Targeting EZH2 Enhances Antigen Presentation, Antitumor Immunity, and Circumvents Anti–PD-1 Resistance in Head and Neck Cancer

Head and neck squamous cell carcinoma (HNSCC) is the sixth most common cancer worldwide (1). Patients with advanced human papillomavirus (HPV)-negative HNSCC have poor outcomes. Programmed cell death protein-1 (PD-1) checkpoint blockade with nivolumab or pembrolizumab has shown promising clinical outcomes in patients with poor prognosis recurrent/metastatic HNSCC (2–6). However, the majority do not benefit with responses limited to 15% to 20% of patients. This resistance to immunotherapy is unexpected given the high mutational burden of cisplatin-resistant tumors and the presence of immune infiltrates that together should predict better response rates (2–6). These findings lead to the question: can combination therapies directed at nonoverlapping pathways improve outcomes of anti–PD-1 therapy in HNSCC?

Clinical responses to mAbs blocking PD-1/PD-L1 signaling are based on the reversal of the dysfunctional exhausted state of tumor-infiltrating T cells. Mechanistically, IFNγ production by tumor-infiltrating T cells induces PD-L1 expression, among other downstream biologic changes, in the tumor microenvironment (7, 8). Deficiency of genes in the IFNγ signaling pathway or HLA class I have been shown to impair the efficacy of immunotherapy (9–11). Chan and colleagues reported that specific patient HLA class I genotypes are associated with response to immune checkpoint blockade (12). In HNSCC, dysfunctional HLA class I antigen processing and presentation has been identified as a key factor contributing to tumor progression and immunotherapy resistance (13–15). Deficiency in the expression of HLA class I components, including B2M and specific HLA alleles, and antigen-processing machinery have been correlated with poor prognosis of HNSCC (16, 17). Together, these observations suggest that approaches which enhance target tumor cell antigen presentation may improve the efficacy of anti–PD-1 therapy in HNSCC.

Enhancer of zeste homolog 2 (EZH2) is the methyltransferase subunit of the polycomb repressive complex 2 (PRC2) that catalyzes histone H3 methylation on lysine 27 (H3K27). This H3K27me3 histone modification can suppress chromatin accessibility and silence downstream gene expression (18, 19). The role of EZH2 in different components of antitumor immunity has been investigated by several groups. Zou and colleagues reported the contribution of EZH2 in regulating T-cell recruiting, Th1 chemokines, CXCL9 and CXCL10 in ovarian cancer cells. Consequently, combination therapy with EZH2 inhibition and adoptive T-cell transfer resulted in better tumor growth suppression than either approach alone in ovarian tumor–bearing mice (20, 21). In regulatory T cells (Tregs), disruption of EZH2 can enhance antitumor immunity by diminishing the suppressive activity of Tregs and enhancing T-cell infiltration in the tumor (22, 23). Combining EZH2 inhibition with anti-CTLA4 resulted in a synergistic impact in mouse bladder carcinoma and melanoma models (23). Sommer and colleagues reported the role of EZH2 in regulating a broad range of pathways impacting immunogenicity in melanoma (24). In addition, EZH2 is also involved in natural killer (NK) cell–mediated tumor eradication in hepatocellular carcinoma by silencing the expression of NK group 2D ligands (25). Very recently, diffuse large B-cell lymphoma showed a strong correlation between EZH2 mutation enrichment and MHC class I and class II expression deficiency. EZH2 inhibition relieved suppression of NLRC5 and CIITA expression leading to increased expression but only in an EZH2-mutated background (26). However, in HNSCC, the specific role of EZH2 in modulating MHC class I antigen presentation and T-cell–mediated antitumor immunity has not been investigated.

