USP22 Deubiquitinates CD274 to Suppress Anticancer Immunity

CD274, also commonly referred as to programmed cell death 1 ligand 1 (PD-L1) and B7 homolog 1 (B7-H1), inhibits T-cell activation and function by directly binding to the immune inhibitory receptor CD279 (PD-1; refs. 1, 2). In addition to being involved in the induction and maintenance of normal immune tolerance, CD274 is often found on tumor cells and acts as a checkpoint for anticancer immunity (3). A number of antibody-based drugs have been developed to block CD274–CD279 interaction, with the intention of recovering the effector response of cytotoxic T cells to tumors (4, 5). However, except for several specific lymphomas and melanomas, the overall response rate for targeting CD279/CD274 remains generally poor, necessitating a further understanding of the mechanisms of resistance to immune-checkpoint inhibition (4–7).

Posttranslational modification (PTM) of CD274 is hypothesized to be a mechanism of resistance to immune-checkpoint blockade. CD274 is regulated by several posttranslational modifications, such as ubiquitination, phosphorylation, glycosylation, and palmitoylation (8–10). CD274 PTM-targeted strategies have achieved important advances in activating antitumor immunity, especially by targeting glycosylated CD274 (11, 12). In this study, we focused on the control of CD274 deubiquitination, which is supposed to be a protective mechanism to maintain its protein stability and its inhibitory functions (13). We identified that USP22 is a novel deubiquitinase (DUB) of CD274 and highlighted the significance of targeting USP22 in liver cancer therapy, especially in rescuing immune inactivation.

Antibodies and proteins were obtained from the indicated sources: anti-USP22 (ab195289, Abcam, for immunoblotting; abs105510, Absin, for IHC), anti-OTUD5 (20087, Cell Signaling Technology), anti-BAP1 (10398-1-AP, Proteintech), anti-VCPIP1 (17802-1-AP, Proteintech), anti-USP9X (14898, Cell Signaling Technology), anti-USP1 (14346-1-AP, Proteintech), anti-CD274 (SAB4301882, Sigma-Aldrich, for immunoblotting; ab205921, Abcam, for IHC; BE0101, Bio X Cell, for therapy), anti-Flag (F1804, Sigma-Aldrich), anti-HA (sc-7392, Santa Cruz), anti-CDKN2A/p16INK4a (80772, Cell Signaling Technology), anti-FZR1 (sc-56312), anti-beta Actin (66009-1-Ig, Proteintech), anti-Rabbit IgG (HRP; GTX221666-01, GeneTex), anti-Mouse IgG (HRP; GTX221667-01, GeneTex), anti-Rabbit IgG (HRP; GTX221666-01, GeneTex), anti-CD3 (100236, APC conjugated, BioLegend), anti-CD8 (100708, PE conjugated, BioLegend), anti-GZMB (515403, FITC conjugated, BioLegend), CD274 GST-fusion protein (25–241 aa; Ag12432, Proteintech), and CD274 His-fusion protein (181–290 aa; Ag27611, Proteintech). Chemicals and other reagents were acquired from the designated suppliers as follows: PR-619 (S7130, Selleckchem), cisplatin (2251, TOCRIS), nocodazole (S2775, Selleckchem), palbociclib (S1116, Selleckchem), MG-132 (S2619, Selleckchem), cycloheximide (0970, TOCRIS), protease inhibitor cocktail (B14001, Bimake), and Lipofectamine 3000 transfection reagent (L3000008, Invitrogen).

To generate USP22 knockout (KO), CD274 KO, and USP22/CD274 double KO (DKO) cell lines, cells were transfected or cotransfected with USP22 Double Nickase Plasmid (human; sc-403660-NIC, Santa Cruz), USP22 Double Nickase Plasmid (mouse; sc-432127-NIC, Santa Cruz), CD274 Double Nickase Plasmid (human; sc-401140-NIC, Santa Cruz), or USP22 Double Nickase Plasmid (human) together with CD274 Double Nickase Plasmid (human), respectively, according to the manufacturer's instructions. In brief, 1.5 × 10⁵ HepG2 or H22 cells were individually seeded in 3 mL antibiotic-free standard growth medium per well in a six-well cell culture plate 12 hours prior to transfection. When 30% confluent, cells were transfected with the indicated plasmids by UltraCruz Transfection Reagent (sc-395739, Santa Cruz). Seventy-two hours after transfection, 1 μg/mL puromycin (ant-pr-1, InvivoGen) was added into growth medium for at least 10 days of selection. After selection, cells were suspended, diluted, and re-seeded to ensure single clone formation. KO efficiency of each single clone was evaluated by Western blot analysis. All the KO and DKO cell lines were maintained in complete medium as described above at 37°C under 5% CO₂.

