HIF-2α is indispensable for regulatory T cell function

Hypoxia-inducible factor 1α (HIF-1α) and HIF-2α are master transcription factors that regulate cellular responses to hypoxia, but the exact function in regulatory T (Treg) cells is controversial. Here, we show that Treg cell development is normal in mice with Foxp3-specific knockout (KO) of HIF-1α or HIF-2α. However, HIF-2α-KO (but not HIF-1α-KO) Treg cells are functionally defective in suppressing effector T cell-induced colitis and inhibiting airway hypersensitivity. HIF-2α-KO Treg cells have enhanced reprogramming into IL-17-secreting cells. We show crosstalk between HIF-2α and HIF-1α, and that HIF-2α represses HIF-1α expression. HIF-1α is upregulated in HIF-2α-KO Treg cells and further deletion of HIF-1α restores the inhibitory function of HIF-2α-KO Treg cells. Mice with Foxp3-conditional KO of HIF-2α are resistant to growth of MC38 colon adenocarcinoma and metastases of B16F10 melanoma. Together, these results indicate that targeting HIF-2α to destabilize Treg cells might be an approach for regulating the functional activity of Treg cells.

The role of HIF-1α in Treg cells has been explored in many studies. HIF-1α has been shown to bind Foxp3 and to promote Foxp3 degradation, thereby inhibiting Treg cell differentiation 41 . HIF-1α regulates T cell metabolism and participates in glycolysis, thereby suppressing Treg cell development 42 . Foxp3 protein stability is increased by an insertional mutation that blocks binding of HIF-1α 43 . Consistent with those findings, VHL-knockout (KO) Treg cells lose their suppressive function and produce excess IFNγ, whereas additional HIF-1α-KO restores Treg cells activity 18 . Similarly, persistent expression of HIF-1α by deleting PHD1, PHD2, and PHD3 in T cells leads to a significant increase in the ratio of IFN-γ + effector T cells to Treg cells 30 . Previously, we also report that Treg cells become highly unstable in vivo in the absence of the E3 ligase deltex1 that downregulates HIF-1α 44 , supporting the inhibitory role of HIF-1α on Treg cells. In addition, iTreg differentiation is inhibited by hypoxia, which can be reversed by HIF-1α deficiency 45 , further confirming the suppressive activity of HIF-1α in Treg cell differentiation. By contrast, a putative hypoxia-responsive element is found on the promoter of Foxp3, and HIF-1α-KO CD4 + CD25 + Treg cells fail to protect mice from effector T-cell-induced colitis 46 . Thus, the exact role of HIF-1α in Treg cell differentiation and action remains uncertain.
Most studies of the function of HIF-2α in immune responses have used myeloid cells. HIF-2α is induced by T helper 2 (Th2) cytokines during M2 macrophage polarization and specifically regulates expression of arginase-1 (ref. 47 ). Lack of HIF-2α in the myeloid lineage results in decreased tumor-associated macrophage infiltration and alleviates tumor progression 48 . Loss of HIF-2α from myeloid cells also increases neutrophil apoptosis and reduces neutrophilic inflammation 49 . However, the function of HIF-2α in the differentiation and function of T cells and Treg cells is unclear 50 .
Here, we show an unexpected role for HIF-2α in Treg cells. Development and the phenotype of HIF-2α-KO tTreg cells is normal, with unchanged in vitro suppressive activity. However, HIF-2α-KO Treg cells have an impaired ability to inhibit effector T cell-induced colitis and airway allergic inflammation. These defects in HIF-2α-KO Treg cells can be partly attributed to elevated HIF-1α expression, and further deletion of HIF-1α restores the suppressive function of HIF-2α-KO Treg cells. Consequently, mice with Foxp3-conditional knockout of HIF-2α are resistant to tumor growth. Our findings demonstrate an unanticipated requirement for HIF-2α in Treg function and suggest a potential approach to regulating Treg activity by targeting HIF-2α in Treg cells.

HIF-2α-KO Treg cells have normal development and phenotype.
To study the role of HIF-2α in T cells, we generated mice with T cell-specific deletion of HIF-2α (Cd4 Cre Hif2a f/f ) and Tregspecific knockout of HIF-2α (Foxp3 Cre Hif2a f/f ). Deletion of HIF-2α was confirmed by RT-PCR ( Supplementary Fig. 1). Thymus and peripheral T cell development was normal in Cd4 Cre Hif2a f/f mice ( Supplementary Fig. 2a, b). Fractions of naïve and memory T cells were not affected by HIF-2α deficiency ( Supplementary  Fig. 2c). T cell proliferation, IL-2 production, and IFN-γ generation stimulated through CD3/CD28 were normal in Cd4 Cre-Hif2a f/f T cells isolated from lymph nodes and spleen ( Supplementary Fig. 3). Therefore, HIF-2α deficiency in T cells does not affect T cell development or T cell activation. We observed a similar outcome for T cell development in Foxp3 Cre-Hif2a f/f mice ( Supplementary Fig. 4). No splenomegaly or lymphadenopathy was observed in any of the Cd4 Cre Hif2a f/f or Foxp3 Cre Hif2a f/f mice.