In this study, we hypothesized that inhibiting EZH2 function may improve the outcome of anti–PD-1 therapy by enhancing antigen presentation in HPV-negative HNSCC. Using both pharmacologic inhibition and CRISPR-Cas9–mediated genome editing, we investigated the role of EZH2 in human HNSCC HLA class I expression and also its role in antigen presentation and T-cell–mediated killing in preclinical models. Extending these data to a novel anti–PD-1 therapy-resistant preclinical model, we found that combinatorial therapy of GSK126, an EZH2 inhibitor, and anti–PD-1 showed significant tumor growth suppression compared with single-agent therapy. Together, these data identify EZH2 as a potential therapeutic target in promoting antigen presentation and antitumor immunity in HNSCC.

Human HNSCC lines CAL-33, CAL-27, SCC-9, and SCC-25 were obtained from ATCC and maintained in the DMEM/Nutrient Mixture F-12 (DMEM/F-12) + GlutaMAX media supplemented with 10% heat-inactivated FBS and 100 U/ml penicillin–streptomycin. Mouse oral squamous cell carcinoma models, MOC1-esc1 and MOC2, were maintained in IMDM/Hams-F12 (2:1) supplemented with 5% heat-inactivated FBS, 100 U/mL penicillin–streptomycin, 5 ng/mL EGF (Millipore), 400 ng/mL hydrocortisone (Sigma Aldrich), and 5 μg/mL insulin (Sigma Aldrich). Cell lines were routinely tested for Mycoplasma and underwent short tandem repeat cell line authentication at the DFCI within 6 months of use. MOC1-esc1, an anti-PD-1–resistant line, is isogenic to the previously described MOC1 model (27). Derivation and further characterization will be described elsewhere (L. Zhou and colleagues; in preparation). C57BL/6 mice (6-week-old, females) were from Taconic and OT-1 mice were purchased from Jackson Laboratory [C57BL/6-Tg (TcraTcrb)1100Mjb].

EPZ6438 and GSK126 were purchased from Selleckchem (S7128) and Chemietek (CT-GSK126), respectively. Recombinant human and mouse IFNγ were purchased from Peprotech (AF-300-02 and AF-315-05). OVA (257-264) SIINFEKL peptide (AS-6-193-1) was purchased from AnaSpec. Primary antibodies against mouse EZH2 (4905; Cell Signaling Technology), β-actin (4967; Cell Signaling Technology), H3K27me3 (9733; Cell Signaling Technology), and Histone H3 (ab1791; Abcam) were used for Western blot analysis. Secondary antibody [IRDye 800CW Goat anti-Rabbit IgG (H + L)] was purchased from LI-COR. EZH2 antibody (39901; Active Motif), H3K27me3 antibody (39055; Active Motif), and Rabbit IgG (15006; Sigma Aldrich) were used in the chromatin immunoprecipitation (ChIP) assay. For in vivo mouse studies, we used a rat antimouse monoclonal anti-PD-1 (RMP1-14, BE0146) and a rat IgG2a isotype control (2A3, BE0089) from BioXCell.

Fluorophore conjugated antibodies specific for human HLA-A, B, C (W6/32, 311403), mouse CD45.2 (104, 109814), mouse H-2K^b (AF6-88.5, 116518), mouse PD-L1 (10F,9G2, 124312), mouse OVA-Kb (25-D1.16, 141603), and 7-AAD Viability Staining Solution (420403) from BioLegend were used in flow cytometry. Rat antimouse CD16/CD32 (2.4G2, 553142) was purchased from BD Biosciences. Mouse tumors were dissociated to single-cell suspension with Tumor Dissociation Kit (130-096-730) and gentleMACS Dissociator (130-093-235) from Miltenyi Biotec. All flow cytometry analyses were performed on a MACSQuant Analyzer 10 (Miltenyi) and interpreted using FlowJo10 (TreeStar).

