Mi-2β-targeted inhibition induces immunotherapy response in melanoma

1Department of Cancer Immunology and Virology, Dana Farber Cancer Institute, Harvard Medical School, Boston, MA 02115, USA 2Department of Pharmaceutical Sciences, College of Pharmacy, University of Arkansas for Medical Science, Little Rock, AR 72205, USA 3Key Laboratory of Bioorganic Phosphorus Chemistry and Chemical Biology, Beijing Advanced Innovation Center for Human Brain Protection, School of Pharmaceutical Sciences, Tsinghua University, Beijing 100084, China 4Adobe Inc., San Jose, CA 95110, USA 5Department of Dermatology, Yueyang Hospital of Integrated Traditional Chinese and Western Medicine, Shanghai University of Traditional Chinese Medicine, Shanghai 200437, China 6Department of Biomedical Research, National Jewish Health, Denver, CO 80206, USA 71 Affiliated Hospital of Integrated Traditional Chinese and Western Medicine, School of Pharmacy, Nanjing University of Chinese Medicine, Nanjing 210023, China 8National Research Center for Translational Medicine, Shanghai State Key Laboratory of Medical Genomics, Rui-Jin Hospital, Shanghai JiaoTong University School of Medicine, 197 Ruijin second Rd,Shanghai 200025, China 9Ludwig Institute for Cancer Research, University of Oxford, Headington, Oxford OX3 7DQ, UK 10Skin Disease Institute, Department of Dermatology, The 2nd Hospital, School of Medicine, Zhejiang University, 88 Jiefang Road, Hangzhou, Zhejiang 310009, China

. For example, chromatin remodeling PBAF was found to contribute to cancer cell immune resistance 21,22 . BRG1, a chromatin-remodeling enzyme, has also been implicated in enhancing IFN-stimulated gene transcription 23 . The overexpression of PRC2, a multiprotein enzyme complex (EZH2, SUz12, EED) regulating the trimethylation of lysine 27 on histone H3 (H3K27me3) 24 is detected in cancer cells and mediates the repression of IFN-γ-stimulated genes. Moreover, EZH2 inhibition enhances T cell-targeting immunotherapies in mouse models of melanoma 25,26 . Interestingly, ARID1A, a member of the SWI/SNF family is found to interact with EZH2 to inhibit IFN-responsiveness gene expression in cancer cells, whose mutations can shape cancer immune phenotype and immunotherapy 27 Here we find that Mi-2β played a key role in regulating adaptive immune response in melanoma. The human Mi-2β protein was discovered as autoantigens in dermatomyositis in 1995 28 . Mi-2β, also named as CHD4 (chromodomain helicase DNA-binding protein 4), is a CHD family remodeling enzyme in the NuRD complex, which include the histone deacetylases 1 and 2 (HDAC1 and HDAC2), RBBP4/RBBP7,  37,38 . We find that Mi-2β silencing induced the immune response of anti-PD-1 antibody treatment in "cold" melanoma in vitro and in vivo, and the effects were directly mediated by factors in IFN-γ signaling, such as Irf1, Cxcl9 and Cxcl10. Moreover, we developed a specific Mi-2β-targeted inhibitor Z36-MP5. Treatment of Z36-MP5 induced response of immune checkpoint blockades in "cold" melanoma in vivo. Our work provides new insights into the epigenetic regulation in adaptive immune response in melanoma and developed a new immune therapeutic strategy in melanoma.

Tumor intrinsic expression of Mi-2β modulates resistance to T cell-mediated killing in melanoma.
Cytotoxic T cells are key effectors to detect and eliminate transformed tumor cells. However, some tumors are lacking T cells infiltration (cold tumor) and subsequently adaptive immune response, including response of the treatment of immune checkpoint blockades 39 . More and more evidence indicate that tumor-intrinsic chromatin regulatory factors are crucial in regulating adaptive immune response in melanoma 15 . We therefore focused on identifying tumor-intrinsic epigenetic factors which are crucial in regulating adaptive immune response in melanoma. To preliminarily identify the key epigenetic factors that regulate cell sensitivity and resistance to T cell-mediated attack in melanoma, we analyzed the hazard ratio of the different epigenetic factors in melanoma with different levels of T cell infiltrations. Tumorintrinsic CD8 levels served as a marker to indicate T cell infiltration 40 . Epigenetic factors were preliminarily recognized as a potential regulator of immune response if its expression level was significantly correlated with hazard ratio in patients with high CD8 T cell infiltration only, but not in patients with low CD8 T cell infiltration. Fifty-five epigenetic factors were identified (Extended Data   Fig. 1a). The melanoma and T cell co-culture system was used to further identify the role of the most correlated genes (n=18) identified in the hazard ratio analysis in regulating T cell mediated cytotoxicity.
