Deltex1 antagonizes HIF-1α and sustains the stability of regulatory T cells in vivo

Application of regulatory T cells (Tregs) in transplantation, autoimmunity and allergy has been extensively explored, but how Foxp3 and Treg stability is regulated in vivo is incompletely understood. Here, we identify a requirement for Deltex1 (DTX1), a contributor to T-cell anergy and Foxp3 protein level maintenance in vivo. Dtx1−/− Tregs are as effective as WT Tregs in the inhibition of CD4+CD25− T-cell activation in vitro. However, the suppressive ability of Dtx1−/− Tregs is greatly impaired in vivo. We find that Foxp3 expression is diminished when Dtx1−/− Tregs are co-transferred with effector T cells in vivo. DTX1 promotes the degradation of HIF-1α. Knockout of HIF-1α restores the Foxp3 stability and rescues the defective suppressive activity in Dtx1−/− Treg cells in vivo. Our results suggest that DTX1 exerts another level of control on Treg stability in vivo by sustaining the expression of Foxp3 protein in Tregs.

C D4 þ CD25 þ regulatory T cells (Tregs) maintain T-cell tolerance and prevent autoimmune diseases by inhibiting T-cell activation [1][2][3][4][5] . The suppressive action of Tregs is known to proceed via many mechanisms, including cytolysis, secretion of suppressive cytokines, competition for cytokines, metabolic disruption of targeted cells and priming of dendritic cells 3,6,7 . Several Treg surface molecules, such as CTLA4, GITR, CD103, LAG3 and FR4, are also involved in the inhibition of target T-cell activation through cell-cell contact [8][9][10] . Tregs are classified into thymus-derived CD4 þ CD25 þ Foxp3 þ Tregs (tTregs), which develop in the thymus before their exit into the peripheral circulation, and induced Tregs (iTregs or converted Tregs), which may be generated by treatment with transforming growth factor (TGF)-b and interleukin (IL)-2 (refs 11,12). Tregs are marked by the expression of Foxp3, a forkhead family transcription factor that is essential for their development and function 13,14 . In addition, the persistent presence of Foxp3 is required to maintain the effector activities of Tregs 15,16 . The expression and epigenetic control of Foxp3 gene have been well characterized 17 . In contrast, regulation on the protein stability of Foxp3 remains poorly understood.
The potential application of Tregs in transplantation, autoimmune diseases and allergy are being extensively examined [18][19][20][21] . How Foxp3 and Treg stability are regulated in vivo remains incompletely understood. Different results are reported for the in vivo Treg stability 22 . Treg cells have been shown to be relatively stable in vivo, with a small population of Tregs susceptible to loss of Foxp3 (refs 23,24). Alternatively, pathogen infection or inflammatory cytokines destabilize Foxp3 protein and decrease the Treg population [25][26][27][28] . Loss of Foxp3 in Tregs results in pathogenic autoimmunity manifested by the generation of inflammatory effector T cells 29,30 . Foxp3 protein is subjected to controlled degradation by several molecules. E3 ubiquitin ligase Stub1 induces the K48-polyubiquitination and degradation of Foxp3 (ref. 28), whereas deubiquitinase USP7 decreases Foxp3 polyubiquitination and increases Foxp3 expression 31 . Hypoxiainducible factor-1a (HIF-1a) binds Foxp3, inducing its degradation and thereby inhibiting Treg development 32 . Interfering with the binding of HIF-1a to Foxp3 increases Foxp3 protein stability and Treg suppressive activity 33 . Given the hypoxic conditions in vivo 34 , a mechanism should exist to prevent Foxp3 protein from HIF-1a-mediated destabilization in vivo.
Deltex (DTX) is a ubiquitin E3 ligase containing N-terminal Notch-binding WWE domains, a proline-rich motif and a RING finger (RF) domain at the C-terminus 35,36 . DTX mediates Notch activation in Drosophila 37,38 , but its exact role in Notch signalling in vertebrates remains unclear. We recently found that DTX1 negatively regulates T-cell activation by targeting MAP kinase kinase kinase 1 (MEKK1) for degradation 39 . DTX1 is a transcriptional target of nuclear factor of activated T cell (NFAT), is induced in T-cell anergy and inhibits T-cell activation in E3-dependent and E3-independent mechanisms 40 . DTX1-deficiency impairs T-cell anergy induction, promotes excess T-cell activation and generates autoimmunity 40 . In this study, we found that Treg-specific knockout of Dtx1 also led to enhanced T-cell activation and autoantibody generation. Surprisingly, DTX1 deficiency did not affect the expression of Foxp3, nor was the suppressive activity of Treg affected in vitro. The in vivo suppressive function of Treg cells was largely impaired in the absence of DTX1, which was attributed to a diminished Foxp3 protein stability in Dtx1 À / À Tregs in vivo. We found that Foxp3 protein stability was enhanced by DTX1, and DTX1 induced degradation of HIF-1a and interfered with the capacity of HIF-1a to degrade Foxp3. Knockout of HIF-1a restored the Foxp3 expression and rescued the defective suppressive activity in Dtx1 À / À Treg cells in vivo. Our results illustrate a novel role for DTX1 in maintaining the protein stability of Foxp3 in vivo and reveal an additional level of control of Treg stability.

