Protein O-GlcNAcylation is upregulated in various cancers [12, 18–21]. OGT is the only known enzyme that mediates O-GlcNAcylation of proteins at the Ser or Thr residues [22], we hypothesized that OGT could serve as an important regulator to regulate cancer cell growth and serve as a biomarker for cancer. Initial analysis of data from The Cancer Genome Atlas (TCGA) dataset in GEPIA2 (http://gepia2.cancer-pku.cn/#index) and found a significant positive correlation between OGT mRNA expression and tumorigenesis in bladder urothelial carcinoma (BLCA), cholangiocarcinoma (CHOL), colon adenocarcinoma (COAD), esophageal carcinoma (ESCA), head and neck squamous cell carcinoma (HNSC), kidney chromophobe (LIHC), lung adenocarcinoma (LUAD), prostate adenocarcinoma (PRAD), rectum adenocarcinoma (READ), sarcoma (SARC), and a stomach adenocarcinoma (STAD) (Supplemental Fig. 1A). Next, we utilized UALCAN database analysis, and found that OGT mRNA expression was compared between 286 COAD samples and 41 adjacent or normal samples, the expression of OGT was significantly increased in COAD at the transcriptional level (Fig. 1A). A significant positive correlation tendency between OGT mRNA expression and individual different tumor stages was observed. The stage IV COAD tissues exhibited the highest expression level of OGT in compared with low stage (Fig. 1B). We also observed a significant positive correlation between OGT mRNA expression and nodal metastasis status (Fig. 1C). Furthermore, we found that the protein level of OGT was also significantly increased in COAD patient samples and amongst individual different stages based on the CPTAC and HPA online database (Fig. 1D-F). Similar results are found in LUAD (Supplemental Fig. 1B-F). These results suggest that high mRNA and protein levels of OGT in tumorigenesis were consistent in different databases.

To determine whether intestinal OGT expression was increased in Apc^min colorectal tumor mouse model, intestinal tissues were collected for western blot and immunohistochemical staining (IHC) analysis. Intestinal OGT protein was markedly increased in Apc^min mouse tumor tissues compared to adjacent or normal tissues (Fig. 1G). As expected, O-GlcNAcylation proteins levels in intestinal tissues were also significantly higher in tumor tissues than adjacent or normal tissues. IHC staining revealed that mouse OGT protein was markedly higher in tumor tissues than in adjacent tissues (Fig. 1H), which was consistent with the expression pattern in human samples. OGT protein was also upregulated in the azoxymethane (AOM)/dextran sodium sulfate (DSS) colorectal tumor model (Fig. 1I-K). Together, these data strongly suggest that OGT may play a critical role in tumorigenesis and serve as a prognostic marker and therapeutic target in cancer treatment.

To determine whether OGT may be an important therapeutic target for tumor treatment, we generated an intestinal epithelial cell-specific Ogt deletion (Ogt^ΔIEC) by crossing Ogt^fl/fl mice with Villin-Cre mice. Ogt^ΔIEC mice were then crossed with Apc^min mice to generate Ogt^ΔIECApc^min mice (Supplemental Fig. 2). To determine whether intestinal OGT expression was decreased in Ogt^ΔIECApc^min mice, intestinal tissues were collected for western blot analysis. Intestinal OGT levels were drastically decreased in Ogt^ΔIECApc^min mice compared to wildtype Apc^min mice (Fig. 2A). Importantly, deletion of the Ogt in intestinal epithelial cells resulted in significantly reduced tumor size and total number of polyps at 20 weeks of age (Fig. 2A), indicating that OGT promotes oncogenic transformation in colorectal tumor in vivo. To gain insight into the role of protein on intestinal carcinogenesis in intestinal cancer, we examined the differences in intestinal carcinogenesis. Hematoxylin and eosin (H&E) staining demonstrated that epithelial inflammation increased in Ogt^ΔIECApc^min mice compared to wildtype Apc^min mice (Fig. 2B). OGT deficiency in intestinal tissues was also associated with significantly elevated gene expression of key pro-inflammatory cytokines like interleukin Il1α, Il6, TNF-α, as well as several interferons and ISGs like Isg15, Mx1, Cxcl10 (Fig. 2C). Furthermore, we found IL-2, IL-6, IL-10, IL-17, IFN-α, IFN-β, IFN-γ and CXCL10 were upregulated compared with control mice (Fig. 2D).