CD8⁺ T cells were isolated from OT-1 mouse splenocytes using CD8α⁺ T-cell Isolation Kit (130-104-075; Miltenyi Biotec). Isolated T cells were cultured in RPMI1640, supplemented with 10% heat-inactivated FBS, 20 mmol/L HEPES (15630080; Gibco), 1 mmol/L sodium pyruvate (11360070; Gibco), 0.05 mmol/L 2-mercaptoethanol, 2 mmol/L L-glutamine (G7513; Sigma), and 100 U/mL penicillin–streptomycin. For activation, Dynabeads Mouse T-Activator CD3/CD28 (11456D; Thermo Fisher Scientific) were added to isolated T cells at a bead-to-cell ratio of 1:2. Recombinant mouse IL2 (402-ML-020/CF; R&D Systems) was added to a final concentration of 20 ng/mL on day 1 after isolation. Medium supplemented with 20 ng/mL IL2 was changed every 2 days. T cells were split every 2 days to keep an approximate cell density of 10⁶ cells/mL. Activated OT-1 T cells were used for coculture killing assay on day 6 to 8 after isolation. For coculture killing assays, mouse IFNγ (20 ng/mL) prestimulated tumor cells were pulsed with 0.02 nmol/L of SIINFEKL peptide for 2 hours at 37°C. Activated OT-1 cells were counted and plated with washed tumor cells at indicated E:T ratios. After 24 hours of coculture, T cells were gently washed off using PBS. Surviving cells were harvested and stained with mouse CD45.2 antibodies to gate out residual T cells. Surviving tumor cells were counted using a MACSQuant Analyzer 10 (Miltenyi).

For the assessment of protein expression levels, RIPA buffer supplemented with protease inhibitor cocktail (11836153001; Sigma Aldrich) was used to homogenize cells. Fifty micrograms of total protein was separated on 4% to 12% SDS-PAGE gels (WG1403A; Invitrogen). Antibodies specific for mouse EZH2 (1:1,000), β-actin (1:1,000), H3K27me3 (1:1,000), and total H3 (1:2,000) were used for probing specific proteins. Secondary antibodies were then used to visualize specific proteins in a Li-Cor Odyssey imaging system.

Human or mouse cells (1,500/well) were plated in 96-well plates (3903; Corning) 1 day before starting drug treatment. Cells were treated with increasing concentrations of GSK126, EPZ6438, or vehicle (DMSO) for 72 hours. Cell viability was analyzed using CellTiter Glo-Luminescent Cell Viability Assay (G7572; Promega). Luminescent signal was measured with an Infinite M200 Pro Multimode microplate reader (Tecan Life Science).

Total RNA was extracted from the cultured cells using RNeasy Mini Kit (74106; Qiagen). One microgram of extracted RNA was used for cDNA synthesis using High-Capacity cDNA Reverse Transcription Kit (4368814; Thermo Fisher Scientific) according to the manufacturer's instructions. TaqMan real-time PCR assays specific for mouse B2m (Mm00437762_m1), H2-K1 (Mm01612247_mH), Cxcl10 (Mm00445235_m1), Gapdh (Mm99999915_g1), human B2M (Hs00187842_m1), HLA-A (Hs01058806_g1), HLA-B (Hs00818803_g1), HLA-C (Hs00740298_g1), CXCL10 (Hs00171042_ m1), and GAPDH (Hs02786624_g1) were purchased from Thermo Fisher Scientific and assays were performed on an ABI Step One Plus for quantifying gene expression levels. Data were analyzed using ΔΔC_t method and normalized to GAPDH.

ChIP assay was performed according to an Abcam protocol (https://www.abcam.com/protocols/cross-linking-chromatin-immunoprecipitation-x-chip-protocol). Briefly, tumor cells (10⁷) were cross-linked with 1% formaldehyde and lysed with SDS lysis buffer. The chromatin extract was sonicated on a Covaris Ultrasonicator. For each precipitation, 5 μg of antibody was incubated with Dynabead Protein A/G (10001D and 10003D; Invitrogen) for 6 hours, and sequentially incubated with sonicated chromatin overnight. ChIP DNA was isolated and quantified. qPCR was performed using SYBR Green with the following primers. Primer sequences for B2M promoter region: primer 1 forward: 5′-GTCCGGATTGGCTGTGAGTT-3′ and reverse: 5′-GAGCCATGCTGACGACTGAA-3′; primer 2 forward: 5′-AGCTAGGAGACTGGTGACGA-3′ and reverse: 5′-AGCCATGCTGACGACTGAAG-3′.