In this co-culture system, B16F10 melanoma cells and the activated Pmel-1 T cells were co-cultured.
Pmel-1 T cells carry a rearranged T cell receptor transgene specific for the mouse homologue of human pre-melanosome protein of gp100 41 , and B16F10 cells are resistant to immunotherapies, including checkpoint blockade antibodies against PD-1 42, 43 . Each candidate gene was silenced by specific gRNA and labeled by GFP in B16F10 cells. The resulted B16F10 cells were mixed with no labeled parent control B16F10 cells (1:1) and then co-cultured with the activated Pmel-1 cells. The number of GFP + cells were detected by flow cytometry to determine B16F10 cell response to cytotoxic T cells (Fig. 1a). Mi-2β, Eif4a1, USP7 or Parp1 silencing significantly induced the response to T cell attack in melanoma cells, and led more than half of melanoma cells to be eliminated by Pmel-1 T cell-mediated killing (Fig. 1b).
Mi-2β was picked for further analysis due to the epidermal inflammation phenotypes in conditional keratinocyte-specific Mi-2β knockout mouse 36 . Mi-2β is a chromatin remodeling enzyme with a SNF2like ATPase domain and plays critical roles in chromatin assembly and genomic stability. To validate the significance of Mi-2β in regulating immune microenvironment in human melanoma, the correlations between Mi-2β mRNA level and CD8A and CD8B mRNA levels were first analyzed in melanoma patients collected in The Cancer Genome Atlas (TCGA). Mi-2β mRNA level was negatively correlated with both CD8A and CD8B mRNA levels (p<0.01) (Fig. 1c). These results indicate that lower Mi-2β expression correlates with enrichment of CD8 T cell infiltration in melanoma. Next, to identify the role of Mi-2β in the immune response in melanoma, the correlations between Mi-2β and GZMB or PRF1 were analyzed. GZMB and PRF1 are crucial for the rapid induction of target cell apoptosis by cytotoxic T lymphocytes (CTL) in cell-mediated immune response 44 . Mi-2β mRNA level was also negatively correlated with both GZMB and PRF1 mRNA level (p<0.01) in melanoma (Fig. 1d). These results suggest that expression levels of Mi-2β are associated with T cell-mediated killing in melanoma. Consistently, the repression of Mi-2β expression were found to correlate with a substantial survival benefits only in melanoma patients with higher CD8 T cell infiltration (p<0.05), but not in melanoma with low CD8 T cell infiltration (Fig.   1e). To further validate the role of Mi-2β in modulating sensitivity to T cell-mediated killing in melanoma, the melanoma-T-cell co-culture system (B16F10/Pmel-1) was used. Mi-2β silencing (Extended Data Fig.   1b) induced T cell-mediated cytotoxicity in vitro (Fig. 1f). All these results demonstrate the critical role of Mi-2β in regulating melanoma resistance to T cell-meditated cytotoxicity. Tumor intrinsic Mi-2β level regulates melanoma sensitivity to the anti-tumor immunotherapy.
Mi-2β silencing synergizes with immune checkpoint blockades to promote anti-tumor immunity. To identify whether Mi-2β depletion induced immune response in B16F10 melanoma cells, mouse graft melanomas with shMi-2β virus-infected B16F10 cells were treated by anti-PD-1 antibodies (10mg/kg) at day 6,9,12,15 and 18 after tumor cell inoculation in immunocompetent C57BL/6 mice. In consistence with the previous reports 18,43 , mice injected with control B16F10 cells with shScramble were not sensitive to anti-PD-1 treatment. However, Mi-2β silencing combined with anti-PD-1 treatment conferred a substantial inhibition on the tumor growth in B16F10 melanoma ( Fig. 2a-b), and subsequently extended the survival of the treated mice (Fig. 2c). Analysis of graft tumor microenvironment by flow cytometry (Extended Data Fig. 2a) shows that increases of CD8 + and CD4 + T cell infiltration were detected in B16F10 tumor graft with Mi-2β silencing, which was strongly augmented by the anti-PD-1 treatment (Fig.   2d).
