Phosphorylation and stabilization of PD-L1 by CK2 suppresses dendritic cell function

Abstract

Targeting immune checkpoints such as programmed cell death 1 (PD-1) and programmed cell death-ligand 1 (PD-L1) has transformed cancer treatment, with durable clinical responses across a wide range of tumor types. However, a high percentage of patients fail to respond to anti-PD-1/PD-L1 treatment. A greater understanding of PD-L1 regulation is critical to improving the clinical response rate of PD-1/PD-L1 blockade. Here, we demonstrate that PD-L1 is phosphorylated and stabilized by casein kinase 2 (CK2) in cancer and dendritic cells. Phosphorylation of PD-L1 at Thr285 and Thr290 by CK2 disrupted PD-L1 binding with speckle-type POZ protein (SPOP), an adaptor protein of the cullin 3 (CUL3) ubiquitin E3 ligase complex, protecting PD-L1 from CUL3-mediated proteasomal degradation. Inhibition of CK2 decreased PD-L1 protein levels by promoting its degradation and resulted in the release of CD80 from dendritic cells to reactivate T-cell function. In a syngeneic mouse model, combined treatment with a CK2 inhibitor and an antibody against T-cell immunoglobulin mucin-3 (Tim-3) suppressed tumor growth and prolonged survival. These findings uncover a mechanism by which PD-L1 is regulated and suggest a potential anti-tumor treatment option to activate dendritic cell function by blocking the CK2-PD-L1 pathway and inhibiting Tim-3.

Introduction

Immune checkpoint blockade (ICB), such as anti-PD-1, anti-PD-L1, and anti-CTLA4, has transformed cancer treatment with unprecedented and durable clinical response in various cancer types and those that are difficult to treat (1); however, ICB therapies are ineffective in a significant percentage of patients (2). Recent studies have reported that expression of PD-L1 on either tumor cells or on tumor-infiltrating immune cells correlates with response to PD-1/PD-L1 blockade (3). To better understand the regulation of PD-L1 protein expression, we sought to identify PD-L1 interacting protein kinases through mass spectrometry (4). We selected kinase targets based on three criteria: 1) the targets can serve as therapeutic targets, eg those that already have FDA-approved drugs or inhibitors in clinical trials; 2) the mediated pathway by the target is novel in regulating PD-L1 expression and 3) the target kinase is relevant in cancer. Casein kinase 2 subunit alpha (CK2α), an enzyme encoded by the CSNK2A1 gene, stood out as the kinase that fits to these criteria (Supplementary Table S1) (4). CK2α is the catalytic subunit of the constitutively active serine/threonine casein kinase 2 (CK2) tetrameric complex that phosphorylates a large number of substrates. CK2 is essential for cell survival. CK2 has been found to be upregulated in all cancers that have been examined and associates with poor prognosis in cancers (5–7). Elevated levels of CK2 have been associated with increased cell growth and proliferation (8). In addition, CK2 also acts as a potent suppressor of apoptosis (8). CK2 downregulation impacts not only cell growth and proliferation but also apoptotic activity in cancer cells, thus CK2 has been proposed as a potentially important target for cancer therapy (9,10). Potent and highly selective CK2 inhibitors have been advanced to clinical trials (11). CK2’s involvement in the regulation of immune responses in the tumor microenvironment (TME) has been investigated. CK2 affects immune responses through regulation of NF-kB, COX-2, JAK/STAT, HIF-1α, AKT, Wnt, ERK, Notch and Ikaros (11). Inhibition of CK2 substantially reduced the amount of polymorphonuclear myeloid-derived suppressor cells (PMN-MDSC) and tumor-associated macrophages (TAM). CK2 inhibition dramatically enhanced the antitumor activity of immune checkpoint receptor blockade using anti-CTLA-4 antibody (12). Our TCGA pan-cancer analysis also found that CK2α was upregulated in multiple cancer types (Supplementary Fig. S1A), and high CK2α correlated with poor prognosis (Supplementary Fig. S1B). The mRNA expression of CK2α was negatively correlated with that of molecules involved in T-cell activation, such as T-cell surface glycoprotein CD8 alpha (CD8A), perforin 1 (PRF1), granzyme A (GZMA), and granzyme B (GZMB) (Supplementary Fig. S1C). These results together suggested that CK2 might play a role in immune suppression and result in poor prognosis. For convenience, we will use CK2 to represent CK2α in the following discussion.

Materials and Methods

Reagents and antibodies

CX4945 (Silmitasertib, A11060) was purchased from AdooQ Bioscience (Irvine, CA, USA), dissolved in DMSO and used at 7.5–10 μM for cell culture work; MG132 was from Sigma (St. Louis, MO, USA), dissolved in DMSO and used at 10 μM for cell culture work; human (431807) and mouse (431007) IL-2 ELISA Kits were from Biolegend (San Diego, CA, USA); and human (430104) and mouse (430804) IFNγ ELISA Kits were from Biolegend (San Diego, CA, USA). The following antibodies were used for Western blotting: PD-L1 (E1L3N 1:1000; Cat. No. 13684, RRID:AB_2864409; Cell Signaling), CK2a (1:1000, A300–198A, RRID:AB_185571; Bethyl), Myc-Tag (1:3000, Cat. No. 2276, RRID:AB_331783; Cell Signaling), FLAG Tag (1:1000, Cat. No. 2368; Cell Signaling), Phospho-Akt-Ser473 (736E11 1:1000; Cat. No. 3787, RRID:AB_331170; Cell Signaling), Phospho-AKT-S129 (1:1000, Ab133458, Abcam), SPOP (1:1000, 16750–1-AP, RRID:AB_2756394; Proteintech), Vinculin (1:5000, 66305–1-Ig; Proteintech), Actin (1:3000, A2066; Sigma), and Tubulin (1:3000, T5168; Sigma). The FLAG tag M2 antibody (Sigma) was used for immunoprecipitation. The anti-PD-L1 antibody for flow cytometry staining was from BD Pharmingen, # 563741 (used at 5 μl for 1×10⁶ cells in 100μl PBS). Mouse antiserum against the phosphorylation site of PD-L1 at Thr 290 were produced with a synthetic phosphopeptide: CDTNSKKQSDTHLEETp. The following antibodies were used for immunohistochemistry (IHC): PD-L1 (1:100, 28–8; Abcam), phospho-T290-PD-L1 (1:50) and CK2a (1:50, A300–198A; Bethyl).