Treg cell development. Moreover, our experiments reveal that HIF-1α may fine-tune the differentiation of iTreg cells in a more complex way than previously recognized.
HIF-2α-KO Treg cells have normal suppressive function in vitro. Next, we determined the in vitro suppressive activity of tTreg cells. Cd4 Cre Hif2a f/f tTreg cells proved as effective as WT (Hif2a f/f ) tTreg cells in inhibiting proliferation, as well as IL-2 and IFN-γ production, of effector T cells. By contrast, HIF-1α-KO tTreg cells were ineffective in inhibiting effector T cell proliferation and secretion of IL-2 ( Fig. 1e-g). Production of IL-10 and TGF-β by Cd4 Cre Hif2a f/f tTreg cells was indistinguishable from that by Cd4 Cre Hif1a f/f and Hif2a f/f tTreg cells (Fig. 1h, i).
HIF-2α-KO tTreg cells have impaired ability to suppress colitis. Treg cells that are defective in inhibiting effector T cells in vivo may display normal in vitro suppressive activity 44,52 . We examined the in vivo suppressive activity of WT, HIF-1α-and HIF-2α-KO tTreg cells. In this assay, adoptive transfer of CD45.1 + WT CD4 + CD25 − effector T cells into CD45.1 + RAG-1-KO mice led to colitis, and co-transfer of CD45.2 + Treg cells that are congenically different from effector T cells was used to evaluate the suppressive activity of Treg cells. Effector T cells-induced colitis was examined by body weight loss, diarrhea and shortened colon length (Fig. 2a-c). Histological examination of hematoxylin and eosin (H&E)-stained colon sections also revealed inflammatory cell accumulation and tissue damage in RAG-1-KO mice that had received CD45.1 + effector T cells (Fig. 2d). Co-transfer of CD45.2 + Hif2a f/f (WT) tTreg cells protected RAG-1-KO mice from CD45.1 + effector T cell-induced body weight loss, colitis, diarrhea, tissue damage and inflammatory infiltration ( Fig. 2a-d). HIF-1α-KO tTreg cells (Cd4 Cre Hif1a f/f ) were as effective as Hif2a f/f tTreg cells in suppressing colitis and associated pathologies in RAG-1-KO mice receiving effector T cells (Fig. 2a-d).
Unexpectedly, a large number of Cd4 Cre Hif2a f/f tTreg cells were unable to inhibit effector T cell-triggered colitis ( Fig. 2a-d). We re-isolated the adoptive-transferred tTreg cells (CD45.2 + ) from RAG-1-KO mice after colitis induction and determined the expression of IL-17 and IFN-γ. We found that the fraction of IL-17 + or IFN-γ + cells among the transferred HIF-2α-KO tTreg cells (CD45.2 + ) had increased relative to WT tTreg cells (Fig. 2e). In addition, recovery of the effector T cells (CD45.1 + ) transferred into RAG-1-KO mice revealed increased IL-17 production (Fig. 2f). A large proportion of RAG-1-KO mice that received Cd4 Cre Hif2a f/f tTreg cells lost body weight and developed colitis, indicating that Cd4 Cre Hif2a f/f tTreg cells have impaired in vivo suppressive activity.
The attenuated inhibitory activity of HIF-2α-KO tTreg cells was not due to decreased Foxp3 expression as proportions of Foxp3 + cells, determined by flow cytometry, were comparable between Hif2a f/f , Cd4 Cre Hif1a f/f and Cd4 Cre Hif2a f/f tTreg cells upon recovery from RAG-1-KO mice (Fig. 2g). Therefore, HIF-2α deficiency, but not HIF-1α deficiency, impairs the in vivo suppressive function of tTreg cells that inhibit colitis.

HIF-2α-KO iTreg cells do not inhibit airway hypersensitivity.
We further extended our study to the in vivo inhibitory activity of HIF-2α-KO iTreg cells. In an allergen-induced airway inflammation assay (Fig. 4a), ovalbumin-induced airway resistance was prevented by administration of WT (Hif2a f/f ) iTreg cells, but not by Cd4 Cre Hif2a f/f iTreg cells (Fig. 4b). Ovalbumn also induced infiltration of leukocytes into bronchoalveolar lavage fluid (BALF) (Fig. 4c). Hif2a f/f iTreg cells decreased the amount of total leukocytes and eosinophils in BALF from ovalbumin-treated mice, yet Cd4 Cre Hif2a f/f iTreg cells failed to exhibit any inhibitory effect (Fig. 4c). Ovalbumin-triggered IL-4 and IFN-γ in BALF was completely suppressed by Hif2a f/f iTreg cells, in contrast to the ineffectiveness of Cd4 Cre Hif2a f/f iTreg cells in the same assay ( Fig. 4d, e). Levels of IL-6 were also fully suppressed by Hif2a f/f iTreg cells, but were only weakly inhibited by Cd4 Cre Hif2a f/f iTreg cells (Fig. 4d). Suppression of IL-5 and IL-13 in BALF was comparable between Cd4 Cre Hif2a f/f and Hif2a f/f iTreg cells. The differential in vivo suppressive activity of WT and HIF-2α-KO iTreg cells was also shown by histology. Ovalbumin re-challenge resulted in extensive inflammatory cell infiltration in perivascular and peribronchial regions, as well as bronchial wall thickening of the airway in mice sensitized with ovalbumin ( Fig. 4f, OVA).