Female C57BL/6 mice (6-week-old; Taconic) were housed in a pathogen-free animal facility and all experiments performed were approved by the Institutional Animal Care and Use Committee of Dana-Farber Cancer Institute. For tumor inoculation, MOC cells were harvested and washed extensively with cold endotoxin-free PBS. Cells (10⁶ MOC1-esc1 and 10⁵ MOC2) were injected subcutaneously into the right flank of each mouse in a volume of 150 μL. Anti–PD-1 therapy was performed by intraperitoneal injections of anti–PD-1 (250 μg/mouse) or isotype control (250 μg/mouse) on day 3, 6, and 9 post-inoculation. GSK126 was dissolved in 20% Captisol (H107; Sigma Aldrich) and administered by intraperitoneal injections 3 times a week from day 6 post-inoculation. Tumor growth was monitored using a digital caliper twice weekly. Tumor volume was calculated using the formula V = (W² × L)/2.

LentiCas9-Blast (Addgene, #52962) was cotransfected with pCMV-VSV-G and psPAX2 into HEK293T cells for lentivirus production using TransIT-293 (MIR2700, Mirus Bio). Virus in the media was collected 48 hours after transfection and MOC1-esc1 cells were transduced for 48 hours and then subjected to Blasticidin (4 μg/mL) selection for 4 days. Pooled MOC1-esc1-cas9 cells were single-cell sorted in 96-well cell culture plate to generate single clones. Cas9 editing efficiency of each clone was tested by assessing the knockout efficiency of B2M single-guide RNA (sgRNA)-expressing lentivirus-transduced cells. H2-K^b class I cell surface expression levels were determined by flow cytometry as a readout of Cas9 efficiency (not shown). Clones #2 and #3 with high editing efficiency were transduced with lentivirus-expressing specific gRNAs targeting EZH2 or ROSA26, respectively. The sequences of specific sgRNAs are listed below.

• EZH2-g1-oligo1: CACCGTATCGTAGTAAGTACCAATG

• EZH2-g1-oligo2: AAACCATTGGTACTTACTACGATAC

• EZH2-g2-oligo1: CACCGAGAGTACATTATAGGCACCG

• EZH2-g2-oligo2: AAACCGGTGCCTATAATGTACTCTC

• ROSA26-oligo1: CACCGCAATCAGCGGAGGCTGCCG

• ROSA26-oligo2: AAACCGGCAGCCTCCGCTGATTGC

Naïve CD8⁺ T cells isolated from OT-1 mouse splenocytes were stained with CFSE Cell Division Tracker Kit (423801; BioLegend) at a final concentration of 5 μmol/L in PBS supplemented with 5% heat-inactivated FBS for 3 minutes in the dark at room temperature. After staining, cells were washed 3 times in PBS with 5% FBS. IFNγ prestimulated tumor cells were plated with carboxyfluorescein diacetate succinimidyl ester (CFSE)-labeled T cells at an E:T ratio of 10:1. After 72 hours coculture, media were harvested for IFNγ ELISA. CD8⁺ T cells proliferation measured by CFSE-dilution was determined using flow cytometry.

Data for HNSCC and lung SCC patients from The Cancer Genome Analysis (TCGA) were obtained from cBioPortal. HPV status of patients was obtained from a PanCancer Atlas study (28). Gene expression correlations between EZH2 and molecules involved in antigen presentation machinery were assessed by Spearman coefficient.

IFNγ levels in the media from T-cell proliferation coculture experiments were measured with the Mouse IFNγ DuoSet ELISA (DY485; R&D Systems) according to the manufacturer's instructions.

Data are plotted as the mean ± SD or mean ± SEM as indicated in specific experiments. The statistical significance was determined by Student t test, one-way ANOVA, and two-way ANOVA using GraphPad Prism. Significance differences P < 0.05, 0.01, and 0.001 are symbolized as *, **, and ***, respectively.

