The lineage stability and suppressive program of regulatory T cells require protein O-GlcNAcylation

Regulatory T (Treg) cells control self-tolerance, inflammatory responses and tissue homeostasis. In mature Treg cells, continued expression of FOXP3 maintains lineage identity, while T cell receptor (TCR) signaling and interleukin-2 (IL-2)/STAT5 activation support the suppressive effector function of Treg cells, but how these regulators synergize to control Treg cell homeostasis and function remains unclear. Here we show that TCR-activated posttranslational modification by O-linked N-Acetylglucosamine (O-GlcNAc) stabilizes FOXP3 and activates STAT5, thus integrating these critical signaling pathways. O-GlcNAc-deficient Treg cells develop normally but display modestly reduced FOXP3 expression, strongly impaired lineage stability and effector function, and ultimately fatal autoimmunity in mice. Moreover, deficiency in protein O-GlcNAcylation attenuates IL-2/STAT5 signaling, while overexpression of a constitutively active form of STAT5 partially ameliorates Treg cell dysfunction and systemic inflammation in O-GlcNAc deficient mice. Collectively, our data demonstrate that protein O-GlcNAcylation is essential for lineage stability and effector function in Treg cells.

R egulatory T (Treg) cells are distinct T lymphocytes that control immunological self-tolerance and homeostasis 1,2 . The lineage-defining transcription factor Forkhead box P3 (FOXP3), together with other transcription regulators, induces Treg cell development in the thymus. T-cell receptor (TCR)derived and interleukin-2 receptor (IL-2R)-derived instructive signals act in two steps to induce the Foxp3 gene expression in developing Treg cells [3][4][5] . Deleting or mutating the Foxp3 gene leads to the scurfy phenotype characterized by multi-organ inflammation in mice [6][7][8] . In mature Treg cells, continued expression of FOXP3 maintains their lineage identity; 9,10 however, a small but significant population of Treg cells may lose FOXP3 expression and acquire effector T-cell activities in normal and particularly inflammatory settings [11][12][13] . Nevertheless, molecular mechanisms controlling FOXP3 protein stability under homeostatic and pathologic conditions are not well understood.
Effector Treg (eTreg) cells are the most biologically potent population of Treg cells 14,15 . Recent studies have demonstrated that pathways that regulate Treg cell development are also required for the formation and function of eTreg cells. Continuous TCR signaling maintains the transcriptional program and suppressive function of eTreg cells, without affecting Foxp3 gene expression 16,17 . IL-2R and downstream STAT5 signaling are also indispensable for eTreg cell differentiation and function by controlling a distinct set of genes that are separable from those regulated by TCR signaling 18 . It is still unclear how Treg cells integrate these pathways to maintain the suppressive program.
Post-translational modification networks exist in Treg cells to rapidly integrate signals from diverse environmental stimuli to modulate Treg cell function accordingly. In this regard, the FOXP3 protein has been intensively investigated. FOXP3 can be regulated by phosphorylation, acetylation, and ubiquitination in response to environmental changes to modulate its protein stability and DNA-binding ability 19 . In recent years, a novel modification was discovered: O-linked N-Acetylglucosamine (O-GlcNAc) modifies intracellular proteins at serine and threonine residues 20 . O-GlcNAcylation is radically different from other types of glycosylation, and, analogous to phosphorylation, plays a central role in signaling pathways relevant to chronic human diseases including cardiovascular disease, diabetes, neurodegeneration, and cancer 21,22 . The enzymes O-GlcNAc transferase (OGT) and O-GlcNAcase (OGA) mediate the addition and removal of O-GlcNAc, respectively. We and others have demonstrated that O-GlcNAc signaling acts as a hormone and nutrient sensor to control many biological processes such as gene transcription, protein stability, and cell signaling [23][24][25][26] .
Earlier studies have shown that T cells express and upregulate O-GlcNAcylation upon immune activation 27 . T cell-specific ablation of OGT resulted in an increase of apoptotic T cells 28 , and blocked T cell progenitor renewal, malignant transformation and peripheral T cell clonal expansion 29 . These data demonstrate that protein O-GlcNAcylation links TCR signaling to T cell differentiation and function; however, the role of O-GlcNAcylation in Treg cells has not been studied.
Here, we demonstrate that protein O-GlcNAcylation is abundant, and is functionally important in Treg cells by modifying FOXP3 and STAT5. Selective ablation of OGT in Treg cells leads to an aggressive autoimmune syndrome in mice as a result of Treg lineage instability and eTreg cell deficiency. On the other hand, pharmacological elevation of protein O-GlcNAcylation enhances the suppressive activity of human Treg cells, which will provide insights to help us better manipulate these cells in patients to treat diseases such as autoimmune disorders, transplant rejection and cancer.