HIF-1α expression is increased in HIF-2α-KO Treg cells. We examined the possible mechanisms by which HIF-2α deficiency impaired the suppressive function of Treg cells. HIF-1α expression ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-18731-y was increased in CD45.2 + Cd4 Cre Hif2a f/f tTreg cells recovered from CD45.1 + RAG-1-KO mice, relative to Hif2a f/f tTreg cells (Fig. 5a). Even though HIF-1α protein was not detectable in naïve T cells, increased Hif1a expression was found in freshly isolated Cd4 Cre Hif2a f/f T cells and tTreg cells (Fig. 5b, c). HIF-1α protein can be induced under normoxic conditions when T cells are continuously activated 39,53 , and we observed a more profound increase in HIF-1α protein in persistently activated Cd4 Cre Hif2a f/f CD4 + T cells compared to the corresponding Hif2a f/f CD4 + T cells (Fig. 5d). Activation by CD3/CD28 under hypoxia induced higher HIF-1α protein expression in HIF-2α-KO tTreg cells than the control tTreg cells (Fig. 5e). Activation of iTreg cells also led to greatly elevated HIF-1α levels in Cd4 Cre Hif2a f/f iTreg cells (Fig. 5f). In addition, inhibition of translation by cycloheximide illustrated that HIF-1α protein stability was increased in Cd4 Cre Hif2a f/f iTreg cells, relative to that of Hif2a f/f iTreg cells (Fig. 5g). We further examined if the HIF-2α-specific inhibitor PT2385 (refs. 54,55 ) exhibited a similar effect on HIF-1α protein as HIF-2α knockout. Treatment of activating WT CD4 + T cells with PT2385 led to dose-dependent upregulation of HIF-1α (Fig. 5h). PT2385 also increased the level of HIF-1α induction in activated WT iTreg cells (Fig. 5i). Treatment of WT tTreg cells with PT2385 also led to enhanced production of IL-17 (Fig. 5j), similar to our results depicted in Fig. 3b. Levels of the HIF-1α transcriptional targets Glut1 and Ccr4 (refs. 56,57 ) were significantly higher in Cd4 Cre-Hif2a f/f tTreg cells (Fig. 5k), along with a weak increase of Pdk1, Srebp1c, and Ccr9, even though the expression of several other HIF-1α targets was comparable between Hif2a f/f and Cd4 Cre Hif2a f/f tTreg cells (Supplementary Fig. 8). Together, these results illustrate that HIF-2α deficiency enhances HIF-1α expression in T cells and Treg cells, suggesting a possible contribution of elevated HIF-1α levels to the impaired suppressive activity of HIF-2α-KO Treg cells. Notably, Hif2a expression was increased in HIF-1α-KO CD4 + T cells and Treg cells (Fig. 5l), suggestive of mutual regulation between HIF-1α and HIF-2α.

HIF-2α-KO in Treg cells confers resistance to tumor growth.
The immunosuppressive role of Treg cells is well known, which can lead to cancer progression. Destabilization of Treg cells by a deficiency of critical Treg-associated factors such as Nrp-1, SOCS1, EZH2, Eos or CARMA1 protects hosts from syngeneic tumor growth 20,21,[31][32][33] . Recent studies have also revealed that upregulation of HIF-1α induces fragility in Treg cells having anticancer immunity 22,30 . We examined if the modest increase in HIF-1α in HIF-2α-KO Treg cells was sufficient to protect a host from cancer growth. We used Foxp3 Cre Hif2a f/f mice in which HIF-2α was only deleted from Treg cells. Inoculation of syngeneic MC38 colon adenocarcinoma led to progressive tumor growth in WT (Foxp3 Cre ) mice (Fig. 7a). By contrast, Foxp3 Cre Hif2a f/f mice were highly resistant to MC38 growth, with four of the eight mice exhibiting pronounced suppression of tumor growth and the other four mice being completely tumor-free (Fig. 7a). However, Foxp3 Cre Hif2a f/f mice were unable to suppress the growth of less immunogenic B16F10 melanoma ( Supplementary Fig. 10). We also examined if Treg-selective knockout of HIF-2α interfered with B16F10 metastasis. Intravenous administration of B16F10 resulted in lung metastases in WT mice (Fig. 7b-d). However, Treg-selective deletion of HIF-2α suppressed melanoma metastases, as revealed by overall lung morphology, numbers of tumor nodules and tissue sections (Fig. 7b-d). Therefore, conditional ablation of HIF-2α in Treg cells is sufficient to protect a host from tumor growth.