For siRNA-mediated knockdown (KD), 5 × 10³ HepG2 cells were seeded in 3 mL of antibiotic-free standard growth medium per well in a six-well tissue culture plate overnight and then individually transfected twice (on days 1 and 2) with 25 nmol/L indicated siRNAs (described below) by Lipofectamine RNAiMAX transfection reagent (13778030, Invitrogen) according to the manufacturer's instructions. Forty-eight hours after transfection, cells were treated as indicated and subjected to further experimentation. CDKN2A/p16INK4a-specific siRNA Oligo Duplex (SR300743) and FZR1-specific siRNA Oligo Duplex (SR324108) were obtained from OriGene Technologies. Nontargeting scramble control RNA (siCtrl; 5′-UUCUCCGAACGUGUCACGUTT-3′) was synthesized by GenePharma. The KD efficiency was verified by Western blot analysis with specific antibodies.

For plasmids, full-length expression cDNA of CD274 (Flag-CD274) was purchased from Sino Biological Inc. (HG10084-NF). Flag-USP22 [wild type (WT) full length], Flag-USP22-C185A (USP22 enzymatically inactive mutant), and Flag-USP22-UCH (USP22 containing only the UCH domain) were gifts from Dr. Jun Cui (School of Life Sciences, Sun Yat-sen University, Guangzhou, Guangdong, China; ref. 14). All constructs were confirmed by DNA sequencing.

According to previously established protocols (15), after two washes with ice-cold PBS, 5 × 10⁵ cells were resuspended in lysis buffer (25 mmol/L HEPES pH 7.5, 150 mmol/L NaCl, 0.25% Triton X-100, 0.25% NP-40, 0.25% CHAPS, 10% glycerol, and 1 × protease inhibitor cocktail) on ice for 1.5 hours, and then centrifuged at 13,000 g for 15 minutes. The supernatants were precleared with 40 μL protein A/G-coupled agarose (sc-2003, Santa Cruz) for 1.5 hours and then incubated with 2 μg of the indicated antibodies, isotype control IgG (co-IgG)–conjugated beads, or 20 μL anti-Flag affinity gel (B23101, Bimake) for another 4 hours at 4°C. After three washes with lysis buffer, immunoprecipitates were boiled in 1 × loading buffer for Western blot analysis.

Protein samples of immunoprecipitation or cell lysates were further resolved by SDS-PAGE with 4% to 20% 15-well SurePAGE Gel (M00657, GenScript) and then transferred onto nitrocellulose membrane (66485, PALL) following standard procedures. Membranes were blocked with 10% skim milk in TBST for 1.5 hours and subsequently incubated with indicated primary antibodies overnight at 4°C according to the manufacturer's recommendation. After three washes with TBST, membranes were incubated with appropriate horseradish peroxidase (HRP)–labeled secondary antibodies. SuperSignal West Femto Maximum Sensitivity Substrate (34095, Thermo Fisher) was used to detect the HRP. Visualized images were obtained using photographic film. The similar settings for exposure time, brightness, contrast, and scanning condition were applied to digital images by using the HP OfficeJet Scanner (Pro 8730, Hewlett-Packard Development Company).

5 × 10⁴ WT and USP22 KO HepG2 cells, or WT HepG2 cells expressing Flag or Flag-USP22 were individually treated with 20 μg/mL cycloheximide (CHX) or vehicle control (DMSO). Cells were collected at the indicated time points and subsequently subjected to Western blot analysis. The half-life of CD274 protein was determined by abundance quantification using ImageJ.

According to previously established protocols (16), to analyze cellular proliferation, 1 × 10⁴ of the indicated cells were seeded in six-well tissue culture plates at day 0 (in triplicate) in 3 mL of normal growth medium. The medium was changed every day. Cell number at the indicated time points was counted after Trypsin digestion using a haemocytometer and recorded for analysis.