Results
Treg-specific deletion of Dtx1 enhances T-cell activation. We previously demonstrated that T cell-specific deletion of Dtx1 (Dtx1 f/f Cd4 Cre ) leads to enhanced T-cell activation and autoantibody formation 40 . To examine the possible involvement of DTX1 in regulatory T-cell function, we generated mice with Treg-specific deletion of Dtx1 by crossing Dtx1 f/f mice with Foxp3 GFP-Cre mice 41 or Foxp3 RFP-Cre mice produced in this study. Treg cells were marked by green fluorescent protein (GFP) or red fluorescent protein (RFP) expression. No difference was found between Foxp3 GFP-Cre mice and Foxp3 RFP-Cre mice, and Foxp3 Cre was used to represent both. The selective deficiency of DTX1 in Foxp3 þ T cells (tTregs), but not in Foxp3 À T cells from Foxp3 Cre Dtx1 f/f mice, was confirmed by immunoblots ( Supplementary Fig. 1a). Similar to mice with systemic and T cell-specific conditional knockout of Dtx1 (ref. 40) thymic development was not disturbed by deficiency of DTX1 in Tregs ( Supplementary Fig. 1b). Populations of splenic CD4 þ and CD8 þ T cells were comparable between control and Foxp3 Cre Dtx1 f/f mice ( Supplementary Fig. 1c). Neither was the naïve and memory T-cell ratio affected by Treg-specific absence of DTX1 ( Supplementary Fig. 1d). However, T-cell proliferation and IL-2 production were elevated in T cells from Foxp3 Cre Dtx1 f/f mice, relative to T cells from Foxp3 Cre mice (Fig. 1a,b). Small  increases in interferon (IFN)-g and IL-17 expression could be detected in naïve Foxp3 Cre Dtx1 f/f T cells ( Supplementary Fig. 1e).
In addition, elevation of anti-dsDNA antibodies and rheumatoid factor (anti-IgG1) was also found in older Foxp3 Cre Dtx1 f/f mice (46-month old; Fig. 1c,d). Notably, the increase in T-cell activation in Foxp3 Cre Dtx1 f/f T cells was less profound than in T cells from Cd4 Cre Dtx1 f/f mice 40 . No increase in anti-histone antibodies was found in Foxp3 Cre Dtx1 f/f mice ( Supplementary  Fig. 1f), in contrast to that seen in Cd4 Cre Dtx1 f/f mice 40 . These results suggest that DTX1-deficiency in Treg accounts for part, but not all, of the phenotypes observed in Cd4 Cre Dtx1 f/f mice, and DTX1 is required for the functional activities of Treg in vivo.
The observations that Treg-conditional knockout of DTX1 led to enhanced T-cell activation and autoantibody production were unexpected, because we have previously found that DTX1 deficiency did not affect the development of tTreg 40 , and the frequency of CD4 þ Foxp3 þ cells in thymus, spleen and lymph node was comparable between Dtx1 À / À mice and littermate control (wild-type (WT)) animals (Fig. 2a). The production of IL-10 and TGF-b in Dtx1 À / À tTreg cells was indistinguishable from that in WT tTreg (Fig. 2b,c). We also determined the expression of several Treg-associated molecules in isolated WT and Dtx1 À / À tTreg cells; CTLA-4, GITR, FR4, LAG3, CD103, ICOS and surface TGF-b expression were all comparable between WT and Dtx1 À / À tTreg cells ( Fig. 2d and Supplementary  Fig. 2a,b). Furthermore, in an in vitro suppression analysis, Dtx1 À / À tTreg cells were as effective as Tregs from WT littermate controls in suppressing the proliferation of T-cell antigen receptor (TCR)-stimulated CD4 þ CD25 À T cells in proportion to the number of tTreg cells added (Fig. 2e). Recent studies have also revealed differences between polyclonal and monoclonal Treg populations in the suppression mechanism 5 . We examined antigen-specific Treg suppression, and found no difference between OT-2 Tregs and OT-2/Dtx1 À / À Tregs in the inhibition of OT-2 effector T cells (Fig. 2f). However, in an in vivo functional assay, Dtx1 À / À tTregs displayed a diminished capacity to suppress T-cell activation. Colitis was induced in Rag1 À / À mice by administration of CD4 þ CD25 À T cells, leading to 20-30% weight loss, rectal prolapse and diarrhoea (Fig. 2g,h). Tissue sections of inflamed colon revealed inflammatory infiltrate, epithelial hyperplasia, crypt cell damage and goblet cell damage (Fig. 2i). Co-injection with WT tTregs effectively repressed the induction of colitis by CD4 þ CD25 À T cells. In contrast, the co-administration of Dtx1 À / À tTregs only weakly reduced the colitis-associated pathogenesis ( Fig. 2g-i), suggesting that the absence of DTX1 greatly diminished the suppressive activities of tTregs in vivo. Therefore, even though the expression of DTX1 is dissociated from the development and in vitro activities of tTregs, DTX1 is indispensable for the in vivo function of tTregs.
