USP22 drives tumor immune evasion and checkpoint blockade resistance through EZH2-mediated epigenetic silencing of MHC-I

Abstract

While immune checkpoint blockade (ICB) therapy has revolutionized the antitumor therapeutic landscape, it remains successful in only a small subset of patients with cancer. Poor or loss of MHC-I expression has been implicated as a common mechanism of ICB resistance. Yet, the molecular mechanisms underlying impaired MHC-I remain to be fully elucidated. Herein, we identified USP22 as a critical factor responsible for ICB resistance through suppressing MHC-I-mediated neoantigen presentation to CD8+ T cells. Both genetic and pharmacologic USP22 inhibition increased immunogenicity and overcame anti-PD-1 immunotherapeutic resistance. At the molecular level, USP22 functions as a deubiquitinase for the methyltransferase EZH2, leading to transcriptional silencing of MHC-I gene expression. Targeted Usp22 inhibition resulted in increased tumoral MHC-I expression and consequently enhanced CD8+ T cell killing, which was largely abrogated by Ezh2 reconstitution. Multiplexed immunofluorescence staining detected a strong reverse correlation between USP22 expression and both β2M expression and CD8+ T lymphocyte infiltration in solid tumors. Importantly, USP22 upregulation was associated with ICB immunotherapeutic resistance in patients with lung cancer. Collectively, this study highlights the role of USP22 as a diagnostic biomarker for ICB resistance and provides a potential therapeutic avenue to overcome the current ICB resistance through inhibition of USP22.

Keywords: Clinical Research; Immunology; Immunotherapy; MHC class 1; Oncology; Ubiquitin-proteosome system.

Figure 1. Usp22 inhibition enhances MHC-I expression.

Indicated cells were transfected with control (WT) or Usp22-specific guide RNAs (Usp22 KO). (A) Immunoblot analysis of MHC-I proteins in WT and Usp22 KO tumor cells. (B) Cell surface expression of H-2Kb and b2M were determined in WT and Usp22 KO cells. (C) Heatmap summarizing for the mRNA expression of genes involved in antigen presentation in WT and KO tumor cells. (D) Cell surface levels of OVA peptide SIINFEKL (pMHC-I) were determined in WT and Usp22 KO MC38/OVA or RM1/OVA cells. (E) Schematic illustration of an in vitro cytotoxicity assay. (F) The viability of WT and Usp22 KO MC38/OVA or RM1/OVA after cocultured with OT-I CD8⁺ T cells. (G and H) OT-I CD8⁺ T cell activation after cocultured with WT and Usp22 KO RM1/OVA or MC38/OVA cells were determined. (I and J) RM1/OVA (I) or MC38/OVA (J) cells were pretreated with or without 20 μM USP22i-S02 for 48 hours and then cocultured with OT-I CD8⁺ T cells. OT-I CD8⁺ T cell activation was determined as in I and J. (K) WT and Usp22 KO RM1/OVA or MC38/OVA cells were cocultured with OT-I CD8 T cells. The concentrations of IFN-γ or TNF-α in the supernatant were determined by ELISA (N = 9). (L) RM1/OVA or MC38/OVA cells were pretreated with USPi-S02 as in I and then cocultured with OT-I cells. The concentrations of IFN-γ and TNF-α in the supernatant were determined by ELISA (N = 9). (M) B2m was deleted by CRISPR in WT and Usp22 KO MC38/OVA and RM1/OVA cells. (N–Q) The effect of B2m deletion on CD8-mediated killing of tumor cells (N) and OT-I CD8⁺ T cell activation was determined as in P and Q. Statistics were calculated by unpaired 2-tailed t test (B, D, and F–L) or 1-way ANOVA followed by Tukey’s test (N–Q). *P < 0.05, **P < 0.01, and ***P < 0.001.

Figure 2. The absence of Usp22 dampens tumor growth by enhancing tumoral infiltrating CD8⁺ T cells.

