualized using Integrative Genomics Viewer (IGV). (E) Correlation of IRF1 and CD274 expression was measured in publically available datasets using Pearson’s correlation coefficient.

Figure 6

BRD4 Is Ubiquitously Required for IFN-γ-Induced PD-L1 (A) Human plasma cell leukemia cell line (ALF1) cells were treated with either single agent or combination of 1 ng/mL IFN-γ and 1 μM JQ1, or DMSO, for 2 hr prior to 3′-mRNA-seq. Heatmap depicts the top 16 IFN-γ induced genes from our common IFN-γ gene signature (Figure S4F). (B) ALF1 cells were treated as in (A) for 24 hr prior to flow cytometry analysis for cell surface PD-L1 expression. Representative data from at least three independent experiments is presented as mean MFI of cells cultured and analyzed in triplicate ± SEM (^∗∗∗p < 0.001, Student’s t test). (C) ALF1 cells were treated as in (A) prior to qPCR analysis of CD274 mRNA levels. Data represent mean of three independent experiments ± SEM (^∗∗∗p < 0.001 two-way ANOVA). (D) HT29, MCF7, SET2, SKBR3, Z119, BV173, HCT116, HEL, and KG1 cells were incubated in the absence or presence of IFNγ (100 U/ml), JQ1 (2.5 μM), or the combination of both for 48 hr prior to analysis of cell surface PD-L1 expression by flow cytometry. Results were calculated as staining index (mean fluorescence intensity of PD-L1 relative to mean fluorescence intensity obtained with an isotype-matched control antibody), expressed as percent of control (control PD-L1 expression = 100%) for each cell line (^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001; two-way ANOVA). (E) Primary CML (chronic phase) CD34⁺/CD38⁻ cells were incubated in the absence or presence of IFN-γ (100 U/ml), JQ1 (2.5μM) or the combination of both for 48 hr. Expression of PD-L1 was determined by FACS. Results were calculated as staining index (mean fluorescence intensity [MFI] of PD-L1 relative to MFI obtained with an isotype-matched control antibody), expressed as percent of control (control PD-L1 expression = 100%) and represent the mean ± SD from four donors (^∗p < 0.05, ^∗∗p < 0.01; two-way ANOVA).

Figure 7

Sustained PD-L1 Expression Reduces the Efficacy of JQ1 (A–C) Eμ-Myc lymphoma ^#1 was transduced with murine stem cell virus expressing murine Pd-l1 (Pd-l1) or empty vector control (MSCV). BFP⁺ cells were isolated by FACS sorting and transplanted into cohorts of C57BL/6 mice (1.5 × 10⁵ cells/mouse), which commenced daily JQ1 therapy (50 mg/kg; 5 d/w) from day 3 post-intravenous lymphoma inoculation. (A) Pd-l1 expression was assessed by flow cytometry on BFP-expressing Eμ-Myc lymphoma cells in the peripheral blood at day 14 post-intravenous inoculation. (B) The percentage of BFP-expressing Eμ-Myc lymphoma cells (as % of total live cells in the blood) was assessed by flow cytometry at day 7 post-intravenous inoculation as a surrogate marker of tumor burden. (C) Kaplan-Meier survival curve representing overall survival of mice bearing Eμ-Myc lymphoma ^#1/MSCV or ^#1/Pd-l1 and treated with JQ1 or DMSO vehicle (n = 6 per treatment group). Overall dosing period is indicated by gray shaded area. (D) Kaplan-Meier survival curves representing cohorts of C56BL/6 (n = 6 per treatment group) bearing Eμ-Myc lymphoma ^#2 treated with JQ1 (50 mg/kg), or DMSO vehicle, via i.p. Injection commencing day 3 post-transplant for a total of 25 doses. Mice received 100 μg of anti-PD-1 (clone RPMI-14) or Rat IgG isotype antibody via i.p. injection on days 5, 10, 15, and 20 post-transplant. (E) Kaplan-Meier survival curves representing cohorts of C56BL/6 (n = 6 per treatment group) bearing Eμ-Myc lymphoma ^#2 treated with JQ1 (50 mg/kg), or DMSO vehicle, via i.p. injection commencing day 3 post-transplant for a total of 25 doses. Mice received 100 μg of anti-4-1BB (clone 3H3) or Rat IgG isotype antibody via i.p. injection on days 5, 8, and 11 post-transplant.