For cell viability assay, 2.5 × 10³ of each of the indicated cells were seeded in 96-well microplates in triplicates in 100 μL of normal growth medium per well. After 48 hours, cells were further incubated with 0.25 mg/mL WST-8 solution at 37°C for 2 hours, followed by the addition of 10 μL of 3% SDS to end the reaction. The absorbance was measured at 450 nm by using the iMark Microplate Reader (168-1130, Bio-Rad), and the percentage of viable cells was calculated and averaged for each well.

For clonality analysis, 500 cells were seeded in six-well plates at day 0 (in triplicate) in 3 mL of regular growth media supplemented with 20% fetal bovin serum (FBS) per well. Cells were maintained at 37°C under 5% CO₂ for 2 weeks, and growth medium was replaced every 2 days. The remaining cells were fixed with 4% cold paraformaldehyde (PFA) for 45 minutes and then stained with 0.1% (w/v) crystal violet (SSI1047-1, SunShineBio), dissolved in 10% methanol, for 2 hours at room temperature (RT). After extensive washing with distilled water, pictures of each replicate were photographed using a digital camera (M10, Leica), and colony numbers were quantified under a microscope (CKX53, Olympus).

To analyze the expression level of CD274, as well as the absolute number and function of tumor-infiltrating lymphocytes (TIL), dissected tumor tissues were cut into small pieces, and further digested with 5 mL mixture of Collagenase/Hyaluronidase (STEMCELL Technologies) and DNase I (STEMCELL Technologies) in Dulbecco's modified Eagle's medium (DMEM) for 1 hour at 37°C to generate single-cell suspension. TILs were isolated and enriched by density-gradient centrifugation using Ficoll solution (Sigma-Aldrich), suspended in DMEM and filtered through a 70-μm strainer, and then lysed with red blood cell lysis buffer for 5 minutes at RT. The purified TILs were fixed for 20 minutes at RT and then incubated with CD16/CD32 antibody (eBioscience) in PBS with 5% BSA for 15 minutes to block FcγR binding (13). After washing, cells were permeabilized and further incubated with fluorescein-conjugated antibodies against CD274 (BioLegend), CD3 (BioLegend), CD8 (BioLegend), and GMZB (BioLegend) in Intracellular Staining Perm Wash Buffer (421002, BioLegend) for 45 minutes at RT. The corresponding isotype IgGs were used for controls. Immunostained cells were washed thrice using PBS with 5% BSA, and subsequently subjected to flow cytometry analysis for calculating absolute and relative quantity of the indicated population by using the BD FACSCanto II Flow Cytometer (BD Biosciences).

Cancer genomics–related data sets in The Cancer Genome Atlas (TCGA, http://cancergenome.nih.gov), International Cancer Genome Consortium (https://icgc.org), Oncomine (https://www.oncomine.org), and Catalogue of Somatic Mutations in Cancer (https://cancer.sanger.ac.uk/cosmic) databases were collected, and used to establish clinical data pools. Cases with unknown or incomplete clinical pathologic information or with lack of prognostic follow-up data were excluded. Three public Web servers, cBioPortal for Cancer Genomics (cBioPortal, http://www.cbioportal.org), Gene-Expression Profiling Interactive Analysis (GEPIA2, http://gepia2.cancer-pku.cn), and Kaplan–Meier plotter (KM-plotter, http://kmplot.com), were individually applied to perform bioinformatics analysis. In brief, cBioPortal was used to visualize and compare gene alteration; GEPIA2 was used to calculate differential expression and correlation of genes; and KM-plotter was used to investigate cancer patient survival.

HEK293T, HepG2, SMMC-7721, Hepa 1-6, and H22 cell lines were individually obtained in the past 2 years from Cell Bank of Chinese Academy of Sciences (Shanghai, China) or KeyGEN BioTECH (Nanjing, China), where Mycoplasma contamination detection and short tandem repeat (STR) profiling were performed for quality and identity guarantee. All the cell lines and derived cells were cultured under standard conditions as specified by suppliers, respectively, in DMEM or RPMI-1640 supplemented with 2 mmol/L L-glutamine, 1% NEAA, 100 units/mL penicillin, 100 mg/mL streptomycin, and 10% FBS and maintained at 37°C incubator with 5% CO₂. No cell lines used in this work were commonly misidentified cell lines according to the database of the International Cell Line Authentication Committee. All the cell lines were freshly thawed from the purchased seed cells, cultured for no more than 2 months, and regularly checked by virtue of their morphologic features to avoid cross-contamination or misuse. The cells were routinely tested for Mycoplasma contamination and independently validated by STR profiling.