Results
FOXP3 is modified and stabilized by O-GlcNAcylation. TCRactivated protein O-GlcNAcylation is critical for T-cell development and function 29 . We found that, similar to CD4 + CD25 − naïve T cells, CD4 + CD25 + FOXP3 + Treg cells displayed abundant expression of OGT and global protein O-GlcNAcylation (Fig. 1a, b), implying a potential role of O-GlcNAcylation in Treg cells. Consistent with findings in T cells, TCR activation further promoted protein O-GlcNAcylation in Treg cells ex vivo (Fig. 1c). We also stimulated naïve T cells with TGFβ to generate induced Treg (iTreg) cells in vitro. Compared with cells only treated with anti-CD3/CD28 beads, iTreg cell showed increased levels of the Ogt gene expression and global protein O-GlcNAcylation ( Supplementary Fig. 1A, B). These data indicate that TCR activates protein O-GlcNAcylation in Treg cells.
The FOXP3 protein is subjected to various posttranslational modifications that are required for lineage maintenance and suppressive function 19 . Thus, we sought to test whether the FOXP3 protein itself could be modified by O-GlcNAcylation. FOXP3 O-GlcNAcylation could be detected when ectopically expressed in human embryonic kidney (HEK) 293 cells (Fig. 1d). OGT overexpression increased levels of total and O-GlcNAcylated FOXP3, while OGA decreased FOXP3 protein expression and its O-GlcNAcylation (Fig. 1d). Similarly, levels of total and O-GlcNAcylated FOXP3 were increased when OGA was inhibited by Thiamet-G (TMG); in contrast, inhibition of OGT by ST045849 decreased both total and O-GlcNAcylated FOXP3 (Fig. 1e). FOXP3 protein degradation induced by the protein synthesis inhibitor cycloheximide (CHX) could be prevented by the proteasome inhibitor MG132 (Fig. 1f), indicating that FOXP3 was degraded through a ubiquitin/proteasome-dependent pathway. OGT overexpression or OGA inhibition increased FOXP3 stability (Fig. 1f-h), while OGA overexpression or OGT inhibition destabilized FOXP3 (Fig. 1g, i). The ubiquitin ligase STUB1 and the deubiquitinase USP7 have been reported to control FOXP3 polyubiquitination and degradation 30,31 . STUB1 overexpression reduced FOXP3 levels, which could not be further decreased by OGA. When USP7 was present, OGT could not further increase FOXP3 protein levels ( Supplementary Fig. 1C). These data suggest that O-GlcNAcylation may counteract ubiquitination to stabilize FOXP3 protein.
To identify sites of FOXP3 O-GlcNAcylation, Flag-tagged FOXP3 was expressed in HEK 293 cells together with OGT, immunopurified with anti-Flag beads, trypsin digested, and analyzed by liquid chromatography with tandem mass spectrometry (LC-MS/MS) using electron transfer dissociation (ETD). Multiple O-GlcNAcylation sites were identified (Supplemental Data 1). Mutating 5 of these sites, including Thr38, Ser57, Ser58, Ser270, and Ser273, to alanine (5A) on the FOXP3 protein significantly blunted its O-GlcNAcylation level (Fig. 2e), reduced its stability (Fig. 2f), and ablated its transcriptional suppression activity induced by OGT ( Supplementary Fig. 2F). We then retrovirally transduced FOXP3 and FOXP3-5A into CD4 + CD25 − conventional T cells and flow cytometric analyses of transduced cells showed that the protein expression level of FOXP3-5A was much lower than that of wildtype FOXP3 (Fig. 2g, h). These results indicate that O-GlcNAcylation is required to stabilize FOXP3.
OGT-deficiency in Treg cells leads to a scurfy phenotype. To directly examine whether OGT regulates mature Treg cell function in vivo, we generated mice with Treg cell-specific deletion of OGT by using Cre recombinase driven by the endogenous Foxp3 locus (Foxp3 YFP−Cre ) to delete the loxP-flanked Ogt gene after FOXP3 was expressed in Treg cells. Of note, the Foxp3 and Ogt genes are located about 40 centimorgans apart on the X Chromosome; thus, we were able to successfully obtain KO mice. Protein O-GlcNAcylation was specifically diminished in Treg cells but not non-Treg CD4 + T, CD8 + T, B, or natural killer cells ( Supplementary Fig. 3A). Compared to Foxp3 YFP-Cre/Y Ogt wt/Y control mice, Foxp3 YFP-Cre/Y Ogt fl/Y male KO mice progressively developed systemic autoimmune lesions including conjunctivitis, dermatitis, hunched posture (Fig. 3a), extensive lymphadenopathy and splenomegaly (Fig. 3b), and loss of body weight (Fig. 3c). KO male mice became moribund at approximately 4 weeks of age (Fig. 3d), and massive lymphocytic infiltration could be seen in colon epithelium, skin epidermis, liver sinusoids, and lung interstitium (Fig. 3e).