Cell culture

HEK293T, MDA-MB-231 (RRID:CVCL_ZZ22), BT-549, PC3, DU145, H460, H1975, Jurkat T, THP-1, EL4, DC2.4, 4T1, EMT6 (NCI-DTP Cat# EMT-6, RRID:CVCL_1923) and B16-F10 cell lines were purchased from American Type Culture Collection (ATCC) and cultured in DMEM high glucose medium supplemented with 10% fetal bovine serum (FBS; GE Healthcare Life Sciences), 100 units of penicillin, and 100 μg/ml streptomycin. All cell lines were routinely tested for mycoplasma contamination and were independently validated by short tandem repeat DNA fingerprinting at The University of Texas MD Anderson Cancer Center (Houston, TX, USA). HEK293T cells were used for packaging of lentiviral cDNA expressing viruses. Medium with secreted viruses were collected at 48 hours after transfection. After filtering, viruses were used to infect cells in the presence of 4 μg/mL polybrene (Sigma-Aldrich). Cells were split 48 hours post infection and selected on puromycin (1 μg/mL), hygromycin B (200 μg/mL) or blasticidin (10 μg/mL) for 3 days.

Co-culture system, IL-2 and IFNγ expression measurement

Human T cells were collected from human peripheral blood mononuclear cells (PBMC; STEMCELL Technologies, Vancouver, BC, Canada). Mouse T cells were collected from C57BL/6 mice. Human and mouse T cells were activated with Dynabeads Human (11161D; Thermo Fisher Scientific) or Mouse (11456D; Thermo Fisher Scientific) CD3/CD28 T Cell Activator and then co-cultured with tumor cells at ratio of 1:5 for 24h as previous reported. Secreted IL-2 and IFNγ in culture medium was detected in triplicate according to the instructions of Human or Mouse IL-2 (# 431807 # 431007) and IFNγ (#430104, #430804) Elisa kit from BioLegend.

Plasmids

A pGIPZ dual expression construct for knockdown and re-expression of FLAG PD-L1 was constructed as described previously (13). The T285A, T290A, T285E, T290E, T285/290AA and T285/290EE mutants were generated via site-directed mutagenesis. Plenti-GFP-CK2 plasmid was purchased from Functional Genomics Core at MD Anderson Cancer Center. The Myc tag was added to the C-terminus of plenti-GFP-CK2 via site-directed mutagenesis. The following pLKO-CK2α shRNA vectors were obtained from Sigma:

Human-shCK2α−1: TRCN0000350294.

Human-shCK2α−2: TRCN0000320928.

Mouse-shCK2α: TRCN0000361110.

Immunoblotting and immunoprecipitation

Cells were lysed in co-immunoprecipitation (co-IP) buffer containing 25 mM Tris pH 7.4, 150 mM NaCl, 0.1% NP-40, 1 mM EDTA, 10% glycerol supplemented with protease inhibitors (Complete Mini, Roche). Equal amounts of protein were resolved by SDS-PAGE and immunoblotted with primary antibodies. For immunoprecipitation, cells were lysed in co-IP buffer. Lysates (1 mg) were mixed with primary antibodies overnight at 4 °C and then pulled down using protein G beads. The immunocomplexes were washed three times with co-IP buffer, resolved by SDS-PAGE, and immunoblotted with antibodies.

Immunohistochemical (IHC) staining

Paraffin-embedded tissue array containing 104 human breast tumor tissues was obtained from Shanghai Jiaotong University. Written informed consent was obtained from patients in all cases at the time of enrollment. IHC staining was performed as described previously (14). Briefly, antigen retrieval was carried out by heating in 0.01 M sodium citrate buffer (pH 6.0). Sections were incubated with 1% H₂O₂ for 30 min at room temperature to block endogenous peroxidase activity. After incubating in normal goat or horse serum for 1 h to block non-specific binding of IgG, sections were incubated with primary antibody at 4 °C overnight. Sections were then incubated for 1 h with biotinylated secondary antibodies (vector) followed by incubation with avidin-biotin peroxidase complex (vector) for 1 h. Samples were developed with 3-amino-9-ethylcarbazole (AEC) and counterstained by Mayer’s hematoxylin. To measure protein expression by IHC, stained sections were analyzed at x400 resolution and scored by H-score system. An H-score (range, 0–300) was calculated as the sum of the product of the highest intensity of staining (0, negative; 1, weak positive; 2, moderate positive; 3, strong positive) and percentage of tumor cells positive (0–100%; with any intensity of positive staining). Two pathologists assessed the immunostaining independently. We set the cut-off value at 50, i.e., H-score >50 as a positive group.

In vitro kinase assay

Kinase assays were performed as described previously (15). GST, GST-PD-L1 wild type, and GST-PD-L1 mutants were purified from E. coli (BL21). Purified recombinant proteins were incubated with 500 units of active CK2 kinase (P6010S, New England Biolabs) in kinase buffer containing 50 mM Tris-HCl at pH 7.6, 10 mM MgCl₂, 2 mM DTT and 0.1 mM EDTA in the presence of 5 μCi of [γ−³²P] ATP and 50 μM cold ATP at 30 °C for 30 min. The kinase reaction was stopped by the addition of SDS sample buffer. The reaction mixture was resolved on SDS-PAGE followed by Western blot analysis with the indicated antibodies. Bands were visualized by autoradiography.

Dot blot analysis of phospho-T290-PD-L1 antibody

Peptides containing phosphorylated (CDTNSKKQSDTHLEETp) or non-phosphorylated (CDTNSKKQSDTHLEET) PD-LI-T290 were synthesized by LifeTein (New Jersey) and blotted at different concentrations in PVDF membrane. The membrane was washed with PBS and blocked with non-fat milk and then incubated with mouse anti-human phospho-T290-PD-L1 antibody. After washing, the membrane was incubated with HRP-conjugated rabbit anti-mouse secondary antibody. After washing, the membrane was developed with ECL. The membrane was stripped out of phospho-T290-PD-L1 antibody and anti-PD-L1 antibody was incubated and detected.