At the same time, a minor, but non-significant, increase in tumor-infiltrating Treg cells was also detected in the B16F10 tumor graft following Mi-2β silencing, which was not inhibited by anti-PD-1 treatment and/or Mi-2β silencing (Extended Data Fig. 2b). Moreover, a minor to medium increase of GZMB expression and upregulation of activation of CD69, IFN-γ, CD25 and CD107 were detected in tumor-infiltrating CD8 + T cells of B16F10 tumor graft with Mi-2β silencing, which were strongly augmented by anti-PD-1 treatment (Fig. 2e-f). All these data indicate that Mi-2β silencing sensitizes tumor cells and confers a more favorable tumor microenvironment to induce an adaptive immune response to immune checkpoint blockades treatment in melanoma.

Loss of Mi-2β induces responses to immune checkpoint blockades in BRaf V600E /Pten null melanoma
in vivo. To further examine whether Mi-2β depletion induced an adaptive immune response in melanoma in vivo, Tyr::CreER;BRaf CA ;Pten lox/lox mice were used for anti-PD-1 antibody experimental treatment. In this mouse strain, induction of Cre-mediated recombination leads to Braf V600E expression and Pten inactivation (BRaf V600E /Pten null ) in cutaneous melanocytes, which results in rapidly progression of malignant melanoma 45 . Mi-2β lox/lox mice 36 were crossed with Tyr::CreER;BRaf CA ;Pten lox/lox mice to deplete Mi-2β in BRaf V600E /Pten null melanoma after the tamoxifen injection. Mice with visible melanomas were randomly treated with either control IgG antibodies (10mg/kg) or anti-PD-1 (10mg/kg) starting at day 9, 12, 15, 18 and 21 after Cre activation (Fig. 3a). The tumor free survival was analyzed. In consistence with the previous reports 20 , BRaf V600E /Pten null melanoma is a kind of "cold" tumor and was insensitive to anti-PD-1 antibody treatment, and there was no significant difference of mouse free survival observed in BRaf V600E /Pten null melanoma with different Mi-2β status (Fig. 3b). IHC staining for the melanoma marker S100 and proliferation marker Ki-67 showed no difference between BRaf V600E /Pten null melanomas with different Mi-2β status (Extended Data Fig. 3a-b). Intriguingly, treatment of anti-PD-1 significantly extended the mouse survival with BRaf V600E /Pten null /Mi-2β null melanoma compared with that of BRaf V600E /Pten null melanoma (Fig. 3b). To further identify whether Mi-2β knockout-induced anti-PD-1 response correlates with T cell activation, tumor-infiltrating lymphocytes (TILs) were measured in BRaf V600E /Pten null melanomas with different Mi-2β status by flow cytometry. The populations of infiltrating CD8 + and CD4 + T cells were minorly increased in TILs of BRaf V600E /Pten null /Mi-2β null melanoma. This increase was significantly augmented by the anti-PD-1 treatment (Fig. 3c-d). At the same time, a minor, but not-significant, increase in the Treg population was also detected in BRaf V600E /Pten null melanoma after Mi-2β knockout. However, the anti-PD-1 treatment did not change Treg cell population in BRaf V600E /Pten null melanomas significantly after Mi-2β knockout (Extended Data Fig. 3c). Moreover, an increase of GZMB expression and upregulation of CD8 + T cell activation markers, such as CD69, IFNγ, CD25 and CD107, were detected in BRaf V600E /Pten null /Mi-2β null melanomas after Mi-2β knockout.
These increase were all further strongly augmented by anti-PD-1 treatment (Fig. 3e-f). Taken together, all these results indicate that loss of Mi-2β in melanocytes activates CTLs to induce response of anti-PD-1 treatment in "cold" melanoma in vivo.

