UHRF1 restrains CD8+ T cell response in the TME via down-regulation of MHC-I expression

To investigate UHRF1’s role in immune evasion by tumor cells, we assessed the status of major immune components within the TME. Importantly, flow cytometric analysis revealed increased CD8+ T and NK cells in UHRF1-deficient LLC-OVA tumors (Fig. 2a). We also detected a significant intratumoral enrichment of CD8+ T cells in Uhrf1-knockout (Fig. 2b) or Uhrf1-knockdown (Supplementary Fig. ^2a) LG1233 tumors. These results suggest that CD8+ T cells may be critical for tumor control in both tumor models. To test this possibility, we performed a CD8+ T cell depletion assay in tumor-bearing immune-competent mice. Indeed, CD8+ T cell ablation completely or partially restored the growth of UHRF1-deficient tumors (Fig. 2c, d, Supplementary Fig. ^{2b, c}), consistent with the negligible effect of UHRF1 deficiency on tumor growth in athymic nude mice (Fig. 1k, m), pointing to a critical role for CD8+ T cells in restricting UHRF1-deficient tumor growth.

Open in a separate window

Fig. 2

UHRF1 restrains CD8+ T cell response through downregulate MHC-I expression.

a, b Flow cytometry analysis of tumor-infiltrating lymphocytes. shNT or shUhrf1 LLC-OVA cells were injected into female C57BL/6 mice, tumors were collected on day 17 for analysis, n=7 (shNT), 8 (shUhrf1) (a); sgNT or sgUhrf1 LG1233 cells were injected into female C57BL/6 mice, tumors were collected on day 8 for analysis, n=10 per group (b). c, d Tumor growth curves of tumor bearing female C57BL/6 mice treated with IgG or anti-CD8 neutralizing antibodies. n=6 (shNT), 7 (shUhrf1+IgG, shUhrf1+anti-CD8)(c), n=7 per group (d). e Flow cytometry assessing IFNγ+, GzmB+ and Ki67+ CD8+ T cells in shNT or shUhrf1 tumors. n=5 (shNT), 4 (shUhrf1). f, g Flow cytometry assessing CD69 expression (f) and cytokine production (g) in OT-I T cells co-cultured with sgNT or sgUhrf1 LG1233-OVA cells. n=3 per group. h Gene set enrichment analysis (GO molecular function) of upregulated genes in sgUhrf1 tumors compared with sgNT tumors. n=3 per group. i Flow cytometry determines the surface level of H2Kb-H2Db in LG1233 sgNT and sgUhrf1 cells. n=4 per group. j Flow cytometry determines the surface level of HLA-ABC in sgNT and sgUHRF1 H2170 cells. n=3 per group. k Flow cytometry measurement of cell surface OVA 257-264 (SIINFEKL) peptide bound to MHC-I complex in LG1233 sgNT and sgUhrf1 cells, with or without 5ng/mL IFNγ for 16h. n=3 per group. Data are mean±SEM (a–g, i–k). n indicates the number of biological (a–e, h) or technical (f, g and i–k) replicates. P values were determined using two-tailed Student’s t test (a, b, e–g, i–k), two-way ANOVA test (c, d) and one sided hypergeometric test then adjusted using multiple-test correction (h). MFI, mean fluorescence intensity; FDR, false discovery rate. Source data are provided as a Source Data file. All the data except h was repeated independently at least twice with similar results.

To further assess the impact of UHRF1 deficiency in tumor cells on CD8+ T cells, we assessed the cytokine production capacity of tumor-infiltrating CD8+ T cells. More IFNγ- and granzyme B (GzmB)-producing CD8+ T cells were observed in UHRF1-deficient tumors (Fig. 2e). Moreover, there were more proliferating CD8+ T cells after UHRF1 deprivation (Fig. 2e). Next, we co-cultured OT-I T cells, which recognize the OVA antigen⁴⁵, with non-targeting control or UHRF1-deficient LG1233-OVA tumor cells (Supplementary Fig. ^2d). Following the co-culture period, OT-I T cells in the UHRF1-deficient group were more readily activated, as indicated by elevated surface levels of CD69 (Fig. 2f), a well-known T cell activation marker^46,47. Meanwhile, the OT-I T cells in UHRF1-deficient group secreted more effector molecules such as IFNγ, TNFα, and GzmB (Fig. 2g). In line with this, UHRF1-deficient tumor cells were more readily to be killed by OT-I cells (Supplementary Fig. ^2e). Together, these data suggest enhanced T-cell activation in the absent of UHRF1.