Dtx1 À / À iTreg cells fail to inhibit allergic inflammation. In addition to tTregs, Tregs could be generated from CD4 þ CD25 À T cells by incubation with anti-CD3/anti-CD28 and TGF-b. Both DTX1 and Foxp3 were induced in CD4 þ CD25 À T cells treated with anti-CD3/anti-CD28 in the presence of TGF-b, with an earlier peak in upregulation of DTX1 than Foxp3 during iTreg differentiation (Fig. 3a). Despite that DTX1 was induced before Foxp3 in iTregs, the absence of DTX1 did not affect generation of iTregs from CD4 þ CD25 À T cells. During the differentiation of iTregs, the induction of Foxp3 was nearly identical between WT and Dtx1 À / À T cells (Fig. 3b). Together with the fact that Foxp3 levels were comparable between WT and Dtx1 À / À tTregs (Fig. 2), it is clear that DTX1 does not regulate the expression of Foxp3 during the development of Tregs in the thymus or periphery. In addition, no differences in the expression of CTLA-4, GITR, FR4 or LAG3 could be found between WT and Dtx1 À / À iTreg cells (Fig. 3c). We also analysed the in vitro suppressive activity of iTregs. Both WT iTregs and DTX1-KO iTregs inhibited the activation of CD4 þ CD25 À T cells to the same extent (Fig. 3d). Similar to tTregs, DTX1 deficiency did not affect the suppressive function of iTregs in vitro.
We further examined the in vivo regulatory function of iTregs. Lung inflammation was examined by bronchoalveolar lavage (BAL). Allergen-induced IL-4 secretion in BAL was suppressed by WT iTregs but was less effectively inhibited by Dtx1 À / À iTregs (Fig. 3e). WT iTregs inhibited ovalbumin (OVA)-mediated secretion of IL-5, IL-6 and IL-13 in BAL, whereas Dtx1 À / À iTregs failed to prevent the production of these cytokines. OVA sensitization also led to massive inflammatory cell infiltration and mucus obstruction of the airway upon OVA re-challenge (Fig. 3f). Pre-administration of WT iTregs prevented this exacerbation of the allergic response (Fig. 3f, OVA þ WT iTreg). In contrast, Dtx1 À / À iTregs cells were unable to suppress the profound immune reaction triggered by allergen re-challenge ( Fig. 3f, OVA þ Dtx1 À / À iTreg). These results illustrate that DTX1 is also required for the in vivo suppressive function of iTregs.
The prominent role of DTX1 in the in vivo Treg effector function prompted further search for possible defects in Dtx1 À / À Treg cells. The expression of Foxo1, Foxo3a, Ikzf4 (encoding Eos), Ebi3 (encoding IL-35 subunit) and Gzmb (encoding granzyme B) was similar in tTreg cells from WT and DTX1-null mice ( Supplementary Fig. 2c-e). We also examined whether the expression of chemokine receptors, required for T-cell homing in vivo, was altered by DTX1 deficiency. The transcripts of Ccr4, Ccr6 and Ccr7, as well as the surface amounts of CCR4, CCR5, and CCR6, were nearly identical between WT and Dtx1 À / À tTreg cells isolated from spleen and mesentery lymph nodes ( Supplementary Fig. 3). DTX1 deficiency thus did not affect the expression of several chemokine receptors on Treg cells. We also examined activation-induced cell death in WT and Dtx1 À / À tTreg. No difference was found in the extent of cell death (Annexin-V þ and PI þ ) induced by CD3 stimulation between these two types of tTregs ( Supplementary Fig. 2f). Because the in vivo conditions for iTreg differentiation is not necessarily optimal, we additionally determined whether the conversion of CD4 þ naïve T cells into iTreg cells at low TGF-b concentration was specifically affected by DTX1 deficiency. Only a very small decease in Foxp3 levels was found in the generation of Dtx1 À / À iTreg cells at suboptimal TGF-b concentrations, relative to WT iTreg cells ( Supplementary Fig. 4a).