Mi-2β-regulated immune response is mainly mediated by IFN-γ signaling pathways. To identify how
Mi-2β regulated immune response is mediated in melanoma, Mi-2β-CRISPR/Cas9-knocked and IFN-γtreated B16F10 cells 46 were used to perform microarray assay. The expressions of 1209 genes were identified to be significantly repressed (>1.5 folds, p<0.05), and the expressions of 1283 genes were identified to be significantly up-regulated (>1.5 folds, p<0.05) after Mi-2β silencing. The deregulated genes identified were further analyzed by Gene Set Enrichment Analysis (GSEA) to identify Mi-2βregulated gene sets and pathways. Interestingly, IFN-γ signal was activated after Mi-2β knockout ( Fig. 4a and Extended Data Table 1-2). IFN-γ production is essential in the response to immunotherapy, especially in patients with melanoma 47,48 . Many of Mi-2β-controlled IFN-γ-responsive genes, such as Cxcl9, Cxcl10, CD74, Irf1, and CD40, functions in T cell chemoattractant, antigen presentation, and T cell targeting and activation (Fig. 4b). Specifically, cytokine expressions, such as Cxcl9, Cxcl10, Cxcl11 and Ccl5 were upregulated by Mi-2β silencing (Fig. 4b), and these cytokines are crucial in inducing and recruiting effector T cells with CXCR3 chemokine receptor into tumor microenvironment to induce antitumor immunity 47,48,49 . Several antigen presentation genes, such as Tap1 and CD74 and some regulators involving in tumor cell immunogenicity, such as Irf1, Icam1 and CD40 were also upregulated by Mi-2β knockout in vitro (Fig. 4b).
To confirm the regulation of Mi-2β on those downstream targets in IFN-γ pathways, the expressions of ISGs in IFN-γ pathway were measured in B16F10 cells with Mi-2β silencing. Mi-2β silencing significantly upregulated the mRNA expressions of Cxcl9, Cxcl10, Cxcl11, Ccl5, Tap1, CD74, Irf1, Icam1, CD40, Fas and PD-L1 (Fig. 4c), and enhanced the paracrine secretions of Cxcl9 and Cxcl10 both before and after addition of IFN-γ ( Fig. 4d-e). In vivo, TIMER analysis 50 indicated that Mi-2β mRNA level negatively correlated with CCL5, CD74 and CD40 mRNA level in melanoma patients collected in TCGA melanoma cohort (p<0.01) (Extended Data Fig. 4a). These data indicate that Mi-2β-regulated immune response is mediated by IFN-γ signaling pathways in melanoma. To identify how Mi-2β involves in the responses of anti-PD-1 treatment, the expression levels of Cxcl9 and Cxcl10 were measured with ELISA assay in melanomas collected in Fig. 2C-D. Upregulation of Cxcl9 and Cxcl10 were detected after Mi-2β silencing and the anti-PD-1 treatment in melanomas ( Fig. 4f-g). In addition, we also measured these factors in the downstream targets of IFN-γ pathways in BRaf V600E /Pten null melanoma collected in Fig. 3B.
Mi-2β, a member of the SNF2/RAD54 helicase family, is a main component of the nucleosome remodeling and deacetylase complex. Mi-2β are highly enhanced at the transcription starting sites and plays an important role in the epigenetic transcriptional repression 51 . To investigate the molecular mechanisms underlying Mi-2β-regulated repression of factors in IFN-γ pathways, chromatin immunoprecipitation (ChIP) assays were performed to identify whether Mi-2β protein binds to the promoters of Cxcl9, Cxcl10 and Irf1. We found that Mi-2β bound to promoters of Cxcl9, Cxcl10 and Irf1, with anti-Stat1 served as a positive control (Extended Data Fig. 4c-e). These data indicate that Mi-2β is directly involved in regulating transcription of Irf1, Cxcl9 and Cxcl10.