(A and B) Effect of Usp22 depletion on tumorigenesis of RM1 cells in C57BL/6J mice. Tumor volume (A), endpoint tumor images and weight (B) are shown. Scale bar: 1 cm. (C and D) Effect of Usp22 depletion on tumorigenesis of RM1 cells in immunocompromised RAG1 knockout mice were determined as in A and B, Scale bars: 1 cm. (E) Flow cytometric analysis of the expression of H-2Kb or b2M on tumoral cells in A. (F and G) Tumoral infiltrating CD4⁺ and CD8⁺ T cells on the total CD45⁺ cells in tumors shown in A were analyzed by flow cytometry (F) or immunofluorescence staining (G). Scale bar: 100 μm. HPF, high powered field. (H–J) The production of granzyme B⁺ (H), IFN-γ⁺ (I), or TNF-α⁺ (J) by CD8⁺ in F. (K and L) Tumor-bearing mice were treated with CD8 depleting antibodies (100 mg) on day 6, 9, and 12. Tumor volume (K), endpoint tumor images, and weight (L) were recorded. Scale bar: 1 cm. (M and N) WT, Usp22 KO, B2m KO or double KO (dKO) RM1 cells were subcutaneously injected into C57BL/6J mice, tumor volume (M), endpoint tumor images, and weight (N) are shown. Scale bar: 1 cm. (O and P) Tumoral infiltrating CD8⁺ T cells (O) or their production of GZMB (P) were analyzed. Statistics were calculated by unpaired 2-tailed t test (B, D, E–J, and L) or 1-way ANOVA followed by Tukey’s test (N–P). Two-way ANOVA with multiple comparisons (A, C, K, and M). *P < 0.05, **P < 0.01, and ***P < 0.001.

Figure 3. USP22 is an EZH2-specific deubiquitinase.

(A and B) Immunoblot analysis of indicated protein levels in WT and KO tumor cells (A) or in tumor cells treated with 20 μM USP22i-S02 (B). (C) Immunofluorescence staining and quantification of EZH2 in WT and KO RM1 tumors. Scale bars: 50 μm. (D) Immunoblot analysis of indicated protein levels in WT and KO cells treated with or without MG132 (10 μM, 8 hours). (E and F) Ch-IP and qRT-PCR analysis for EZH2 (E), H3K27me3 (F) enrichment in B2m or H-2K1 genes promoter in WT and KO cells. (G) Analysis of USP22 interaction with PCR2 complex proteins by Co-IP and immunoblot. WCL, whole cell lysates. (H) Analysis of USP22 interaction with EZH2 in transiently transfected HEK-293T cells. (I) Recombinant GST/GST-USP22 were purified from bacteria and incubated with 4T1 cell lysate overnight. The binding proteins were analyzed by immunoblot. (J) Schematic illustration of USP22 and its truncated mutants. (K and L) Analysis of EZH2 interaction with USP22 and its mutants in transiently transfected HEK-293T cells. (M) Molecular docking analysis of the interaction between USP22 and EZH2. (N) EZH2 ubiquitination was determined from HEK293T cells in the presence of transient transfection of Myc-USP22/C185A, HA-ubiquitin. (O) Indicated cells were pretreated with 10 μM MG132 for 8 hours, EZH2 ubiquitination was determined. (P) HEK-293T cells cotransfected with FLAG-EZH2 and Myc-USP22 or its C185A mutant. After 24 hours’ transfection, cells were treated with 20 mg/mL cycloheximide (CHX) for the indicated time points and indicated protein levels were determined. (Q) RM1 KO cells were transfected with Usp22/C185A mutant. EZH2 protein stability was determined as in P. (R) EZH2 protein stability in WT and KO RM1 and MC38 cells were determined as in P. Statistics were calculated by unpaired 2-tailed t test (C, E, and F) or 2-way ANOVA with multiple comparisons (P and Q). *P < 0.05, **P < 0.01, and ***P < 0.001.

