Tumour suppressor death-associated protein kinase targets cytoplasmic HIF-1α for Th17 suppression

Death-associated protein kinase (DAPK) is a tumour suppressor. Here we show that DAPK also inhibits T helper 17 (Th17) and prevents Th17-mediated pathology in a mouse model of autoimmunity. We demonstrate that DAPK specifically downregulates hypoxia-inducible factor 1α (HIF-1α). In contrast to the predominant nuclear localization of HIF-1α in many cell types, HIF-1α is located in both the cytoplasm and nucleus in T cells, allowing for a cytosolic DAPK–HIF-1α interaction. DAPK also binds prolyl hydroxylase domain protein 2 (PHD2) and increases HIF-1α-PHD2 association. DAPK thereby promotes the proline hydroxylation and proteasome degradation of HIF-1α. Consequently, DAPK deficiency leads to excess HIF-1α accumulation, enhanced IL-17 expression and exacerbated experimental autoimmune encephalomyelitis. Additional knockout of HIF-1α restores the normal differentiation of Dapk−/− Th17 cells and prevents experimental autoimmune encephalomyelitis development. Our results reveal a mechanism involving DAPK-mediated degradation of cytoplasmic HIF-1α, and suggest that raising DAPK levels could be used for treatment of Th17-associated inflammatory diseases.

The Hif1a transcript is constitutively expressed in T lymphocytes, and the HIF-1a protein is detected after T-cell receptor (TCR) stimulation under hypoxic conditions 24,25 . HIF-1a is highly expressed in Th17 cells 26,27 , priming at physiological oxygen tension in the presence of inflammatory cytokines. HIF-1a plays a prominent role in Th17 cell differentiation 26,27 by activating the transcription of Rorc (RORgt), and it helps recruit CBP/p300 to the RORgt transcription complex but does not directly bind to the IL-17 promoter 27 . Additionally, HIF-1a increases glycolysis by inducing the expression of glycolytic enzymes, which further contributes to Th17 development 26,28 . HIF-1a also contributes to the survival of Th17 cells by coordination with Notch to enhance Bcl-2 expression 29 . In contrast, targeted degradation of HIF-1a by miR-210 negatively regulates Th17 differentiation 30 .
HIF-1a promotes carcinogenesis and is a prominent cancer target 18,19 . Various HIF-1a inhibitors have been identified and are currently being studied for their efficacy in cancer therapy 18,19,31,32 . Presumably, HIF-1a inhibitors could also be used for treatment of Th17-mediated inflammatory diseases. However, HIF-1a is essential for oxygen homoeostasis, and curtailment of the protective effects of HIF-1a by HIF-1a inhibitors may limit their application.
Death-associated protein kinase (DAPk/DAPK) is a multidomain serine/threonine kinase regulated by calcium 33,34 . DAPK belongs to the DAPK family, which also contains DAPK-related protein 1 and zipper-interacting protein kinase (also called DAPK3), both of which share 80% identity in their kinase domains with DAPK 33 . The DAPK family also contains two distantly related kinases: DAPK-related apoptosis inducing kinase 1 and 2 (DRK1 and DRK2) 35 . DAPK family members are pro-apoptotic proteins and function as tumour suppressors, and are specifically downregulated in many types of cancer [36][37][38][39][40][41] . DAPK participates in a wide variety of cellular events-including apoptosis, autophagy, membrane blebbing and stress fibre formation-that contribute to its tumour suppressor functions. In T lymphocytes, DAPK inhibits T-cell activation by suppressing TCR-induced nuclear factor (NF)-kB activation 42 . DAPK is induced by TGF-b (ref. 43), and is present in the early precursors of Th17, but the role of DAPK in Th17 immune cells is unclear.
In the present study, we found that DAPK negatively regulates Th17 differentiation. DAPK deficiency leads to preferential Th17 differentiation and exacerbated EAE induction. During the differentiation of Th17, the presence of DAPK is accompanied by downregulation of HIF-1a. We further found that, in contrast to the exclusive nuclear localization of HIF-1a in most other cells, HIF-1a is located in both the cytoplasm and nucleus of T cells, allowing the cytosolic binding of DAPK and subsequent HIF-1a degradation. Our results illustrate a novel mechanism of Th17 regulation by downregulating cytoplasmic HIF-1a, and suggest the therapeutic potential of increasing DAPK levels in Th17-mediated inflammatory diseases.