As antigen presentation defects have been identified as a common feature in HNSCCs, we sought to identify the relationship between EZH2 expression and antigen presentation pathway components. Analysis of 522 HNSCC HPV-negative tumors from TCGA (28) showed a significant inverse correlation between the EZH2 expression levels and major MHC class I antigen presentation molecules, including β2M, HLA-A, HLA-B, HLA-C, and HLA-E (Fig. 1), highlighting a potential regulatory function of EZH2 on antigen presentation in HNSCC. Interestingly, this negative correlation was also seen in 483 HPV-negative lung squamous cell carcinomas (Supplementary Fig. S1).

To test the hypothesis that targeting EZH2 could promote antigen presentation in HNSCC, we used two highly selective EZH2 inhibitors, GSK126 and EPZ6438 (22, 29, 30). Human HNSCC cell lines (CAL27, CAL33, SCC25, and SCC9) were treated with increasing concentrations of GSK126 or EPZ6438 to test the impact of EZH2 inhibition on cell viability. On the basis of the various sensitivities of these cell lines to EZH2 inhibitors, we selected the 10 μmol/L concentration for both GSK126 and EPZ6438, at which neither inhibitor showed significant cell growth inhibition (Fig. 2A). Although GSK126 increased IFNγ-induced HLA expression in three of four lines tested, EPZ6438 increased both basal and IFNγ-induced HLA cell surface expression levels compared with DMSO control in all cell lines (Fig. 2B). Given the known function of EZH2 as a methyltransferase that silences target genes, we quantified the mRNA expression changes of MHC class I antigen presentation genes, including B2M, HLA-A, HLA-B, and HLA-C after treatment with EZH2 inhibitors. Despite different sensitivities to EZH2 inhibition, major class I antigen presentation genes were upregulated by at least one EZH2 inhibitor under either basal or exogenous IFNγ-stimulated conditions or both in all tested cell lines (Fig. 2C; Supplementary Fig. S2). Therefore, EZH2 regulated mRNA levels of antigen presentation genes. Consistent with findings in ovarian cancer (20), we also observed a significant induction of CXCL10 expression in EZH2 inhibitor–treated cells compared with DMSO control that was dramatically enhanced in combination with IFNγ (Fig. 2C). Therefore, EZH2 inhibition influenced MHC class I expression and CXCL10 expression in human HNSCC lines.

Aiming at reversing immunotherapy resistance in HNSCC, we tested the effect EZH2 inhibition in MOC1-esc1, a syngeneic anti–PD-1–resistant mouse model. When assayed for impact on cell viability assay, significant cytotoxicity in MOC1-esc1 was not seen with either EZH2 inhibitors at the 10 μmol/L dose (Fig. 3A). We then used GSK126 or EPZ6438 to determine the effect of EZH2 inhibition on antigen presentation in MOC1-esc1. Consistent with the data in human HNSCC, either treatment significantly upregulated MHC class I cell surface protein levels in combination with IFNγ (Fig. 3B and C). To confirm that the role of EZH2 in antigen presentation modulation was not model specific, we performed similar experiments using MOC2, another immunotherapy-resistant mouse HNSCC model. Consistent with the results in MOC1-esc1, EZH2 inhibitor–treated MOC2 cells showed enhanced MHC class I expression (Fig. 3C). To assess whether this increased expression also resulted in increased antigen presentation, we assessed levels of MHC class I bound SIINFEKL. EZH2 inhibition also increased the presentation of SIINFEKL peptide by H2-K^b on the MOC1-esc1 and MOC2 cell surface confirming the functional upregulation of antigen presentation (Fig. 3D). Together, these data showed that EZH2 inhibition increased antigen presentation in anti–PD-1–resistant HNSCC cells.