For cell-based analysis of USP22 deubiquitinating CD274 in vivo, 1 × 10⁶ WT and USP22 KO HepG2 cells were individually cotransfected with HA-Ub and Flag-USP22 or Flag expression vectors. Seventy-two hours after transfection, cells were treated with 20 μmol/L MG132 for 8 hours to accumulate ubiquitinated proteins before being harvested, lysed with RIPA buffer (50 mmol/L Tris-HCl pH 6.8, 150 mmol/L NaCl, 10 mmol/L NaF, 10 mmol/L DTT, and 10% glycerol and proteinase inhibitors) containing 1% SDS with mild sonication, and boiled at 95°C for 10 minutes. The denatured cell lysates were further diluted with SDS-negative lysis buffer to reduce SDS to 0.2% and subsequently subjected to immunoprecipitation with anti-CD274 antibody at 4°C overnight, followed by Western blot analysis with indicated antibodies. Endogenous ubiquitination levels of CD274 were determined by anti-HA antibody.

For the deubiquitination assay reactions in vitro, HEK293T cells were cotransfected with Flag-CD274 and HA-Ub expression vectors to prepare ubiquitinated CD274 protein as the substrate. After treatment with 20 μmol/L proteasome inhibitor MG132 for 8 hours, ubiquitinated CD274 protein was immunoprecipitated from the cell extracts using Flag affinity beads as described before. To exclude nonspecific ubiquitin-modified species from the CD274 complex, the immunoprecipitates were washed extensively using the ubiquitination wash buffer (50 mmol/L Tris-HCl pH 6.8, 150 mmol/L NaCl, 0.5% Triton X-100, 0.5% NP-40, 0.5% DCA, 1 M Urea, 1 mmol/L NEM, and protease inhibitors) and subsequently eluted by Flag peptides (F4799, Sigma-Aldrich) in BC100 buffer (25 mmol/L Tris-HCl pH 7.8, 100 mmol/L NaCl). The Flag-USP22, CA mutant, and vehicle Flag control proteins were also individually expressed and purified as above. The ubiquitinated CD274 protein was incubated with 200 ng USP22 protein, CA mutant or Flag tag, respectively, in the deubiquitination buffer (50 mmol/L Tris-HCl pH 8.0, 50 mmol/L NaCl, 1 mmol/L EDTA, 10 mmol/L DTT, and 5% glycerol) at 37°C for 2 hours. The resulting reactions were subjected to Western blot analysis, and CD274 ubiquitination was detected with anti-HA antibody.

All animal studies were approved and supervised by the Institutional Animal Care and Use Committee (IACUC) of the Institute of Life Sciences at Southeast University (SEU). All animal experiments strictly adhered to protocols, policies, and ethical guidelines formulated by our IACUC. NOD CRISPR Prkdc Il2r-Gamma triple-immunodeficient mice [NOD-Prkdc^em26Cd52Il2rg^em26Cd22/Nju (NCG), female, 5 weeks old], created by sequential CRISPR-Cas9 editing of the Prkdc and Il2rg loci in the NOD/Nju mouse, were obtained from Nanjing Biomedical Research Institute of Nanjing University, Nanjing, Jiangsu, China (17); C57BL/6 mice (female, 5 weeks old) were obtained from Comparative Medicine Center, Yangzhou University (Yangzhou, China). All mice were housed in specific pathogen-free conditions in the animal facility of the School of Medicine at SEU, maintained with constant temperature and humidity under 12-hour light–dark cycles, and received food and water ad libitum. Mice were allowed to acclimate to the animal facility for 3 weeks prior to experimental use. For humane care, tumors did not exceed the limit of neoplastic lesions (20% of mouse body mass or 20 mm in the longest axis).

For tumorigenesis experiments, 2 × 10⁵ WT and USP22 KO H22 cells were prepared in 100 μL PBS and individually inoculated subcutaneously (s.c.) into the right flank of 8-week-old female immunodeficient NCG mice or immunocompetent C57BL/6 mice. After inoculation, mice were monitored daily and weighed twice weekly. When tumors became visible, tumor growth was routinely recorded at the indicated times by caliper measurement and calculated as tumor surface size (longest dimension × perpendicular dimension). Tumor weight was determined after sacrifice at the indicated day.