DAPK inhibits T-cell activation and attenuates EAE.
Our previous studies showed that expression of the dominant negative mutant of DAPK, [K42A]DAPK, leads to increased IL-2 production and NF-kB activation in T-cell lines 42 . Extending these studies, we examined T cells from the Dapk À / À mouse 44 . T-cell development was not affected by DAPK deficiency, as illustrated by the normal thymic and peripheral T-cell populations in the Dapk À / À mouse ( Supplementary Fig. 1a,b). Following TCR stimulation, Dapk À / À T-cell proliferation was enhanced relative to the wild-type (WT) T cells ( Supplementary  Fig. 2a). This was associated with increased IL-2 and interferon (IFN)-g production in naive Dapk À / À T cells ( Supplementary  Fig. 2b). An increase in IL-17 was also observed in Dapk À / À T cells (Supplementary Fig. 2c). In contrast, the production of IL-4 in primary T cells was not affected by DAPK deficiency (Supplementary Fig. 2c). Additionally, IFN-g production was similar between WT and Dapk À / À T cells when cultured in Th0 conditions ( Supplementary Fig. 2d). A prominent increase in NF-kB activation, indicated by the enhanced nuclear translocation of p65 (RelA), was also observed in Dapk À / À T cells stimulated through the TCR ( Supplementary Fig. 2e).
We further examined whether the enhanced T-cell activation in Dapk À / À mice led to increased susceptibility to autoimmune diseases, using EAE as our model. After priming with myelin oligodendrocyte glycoprotein (MOG) peptide, disease onset was earlier in Dapk À / À mice than their WT littermates (Fig. 1a). Additionally, the severity of encephalomyelitis was higher in Dapk À / À mice than their WT counterparts. This was correlated with an increase in mononuclear cell infiltration and demyelination in spinal cords from primed Dapk À / À mice (Fig. 1b,c). We also isolated T cells from the primed mice and determined their reactivity towards antigen. Consistent with exacerbated EAE generation, the response to MOG peptide in Dapk À / À T cells was nearly double that of control T cells (Fig. 1d).
We further used transgenic mice with T-cell-specific expression of [DCAM]DAPK 42 , the constitutively active form of DAPK, to analyse EAE induction. T-cell development was normal in [DCAM]DAPK-transgenic mice ( Supplementary Fig. 3a,b). We increased the dose of MOG peptide for a more profound EAE response in the control mice, allowing us to differentiate any inhibitory effect from the DAPK transgene. The expression of [DCAM]DAPK in T cells suppressed EAE generation ( Supplementary Fig. 3c). The EAE symptoms were attenuated in [DCAM]DAPK-transgenic mice in comparison with WT littermate controls. Reduced responses to MOG peptide were also found for T cells isolated from [DCAM]DAPK-transgenic mice ( Supplementary Fig. 3d). Together, these results demonstrate that DAPK deficiency enhanced EAE generation, whereas T-cellspecific expression of active DAPK suppressed EAE induction, suggesting that DAPK in T cells negatively regulates EAE pathogenesis.
Both Th17 and Th1 contribute to the development of EAE and multiple sclerosis 6 . Notably, IL-17 production, but not IFN-g generation, was increased in MOG-primed Dapk À / À T cells (Fig. 1e), suggesting a preferential expansion of DAPK-deficient Th17 cells. We examined the expression of IFN-g and IL-17 in mononuclear cells infiltrated into the spinal cords of EAEinduced mice. During the peak of EAE induction, IFN-g þ cells were more abundant than IL-17 þ cells in spinal cords from WT mice ( Supplementary Fig. 4a). Conversely, more IL-17 þ cells than IFN-g þ cells were detected in the spinal cords from Dapk À / À mice with encephalomyelitis ( Supplementary Fig. 4a). Quantitation shows an increased IL-17 þ :IFN-g þ cells ratio in EAE-inducing Dapk À / À mice ( Supplementary Fig. 4b).
We further examined the induction of EAE in Rag1 À / À mice through adoptive transfer of 2D2 T cells. The adoptive transfer of MOG-specific 2D2 cells that were differentiated into Th17 cells allows for assessment of the effect of DAPK on the induction of antigen-specific and Th17-specific EAE. Dapk À / À 2D2 Th17 cells triggered earlier disease onset and induced more severe encephalomyelitis than WT 2D2 Th17 cells (Fig. 1f). In contrast, the EAE onset and disease severity induced by Dapk À / À 2D2 Th1 cells and WT 2D2 Th1 cells were similar (Fig. 1g). Together, these results suggest that DAPK deficiency leads to an increase in encephalitogenic Th17 cells, but not encephalitogenic Th1 cells.