To unbiasedly assess the signaling pathways affected by UHRF1, we performed RNA-sequencing on WT and UHRF1-deficient bulk tumor tissues. Interestingly, among the upregulated genes in UHRF1-deficient tumors, the top-ranked pathways were T cell receptor binding, peptide antigen binding, and MHC-I protein binding (Fig. 2h). Given that impaired tumor antigen processing and presentation is a principal mechanism by which tumors evade immune recognition and elimination by CD8+ cytotoxic T cells⁴⁸, we hypothesized that UHRF1 may modulate MHC-I expression, consequently affecting T cell activation. Indeed, we noticed an increased expression of H2k1 in RNA-seq data (Supplementary Fig. ^2f). We further performed flow cytometry to measure the level of MHC-I molecules on tumor cell surface. The results showed that H2Kb-H2Db and HLA-ABC are significantly increased in UHRF1-deficient mouse and human tumor cells compared with non-targeting (NT) control cells (Fig. 2i, j, Supplementary Fig. ^{2g, h}). Consequently, when the OVA peptide was loaded onto LG1233 tumor cells, UHRF1-deficient LG1233 tumor cells presented more OVA peptide on the tumor cell surface (Fig. 2k, Supplementary Fig. ²ⁱ). Furthermore, in the presence of IFNγ, which is known to enhance cell surface MHC-I expression in the TME⁴⁹, UHRF1-deficient tumor cells exhibited greater levels of OVA antigen presented on cell surface (Fig. 2k). To evaluate the role of UHRF1 on the antigen presentation of endogenous self-antigen, we expressed full-length OVA protein in LG1233 cells and detected SIINFEKL epitope presentation. As expected, more SIINFEKL epitope was presented on the UHRF1-deficient tumor cell surface (Supplementary Fig. ^2j), indicating enhanced antigen processing and presentation following UHRF1 elimination. These data collectively suggest that UHRF1 may induce immune escape of tumor cells from recognition by CD8+ T cells, potentially through downregulation of MHC-I. Additionally, we also observed an enhanced type I interferon response, and increased endogenous retroviruses (ERV) gene expression (Supplementary Fig. ^2k), both of which can induce innate immune response^50,51.

Cytoplasmic UHRF1 mediates MHC-I degradation in tumor cells within the TME

To systematically identify proteins interacting with MHC-I, which is widely presented on all nucleated cells⁵², we transfected HEK293T cells with Flag-HLAA, followed by affinity purification of HLAA-containing protein complexes using anti-Flag antibody and subsequent mass spectrometry analysis. Interestingly, UHRF1 was among the top interacting proteins (Fig. 3a). To confirm this finding, we co-transfected Flag-Uhrf1 and HA-H2k1 plasmids into HEK293T cells. Co-immunoprecipitation (Co-IP) assays revealed that UHRF1 interacts with H2K1 (Fig. 3b). In addition to its role as an epigenetic regulator, UHRF1 is also recognized as an E3 ligase capable of ubiquitinating and degrading nuclear substrates, with several reports substantiating this biochemical function^42,53–55. Given that proteolytic degradation is a well-known mechanism for restraining MHC-I expression^56–58, we reasoned that UHRF1 may bind to and ubiquitinate MHC-I, leading to its subsequent degradation. To test this postulation, we expressed HA-H2k1 in both UHRF1-sufficient and -deficient tumor cells. Subsequent Co-IP assays revealed that ubiquitination levels of H2K1 were significantly reduced in UHRF1-deficient cells (Fig. 3c), suggesting that UHRF1 possesses the biochemical capability to facilitate MHC-I ubiquitination and degradation.

Open in a separate window

Fig. 3

Cytoplasmic UHRF1 mediates MHC-I degradation.

a Affinity purification of Flag-HLAA using antibody against Flag-tag. The top hits from mass spectrometry analysis were shown. b Interaction of mouse UHRF1 and H2K1 was determined by Co-immunoprecipitation (Co-IP). HEK293T cells were transfected with Flag-Uhrf1 in combination with HA-H2k1. c Wild-type and Uhrf1-knockdown LLC-OVA cells were transfected with HA-H2k1. Ubiquitination of H2K1 was examined by Co-IP. d Representative images of UHRF1 expression by IHC staining from LUSC and LUAD patient samples and corresponding adjacent normal tissues. UHRF1 monoclonal antibody (H-8) from Santa Cruz (Cat# sc-373750) was used for the IHC staining. Scale bars, 25µm. e Quantification of the cytoplasmic UHRF1 positive cells by Qupath from LUSC (n=90) and LUAD (n=86) patients with corresponding adjacent normal tissues. f Representative confocal images showing UHRF1 localization in cultured LG1233 cells (upper panel), LG1233 tumor tissues developed in C57BL/6 mice (middle panel) and normal spleen tissues (lower panel). Nuclei were stained with DAPI. Scale bars, 10µm. n indicates the number of biological replicates. Data are mean±SEM (e). P values were determined using two-tailed Student’s t test (e). Data in (b, c, f) were repeated independently at least twice with similar results. Source data are provided as a Source Data file.