We then analyzed the lymphocyte compartment at the age of 2 weeks, before an autoimmune phenotype was overtly apparent.
The absolute numbers of CD4 + and CD8 + T cells were increased in the lymph nodes (LNs) of Foxp3 YFP-Cre/Y Ogt fl/Y mice ( Supplementary Fig. 3B). The percentage of effector/memory cells (CD44 hi CD62L lo ) within the CD4 + and CD8 + compartments were consistently higher in both the LNs and the spleen of Foxp3 YFP-Cre/Y Ogt fl/Y KO mice than those in Foxp3 YFP-Cre/Y Ogt wt/Y controls (Fig. 3f, g). B cell frequency in the LNs and levels of IgG, IgM, and free kappa and lambda chains in the serum were upregulated in KO mice ( Supplementary Fig. 3C, D). Moreover, the frequencies and absolute numbers of CD4 + subsets including T helper (Th) 1, Th2, and Th17 were all increased ( Fig. 3h-k and Supplementary Fig. 3E-G). However, we could only observe significantly increased expression of interferon-γ (IFNγ) in CD4 + FOXP3 − T cells in Foxp3 YFP−Cre/Y Ogt fl/Y KO mice (Fig. 3l). Expression of IL-4, IL-5, IL-13, and IL-17 remained unchanged ( Fig. 3l and Supplementary Fig. 3H, I). These observations reveal an excessive Th1-dominant inflammatory response in Foxp3 YFP−Cre/Y Ogt fl/Y mice.
O-GlcNAcylation stabilizes the Treg cell lineage. We then sought to characterize Treg cells directly. 2-week-old mice showed no difference in the Treg cell frequency among CD4 + TCRβ + cells in the LNs or spleen (Fig. 4a, b). The absolute number of Treg cells in the LNs of Foxp3 YFP-Cre/Y Ogt fl/Y mice even surpassed those in Foxp3 YFP-Cre/Y Ogt wt/Y controls (Supplementary Fig. 4A, B). In the thymus, we did not observe any reduction in the frequency or number of Treg cells (Supplementary Fig. 4C, D). In addition, there was no difference in Treg  Fig. 4E, F). These data suggest that OGT is dispensable for the development of thymic Treg cells.
Similar to findings in induced Ubc-Cre/ERT2 + Ogt fl/Y cells ex vivo (see Fig. 2a-d), Treg cells from 2-week-old Foxp3 YFP-Cre/Y Ogt fl/Y KO mice had less FOXP3 protein expression than those from control mice (Fig. 4c, d). Treating isolated Treg cells from Foxp3 YFP-Cre/Y Ogt wt/Y and Foxp3 YFP-Cre/Y Ogt fl/Y mice with CHX, we were able to find decreased FOXP3 protein stability in OGTdeficient Treg cells ( Supplementary Fig. 4G). In adult Foxp3 YFP-Cre/wt Ogt fl/fl female mice, in which OGT-sufficient (YFP -) and OGT-deficient (YFP + ) Treg cells co-existed because of random inactivation of the X chromosome, YFP + OGT-deficient Treg cells displayed similar abundance as those YFP + OGT-sufficient Treg cells in Foxp3 YFP-Cre/wt Ogt wt/fl females ( Supplementary  Fig. 4H). In Foxp3 YFP-Cre/wt Ogt fl/fl females, FOXP3 expression was lower in OGT-deficient Treg cells than in OGT-sufficient Treg cells (Fig. 4e). These data demonstrate that OGT is required for FOXP3 protein stability in Treg cells.
