RCOR2 Inhibition May Improve Cancer Treatment Outcomes

RCOR2 is upregulated primarily in tumor cells across human cancers. The proteomic analysis of human tissues revealed that RCOR2 expression was restricted to embryonic stem cells and a few human adult tissues, including colon, rectum, brain, and heart (Figure 1A). Intriguingly, we found widespread upregulation of RCOR2 mRNA in various types of human cancers in The Cancer Genome Atlas (TCGA) cohort (Figure 1B), predominantly expressed in malignant tumor cells (Figure 1C). RCOR2 protein upregulation was confirmed in murine MMTV-PyMT mammary tumors and human triple-negative breast cancer (Figure 1, D–F). Kaplan-Meier analysis of the TCGA cohort revealed that high levels of RCOR2 were significantly associated with worse disease-free interval and progression-free interval in breast cancer patients (Figure 1, G and H). These findings indicate that RCOR2 expression is awakened in tumors and may play a critical role in cancer development.

RCOR2 is upregulated in cancer cells and predicts poor survival in breast cancer patients. (A) Mass spectrometry analysis of RCOR2 protein levels in human tissues. Data were retrieved from ProteomicsDB. (B) mRNA expression analysis of RCOR2 across various types of human tumors and normal tissues from TCGA. P values were calculated by unequal-variance t test. Data were retrieved from UALCAN. N/A, not applicable; NS, not significant. (C) Single-cell RNA-Seq analysis of RCOR2 in tumors. Data were retrieved from TISCH2. (D) Immunoblot analysis of RCOR2 and actin proteins in normal mammary gland and MMTV-PyMT mammary tumors from mice. (E and F) Representative RCOR2 IHC in human triple-negative breast tumors and adjacent benign tissues (E); staining is quantified with H-score (F). *P < 0.05 by paired 2-tailed Student’s t test. Scale bars: 50 μm. (G and H) Kaplan-Meier survival analysis for patients with breast cancer by log-rank test. Patients were divided by median expression levels of RCOR2 mRNA. Data were retrieved from TCGA. iBAQ, intensity-based absolute quantification.

Tumoral RCOR2 inhibits cytotoxic T cell infiltration to promote tumor growth in mice. To determine a role of RCOR2 in tumor progression, we crossed Rcor2-floxed mice with K14-Cre and MMTV-PyMT transgenic mice and monitored mammary tumor growth in mice over 5 months. RCOR2 protein was depleted by K14-Cre in PyMT tumors harvested from homozygous Rcor2-floxed mice but not wild-type and heterozygous mice (Figure 2A). Homozygous deletion of Rcor2 significantly inhibited PyMT mammary tumor growth in mice (Figure 2B). To validate the results from the genetically modified mammary tumor mouse model, we conducted allograft experiments by implanting parental and RCOR2-knockout (KO) murine tumor cells into the mammary fat pad of female BALB/c mice or the flank of male C57BL/6J mice. RCOR2 KO significantly inhibited growth of MC38 colorectal tumors and TUBO mammary tumors in the syngeneic mouse models (Figure 2, C and D, and Supplemental Figure 1, A and B; supplemental material available online with this article; https://doi.org/10.1172/JCI188801DS1). To our surprise, the inhibitory role of RCOR2 KO in murine tumor growth was abolished in immunodeficient NSG mice (Figure 2E and Supplemental Figure 1C). We further confirmed that RCOR2 KO1 or KO2 did not inhibit human tumor growth in NSG mice orthotopically implanted with 2 million human MDA-MB-231 breast cancer cells (Supplemental Figure 1, D–H). Consistently, RCOR2 KO1 or KO2 had no effect on breast cancer cell proliferation and colony growth in vitro (Supplemental Figure 1, I–K). These results indicate that RCOR2 promotes tumor growth in a manner that relies on the host’s immune system.