Development of Z36-MP5 as a specific small molecule inhibitor targeting Mi-2β. Given the pivotal role of Mi-2β in regulating immune response, targeting Mi-2β would represent a potential therapeutic strategy in melanoma immunotherapy, especially in combination with immune checkpoint blockades. To screen small molecules that inhibit Mi-2β activity, Homology Modeling was carried out using Structure Prediction Wizard in Prime 52,53 . Mi-2β belongs to the CHD family of chromatin remodelers, which share the highly conserved ATPase/helicase domains 54,55 . The Homology Model of Mi-2β was generated using the yeast CHD1 structure (PDB code: 3MWY) as template and the receptor sequence was obtained from Uniprot 56 , which clearly depicted the interaction of Mi-2β binding pocket and ATP (Extended Data Fig.   5a). Virtual screening was done with enzyme hinge region ligands database and nucleoside mimetic database from Enamine. All ligands of ~23,010 compounds were docked to the ATP binding site using SP docking and post-processed with Prime MM-GBSA. The ligands with methyldihydroimidazopyridinone structure were predicted to bind best to ATP warhead binding region of Mi-2β. To biochemically analyze the inhibitory activity of those inhibitors, a Fluorescence Resonance Energy Transfer (FRET)-based nucleosome repositioning assay 57, 58 was designed and modified using recombinant purified human Mi-2β protein to screen an in-house library of small molecular compounds with methyldihydroimidazopyridinone structure (Fig. 5a). Briefly, the recombinant nucleosome substrates consist of a Cy5-labeled human histone octamer (H2A T120C-Cy5) wrapped with 5' Cy3-labeled DNA, which contains a terminal nucleosome 601 positioning sequence. The 601 sequence provides the most preferred locations on DNA for histone octamer thermodynamically 59 . The FRET signaling was monitored by exciting the nucleosomes at the Cy3 absorption maximum and measuring the Cy5 emissions.
FRET signaling is at a maximum level at the assembled starting point. The chromatin remodeler Mi-2β modulates histone octamer to move along the DNA in the presence of ATP. Therefore, Cy3-labeled DNA 5' end is moved away from the Cy5-labeled octamer and consequently the FRET signal is decreased (Fig.   5a). The reaction conditions of nucleosome repositioning were modified through multiple rounds of optimization and validation (Extended Data Fig. 5b-c). Z36 was initially identified as the best hit with IC50 values of 6.971 ± 2.072 µM (Extended Data Fig. 5d). Structure Activity Relationship (SAR) studies were further used to improve the specificity and efficacy of Z36 for Mi-2β inhibition. Through iterative rounds of structure-activity optimization and in vitro assay screens, Z36-MP5 (Fig. 5b) was identified to have a high inhibitory activity on Mi-2β function where it was predicted to docked into the ATP binding pocket of Mi-2β ( Fig. 5c), with its methyl group extended to a solvent-exposed channel lined with the side  (Fig. 5d), ~85 folds more inhibitory potential than the original compound Z36.
Moreover, an ATP acyl phosphate probe assay 60 was performed by ActivX Biosciences to profile of Z36-MP5 inhibition on ATPases in native cell lysates, in which the protein-protein interactions remained intact.
Z36-MP5 showed less than 35% inhibition at a concentration of 1 µM against a panel of 233 diverse ATPases (Extended Data Table 3). These results suggest that Z36-MP5 has a high Mi-2β ATPase selectivity and specificity. inhibited Mi-2β function to recover target gene expressions, such as Cxcl9, Cxcl10 and Irf1 (Fig. 5f) in B16F10 cells, indicating its high inhibitory capacity. We also performed the co-culture assay of B16F10 and activated Pmel-1 T cells to identify whether Z36-MP5 stimulation activates T cell mediated cytotoxicity. Z36-MP5 stimulation significantly induced T cell-mediated killing of B16F10 cells (Fig. 5g).
Importantly, monitoring mouse weight (Extended Data Fig. 5e) and organ tissue histological staining (Extended Data Fig. 5f) showed Z36-MP5 treatment was tolerated without significant toxicity in Z36-MP5 treatment alone induced a moderate increase in the CD8 + T cell TILs in graft melanomas that was augmented by combining with anti-PD-1 therapy ( Fig. 6d and Extended Data Fig. 6a). However, the population of CD4 + T cell and Treg cells were not changed significantly by either the individual or combinational treatments (Extended Data Fig. 6b-c). An upregulation of GZMB expression in tumor-infiltrating CD8 + T cells was detected in tumors treated with Z36-MP5, as well as the activation markers CD69, IFN-γ, CD25 and CD107, which increase was also augmented by the combinational treatment of anti-PD-1 ( Fig. 6e-f). These results indicate that Z36-MP5 represents an effectively combinational therapeutic option of anti-PD-1 treatment in melanoma.