Because iTregs may not be able to maintain stable Foxp3 expression in vivo 42,43 , tTregs were used for further investigation of in vivo stability. To examine the consequences of defective Dtx1 À / À tTregs in vivo, WT or Dtx1 À / À tTreg cells were transferred into Rag1 À / À mice, and the fraction of CD4 þ T cells with Foxp3 expression in spleen and lymph nodes was determined 1 week after transfer. The recovered CD4 þ Foxp3 high population was similar for WT and Dtx1 À / À tTregs (Fig. 4a). The CD4 þ T-cell frequency repopulating different lymph tissues of Rag1 À / À mice was comparable after the transfer of WT and Dtx1 À / À tTregs (Fig. 4b). We then examined the effect of CD4 þ CD25 À T-cell co-transfer on the in vivo stability of tTregs. CD45.2 þ tTregs were co-administered with CD45.1 þ effector T cells into Rag1 À / À mice. One week later, the CD45.2 þ Foxp3 high populations in spleen and lymph nodes of the recipient mice were quantitated. A large percentage of WT tTreg cells retained high levels of Foxp3 expression when co-transferred with CD4 þ CD25 À T cells (Fig. 4c). In contrast, the Foxp3 high population from mice receiving Dtx1 À / À tTregs was profoundly reduced (Fig. 4c). In contrast to the stability of WT Tregs in vivo, DTX1 is required for the in vivo stability of Tregs when T effector cells were present. In addition, small increases in the expression of IFN-g, IL-17, TNF-a and IL-4 in the Dtx1 À / À tTreg cells recovered from Rag1 À / À mice were found ( Supplementary Fig. 4b), suggesting that some portions of the tTreg population may be converted into effector cells in vivo. DTX1 promotes HIF-1a degradation. Because the in vivo stability of Foxp3 was compromised in the presence of co-transferred CD4 þ CD25 À T cells, we speculate that inflammation associated with effector T cells contributes to Foxp3 downregulation. We examined whether Dtx1 À / À Tregs behaved differently from WT Tregs in the inflammatory mucosal environments where colitis and allergic airway responses take place. Retinoic acid is known to promote the differentiation of Tregs in vivo at mucosal sites that are often hypoxic 44,45 . WT and Dtx1 À / À CD4 þ T cells responded similar to retinoic acid in TGF-b-induced iTreg differentiation ( Supplementary Fig. 5a). Dtx1 À / À iTregs generated in the presence of retinoic acid produced IL-10 and TGF-b indistinguishably from WT iTregs ( Supplementary Fig. 5b) and suppressed the activation of effector T cells similar to retinoic acid-treated WT iTregs, as measured by the secretion of IFN-g and IL-2 by effector T cells ( Supplementary Fig. 5c). DTX1 deficiency thus did not affect the ability of retinoic acid to stabilize the Treg population.
Recent studies demonstrate that Foxp3 protein could be destabilized by incubation with inflammatory cytokines in vitro [25][26][27][28] . We examined whether DTX1 regulated the stability of Foxp3 in Treg cells encountering inflammatory cytokines. WT and Dtx1 À / À tTreg cells were treated with inflammatory cytokines and lipopolysaccharide (LPS). Similar to those reported, tTregs displayed variable sensitivity to different inflammatory cytokines in vitro, shown by reduction in the Foxp3 protein levels ( Supplementary Fig. 6a). No difference in inflammatory cytokine-induced Foxp3 downregulation was observed between WT and Dtx1 À / À tTregs ( Supplementary  Fig. 6a).