Studies have shown that administration of inflammatory Treg cells inhibits tumor growth in vivo 20 . We examined if inhibition of HIF-2α in Treg cells also conferred tumor-suppressive activity on Treg cells. Treg cells were treated with PT2385 and then intravenously delivered into mice pre-implanted with MC38 cancer. We found that PT2385-treated Treg cells inhibited the growth of MC38 cancer cells (Fig. 7e). These results suggest that administering HIF-2α-KO Treg cells may be a therapeutically viable approach to suppressing established tumors. 1 + CD4 + CD25 − effector T (Teff) cells (4 × 10 5 ) were administered into sexmatched CD45.1 + RAG-1-KO mice with or without CD45.2 + tTreg cells (1 × 10 5 ) originating from Hif2a f/f , Cd4 Cre Hif1a f/f , or Cd4 Cre Hif2a f/f mice. Body weight a and colitis scores b were assessed. Four out of the 10 RAG-1-KO mice that received Cd4 Cre Hif1a f/f tTreg cells did not lose weight, but their data have been included in a, b. Open circle, PBS; black solid circle, Teff; red solid circle, Teff + Hif2a f/f Treg; grey inverted square, Teff + Cd4 Cre Hif1a f/f Treg, blue square, Teff + Cd4 Cre Hif2a f/f Treg. Data represent a combination of two independent experiments, and were confirmed by another two independent experiments, each of which individually presented a pattern similar to that of the combined dataset. **P = 0.007 (a), 0.0012 (b), by repeat measures twoway ANOVA with Tukey's multiple comparisons test, with the Greenhouse-Geisser correction. ns, not significant. Five weeks after T-cell transfer, mice were sacrificed and colons were isolated c, and stained with hematoxylin and eosin (H&E) d. Micrographs are representative of the mice in each group. Bar indicates 100 μm. Results were reproduced in four independent experiments. e-g CD45.2 + Hif2a f/f or Cd4 Cre Hif2a f/f tTreg cells were administered into CD45.1 + RAG-1-KO mice together with CD45.1 + Teff cells. Mice were sacrificed 5 weeks later and CD4 + T cells from lymph nodes were isolated. CD4 + T cells were activated by TPA/A23187 (50/500 ng ml −1 ), Foxp3 + IL-17A + and Foxp3 + IFN-γ + subpopulations in CD45.2 + tTreg cells (e) and CD45.1 + Teff cells (f), and Foxp3 + frequency in CD45.2 + tTreg cells (g) were determined. PB Pacific blue. Left panels (e, f), representative staining plots. Right panel e, f, percentages of Foxp3 + IL-17A + and Foxp3 + IFN-γ + cells. Data are presented as the mean ± SEM, n = 3 (Hif2a f/f ) or 7 (Cd4 Cre Hif2a f/f ) e-g. Red circle, Hif2a f/f ; blue square, Cd4 Cre Hif2a f/f . *P = 0.019 (IL-17A) and 0.024 (IFN-γ) for e, P = 0.0451 (IL-17A) and 0.328 (IFN-γ) for f, P = 0.0734 g by twotailed unpaired t-test. Source data are provided as a Source data file.

Discussion
In the present study, we characterize the role of HIF-1α and HIF-2α in the development and function of Treg cells. We find that HIF-1α and HIF-2α exhibit distinct roles in Treg cells and identify an unexpected function of HIF-2α. The role of HIF-1α in Treg cells has been elucidated in many studies, but remains controversial. HIF-1α has been implicated in the expression of Foxp3 due to the presence of a putative HIF-responsive element at the promoter of the Foxp3 gene 46 . However, HIF-1α binds Foxp3 protein and downregulates Foxp3 (refs. 41,43 ), and also tunes up glycolysis to suppress Treg development 42 . Here, we find that HIF-1α knockout does not affect the development or expression of Foxp3 in tTreg cells (Fig. 1a). Our data on differentiation of iTreg cells from HIF-1α-KO CD4 + T cells (Fig. 1c, d) clearly support an inhibitory role of HIF-1α in Foxp3 expression, as previously reported 41,43 . Foxp3 levels (mean fluorescence  Hif2a f/f (WT) and Cd4 Cre Hif2a f/f tTreg cells were activated with plate-bound anti-CD3/CD28 (4/2 μg ml −1 ) and IL-2 (Th0), with an additional 50 ng/ ml of IL-12 (Th1), or an additional IL-6 (50 ng ml −1 ) plus IL-1 (IL-1α and IL-1β, 20 ng ml −1 each) (Th17) for 5 days. IFN-γ a and IL-17A b levels were measured in supernatants collected from tTreg cells reactivated by TPA/A23187 (20/200 ng ml −1 ) for 16 h. Data a, b are expressed as mean ± SD, n = 3. The frequency of Foxp3 + cells was determined by intracellular staining, and representative plots of four independent experiments are shown in the left panel of c. Right panel c, data are expressed as mean ± SEM, n = 4. Red circle, Hif2a f/f ; blue square, Cd4 Cre Hif2a f/f . **P = 0.0057 (Th0), 0.0024 (Th1); ***P < 0.001 (Th17) (b). d, e Increased conversion of HIF-2α-KO iTreg cells into IL-17-producing cells. Hif2a f/f (WT) and Cd4 Cre Hif2a f/f iTreg cells, generated as described in Fig. 1c, were rested for 2 days and then reactivated with plate-bound anti-CD3/CD28 (4/2 μg ml −1 ) under Th1-, Th2-, and Th17priming