After tamoxifen administration, mice with visible melanomas were randomly treated with Z36-MP5 (30mg/kg) once a day starting at day 9 and/or anti-PD-1 (10mg/kg) five times at day 9, 12, 15, 18 and 21 after Cre activation. Z36-MP5 in combination with the anti-PD-1 antibody treatment significantly extended the mice tumor survival in BRaf V600E /Pten null melanoma mice (Fig. 6g). However, Z36-MP5 or anti-PD-1 treatment alone cannot extend the tumor free survival in BRaf V600E /Pten null mice, which is consistent with the previous reports that BRaf V600E /Pten null melanoma was insensitive to anti-PD-1 treatment 20 (Fig. 6g). To identify the role of Z36-MP5 treatment in regulating tumor immune microenvironment, TILs were collected and assayed with flow cytometry. Z36-MP5 treatment alone moderately induced CD8 + T cell population, which was further augmented by anti-PD-1 treatment (Fig.   6h). However, the CD4 + T cell and Treg populations in BRaf V600E /Pten null mouse melanomas were not affected by either Z36-MP5 alone or in combination with anti-PD-1 treatment in BRaf V600E /Pten null melanoma (Extended Data Fig. 6d-6e). An increased expression of GZMB, CD69, IFN-γ, CD25 or CD107 in CD8 + T cells was detected in BRaf V600E /Pten null melanoma, and their induction was further augmented by the anti-PD-1 treatment (Fig. 6i-6j and Extended Data Fig. 6f). These data indicate that Z36-MP5 treatment confers a more favorable tumor microenvironment to cytotoxic CD8 + T cells for overcoming the resistance of melanoma to anti-PD-1 treatment.

Discussion
Given heterogeneity of cancer cells and dynamic evolvement of tumor microenvironment, identifying and demonstrating the potential regulatory factors, which target and modulate interferon signaling pathways and antigen presentation, will be promising in inducing response, or resistance recovery in cancer immunotherapy. Tumor cell-intrinsic resistance mechanisms of immunotherapies are deeply explored and identified, including processing and presentation of neoantigen by the major histocompatibility complex (MHC) 61,62 and absence of pre-existing T cell infiltration caused by a lack of T cell-recognized antigens 63 or MHC 14 . Melanoma-intrinsic Wnt pathway was demonstrated to contribute to a lack of T cellmelanoma recognition to prevent anti-tumor immunity 64 . The alteration of antigen presentation also regulates the interaction and recognition of tumor cell and T cell recognition, and of interferon signaling pathways to induce resistance of immunotherapy 7 .
Over recent years there has been increasing evidence that some chromatin regulatory factors are crucial in regulating resistance to anti-PD-1 antibody treatment in melanoma, 15 , such as EZH2 26 and ARID1A 27 . EZH2 inhibition enhances T cell-targeting immunotherapies in vivo 25,26 whereas ARID1A interacts with EZH2 to inhibit IFN-response gene expression in cancer cells 27  CD74 plays a role in cross-presentation on HLA class I molecules to contribute cytotoxic T cell antitumor response 72 . One recent report indicates that the acquired resistance of anti-PD-1 treatment is associated with defects in interferon-receptor signaling (mutations in JAK1/2) and antigen presentation (B2M) 47 .
JAK1/2 loss-of-function mutations result into a lack response to IFN-γ, also causing a primary resistance to PD-1 blockade therapy 73 .
Our microarray data indicates that other signaling pathways, including TNF, NF-κB and PD-L1/PD-1 signaling pathways are also regulated by Mi-2β and contribute to Mi-2β-regulated immune responses. The selective reducing TNF cytotoxicity threshold has been demonstrated to increase the response to immunotherapy in a complementary research with genome-wide CRISPR/Cas9 screen 74 . With a pooled in vivo genetic screening approach using CRISPR-Cas9 genome editing, genes involved in NF-κB signaling are also identified to be a resistant mechanism to immunotherapy 75 . In addition, decoupling NF-κB signaling from cell dying of necroptosis or inflammatory apoptosis reduces CD8 + T cell cross-priming efficiency and anti-tumor immunity, suggesting a possible mechanism for NF-κB role in orchestrating immunotherapy 76 . Metastatic melanoma was reported to release extracellular vesicles with PD-L1 on their surface, which suppresses CD8 T cell anti-tumor immunity 77 .