Generation of human mature dendritic cell from THP-1 cells

Following literature (16), THP-1 cells were suspended in serum free culture medium and plated at a concentration of 2 × 10⁵ cells/ml. After incubation with rhIL-4 (200 ng/ml), rhGM-CSF (100 ng/ml), rhTNF-α (20 ng/ml), and ionomycin (200 ng/ml) for 72 hours, total RNA of cells was harvested for real-time PCR analysis. The markers of dendritic cells were examined by real-time PCR using primers listed as following:

CD83-F 5’-AAGGCCCTATTCCCTGAAGA-3’

CD83-R 5’-CTCTGTAGCCGTGCAAAACA-3’

CD80-F 5’-ATGCTGCCTGACCTACTGCT-3’

CD80-R 5’-GGTCAATTGCAAATGGAGGT-3’

CD86-F 5’-TGGAACCAACACAATGGAGA-3’

CD86-R 5’-AAAAAGGTTGCCCAGGAACT-3’

T cell killing assay

Spleen from OT-I TCR transgene mice were collected and disrupted in Hank’s balanced salt solution (HBSS) containing 2% FBS and 1 mM EDTA. Aggregates and debris were removed by passing cell suspension through a 70-μm mesh nylon strainer. Tumor cell-specific T cells were then expanded in vitro by adding ovalbumin (OVA; 2 μg/ml) and interleukin (IL-2; 30 U/ml) for 5 days. Plasmid expressing OVA (pBlueRIP.TfrOVA was a gift from Francis R. Carbone (Addgene plasmid # 69596; http://n2t.net/addgene: 69596; RRID: Addgene_69596)) (17) was transfected into B16 Flag-PD-L1 cells and selected on G418 antibiotic for 2 weeks. Tumor cells were seeded in 12-well plates at a density of 1 × 10⁵ cells per well overnight followed by the addition of T cells at density of 5 × 10⁵ to each well using the ratio of 1:5. After two-day co-culturing, the cells were gently washed with PBS and ice methanol added to each well for incubation at –20 °C overnight. The wells were then stained with crystal violet for 2h at room temperature and washed with PBS. For the quantitative analysis, the crystal violet dye was solubilized with 30% acetic acid and measured by absorbance at 570 nm with microplate reader (Bio-Rad). For co-culture system of dendritic cell and T cell killing, C57BL/6 mice bone marrow dendritic cells (Cell Biologics, Chicago) were treated with or without CK2 inhibitor for 24h. Then dendritic cells were co-cultured in the presence of OVA with OT-I T cells and B16 Flag-PD-L1-OVA cells for three days. The wells were then stained with crystal violet for 2h at room temperature and washed with PBS. For the quantitative analysis, the crystal violet dye was solubilized with 30% acetic acid and measured by absorbance at 570 nm with microplate reader (Bio-Rad).

Database analysis

A user-interactive website (http://timer.cistrome.org/) was used to show the CK2 gene expression between tumor and normal tissue (18). Another user-interactive website (http://tide.dfci.harvard.edu/) was used to show the correlation between CK2 expression and patient survival (18). And several representative cancer types, like neuroblastoma, melanoma, TNBC, kidney cancer, lung cancer and pancreatic cancer were selected to show the correlation. TCGA data of skin cutaneous melanoma (TCGA, Pan Cancer Atlas) cohort was obtained from cBioPortal (RRID:SCR_014555) database (http://www.cbioportal.org/) to explore the correlation between CK2 mRNA expression and T cell killing marker mRNA expression, such as CD8, PRF1, GZMA, GZMB (19).

Animal studies

The BALB/c, C57BL/6 and nude mice (6–8 weeks old, female) were purchased from Jackson Lab. Studies were conducted under the guidelines approved by the Institutional Animal Care and Use Committee (IACUC). Mouse mammary tumor 4T1 (5 × 10⁴) and EMT6 (1 × 10⁵) cells were suspended in 50 μl medium mixed with 50 μl Matrigel basement membrane matrix (BD Biosciences) and injected into the second pair of mammary fat pads of BALB/c mice. EMT6 cells were also injected into nude mice using same method. Mouse melanoma B16F10 cells (5 × 10⁴) were injected into subcutaneous tissue of C57BL/6 mice. Tumor size was measured every two days using a caliper and calculated using the formula length × width²/2. When tumor size reached 50–100 mm³, mice were randomly grouped and administered treatment. For CK2 inhibitor treatment, mice were administered 75 mg/kg CX4945 (dissolved in 0.1% Tween 80 + 0.5% Sodium carboxymethyl cellulose (NaCMC) dissolved in 20% hydroxypropyl-beta-cyclodextrin) by oral gavage twice a day until the tumor size reached limit. For Tim-3 antibody treatment, mice were injected intraperitoneally 100 μg of Tim-3 antibody (clone B8.2C12; Bio X Cell) or control rat IgG (clone HRPN; Bio X Cell) on days 6, 11, and 16. Tumor weight was measured every two days. On day 21, tumor tissue (N = 3/group) was harvested and prepared for single cell suspension for further analysis. The remaining mice (N = 6/group) were monitored until the tumor size reached 1,500 mm³, then euthanized. For toxicity studies, blood (300 μl) was collected from the orbital sinus of mice using a microhematocrit blood tube at the end point of experiment. The blood was subjected to biochemical analysis for liver marker enzymes alanine transaminase (ALT) and aspartate transaminase (AST) and kidney marker by-products creatinine and blood urea nitrogen (BUN) to evaluate treatment toxicity by COSBA INTERGRA 400 plus (Roche Diagnostics, Rotkreuz, Switzerland) at The Department of Veterinary Medicine and Surgery, The University of Texas MD Anderson Cancer Center.