In the hypoxic or inflammatory tissues, HIF-1a is one of the few proteins suggested to destabilize Foxp3 by binding Foxp3 and stimulating the ubiquitination of Foxp3, leading to reduction in Treg population 32 . We identified an antagonism between DTX1 and HIF-1a, observing that expression of increasing amounts of DTX1 resulted in decreased HIF-1a protein levels (Fig. 5a).  We further detected an interaction of DTX1 with HIF-1a. Immunoprecipitation of HIF-1a brought down DTX1 in 293T cells (Fig. 5f). The interaction between DTX1 and HIF-1a was independent of the RF of DTX1, as shown by the association of HIF-1a with DTX1DRF (Fig. 5f). The N-terminal WWE domain, in contrast, was required for the interaction between HIF-1a and DTX1 (Fig. 5f). No interaction between DTX1 and Foxp3, however, was detected ( Supplementary Fig. 6b). Despite the binding of DTX1 to HIF-1a, DTX1 did not directly induce the ubiquitination of HIF-1a, as illustrated by in vitro ubiquitination analysis ( Supplementary Fig. 7). The stability of HIF-1a is regulated by prolyl hydroxylase domain protein (PHD)-mediated hydroxylation at Pro402 and Pro564, located in the O 2 -dependent degradation domain (ODD) region 46 . Knockdown of PHD2 prevented DTX1-induced HIF-1a degradation in DO11.10 T cells (Fig. 5g), suggesting that DTX1-induced HIF-1a downregulation is mediated by a PHD2-dependent pathway. MG132 restored HIF-1a in normoxic T cells (Fig. 5c), whereas DTX1 deficiency decreased the proline hydroxylation of HIF-1a in MG132-treated Jurkat T cells (Fig. 5h). Prolyl hydroxylated HIF-1a is recognized by the von Hippel-Lindau (VHL)-containing E3 complex and ubiquitinated for proteasome degradation 46 . By using the induction of HIF-1a in normal T cells by TCR/CD28 ligation under normoxic conditions, we found that DTX1 deficiency decreased HIF-1a ubiquitination in CD4 þ T cells activated through CD3/CD28 (Fig. 5i). We also observed that CD3/CD28induced HIF-1a expression was increased in Dtx1 À / À iTreg cells relative to WT iTreg cells (Fig. 5j). DTX1 did not affect the protein stability of PHD2 and VHL (Supplementary Fig. 8). We further found an association between DTX1 and PHD2 ( Supplementary Fig. 9a). DTX1, but not DTX1DRF, inhibited the oligomerization of PHD2 (Supplementary Fig. 9b). How the suppression of PHD2 oligomerization by DTX1 leads to HIF-1a degradation is being delineated. Together, these results suggest that DTX1 promotes HIF-1a degradation by binding to HIF-1a and increasing PHD2-mediated proline hydroxylation and the subsequent ubiquitination of HIF-1a in T cells. DTX1 increases Foxp3 levels by antagonizing HIF-1a. We confirmed previous reports showing overexpression of HIF-1a resulted in a decrease in the levels of Foxp3 (Fig. 6a). In contrast, DTX1 expression led to an elevation of Foxp3 protein contents in 293T cells (Supplementary Fig. 6c). Hypoxia-induced HIF-1a upregulation was accompanied by a reduction in Foxp3 protein, whereas the expression of DTX1 prevented hypoxia-induced Foxp3 downregulation in DO11.10 T cells (Fig. 6b). We also examined the effect of hypoxia treatment on Foxp3 protein levels in tTregs. Foxp3 contents were comparable between control and Dtx1 À / À tTregs incubated with IL-2 under normoxic conditions (Fig. 6c). HIF-1a was induced in tTreg cells when switched to hypoxic conditions. DTX1 deficiency increased hypoxia-mediated HIF-1a upregulation in tTregs, with a concomitant decrease in Foxp3 (Fig. 6c). Together, these results suggest a protective effect of DTX1 against HIF-1a-mediated Foxp3 instability.

Discussion
T cell-specific deletion of Dtx1 results in increased T-cell activation and autoantibody generation 40 . In the present study, we further found that Treg-specific deletion of Dtx1 led to enhanced T-cell activation and elevated autoantibody production ( Fig. 1), suggesting that the presence of DTX1 in Tregs is also required for maintenance of full T-cell immune tolerance. The phenotypes of Foxp3 Cre Dtx1 f/f mice were not as profound as those in Cd4 Cre Dtx1 f/f mice. The increase in T-cell proliferation was modest in Foxp3 Cre Dtx1 f/f T cells (Fig. 1a) relative to a twofold enhanced T-cell proliferation in Cd4 Cre Dtx1 f/f T cells 40 . In addition, the elevated anti-histone antibodies in Cd4 Cre Dtx1 f/f mice were not detectable in Foxp3 Cre Dtx1 f/f mice ( Supplementary  Fig. 1f). Therefore, the phenotypes previously observed in Cd4 Cre Dtx1 f/f mice 40 are likely a combinatory consequence of the loss of T-cell anergy and impairment of Treg activity.