Targeted therapies have significantly improved clinical outcomes in patients with various cancers including BRAF and MEK/ERK inhibitors in metastatic melanoma 78,79,80 . Targeted      Z-factor was used to determine the assay quality (Z-factors above 0.5 represent an assay with an excellent quality). In the optimization assay procedure, the wells without Mi-2β was defined as 100% inhibition controls, and that containing Mi-2β was regarded as the 0% inhibition controls. The FRET signaling in each well was detected and Cy3/Cy5 ratio was calculated. Then the average (represented as µ) and standard deviations (represented as σ) of the ratios were calculated too. The Z-factor equation is Z-factor = 1 -3 × (σ 0%Inhibition + σ 100%Inhibition ) / (µ 0%Inhibition -µ 100%Inhibition ). The Z-factor was 0.729 for Mi-2β, which confirmed the optimization of assay conditions including enzyme concentration, ATP concentration and the reaction time.

Homology modeling and screening
Homology

H-NMR spectral data were recorded on Varian
Mercury 400 NMR spectrometer, and 13 C-NMR was recorded on Varian Mercury 126 NMR spectrometer at ambient temperature. Chemicals shifts (δ) were reported in ppm, coupling constants (J) were in hertz, and the splitting patterns were described as follows: s for singlet; d for doublet; t for triplet; q for quartet; and m for multiplet. Mass spectrometry was conducted using a Thermo Fisher LCQ-DECA spectrometer (ESI-MS mode). All tested compounds were purified to ≥95% purity as determined by high performance liquid chromatography (HPLC). Step f: synthesis of 5-bromo- To a suspension of 5-bromo-N-methyl-2-nitroaniline (7, 3.0 g, 13.0 mmol) and ammonium chloride ( [M] + . Step

Quantification statistical analysis
Animals were grouped randomized. The qualification experiments were blinded by investigators. All samples or animals were included in analysis. All quantitative data were presented as the mean ± SD or SEM of at least three independent experiments. The unpaired, two tailed t-test Comparisons were performed between two groups. Statistical tests were done with biological replicates. The Kaplan-Meier survival curves for survival curve were compared using the log-rank test. p < 0.05 was considered statistically significant. *p < 0.05, ** p < 0.01, *** p < 0.001, n.s., not significant.  Tumor-infiltrating lymph cells in graft tumor were measured by flow cytometry. d, The population of CD8 + and CD4 + T cells were gated within CD45 + T cells. e, The Granzyme B expression in CD8 + T was measured and quantified with flow cytometry. f, Expression of activation markers of CD8 + T cells were measured by flow cytometry assay. MFI represents mean fluorescence intensity. Log-rank test was used to determine statistical significance of P value for mouse Kaplan-Meier survival curves. Values represent mean ± SEM. *P < 0.05, **P < 0.01, *** P < 0.001.

Fig. 3. Mi-2β deficiency induces responses to anti-PD-1 treatment for melanoma in vivo
a, A schematic for experimental strategy with anti-PD-1 treatment on genetically engineered melanoma mouse model. Mice carrying conditional alleles of Tyr::CreER;BRaf CA ;Pten lox/lox or Tyr::CreER;BRaf CA ;Pten lox/lox Mi-2β lox/lox were administered with tamoxifen for constant 5 days to activate CreER to cause melanocyte-specific conversion of Braf CA to Braf V600E , and the conversion of the Pten lox/lox and Mi-2β lox/lox alleles to null alleles, which expressed proteins of BRaf V600E /Pten null or BRaf V600E /Pten null /Mi-2β null , respectively. Mice with measurable tumors were randomly treated with either control IgG (10mg/kg) or anti-PD-1 (10mg/kg) antibodies by i.p. administration at day 9, 12, 15, 18 and 21 after Cre activation. b, Mouse survival of BRaf V600E /Pten null mice treated with IgG (n=6) or anti-PD-1 (n=7), and of BRaf V600E /Pten null /Mi-2β null mice treated with IgG (n=9) or anti-PD-1 (n=11). Log-rank test was used for P value calculation. c-d, TILs were assayed with flow cytometry assay for the population of CD8 + cells (c) and CD4 + T cells (d) gated within CD45 + T cells. e, Granzyme B expression in CD8 + T was determined and quantified with flow cytometry. f, Expression of activation markers on CD8 + T cells were determined with flow cytometry assay. MFI represents mean fluorescence intensity. Values represent mean ± SEM. *P < 0.05, **P < 0.01, *** P < 0.001.  