However, UHRF1 has been defined to function primarily in the nucleus, and the known substrates of its E3 ligase are nuclear proteins^42,53–55. To address this intriguing question, we closely examined the expression pattern of UHRF1 protein in tumor tissue sections derived from NSCLC patients. We were surprised to find that 70% of tumor samples exhibit cytoplasmic UHRF1 staining in cancer cells in contrast to normal cells in the adjacent regions (Fig. 3d, e). To validate this highly unusual observation, we stained tissue samples from another lung cancer cohort using an independent UHRF1 antibody from a different vendor. Consistently, we observed more than 70% of patient samples from this cohort with cytoplasmic UHRF1 localization (Supplementary Fig. ^{3a, b}). These results indicate that tumor cells not only highly express UHRF1 but also exhibit aberrant cytoplasmic localization detected in tissue samples derived from NSCLC patients. To probe this question further, we conducted immunofluorescence analysis in cultured LG1233 cells and found that UHRF1 was primarily located in the nucleus of the tumor cells; in contrast, UHRF1 was detected predominantly in the cytoplasm of LG1233 tumor tissues isolated from mice (Fig. 3f), consistent with IHC results from human cancer samples. Meanwhile, UHRF1 was consistently found in the nuclei of normal cells in mouse tissues (Fig. 3f). These observations suggest that cytoplasmic localization of UHRF1 occurs only in tumor cells grown in the TME, where UHRF1 could interact with MHC-I and cause its degradation.

Phosphorylation of UHRF1 regulates its cytoplasmic localization in response to factors in the TME, such as TGF-β

Previous studies found that phosphorylation of UHRF1 could be linked to its sequestration in the cytoplasm^59,60. To test whether the TME-induced cytoplasmic translocation of UHRF1 is regulated by phosphorylation, we generated an antibody specifically recognizing phosphorylated UHRF1 on a specific serine residue (Ser661 in human and Ser656 in mouse; hereafter termed pUHRF1), based on findings from a zebrafish study⁶⁰. We confirmed the specificity of this pUHRF1 antibody using HEK293T cells overexpressing Flag-tagged wild-type (WT) UHRF1 or Flag-tagged S661A mutant UHRF1, which is non-phosphorylable at this residue⁶⁰. Western blot analysis showed that, whereas anti-UHRF1 antibody recognized both WT UHRF1 and the S661A mutant, the anti-pUHRF1 antibody exclusively recognized WT UHRF1 (Supplementary Fig. ^4a). Immunofluorescent imaging analysis with the anti-pUHRF1 antibody further revealed a robust signal in the cytoplasm of tumor samples, in contrast with weak staining in normal lung tissues (Fig. 4a). Consistently, when we overexpressed Flag-UHRF1 (S661A, resistant to phosphorylation) and Flag-UHRF1 (S661D, mimicking phosphorylation) into A549 human lung cancer cells, significantly higher cytoplasmic Flag signal was detected in Flag-UHRF1 (S661D) overexpressing cells, while Flag-UHRF1 (S661A) was mainly located in the nuclei (Supplementary Fig. ^4b). These findings are consistent with previous reports showing that UHRF1 phosphorylated at Ser661 (mouse Ser656) is enriched in the cytoplasm^59,60. Since we found that UHRF1 physically interacts with the MHC-I molecule (Fig. 3a, b), we next tested whether this interaction is regulated by UHRF1 phosphorylation. To investigate this, we co-transfected HA-HLAA with either wild-type, S661A, or S661D Flag-UHRF1 and performed Co-IP analysis. This revealed that the phosphomimic S661D mutation significantly enhances the interaction between UHRF1 and MHC-I (Fig. 4b). Furthermore, immunofluorescence staining showed the co-localization of phosphorylated UHRF1 and MHC-I (Fig. 4c). Moreover, MHC-I levels on the surface of Flag-UHRF1 (S661D) overexpressing cells are reduced as compared with Flag-UHRF1 (S661A) overexpressing cells (Fig. 4d). These data suggest that UHRF1 phosphorylation promotes its interaction with MHC-I to facilitate MHC-I degradation. Functionally, we found that overexpression of Uhrf1 (S656D) in mouse LG1233 tumor cells notably enhanced tumor growth in immune-competent host, compared with the phosphorylation-resistant form Uhrf1 (S656A) (Fig. 4e, Supplementary Fig ^4c), likely through more effective suppression of MHC-I expression on cell surface. To evaluate the expression profile of pUHRF1 in human lung cancer, we stained patient samples with the antibody against pUHRF1 (S661). The expression pattern of pUHRF1 (S661) was similar among patients with stage I, II and III lung cancer and colon cancer, suggesting that the pUHRF1-mediated immune evasion may be involved in both early tumor development and advanced progression (Supplementary Fig. ^4f). We next explored the relationship between pUHRF1 and MHC-I in patient samples. A negative correlation between pUHRF1 and HLA-ABC expression was observed in both lung cancer and colon cancer patient samples (Fig. 4f, g, Supplementary Fig. ^{4d, e}). Together, these results indicate that phosphorylated UHRF1 localized in the cytoplasm of tumor cells within the TME suppresses MHC-I to impair antigen presentation and curtail anti-tumor immunity.

