Epstein-Barr virus-encoded EBNA2 alters immune checkpoint PD-L1 expression by downregulating miR-34a in B-cell lymphomas

Abstract

Cancer cells subvert host immune surveillance by altering immune checkpoint (IC) proteins. Some Epstein-Barr virus (EBV)-associated tumors have higher Programmed Cell Death Ligand, PD-L1 expression. However, it is not known how EBV alters ICs in the context of its preferred host, the B lymphocyte and in derived lymphomas. Here, we found that latency III-expressing Burkitt lymphoma (BL), diffuse large B-cell lymphomas (DLBCL) or their EBNA2-transfected derivatives express high PD-L1. In a DLBCL model, EBNA2 but not LMP1 is sufficient to induce PD-L1. Latency III-expressing DLBCL biopsies showed high levels of PD-L1. The PD-L1 targeting oncosuppressor microRNA miR-34a was downregulated in EBNA2-transfected lymphoma cells. We identified early B-cell factor 1 (EBF1) as a repressor of miR-34a transcription. Short hairpin RNA (shRNA)-mediated knockdown of EBF1 was sufficient to induce miR-34a transcription, which in turn reduced PD-L1. MiR-34a reconstitution in EBNA2-transfected DLBCL reduced PD-L1 expression and increased its immunogenicity in mixed lymphocyte reactions (MLR) and in three-dimensional biomimetic microfluidic chips. Given the importance of PD-L1 inhibition in immunotherapy and miR-34a dysregulation in cancers, our findings may have important implications for combinatorial immunotherapy, which include IC inhibiting antibodies and miR-34a, for EBV-associated cancers.

Fig. 1

PD-L1 expression in EBV-infected and EBNA2-transfected BL and DLBCL cell lines. a EBNA2 and PD-L1 expression in Mutu I (latency I expressor) and its latency III-expressing counterpart. Furthermore, expression of PD-L1 and EBNA2 is shown in two additional BLs with resident viral genomes. Daudi carries an EBNA2-deleted EBV strain and Jijoye is EBNA2-positive BL. b Two GC DLBCLs, namely U2932 and SUDHL5, were infected with a recombinant Akata strain of EBV. Total cell lysates were electrophoresed and EBNA2 expression was verified by immunoblotting using PE2 monoclonal antibodies. PD-L1 expression was analyzed using rat monoclonal antibodies. c Two BL cell lines, Oma4 and DG75, were infected with the recombinant Akata virus and tested for EBNA2 and PD-L1 expression. β-actin is used as loading control. d PD-L1, EBNA2, and LMP1 expression in transfected U2932 DLBCL and e BL41 is an EBV-negative BL. The BL41K3 derivatives are estrogen-inducible EBNA2 transfectants. PD-L1, EBNA2, and β-actin expression was analyzed before and after β-estradiol treatment. β-actin is used as loading control

Fig. 2

EBNA2 expression decreases miR-34a in transfected B lymphoma cells by affecting its transcription. miR-34a (top panels) and pre-miR-34a expression (middle panels) was verified by real-time qPCR in (a) U2932 DLBCL transfected with EBNA2 and (b) BL41K3 are transfected with estrogen-inducible EBNA2. miR-34a promoter carrying Luc reporter activity was analyzed in U2932 and BL41 and their EBNA2-expressing derivatives (lower panels). The figure shows standard deviation (SD) of the average of three different experiments. Each experiment was performed in biological triplicates for each sample and repeated three times. p values were calculated with unpaired t test. In all cell lines miR-34a expression (****) p < 0.0001. Pre-miR-34a: U2932 EBNA2 cl-1 (**) p = 0.0020 and cl-2 (**) p = 0.0027, BL41K3 (****) p < 0.0001. MiR-34a promoter: U2932 EBNA2 cl-1 (***) p = 0.0002 and cl-2 (**) p = 0.0011, BL41K3 (****) p < 0.0001

Fig. 3

miR-34a targets 3′UTR of PD-L1 in EBNA2-transfected U2932 cells and site-directed mutagenesis of its seed sequence abrogates its binding to PD-L1 3′UTR. a Luciferase reporter construct containing wild-type 3′UTR PD-L1 was transfected in presence of either miR-34a inhibitor in MPA vector control transfectants or miR-34a mimic in U2932 EBNA2-expressing clone. Each transfection was performed in triplicate. For U2932 EBNA2 cl-1 is (*) p = 0.0172. b The specificity of miR-34a binding to its seed sequence in 3′UTR of PD-L1 was confirmed by mutating seed sequence with site-directed mutagenesis. The mimic miR-34a bound to the wild-type 3′UTR of PD-L1 and reduced luc activity. The inhibitory effect of mimic miR-34a was abrogated when its seed sequence in PD-L1 3′UTR was mutated. Each transfection was performed in triplicate. (**) p = 0.0026 refers to U2932 MPA Vector and (**) p = 0.0021 refers to U2932 EBNA2 cl-1. p values were calculated with unpaired t test

Fig. 4

Overexpression of miR-34a in EBNA2-expressing U2932 downregulates PD-L1. a miR-34a mimic was transfected into U2932 MPA vector and EBNA2-expressing clone 1. The transfected cells were processed for PD-L1 expression by flow cytometry. The lower panel indicates an average of three flow cytometry experiments and its significance. (*) p = 0.0109. b The effect of miR-34a reconstitution on PD-L1 expression in the vector control and EBNA2-transfected U2932 cells was also confirmed by immunoblotting. An average of the corresponding densitometric analysis on three such experiments is shown in the lower panel (**) p = 0.0055. β-actin served as loading control