Figure 4. USP22 attenuates antitumor immunity partially through protecting EZH2 from degradation.

(A–C) Indicated cancer cells were treated with 10 ng/mL IFN-γ for indicated time points. The expression of indicated proteins was determined. (D) HEK-293T cells were cotransfected with FLAG-EZH2 and Myc-USP22 and then treated with 10 ng/mL IFN-γ for the indicated times. The interaction between USP22 and EZH2 was determined. (E) Tumor cells were isolated based on membrane b2M expression. Indicated protein expression was determined. (F) MC38/OVA or RM1/OVA were isolated according to cell surface pMHC-I. Indicated protein expression was determined. (G) OT-I CD8⁺ T cells were isolated from OT-I mice and cocultured with Usp22-deficient RM1/OVA or MC38/OVA cells with or without Ezh2, Ezh2 F667I, or ΔSET mutant reconstitution for 8 hours at the ratio of 1:1 in the presence of CD28 blocking antibodies treatment. Quantification data of flow cytometric analysis of percentages of GZMB⁺, IFN-γ⁺, and TNF-α⁺ producing CD8⁺ T cells are shown. (H) Cell viability of indicated cells after coculturing for 48 hours. (I) Living tumor cells were collected after coculture with naive OT-I CD8⁺ T cells for 48 hours at a ratio of 1:1 in the presence of CD28 blocking antibody treatment. Indicated protein levels were determined. (J and K) MC38 cells with Ezh2, Ezh2 F667I, or ΔSET mutant reconstitution in the setting of Usp22 depletion were inoculated into immunocompetent mice. Tumor volume (J) and endpoint mass (K) of indicated tumors were recorded. (L) The expression of b2M and H-2Kb on indicated tumor cell surface. (M and N) The frequencies of tumoral-infiltrating CD8⁺ T cells (M) or GZMB⁺ producing CD8⁺ T cells (N) from indicated MC38 tumors. Statistics were calculated by 1-way ANOVA followed by Tukey’s test (G, H, and K–N). Two-way ANOVA with multiple comparisons (C and J). *P < 0.05, **P < 0.01, and ***P < 0.001.

Figure 5. Analysis of USP22/EZH2/b2M expression in multiple types of cancers.

(A) Representative images of multiplexed immunofluorescence staining of USP22/EZH2/b2M/CD8 in breast cancer tissues with different USP22 intensity. Scale bar: 100 μm. H-score, histochemical scoring assessment. (B) The correlation between USP22 with b2M, USP22 with CD8, and USP22 with EZH2. (C) Quantification of tumoral-infiltrated CD8⁺ T cells, EZH2, or b2M intensity in breast cancer tissues with different USP22 intensity. Patients were classified into the USP22 intensity low or high group. The median value was used as cutoff. (D) Immunohistochemical staining of USP22/EZH2/b2M/CD8 in a prostate cancer tissue microarray. Scale bar: 200 μm. (E) The proportion of tumor-infiltrating CD8⁺ T cells or b2M and EZH2 intensity in different intensity cohorts. Patients were classified into the USP22 intensity low or high group. The median value was used as cutoff. (F) Immunohistochemical staining of USP22/EZH2/b2M/CD8 in colorectal tissue microarray containing 80 paired benign or colorectal cancer tissues. Scale bar: 200 μm. (G) The proportion of tumoral-infiltrated CD8⁺ T cells or b2M and EZH2 intensity in low or high USP22 intensity cohorts. Patients were classified into the USP22 intensity low or high group. The median value was used as cutoff. (H) Correlations between the mRNA levels of USP22 and B2M in breast cancer cell lines from Cancer Cell Line Encyclopedia (CCLE). (I) Correlations between the mRNA expression of USP22 and CD8 infiltration score in prostate or colorectal cancer from TCGA database. Statistics were calculated by unpaired 2-tailed t test (C, E, and G), 2-tailed Pearson correlation test (B, H, and I). *P < 0.05, **P < 0.01, and ***P < 0.001.