The normal development of tTregs in Dtx1 À / À mice and the comparable Foxp3 levels in Dtx1 À / À iTregs indicate that DTX1 does not directly participate in the expression of Foxp3. In addition, DTX1 deficiency did not affect the expression of Tregassociated molecules including CD25, CTLA-4 and GITR, which are also transcriptional targets of Foxp3. Unlike transcription factors like NFAT that are associated with Foxp3 for the expression of targets such as Il2ra and Ctla4 (ref. 47), DTX1 is apparently not involved in the expression of genes through coordination with Foxp3.
We also demonstrated that Treg-specific deficiency in DTX1 impaired the suppressive function of tTregs and iTregs in vivo, but did not affect the ability of Tregs to inhibit T effector cell activation in vitro. Recent studies have shown that deficiency in STAT3 or Ubc13 does not affect the frequency of tTregs, the expression level of Foxp3 or the in vitro suppressive activity of Tregs, but profoundly alters the in vivo suppressive activity of tTregs 27,48 . For Treg function, various mechanisms have been identified for the Treg-mediated inhibitory mechanism in vitro.
However, whether these mechanisms also operate in vivo remains unclear 7 . Our results thus provide additional evidence that the in vivo function of Tregs could be regulated by mechanisms not detectable in vitro.
We further found that the defective in vivo suppressive activity by Dtx1 À / À Treg cells could be attributed to the instability of Foxp3 protein. The expression and epigenetic control of the Foxp3 gene have been well characterized 17 . In contrast, the regulation on the Foxp3 protein stability is incompletely understood. Loss of Foxp3 in Tregs leads to generation of autoreactive effector T cells 29,30 . Two ubiquitin-modifying enzymes, Stub1 and USP7, act directly on Foxp3 protein stability. LPS and proinflammatory cytokines induce expression of E3 ubiquitin ligase Stub1, whereas Stub1 binds Foxp3 and induces Foxp3 K48-polyubiquitination and degradation in an Hsp70-dependent manner 28 . In contrast, the binding of deubiquitinase USP7 to Foxp3 promotes the removal of polyubiquitination from Foxp3 and increases Foxp3 protein levels 31 . In addition, non-ubiquitin-modifying enzymes like HIF-1a bind Foxp3 and promote Foxp3 degradation in PHD 2dependent processes 32 , whereas Sirtuin 1 interacts with Foxp3 and promotes Foxp3 ubiquitination and degradation by deacetylation of Foxp3 (ref. 49).
In the present study, we demonstrated a new mechanism: the E3 ligase DTX1 can increase the in vivo stability of Foxp3 by degradation of HIF-1a protein. DTX1 did not bind Foxp3 ( Supplementary Fig. 6b), yet DTX1 increased Foxp3 stability (Supplementary Fig. 6c). Unlike Stub1 (ref. 28), DTX1 is not involved in LPS-induced destabilization of Foxp3 protein in Tregs ( Supplementary Fig. 6a). DTX1 protected Foxp3 from hypoxiainduced Foxp3 downregulation (Fig. 6b,c) by promoting HIF-1a destabilization. We found that DTX1 bound and promoted HIF-1a degradation in a PHD2-dependent process (Fig. 5e-g), leading to a profound reduction of the HIF-1a protein half-life (Fig. 5b). Notably, DTX1 did not directly promote the ubiquitination of HIF-1a ( Supplementary Fig. 7). Instead, DTX1 increased the proline hydroxylation of HIF-1a. The exact molecular processes underlying the binding of DTX1 to HIF-1a and that lead to accelerated HIF-1a degradation are currently being examined. Our results suggest that DTX1-HIF-1a-Foxp3 represents a pathway different from those of inflammatory cytokines in the regulation of the protein stability of Foxp3. DTX1 belongs to a family of NFAT-dependent E3 ligases that includes Cbl-b, ITCH and GRAIL and that are involved in T-cell anergy 50 . Similar to DTX1, deficiency in ITCH and GRAIL does not affect the development of tTregs and iTregs or expression of Foxp3 (refs 51,52). The development of tTregs and suppressive function of Cblb À / À Tregs are also normal 53 . Notably, these E3 ligases regulate the functional integrity of Tregs cells using different mechanisms. Itch À / À Treg cells secrete excess Th2 cytokines that lead to chronic T-cell activation, skin lesions and antigen-induced airway hypersensitivity 51 . GRAIL knockout in Tregs leads to attenuated suppressive function 52 . Cbl-b-deficiency is accompanied by elevated PI-3 kinase activation, reduced nuclear presence of Foxo3 and Foxo1, and attenuated Foxp3 expression in Tregs 54 . Together, DTX1, Cbl-b, ITCH and GRAIL represent a class of E3 ligases that are not only essential for T-cell anergy, but also are required for the functional integrity of Tregs.