Next, we asked whether the EZH2 inhibition based enhanced MHC class I antigen presentation sensitized tumor cells to T-cell–mediated killing. Note that, in addition to pharmacologic targeting and to further study the role of EZH2 in T-cell/tumor cell interactions, we generated CRISPR-CAS9 genetically deleted EZH2 MOC1-esc1. We used two independent clones of CAS9-expressing MOC1-esc1 to knockout EZH2 with two different sgRNA to rule out clone-specific or off-target effects. Immunoblotting for EZH2 was performed and showed attenuated EZH2 expression relative to ROSA26 control targeted lines (Fig. 4A). We performed a 2D coculture assay using OT-1 CD8⁺ T cells that were in vitro activated and expanded with SIINFEKL peptide antigen pulsed tumor cells as targets. GSK126 or EPZ6438 treatment sensitized MOC1-esc1 cells to T-cell–mediated killing (Fig. 4B). Genetic ablation of EZH2 dramatically sensitized tumor cells to T-cell–mediated killing in both clones in comparison with their parental lines and ROSA26 targeting controls (Fig. 4B; Supplementary Fig. S6). Consistent with the inhibitor treatment experiment results, loss of EZH2 significantly increased MHC class I cell surface expression levels, which again were enhanced in combination with IFNγ without impacting PD-L1 (Fig. 4C and D), indicating the specificity of this regulation on antigen presentation. Therefore, targeting of EZH2 sensitized tumor cells to T-cell–mediated killing.

To start to define the mechanism of EZH2 regulation of antigen presentation, we tested H3K27me3 levels in GSK126- or EPZ6438-treated cells. As expected, inhibition of EZH2 resulted in dramatic decrease of global H3K27me3 levels, without affecting the protein expression levels of EZH2 (Fig. 5A). In addition, the mRNA levels of both B2M and H2-K1 were significantly upregulated by EZH2 inhibition (Fig. 5B), suggesting that the regulation of EZH2 on antigen presentation is conserved between human and mouse (Fig. 2B and C; Supplementary Fig. S2). Interestingly, CXCL10 expression was not induced by EZH2 inhibition in this mouse model (Fig. 5B).

To test the hypothesis that EZH2 regulates H3K27me3 occupancy on the promoter region of B2M, MOC1-esc1 cells were treated with EPZ6438 or DMSO and subjected to ChIP with antibodies against EZH2, H3K27me3, or IgG control. To test the occupancy of EZH2 and H3K27me3 modification on B2M promoter, DNA samples from ChIP were quantified using two independent primers specific for B2M promoter region. The results of ChIP assay followed by qPCR showed that EPZ6438 treatment did not affect the binding of EZH2 on B2M promoter region (Fig. 5C), which is in agreement with the mode of action of EPZ6438 as an S-adenosyl-I-methionine (SAM)-competitive inhibitor. Importantly, EPZ6438 treatment significantly reduced H3K27me3 enrichment on the B2M promoter region (Fig. 5D).

Next, we asked whether the enhanced antigen presentation in EZH2-deficient tumor cells can promote the function of antigen-specific T cells. To answer this question, we analyzed CD8⁺ T-cell proliferation and IFNγ production after coculture of SIINFEKL peptide pulsed tumor cells with naïve OT-1 T cells. These data show that both T-cell proliferation and IFNγ production were significantly higher in EZH2-deficient tumor cells compared with ROSA26 control (Fig. 6A–C). To investigate whether EZH2 inhibitor can improve the outcome of anti-PD-1 therapy, we used resistant MOC1-esc1 to assess the effect of combination therapy. MOC1-esc1 tumor growth in immunocompetent wild-type mice was significantly attenuated by the combination of GSK126 (50 mg/kg) and anti-PD-1 but not by either agent alone (Fig. 6E). To test the role of EZH2 inhibition in modulating antigen presentation of tumor cells in vivo, we analyzed the tumor samples with these treatment conditions (Fig. 6F and G). Tumor cell MHC class I cell surface expression levels were increased by GSK126 monotherapy compared with the control group (Fig. 6F). Similarly, anti–PD-1 monotherapy also increased the MHC class I level compared with control (Fig. 6F). Interestingly, the combination of GSK126 and anti–PD-1 resulted in an additive increase of MHC class I (Fig. 6F). Total CD45⁺ immune infiltration in the tumors was not affected by either monotherapy or combinatorial therapy (Fig. 6G; Supplementary Fig. S8). In summary, enhanced antigen presentation induced by targeting of EZH2 promoted proliferation and IFNγ production of antigen-specific T cells and combination therapy of GSK126 and anti–PD-1 resulted in attenuation of tumor progression in an anti–PD-1–resistant mouse model of HNSCC.