To assess immunogenicity, 2 × 10⁵ WT and USP22 KO H22 cells in 100 μL of PBS were injected s.c. into the opposite flanks of 8-week-old female immunocompetent C57BL/6 mice at the same time. Tumor growth was regularly reported as described above, and tumor weight was determined after sacrifice.

Once tumors reached 35 to 45 mm², mice were treated intraperitoneally (i.p.) with 100 μg anti-CD274 antibody (clone: 10F.9G2, #BE0101, Bio X Cell) or an equivalent isotype-same IgG as vehicle control (co-IgG), respectively, every 5 days. Tumor growth and tumor weight were monitored as indicated. Mice were sacrificed when signs of ulceration in tumor were evident or when tumors reached maximum permitted area.

Once tumors reached 35 to 45 mm², mice were individually treated i.p. with 150 μmol/L cisplatin (CDDP) or equivalent vehicle control (PBS) every 5 days. Tumor growth and tumor weight were monitored as indicated. Mice were sacrificed when signs of ulceration in tumor were evident or when tumors reached the maximum permitted area.

Human hepatocellular carcinoma (HCC) cDNA bank and tissue microarray (TMA) were generated in-house using samples acquired from HCC patients who underwent surgery between 2013 and 2017. For the cDNA bank, total RNA was extracted using the TRIzol reagent (Takara) and quantified by NanoDrop 2000 (Thermo Fisher Scientific). Reverse transcription was performed following the manufacturer's protocol of the Prime Script Reagent RT Kit (Takara), and 200 ng cDNA template was dispensed in 384-well plates in triplicate to generate an HCC cDNA bank. The protocol of this study was approved by the ethical committee of our hospital, and written informed consent was obtained from all patients. Quantitative polymerase chain reaction (PCR) was performed using the specific primers (USP22 forward: 5′-CCATTGATCTGATGTACGGAGG-3′, USP22 reverse: 5′-TCCTTGGCGATTATTTCCATGTC-3′; CD274 forward: 5′-GCCGACTACAAGCGAATTAC-3′, CD274 reverse: 5′-TCTCAGTGTGCTGGTCACAT-3′; GAPDH forward: 5′-AACAGCAACTCCCATTCTTC-3′, GAPDH reverse: 5′-TGGTCCAGGGTTTCTTACTC-3′, generated by XY Biotechnology) in a ViiA 7 Real-Time PCR System (Thermo Fisher Scientific) as previously described (18). Relative expression of CD274 and USP22 was calculated using the delta-delta Ct method, and normalized by GAPDH. For TMA, multiple formalin-fixed paraffin-embedded tissues were punctured and reorganized to generate a microarray. IHC and protein expression scoring were performed as previously described (19). In brief, the expression of CD274 and USP22 was semiquantified as negative (score 0), low (score 1), medium (score 2), or high (score 3).

No statistical methods were used to predetermine sample size, and no samples, mice, or data points were excluded from the reported analyses. Samples were not randomized to experimental groups except for the allocations in mice experimets. The investigators were not blinded to allocation during experiments and outcome assessment.

Statistical analyses were performed using Microsoft Excel 2015 (Microsoft Corporation) and GraphPad Prism 5 (GraphPad Software Inc.) software to assess the differences between experimental groups. The variability within each group was quantified from at least three technical or biological replicates, and presented as means ± SD or SEM, respectively. Statistical significance was determined by two-tailed Student t test or one-way ANOVA analysis. The Tukey multiple test was used for comparison between groups, the Kaplan–Meier and log-rank tests were used for the survival analysis, and the Cox regression model was used for the single-factor and multifactor analysis. Pearson correlation with a linear regression model was conducted to evaluate the correlation of CD274 and USP22 expression in human samples. The detailed P value and other statistical indexes, including R, HR, and log-rank P, were individually defined and calculated in each panel as indicated. *, P < 0.05; **, P < 0.01, as compared with the negative, untreated, or scrambled control.

All Western blots were representative of at least three independent experiments. All mouse studies contained 5 individuals per group and independently performed twice, for a total of 10 mice. All other graphs have been reproduced more than five times.