Dapk
Elevated RORct and RORa in Dapk À / À Th17 cells. We then determined whether the increased production of IL-17 correlated with enhanced induction of the transcription of Il17 genes. Quantitative PCR confirmed that the expression of Il17a and Il17f was elevated in DAPK-deficient Th17 cells relative to the WT Th17 cells (Fig. 2f), implying that enhanced IL-17 generation could be partly attributed to increased expression of their mRNAs. Induction of Il21 and Il23r-the cytokine and receptor preferentially expressed in Th17 cells, respectively-were also increased in Dapk À / À Th17 cells. In addition, levels of Rorc and Rora-the master transcription factors determining the expression of IL-17-were twofold higher in DAPK-null Th17 cells compared with control Th17 cells (Fig. 2f), which correlated with the elevated protein levels of RORgt in Dapk À / À Th17 cells (Fig. 2h). As a control, in the activated Th0 cells, Il17a, Il17f, Il21, WT and Hif1a À / À Th17 cells were re-stimulated and expression of IL-17A and IFN-g determined. (b) Enhanced HIF-1a protein levels in Dapk À / À Th17 cells. HIF-1a protein contents were determined in Th17 cells differentiated for 3 days (left). HIF-1a levels were quantified and normalized to tubulin (right).
Mean±s.e.m., n ¼ 3. **Po0.01 for unpaired t-test. (c) DAPK-knockout does not affect HIF-1a mRNA induction during Th17 differentiation. The quantities of Hif1a transcript were determined in WT and Dapk À / À T cells in the course of Th17 differentiation. Mean ± s.e.m., n ¼ 3. (d) DAPK induces HIF-1a protein downregulation in normal T cells. Activated primary T cells were transduced with DAPK-FLAG. GFP þ T cells were sorted, incubated under hypoxic conditions for 48 h, and the levels of DAPK-FLAG and HIF-la were determined. (e) DAPK reduces HIF-1a protein stability. HEK293T cells were transfected with DAPK and HIF-la, treated with CHX (50 ng ml À 1 ) 48 h later, and the levels of HIF-1a and DAPK were measured. HIF-1a levels were quantified.
In parallel experiments, T cells were primed with IL-12 for Th1 differentiation. Supplementary Fig. 5d illustrates that the development into Th1 cells was comparable between WT and Dapk À / À T cells. Consistent with this result, the expression of Tbx21 (T-bet) and Ifng was not affected by the deficiency of DAPK ( Supplementary Fig. 5e). Similarly, differentiation into Th1 cells was not altered by the [DCAM]DAPK transgene ( Supplementary Fig. 5f).
We also determined whether activation of the IL-17 promoter by RORgt and RORa was regulated by DAPK. To exclude interference from endogenous DAPK, DAPK was knocked down in 293T cells by siRNA targets to its 5 0 UTR ( Supplementary  Fig. 6a). IL-17P-Luc was activated in 293T cells by co-transfection with RORgt ( Supplementary Fig. 6b Fig. 6c). Therefore, DAPK likely inhibits Th17 development at two sequential stages, that is, expression of both RORgt and RORa as well as RORgt/RORamediated IL-17 expression.
DAPK downregulates HIF-1a in Th17 differentiation. It has been previously shown that HIF-1a promotes Th17 differentiation 26,27 . Consistent with this observation, diminished Th17 differentiation was found in Hif1a À / À T cells (Fig. 3a). HIF-1a is known for its participation in carcinogenesis 18,19 , while DAPK is a tumour suppressor. Therefore, we investigated whether DAPK directly modulated the expression of HIF-1a. HIF-1a was upregulated in Dapk À / À Th17 cells (Fig. 3b), which correlated with the enhanced IL-17 expression in the knockout T cells. The increased HIF-1a protein in Dapk À / À T cells was not due to an increase in Hif1a mRNA, as Hif1a transcript levels were comparable between WT and Dapk À / À T cells during the course of Th17 differentiation (Fig. 3c). Because HIF-1a expression is regulated by the mTORC1 pathway 24,47,48 , we examined whether DAPK deficiency affected mTORC1 signalling in T cells. Expression and phosphorylation of mTOR were similar between WT and Dapk À / À Th17 cells ( Supplementary Fig. 7a). Further, activation of the mTOR substrates p70 S6 kinase and 4E-BP1 in Th17 cells was not affected by DAPK deficiency (Supplementary Fig. 7a). Similarly, p70 S6 kinase-mediated phosphorylation of ribosomal S6 at S235 and S240 was comparable between WT and Dapk À / À T cells ( Supplementary  Fig. 7b,c). In addition, overexpression of DAPK downregulated the levels of HIF-1a in normal T cells cultured under hypoxic conditions (Fig. 3d). Together, these results suggest that DAPK targets HIF-1a at the post-transcriptional level.