In addition, as significant numbers of patients are resistant to immunotherapy due to poor immune infiltration, we asked whether the efficacy of GSK126 and anti-PD-1 combinatorial therapy requires preexisting immune infiltration in tumors. We next assessed for combination therapy impact in the MOC2 model that displays an immune “desert” phenotype with negligible T-cell infiltration and aggressive in vivo growth (31). Consistent with previous data (31), MOC2 tumors are resistant to anti–PD-1 treatment (Fig. 6H). Neither GSK126 monotherapy nor the combination of GSK126 and anti–PD-1 suppressed tumor growth (Fig. 6H). Although the growth phenotype was not altered, we assayed for in vivo MOC2 tumor cell class I expression with different treatments. Neither the control nor the monotherapy groups displayed alteration in surface class I expression. However, consistent with our observation in MOC1esc1 tumors, the combination group induced a significant higher MHC class I on tumor cells in vivo. As for MOC1-esc1, there was minimal impact on CD45⁺ T-cell infiltration. Therefore, EZH2 inhibition and anti–PD-1 combinatorial therapy upregulates tumor cell MHC Class I expression in vivo, whereas the therapeutic efficacy may require preexisting immune infiltration.

To define approaches to improve outcomes of anti-PD-1 therapy in HNSCCs, we identified that targeting the histone methyltransferase EZH2 enhanced antigen presentation in both human and mouse HNSCC lines. Mechanistic studies revealed that EZH2 inhibition decreased the enrichment of H3K27me3 on the promoter region of B2M. The enhanced antigen presentation induced by targeting EZH2-sensitized tumor cells to T-cell–mediated killing, resulted in higher T-cell proliferation and IFNγ production in vitro. These findings translated to a synergistic impact of combination GSK126 and anti-PD-1 in attenuating the tumor progression of an anti-PD-1–resistant model of HNSCC and showed that EZH2 inhibition can enhance tumor cell Class I expression in vivo including in highly resistant models. Our data define EZH2 as a therapeutic target to improve the outcome of anti-PD-1 therapy in HNSCC.

EZH2 inhibition has been proposed as a therapeutic strategy in cancers with frequent gain-of-function mutation or over expression of EZH2, including melanoma, ovarian cancer, and lymphoma to induce apoptotic cell death (32–34). In HNSCC, overexpression of EZH2 has been correlated with poor prognosis, recurrence, and lymph node metastasis in HNSCC (35–37). Saunders and colleagues reported that targeting EZH2 in HNSCC attenuated tumor growth by inducing differentiation gene expression via modulating H3K27me3 occupancy on the promoter regions of genes involved in squamous differentiation (38). To test the effect of EZH2 inhibition on human HNSCC tumor progression, Cal27 was implanted in immunocompromised NOD/SCID mice, in which setting the interaction between tumor cells and host immunity was eliminated to support the human tumor growth in mice (38). These previous studies highlighted EZH2 inhibition as a potential therapeutic target in HNSCC by modulating epigenetic silencing of genes involved in tumor progression. In our study, we first demonstrated the effect of EZH2 inhibition on increasing antigen presentation is conserved in human and mouse HNSCC lines. Subsequently, a syngeneic mouse model was used to investigate the in vivo therapeutic potential of EZH2 inhibition independent from direct cytotoxicity of EZH2 on tumor cells. Thus, this study provides a new line of evidence in support of EZH2 inhibition as a new combination immunotherapy target in promoting HNSCC antitumor immunity to overcome checkpoint blockade resistance.