RCOR2 promotes tumor immune evasion in mice. (A and B) Immunoblot (A) and weight (B) of mammary tumors in MMTV-PyMT+/– K14-Cre+/– Rcor2+/+, MMTV-PyMT+/– K14-Cre+/– Rcor2+/fl, and MMTV-PyMT+/– K14-Cre+/– Rcor2fl/fl mice. (C) Immunoblot analysis of RCOR2 protein in parental and RCOR2-KO1 or -KO2 MC38 cells. (D and E) Growth of parental and RCOR2-KO1 or -KO2 MC38 tumors in C57BL/6J (D) and NSG (E) mice. (F) Flow cytometry analysis of CD8+ T cells (CD45+CD3e+CD8+), CD4+ T cells (CD45+CD3e+CD4+), B cells (CD45+B220+), regulatory T cells (CD45+CD3e+CD4+FOXP3+), myeloid-derived suppressor cells (MDSCs; CD45+CD11b+Gr-1+), macrophages (CD45+CD11b+F4/80+), and dendritic cells (CD45+CD11c+F4/80–) in parental and RCOR2-KO1 MC38 tumors (n = 5). (G and H) CD8 and CD4 IHC in MMTV-PyMT+/– K14-Cre+/– Rcor2+/+ and MMTV-PyMT+/– K14-Cre+/– Rcor2fl/fl tumors (G); the percentage of T cells is quantified (H) (n = 4). Scale bars: 25 μm. (I) Growth of parental and RCOR2-KO1 MC38 tumors in C57BL/6J mice treated with IgG or anti-CD4 and anti-CD8 neutralizing antibodies. (J) Growth of parental and RCOR2-KO1 MC38 tumors in C57BL/6J and CD4-KO mice. Data represent mean ± SEM. P values were determined by 1-way ANOVA with Tukey’s test (B), 2-way ANOVA with Tukey’s test (I and J) or Dunnett’s test (D and E), and 2-tailed Student’s t test (F and H). *P < 0.05; **P < 0.01; ****P < 0.0001.

We next conducted immune cell profiling by flow cytometry to comprehensively assess the effect of tumoral RCOR2 on the composition and abundance of immune cell subsets within tumors (Supplemental Figure 2, A and B). The percentage of intratumoral lymphocytes including CD4+ T cells, CD8+ T cells, and B cells was significantly increased in RCOR2-KO1 MC38 tumors as compared with their control tumors (Figure 2F). A similar effect on CD4+ and CD8+ T cell infiltration was observed in RCOR2-KO1 TUBO tumors (Supplemental Figure 2C). Immunohistochemistry (IHC) analysis further confirmed increased infiltration of CD4+ and CD8+ T cells in RCOR2-KO PyMT mammary tumors compared with wild-type tumors (Figure 2, G and H). In contrast, RCOR2 KO1 had no effect on infiltration of regulatory T cells and myeloid cells including myeloid-derived suppressor cells, macrophages, and dendritic cells in MC38 and TUBO tumors (Figure 2F and Supplemental Figure 2C). These results indicate that tumoral RCOR2 shapes the lymphocyte landscape in the tumor microenvironment.

To determine whether loss of CD4+ and/or CD8+ T cells is necessary for RCOR2-mediated tumor growth, we administered anti-CD4 antibody, anti-CD8 antibody, anti-CD4/CD8 antibodies or control antibody isotype intraperitoneally into tumor-bearing mice to deplete CD4+ and CD8+ T cells. Depletion of CD4+ T cells, CD8+ T cells, or both with anti-CD4/CD8 neutralizing antibodies effectively restored RCOR2-KO1 tumors in the murine TUBO mammary tumor model (Supplemental Figure 2D). More robust rescue of RCOR2-KO1 tumors was observed in the MC38 tumor mouse model when mice were cotreated with anti-CD4/CD8 antibodies (Figure 2I). Notably, genetic deletion of CD4+ T cells greatly promoted MC38 tumor growth and abolished tumor reduction conferred by RCOR2 KO in mice (Figure 2J), supporting an inhibitory role of CD4+ T cells in RCOR2-induced tumor growth, either directly or indirectly through their regulation of other immune components. Collectively, these results indicate that RCOR2 promotes tumor growth through reducing infiltration of CD4+ and CD8+ T cells.