Based on previous studies, we found that several interferons and several interferon-stimulated genes, such as Isg15, Mx1, IFN-α, IFN-β and Cxcl10 were upregulated. We hypothesized that OGT deficiency may activate the type I IFN pathway. First, we generated Ogt knock outs of murine colorectal carcinoma models like MC38, LLC, B16-OVA and of a human colorectal cell line, HT29. We then evaluated the expression changes of Ifna4, Ifnb1, Isg15, Mx1 and found that all the genes except Ifna4 mRNA expression were significantly increased in all cultured Ogt knockout cells (Fig. 3A-D). Next, we investigated how the type I IFN pathway in Ogt knockout cells was activated [23, 24]. We found that while the phosphorylation of STAT1, TBK1 and IRF3 is increased, the STING expression is reduced in Ogt knockout MC38, LLC, B16-OVA and HT29 cells (Fig. 3E-F). To determine whether this effect is specific for Ogt knockout, we stably expressed exogenous OGT in Ogt knockout cells and found that the type I IFN pathway activation effect is abolished (Fig. 3F). The type I IFN pathway is typically activated by both viral RNA and dsDNA. In order to eliminate the type I IFN pathway activate independent on RNA but dependent on dsDNA, we knocked out MAVS protein in Ogt^−/− cells, which is a RNA pivotal adaptor protein activate the downstream protein kinase TBK1, IFN pathway and ISGs in Ogt^−/− cells (24), and we found that knockout of MAVS in Ogt^−/− cells has no effect on either Isg15, Mx1, Ifna4 and Ifnb1 expression or the type I IFN pathway (Fig. 3G-H).

We further investigated the mechanism of the type I IFN signaling activation and found that Ogt knockout induced the reduction of the STING. This result is similar to previous studies [25, 26]. The phenotype of Ogt knockout implies that STING is involved in the activation of the type I IFN pathway. We hypothesized that OGT negatively regulates type I IFN through the cGAS/STING pathway. Next, we produced the OGT/cGAS and OGT/STING double knockout cells both in MC38, HT29 and B16-OVA. Surprisingly, these type I IFN and Cxcl10 signals disappeared in OGT/cGAS and OGT/STING double knockout cells both in mRNA and protein level (Fig. 3I-L).

The type I IFN signaling is a key pathway that promotes antigen presentation and DC activation [27–29]. Phagocytosis of extracellular tumor DNA by DCs triggers the activation of the cGAS-STING-IFN pathway [30, 31]. To determine whether the increased cGAS-STING-IFN signaling in Ogt knockout cells provides an activated signal for single epitope-specific CD8⁺ T cell priming by antigen-presenting cells (APCs), we added the supernatant from B16-OVA tumor cells into the co-culture system of BMDCs and OT-I cells. The results showed that BMDCs pre-cultured with Ogt^+/+ or Ogt^−/− B16-OVA tumor cells supernatant provided a potent activated signal for optimal single epitope-specific T cell proliferation (Fig. 3M), while cGAS or STING deficiency in Ogt^−/− tumor cells diminished such effects (Fig. 3N). Further, we added the supernatant from Ogt^+/+ or Ogt^−/− B16-OVA tumor cells into the co-culture system of Ifnar1 knockout BMDCs and OT-I cells. The cell proliferation was abolished when treated with Ifnar1^-/-BMDCs pre-cultured with Ogt^+/+ or Ogt^−/− B16-OVA tumor cells supernatant. (Supplemental Fig. 3). These results demonstrated that Ogt deficiency activate antitumor CD8⁺ T cells response is dependent on the type I IFN signal in dendritic cells and in a manner dependent on the tumor-cell-intrinsic cGAS-STING pathway.

Because the accumulation of cytosolic DNA is a consequence of nuclear DNA damage, which can activate immune response. The cGAS senses cytoplasmic DNA as a consequence of nuclear DNA damage [32]. Previous studies showed that the STING is degraded while cGAS-STING is activated [25]. As mentioned above, the STING degraded in Ogt knockout cells. We hypothesized that whether the cytosolic DNA is accumulated in Ogt knockout cells, we stained cytosolic double-strand DNA (dsDNA) with PicoGreen, a widely used immunofluorescence staining that selectively binds to dsDNA [33, 34], and found that a significantly higher percentage of Ogt deficiency cells than control cells in two different Ogt knockout MC38 and LLC cell clones, respectively (Fig. 4A and Supplemental Fig. 4A). In addition, we also stained dsDNA with anti-dsDNA immunofluorescence and the results were similar to PicoGreen staining (Fig. 4B and Supplemental Fig. 4B). We next assessed the phosphorylation of H2AX at Ser 139 (γH2AX), an indirect marker of DNA DSBs in the cell lines [35]. Comparable expression levels of H2AX, we found that γH2AX was dramatically increased in Ogt knockout MC38, LLC, HT29 and B16-OVA cells (Fig. 4C and 4F). We also used the anti-γH2AX immunofluorescence staining and comet assays, a classical quantifying and analyzing DNA damage, and found that Ogt knockout significantly induced immense DNA strand breakage in MC38 cells (Fig. 4D-E). Finally, the rescued results showed that exogenous OGT expression abolished the γH2AX expression both in MC38, LLC, HT29 and B16-OVA cells and DNA damage in MC38 cells (Fig. 4F-H). Together, these data indicated that Ogt knockout caused DNA damage and induced the cytosolic DNA accumulation.