An opposing role of HIF-1a in Foxp3 regulation has also been reported in which HIF-1a transactivates the Foxp3 promoter and stimulates Foxp3 expression 55 . In the present study, the stability of Foxp3 was assessed in Treg cells isolated from the spleen and lymph nodes, representing peripheral fully differentiated Tregs. Our study therefore may have excluded the developmental stage of Tregs in which HIF-1a is reported to participate in Foxp3 expression 55 . However, we found that Hif1a À / À tTregs retained the ability to inhibit colitis ( Supplementary Fig. 13), in contrast to the impaired suppressive function of Hif1a À / À tTregs in the same model reported by Clambey et al. 55 We noticed that Hif1a À / À tTregs were less effective than WT tTregs in the readout of colitis score (Fig. 6f), but not in the histology analysis ( Supplementary Fig. 13). There was likely difference between Hif1a À / À tTregs and WT tTregs. Further studies will be required to determine the exact window to detect the effects of HIF-1aknockout in the inhibitory activity of tTregs.
HIF-1a is induced by inflammatory stimuli, cytokines and growth factors even in normoxic conditions 56 . HIF-1a is expressed in T cells that have infiltrated inflammatory tissues 57 . HIF-1a interacts with Foxp3 and promotes Foxp3 degradation 32 . Blocking the binding of HIF-1a to Foxp3 increases Foxp3 levels in Tregs, resulting in enhanced Treg suppressive activities 33 . HIF-1a has been shown to destabilize Foxp3 protein in fully differentiated Tregs in hypoxic or inflammatory states in vitro 25,28 . This evidence supports the possibility that HIF-1a also mediates the downregulation of Foxp3 in Tregs in a physiological inflammatory environment. We found that DTX1 was able to increase Foxp3 stability through HIF-1a downregulation. In Dtx1 À / À iTregs, HIF-1a induced by TCR stimulation was clearly increased (Fig. 5j), indicating an association between the dysfunctional Dtx1 À / À Tregs and upregulated HIF-1a. In addition, knockout of HIF-1a in Dtx1 À / À tTreg and Dtx1 À / À iTreg cells restored Foxp3 expression and its suppressive activity in vivo (Fig. 6e-h and Supplementary Figs 11-14). Our results therefore suggest that the function of Tregs in vivo requires factors like DTX1 to sustain stable expression of Foxp3 under hypoxic or inflammatory conditions. DTX1 represents another level of control of Treg function by regulating the protein stability of Foxp3 against the destabilizing effect of HIF-1a in vivo. Notably, Tregs are present at high frequency in tumour tissues and are associated with suppression of effector T cells within the tumour microenviroment 58,59 . Tumours contain hypoxic regions, with increased HIF-1a protein levels 46 . Given the ability of HIF-1a protein to degrade Foxp3 (refs 32,33), we propose that DTX1 may also contribute to the stability of tumour-infiltrated Tregs by antagonizing HIF-1a.
Tregs are one of the major components in immune tolerance maintenance [3][4][5][6][7] . Administration of Tregs to restore immune tolerance has been explored as a potential therapeutic approach in transplantation, autoimmune diseases and allergy [18][19][20][21] . The present study illustrates a new level of regulation on the stability of Tregs by antagonizing HIF-1a. It may be noted that DTX1 is unlikely the only protein that participates in the downregulation of HIF-1a. Given the diverse proinflammatory stimuli Tregs may encounter in their lifespan, DTX1 may coordinate with several stabilizing factors to confer full protection on Foxp3 protein expression in Tregs. Identification of these stabilizing factors and further characterization of the molecular mechanism underlying the Foxp3-stabilizing effect of DTX1 will help provide a full molecular picture on the physiological regulation of Tregs, as well as establish the basis for the application of Tregs in immunotherapy.
For the transfection of 293T cells, 1 Â 10 6 293T cells were seeded overnight. Plasmid DNA (1-4 mg) and OmicsFect in vitro Transfection Reagent (2-8 ml, in a ratio of 1 mg DNA to 2 ml reagent; Omics Biotechnology) were diluted separately in 250 ml serum-free DMEM medium. The diluents were mixed slowly, set at room temperature for 15 min and were then added to 293T cells. 293T cells were analysed 24 h later. Transfection of Jurkat and DO11.10 T cells was conducted by electroporation on a Neon Transfection System (MP-100 Microporator, Invitrogen). A total of 1 Â 10 7 T cells were washed twice with PBS and resuspended in 100 ml resuspension R buffer (Invitrogen). A measure of 500 pmole short interfering RNA or 10 mg plasmid DNA were added to cells, and T cells were electroporated at 1,700 V, 30 ms, 1 pulse. T cells were then resuspended in 10 ml antibiotic-free complete RPMI medium, and cultured for 24-72 h before further analysis.