Open in a separate window

Fig. 4

Phosphorylated UHRF1 is localized to the cytoplasm in response to TGF-β within the TME.

a Representative confocal images showing pUHRF1 (green) in LG1233 tumor tissues and normal lung tissues. Scale bars, 25µm. b The UHRF1-HLAA interaction was examined by Co-IP. HEK293T cells were co-transfected with HA-HLAA and Flag-UHRF1 (wild-type, S661A or S661D). c Localization of HLA-ABC with pUHRF1. Dashed box regions were magnified. Scale bar, 25µm. d Surface level of HLA-ABC in A549 cells with overexpressed Flag-UHRF1 (S661A) or Flag-UHRF1 (S661D), n=5 per group. e LG1233 cells with S656A or S656D Uhrf1 expression were subcutaneously injected into male C57BL/6 mice. Tumor volume was monitored. Uhrf1 (S656A), n=7; Uhrf1 (S656D), n=8. f Representative IHC images of pUHRF1 and HLA-ABC staining, showing low pUHRF1 with high HLA-ABC (patient 1) and high pUHRF1 with low HLA-ABC (patient 2). Scale bars, 50µm. g Pearson correlation of pUHRF1 expression with HLA-ABC level in lung cancer samples (n=90). h Representative confocal images showing LG1233 cells treated with or without 200ng/mL TGF-β1 for 72h, stained for UHRF1 (green) and DAPI (blue). Scale bars, 25µm. i Representative confocal images showing pUHRF1 (red) localization in sgNT and sgTgfbr2 LG1233 tumor tissues. Scale bars, 25µm. j Tumor growth curves of sgNT or sgTgfbr2 LG1233 cells injected into male C57BL/6 mice. n=6 per group. Data are mean±SD (d) and mean±SEM (e, j). n indicates the number of biological (e, j) or technical (d) replicates. For (g), n=18 of biological samples, five technical replicates in per biological sample were included. P values were determined using two-tailed Student’s t test (d) and two-way ANOVA test (e, j), the correlation coefficient (r) and P value in (g) were determined using two-tailed Pearson correlation analysis. Data in (a–d, h, i) were repeated independently twice with similar results. MFI, mean fluorescence intensity. Source data are provided as a Source Data file.

The observation that UHRF1 cytoplasmic translocation happens only when tumor cells are grown in vivo strongly indicate that factors in the TME are responsible to stimulate its phosphorylation and translocation. In this regard, a previous report indicated that UHRF1 is sequestered outside the nucleus upon TGF-β stimulation during Treg differentiation⁵⁹. Given the significant role of TGF-β in the TME across various solid tumor types^61,62, we postulated that TGF-β may facilitate UHRF1’s nuclear-to-cytoplasmic translocation. To evaluate this, we treated LG1233 cells with recombinant mouse TGF-β protein for 72h, and found a subset of tumor cells to display UHRF1 localization in the cytoplasm (Fig. 4h), which was accompanied by an increased UHRF1 phosphorylation (Supplementary Fig. ^4g). To further confirm TGF-β’s role in regulating UHRF1 cytoplasmic translocation, we employed the CRISPR-Cas9 technology to delete Tgfbr2 in LG1233 cells (Supplementary Fig. ^4h)⁶³. Upon subcutaneous injection into mice, the loss of TGF-β signaling in tumor cells markedly reduced the amount of phosphorylated UHRF1 in the cytoplasm (Fig. 4i). Furthermore, Tgfbr2-KO in LG1233 tumor cells notably inhibited tumor growth in syngeneic mice compared with NT control cells (Fig. 4j). Collectively, these results suggest that TGF-β present in the TME could be one of the factors to induce cytoplasmic translocation of UHRF1, likely by initiating its phosphorylation. The phosphorylated UHRF1 then interacts with and facilitates the degradation of MHC-I, thereby promoting immune evasion and tumor growth. It is possible that other factors present within the TME could also contribute to the aberrant UHRF1 localization, either individually or in concert with TGF-β.

Cell culture

A549, H2170, B16 and HEK293T cell lines were acquired from Duke University Cell Culture Facility, with original sourcing from ATCC. LLC-OVA cells were generated as previously described⁸⁷. LG1233 cells were provided by T.Jacks (Massachusetts Institute of Technology). MC38 cells were purchased from Kerafast. Mycoplasma contamination was tested by Universal Mycoplasma Detection Kit (ATCC, 30-1012K). A549, B16, MC38, HEK293T, LLC-OVA and LG1233 cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum and 100U/mL penicillin-streptomycin. H2170 cells were cultured in RPMI-1640 medium supplemented with 10% fetal bovine serum and 100U/mL penicillin-streptomycin. All cells were cultured at 37°C in a humidified atmosphere with 5% CO₂.

Stable cell lines

HEK293T cells with approximately 70% confluency were transfected with vectors together with lentiviral packaging vectors PAX2 and VSV-G using Lipofectamine 2000 (Thermo Fisher Scientific, 11668019). Viral particles were collected and filtered after transfection for 48h. For infection, lentiviral supernatant was added to cultured cells with 8µg/mL Polybrene (Sigma-Aldrich, TR-1003). After incubation for 24h, infected cells were selected for 3 to 5 days with 2µg/mL puromycin (Thermo Fisher Scientific, A1113803). To generate Uhrf1 knockdown cell line, shRNA targeting Uhrf1 was purchased from Sigma-Aldrich. (Mouse-shUhrf1, 5’- CTGTAGCTCCAGTGCCGTTAA-3’). To generate knockout cell lines, sgRNAs were designed based on webtool (https://chopchop.cbu.uib.no/)⁸⁸ and cloned according to Dr. Feng Zhang laboratory’s protocol^89,90. The sequences of sgRNAs used in this study are:

mouse-sgUhrf1: 5′-CATCATAATCGAAGAGTGTG-3′,

human-sgUHRF1: 5′-CGGGGCTTCTGGTACGACG-3′,

mouse-sgTgfbr2: 5′-ACCTGCAGGAGTACCTCACG-3′

To generate stable LG1233 cell line with OVA expression, wild-type OVA (N4) and its Q4 mutant were cloned into pWPT (Addgene, #12255), LG1233-OVA(N4)-T2A-GFP cells and LG1233-OVA(Q4)-T2A-Scarlet cells were sorted by FACS for 2 rounds with purity more than 95%.

Tumor models

In the mouse tumor experiments, 2 × 10⁵ LLC-OVA; 1 × 10⁵, 2 × 10⁵ or 1 × 10⁶ LG1233 (-OVA); 1 × 10⁶ MC38 and 2 × 10⁵ B16 cells were subcutaneously injected into C57BL/6 mice (6 to 8 weeks old). 2 × 10⁵ LLC-OVA cells were subcutaneously injected into athymic nude mice (6 to 8 weeks old). The specific gender of the mice, along with the number of cells injected for each experiment, are detailed in the corresponding figure legends. Both male and female mice were used in animal studies. Tumor volumes were calculated using the following formula: 0.52 × a × b², where a is the larger diameter and b is the smaller diameter of the tumor mass. The permitted maximal tumor size was 2000 mm³ by Duke University Institutional Animal Use and Care Committee. Mice were euthanized once tumor size reached 2000 mm³. In some instances, malignant tumors exhibited explosive growth, potentially exceeding 2000 mm³. Mice with oversized tumors were euthanized with CO₂ immediately upon discovery.

In vivo lymphocyte depletion and ICB therapy

Depletion of lymphocytes in C57BL/6 mice was conducted by injection of neutralizing antibodies. In brief, mice were given intraperitoneal injection of antibodies against CD8 (100µg, twice a week, Bio X Cell, BE0223) or recommended isotype control (Bio X Cell, BE0088). All deletion assays were initiated one day before tumor cell inoculation. Specific cell deletion was confirmed by flow cytometry at experimental endpoints. For ICB therapy treatments, tumor-bearing mice were intraperitoneally injected with 200µg anti-mouse PD-1 antibodies (Bio X Cell, BE0033-2) or IgG isotype control (Bio X Cell, BE0290); and 30µg anti-mouse CTLA-4 antibodies (Bio X Cell, BP0164) or IgG isotype control (Bio X Cell, BP0086) on the indicated days.

Molecular cloning

The coding sequences for mouse Uhrf1 and human UHRF1 were purchased from Horizon. Mutant variants of human UHRF1 (S661A, S661D) were generated by PCR. The coding sequence for HLAA was cloned from Addgene plasmid (#85162)⁹¹. The coding sequence for H2k1 was cloned from cDNA of mouse embryonic fibroblast (MEF) cells. These coding sequences were subsequently cloned into pSIN-lentiviral vectors using the Gibson Assembly (New England Biolabs). Details of the primers used are listed in Supplementary Table ¹.

Adoptive T cell transfer

For the adoptive transfer of OT-I T cells, lymph nodes and spleens from OT-I TCR transgenic mice were mechanically dissociated. Red blood cells were lysed using ACK lysing buffer (Quality Biological). After isolation, OT-I T cells were activated in vitro for 48h with 1µg/mL OVA N4 or Q4 peptide, then cultured with 50U/mL hIL-2 (Peprotech, 200-02). 2 × 10⁵ LG1233-OVA (N4 or Q4) tumor cells were subcutaneously injected into male C57BL/6 mice. On day 2, each tumor-bearing mouse received 2mg cyclophosphamide (Sigma-Aldrich, C7397) for lymphodepletion before T cell transfer. On day 4, mice were intravenously injected with 1 × 10⁶ OT-I T cells per mouse. 10 days after inoculation, mice were treated with isotype control antibody (10mg/kg, Bio X cell, BE0083) or anti-TGF-β (10mg/kg, Bio X Cell, BE0057) as previously described⁹². Briefly, antibodies were administered 3 times a week for 3 weeks. The first dose was delivered intravenously, and subsequent doses were given intraperitoneally.