CyTOF analysis

Mouse Tumor Dissociation Kit (130–096-730, Miltenyl Biotec) and gentleMACSOcto Dissociator (130–096-427, Miltenyl Biotec) were used to digest the excised tumors from mice. Cells were then blocked with CD16/CD32 (40477, 1:50; BioLegend, San Diego, CA, USA) antibody followed by incubation with a mixture of metal-labeled antibodies (Supplementary Table S2) for 1 h at room temperature. After incubation, cells were washed with wash buffer three times. Cisplatin (195Pt, Fluidigm) was used as a marker to detect dead cells. Cell-ID Ir-intercalator (Fluidigm, San Francisco, CA, USA) was incubated overnight at 4 °C. The samples were analyzed using the Helios System (Fluidigm). Data were processed by the FlowJo (RRID:SCR_008520) software and Cytobank (RRID:SCR_014043 ) (Cytobank, Inc. Santa Clara, CA, USA).

Statistical analysis

Statistical analyses were performed using SPSS (RRID:SCR_002865 )(V24; SPSS). Fisher’s exact test and Spearman’s rank correlation coefficient were used to compare correlation. Kaplan-Meier survival curves were compared using the log rank test. A p value < 0.05 was considered statistically significant.

Data Availability

Data were generated by the authors and included in the article

The data generated in this study are available within the article and its supplementary data files.

Results

CK2 positively regulates PD-L1 expression

To investigate whether PD-L1 is regulated by CK2, we asked whether alteration of CK2 activity affects PD-L1 protein levels. First, we treated several different cancer cell lines with a clinically used CK2 inhibitor, CX4945 (20) at a concentration of 10 μM for 24 hours, whose activity was validated by the decrease in phosphorylation of AKT, a known CK2 downstream target (21). PD-L1 expression decreased in cells treated with CX4945 (Fig. 1A). As expected, treatment with CX4945 decreased the phosphorylation of AKT at multiple sites, such as S129 and S473. The S129 position is directly phosphorylated by CK2, whereas phosphorylation at S473 dictates the overall activity of AKT. As shown in Figure 1A, a decrease in the phosphorylation at the position S473 was validated in almost all the cell lines tested. P-AKT-S473 reflects the activity of AKT and phosphorylation of S129 activates AKT. The data indicate that CK2 inhibitor can downregulate both p-AKT-S129 and p-AKT-S473 levels. For all the subsequent experiments, we used p-AKT-S473 to reflect the global CK2 activity when cells were treated by CK2 inhibitor. There was no significant difference of cell viabilities after treatment with CX4945 at 10 μM for 24 hours (Supplemental Fig. S2A). In cells stably expressing exogenous FLAG-PD-L1 which is not driven by endogenous PD-L1 promoter for transcription, CK2 inhibitor also decreased PD-L1 level (Fig. 1B), suggesting that CK2 upregulates PD-L1 protein expression via a non-transcriptional regulatory such as posttranslational modification. As expected, knocking down endogenous CK2 by shRNA reduced both cell surface (Fig. 1C) and overall (Fig. 1D) PD-L1 protein levels. Consistently, ectopic expression of Myc-tagged CK2 increased PD-L1 protein levels (Fig. 1E). These results implied that CK2 upregulates PD-L1 protein expression.

Figure 1. CK2 positively regulates PD-L1 expression.

(A) Various cancer cell lines were treated with 10 μM CX4945 for 24 hours. Cell lysates were subjected to Western blot with the indicated antibodies. Data are from three repeated experiments. (B) Human and mouse cell lines stably expressing ectopic PD-L1 were treated 10 μM CX4945 for 24 hours. Cell lysates were subjected to Western blot with the indicated antibodies. Data are from three repeated experiments. (C) Relative PD-L1 cell surface expression in control or CK2α-knockdown BT549-PD-L1 cells. Data are mean ±s.d., N = 3 replicates, **P<0.01. (D) Western blot analysis of CK2 and PD-L1 expression in control or CK2α-knockdown cells. Data are from three repeated experiments. (E) Western blot analysis of PD-L1 expression in control and cells ectopically expressing Myc-tagged CK2. Data are from three repeated experiments.

CK2 phosphorylates and stabilizes PD-L1

Since inhibition of CK2 decreases PD-L1, we next asked whether CK2, a constitutively active serine/threonine kinase, may phosphorylate PD-L1 and increase its stability. We first determined whether CK2 physically interacts with PD-L1 by reciprocal co-immunoprecipitation. The data revealed an association between FLAG-tagged PD-L1 and Myc-tagged CK2 (Fig. 2A and 2B) and endogenous association between PD-L1 and CK2 (Fig. 2C). We then carried out in vitro kinase assay using active CK2 kinase and GST and GST-PD-L1 as substrates and showed that CK2 phosphorylated PD-L1 but not GST (Fig. 2D). We further performed in vitro kinase analysis to identify CK2 phosphorylation sites on PD-L1. By mass spectrometry and CK2 consensus phosphorylation site prediction, both Thr 285 and Thr 290 of PD-L1 were identified as potential phosphorylation sites. Then we validated these two sites by mutagenesis. In vitro kinase assay showed that both Thr 285 and Thr 290 of PD-L1 were phosphorylated by CK2 as mutation of these two threonine residues to alanine completely abolished PD-L1 phosphorylation signals by CK2 (Fig. 2E). These results suggested that PD-L1 Thr 285 and Thr 290 are CK2 phosphorylation sites.

Figure 2. CK2 phosphorylates and stabilizes PD-L1.