CXCL9  CXCL10  CXCL11  CCL5  TAP1  CD274  CD74  BST2  TAPBP  PSMB10  IRF1  ICAM1  IL15  IL15RA  STAT2  CD40  FAS  IFIT2  TNFAIP3  BTG1  NAMPT  NFKBIA  TNFAIP6  PTGS2  TNFAIP2  ISG20  IRF7  NMI  WARS  PSMB8  TRAFD1  UBE2L6  PSMB9  CMPK2  MX1  RSAD2  BATF2  IFI35  PARP14  HERC6  RNF31  DDX60  PARP12  TRIM21  OASL  PNPT1  EPSTI1  TRIM14  MT2A  PIM1  TNFSF10  LCP2  PNP  MTHFD2  CFB  PTPN2  C1R  PLA2G4A  NOD1  ISOC1  IL18BP  SAMHD1  XAF1  NLRC5  PTPN6  PLSCR1  IRF8  CD38  CFH  PTPN1  CCL2  EIF4E3  ST3GAL5  CASP3 mRNA Relative expression      Fig. 1. Hazard ratio of epigenetic factors dependent on CD8 T cell infiltration a, Hazard ratio of epigenetic factor in melanoma patients depending on level of CD8 T infiltration. All patients in TCGA melanoma were divided into CD8 high or CD8 low groups based on CD8A median expression. The hazard ratio and P values were calculated. The genes (n=55), whose mRNA expression levels significantly correlated with hazard ratio in patients with high CD8 T cell infiltration only, but not in patients with low CD8 T cell infiltration, were shown. b, Western blot assay showing the efficiency of shMi-2β knockdown in B16F10 cells.   Fig. 3. Analysis of Mi-2β deficient melanoma tumor a-b, The melanomas from BRaf V600E /Pten null mice and BRaf V600E /Pten null /Mi-2β null mice were prepared and processed for immunohistochemistry staining to detect the expression of melanoma marker of S100 (a), and tumor proliferation marker of Ki-67 (b). c, Treg cells in TILs were assayed by flow cytometry assay for the population within CD45 + T cells for each groups. Values represent mean ± SEM. Scale bar=200 µm. n.s. represents no significance.  Fig. 4. Mi-2β directly regulates inflammatory genes a, Plots showed the Spearman's correlation between Mi-2β mRNA level and CCL5, CD74 or CD40 mRNA expression level in RNA-seq data in TCGA SKCM-Metastasis (n=368). b, The Mi-2β-regulated downstream target genes in IFN-γ signaling were measured in BRaf V600E /Pten null and BRaf V600E /Pten null /Mi-2β null melanoma in mice treated with IgG control and/or anti-PD-1 with RT-qPCR assay. Values represent mean ± SEM. c-e, ChIP assays were performed to detect Mi-2β binding on the promoter of Cxcl9 (c), Cxcl10 (d) and Irf1 (e) genes in both shScramble and Mi-2β knockdown B16F10 cells, and IP with anti-Stat1 was used as the positive binding control. Values represent mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001.  Fig. 5. In vitro assays for Mi-2β inhibitors a, The candidate protein structure for homology modelling. 3MWY depicted the interaction of ATP and its binding pocket. b, The FRET-based nucleosome repositioning assays were performed with different concentrations of Mi-2β and a non-limiting ATP concentration (1 mM) for the indicated incubation time.

Extended Data
Values represent mean ± SD. c, The ATP titration (concentrations ranging from 0.1 to 300 µM) was performed with the FRET-based nucleosome repositioning assays. The Michaelis-Menten equation was performed to calculate the apparent ATP Km, with the ATP Km of 11.54 µM. d, The inhibitory activity of Z36 for Mi-2β chromatin modulatory activity, measured as fold changes of Mi-2β activity treated with control vehicle. Data presents as means ± SD. e, The body weight changes of C57BL/6J mice treated with Z36-MP5 (30mg/kg/day) for 2 weeks. Data are mean ± SD (n=5). f, H&E staining of tissues in C57BL/6J mice treated with or without Z36-MP5 (30mg/kg/day) for 2 weeks. Scale bar=200 µm. g, Blood concentration profiles of Z36-MP5 after a single-dose intraperitoneal injection into 3 male Sprague-Dawley (SD) rats. Values represents the mean ± SD.