Western blots

Total protein was extracted and lysed in RIPA buffer (Thermo Fisher Scientific) with 1% protease and phosphatase inhibitor cocktail (Thermo Fisher Scientific), and protein concentrations were determined using the BCA protein assay kit (Thermo Fisher Scientific). Equal amounts of protein in the lysates were boiled in SDS loading buffer. Protein samples were loaded and fractionated on a 4–12% SDS-PAGE gel and then transferred to a polyvinylidene difluoride membrane, which was subsequently blocked in 5% skim milk in Tris-buffered saline with 0.1% Tween-20 (TBST buffer). Primary antibodies against total proteins or phosphorylated forms of protein were applied, followed by secondary antibodies (Thermo Fisher Scientific) incubation and visualized using an enhanced chemiluminescent horseradish peroxidase substrate (Thermo Fisher Scientific). A protein ladder (Thermo Fisher Scientific) was used to determine molecular mass in the western blot analysis. The western blot primary antibodies, their source and dilution information as follows. Anti-UHRF1(Santa Cruz Biotechnology, sc-373750, 1:1000); Anti-Flag (Sigma, F1804, 1:5000); Anti-HA (Sigma, H3663, 1:2000); Anti-TGFBR2 (Santa Cruz Biotechnology, sc-17791, 1:500), Anti-β-Actin (Proteintech, 66009-1-Ig, 1: 5000), anti-Ubiquitin (Cell Signaling Technology, 43124, 1:1000), GAPDH (Santa Cruz Biotechnology, sc-365062, 1:5000). The polyclonal antibody recognizing human phosphorylated Ser-661(mouse Ser-656) in UHRF1 was generated by LifeTein, LLC, as previously reported⁶⁰. The antigen used was a synthetic phosphoserine peptide corresponding to residues 654-659 of human UHRF1, denoted as CGPSRAG(pS)PRRTSKKT-KLH.

Immunoprecipitation

Cells were transfected with the indicated constructs for 24h, washed briefly, and then lysed on ice using lysis buffer (Cell Signaling Technology, 9806) for 1h. Lysates containing 1mg of total protein were precleared with magnetic Protein G Dynabeads (Thermo Fisher, 10003D) and incubated overnight at 4°C with either 2μg of the indicated antibodies or with an isotype control. The immunoprecipitates were collected with magnets, and the pellets were washed with lysis buffer and stored in 2× sample buffer prior to western blot analysis. Mass spectrometry analysis was conducted at the Proteomics and Metabolomics Core Facility at Duke University. IP samples were analyzed by LC-MS using an Orbitrap Fusion Lumos mass spectrometer. Database searches were performed against the SwissProt Homo sapiens database. A total of 2 samples were analyzed, including one IgG control and one Flag-HLAA pull down sample.

Flow cytometry

For in vitro experiments, cultured cancer cells were resuspended and stained with APC anti-mouse H-2Kb/H-2Db (28-8-6) (Biolegend, 114614, 1:100), FITC anti-human HLA-ABC (W6/32) (Biolegend, 311404, 1:100) and APC anti-mouse H-2Kb bound to SIINFEKL (25-D1.16) (Biolegend, 141605, 1:100) at 4°C for 30min in the dark. After washing with PBS, the surface expression of MHC-I or the MHC-I bound endogenous OVA peptide was analyzed on an BD FACS Canto flow cytometer using BD FACS DIVA SOFTWARE (Duke University Flow Cytometry Shared Facility), and data were analyzed using FlowJo. To analyze tumor-infiltrating lymphocytes, murine tumor tissues were mechanically dissociated and digested with 1mg/mL collegenase type I (GIBCO, 17100-017) by incubating at 37°C for 60min. The cells were filtered with a 70µm cell strainer. Red blood cells were removed with ACK lysis buffer. Single cell suspensions were stained with NearIR-LIVE/DEAD fixable dead cell dye (Invitrogen, 1:2000), followed by incubation with the following cell surface antibodies for 30min at 4°C in the dark. After washing with PBS, cells were then subjected to flow cytometry analysis as described above. The following antibodies were used to identify immune cell populations in TME: FITC anti-mouse CD45 (30-F11) (Biolegend, 103108, 1:100), PE anti-mouse TCR vβ8.3 (1B3.3) (BioLegend, 156304, 1:100), PE/Cyanine7 anti-mouse CD4 (RM4-5) (Biolegend, 100528, 1:100), PerCP/Cyanine5.5 anti-mouse CD8β (YTS156.7.7) (Biolegend, 126610, 1:100), APC anti-mouse NK-1.1 (PK136) (Biolegend, 108710, 1:100), PE/Cyanine7 anti-mouse CD69 (H1.2F3) (BioLegend, Cat. 104511, 1:100). For intracellular staining, the Intracellular Fixation & Permeabilization Buffer Set (eBioscience™, 88-8824-00) was used to fix and permeabilize cells, followed by staining with Pacific Blue anti-mouse IFN-γ (XMG1.2) (Biolegend, 505818, 1:100), PE/Cyanine7 anti-human/mouse Granzyme B (QA16A02) (Biolegend, 372214, 1:100), APC anti-mouse TNFα (MP6-XT22) (BioLegend, 506308, 1:100), Pacific Blue anti-mouse Ki-67 (16A8) (BioLegend, 652421, 1:100). Flow cytometry gating strategies are shown in Supplementary Fig. ⁷.