For retroviral infection of T cells, Foxp3 or DTX1 were subcloned into pGCIRES-YFP, a gift from Dr Gina Costa (Stanford University). Retroviruses were produced using Phoenix-Eco cells (gifts of Dr Garry P Nolan, Stanford University) 39 . DO11.10 T cells were mixed with retroviruses-containing supernatants with polybrene (1 mg ml À 1 ), and spin-infected at 1,900 r.p.m. for 45 min at 30°C. Two days after infection, YFP-expressing DO11.10 cells were isolated by sorting on FACSVantage SE (Becton Dickinson).
For immunoprecipitation, whole-cell lysates were incubated with 1 mg specific antibody overnight at 4°C. This was followed by incubation with Protein G Mag Sepharose (10 ml; GE Healthcare) for 2 h at 4°C. The beads were washed five times and analysed by SDS-polyacrylamide gel electrophoresis (SDS-PAGE).
For immunoblots, lysates or immunoprecipitates resolved by SDS-PAGE were transferred to polyvinylidene difluoride membrane (Millipore) with transfer buffer (30 mM Tris, 250 mM glycine, 1 mM EDTA, 20% methanol) at 4°C for 2 h at 400 mA. Membranes were blocked with 5% non-fat milk in 0.1% TBST (10 mM Tris-HCl pH 8.0, 150 mM NaCl, 0.1% Tween-20) at room temperature for 1 h. The membrane was then incubated with primary antibodies overnight at 4°C. After washing, the membrane was incubated with horseradish peroxidase-conjugated secondary antibodies in blocking buffer for 1.5 h at room temperature. After washing three times, membranes were developed using ECL western blot detection reagents (GE Healthcare), and signals detected by X-ray film. Immunoblot images have been cropped for presentation. Full-size images are presented in Supplementary Figs 15-17.
Induced Treg (iTreg) cells were generated from CD4 þ CD25 À T cells by treating with immobilized anti-CD3 (5 mg ml À 1 ) and anti-CD28 (1 mg ml À 1 ) in the presence of TGF-b (5 ng ml À 1 ) for 5-7 days. CD4 þ CD25 þ cells were isolated by MACS at day 5 and used for in vitro and in vivo suppression analysis. For expression of Treg-associated molecules, iTreg cells after 6-day treatment were first stained with antibodies specific for CD103, CTLA-4, FR4, GITR, LAG3 or TGF-b, followed by secondary antibody. Cells were then permeabilized, stained with FITCconjugated anti-Foxp3 and the expression of each marker in the Foxp3 þ population was analysed by FACS.
Induction of colitis in Rag1 À / À mice. CD4 þ CD25 À T cells (4 Â 10 5 ) from female mice were administered by intraperitoneal injection into female RAG1 À / À mice (8-to 10-week-old) with or without 1 Â 10 5 tTregs from female mice. Colitis was assessed weekly. The score for wasting was as follows: 0, no wasting; 1 for up to 10% initial body weight loss and 2 for 410% initial body weight loss. Rectal prolapse was scored as 0 for absence and 1 for presence. Diarrhoea was scored as 0 for absence, 1 for soft stool and 2 for watery/bloody stool. Hunching/bristled fur/ skin lesions were scored as 0 for absence and 1 for presence. At 7 weeks after T-cell transfer, mice were killed and blood was removed by cardiac perfusion with PBS. Colons were removed, cleaned by PBS, fixed in 4% paraformaldehyde and then embedded in paraffin. Sections were stained with haematoxylin and eosin or PAS staining system.
Allergen-induced airway inflammation. C57/BL6 mice (8-to 10-week-old) were intravenously injected with or without 2.5 Â 10 6 iTreg cells or PBS on day 0, and sensitized with two intraperitoneal doses of 50 mg OVA in 2 mg alum solution 62 on day 1 and day 14. Mice were challenged daily for 30 min with OVA by placing them in a chamber where an aerosolized solution of 1% OVA in saline (0.9% NaCl) was generated by a Nebulizer 646 in aerosol therapy system (DeVilbliss) from days 21 to 24. On day 25, mice were killed, perfused with PBS and lungs were fixed with 4% PFA followed by staining with haematoxylin and eosin and PAS.