We next examined the effect of DAPK on the stability of HIF-1a. Cycloheximide-induced HIF-1a protein instability was enhanced by the presence of DAPK in 293T cells (Fig. 3e), while HIF-1a protein stability was increased in DAPK-deficient Th17 cells (Fig. 3f), suggesting that DAPK promotes HIF-1a protein destabilization. DAPK-induced HIF-1a protein downregulation was inhibited by MG132 in 293T cells, but not by NH 4 Cl, leupeptin or chloroquine (Fig. 3g). In addition, MG132 treatment increased HIF-1a to similar levels in WT and Dapk À / À Th17 cells (Fig. 3h). These results suggest the involvement of the proteasome degradation pathway in DAPK-mediated HIF-1a downregulation in Th17 cells.
In a separate experiment, we examined whether HIF-1a protein expression could be induced by prolonged TCR activation in T cells under normoxic conditions (Fig. 3i). CD3-induced HIF-1a protein expression was much higher in Dapk À / À T cells than WT T cells (Fig. 3i), further illustrating the ability of DAPK to downregulate HIF-1a protein.
DAPK interacts with cytoplasmic HIF-1a in T cells. To isolate the mechanism underlying DAPK-dependent HIF-1a degradation, we next examined whether there was a direct interaction between DAPK and HIF-1a. DAPK is a cytosolic protein whereas HIF-1a is known for its predominant nuclear localization, as shown in hypoxic HeLa cells (Fig. 4a). HIF-1a was induced in T cells through prolonged TCR stimulation (Fig. 3i), followed by immunoblotting to determine its localization. Some HIF-1a was present in the cytosolic fraction of activated T cells (Fig. 4b).
Similarly, the presence of HIF-1a was detected in the cytoplasmic fraction of Th17 cells (Fig. 4c). The cytosolic localization of HIF-1a in Th17 cells, but not in HeLa cells, was further confirmed by confocal microscopy (Fig. 4d). In addition, DAPK co-localized with HIF-1a in the cytoplasm of hypoxic Jurkat cells (Fig. 4e), as well as in the cytosol of normoxic Th17 and Th0 cells (Fig. 4f).
Since HIF-1a and DAPK co-localized to the cytoplasm, we looked for (and found, see Fig. 4g) their direct interaction by co-immunoprecipitation. The cytoskeleton-binding domain of DAPK was required for the association with HIF-1a (Fig. 4h), while the N-terminal domain of HIF-1a interacted with DAPK ( Supplementary Fig. 9a). The association of DAPK with endogenous HIF-1a was also observed in hypoxic T cells (Fig. 4i). Together, these results suggest that DAPK interacts with HIF-1a in the cytosol of T cells, and DAPK may regulate the protein stability of HIF-1a through this direct binding. We also examined whether the kinase domain of DAPK participates in the downregulation of HIF-1a. An N-terminal DAPK fragment containing the kinase domain and ankyrin repeats weakly reduced HIF-1a expression, while a C-terminal DAPK segment from the cytoskeleton binding to death domain effectively promoted HIF-1a degradation (Supplementary Fig. 9b). In addition, we found that HIF-1a is not a kinase substrate of DAPK since recombinant HIF-1a was not phosphorylated by DAPK in vitro ( Supplementary Fig. 9c,d). Therefore, even though DAPK binds HIF-1a and induces HIF-1a protein instability, the kinase domain of DAPK appears to be disposable in HIF-1a downregulation.