Continuous expression of FOXP3 maintains the Treg cell identity. In Foxp3 YFP-Cre/Y Ogt fl/Y KO mice, a significant proportion of YFP + CD25 + cells lost FOXP3 expression when compared to their Foxp3 YFP-Cre/Y Ogt wt/Y counterparts (Fig. 4f), suggesting that OGT-deficient cells tend to become the so-called ex-Treg cells or latent Treg cells 12 . To directly trace the Foxp3 lineage, we crossed inducible Foxp3 eGFP-Cre-ERT2/Y Ogt fl/Y mice to the Rosa26 tdTomato Cre-reporter line, in which a loxP-flanked STOP cassette preventing transcription of the tdTomato protein was inserted into the Rosa26 locus 32 . Tamoxifen-containing diet feeding induced Treg cell-specific ablation of protein O-GlcNAcylation Gating on the CD4 + GITR + tdTomato + population, we found that more FOXP3 -tdTomato + ex-Treg cells emerged in Foxp3 eGFP-Cre-ERT2/Y Ogt fl/Y Rosa26 tdTomato/wt KO mice than in Foxp3 eGFP-Cre-ERT2/Y Ogt wt/Y Rosa26 tdTomato/wt controls (Fig. 4g). These ex-Treg cells were prone to express the Th1 transcription factor T-BET and the Th2 transcription factor GATA3, but not the Th17 transcription factor RORγT ( Fig. 4h, i). These results show that OGT-mediated protein O-GlcNAcylation contributes to the maintenance of Treg lineage stability.

O-GlcNAc is indispensable for the effector Treg development.
In mature Treg cells, IL-2 and TCR signaling are not only required for lineage stability, but also for suppressive function, partially via the AKT-mTOR axis 33 . The intensity of ribosomal protein S6 phosphorylation but not AKT S473 phosphorylation was decreased in OGT-deficient Treg cells ( Supplementary Fig. 6A), suggesting a mTORC1-specific defect. An in vitro assay of Treg's ability to abrogate lymphocyte proliferation did not reveal any difference between OGT-sufficient and OGT-deficient cells ( Supplementary Fig. 6B), indicating that OGT is only indispensable for part of the immune functions regulated by Treg cells and/or compensatory mechanisms may be evolved. Nonetheless, in Foxp3 YFP-Cre/Y Ogt fl/Y mice, we found that the abundance of CD44 hi CD62L lo effector Treg (eTreg) cells was significantly lower than Foxp3 YFP-Cre/Y Ogt wt/Y counterparts (Fig. 5a). Strikingly, eTreg cells that express signature effector molecules such as KLRG1 34 , PD-1 35 , and CD73 36 were almost eliminated in Fox-p3 YFP-Cre/Y Ogt fl/Y mice ( Fig. 5b-d). Attenuated IL-2/STAT5 signaling in OGT-deficient Treg cells.
It seemed unlikely to us that the mild downregulation of FOXP3 protein stability could account for the absence of eTreg cell formation, as FOXP3 binds to the same set of regulatory elements in both resting Treg cells and eTreg cells 37 . To gain comprehensive insight into the OGT-dependent transcriptional program in Treg cells, we performed RNA-sequencing of isolated YFP + Treg cells from Foxp3 YFP-Cre/wt Ogt wt/fl and healthy Foxp3 YFP-Cre/wt Ogt fl/fl females to avoid secondary changes in gene expression caused by inflammation. We were able to identify 269 differentially expressed genes including 154 downregulated and 115 upregulated with p values less than 0.01 (Fig. 6a). eTreg cell markers such as klrg1, S100a4, Gzmb, and Ccr2 were downregulated by the loss of O-GlcNAcylation (Fig. 6a). The transcription factor B lymphocyte-induced maturation protein (BLIMP)-1 is common to all eTreg cells 38 , and we found that BLIMP-1-upregulated genes were enriched in OGT-sufficient Treg cells, while BLIMP-1-downregulated genes were enriched in OGT-deficient Treg cells (Fig. 6b, c), suggesting that OGT maintains a transcriptional program similar to that of BLIMP-1 + eTreg cells. Ingenuity Pathway Analysis (IPA) identified 9 potential upstream regulators which were all predicted to be inhibited, and the top one was IL-2    Supplementary  Fig. 7D). The increase in the expression of CD25 and IL-10 induced by IL-2 in OGT-sufficient cells was also inhibited when OGT was deleted ( Supplementary Fig. 7E, F). STAT5, acting downstream of IL-2, is indispensable for the formation of KLRG1 + terminally differentiated Treg cells and their suppressive function 18,34 . O-GlcNAcylation of STAT5 has been shown to promote its oligomerization and transcriptional activity in cooperation with tyrosine phosphorylation in neoplastic cells 40 . Consistent with the defective IL-2 response, many STAT5-target genes were downregulated in OGT-deficient Treg cells (Fig. 6k). However, the IL-2-activated tyrosine phosphorylation of STAT5 (pY-STAT5) in Treg cells was not affected by the loss of OGT ( Supplementary Fig. 8A, B), again supporting the notion that O-GlcNAcylation can act independently of tyrosine phosphorylation 40 . O-GlcNAcylation seemed not to regulate the cellular localization of STAT5 either (Supplementary Fig. 8C). Retroviral expression of a constitutively active STAT5A (cS5) in Treg cells increased the expression of the STAT5-target gene Socs1 and Socs3, whereas the O-GlcNAc-deficient cS5-T92A diminished such effect 40 (Fig. 6l, m). Moreover, in OGT-deficient Treg cells, reconstitution of cS5 but not cS5-T92A increased Socs1 expression (Supplementary Fig. 8D), suggesting that STAT5 O-GlcNAcylation is important for its transcriptional activity. Collectively, these data demonstrate that the deficiency in protein O-GlcNAcylation results in attenuated IL-2/STAT5 activity in Treg cells.