RCOR2 increases intrinsic cancer cell plasticity to promote tumor development in mice. Homozygous deletion of Rcor2 significantly decreased incidence and numbers of murine PyMT tumors in mice (Figure 3, A and B), indicating that RCOR2 promotes tumor initiation. To determine whether RCOR2 controls cancer cell plasticity leading to tumor initiation, we isolated aldehyde dehydrogenase–high (ALDHhi) breast cancer stem cells (BCSCs; Lin–CD90–ALDHhi) from PyMT mammary tumors by flow cytometry and found elevated RCOR2 protein in this cell population compared with Lin–CD90–ALDHlo non-BCSCs (Figure 3C). Loss of RCOR2 blocked PyMT tumorsphere formation ex vivo and reduced ALDHhi BCSCs in PyMT tumors in vivo (Figure 3, D–G).

RCOR2 enhances tumor cell plasticity to promote cancer development. (A and B) Tumor-free period (A) and mammary tumor number (B) of MMTV-PyMT+/– K14-Cre+/– Rcor2+/+, MMTV-PyMT+/– K14-Cre+/– Rcor2+/fl, and MMTV-PyMT+/– K14-Cre+/– Rcor2fl/fl mice. (C) Immunoblot analysis of RCOR2 protein in non-tumor-initiating cells (Lin–CD90–ALDHlo) and tumor-initiating cells (Lin–CD90–ALDHhi) isolated from MMTV-PyMT tumors. (D and E) Tumorsphere formation assay of MMTV-PyMT+/– K14-Cre+/– Rcor2+/+, MMTV-PyMT+/– K14-Cre+/– Rcor2+/fl, and MMTV-PyMT+/– K14-Cre+/– Rcor2fl/fl tumors. Representative tumorsphere images are shown in D. Tumorsphere numbers are quantified in E (n = 5). (F and G) Flow cytometry analysis (F) and quantification (G) of tumor-initiation cells in MMTV-PyMT+/– K14-Cre+/– Rcor2+/+, MMTV-PyMT+/– K14-Cre+/– Rcor2+/fl, and MMTV-PyMT+/– K14-Cre+/– Rcor2fl/fl tumors. Representative gating is shown in F. ALDHhi cells are quantified in G. (H) Immunoblot analysis of RCOR2 protein in parental, RCOR2-KO1, and RCOR2-rescue MDA-MB-231 cells. (I and J) Mammosphere formation assay of parental, RCOR2-KO1, and RCOR2-rescue MDA-MB-231 cells. Representative mammosphere images are shown in I. Mammosphere numbers are quantified in J (n = 3). (K) Limiting dilution assay of parental and RCOR2-KO1 or -KO2 MDA-MB-231 cells in NSG mice. (L) Growth of parental and RCOR2-KO1 or -KO2 MDA-MB-231 tumors in NSG mice. (M and N) Tumorsphere formation assay in parental and RCOR2-KO1 or -KO2 MDA-MB-231 tumors. Representative tumorsphere images are shown in M. Tumorsphere numbers are quantified in N (n = 5). (O and P) Aldefluor assay (STEMCELL Technologies) in parental and RCOR2-KO1 or -KO2 MDA-MB-231 tumors. Representative flow cytometry gating is shown in O. ALDHhi cells are quantified in P (n = 5). Data represent mean ± SEM. P values were determined by 1-way ANOVA with Tukey’s test (B, E, G, and J) or Dunnett’s test (N and P), 2-way ANOVA with Dunnett’s test (L), log-rank (Mantel-Cox) test (A), and χ2 test (K). *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. Scale bars: 100 μm (D, I, and M).