DAPK induces PHD2-mediated HIF-1a degradation. The inclusion of ubiquitin enhanced the ability of DAPK to downregulate HIF-1a ( Supplementary Fig. 10a), illustrating the involvement of the ubiquitin-dependent degradation process. The stability of HIF-1a protein is regulated by proline hydroxylation, a step necessary for its subsequent binding to VHL protein leading to ubiquitination and proteasome degradation [18][19][20]22 . Therefore, we sought to identify the degradation pathways responsible for DAPK-regulated HIF-1a stability. Deletion of the HIF-1a oxygen-dependent degradation domain containing the proline residues necessary for hydroxylation conferred resistance to DAPK-induced degradation (Fig. 5a). Similarly, inhibition of PHD1/2, which hydroxylates HIF-1a protein, by deferoxamine mesylate (DFX) restored the expression levels of HIF-1a protein in DAPK-expressing T cells and in 293T cells ( Fig. 5b; Supplementary Fig. 10b). This result was consistent with the decreased hydroxylation of HIF-1a protein at proline 564 in Dapk À / À T cells (Fig. 5c) and the increased proline 564 hydroxylation in DAPK-overexpressing Jurkat cells (Fig. 5d). Mutation of the hydroxylation sites Pro402 and Pro564 on HIF-1a into alanine (2PA) conferred resistance of HIF-1a protein to DAPK-induced degradation in T cells and in HEK293T cells  Supplementary Fig. 10c). We confirmed that DAPK deficiency did not affect the levels of PHD1 and PHD2 in T cells (Fig. 5f). However, knockdown of PHD2 prevented DAPKinduced HIF-1a degradation in T cells (Fig. 5g). Taken together, these results suggest that DAPK-induced HIF-1a downregulation is partly mediated by PHD2-triggered proline hydroxylation on HIF-1a. We further detected an interaction between DAPK and PHD2 in T cells (Fig. 6a). Co-localization of DAPK with PHD2 was observed in Th17 cells by confocal microscopy (Fig. 6b). In addition, PHD2 co-localized with HIF-1a in the nucleus and cytoplasm of Th17 cells (Fig. 6c). We also identified the N-terminal part (aa 1-180) of PHD2 as the DAPK-interacting region (Fig. 6d). Addition of recombinant DAPK protein enhanced the interaction of GST-HIF-1a and FLAG-PHD2 in an in vitro binding assay (Fig. 6e). Therefore, by binding to both HIF-1a and PHD2, a possible role for DAPK is to enhance the association of PHD2 with HIF-1a and increase HIF-1a degradation.
HIF-1a knockout inhibits Dapk À / À Th17 cell differentiation. We hypothesized that if the excess generation of IL-17 in Dapk À / À T cells is due to enhanced HIF-1a-mediated Th17 differentiation, the additional deletion of HIF-1a should reduce IL-17 production to normal levels in Dapk À / À T cells. We generated mice with Dapk À / À Hif1a À / À by crossing mice with T-cell-specific deletion of HIF-1a (Cd4 Cre X Hif1a f/f ) with Dapk À / À mice. Deletion of HIF-1a and/or DAPK in T cells from each genotype was confirmed by immunoblotting (Fig. 7a). In agreement with our hypothesis, we found that while Dapk À / À T cells exhibited enhanced differentiation of naive CD4 þ T cells into Th17 cells, Dapk À / À Hif1a À / À T cells displayed levels similar to WT T cells (Fig. 7b). Knockout of HIF-1a also decreased the levels of RORgt in Dapk À / À Th17 cells (Fig. 7c). Therefore, elevated HIF-1a levels partly account for the increased IL-17 expression in Dapk À / À T cells. We further examined whether selective deletion of HIF-1a in T cells attenuated the exacerbated EAE generation in Dapk À / À mice. The severe EAE disease scores in Dapk À / À mice were largely attenuated by T-cell-selective deletion of HIF-1a (Fig. 7d), supporting that DAPK acts upstream of HIF-1a in EAE development. T-cell-specific deletion of HIF-1a did not affect the development and suppressive activity of nTreg in Dapk À / À mice ( Supplementary Fig. 11a,b). The reduced development of Dapk À / À iTregs was restored by HIF-1a knockout ( Supplementary Fig. 11c), yet the impaired suppressive activity was not corrected in Dapk À / À Hif1a À / À iTregs ( Supplementary Fig. 11d). Together, these results suggest that DAPK-HIF-1a interaction contributes little to maintain normal Treg functioning, in contrast to a prominent role of DAPK-HIF-1a antagonism in Th17 cells.