(A) Lysates from PD-L1 stable cells were immunoprecipitated (IP) by FLAG antibody and subjected to Western blot using CK2 or FLAG antibody. Data are mean ±s.d., N = 3 replicates, *P<0.05. (B) Lysates from stable cells co-expressing PD-L1 and CK2 were immunoprecipitated using anti-Myc antibody and subjected to Western blot using Myc or FLAG antibody. Data are from three repeated experiments. (C) Lysates from H460 cells were immunoprecipitated by CK2 antibody and subjected to Western blot using PD-L1 or CK2 antibody. Data are from three repeated experiments. (D) In vitro kinase assay. GST or GST–PD-L1 was incubated with CK2 and [γ−³²P] ATP. Kinase reactions were subjected to SDS-PAGE and stained with Coomassie blue. Data are from three repeated experiments. (E) In vitro kinase assay. Wild-type GST–PD-L1 (WT), GST–PDL1^T285A, GST–PDL1^T290A, or GST–PDL1^T285/T290AA was incubated with CK2. Kinase reactions were subjected to SDS-PAGE. Phosphorylation of PD-L1 was visualized by autoradiography. Protein loading was assessed by Western blot by detection of PD-L1. Data are from three repeated experiments. (F) Western blot analysis of lysates from stable BT549 transfectants established by transfection of cells with equal amount of plasmids encoding FLAG–PD-L1-WT (PD-L1^WT), FLAG-PD-L1-T285A/T290A (PD-L1^2A) or FLAG-PD-L1-T285E/T290E (PD-L1^2E). Equal amount of plasmids encoding GFP was co-transfected with PD-L1 plasmids to ensure equal amount of plasmid transfection. Data are mean ±s.d., N = 3 replicates, * P<0.05. (G) BT549 cells stably expressing PD-L1^WT and PD-L1^2A were treated with 10 μM CX4945 for 24 hours and subjected to Western blot analysis with the indicated antibodies. Data are from three repeated experiments. (H) BT549 cells stably expressing PD-L1^WT and PD-L1^2A treated with 20 μM cycloheximide for different time intervals were analysed by Western blotting. Data are from three repeated experiments. (I) Densitometry results for PD-L1^WT and PD-L1^2A after cycloheximide treatment in (H) were plotted, and the half-lives of PD-L1 were determined. (J) H460 cells were treated with 10 μM MG132, with or without 10 μM CX4945 for 24 hours and subjected to Western blot analysis with the indicated antibodies. Data are from three repeated experiments.

Next, we asked whether CK2 stabilizes PD-L1 by phosphorylation. To this end, we generated cell lines that stably express ectopic wild type PD-L1 (PD-L1^WT), phosphorylation-deficient T285A/T290A double mutant (PD-L1^2A) and phosphorylation-mimic T285E/T290E double mutant (PD-L1^2E). We found that PD-L1 expression levels were higher in cells expressing phosphorylation-mimic PD-L1^2E than in those expressing PD-L1^WT (Fig. 2F; lane 3 vs. lane 1). Likewise, PD-L1 expression levels were lower in phosphorylation-deficient PD-L1^2A than in cells expressing PD-L1^WT (Fig. 2F; lane 2 vs. lane 1). In addition, inhibition of CK2 with CX4945 decreased PD-L1 levels in cells expressing PD-L1^WT but not in those expressing PD-L1^2A (Fig. 2G). To test that phosphorylation of PD-L1 by CK2 indeed stabilize PD-L1, we measured the half-lives of PD-L1^WT and PD-L1^2A. Cells expressing PD-L1^WT and PD-L1^2A were treated with cycloheximide (CHX) to inhibit protein synthesis, followed by the measurement of remaining PD-L1. As shown in Fig. 2H and 2I, PD-L1^WT had a half-life of about16 hours, whereas PD-L1^2A had a half-life of about 6 hours. These results suggested that phosphorylation of PD-L1 at Thr 285 and Thr 290 by CK2 stabilizes PD-L1.

To further validate that CK2 indeed phosphorylates PD-L1 in vivo, we generated an antibody that specifically recognizes PD-L1 phosphorylation at Thr 290 and did not recognize non-phosphorylated PD-L1 peptide (Supplementary Fig. S3A). Treatment with CK2 inhibitor decreased phospho-T290 signal, suggesting that the phospho-T290 antibody is specific and CK2 is responsible for the phosphorylation of PD-L1 (Fig. 2J).

CK2 regulates PD-L1 function through SPOP

To understand how CK2 regulates PD-L1 expression, we treated cells with proteasome inhibitor MG132. The data showed that MG132 restored CX4945-mediated PD-L1 downregulation (Fig. 3A), suggesting that CK2 inhibitor downregulates PD-L1 through proteasome degradation. In addition, knockdown of CK2 by shRNA enhanced PD-L1 ubiquitination (Fig. 3B), suggesting that CK2 may protect PD-L1 from ubiquitination. Interestingly, SPOP has been reported as the adaptor protein for the binding of CUL3-based E3 ubiquitin ligases to PD-L1 and responsible for PD-L1 degradation (22). Moreover, the region of PD-L1 spanning amino acids 283–290 has been proposed as the binding motif for SPOP (22). Given that PD-L1 phosphorylation sites, Thr 285 and Thr 290, are located in this motif, we speculated that their phosphorylation by CK2 may disrupt the binding of PD-L1 to SPOP, thereby protecting PD-L1 from ubiquitination and degradation. To test this hypothesis, we first examined the binding of PD-L1 to SPOP in the absence of CK2. Indeed, knocking down CK2 increased the binding of PD-L1 to SPOP (Fig. 3C) and PD-L1 ubiquitination (Fig. 3B). Next, we compared the binding of PD-L1 to SPOP in PD-L1^WT, PD-L1^2A and PD-L1^2E stable cell lines. We showed that phosphorylation-deficient mutant PD-L1^2A and phosphorylation-mimic mutant PD-L1^2E exhibited stronger and weaker binding to SPOP, respectively, compared with PD-L1^WT (Fig. 3D). Taken together, these results indicated that phosphorylation of PD-L1 at Thr 285 and Thr 290 by CK2 disrupts its binding to SPOP and protects it from ubiquitination and degradation.

Figure 3. CK2 regulates PD-L1 function through SPOP.