Immunofluorescence

For immunofluorescence analysis, frozen tumor samples were sectioned into 8µm slices. Cultured cells were seeded on a sterile glass coverslip and maintained in complete medium. LG1233 cells were treated with or without 200ng/mL recombinant mouse TGF-β1 protein (R&D, 7666-MB-005/CF) for 72h prior to IF staining. Briefly, tumor tissues or culture cells were fixed with 4% paraformaldehyde (PFA) for 15min at room temperature. After fixation, samples were permeabilized using 0.1% Triton X-100 in PBS for 10min and then blocked with 5% BSA for 1h at room temperature. The cells or tissue sections were incubated with the primary antibodies against UHRF1 (Santa Cruz Biotechnology, sc-373750, 1:200), pUHRF1 (LifeTein, customized, 1:200), anti-HLA-ABC (Proteintech, Cat. 15240-1-AP, 1:200) and Flag (Sigma, F1804, 1:200) at 4°C overnight. Subsequently, samples were incubated with fluorescently-labeled secondary antibodies for 1h at room temperature. Tissues or cells were counterstained with DAPI. Finally, slides were mounted using ProLong Diamond antifade mountant (ThermoFisher, P36965) and images were captured with a Leica SP5 inverted confocal microscope.

Immunohistochemistry

Lung cancer tissue arrays in Fig. 3 were obtained from Shanghai Outdo Biotechnology Company Ltd. Lung cancer tissue arrays in Fig. 4 and colon cancer tissue arrays in Supplementary Fig ⁴ were purchased from TissueArray.com. Lung cancer tissue sections in Supplementary Fig ³ were obtained from West China Hospital, Chengdu, China, which was approved by the ethics committees of West China Hospital, Chengdu, China. Briefly, slides were heated until tissue transparency was achieved, followed by sequential xylene and ethanol washes, and a rinse in water. Antigen retrieval was performed using Rodent Decloaker buffer (Bio Care Medical) in a pressure cooker set to 95°C for 20min. Next, 3% hydrogen peroxide solution (VWR, BDH7690-1) was used to block endogenous peroxidases and subsequently washed with 0.1% TBS-T (Tween-20) wash buffer. Slides were then blocked with 5% BSA for 1h at room temperature. Primary anti-UHRF1 (Santa Cruz Biotechnology, sc-373750, 1:100), anti-UHRF1 (Abnova, H00029128-M01, 1:100), anti-pUHRF1 (LifeTein, LLC, customized, 1:100) and anti-HLA-ABC (Proteintech, 15240-1-AP, 1:1000) antibodies were diluted accordingly in the blocking buffer and incubated overnight at 4°C. Next day, secondary antibodies were used to incubate at room temperature for 1h. After wash 4 times with TBST buffer, DAB chromogen (DAKO, K3468) was added to slides and monitoring for brown color development. After the desired staining intensity was achieved, slides were rinsed with distilled water. Hematoxylin was applied and slides were then treated with a bluing reagent, dehydrated, and mounted using Cytoseal (Thermo Scientific, 8312-4). The images were captured by using Zeiss Axio Imager Z2. QuPath was used for quantification, images were analyzed using the H-DAB staining option, with cytoplasmic DAB thresholds set at >0.3. A minimum of 100 cells per slide were analyzed.

T-cell activation assay

CD8+ T cells were isolated from lymph nodes of OT-I mice. OT-I T cells were co-cultured with sgNT or sgUhrf1 LG1233-OVA cells at a ratio of 1:1. 16h after co-culture, T cells were stained with anti-CD69-PE/Cy7 (Biolegend,105512, H1.2F3, 1:100). For intracellular staining, T cells were fixed and permeabilized, followed by staining with Pacific Blue anti-mouse IFN-γ(XMG1.2) (Biolegend, 505818, 1:100), PE/Cyanine7 anti-human/mouse Granzyme B (QA16A02) (Biolegend, 372214, 1:100), APC anti-mouse TNFα (MP6-XT22) (BioLegend, 506308, 1:100). Cells were acquired by flow cytometry analysis.

LDH release assay

Cytotoxicity was assessed using the LDH release assay (Promega, G1780) according to the manufacturer’s instructions. Briefly, LG1233-OVA target cells were seeded at a density of 6000 cells/well in a 96-well plate. OT-1 T cells were added at effector-to-target (E:T) ratio of 1:1 and the co-cultures were incubated for 8h. Subsequently, the culture supernatants were carefully harvested to avoid disturbing the cell pellet. LDH release, indicative of target cell lysis, was quantified using the enzymatic assay provided in the kit. Cytotoxicity percentage was calculated using % Cytotoxicity = [(Experimental−Effector Spontaneous−Target Spontaneous)/(Target Maximum−Target Spontaneous)] × 100%, where Experimental represents LDH activity in wells with both effector and target cells, Effector Spontaneous is the background LDH release from effector cells alone, Target Spontaneous from target cells alone, and Target Maximum from target cells lysed with detergent to determine maximum LDH release. All assays were performed in quadruplicate to ensure reproducibility and statistical reliability.

Antigen presentation assay

sgNT and sgUhrf1 LG1233 cells were treated with 5ng/mL recombinant murine IFNγ protein (Peprotech, 315-05) for 16h. After treatment, the cells were collected and incubated with 1µg/mL OVA(SIINFEKL) peptide in a 37°C water bath for 1h. Subsequently, the cells were washed twice with PBS and stained with APC-conjugated anti-mouse SIINFEKL bound H-2Kb (25-D1.16) (Biolegend, 141605, 1:100) at 4°C for 1h. After wash with PBS, the samples were proceeded for flow cytometry analysis.