STAT5B-CA partially rescues Treg cell dysfunction. Treg cellspecific deletion of OGT resulted in a scurfy phenotype with comparable early onset and disease severity to those observed upon IL-2R or STAT5 ablation 18 . Expression of a constitutively   active form of STAT5B-CA partially rescues IL-2R deficient Treg function 18 . We next sought to determine whether STAT5 activation could restore Treg cell function in Foxp3 YFP-Cre/Y Ogt fl/Y mice by crossing them to a Cre-inducible line overexpressing STAT5B-CA at the Rosa26 locus (Rosa26 Stat5b-CA ) 18 . Remarkably, the Foxp3 YFP-Cre/Y Ogt fl/Y Rosa26 Stat5b-CA/wt mice with STAT5B-CA overexpression specifically in OGT-deficient Treg cells alleviated skin inflammation (Fig. 7a), reduced sizes of the LNs and spleen (Fig. 7b), and prolonged lifespan (Fig. 7c), when compared to Foxp3 YFP-Cre/Y Ogt fl/Y mice. Consistent with the role of STAT5 in regulating Foxp3 transcription and Treg cell development 3,18,41 , FOXP3 protein expression was restored to a level comparable to wildtype mice and the frequency of Treg cells was further increased when STAT5B-CA was present (Fig. 7d, e, and Supplementary Fig. 9A). Meanwhile, the eTreg cell population was significantly boosted to a level comparable to wildtype mice, shown as CD44 hi CD62L lo (Fig. 7f and Supplementary  Fig. 9B) and CD44 + KLRG1 + subsets ( Fig. 7g and Supplementary  Fig. 9C). However, the expression levels of CD73 and PD-1 on a per-cell basis did not change ( Supplementary Fig. 9D), suggesting a partial rescue in the effector function of Treg cells by STAT5B-CA. Moreover, the fraction of YFP + OGT-deficient Treg cells that lost FOXP3 expression as a result of protein instability was substantially inhibited by the presence of STAT5B-CA ( Fig. 7h and Supplementary Fig. 9E), indicating that STAT5B-CA also restored Treg lineage stability. Accordingly, STAT5B-CA overexpression reduced the number of total CD4 + and CD8 + T cells (Fig. 7i, j) and the frequency of memory/effector CD4 + and CD8 + T cells in Foxp3 YFP-Cre/Y Ogt fl/Y Rosa26 Stat5b-CA/wt mice (Fig. 7k, l, and Supplementary Fig. 9F). In addition, STAT5B-CA overexpression prevented the increase in Th1 and Th2 cells caused by loss of O-GlcNAcylation in Treg cells (Fig. 7m-o and Supplementary Fig. 9G) and reduced the production of IFNγ by CD4 + and CD8 + T cells ( Fig. 7p and Supplementary Fig. 9H, I). Collectively, these data show that restoring STAT5 signaling in OGT-deficient Treg cells improves lineage stability and effector function, thereby delaying the autoimmune responses in mice.