To validate the role of RCOR2 in cancer cell plasticity observed in a murine mammary tumor model, we generated BCSC-enriched mammospheres from human breast cancer cells. In line with murine tumors, RCOR2 protein levels were remarkably increased in MDA-MB-231 and MCF-7 mammospheres compared with their monolayers with scarcely detectable BCSCs (Supplemental Figure 3, A and B). Forced expression of RCOR2 significantly increased formation of MDA-MB-231 mammospheres (Supplemental Figure 3, C–E). In contrast, RCOR2 KO1 or KO2 decreased the number of MDA-MB-231 and MCF-7 mammospheres (Supplemental Figure 3, F–J). The RCOR2 loss-of-function effect was specific, as re-expression of RCOR2 could partially restore formation of RCOR2-KO1 mammospheres (Figure 3, H–J). We further showed that RCOR2 KO1 or KO2 significantly decreased the proportion of ALDHhi BCSCs in MDA-MB-231 and MCF-7 cells as well as in MDA-MB-231 mammospheres (Supplemental Figure 3, K–P). CD44+CD24–EpCAM+ BCSC populations were also decreased by RCOR2 loss in MDA-MB-231 and MCF-7 cells (Supplemental Figure 3, Q and R). Collectively, these results indicate that RCOR2 is strongly expressed in ALDHhi BCSCs and is sufficient and necessary for cancer cell plasticity.

To determine whether RCOR2 controls tumor cell plasticity to promote tumor development, we performed limiting dilution assay in NSG mice. Parental and RCOR2-KO1 or -KO2 MDA-MB-231 cells with 3 cell numbers of 40, 200, and 1,000 were orthotopically implanted into the mammary fat pad of female NSG mice. RCOR2 KO1 or KO2 significantly decreased the tumor incidence in NSG mice (Figure 3K). Similar results were observed in the MCF-7 xenograft mouse models (Supplemental Figure 3S). Notably, RCOR2 KO significantly inhibited MDA-MB-231 tumor growth in NSG mice when a limited number of cancer cells were implanted (Figure 3L). We confirmed reduced ALDHhi BCSCs within tumors and ex vivo tumorsphere formation by RCOR2 KO1 or KO2 (Figure 3, M–P). Collectively, these results indicate that RCOR2 enhances cancer cell plasticity to promote tumor development.

RCOR2 activates Wnt/β-catenin signaling but suppresses CIITA/MHC-II signaling in cancer cells through two distinct epigenetic programs. To determine the mechanism by which tumoral RCOR2 promotes tumor cell plasticity and immune evasion, we assessed RCOR2 transcriptome in MDA-MB-231 cells by RNA sequencing (RNA-Seq). Four hundred eighty-five genes were induced whereas 289 genes were repressed by RCOR2 (FDR < 0.05; log counts per million > 0; |fold change| > 1.5; Figure 4, A–D). Reactome pathway analysis of these differentially expressed genes revealed that activation of the Wnt/β-catenin signaling pathway and inhibition of the interferon signaling pathway were shared in both RCOR2-KO1 and -KO2 cells (Figure 4, E and F). Reverse transcription–quantitative polymerase chain reaction (RT-qPCR) assay confirmed repression of two Wnt ligands, WNT5A and WNT10B, and induction of two negative regulators of the Wnt/β-catenin pathway, RNF43 and CXXC4, in RCOR2-KO1 and -KO2 MDA-MB-231 cells (Supplemental Figure 4A). However, re-expression of RCOR2 caused derepression of RNF43 only in RCOR2-KO cells (Figure 4G). RNF43 protein levels were also increased by RCOR2 loss in MDA-MB-231 cells and PyMT tumors (Figure 4H and Supplemental Figure 4, B and C). These results indicate that RCOR2 represses RNF43 expression in cancer cells.