The protein stability of CD274 is under the control of Cullin 3 (CUL3)–Speckle-Type POZ Protein (SPOP), CKLF-Like MARVEL Transmembrane Domain-Containing Protein 4/6 (CMTM4/6) and COP9 Signalosome Subunit 5 (CSN5) through the ubiquitin–proteasome pathway (13, 20–22). To investigate whether CD274 degradation can be reversed by DUBs, we used a broad-range DUB small-molecule inhibitor (PR-619) to treat 293T cells stably expressing Flag-CD274 (23), which reduced the CD274 expression in both dose- and time-dependent manners (Fig. 1A). Based on the reported inhibitory effects of PR-619 on different DUBs (23), we chose several highly sensitive DUBs as candidates to conduct a binding-based miniaturized screening, and found that Flag-CD274 specifically and strongly interacted with USP22 (Fig. 1B). USP22–CD274 interactions occur under physiologic conditions, validated here by endogenous immunoprecipitation in HepG2, SMMC-7721, Hepa 1-6, and H22 cells (Fig. 1C). In vitro pulldown assays with purified recombinant proteins demonstrated that USP22 directly bound to the cytoplasmic tail and transmembrane region of the C terminal of CD274 (Fig. 1D). CD274 has three different functional domains in its C terminal (24); thus we defined the precise sites to which USP22 bound. Coimmunoprecipitation results strongly suggested that USP22 interacted with the last five amino acids (HLEET) of CD274 (Supplementary Fig. S1A), and the potential binding motif (XX)-EE/DD/DE-(X) was evolutionarily conserved in CD274 and other previously identified substrates of USP22 (Supplementary Fig. S1B).

Because USP22 is a well-established ubiquitin-specific protease (25–28), we hypothesized that CD274 may be its potential substrate. We depleted endogenous USP22 via CRISPR–Cas9-mediated KO in a USP22-amplified HCC cell line (HepG2; ref. 29). To exclude off-target or selection-induced artificial effects, we generated three individual USP22 KO clones and found that all of them presented decreased CD274 expression (Fig. 1E). In contrast, overexpression of USP22 led to increased protein level of CD274 (Fig. 1F). Given that the stability of CD274 is negatively regulated by cell cycle–dependent cyclin D–CDK4 kinase (20), and USP22 is also involved in cell-cycle progression (26), we further tested CDK4-associated influences on USP22–CD274 regulation. Treatment with nocodazole (rapidly reversible inhibitor of microtubule polymerization) or palbociclib (CDK4/6-selective inhibitor) did not impede USP22 depletion–induced CD274 downregulation (Supplementary Fig. S2A and S2B). Similar results were obtained in specific siRNA-mediated interference of the CDK4 signaling pathway. Briefly, KD of CDKN2A (also known as p16 ARF, the upstream inhibitor of cyclin D–CDK4) or FZR1 (also known as CDH1, the downstream adaptor protein of cyclin D–CDK4) had no impact on USP22 overexpression–caused CD274 upregulation, although its basal levels were downregulated as reported after KDs (Supplementary Fig. S2C and S2D; ref. 20). Together, these data suggest that USP22 stabilizes CD274 in a CDK4-independent manner.

To determine if protein-protective effects depended on the catalytic activity of USP22, WT, catalytically inactive mutant (C185A), and UCH-contained domain (UCH) of USP22 were individually transfected into USP22 KO HepG2 cells (14). Only reexpression of WT USP22 was able to rescue CD274 level in USP22-deficient cells, suggesting that USP22 regulates CD274 protein stability through its deubiquitination-associated enzymatic activity (Fig. 1G). To understand the positive impact of USP22 on CD274 protein expression, we examined whether USP22 affects CD274 degradation through proteolysis. Altered CD274 levels in cells with USP22 depletion or overexpression could be restored by incubation with MG132 (a proteasome inhibitor; Fig. 1H and I). These results indicate that USP22 prevents proteasomal degradation of CD274. Consistent with this conclusion, the half-life of CD274 was markedly shorter in USP22 KO cells (Fig. 1J) but largely extended in USP22 overexpression cells (Fig. 1K). We sought to determine whether USP22 catalyzed the deubiquitination of CD274 and found that USP22 depletion induced an increase of ubiquitinated CD274 in vivo, which was largely reduced after rescue of USP22 (Fig. 1L). Finally, deubiquitylation assays validated that WT USP22, but not the inactive CA mutant, could directly deubiquitinate purified ubiquitinated CD274 protein in vitro (Fig. 1M). Collectively, these results demonstrated that USP22 is a bona fide DUB of CD274.