Discussion
In this study, we identified a negative role of DAPK in Th17 development. In Dapk À / À T cells, generation of IL-17A, IL-17F, IL-21 and IL-23R were highly elevated. We further found that the levels of RORgt and RORa-the transcription factors dictating IL-17 expression-were elevated in DAPK-deficient Th17 cells; even though activation of the Rorc inducer STAT3 was similar between WT and Dapk À / À T cells ( Fig. 2i; Supplementary  Fig. 5g). Instead, increases in the expression of RORgt and RORa could be partly attributed to elevated levels of HIF-1a protein.
In addition, phosphorylation of mTOR, p70 S6 kinase, 4E-BP1 and ribosomal S6 were comparable between WT and Dapk À / À Th17 cells ( Supplementary Fig. 7), suggesting that the increased levels of HIF-1a in Dapk À / À Th17 cells are not caused by increased mTOR signalling. We further found that DAPK promoted the destabilization of HIF-1a protein in a proteasome-dependent manner (Fig. 3). DAPK expression is induced by TGF-b (ref. 43), which is commonly used for Th17 differentiation 10,11 . This may partly explain why DAPK levels were higher in Th17 cells than Th1 cells (Fig. 2a). The lower DAPK expression in Th1 cells likely correlated with the weak effect of DAPK deficiency in Th1 cells. Even though naive Dapk À / À T cells displayed increased production of IFN-g ( Supplementary Fig. 2b), IFN-g expression was not enhanced in re-stimulated DAPK-deficient T cells, including T cells optimized for Th0 and Th1 differentiation ( Supplementary Figs 2d and 5d). In addition, IFN-g production was comparable in MOG-primed T cells from WT and Dapk À / À mice (Fig. 1e). Similarly, transgenic expression of DAPK did not affect Th1 development (Supplementary Fig. 5f).
DAPK deficiency led to enhanced sensitivity to EAE induction (Fig. 1a-c). We found that the ratio of IL-17-expressing cells to IFN-g-expressing cells was significantly increased in spinal cords from Dapk À / À mice during peak EAE induction ( Supplementary Fig. 4), suggesting enhanced spinal cord infiltration of Dapk À / À Th17 cells. We also observed that transfer of Dapk À / À 2D2 Th17 cells induced highly exacerbated EAE in comparison to WT Th17 cells (Fig. 1f). DAPK-deficient Th17 cells thus displayed increased encephalitogenic activity relative to WT Th17 cells. In contrast, EAE induction by Dapk À / À 2D2 Th1 cells was similar to that induced by WT 2D2 Th1 cells (Fig. 1g). Therefore, DAPK deficiency correlates with preferential induction of pathogenic Th17.
We have previously shown that DAPK specifically suppresses TCR-induced NF-kB activation 42 , as confirmed here by the elevated RelA nuclear translocation in Dapk À / À T cells (Supplementary Fig. 2e). NF-kB activation in T cells is required for the generation of autoimmune encephalomyelitis; mice deficient in NF-kB1 (p50) are resistant to the induction of EAE 50 , whereas PKCy deficiency prevents T-cell activation and the pathogenesis of EAE 51 . NF-kB has been shown to directly activate the Hif1a promoter 52 . In the present study, we found that induction of Hif1a transcripts during Th17 differentiation was not affected by Dapk deficiency (Fig. 3c), indicating that enhanced NF-kB activation in Dapk À / À T cells did not lead to a further increase in Hif1a expression. We speculate that the levels of NF-kB activation in WT T cells were already sufficient for optimal transcriptional activation of Hif1a mRNA.
A previous study of DRAK2 showed that deletion of the gene Stk17b2, which encodes DRAK2, results in complete resistance to EAE induction 35 . DRAK2 knockout is accompanied by increased T-cell activation, similar to Dapk-deficient T cells 35 . However, TCR-induced NF-kB activation is selectively enhanced in Dapk À / À T cells ( Supplementary Fig. 2e), but is attenuated in DRAK2-null T cells 35 . In addition, DRAK2 consists only of a catalytic domain, lacking the cytoskeleton-binding domain 33 by which DAPK interacts with HIF-1a (Fig. 4h). We also found that a DAPK kinase domain fragment was not efficient in promoting HIF-1a degradation (Supplementary Fig. 9b). It is likely that the different roles in EAE induction of DAPK and DRAK2 could be partly attributed to their differential effects on NF-kB activation and HIF-1a downregulation.