RCOR2 activates Wnt/β-catenin by repressing RNF43 and inhibits immune response by repressing CIITA and MHC-II. (A and B) Volcano plots of RCOR2 target genes in MDA-MB-231 cells (n = 2). (C and D) Venn diagrams of RCOR2 activated (C) and repressed (D) gene numbers in MDA-MB-231 cells (n = 2). (E and F) Reactome pathway analysis of RCOR2 target genes in MDA-MB-231 cells (n = 2). (G) RT-qPCR analysis of indicated mRNAs in parental, RCOR2-KO, and RCOR2-rescue MDA-MB-231 cells (n = 3). (H) Flow cytometry analysis of RNF43 protein in parental and RCOR2-KO1 or -KO2 MDA-MB-231 cells. (I) RT-qPCR analysis of indicated mRNAs in parental and RCOR2-KO1 or -KO2 MDA-MB-231 cells treated with 0.1 ng/mL IFN-γ for 24 hours (n = 3). (J) Immunoblot analysis of indicated proteins in parental and RCOR2-KO1 or -KO2 MDA-MB-231 cells treated with 0.1 ng/mL IFN-γ for 24 hours. (K) Immunoblot analysis of indicated proteins in parental, RCOR2-KO1, CIITA-KO, and RCOR2/CIITA–DKO TUBO cells treated with 1 ng/mL IFN-γ for 24 hours. (L) Immunoblot analysis of indicated proteins in parental, RCOR2-KO1, CIITA-KO, and RCOR2/CIITA–DKO MC38 cells treated with 5 ng/mL IFN-γ for 24 hours. (M) Representative immunostaining of I-A/I-E in parental, RCOR2-KO1, CIITA-KO, and RCOR2/CIITA–DKO MC38 cells treated with 5 ng/mL IFN-γ for 24 hours. Scale bars: 10 μm. Data represent mean ± SEM. P values were determined by 1-way ANOVA with Tukey’s test (G) or Dunnett’s test (I). *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

By searching differentially expressed genes involved in interferon signaling pathways from our RNA-Seq dataset (Figure 4, A–D), we found that a family of MHC-II heavy chain genes and their transcriptional coactivator CIITA were repressed by RCOR2, which was validated in multiple cancer cell lines by RT-qPCR assay (Figure 4I and Supplemental Figure 4, D and E). Protein levels of CIITA and MHC-II molecules were also elevated in RCOR2-depleted cancer cells following IFN-γ treatment and in PyMT tumors, as shown by immunoblot, flow cytometry, and/or immunostaining assays (Figure 4J and Supplemental Figure 4, F–J). CIITA KO counteracted RCOR2 KO1–induced MHC-II molecules in TUBO and MC38 cells (Figure 4, K–M), suggesting that RCOR2 indirectly reduces MHC-II expression in tumor cells by repressing CIITA. We further found that RCOR2 had no effect on MHC-I expression in cancer cells (Figure 4, A and B, and Supplemental Figure 4E). Together, these findings indicate that RCOR2 specifically induces MHC-II silencing in cancer cells through suppression of CIITA.

Two types of histone modifiers, LSD1 and HDAC1/2, are associated with RCOR2 in the complex (36). Treatment with an LSD1 inhibitor, GSK-LSD1 (50 μM), significantly induced the expression of RNF43, but not CIITA and MHC-II heavy chain genes, whereas an HDAC inhibitor, Trichostatin A (TSA) (0.2 μM), had an opposite effect on the expression of these genes in MDA-MB-231 cells (Figure 5, A and B). These results were confirmed by genetic KO of LSD1, HDAC1, or HDAC2 in MDA-MB-231 cells treated with or without 0.1 ng/mL IFN-γ (Figure 5, C–F). We further found that GSK-LSD1 treatment blocked RCOR2-induced RNF43 repression in MDA-MB-231 cells (Figure 5G), whereas TSA treatment caused CIITA derepression in MDA-MB-231 cells overexpressing RCOR2 (Figure 5H). These findings indicate that RCOR2 suppresses RNF43 and CIITA through LSD1 and HDAC1/2, respectively.