To investigate the potential pathophysiologic significance of the USP22–CD274 axis, we generated CD274-overexpressed USP22 KO cells, as well as USP22 and CD274 DKO cells. We found neither overexpression nor depletion of CD274 had any significant promoting or inhibitory effects on USP22-controlled cell proliferation (Supplementary Fig. S3A and S3B), viability (Supplementary Fig. S3C and S3D), or colony formation (Supplementary Fig. S3E and S3F). Thus, CD274 is not required for USP22-mediated cancer cell growth in vitro. We analyzed the correlation of USP22 and CD274 in clinical cancer cases in the 68,054 samples from 65,852 patients in 228 cancer studies from TCGA, and found that USP22 and CD274 have frequent genetic alterations (1.5% and 1.9%, respectively) and significant occurrence tendencies (P < 0.001; Fig. 2A). USP22 was only highly expressed in liver HCC (LIHC) and cholangiocarcinoma (CHOL; Supplementary Fig. S4A), with 1.5% and 0.3% genetic alterations (Supplementary Fig. S4B), respectively, but not other cancer types (Supplementary Fig. S4C). These data suggest that liver cancer is an ideal model to study USP22–CD274 regulation in vivo.

To explore the impact of USP22 on tumor growth with or without a functional immune system and in response to immune-checkpoint blockade and conventional chemotherapy, we utilized WT and USP22 KO H22 mouse liver cancer cell lines (30). We observed that in severe immunodeficient NCG mice, USP22 KO cell–derived tumor growth was very similar to WT control cells, although differences in tumor size and weight were statistically significant at day 30 (Fig. 2B–D). In contrast, we found USP22 KO largely limited tumor growth in immunocompetent C57BL/6 mice, which indicates that USP22 plays a critical role in tumor immune evasion (Fig. 2E–G). To further evaluate USP22-suppressed immunogenicity, paired WT and USP22 KO cells were individually injected s.c. into the opposite sides of same immunocompetent C57BL/6 mice (Fig. 2H). Results showed that the growth of WT tumors was similar to USP22 KO tumors (Fig. 2I and J). Because USP22 is a DUB of CD274, its depletion-induced destabilization of CD274 may improve CD274-targeted immunotherapy. As expected, the therapeutic efficiency of anti-CD274 antibody was significantly improved in USP22 KO tumors compared with WT tumors, indicating that USP22 depletion facilitates antitumor immune response (Fig. 2K–M). To investigate its synergistic effects between USP22 depletion and conventional chemotherapy, C57BL/6 mice bearing WT and USP22 KO tumors were treated with cisplatin (CDDP). USP22 depletion improved the therapeutic efficacy of CDDP compared with WT tumors (Fig. 2N–P). Depletion of USP22 in CD274 KO H22 cells was still able to suppress tumor growth in immunocompetent mice, which implies that the CD274-independent roles of USP22 cannot be completely ignored (Supplementary Fig. S5A–S5C). Taken together, these results demonstrate that USP22 depletion can potentiate anticancer immunity.

As CD274 has great influence on TILs, which are crucial for the efficacy of cancer immunotherapy (31, 32), we tested the CD274 expression and absolute number and function of TILs in our tumor models. As expected, CD274 expression levels were significantly downregulated in tumors derived from USP22 KO H22 cells in NCG mice (Supplementary Fig. S6A), C57BL/6 mice (Supplementary Fig. S6B), C57BL/6 mice coinoculated with USP22 KO and WT tumors (Supplementary Fig. S6C), during CD274-targeted immunotherapy (Supplementary Fig. S6D), or during CDDP-based chemotherapy (Supplementary Fig. S6E). Absolute numbers of CD3⁺ TILs and relative ratios of CD8⁺ granzyme B (GZMB)⁺ cells in CD3⁺ TILs were largely increased in USP22 KO tumors from individually inoculated mice (Fig. 3A and B), mice coinoculated with USP22 KO and WT tumors (Fig. 3C and D), during immunotherapy (Fig. 3E and F), or during chemotherapy (Fig. 3G and H). In addition to molecular and cellular validation, bioinformatics analyses also revealed that USP22 positively correlates with negative immune signatures in clinical LIHC, such as CD274 (Supplementary Fig. S7A), exhausted T-cell signatures (Supplementary Fig. S7B), effector regulatory T-cell (Treg) signatures (Supplementary Fig. S7C), and resting Treg signatures (Supplementary Fig. S7D).