Consistent with other studies, inhibition of EZH2 significantly induced the expression of CXCL10 in human HNSCC lines (Fig. 2C; ref. 20). CXCL10 is a Th1-chemokine CXCL10 that is a key ligand regulating T-cell trafficking into the tumor microenvironment through CXCR3 (39, 40). However, CXCL10 expression was not increased by EZH2 inhibition in the HNSCC mouse model (Fig. 5B), suggesting the regulation of CXCL10 by EZH2 is not conserved in human and mouse models used in this study. Although speculative, this suggests that in human HNSCC, the effect of EZH2 inhibition on antitumor immunity might include enhanced antigen presentation and additionally, higher immune infiltration in tumors.

One limitation of this study is the lack of well-defined endogenous antigen and antigen-specific T cells in the MOC1-esc1 HNSCC model. Therefore, we employed a commonly used model antigen SIINFEKL peptide from ovalbumin and OT-1 mouse-derived antigen-specific T cells for in vitro studies (41). Supporting the use of this approach, CRISPR-CAS9 knockout of B2M or PD-L1 in MOC1-esc1 resulted in resistance or sensitivity to T-cell–mediated killing, respectively (data not shown). Therefore, this 2D coculture system retained the relevant function of EZH2 in T-cell recognition to tumor cells. In addition, as EZH2 impacts a broad range of pathways in vivo, dissecting the specific immune contribution will require further studies. Our work supports that EZH2 has the capacity to directly regulate tumor cell Class I in vitro and that inhibition in vivo results in a similar end result of enhanced Class I expression. Further mechanistic dissection of in vivo antigen presentation is required.

Others have studied the contribution of EZH2 in augmenting tumor immunogenicity including antigen presentation regulation in melanoma and lymphoma, respectively (24, 26). Our study focused on HNSCCs and defined further details of antigen presentation regulation. Differences in HNSCC include that the frequency of EZH2 overexpression in HNSCC is relatively low compared with that in cutaneous melanoma, lymphoma, and ovarian cancers. In addition, there are no EZH2 gain-of-function mutations identified in HNSCC (19). Therefore, the sensitivity of HNSCC to EZH2 inhibition induced apoptosis may be relatively low. The function of EZH2 inhibition in promoting antigen presentation thus identifies a disease-specific contextual opportunity in which to consider EZH2 inhibitor as an immunotherapy agent in HNSCC.

Interestingly, the inverse correlation of EZH2 and MHC class I antigen presentation molecule expression levels in HNSCC TGCA dataset is also found in lung squamous cell carcinoma, suggesting the function of EZH2 in regulating antigen presentation may be conserved in squamous cell carcinomas from other tissues. On the basis of our results, we propose that combination EZH2 inhibition with anti-PD-1 therapy may be beneficial for patients with HNSCC. Studies using different EZH2 inhibitors and additional HNSCC preclinical models are needed to further confirm our findings. In addition, the effect of EZH2 in regulating antigen presentation in professional antigen-presenting cells, such as macrophages and dendritic cells, also requires further investigation.

R. Uppaluri is a consultant for and has received research support from Merck. No potential conflicts of interest were disclosed by the other authors.

Conception and design: L. Zhou, R. Uppaluri

Development of methodology: L. Zhou, T. Mudianto, R. Uppaluri

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L. Zhou, X. Ma, R. Riley, R. Uppaluri

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L. Zhou, R. Uppaluri

Writing, review, and/or revision of the manuscript: L. Zhou, R. Uppaluri

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L. Zhou, X. Ma, R. Uppaluri

Study supervision: R. Uppaluri

This study was supported by the NIH/NIDCR (DE024403 and DE027736) to R. Uppaluri. The authors thank Drs. Zhong Wu and Deli Hong for technical assistance with ChIP.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.