A novel finding in the present study is that DAPK interacts with HIF-1a and promotes the degradation of HIF-1a in T cells. We reveal that, in contrast to the predominant nuclear localization of HIF-1a in HeLa cells, HIF-1a protein was present in both the cytoplasm and nucleus of T cells (Fig. 4a-f). The cytosolic location of HIF-1a protein permits direct interaction with, and subsequent destabilization by, DAPK. Therefore, DAPK exhibits an additional ability to regulate HIF-1a protein in T cells by its ability to associate with HIF-1a. We further show that DAPK increased the proline hydroxylation of HIF-1a protein (Fig. 5c,d), and HIF-1a proteins lacking the hydroxylation sites were resistant to DAPK-induced degradation (Fig. 5a,e). Knockdown of PHD2 also eliminated the ability of DAPK to downregulate HIF-1a (Fig. 5g), suggesting that the binding of DAPK to HIF-1a protein leads to proline hydroxylation of HIF-1a and subsequent HIF-1a ubiquitination by VHL.
HIF-1a is pivotal for Th17 differentiation. The specific downregulation of HIF-1a protein by DAPK leads to preferential enhancement of Th17 development in the absence of DAPK-null T cells. This is further supported by the fact that additional HIF-1a knockout eliminated the excess Th17 differentiation and encephalitogenic activity of Dapk À / À T cells (Fig. 7). HIF-1a protein is also known to promote Foxp3 degradation and negatively regulates Treg differentiation 27,49 . Despite that differentiation and functioning of iTreg from Dapk À / À T cells were impaired (Supplementary Fig. 8c,d), knockout of HIF-1a did not restore functioning of Dapk À / À iTregs ( Supplementary  Fig. 11d), suggesting a minor role of DAPK-HIF-1a antagonism in Treg cells. Therefore, the ability of DAPK to inhibit HIF-1a predominantly regulates Th17 cells, and impaired Treg function do not seem to contribute to the enhanced Th17 development and severe EAE disease observed in Dapk À / À mice. Why Th17 and Tregs display different susceptibility to regulation by the DAPK-HIF-1a interaction is currently being investigated.
Recent studies have revealed that hypoxia promotes the expression of miR-103/107 that target DAPK for downregulation in cancer cells 53 . HIF-1a induces the expression of KLHL20, which coordinates with Cul3 for ubiquitination and proteasome degradation of DAPK in several cancer cell lines 54,55 . Together with the results from the present study, we propose a mutual antagonism between HIF-1a and DAPK. We suggest that the continuous cross-interaction between HIF-1a and DAPK, modulated by specific environmental cues, may determine the patho-physiological status of Th17 cells. DAPK is a tumour suppressor known to be downregulated in many types of cancers [36][37][38][39][40][41] . Our results illustrate that DAPK is not only a tumour suppressor, but also plays a role in the prevention of Th17-mediated autoimmune diseases like EAE. DAPK suppresses tumour formation and metastasis through its promotion of apoptosis, induction of autophagy and regulation of cell motility 56 . In the present study, we have demonstrated that DAPK inhibits EAE development by inducing HIF-1a degradation and inhibiting pro-inflammatory Th17 cell differentiation. Notably, HIF-1a is also known for its prominent role in tumorigenesis, with levels of HIF-1a correlating with poor prognoses for various cancer types 18,19 . Even though HIF-1a is exclusively localized in the nucleus of most cancer cells, it will be interesting to examine whether the ability of DAPK to promote HIF-1a degradation in the cytoplasm also contributes to the tumour suppressing function of DAPK in selective cancer types. We found that HIF-1a was present in the cytoplasm of T-cell leukaemia Jurkat cells and was subjected to DAPK regulation (Figs 4e and 5e), suggesting that DAPK likely exhibits a tumour suppressor effect in T leukaemia cells by directly binding to the cytosolic HIF-1a to prompt HIF-1a downregulation.
Th17 cells contribute to autoimmune diseases-including multiple sclerosis, rheumatoid arthritis, inflammatory bowel diseases, psoriasis and autoimmune diabetes 1-3 , and Th17 is a target for immunotherapy in these inflammatory diseases. In addition to promoting Th17 differentiation, HIF-1a contributes to the inflammatory activities of different innate immune cells including macrophages and neutrophils 22 . Therefore, targeted HIF-1a degradation by DAPK upregulation is a potential novel approach in the treatment of Th17-associated inflammatory diseases.