Tools for the prediction, diagnosis, and prognosis of liver cancer are lacking (33). To determine if USP22 may be a prognostic marker for liver cancer, we performed Kaplan–Meier survival analyses and found that the high expression of USP22 is correlated to worse overall survival (OS) of liver cancer patients [HR, 1.71; 95% confidence interval (CI), 1.13–2.58; P < 0.01; Fig. 4A], as well as those who received sorafenib treatment (HR, 4.34; 95% CI, 0.93–20.3; P < 0.05; Fig. 4B). The survival correlation of USP22 was not related to hepatitis virus infection (HR, 1.66; 95% CI, 0.79–3.49; P = 0.18; Fig. 4C) or alcohol consumption (HR, 1.85; 95% CI, 0.82–4.16; P = 0.13; Fig. 4D). To further confirm the role of USP22 in the regulation of CD274 in human samples, we tested the expression of both genes in HCC. At the protein level, the expression of USP22 was strongly correlated with that of CD274 (Fig. 4E). Intriguingly, there was a positive correlation between USP22 and CD274 expression in mRNA levels using an in-house HCC cDNA bank (Fig. 4F). Collectively, we revealed a novel link between USP22, CD274 and cancer immunity, and have proposed a working model to summarize our findings (Supplementary Fig. S8).

As an established oncoprotein, the tumor-promoting functions of USP22 have been attributed to its protective or synergetic effects on a series of downstream regulatory factors, such as SAGA, CCNB1, and KDM1A (25, 26, 28). For instance, USP22 catalyzes H2B deubiquitination and mediates MYC-activated transformation through SAGA complex (26). USP22 can also facilitate cell-cycle progression and colorectal tumorigenesis by targeting CCNB1 (25). USP22 also contributes to glioblastoma tumorigenesis via stabilizing KDM1A (28). Although the direct and intracellular roles of USP22 in driving tumor growth have been clearly described as above, its potential influence on the microenvironmental immune system are obviously neglected, especially in liver cancer. In this study, we found that USP22 acted as a novel DUB of CD274 in liver cancer, deubiquitinated CD274 to stabilize its protein expression, and thus caused cancer immune resistance.

Accumulating evidence suggests that CD274 expression levels on tumor cells can predict the therapeutic efficacy of CD279/CD274-targeted immunotherapy (34). However, USP22 depletion not only improved the therapeutic efficacy of CD274-targeted immunotherapy but also potentiated CDDP-based chemotherapy, indicating the complicated roles of the USP22–CD274 axis in the efficacy of cancer therapy. Given that blocking or downregulating CD274 expression synergizes with CTLA4-targeted cancer therapy (13, 35), further investigation into the synergistic effects of USP22 depletion and other immune checkpoint blockade (non–CD274 based) is needed. In addition to targeting transcriptional activators of CD274, such as JAK, MYC, CDK5, and RAS (36–39), or the combination with conventional chemotherapy or targeted therapies (40–42), USP22 inhibition may be another promising strategy for improving immunotherapeutic efficacy.

No potential conflicts of interest were disclosed.

Conception and design: X. Huang, Q. Zhang, T. Liang, X. Bai

Development of methodology: X. Huang, Q. Zhang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): X. Huang, Q. Zhang, Y. Lou, X. Zhao, X. Zhang, S. Li, Q. Chen

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): X. Huang, Q. Zhang, X. Zhao, L. Wang, X. Zhang, T. Liang

Writing, review, and/or revision of the manuscript: X. Huang, Q. Zhang, Y. Zhao, T. Liang

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): X. Huang, Q. Zhang, J. Wang, Q. Chen

Study supervision: X. Huang, T. Liang

X. Huang would like to express deepest thanks to Guido Kroemer for the cancer immunity–associated technological training, ideological inspiration, and moral edification in his lab. This work was supported by grants from the National Natural Science Foundation of China (31970696 and 81502975, to X. Huang; 81830089, to T. Liang; and 81871320, to Q. Zhang), the China Postdoctoral Science Foundation (2016T90413 and 2015M581693; to X. Huang), and the SEU-Alphamab Joint Center (SA2015001; to X. Huang). The research was funded in part by Jiangsu Planned Projects for Postdoctoral Research Funds (1501002A; to X. Huang), Fundamental Research Funds for the Central Universities (2242016R20027 and 2242016K41045; to X. Huang), and Zhejiang Provincial Program for the Cultivation of High-level Innovative Health Talents (to X. Bai).