O-GlcNAc promotes the suppressive program of Tregs. The therapeutic potential of Treg cells for autoimmune disorders and graft-versus-host disease (GVHD) has been well characterized in preclinical and early clinical studies 42,43 . We then asked whether enhancing protein O-GlcNAcylation could augment the suppressive function of Treg cells. We first treated mouse Treg cells with TMG to inhibit OGA (Fig. 8a), the enzyme removing O-GlcNAc moieties from proteins 44 , and found that elevating protein O-GlcNAcylation by TMG increased the level of FOXP3 protein on a per cell basis (Fig. 8b, c) and the number of CD4 + FOXP3 + Treg cells ( Supplementary Fig. 10A), again demonstrating that O-GlcNAcylation stabilizes FOXP3 and Treg cell lineage. In addition, TMG treatment enhanced the expression of STAT5-target genes Socs1 and Socs3 (Fig. 8d, e), and the expression of activation markers including Klrg1, Pd-1 and, to a lesser extent, Cd73 and Cd44 in mouse Treg cells (Fig. 8f).   Finally, we purified CD4 + CD25 + CD127 − CD45RA + naïve human thymus-derived Treg (tTreg) cells from peripheral blood mononuclear cells (PBMNCs) 45 . OGA inhibition by TMG significantly increased intracellular O-GlcNAc levels ( Fig. 8g) but had no effect on FOXP3 expression or tTreg cell expansion ( Fig. 8h and Supplementary Fig. 10B), indicating that the endogenous O-GlcNAcylation was sufficient to maintain human FOXP3 stability. Nevertheless, we observed a significant increase in the expression of SOCS3 gene (Fig. 8i) and the suppressive activity of human Treg cells when treated with TMG (Fig. 8j,   and Supplementary Fig. 10C). Collectively, these results reveal a fundamental role for O-GlcNAc signaling in supporting the suppressive program of mouse and human Treg cells.

Discussion
In mature Treg cells, FOXP3 expression and signals from TCR and IL-2R are continuously required for Treg lineage stability and suppressive function 9,10,[16][17][18]46 [47][48][49] . Treatment of Treg cells with a pan-deubiquitinase inhibitor reduces FOXP3 protein levels and decreases suppressive activity 31 . Ubiquitination dynamics of FOXP3 is dictated by the ubiquitin ligase STUB1 and the deubiquitinase USP7 30,31 . We postulate that O-GlcNAcylation inhibits STUB1-mediated polyubiquitination and/or facilitates USP7-mediated deubiquitination to stabilize FOXP3. By identifying and mutating O-GlcNAc sites on the FOXP3 protein, we were able to show reduced FOXP3 stability when O-GlcNAcylation was ablated (Fig. 2), suggesting a potential direct interplay between O-GlcNAcylation and ubiquitination. Nonetheless, O-GlcNAc signaling can modulate proteasome activity and other posttranslational modifications to indirectly control FOXP3 protein abundance 26 . Fig. 6 Attenuated IL-2/STAT5 signaling in OGT-deficient Treg cells. a Hierarchical clustering of top differentially expressed genes (60 genes with FDR q values less than 0.1) between OGT-sufficient and OGT-deficient Treg cells using Morpheus. b, c Enrichment plots for up-regulated (b) and down-regulated (c) genes in Blimp1 + eTreg cells from the Gene Set Enrichment Analysis (GSEA) of differentially expressed genes between OGT-sufficient and OGTdeficient Treg cells. d Ingenuity Pathway Analysis (IPA) of predicted upstream regulators for observed changes in gene expression. e-j Foxp3 YFP-Cre/wt Ogf fl/fl female mice were injected with PBS or the IL-2 immune complex (n = 2-5) for 3 consecutive days and Treg cells were analyzed at day 6. Flow cytometry of CD44 + KLRG1 + (e), CD4 + GZMB + (f) and CD4 + BLIMP-1 + (g) Treg cells among YFP − OGT-sufficient and YFP + OGT-deficient Treg cells (CD4 + CD25 + GITR + ) in LNs and spleen and corresponding frequencies in h-j. k Expression of STAT5-target genes in OGT-sufficient and OGT-deficient Treg cells. l-m mRNA levels of Socs1 (l) and Socs3 (m) in OGT-sufficient Treg cells after retrovirus infection as indicated, n = 3 each group. Data are shown as mean ± s.e.m. *p < 0.05; **p < 0.01; ***p < 0.001 by two-way ANOVA (h-j) and one-way ANOVA (l, m) Foxp3 YFP-Cre/Y Ogt fl/Y mice developed a scurfy phenotype at a similar pace as observed in mice with germline Foxp3 deficiency 6-8 . When OGT was absent, FOXP3 stability was only modestly reduced ex vivo and in vivo (Figs. 2 and 4). Thus, FOXP3 instability is only one of the contributors to Treg cell dysfunction observed in Foxp3 YFP-Cre/Y Ogt fl/Y mice. In Treg cells with TCR ablation, there was also a slight reduction in FOXP3 protein expression, without affecting Foxp3 transcription 16,17 . It would be interesting to determine whether TCR signaling controls FOXP3 stability through protein O-GlcNAcylation.