Previous studies in Fig. 3F and Fig. 4F showed that expression of exogenous OGT abolished the type I IFN and γH2AX signals. To explore the mechanism how OGT regulate the IFN and γH2AX production. Next, we use GFP agarose immunoprecipitation and liquid chromatography coupled to tandem MS (LC-MS/MS) to identify different proteins that interacted with OGT in human colonic cancer cell line HT29. Interestingly, HCF-1 was the most enriched protein in the precipitates from OGT restored HT29 knockout cells compared to OGT knockout cells reconstituted with empty vector, based on the number of peptides (indicating the identification confidence) and the number of peptide-spectrum matches (PSMs, indicating the abundance) (Fig. 5A and Supplemental Table 4). Previous studies showed that the human epigenetic cell-cycle regulator (HCF-1) undergoes an unusual proteolytic maturation by OGT cleavage and process resulting in stably associated HCF-1^N1011 and HCF-1^C600 subunits that regulate different aspects of the cell cycle [36]. We found similar results in this study by using co-immunoprecipitation assay (Fig. 5B-C). We rescued the HCF-1 cleavage phenotype by expression of exogenous OGT (Fig. 5D). To clarify how to OGT regulate function of HCF-1, we transfected empty vector (EV), the full-length HCF-1-HA (HCF-1^FL), HCF-1^1-1011-HA (HCF-1^N1011), HCF-1^1-450-HA (HCF-1^N450), HCF-1^450-1011-HA (HCF-1^N450-1011) and HCF-1^1436-2035-HA (HCF-1^C600) and myc-OGT and found that OGT can bind HCF-1^C600 (Fig. 5E), furthermore OGT directly bind the HCF-1^C600 (Supplemental Fig. 5A-B). To further understand the physiological function of HCF-1, we transfected EV, HCF-1^FL, HCF-1^N1011and HCF-1^C600 into MC38 control and OGT knockout cells, respectively. As shown in Fig. 5F, transfected HCF-1^FL and HCF-1^N1011 didn’t rescue the gene expression differences, such as Isg15, Mx1 and Ifnb1, but HCF-1^C600 can restore the gene expression differences. These results were confirmed by using western blot and anti-dsDNA immunofluorescence staining (Fig. 5G-H). Overall, these data indicated that OGT regulates the HCF-1 cleavage and maturation, HCF-1^C600 can eliminate the cytosolic DNA accumulation, DNA damage, the type I IFN activation and restrain cGAS-STING-mediated DNA sensing.

We examined whether inhibiting Ogt can delay the tumor growth and prolong the survival. To further verify these effects, we utilized transplanted tumor model. We found that the Ogt deficiency MC38 colorectal tumor cells have no obvious inhibit cell growth in vitro (Supplemental Fig. 6A). However, we found that a significant delay in tumor growth, tumor weight and prolong mice survival compared to the control group (Fig. 6A-B). We also used lewis lung carcinoma (LLC) cells and the B16-OVA melanoma cells because they represent an aggressive murine tumor model and are highly resistant to various immunotherapies, similar results were shown in Fig. 6C-D and Supplemental Fig. 6B-E. These results showed that Ogt deficiency could delay tumor growth and prolong survival of mice in MC38, LLC, and B16-OVA tumor model. We next examined whether the enhanced antitumor function is related to tumor microenvironment (TME). Consistently, tumors from C57BL/6 immunocompetent mice bearing Ogt^−/− MC38 tumors compared with MC38 control tumors, showed higher proportion of CD8⁺ and CD4⁺ T cells, functional CD8⁺ IFN-γ⁺, CD8⁺ TNF-α⁺ and CD8⁺ IFN-γ⁺ TNF-α⁺ double positive T cells, but not CD45⁺, CD11b⁺ CD11c⁺, CD11b⁺ F4/80⁺, CD11b⁺ Ly6C⁺ and Treg cells (Fig. 6E-H and Supplemental Fig. 7D-H). The similar results were also observed in mice challenged with LLC or B16-OVA cells (Supplemental Fig. 7I-P).