conditions as described in a. Re-programmed iTreg cells were activated by TPA/A23187, and IFN-γ (d) and IL-17A (e) production was quantitated. Data are expressed as mean ± SEM, n = 6. *P = 0.011 (e). f Normal HIF-2α-deficient T helper cell differentiation. Naïve T cells (CD4 + CD25 − CD44 lo CD62L hi ) from Hif2a f/f and Cd4 Cre Hif2a f/f mice were differentiated under either Th1-, Th2-, or Th17-polarizing conditions, as described in Methods, for 5 days. T cells were then re-stimulated with TPA/A23187 (50/500 ng ml −1 ), and expression of IFN-γ, IL-4, and IL-17 was determined by intracellular staining. Numbers represent the percentage of cells positive for IFN-γ, IL-4, or IL-17. Left, representative plots of three independent experiments are shown. Right, data are expressed as mean ± SEM. n = 3 (IFN-γ + ) or n = 4 (IL-4 + or IL-17 +) . ***P < 0.001. P-value was determined by two-tailed unpaired t-test (a-f). Source data are provided as a Source data file. intensity, MFI) in iTreg cells are higher in HIF-1α-KO iTreg cells relative to control when differentiated under normoxic conditions (Fig. 1c). For iTreg cells differentiated under hypoxic conditions, when HIF-1α protein is induced, both the proportion of Foxp3 + cells and Foxp3 MFI are significantly elevated in Cd4 Cre Hif1a f/f iTreg cells at higher concentrations of TGF-β (Fig. 1d). Similarly, both those attributes are higher in Foxp3 Cre Hif1a f/f iTreg cells developed under hypoxia than for Foxp3 Cre and Foxp3 Cre Hif2a f/f iTreg cells ( Supplementary Fig. 6), indicating that the suppressive effect of HIF-1α persists after Foxp3 induction. However, the differentiation of WT (Hif2a f/f ) iTreg cells at suboptimal conditions (0.5 ng ml −1 TGF-β) under hypoxia was greater than for Cd4 Cre Hif1a f/f and Cd4 Cre Hif2a f/f iTreg cells, but not at higher TGF-β concentrations (Fig. 1d, indicated by arrow). This scenario is analogous to a previous report showing that hypoxia increases WT iTreg differentiation in a similar setting

Ccr4 Glut1
Hif1a CD4 + T NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-18731-y ARTICLE NATURE COMMUNICATIONS | (2020) 11:5005 | https://doi.org/10.1038/s41467-020-18731-y | www.nature.com/naturecommunications (CD3/CD28, IL-2, and 0.75 ng ml −1 TGF-β), which led to the suggestion that hypoxia activates Foxp3 expression through HIF-1α and the HIF-responsive element on the Foxp3 promoter 46 . Notably, enhanced hypoxia-mediated WT iTreg differentiation at low TGF-β was not observed in iTreg cells differentiated from Foxp3 Cre T cells, relative to Foxp3 Cre Hif1a f/f and Foxp3 Cre Hif2a f/f iTreg cells ( Supplementary Fig. 6), suggesting that the enhancing effect of hypoxia is limited to Foxp3 expression as it is no longer detectable after Foxp3 is expressed. Therefore, hypoxia may promote Foxp3 expression in vitro, but that effect can only be detected in a narrow timeframe. Together with our finding that HIF-1α mediates suppression of Foxp3 levels at higher TGF-β concentrations under hypoxia ( Fig. 1d and Supplementary Fig. 6), we assert that HIF-1α exerts two opposing effects on Foxp3 expression. In iTreg cells differentiated under sub-optimal conditions in which levels of Foxp3 are low, HIF-1α may promote Foxp3 expression. Upon Foxp3 being fully expressed, HIF-1α interacts with Foxp3 protein to downregulate HIF-1α protein expression. Thus, our results may help reconcile the two longstanding opposing views 41,43,46 for how HIF-1α regulates Foxp3 expression. However, in contrast to the impaired ability of HIF-1α-KO tTreg cells to suppress colitis 46 , we observed nearly comparable capacity between WT and HIF-1α-KO tTreg cells to inhibit effector T cell-induced colitis (Fig. 2a-d). Therefore, HIF-1α-KO tTreg cells are fully functional in vivo.
Recent studies have revealed that the in vivo inhibitory activity of many different Treg cells is distinct from their in vitro suppressive activities (reviewed in ref. 52 ). As discussed above, Foxp3 expression is not affected by HIF-1α deficiency in vivo, but may be HIF-1α-dependent in vitro in particular contexts. It is possible that the difference between the suppressive activity of HIF-1α-KO Treg cells in vitro and in vivo is linked to HIF-1α-dependent in vitro Foxp3 expression, since persistent Foxp3 expression is required for Treg inhibitory activity 5,6 . It is also possible that during an in vitro suppressive analysis comprising only Treg cells, presenting cells and effector T cells, the inhibitory effect is quantitated from the close proximity between Treg cells and the target cells, but such bystander suppression would not be measured in an in vivo suppression assay that involves specific cognation 58 . Further studies are needed to examine these possibilities.