RNA-seq

1 × 10⁶ sgNT or sgUhrf1 LG1233-OVA cells were subcutaneously injected into female C57BL/6 mice. Tumor tissues were collected on day 27. Total RNA was extracted using ZYMO RESEARCH kit (Direct-zol RNA miniprep, cat. No. R2052) and dissolved in RNase-free water. Library construction and RNA sequencing were performed by Novogene. Briefly, mRNA was purified, and first strand cDNA was synthesized using random hexamer primers followed by the second stand cDNA synthesis. After end repair, A-tailing, adapter ligation, size selection, amplification, and purification, quantified libraries were pooled and sequenced on Illumina platform (NovaSeq PE150).

For RNA-seq data analysis, fastp v0.20.157 was used for read trimming and STAR v2.7.5a58 for mapping to the Gencode GRCm38 genome. Quality filtering was applied with Samtools v1.1859. Gene counts were derived using featureCounts v2.0.160 and analyzed with DESeq2 v1.36.061, filtering for genes with >5 reads in >3 samples. P values were adjusted using the Benjamini-Hochberg method. Significant DEGs (adjusted p<0.05) underwent hierarchical clustering with the R package pheatmap v1.0.12 and volcano plot visualization with EnhancedVolcano v1.14.0. GO enrichment was performed with R package clusterProfiler v4.4.462.

TCRβ sequencing

For LG1233-OVA tumor model, 1 × 10⁶ LG1233-OVA sgNT or sgUhrf1 cancer cells were inoculated into female C57BL/6 mice. Tumors were collected on day 27. For LLC-OVA tumor model, 2 × 10⁵ shNT or shUhrf1 LLC-OVA cells were inoculated into male C57BL/6 mice and subsequently treated with either 30µg IgG or CTLA4 antibodies on days 7, 10 and 14. Tumors were collected on day 17. Total RNA was extracted from the bulk tumor tissues for TCR library construction using the Direct-zol RNA MiniPrep kit (Zymo reseach, 11-331). In brief, 1µg of purified RNA was converted into cDNA, followed by amplification of the CDR3 region of rearranged TCRB loci using a multiplex PCR system. A set of forward primers, each tailored for specific TCR Vβ segments, and a reverse primer for the TCRB constant region to produce amplicons spanning the entire CDR3 region. The PCR products were run on a 2.5% agarose gels, targeted bands were purified using QIAquick Gel Extraction kit, and then sent for sequencing at on an Illumina platform at Novogene (NovaSeq PE150). Primers for TCRβ sequencing are provided in Supplementary Table ¹. For data analysis, raw sequencing data was preprocessed and aligned to TCR gene segments using MiXCR software⁹³. A two-step assembly was applied to create an extensive profile of TCR clonotypes, focusing on the CDR3 region. Clonotypes were expanded, error-corrected, and quantified. Diversity and Similarity within TCR repertoires were measured using Shannon entropy and Jaccard indices, respectively, based on the analysis of amino acid sequences. The diversity and similarity analysis were conducted using R immunarch package.

Statistics and reproducibility

No statistical method was used to predetermine sample size. Experimental sample size was chosen based on commonly accepted standards. We excluded one outlier sample in Fig. 5c by performing Grubbs’ test which calculated with GraphPad. Mice were randomly assigned to each group, other experiments were not randomized. The investigators were not blinded to allocation during experiments and outcome assessment. Statistical analysis was conducted using GraphPad Prism 8 software. P values less than 0.05 were considered statistically significant and are presented in the figures by the value. Two-way ANOVA was used for multiple comparison in tumor growth experiments. Log-rank tests were used for survival analyses. Two-tailed Student’s t-tests or Wilcox tests were used to compare between two groups. The correlation coefficient (r) and P value were obtained from Pearson correlation analysis. Chi-square test was used for comparison of the different percentage of expanded clones in different groups.

The authors thank Dr. Guohong Li at Wuhan University for providing assistance during manuscript revision. We thank Dr. Yongsheng Wang at West China Hospital for help with IHC staining for human lung cancer tissues. We thank the staff of the Duke University Flow Cytometry Shared Resource, Duke Light Microscopy Core Facility, and the Proteomics and Metabolomics Core Facility for help with data acquisition. We thank Dr. Zhao Zhang from Duke University for assistance with drawing illustrations. Schematics in Fig.​Fig.5d5d (BioRender.com/u82q392) and Supplementary Figs. ^2d (BioRender.com/e87x601), 2i (BioRender.com/k25b929), 5a (BioRender.com/o76j936), 5c (BioRender.com/j18n557) and 6 (BioRender.com/y71m947) were created in BioRender. Tan, L. (2024) with modifications and released under a Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International license. This work was supported by R01-CA233205 from the NIH to X.-F.W. and Q.-J.L. Q.-J.L. is supported by core research grants provided to the IMCB and SIgN by the BMRC, A*STAR and National Research Foundation (NRF) Singapore under the NRF Investigatorship (NRFI09-0016).