RCOR2 inhibits RNF43 and CIITA expression via distinct epigenetic mechanisms. (A and B) RT-qPCR analysis of indicated mRNAs in MDA-MB-231 cells treated with 50 μM GSK-LSD1 (A) or 0.2 μM TSA (B) for 48 hours (n = 3). (C) RT-qPCR analysis of indicated mRNAs in parental and LSD1-KO MDA-MB-231 cells treated with or without 0.1 ng/mL IFN-γ for 24 hours (n = 3). (D) Immunoblot analysis of indicated proteins in parental and LSD1-KO MDA-MB-231 cells. (E) Immunoblot analysis of indicated proteins in parental, HDAC1-KO, and HDAC2-KO MDA-MB-231 cells. (F) RT-qPCR analysis of indicated mRNAs in parental, HDAC1-KO, and HDAC2-KO MDA-MB-231 cells treated with or without 0.1 ng/mL IFN-γ for 24 hours (n = 3). (G) RT-qPCR analysis of indicated mRNAs in MDA-MB-231 cells overexpressing empty vector (EV) or RCOR2 treated with DMSO or 50 μM GSK-LSD1 for 48 hours (n = 3). OE, overexpression. (H) RT-qPCR analysis of indicated mRNAs in MDA-MB-231 cells overexpressing EV or RCOR2 treated with DMSO or 0.2 μM TSA for 24 hours and in combination with 0.1 ng/mL IFN-γ for another 24 hours (n = 3). (I and J) Genome browser snapshots of HA, HDAC1, and LSD1 binding peaks, highlighted in gold-yellow, on RNF43 (I) and CIITA (J) in control, RCOR2-OE, and RCOR2-KO MDA-MB-231 cells (n = 2). (K and L) ChIP-qPCR assay showing relative H3K4me2 (K) and H4K16Ac (L) occupancy on RNF43 and CIITA in parental and RCOR2-KO MDA-MB-231 cells. Data represent mean ± SEM. P values were determined by 1-way ANOVA with Tukey’s test (G and H) or Dunnett’s test (C and F), 2-way ANOVA with Tukey’s test (K and L), and 2-tailed Student’s t test (A and B). *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

To support epigenetic regulation of RNF43 and CIITA by the RCOR2 complex, we next performed ChIP sequencing (ChIP-Seq) in MDA-MB-231 cells overexpressing HA-RCOR2 and detected 2 strong RCOR2 binding peaks at the second intron of the RNF43 gene and 3 strong RCOR2 binding peaks at the promoter of the CIITA gene (Figure 5, I and J). RCOR2 occupancies were detected at the genome nearest to HLA-DMA and HLA-DMB, but not other MHC-II heavy chain genes (Supplemental Figure 5, A–C), further supporting indirect repression of MHC-II by RCOR2. Consistently, RCOR2 KO1 selectively increased H3K4me2 enrichment on RNF43 intron 2 and H4K16ac enrichment on the CIITA promoter in MDA-MB-231 cells (Figure 5, K and L). Both LSD1 and HDAC1 were colocalized with RCOR2 at RNF43 and CIITA, but their enrichment was not affected by RCOR2 KO1 in MDA-MB-231 cells (Figure 5, I and J), suggesting that RCOR2 is not involved in recruitment of LSD1 and HDAC1 to RNF43 and CIITA genes and that the enzymatic activity of LSD1 and HDAC1/2 is selectively stimulated on RNF43 and CIITA. Together, these findings indicate that RCOR2 reduces H3K4me2 and H4K16ac to suppress the expression of RNF43 and CIITA, respectively.

Tumoral CIITA/MHC-II silencing is responsible for RCOR2-induced tumor immune evasion. Next, we studied whether CIITA silencing regulates RCOR2-induced tumor immune evasion. Parental, RCOR2-KO1, CIITA-KO, and RCOR2/CIITA–double-KO (DKO) TUBO cells were orthotopically implanted into the mammary fat pad of female BALB/c mice. CIITA KO reversed tumor reduction conferred by RCOR2 loss in mice, even though CIITA KO alone had no effect on tumor growth (Figure 6A). Similar results were observed in the MC38 tumor mouse model (Figure 6B). Increased infiltration of CD4+ and CD8+ T cells was also reversed in tumors when CIITA was co-deleted with RCOR2 (Figure 6, C–E). These results indicate that CIITA silencing is responsible for RCOR2-induced T cell evasion and tumor growth in syngeneic mouse models.