How HIF-2α regulates Treg differentiation and function had been unclear. We found that the in vitro suppressive function of tTreg cells and iTreg cells was not affected by HIF-2α deficiency, but HIF-2α-KO tTreg cells were functionally defective in suppressing effector T cell-induced colitis and in inhibiting airway inflammation. HIF-2α-KO tTreg cells and HIF-2α-KO iTreg cells that were stimulated in vitro secreted higher levels of IL-17, but not IFN-γ, relative to WT tTreg cells (Fig. 3b, e). By contrast, we observed increased expression of IL-17 and IFN-γ in HIF-2α-KO tTreg cells recovered from RAG-1-KO mice with ongoing colitis (Figs. 2e, 6c). It is known that Treg-specific deletion of a number of genes affects Treg inhibitory function in vivo but not in vitro. For example, in mice with Treg-selective deletion of CD28, Ubc13, Helios or Ezh2, impaired Treg in vivo suppressive activity is accompanied by development of spontaneous autoimmune diseases, yet in vitro Treg inhibitory function remained normal [59][60][61][62] . Mechanisms involved in the in vivo inactivation of these Treg cells include loss of competitiveness due to a survival disadvantage (CD28, Helios), reduced Treg stability (Ubc13), or destabilization of Foxp3-driven gene expression (Ezh2). Moreover, in vitro co-culture of Treg cells may not fully recapitulate the in vivo inflammatory microenvironments in which Treg cells exert their action. We speculate that the in vivo instability of HIF-2α-KO Treg cells likely contributes to the loss of their suppressive function in vivo but not in vitro. It has also been suggested that proliferating memory-type Treg cells are the main Treg populations responsible for suppression in vivo, but their functions have not been determined in an in vitro suppression assay 52 . Accordingly, HIF-2α is likely involved in the functions of memory-type Treg cells but not in other subsets of Treg cells. Further characterization will help establish the validity of that scenario.
expression of HIF-1α in Treg cells 18 , with this latter resulting in massive inflammation, excess IFN-γ production, and premature death.
In the present study, we have revealed crosstalk between the expression of HIF-1α and HIF-2α in Treg cells. Upregulation of HIF-1α in HIF-2α-KO Treg cells is mediated by both transcriptional and post-transcriptional mechanisms. Hif1a transcript was upregulated in naïve HIF-2α-KO T cells and tTreg cells relative to naïve WT tTreg cells (Fig. 5b, c), indicating that HIF-2α inhibits Hif1a transcription. Interestingly, similar upregulation of Hif2a was found in HIF-1α-KO T cells and tTreg cells (Fig. 5l). Part of the inhibitory effect of HIF-2α may be attributable to its transcriptional activity given that treatment of WT Treg cells with PT2385, which blocks the dimerization of HIF-2α and HIF-1β, recapitulated some of the phenotypes of HIF-2α-KO  Fig. 6 HIF-1α KO restores the suppressive activity of HIF-2α-KO Treg cells. a, b CD45.1 + RAG-1-KO mice were administered with CD45.1 + CD4 + CD25 − Teff cells (4 × 10 5 ) with or without CD45.2 + Hif2a f/f , Cd4 Cre Hif2a f/f or Cd4 Cre Hif1a f/f Hif2a f/f tTreg cells (1 × 10 5 ). Body weight a and colitis scores b were assessed. Three out of the 12 RAG-1-KO mice that received HIF-2α-KO tTregs did not lose weight, but their data are included in a, b. Data represent a combination of three independent experiments. PBS, n = 9; Teff, n = 7; Teff + Hif2a f/f Treg, n = 7; Teff + Cd4 Cre Hif2a f/f Treg, n = 12; Teff + Cd4 Cre Hif1a f/f Hif2a f/f Treg, n = 8. ****P < 0.0001, as determined by repeat measures two-way ANOVA with Tukey's multiple comparisons test, with the Greenhouse-Geisser correction. c, d Mice from a were sacrificed 5 weeks later and CD45.2 + CD4 + T cells were isolated, re-stimulated with TPA/A23187, and then expression of Foxp3, IFN-γ and IL-17A was determined. (c, left), representative plots; (c, right), percentages of Foxp3 + IL-17A + and Foxp3 + IFN-γ + cells; (d), Foxp3 + fraction among the CD4 + CD45.2 + T cell population. Data are presented as mean ± SEM. Hif2a f/f , n = 4 c, d; Cd4 Cre Hif2a f/f , n = 4 c, 8 d; Cd4 Cre Hif1a f/f Hif2a f/f , n = 4 (c), 8 (d). e-g B6 mice were treated with Hif2a f/f , Cd4 Cre Hif2a f/f and Cd4 Cre Hif1a f/f Hif2a f/f iTreg cells. R L in response to methacholine was measured (e, left) and data is represented as mean ± SEM (e, right). Amounts of IL-4, IL-6 (f), and IFN-γ g in the BALF were determined. PBS, n = 3 (e),12 (f, g); OVA, n = 3 (e), 9 (f, g); OVA + Hif2a f/f iTreg, n = 5 (e), 13 (f), 6 (g); OVA + Cd4 Cre Hif2a f/f iTreg, n = 8 (e), 8 (f), 6 (g); OVA + Cd4 Cre Hif1a f/f Hif2a f/f iTreg, n = 12 (e), 10 (f, g). Data