(A) Western blot analysis of BT549-PD-L1 cells treated with CX4945 (10 μM, 24 hour) with or without MG132 (10 μM). Data are mean ±s.d., N = 3 replicates, ** P<0.01, n.s., not significant. (B and C) Lysates from BT549-PD-L1 stable cells expressing control or CK2 shRNA treated with MG132 (10 μM) were immunoprecipitated using FLAG antibody and subjected to Western blot analysis with K48-ubiquitin (B) or SPOP (C) antibody. Data are from three repeated experiments. (D) Lysates from stable cells expressing PD-L1^WT, PD-L1^2A or PD-L1^2E were immunoprecipitated using FLAG antibody and subjected to Western blot analysis with SPOP antibody. Data are from three repeated experiments. (E) Representative images of CK2, phospho-T290-PD-L1 and PD-L1 immunohistochemistry (IHC) staining of human breast tumor specimens. (F) Kaplan-Meier survival curves with low or high phospho-T290-PD-L1 antibody IHC staining in patients with breast cancer. P = 0.038, Log-rank test. (G) Human primary T cells isolated from PBMC were stimulated with CD3/CD28 and co-cultured with BT549-PD-L1 with or without CK2 knockdown. IL2 and IFNγ production was assayed by ELISA. Live T cells were collected and counted. Data are mean ±s.d., two-tailed t-test (N = 3 replicates). (H) Mouse primary T cells from C57BL/6 mice were stimulated with CD3/CD28 and co-cultured with B16-PD-L1 with or without CK2 knockdown. IL2 and IFNγ production was assayed by ELISA. Live T cells were collected and counted. Data are mean ± s.d., two-tailed t-test (N = 3 replicates). (I) B16 cells stably expressing ovalbumin (OVA) with or without CK2 knockdown were co-cultured with T cells collected from OT-I TCR transgene mice for 3 days. Live cells were stained with crystal violet. The crystal violet dye was solubilized and measured by absorbance at 570 nm. Data are mean ±s.d., two-tailed t-test (N = 3 replicates).

To establish clinical relevance of our findings, we analyzed the correlation between CK2 and PD-L1 expression in human breast tumor specimens by IHC staining. The results indicated a positive correlation between CK2 and PD-L1 expression (P = 0.013) (Table 1 and Fig. 3E). In addition, using specific phospho-T290-PD-L1 monoclonal antibody we generated, we showed that CK2 was positively correlated with phospho-T290-PD-L1 (P = 0.025) (Table 2 and Fig. 3E). Moreover, high phospho-T290-PD-L1 was associated with poor overall survival in patients with breast cancer (P = 0.038) (Fig. 3F). These results suggested that the identified pathway is pathologically relevant.

Table 1.

CK2 positively correlates with PD-L1 staining in human breast cancer tissues

			PD-L1		Total
			Low	High
		Low	28	10	38
	CK2	High	30	34	64
Total			58	44	102
			P=0.013

Table 2.

CK2 positively correlates with p-T290-PD-L1 staining in breast cancer tissues

			p-T290-PD-L1		Total
			Low	High
		Low	18	20	38
	CK2	High	17	49	66
Total			35	69	104
			P=0.025

Cytokine production is an important indicator of T-cell function. To test whether regulation of PD-L1 by CK2 affects T-cell function, we co-cultured human primary T cells with BT549-PD-L1 cells and assessed their IL-2 and IFNγ production by ELISA. After stimulation of primary T cells by CD3/CD28, IL-2 and IFNγ production significantly increased (Fig. 3G). Primary T cells co-cultured with BT549-PD-L1 cells had lower levels of IL-2 and IFNγ which was restored when we knocked down CK2 (Fig. 3G). Similar results were observed in mouse primary T cells co-cultured with mouse B16-PD-L1 cells (Fig. 3H). The proliferation of T cells after CD3/CD28 stimulation, co-culture with tumor cell and CK2 knockdown showed similar pattern with IL-2 and IFNγ production (Fig.3 G and H). These results suggested that CK2 stabilizes PD-L1 and inhibits cytokine production by activated T cells.

To study the function of CK2-mediated PD-L1 phosphorylation in an antigen-specific T cell killing context, we used the well-established OT-I TCR transgene C57BL/6-Tg (TcraTcrb) mice, whose CD8+ T cells specifically and almost exclusively express an ovalbumin (OVA)-specific TCR (23). To this end, B16 cells stably expressing the OVA antigen were co-cultured with OT-I T cells. We found that after co-culture, T cells killed B16-OVA cells, and the killing effect was enhanced when CK2 was knocked down in B16-OVA cells (Fig. 3I). These results suggested that CK2 inhibits antigen-specific T-cell function through phosphorylation of PD-L1.

CK2 inhibitor suppress tumor growth by activating dendritic cell function

Because inhibition of CK2 downregulated PD-L1 protein expression, we asked whether CK2 inhibitor (CK2i), through downregulating PD-L1 level, may inhibit tumor growth in B16F10 mouse tumor model. Indeed, treatment with higher dose of CK2i (75mg/kg) inhibited tumor growth (P<0.01) (Fig. 4A). To learn in detail the immune cell populations affected by CK2i, we performed mass cytometric (CyTOF) analysis by collecting fresh tumor tissues treated with vehicle and CK2i. Among immune cell populations, CD11C+MHC II+CD80+ F4/80- dendritic cells (DCs) are the most elevated cell population by CK2i treatment (Fig. 4B and 4C). Recent findings suggest that DCs can be an important target of PD-L1 blocking antibody (24). PD-L1 is expressed much more abundantly than CD80 on tumor-associated DCs. And interaction between PD-L1 and CD80 shields CD80 of DCs to interact with CD28 of T cells, which is required to prime T cell activity, resulting in inactivation of T cells. Blocking PD-L1 on DCs relieves CD80, allowing the CD80/CD28 interaction to enhance T cell priming (24). Our CyTOF results suggest that CK2i might relieve CD80 and reinvigorates DCs function through downregulation of PD-L1 on DCs.

Figure 4. CK2 inhibitor suppress tumor growth by activating dendritic cell function.