Fig. 5

EBNA2 suppresses miR-34a transcription through EBF1. a Prediction of EBF1 binding motifs at the miR-34a promoter in human reference genome hg38, coordinates chr1:9181678-9182943, was performed with the JASPAR database and visualized with IGV. EBNA2 peak overlaps with the second of the predicted EBF1 binding sites, highlighted in green square, on the miR-34a promoter in dataset GSM2039170. b Knockdown of EBF1 was verified by Q-PCR in the U2932 cell line and U2932 EBNA2 cl1 transduced with pLK0.1 lentiviral vectors which carry shEBF1 and control shRNA. Upon EBF1 depletion (***p = 0.0004) in U2932 EBNA2 cl1, expression of mature miR-34a (**p = 0.0023) and pre-miR-34a (**p = 0.0024) was derepressed and consequently PD-L1 expression decreased (***p = 0.0008). Q-PCRs were performed with three biological and technical triplicates for each sample. c MiR-34a promoter activity was enhanced upon EBF1 knockdown in U2932 EBNA2 cl1, after 48 h. Luciferase assay was performed three times and each sample was in triplicate. (*p = 0.0105). Statistical analysis was performed using an unpaired t test (Prism-7)

Fig. 6

miR-34a relieves suppression of immunogenicity induced by EBNA2 as measured in MLR and 3D biomimetic microfluidic coculture devices. a T cells were activated in plates coated with anti-CD3/anti-CD28 antibodies for 72 h. Irradiated targets U2932 MPA vector and U2932 EBNA2 cl-1 were cocultivated with activated T cells (effector). The effector-target ratio was 1:10. The target cells were transfected with mimic control or miR-34a mimic 24 h prior to cocultivation with the effector cells. The coculture was carried out for 48 h and the cells were stained for CD4/CD8 and IFN-γ and processed for flow cytometry. Data are expressed as mean ± SD. The p values for T-cell activation without stimulators are (*) p ≤ 0.05 for CD8 and CD4. In MLR with U2932 stimulators, the statistical significance is (*) p = 0.028 for CD8 and (**) p = 0.0081 for CD4. Three different experiments were performed with PBMCs isolated from three different donors. b Three-dimensional biomimetic microfluidic coculture devices: Four million/ml U2932 EBNA2 cells tranduced with lentiviral vector controls were introduced into the microfluidic devices. In coculture experiments (right panel in b), devices were first seeded with U2932 EBNA2 cells and were incubated for 24 h at 37 °C, followed by activated T cells seeding. Immunostaining was performed after 48 h of cocultivation. bi: Representative confocal images of U2932 EBNA2 cl-1 transduced with GFP-lentiviral vector control in the absence or presence of activated T cells; bii Four million/ml U2932 EBNA2 cl 1, tranduced with miR-34a lentivirus, were introduced in the collagen/fibronectin devices either alone (left panel) or cocultivated with previously activated T cells (right panel). The cocultivation of target, miR-34a transduced U2932 EBNA2 cells with activated T cells, was carried out for 48 h before immunostaining. miR-34a containing lentivirus-transduced U2932 EBNA2 cl-1 cells were stained with anti-GFP antibody (green), activated CD8/CD4 T cells were stained with anti-CD8 and anti-CD4 (magenta), apoptotic U2932 EBNA2 cl-1 cells were visualized with anti-caspase-3 antibody (red), and nuclei were counterstained with DAPI (blue). The overlap of miR-34a transduced U2932 EBNA2 cells (GFP positive) and caspase-3 (red) indicates tumor cell death (merged images, yellow). Scale bar = 100 μm, magnification of the insets = scale bar 20 μm. c Arbitrary Caspase-3 units were calculated for each experimental condition. Statistically significant caspase-3 positive cells were observed only in miR-34a tranduced EBNA2 expressors; Data expressed as mean ± SEM; (****) p < 0.0001. N = 4 fields (3–4 devices for each experimental condition). Shown is one representative experiment out of four performed. Statistical analysis was performed with Prism-7 software using unpaired T test

Fig. 7

PD-L1 expression in DLBCL clinical tissues. a Three non-GC DLBCL patient samples representing the three ABC DLBCL categories, out of a total of 21, stained for PD-L1 are shown. Paraffin sections were immunostained for PDL1 using an automated immunostainer (DAKO, Glostrup, Denmark). As control for PDL-1 immunostaining, sections from paraffin-embedded human lung carcinoma were used. b The stained tissue sections were digitalized at a ×40 magnification using Aperio Scan Scope. The percentage positivity was calculated by counting positive cells in three squared areas measuring 50,000 μm² from each clinical sample in (a) above. The number of positive cells was determined using Aperio software IHC Membrane v1. This algorithm detects membrane staining for individual tumor cell in the selected regions and quantifies the intensity and completeness of the membrane staining. Unpaired t test was applied to demonstrate that differences in % total PD-L1-positive cells and % cells with strong staining intensity were statistically significant, (*) p = 0.0125, (**) p = 0.0040