RCOR2 promotes tumor immune evasion by suppressing CIITA and MHC-II. (A and B) Growth of parental, RCOR2-KO1, CIITA-KO, and RCOR2/CIITA–DKO TUBO (A) or MC38 (B) tumors in BALB/c or C57BL/6J mice. (C–E) CD4 and CD8 IHC in parental, RCOR2-KO1, CIITA-KO, and RCOR2/CIITA–DKO MC38 tumors (C); the percentage of T cells is quantified (D and E; n = 5). Scale bars: 25 μm. (F) Scheme of MHC-II KO using CRISPR/Cas9. (G) Genotyping of MHC-II KO in parental, RCOR2-KO1, MHCII-KO, and RCOR2/MHCII–DKO MC38 cells. (H) Representative immunostaining of I-A/I-E in parental, RCOR2-KO1, MHCII-KO, and RCOR2/MHCII–DKO MC38 cells treated with 5 ng/mL IFN-γ for 24 hours. Scale bars: 10 μm. (I) Growth of parental, RCOR2-KO1, MHCII-KO, and RCOR2/MHCII–DKO MC38 tumors in C57BL/6J mice. (J–L) CD4 and CD8 IHC in parental, RCOR2-KO1, and RCOR2/MHCII–DKO MC38 tumors (J); the percentage of T cells is quantified (K and L; n = 5). Scale bars: 25 μm. Data represent mean ± SEM. P values were determined by 1-way ANOVA with Tukey’s test (D, E, K, and L) and 2-way ANOVA with Tukey’s test (A, B, and I). **P < 0.01; ***P < 0.001; ****P < 0.0001.

To further determine whether loss of MHC-II–mediated antigen presentation controls RCOR2-mediated immune escape, we deleted all five of the classic mouse MHC-II heavy chain genes in parental and RCOR2-KO1 MC38 tumor cells using the CRISPR/Cas9 technique (Figure 6F). A genotyping test showed that all five MHC-II heavy chain genes were deleted from one allele in both parental and RCOR2-KO1 MC38 cells (Figure 6G), which was sufficient to deplete their proteins (Figure 6H). MHC-II protein depletion completely abolished tumor reduction caused by RCOR2 loss in mice (Figure 6I), which phenocopied CIITA loss (Figure 6, A and B). Increased infiltration of CD4+ and CD8+ T cells was also reversed in RCOR2/MHC-II–DKO tumors (Figure 6, J–L). These results indicate that MHC-II silencing is responsible for RCOR2-induced T cell evasion and tumor growth in mice.

To determine whether RCOR2 impairs cytotoxicity of CD4+ T cells through CIITA/MHC-II silencing, we performed CD4+ T cell killing assay by coculturing CD4+ T cells isolated from OT-II mouse spleen with parental or RCOR2-KO1 MC38 cells pretreated with the OVA323-39 peptide at the ratio of 10:1. The number of dead RCOR2-KO1 MC38 cells, which are shown in yellow, was significantly increased after coculture with CD4+ T cells as compared with parental MC38 cells, which was prevented by loss of CIITA or MHC-II (Figure 7, A and B). Under conditions of coculture with RCOR2-KO1 MC38 cells, CD4+ T cells expressed higher mRNA levels of cytotoxic cytokines, including IFN-γ and TNF-α and the T cell fate activator IL-2, than those in the other 3 coculture groups (Figure 7C). Consistently, we showed that loss of tumoral RCOR2 significantly increased granzyme B–expressing (GzmB-expressing) CD4+ and CD8+ T cells in MC38 tumors, which was reversed by co-deletion of CIITA or MHC-II (Figure 7, D–G), suggesting that tumoral RCOR2 impedes activation of cytotoxic CD4+ and CD8+ T cells in tumors through CIITA/MHC-II silencing. Collectively, these findings indicate that RCOR2 downregulates MHC-II–mediated antigen presentation in cancer cells, leading to tumor escape from T cell immunosurveillance.