(A) C57BL/6 mice inoculated with B16F10 cells were treated with vehicle or CK2 inhibitor by oral gavage twice a day until the tumor size reached limit. Data are mean ±s.d., N = 3 mice per group, **P<0.01. (B) CyTOF analysis of B16F10 tumors in mice after drug intervention using the Cytobank software platform. viSNE map of dendritic cells (DCs) population in B16F10 tumors was shown. Cells are color coded in the map based on the expression intensity of the indicated markers. (C)The percentage of dendritic cells was determined using CD11C+MHC II+CD80+ F4/80- as markers. Data are mean ±s.d., N = 3 mice per group. (D) BALB/c mice inoculated with EMT6 cells were treated with CK2 inhibitor at 75mg/kg by oral gavage twice a day until the tumor size reached limit. Tumor size was measured every two days. Data are mean ±s.d., N = 4 mice per group, ***P<0.001. (E) Nude mice inoculated with EMT6 cells were treated with CK2 inhibitor at 75mg/kg by oral gavage twice a day until the tumor size reached limit. Tumor size was measured every two days. Data are mean ±s.d., N = 4 mice per group, n.s.: not significant. (F) CyTOF analysis of EMT6 tumors in mice from (D) using the Cytobank software platform. viSNE map of DCs population in EMT6 tumors was shown. (G) The percentage of DCs was determined using CD11C+MHC II+CD80+ F4/80- as markers. Data are mean ±s.d., N = 3 mice per group. (H) Human dendritic cells differentiated from THP-1 monocytes were treated with 7.5 μM and 10 μM CX4945 for 24 hours. and subjected to Western blot analysis with the indicated antibodies. Data are from three repeated experiments. (I) Mouse dendritic cell line, DC2.4, was treated with 10 μM CX4945 for 24 hours and subjected to Western blot analysis with the indicated antibodies. Data are from three repeated experiments. (J) Mouse dendritic cells from C57BL/6 mouse, after treatment with or without 10 μM CX4945 for 24 hours, in the presence of ovalbumin (OVA), were co-cultured with T cells from OT-I mice and B16-PD-L1 cells stably expressing OVA for 3 days. Live tumor cells were stained with crystal violet. The crystal violet dye was solubilized and measured by absorbance at 570 nm. Data are mean ±s.d., two-tailed t-test (N = 3 replicates). (K) A proposed model of T cell inhibition by CK2-mediated PD-L1 phosphorylation in dendritic cells. Left: CK2 phosphorylates PD-L1, disrupts its binding to SPOP-CUL3 E3 ligase, preventing its ubiquitination and degradation. In addition to binding to PD-1 on the surface of T cells to inhibit T cell, stabilized PD-L1 binds to CD80 in dendritic cells, blocking CD80 interaction with CD28 on the cell surface of T cells. Without CD28 coactivation, T cell cannot be fully activated. Right: CK2 inhibitor, CX4945, inhibits PD-L1 phosphorylation, promoting its ubiquitination and degradation by SPOP-CUL3. Relived CD80 from PD-L1 inhibition interacts with CD28 and fully activates T cells.

To investigate whether this is a general finding in tumor rather than tumor type specific, we used another mouse tumor model, EMT6 mouse mammary tumor model. And to confirm the efficacy of CK2 inhibitor is mediated by T cells, we compared the antitumor effects of CK2i between immunocompetent BALB/c mice and nude mice which lack T cell. The results showed that treatment with CK2i significantly inhibited EMT6 tumor growth in BALB/c mice (P<0.001) (Fig. 4D), but not in nude mice (P>0.05) (Fig. 4E), suggesting that the antitumor effect of CK2i is mediated by T cells. Next, we performed CyTOF analysis by collecting fresh tumor tissues, spleen and draining lymph node from the BALB/c mice treated with vehicle and CK2i. As shown in Fig. 4F and 4G, in the tumor tissues, CD11C+MHC II+CD80+ F4/80- dendritic cells (DCs) are the most elevated cell population by CK2i treatment. The data is consistent with B16F10 mouse tumor data, suggesting CK2i exerts its anti-tumor effect through the expansion of dendritic cell population. There is no change of CD11C+MHC II+CD80+ F4/80- dendritic cells populations in the draining lymph node (Supplementary Fig. S4A) and spleen (Supplementary Fig. S4B) after CK2i treatment, suggesting that CK2i mainly affects tumor-associated dendritic cells.

To answer the question whether CK2 might use the same mechanism to upregulate PD-L1 expression in DCs as in tumor cells, we treated human DCs differentiated from THP-1 monocytes with CK2 inhibitor and found that CK2 inhibitor indeed inhibited PD-L1 phosphorylation and downregulated PD-L1 expression in DCs (Fig. 4H and Supplementary Fig. S4C). Similar results were also found in mouse DCs (Fig. 4I). There was no change of another PD-1 ligand, PD-L2, after CK2 inhibitor treatment (Fig. 4H and 4I). Next, we asked whether CK2 inhibitor may enhance DCs function and thus more efficient at priming T cells. Co-culture of DCs from C57BL/6 mice with OT-I T cells enhanced killing of B16-OVA tumor cells (P<0.01) (Fig. 4J), and CK2 inhibitor treatment of DCs before co-culture further enhanced OT-I T cells killing effect (P<0.001) (Fig. 4J). These results suggest that CK2 inhibitor may enhance DCs function through downregulation of PD-L1 on DCs (Fig. 4K, working model).

CK2 inhibitor synergizes with Tim-3 blockade therapy to suppress tumor growth

Next we asked whether combination with immune checkpoint inhibitors might enhance CK2i effect on DCs. Tim-3 has been shown to be highly expressed in tumor-associated DCs and suppresses immune responses (25). We explored the combination therapy of CK2i and antibody against Tim-3. When we treated mice bearing 4T1 mammary tumors with CK2i alone or in combination with Tim-3 antibody, the combined treatment induced better tumor growth inhibition (Fig. 5A and Supplementary Fig. S5A) and led to longer survival (Fig. 5B) than either agent alone. We also evaluated the toxicities of single-drug and combination treatment and showed that indicators of liver (AST and ALT) and kidney (BUN and creatinine) functions were within the normal range and there was no difference of mice body weight in different treatment groups. (Supplementary Fig. S5B). These findings indicated that CK2i and anti-Tim-3 combination therapy showed efficacy in inhibiting tumor growth and well tolerated in animal experiments.

Figure 5. CK2 inhibitor synergizes with Tim-3 blockade therapy to suppress tumor growth.