To determine whether Ogt deficiency is dependent upon adaptive immune system, we inoculated MC38 control and Ogt knockout cells into immunodeficient Rag2^−/− mice and tracked tumor growth. The difference of tumor growth rate disappeared between MC38-Ogt knockout and control ones (Fig. 6I), indicating their association with an impaired immune response. We then postulated that Ogt knockout might have potent antitumor effects in vivo through CD4⁺ T and CD8⁺ T cells. In order to test this hypothesis, we performed antibody-mediated CD4⁺ T or CD8⁺ T cells depletion in Ogt^−/− and control tumor-bearing mice and examined the tumor growth and survival. Depletion CD8⁺ T cells dramatically enhanced the tumor growth in Ogt^−/− tumor-bearing mice, compared to the isotype antibody treatment group, and their antitumor activity of Ogt deficiency disappeared, both in tumor volume, weight and survival curve (Fig. 6J-K). However, depletion CD4⁺ T cells have no obvious effect both in tumor volume and weight in Ogt^−/− tumor-bearing mice, compared to the isotype antibody treatment group (Supplemental Fig. 8A). These results implied that CD8⁺ T cells mediated the inhibitory effect of Ogt deficiency on tumor progression.

Next, to determine whether this effect is specific for OGT knockout, we performed this experiment in vivo using MC38 tumor-bearing rescue model. As we expected, decreased tumor growth and prolong mice survival phenotype disappeared both in OGT rescued cells (Fig. 6L-M) and in OGT/cGAS or OGT/STING double knockout tumors (Fig. 6N-O). Furthermore, flow cytometry results showed that the proportion of CD4⁺ T and CD8⁺ T cells, CD8⁺ IFN-γ⁺ and CD8⁺ TNF-α⁺ disappeared both in OGT rescued cells (Supplemental Fig. 8B-D) and in OGT/cGAS or OGT/STING double knockout tumors in vivo (Fig. 6P and Supplemental Fig. 8E-F). These results proved that OGT deficiency induces the cGAS/STING and activates the type I IFN pathway.

Increasing evidences support that intratumoral infiltration of CD8⁺ T cells dictates the response to immune checkpoint blockade (ICB) therapy and its efficacy on various cancers [37, 38]. Blocking PD-L1 can restore the anti-tumor immune function and enhance the antitumor immunity by promoting CD8-positive T-cell infiltration, which is widely used in clinical immunotherapy [39, 40]. Because Ogt deficiency induced tumor cell-intrinsic immune response to recruit CD8⁺ T cells into MC38, LLC and B16-OVA cells, we hypothesized that Ogt deficiency potentiated enhance the efficacy of PD-L1 blockade in vivo. To test this hypothesis, we carried out the combination treatment of Ogt knockout and neutralizing antibody (anti-PD-L1). MC38 and LLC tumor growth was significantly delayed in tumor-bearing mice treated with PD-L1 antibody compared to isotype control (Fig. 6Q and Supplemental Fig. 8G), which translated into extended survival (Fig. 6R and Supplemental Fig. 8H). Deletion of Ogt synergized with PD-L1 blockade treatments to improve antitumor immunity. We next assessed the potential relevance of OGT in human cancer immunity. We first analyzed gene expression profiles of cancer patients from TCGA database and survival, we found that low OGT expression was associated with improved overall survival (OS) and progression free survival (PFS) in patients with COAD (Fig. 6S-T). Using TIMER2.0 (http://timer.cistrome.org) analysis, we found that OGT expression negatively correlated with CD8⁺ T cell infiltration (Fig. 6U). For further analysis, we found that a set of genes associated with immune response was robustly regulated in OGT high and OGT low patients. The response interferon-gamma, interferon-gamma production, cellular defense response, regulation of inflammatory response, acute inflammatory response is upregulated; DNA mismatch repair is a downregulated processe, as shown by Gene Ontology (GO) enrichment and pathway analysis (Supplemental Fig. 9A). Gene Set Enrichment Analysis (GSEA) showed that T cell activation, response to interferon-gamma, interferon-gamma production, antigen processing and presentation, interleukin-1/12 production, dectin-1 mediated noncanonical NF-κB signaling are negatively correlated with OGT expression (Supplemental Fig. 9B-H), while mismatch repair, covalent chromatin modification and DNA repair complex are positively correlated with OGT expression (Supplemental Fig. 9I-K). Of our most interest, we found that CD8A, IFNG, ISG15, MX1, CD274 and CXCL10 expression are negatively correlated with OGT expression (Supplemental Fig. 9L-Q). These data suggest a potential involvement of OGT deficiency and antitumor immunity in patients with cancer.