In conventional T (Tconv) cells, O-GlcNAcylated proteins are abundant and dynamically regulated by TCR activation [50][51][52] . These proteins are involved in many biological processes, including TCR signaling (LCK, ZAP70, STAT3, c-JUN, NFAT, etc), transcriptional and translational processes, and mRNA processing. Compared to Tconv cells, Treg cells possess a unique proteomic signature to accommodate specific metabolic requirements and regulatory function 53,54 . It is unclear if O-GlcNAcylated proteins identified in Tconv cells can be modified in Treg cells. Future O-GlcNAc proteomics is needed to not only profile O-GlcNAcylated proteins in response to TCR activation but also identify other potential targets that could regulate Treg cell function. Nevertheless, it is unlikely that TCR signaling is a major downstream pathway that mediates the effect of O-GlcNAcylation, as the activation of immune system in Treg-specific TCR deletion mice was much milder than we observed in Foxp3 YFP-Cre/Y Ogt fl/Y mice 16,17 .
Through gene expression profiling, we predicted that the IL-2/STAT5 signaling pathway was inhibited in OGT-deficient Treg cells. In accordance with the prediction, we found that eTreg cells failed to expand in response to IL-2 immune complexes when OGT was ablated and the scurfy phenotype in Foxp3 YFP-Cre/Y Ogt fl/Y mice could be partially ameliorated by a constitutive-active form of STAT5B. In OGT-deficient Treg cells, even though the expression of STAT5 target genes was significantly downregulated, we did not observe any change in levels of pY-STAT5, suggesting that O-GlcNAcylation can function independently of tyrosine phosphorylation. This is consistent with a recent study showing that O-GlcNAcylation and tyrosine phosphorylation act together to activate STAT5depedent malignancies 40 . Although O-GlcNAcylation at threonine 92 is required for strong pY-STAT5A upon IL-3 stimulation, erythropoietin-activated pY-STAT5A and IL-2-stimulated T cell proliferation were comparable between cells expressing constitutively active STAT5A (cS5) and cS5-T92A that lacks O-GlcNAcylation. In addition, there was no significant change in pY-STAT5B when introducing the T92A mutant to human STAT5B or the hyperactive N642H derivative 40 . These findings indicate that O-GlcNAcylation and IL-2-induced pY-STAT5 have separate but additive effects on STAT5 function.
In TCR-triggered naïve T cells, glucose and glutamine flux into the hexosamine biosynthetic pathway (HBP) to generate UDP-GlcNAc, which fuels the protein O-GlcNAcylation 29 . O-GlcNAcylation is critical for self-renewal and malignant transformation of T cell progenitors, thymocyte positive selection, and clonal expansion of peripheral T cells 29 . It is still unclear how TCR-activation promotes the HBP flux. In addition, Treg cells and CD4 + Tconv cells program distinct cellular metabolism 55 . Tconv cells are dependent on aerobic glycolysis and glutaminolysis to fulfill their bioenergetic and biosynthetic demand of proliferation 56 . However, Treg cells couple TCR and IL-2R signals with mTORC1-dependent lipid biosynthesis to establish their functional program 33 . It is thus unlikely that TCR-activated protein O-GlcNAcylation in Treg cells is a result of increased glucose flux. Recently, our findings in hepatocytes have indicated that Ca 2+ -dependent CaMKII phosphorylates and activates OGT 57 . Ca 2+ release is induced downstream of TCR signaling and the Ca 2+ -dependent serine-threonine phosphatase Calcineurin, by dephosphorylating and mediating nuclear translocation of NFAT, is crucial for Treg cell development 58,59 . It is warranted to test whether Ca 2+ signaling mediates TCR-activated O-GlcNAcylation via modulating OGT activity.
Treg cells restrain autoimmune disease, suppress inflammation, prevent organ transplant rejection, and contribute to immune dysfunctions in some infections and cancers 60,61 . However, challenges for Treg cell immunotherapy exist as Treg cells display functional instability and heterogeneity. Our study has identified OGT and protein O-GlcNAcylation as a regulatory hub connecting TCR, FOXP3, and IL-2R to support Treg cell lineage stability and suppressive function. More importantly, pharmacological elevation of O-GlcNAcylation by inhibiting OGA augments the suppressive activity of human Treg cells, thus targeting O-GlcNAcylation can facilitate the translation of Treg cell therapy into the clinic to treat diseases such as autoimmune disorders, inflammation, and GVHD.