(A) BALB/c mice inoculated with 4T1 cells were treated with CK2 inhibitor at 75mg/kg by oral gavage twice a day, Tim-3 antibody (100 μg) injected intraperitoneally on days 6, 11, and 16, or their combination. Tumor size was measured every two days. Data are mean ±s.d., N = 6 mice per group. (B) Survival analysis of mice from panel (A). Data are mean ±s.d., N = 6 mice per group. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, Student’s t-test and Log-rank test. (C) CyTOF analysis of 4T1 tumors in mice after drug intervention. viSNE map was shown. (D) The percentage of DCs was determined using CD11C+MHC II+CD80+ F4/80- as markers. Data are mean ±s.d., N = 3 mice per group. (E) The percentage of granzyme B positive CD8 T cells was determined using CD8+Granzyme B+ as markers. Data are mean ±s.d., N = 3 mice per group. (F) CyTOF analysis of memory T cell population in re-challenged tumors. viSNE map of memory T cell populations was shown. (G, H) The percentage of memory T cell was determined using CD8+CD44+ and CD4+CD44+ as markers. Data are mean ±s.d., N = 2 mice in control group, N = 3 mice in treatment group.

To better understand how the combination therapy of CK2i and anti-Tim-3 affects the tumor microenvironment, we performed CyTOF using samples collected from single agent and combined groups by a panel of antibodies (Supplementary Table S2). CyTOF analysis showed that the percentage of CD11C+MHC II+CD80+F4/80-DCs was increased in the combination treatment group (Fig. 5C and 5D). In addition, CD8+Granzyme B+ T cell population was increased in the combination group (Fig. 5C and 5E), suggesting enhanced cytotoxic CD8 T cell function. These results suggested that the combination of CK2i and anti-Tim-3 activated DCs in tumor microenvironment possibly via downregulation of PD-L1 and inhibition of Tim-3 in DCs. Recent findings suggest that PD-L1 on host cells such as DCs and macrophages is essential for PD-L1 blockade-mediated tumor regression (26–27), our results also support such conclusion.

We observed tumor regression in one mouse (1/6) in the CK2i group and two mice (2/6) in the combination group (Fig. 5B). Thus, we re-challenged those three mice with tumor regression as well as in two normal mice with 4T1 tumor cell injection. The tumors in normal mice were palpable on day 5 whereas tumors in re-challenged mice were not (Supplementary Fig. S5C), suggesting boosted memory T-cell function in the treatment group. We collected the spleen from mice, and found that the percentage of CD8+CD44+ and CD4+CD44+ memory T cells indeed increased in the treatment group (Fig. 5F-H). These results suggested that CK2 inhibitor suppresses tumor growth and synergizes with anti-Tim-3 therapy, and their combination likely induces antitumor effects via activation of DC cells and prolongs mouse survival by boosting memory T cells.

Discussion

Expression of PD-L1 on cancer cells and immune cells is correlated with a more durable objective response rate to PD-1/PD-L1 antibodies, which highlights the importance of deeply understanding how PD-L1 is regulated (28). Posttranslational modifications including phosphorylation play important roles in the regulation of PD-L1 protein stability and protein-protein interactions. Both serine/threonine and tyrosine phosphorylation of PD-L1 have been reported (4, 13, 29–30). Interestingly, Thr 290 of PD-L1 has been reported to be phosphorylated (31), but its function in PD-L1 regulation is not known. In this study, we demonstrated that Thr 290 of PD-L1 was phosphorylated by CK2. Phosphorylation of PD-L1 at both Thr 290 and Thr 285 by CK2 disrupts PD-L1 binding to SPOP which in turn stabilizes PD-L1 by protecting PD-L1 from ubiquitination and degradation. CK2 inhibitor can downregulate PD-L1 in both tumor cells and DCs, thus activates T cells not only through the traditional mechanism, namely disruption of PD-L1 and PD-1 interaction between tumor and T cells, but also releasing CD80 from DCs to prime T cells by interacting with CD28 of T cells. Recent findings demonstrate that PD-L1 expressed in host cells such as DCs and macrophages plays an essential role in checkpoint blockade therapy and DCs can be an important target of PD-L1 blocking antibody (24, 26–27). Our findings suggest a potential anti-tumor approach by activating DC function by blocking the CK2-PD-L1 pathway and inhibiting Tim-3.

Triple-negative breast cancer (TNBC) is a highly aggressive subtype of breast cancer that initially responds to chemotherapy but eventually develops resistance. This presents a major clinical challenge as there are currently few effective options for the treatment of TNBC. We used TNBC as a model system to study the role of CK2 in immune system. The result suggested that our findings are not restricted to TNBC but also applied to general tumor types, including lung cancer, prostate cancer and melanoma.

The effect of CK2 inhibitor on modulating myeloid cells in the tumor microenvironment has been investigated (12). Hashimoto et al found that potent and selective CK2 inhibitors, BMS-211 and BMS-595, significantly reduced the amount of polymorphonuclear myeloid-derived suppressor cells (PMN-MDSC) in the spleen and tumor-associated macrophages (TAM) in the tumor (12). These findings, together with our findings of CK2 inhibitor’s effect on tumor and dendritic cells, suggest that CK2 inhibitor may exert its anti-tumor effect through multiple ways. Interestingly, Hashimoto et al showed that dendritic cell (DC) was decreased in the tumor (12). That is different to our results. The different markers used for distinguishing DC might explain the difference of the two study. In Hashimoto’s paper, the authors used CD11C+MHC II+ as markers for DC in flow cytometry analysis, whereas in our study we used CD11C+MHC II+CD80+ F4/80-as markers for DC in CyTOF analysis. Since in tumors, macrophages can also express CD11C and MHC II (32), some CD11C+MHC II+ cell population might belong to macrophages, and tumor-associated macrophages were significantly decreased in the study (12), the decrease of CD11C+MHC II+ population in the tumor may be partially due to decreased macrophage population. Another explanation for the difference is that the CK2 inhibitors used are different. We used CX4945, whereas in Hashimoto’s paper the CK2 inhibitors are BMS-211 and BMS-595. Although all of these are potent and selective CK2 inhibitors, their biological profile might be different.

Significance:

This work identifies a role for CK2 in immunosuppression by phosphorylation and stabilization of PD-L1, identifying CK2 inhibition as an immunotherapeutic approach for treating cancer.