Figure 6. USP22 expression links with ICB resistance.

(A) Representative images of multiplexed immunofluorescence staining of USP22/EZH2/b2M/CD8 in 32 pretreatment biopsies taken from individual patients who received anti-PD-1 antibody therapy. Scale bar: 100 μm. (B) Kaplan-Meier plot of progression-free survival (PFS) for 32 patients with NSCLC who did or did not respond to anti-PD–1 antibody therapy. (C) Patients were divided into USP22 low or high groups according to USP22 expression. Frequency of responder or nonresponder with high or low USP22 expressions are shown. R, responder; NR, nonresponder. (D) Quantification data of USP22/EZH2/b2M intensity in biopsies from anti-PD–1 responders or nonresponders. (E) Kaplan-Meier plot of PFS for patients treated with anti-PD–1 in USP22 low versus high group. Patients were classified into the USP22 low or high groups, with the median expression value across all the samples used as the cutoff. (F and G) Pearson correlation analyses between indicated proteins expression in biopsies from patients who did or did not respond to anti-PD–1 therapy. (H) The mRNA expression of USP22 in pretreatment biopsies from patients with triple negative breast cancer who received anti-PD–1 therapy. Clinical responses were classified in the original studies GSE173839. Correlations between the mRNA expression of USP22 and B2M are shown. (I) The mRNA expression of USP22 in pretreatment biopsies with melanoma who received anti-PD–1 therapy. Clinical responses were classified in the original studies GSE91061. Correlations between the mRNA expression of USP22 and B2M are shown. Statistics were calculated by unpaired 2-tailed t test (D, H and I (left panel)), Fisher exact test (C), Log rank t test (B and E), 2-tailed Pearson correlation test (F and G, and H and I (right panel)). *P < 0.05, **P < 0.01, and ***P < 0.001.

Figure 7. Targeting USP22 overcomes ICB resistance.

(A and B) Effects of administration of anti-PD–1 on 4T1 or 4T1R tumor growth (A) and weight (B). Scale bar: 1 cm. Image of 4T1 or 4T1 R tumors treated with or without anti-PD–1 are shown. (C and D) Effects of Usp22 deficiency on 4T1 or 4T1R tumor growth (C) and weight (D). (E and F) Effects of S02 or anti-PD–1 in 4T-1 R tumor growth. Mice were randomly grouped into 4 groups and administered with 10 mg/kg USP22i-S02 and/or 100 mg anti-PD–1. Mice were treated with 10 mg/kg S02 daily from day 4 to day 9, or were given a combination treatment with 100 μg anti-PD-1 antibodies every other day from day 4 to day 8. (G–I) Representative flow cytometric images and quantification data of cell surface b2M (G), H-2Kd (H), or PD-L1 (I) MFI in indicated tumor cells. (J) Representative flow cytometric images and quantification of FoxP3 MFI. (K) Representative images of flow cytometric analysis and quantification of frequencies of Tregs among total CD4⁺ lymphocytes in indicated tumors. (L) Quantification of frequencies of CD8⁺ T cells among tumor-infiltrating CD45⁺ lymphocytes in indicated tumors. (M and N) Representative flow cytometric images and quantification of frequencies of GZMB- (M) or IFN-γ– (N) producing tumor-infiltrating CD8⁺ T cells in indicated tumors. (O) Proposed working model showing that USP22 inhibition enhances antitumor immunity through increases in EZH2 proteasomal mediated degradation and MHC-I mediated. CD8⁺ T cells recognition and killing. Pharmacological USP22 inhibition overcomes immune checkpoint blockade resistance. Statistics were calculated by 1-way ANOVA followed by Tukey’s test (B, D, and F–N) or 2-way ANOVA with multiple comparisons (A, C, and E). *P < 0.05, **P < 0.01, and ***P < 0.001.