are presented as mean ± SEM (f, g). *P < 0.05, **P < 0.01, ***P < 0.001 (c-g), as determined by two-tailed unpaired t-test (c, d, f, g), or two-way ANOVA with Tukey's multiple comparisons test e. Source data are provided as a Source data file. tTreg cells (Fig. 5j). HIF-1α protein is also subjected to various modifications including phosphorylation, sumoylation, deacetylation and ubiquitination that regulate its stability 65 . Absence of HIF-2α led to increased HIF-1α protein stability in iTreg cells (Fig. 5g). HIF-2α regulates over 1000 genes that are HIF-1αindependent 64 , so the modulation of HIF-1α protein stability by HIF-2α deficiency could be mediated by HIF-2α target gene products, but not necessarily by HIF-2α itself. Further works are required to map out exactly how HIF-2α modulates HIF-1α expression and stability. The inhibitory activities of Treg cells contribute to immunosuppressive microenvironments in tumors 26,27 . Depletion or destabilization of Treg cells represents a current approach to reactivating anticancer immunity 29 . HIF-1α promotes Foxp3 instability and degradation, induces IL-17 expression, stimulates IFN-γ production, and drives Treg fragility 22,30,41,43,44 . Through these activities, HIF-1α expression in Treg cells confers anticancer immunity, which protects hosts from tumor growth. Selectively increased HIF-1α expression in Treg cells could therefore be considered a therapeutic approach to treating cancer because it would destabilize Treg cells and convert them into inflammatory T cells. However, high HIF-1α expression in T cells or Treg cells would likely induce lethal autoimmunity in the host 18,30 . In the present study, we found that a moderate increase of HIF-1α expression in HIF-2α-KO Treg cells did not result in autoimmunity in Cd4 Cre Hif2a f/f or Foxp3 Cre Hif2a f/f mice. Interestingly, that moderate HIF-1α expression level in Treg cells was sufficient to protect the host from MC38 colon adenocarcinoma and metastatic B16F10 invasion (Fig. 7). We have further demonstrated that the HIF-2α inhibitor PT2385 promotes upregulation of HIF-1α in T cells (Fig. 5h, i), and that adoptive transfer of PT2385-treated Treg cells was able to inhibit growth of MC38 (Fig. 7e). Therefore, our results suggest a potential anticancer therapeutic application of moderately increasing HIF-1α expression in Treg cells to downregulate HIF-2α. HIF-2α has been shown to be overexpressed and is the oncogenic driver of clear cell renal cell carcinoma (ccRCC), and targeting of HIF-2α by HIF-2α-specific inhibitors represents a promising therapeutic strategy for treating ccRCC 54,55 . Our results may indicate that destabilization of Treg cells by HIF-2α inhibitors contributes to the anticancer effect by upregulating HIF-1α. Further experiments will help determine the exact mechanism by which reprogramming of Treg cells by HIF-2α-specific inhibitors can aid in the treatment of cancers.
In summary, we have demonstrated crosstalk between the expression of HIF-2α and HIF-1α in Treg cells, which contributes to an unexpected role of HIF-2α in Treg stability. Therapeutic applications of ex vivo-expanded Treg cells have been explored extensively [23][24][25] , including manipulation of specific gene expression in Treg cells. HIF-1α and HIF-2α are known for their prominent roles in autoimmune diseases, immunity and cancer, and reagents to modulate their expression are being examined in clinical trials or are being developed. Our results indicate that controlled expression of HIF-2α and HIF-1α in Treg cells could be used to maintain the suppressive activity of Treg cells or to generate fragile Treg cells. Further studies should pursue this research direction to optimize the therapeutic potential of Treg cells in different applications.

Methods
Preparation and characterization of Treg cells. CD4 + CD25 + thymus-derived regulatory T (tTreg) cells were purified from peripheral lymph nodes and spleens by sorting on a FACSAria II SORP system. Intracellular staining of Helios and Foxp3 was performed on paraformaldehyde-fixed splenocytes pre-labelled for CD4 and CD25 using the Foxp3/Transcription Factor Staining Buffer Set (Thermo Fisher Scientific Inc., Madison). Neuropilin-1 (Nrp-1) staining was conducted with anti-Nrp-1. Naïve IL-10 and TGF-β cytokine secretion from tTreg cells was estimated by ELISA as per the manufacturer's instructions (Thermo Fisher) upon stimulating tTreg cells with immobilized anti-CD3 (4 µg ml −1 ) and anti-CD28 (2 µg ml −1 ) for 48-96 h. Gating and sorting strategies for tTreg cells purification used in in vitro and in vivo functional analyses are described in Supplementary Fig. 11a.