Based on previous studies, we know that Ogt knockout causes DNA damage, accumulates cytosolic DNA, induces cGAS-STING pathway and activates antitumor immunity. Here we speculated that OGT inhibitor may cause DNA damage and activate antitumor immunity. OSMI-1 is a small molecule inhibitor of OGT that does not significantly affect other glycosyltransferases and is active in a very low doses [41]. The cell proliferation exhibited no obvious difference in different concentrations in MC38 and LLC cells in vitro assay (Supplemental Fig. 10A-B). We stained cytosolic dsDNA with anti-dsDNA and found that treatment with OSMI-1 could significantly induce a high percentage of cytosolic DNA accumulation (Fig. 7A and Supplemental Fig. 10C). We next examined the DNA damage and found that γH2AX was obviously increased in OSMI-1 treated cells (Fig. 7B-C). We also performed the anti-γH2AX immunofluorescence staining and found that OSMI-1 significantly induced immense DNA strand breakage in treated cells (Fig. 7D and Supplemental Fig. 10D).

As we all know, the presence of cytosolic DNA could trigger activation of cGAS-STING pathway [42]. To investigate whether OSMI-1 activated cGAS/STING pathway, we examined activation of major regulators of the pathway in OSMI-1-treated MC38 and LLC cells, as indicated by increased phosphorylation of STAT1, TBK1, and IRF3 and reduced STING expression (Fig.7B-C), which is consistent with Ogt knockout cells. As mentioned earlier, deletion of Ogt synergized with PD-L1 blockade treatments to improve antitumor immunity. We next sought to determine whether OSMI-1 enhanced the antitumor immune effect of anti-PD-L1 antibody in vivo by using MC38 tumor-bearing model. To our surprise, similar to anti-PD-L1 therapy, OSMI-1 alone significantly inhibited MC38 tumor growth and survival, and the combination of OSMI-1 and anti-PD-L1 therapy resulted in superior tumor suppression compared with monotherapy (Fig. 7E-F). Flow cytometry results showed that proportion of CD4⁺ T and CD8⁺ T cells was increased both in OSMI-1, anti-PD-L1 treatment alone and combined treatment with OSMI-1 and anti-PD-L1 antibody (Fig. 7I), production of IFN-γ and TNF-α were significantly enhanced in intratumoral CD8⁺ T cells not only combined treatment with OSMI-1 and anti-PD-L1 antibody, but also single-agent OSMI-1 treatment (Fig. 7J-K). This pharmacological inhibition model is consistent with MC38 Ogt knockout tumor-bearing mice model.

Furthermore, we treated mice bearing LLC tumors, because it represents a most aggressive murine tumor model and are highly resistant to various immunotherapies. Our results showed that treatment with single-agent OSMI-1 can slightly inhibited LLC tumor growth and survival. Combined treatment with OSMI-1 and anti-PD-L1 antibody caused significantly greater tumor suppression than either monotherapy (Fig. 7F-G). We also found that proportion of CD4⁺ T and CD8⁺ T cells was significantly increased in combined treatment with OSMI-1 and anti-PD-L1 antibody. However, single-agent OSMI-1 or anti-PD-L1 antibody treatment cannot significantly increased the proportion of CD4⁺ T and CD8⁺ T cells (Fig. 7L). Production of IFN-γ and TNF-α were significantly enhanced in intratumoral CD8⁺ T cells in combined treatment with OSMI-1 and anti-PD-L1 antibody, but not in single-agent OSMI-1 or anti-PD-L1 treatment (Fig. 7M-N). This model is also consistent with Ogt knockout tumor-bearing mice model. In summary, OGT inhibitor OSMI-1 induces DNA damage and cytosolic DNA accumulation, activates cGAS/STING pathway. Combined OGT inhibitor with anti-PD-L1 antibody markedly suppressed tumor growth and increased CD8⁺ T cells and production of IFN-γ and TNF-α in tumor. These results demonstrated a pivotal role of OGT inhibition in augmenting the antitumor immune response of ICB. Given the increasing importance of immunotherapy for the management of patients with that OGT inhibitors, combined with anti-PD-L1 blockade may offer a particularly attractive strategy for the treatment of colorectal and lung cancer, which are instrumental in turning ‘cold tumors’ into ‘hot tumors’.