Methods
Mice. Foxp3 YFP-cre mice (stock number 016959) and Foxp3 eGFP-Cre-ERT2 mice (stock number 016961) were purchased from the Jackson Laboratory. Rosa26 Stat5b-CA mice were provided by Dr. Alexander Rudensky at Memorial Sloan Kettering Cancer Center 18 . Ogt fl/fl Mice (Jackson Laboratory stock number 004860) were kindly provided by Dr. Xiaoyong Yang at Yale University. Rosa26 tdTomato reporter mice (Jackson Laboratory stock number 007909) were kindly provided by Dr. Jop van Berlo at University of Minnesota. Ubc CreER+ mice (Jackson Laboratory stock number 007001) were kindly provided by Dr. Doug Mashek at University of Minnesota. All mice were on C57BL/6 background. Foxp3 YFP-Cre/Y Ogt fl/Y male mice were used at 2-week old unless specific noted, with the same age Foxp3 YFP-Cre/Y Ogt wt/Y as control. Other mice were used at 6-12 weeks old unless specific noted. Mice were free to access water and fed on a regular chow or Tamoxifen food (Teklad, TD.130860) as indicated. All procedures were approved by the Institutional Animal Care and Use Committee at the University of Minnesota. All relevant ethical regulations for animal testing and research were complied with.
For histology, mouse tissues were dissected and fixed in 10% buffered formalin. Sectioning and haematoxylin & eosin (H&E) staining were performed by the Comparative Pathology Shared Resource of UMN.
Human participants. Nonmobilized PB leukapheresis products were collected from normal adult volunteers with Food and Drug Administration (FDA)-approved/cleared apheresis instruments at Memorial Blood Center (St. Paul, MN) 45 . Written informed consent was obtained from all participants with approval from the University of Minnesota Institutional Review Board.
Naïve T cell isolation and in vitro Treg cell induction. Naïve CD4 + T cells were purified from lymph nodes and spleen according to the protocol provided by the manufacturer (Invitrogen, catalog no. 11461D). Plates were coated with 10 μg/ml of anti-mouse CD3ε antibody (BioLegend catalog no. 100314) for 4 h at 37°C. Before plating the cells, plates were washed once with PBS. Naïve CD4 + T cells were suspended at the concentration of 2 × 10 6 /ml in X-Vivo15 serum-free medium (Fisher Scientific, catalog no. 04-744Q) with 2 μg/ml anti-mouse CD28 antibody (BioLegend catalog no. 102112) in the presence or absence of mouse TGF-β (BioLegend catalog no. 763102). Cells were cultured for 2 days followed with retroviral transduction or for 5 days before the collection for RNA and immunoblot analysis 62 .
Flow cytometry. For surface markers, cells were stained in PBS containing 0.5% (wt/vol) BSA with relevant antibodies at 4°C for 30 min. For analysis of intracellular markers, cells were first fixed with Fixation/Permeabilization buffer (ThermoFisher, catalog no. 00-5123) at 4°C for 30 min and then stained in Permeabilization Buffer (ThermoFisher, catalog no. 00-8333) with relevant antibodies at 4°C for 30 min. For detection of Annexin V population, cells were suspended in 100 μl 1× Annexin V binding buffer after surface staining, added 5ul APC-Annexin V at room temperature for 15 min, and then added 400 μl 1× Annexin V binding buffer for flow cytometry measurement. For pY-STAT5 signaling detection, lymphocytes were stimulated with IL-2 (0.5 ug/ml in PBS) (BioLegend, catalog no. 575402) at 37°C for 20 min, fixed with 3% Paraformaldehyde (formaldehyde) aqueous solution (Electron Microscopy Sciences, catalog no. 15710) at 1:1 ratio at room temperature for 10 min, washed with PBS, resuspended in 100% cold methanol at 4°C for 10 min and finally stained with antibodies for FACS. For intracellular cytokine staining, cells were stimulated with PMA (50 ng/ml, Sigma, catalog no. P1585) and Ionomycin (500 ng/ml, Sigma, catalog no. I3909), together with GolgiPlug (BD Biosciences, catalog no. 555029) for INFγ, IL-5, IL-13, and IL-17A or GolgiStop (BD Biosciences, catalog no. 554724) for IL-4 at 37°C for 5 h before being stained. Antibody information were shown in Supplementary Table 2. Flow cytometry data were acquired on BD Fortessa X-20 and analyzed with FlowJo (v10.5.3). Gating strategies were shown in Supplementary Fig. 11.
Peaklists were generated using PAVA (UCSF) and searched using Protein Prospector 5.23.0 against the human SwissProt database (downloaded 9/6/2016) and a randomized concatenated database with the addition of the recombinant FoxP3 recombinant sequence. Cleavage specificity was set as Trypisn allowing 2