In vivo iTreg suppressive assay. WT mice were intravenously injected with 2.5 × 10 6 iTreg cells or PBS on day 0, and then sensitized by intraperitoneal injections of 50 μg ovalbumin (OVA) emulsified in 2 mg of aluminum hydroxide (77161, Pierce Chemical Co., Rockford, IL) on day 1 and day 14. Three weeks later, mice were challenged by inhalation (30 min daily for four consecutive days) of aerosolized 1% OVA in saline (0.9% NaCl) generated by a Nebulizer 646 in an aerosol therapy system (DeVilbliss). To evaluate lung function, an allergy-induced airway hyperreactivity (AHR) assay was performed. Briefly, mice were anesthetized with pentobarbital (P3761, Sigma-Aldrich) at 100 mg/kg body weight after the last OVA challenge. Anesthetized mice were tracheotomized, intubated, and mechanically ventilated at a tidal volume of 0.20 mL and a frequency of 150 breaths/min. Resistive index (RI) values were recorded at baseline and following a 10 s exposure to an increasing concentration of methacholine (A2251, Sigma-Aldrich) (0, 6.125, 12.5, 25, and 50 mg ml −1 ) through the FinePointe RC System (Buxco Research Systems, Wilmington, NC). Immediately after AHR measurement, bronchoalveolar lavage fluid (BALF) was collected by three instillations of 1 mL cold saline, and amounts of IL-4, IL-5, IL-6, IL-13 and IFN-γ in BALF from the first collection were measured by cytometric bead assay (BD Biosciences, San Jose, CA). The relative number of different types of leukocytes in BALF was determined from differentials based on 200 cells by Differential Quik III Staining (Dade Behring, Deerfield, Ill). Results are given as cells ml −1 in BALF.
Quantitative PCR. Total RNA from CD4 + T or tTreg cells was collected using a GENEzol triRNA Pure Kit (Geneaid, Taiwan). The eluted RNA was reversetranscribed using SuperScript III Reverse Transcriptase (Thermo Fisher). cDNA was synthesized from 1 μg of RNA from each sample and analyzed for expression of HIF-1α, HIF-2α or HIF-1α target genes on a LightCycler 480 Real-Time PCR System using software v1.5.1.621.5.1.62 (Roche, Germany). The primers used are listed in Supplementary Table 1. The PCR protocol was 95°C for 10 min, followed by 45 cycles of 95°C for 10 s, 60°C annealing for 10 s, and 72°C extension for 8 s. Samples were normalized to Actin expression.
Cell lysates and immunoblots. To prepare whole-cell lysates, cells were lysed by whole-cell extract buffer (25 mM HEPES pH 7.7, 300 mM NaCl, 0.1% Triton X-100, 1.5 mM MgCl 2 , 0.2 mM EDTA, 0.1 mM Na 3 VO 4 , 50 mM NaF, 0.5 mM dithiothreitol and 10% glycerol) on ice for 30 min. Debris was removed by centrifuging on an Eppendoff microfuge (13,200 r.p.m., 20 min, 4°C). Protein concentrations in the supernatants were quantitated by Bio-Rad protein assay. Protein samples were resolved by SDS polyacrylamide gel electrophoresis, before being transferred to PVDF membrane in transfer buffer (30 mM Tris, 250 mM glycine, 1 mM EDTA, 20% methanol) at room temperature for 1.5 h at 400 mA. The membrane was blocked with blocking buffer (5% non-fat milk in 10 mM Tris-HCl pH 8.0, 150 mM NaCl, 0.05% Tween 20) at room temperature for 1 h. The treated membrane was incubated with primary antibodies overnight at 4°C, washed, and incubated with horseradish peroxidase-conjugated secondary antibodies at room temperature for 1.5 h. The membrane was developed using an ECL Western blotting detection kit. Luminescence was detected by X-ray film. Western blot images have been cropped for presentation. Full size images are presented in Supplementary Fig. 13.
Implantation of tumor cells. B16F10 melanoma cells were obtained from ATCC (CRL-6475, Manassas, VA). MC38 colon adenocarcinoma cells were purchased from Kerafast (CVCL_B288, Boston, MA). MC38 and B16F10 cells were cultured in DMEM (Life Technologies, Carlsbad, CA) supplemented with 10% FCS (Life Technologies), 1% Penicillin-Streptomycin, 1% L-Glutamine, and 1% HEPES. A total of 1 × 10 5 tumor cells was suspended in 100 μl PBS and inoculated subcutaneously into the right hind flank of WT (Foxp3 Cre ) or Foxp3 Cre Hif2a f/f mice. Tumor size was determined by measuring tumor length and width, using volume as readout. Volumes (V) were calculated using the equation: V = L × W 2 /2, where L is the long diameter and W is the short diameter. To evaluate the effect of HIF-2α-suppressed iTreg cells, C57BL/6 mice were lightly irradiated (2 Gy), followed by subcutaneous transplantation with 5 × 10 5 MC38 cells. Sorted CD4 + CD25 + iTreg cells were untreated or treated with 20 μM of the HIF-2α-specific inhibitor PT2385 (HY-12867, Medchem Express, Monmouth Junction, NJ) for 16 h and then adoptively transferred into mice on days 3, 6 and 10 after tumor implantation.
Evaluation of B16F10 lung metastasis. Lung metastases were established in WT or Foxp3 Cre Hif2a f/f mice by injecting 2 × 10 5 B16F10 melanoma cells in 100 μL PBS via the tail vein. On day 12, the mice were euthanized and then perfused with PBS. Numbers of tumor nodules were counted. Lungs were then fixed in 4% paraformaldehyde followed by 70% alcohol for histologic evaluation. Lung sections were stained with hematoxylin and eosin (H&E) to visualize lung architecture and malignant nodules.