The critical role of HIF-1a in carcinogenesis and metastases has led to the identification of drugs that inhibit HIF-1a (refs 18,19,21,31,32). Conceivably, HIF-1a inhibitors are also potential therapeutic agents for HIF-1a-mediated diseasesincluding pulmonary arterial hypertension, hereditary erythrocytosis, obstructive sleep apnea, ocular neovascularization and traumatic shock 19,20 -as well as Th17-mediated inflammatory diseases. However, HIF-1a is critical for oxygen homoeostasis, and is protective in coronary artery disease, peripheral artery disease, wound healing, colitis and organ transplantation rejection [19][20][21] . Thus, the therapeutic benefits of HIF-1a inhibitors are expected to be offset by the loss of HIF-1amediated protective responses. In the present study, we used Th17 as an example to demonstrate an alternative approach to downregulating HIF-1a. The cytoplasmic presence of HIF-1a permits its interaction with DAPK in the cytosol of T cells, leading to HIF-1a downregulation. DAPK-induced cytosolic HIF-1a degradation presumably does not operate in cells exhibiting exclusively nuclear HIF-1a, leaving the HIF-1a in those cells unaffected. Compounds that activate DAPK have been identified 41 , and further improvements in the specificity of drugs for DAPK induction may lead to reagents capable of effectively downregulating HIF-1a in DAPK-dependent processes. These drugs could potentially be used for treatment of Th17-mediated inflammatory diseases and T-cell leukaemia. Future investigations will help validate the therapeutic approach of inhibiting HIF-1a pathogenic activity in T cells through DAPK upregulation, without affecting the HIF-1a in other cells.
Histology on spinal cord sections. Mice were perfused with phosphate buffer saline (PBS) followed by 4% (v/v) paraformaldehyde (PFA) in PBS. Spinal cords were isolated and fixed with 4% (v/v) PFA in PBS overnight at 4°C. Spinal cords were dehydrated with 30% sucrose in PBS for 48 h at 4°C. Spinal cords were embedded in optimal cutting temperature compound and serial cryosections (30 mM thick) were acquired. Tissue sections were stained with hematoxylin & eosin or luxol fast blue. Images were obtained on a Zeiss Axio Imager Z1 microscope (Jena, Germany) with an objective lens of plan-Apochromat 10 Â /0.45 M27 and the camera of Axiocam 506. The channel was TL Brightfield and the depth of focus was at 5.43 mM.
Western blot. Cell lysates or immunoprecipitates were resolved by SDS-PAGE and were transferred to PVDF membrane (Millipore) using transfer buffer (30 mM Tris, 250 mM glycine, 1 mM EDTA, 20% methanol) at 4°C for 2 h at 400 mA. The membrane was treated with block buffer (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, followed by horseradish peroxidase-conjugated secondary antibodies in blocking buffer at room temperature for 1.5 h. The membrane was developed using ECL western blotting detection kit (GE Healthcare), and the chemiluminescence detected by X-ray film. Western blot images have been cropped for presentation. Full size images are presented in Supplementary Figs 12-17. cells were collected and lysed with WCE buffer (25 mM pH 7.7 HEPEs, 300 mM NaCl, 0.1% Triton X-100, 1.5 mM MgCl 2 , 0.2 mM EDTA, 0.1 mM Na 3 VO4, 50 mM NaF and 0.5 mM DTT). Recombinant FLAG fusion proteins were purified with anti-FLAG M2 agarose. Purified recombinant proteins were incubated in PBS at 4°C for 4 h, and then incubated with anti-GST antibody overnight at 4°C. Samples were incubated with protein G magnetic beads (GE) at 4°C for 1 h. The immune complexes were washed with WCE buffer and examined by western blot.
Statistics. Our data were randomly collected but were not blinded. We did not exclude any data in this study. Our data mostly meet the assumption of the tests (normal distribution). Microsoft Office Excel and Prism 5.0 (GraphPad software) were used for data analysis. Unpaired two-tailed Student t-tests were used to compare results from between two groups. Data were presented as mean with standard deviation (s.d.) or standard error of the mean (s.e.m.), as indicated in the figure legend.
Data availability. All relevant data are available from the authors upon request.