Endothelial PAS domain protein 1 activates the inflammatory response in the intestinal epithelium to promote colitis in mice.

Hypoxia is a well-conserved transcriptional stimuli leading to an adaptive increase in the cellular response to limited oxygen availability. Inflammatory foci are hypoxic, and this focal hypoxia can drive the inflammatory cascade 1 , 2 . The significance of this bidirectional crosstalk and the mechanistic role of hypoxia in inflammation are currently unclear. Hypoxia-induced signal pathway is transcriptionally mediated by hypoxiainducible factor (HIF) 3 , 4 . In oxygen rich conditions, prolyl hydroxylases (PHD) hydroxylate HIF leading to its degradation by the von Hippel-Lindau tumor suppressor protein (VHL) coupled to the E3 ubiquitin ligase complex. Conversely, in conditions of low cellular O 2 HIF is not hydroxylated and stabilized. Two transcriptionally active HIFs have been identified, HIF-1α and endothelial PAS domain protein 1 (EPAS1, also referred to as HIF-2α). They share a 48% sequence identity and regulate overlapping and distinct sets of genes critical in the adaptation to hypoxic environments. A robust activation of HIF-1α and EPAS1 is observed in the intestinal epithelium from inflammatory bowel disease (IBD) patients 5 . IBD is an inflammatory disease of the intestine, which is grouped into two major types, Crohn's disease (CD) and ulcerative colitis (UC). While the etiology of CD and UC are not known, the dysregulation in intestinal epithelial barrier function and the mucosal immune response are critical in the pathogenesis of IBD. Recent studies demonstrate that HIF-1α in intestinal epithelial cells is critical in maintaining the intestinal epithelial barrier following inflammation 6 - 10 . Moreover, pharmacological activation of HIF signaling using PHD inhibitors is protective in mouse models of IBD 11 , 12 . Therefore, it is still unclear the mechanism by which hypoxia exacerbate the inflammatory cascade 2 , 13 . In this manuscript, using novel mouse models of intestinal HIF signaling, we demonstrate that EPAS1 is highly activated in epithelial cells from IBD patients and mouse models of colitis. Genetic activation of EPAS1 in epithelial cells leads to an increase in epithelial-derived pro-inflammatory mediators, which are critical in the initiation and progression of IBD.

Vhl F/F , Vhl ΔIE , Epas1 F/F , Epas1 ΔIE , Vhl F/F / Hif-1 α F/F , Vhl ΔIE / Hif-1 α ΔIE , Vhl F/F / Epas1 F/F , Vhl ΔIE / Epas1 ΔIE and ODD-luciferase mice were described previously 13 , 14 , 18 , 19 , 53 . LSL-HIF1 dPA and LSL-HIF2 dPA mice are previously described 20 . The LSL-HIF1 dPA and LSL-HIF2 dPA mice were crossed with murine villin promoter driven cre mice 21 to generate intestinal epithelial specific HIF-1α and EPAS1 overexpressing mice ( Hif-1α LSL/LSL and Epas1 LSL/LSL ), respectively. Hif-1 α LSL/LSL and Epas1 LSL/LSL mice were crossed to each other to generate intestinal epithelial specific HIF-1α and EPAS1 double overexpressing mice ( Hif-1 α LSL/+ / Epas1 LSL/+ ). For DMOG study, 6- to 8-week-old ODD-luciferase mice (n=3) were intraperitoneally (i.p.) injected with 400 mg/kg DMOG daily or saline for two days, a regimen previously shown to stabilize HIF-1α in vivo 11 . For TNF-α inhibition study, 6- to 8-week-old Vhl F/F and Vhl ΔIE mice were subcutaneously injected with 5 mg/kg Enbrel (Immunex Corporation, Thousand Oaks, CA) 4 hours prior to C. rod infection, and at day 1 and day 4 following C. rod infection or i.p. injected with 10 mg/kg Remicade (Janssen Biotech, Inc., Horsham, PA) one day and 4 hours prior to and every other day following C. rod infection or DSS treatment. All mice were maintained in standard cages in a light and temperature-controlled room and were given standard chow and water ad libitum. All animal studies were carried out in accordance with Institute of Laboratory Animal Resources guidelines and approved by the University Committee on the Use and Care of Animals at the University of Michigan (UCUCA approval number: 10299).

In order to test the presence of hypoxia in IBD, the oxygen dependent degradation (ODD)-luciferase hypoxia reporter mouse model was assessed following induction of acute colitis 14 . The ODD-luciferase mice were treated with Citrobacter rodentium ( C. rod ) a model of enteric infection-induced colitis or dextran sulfate sodium (DSS) an injury-induced colitis model 15 , 16 . In vivo imaging demonstrated that the luciferase signal is increased in the abdomen of C. rod or 3% DSS treated mice ( Figure 1A ). This increase was consistent with a robust activation of luciferase signal in excised colons from the ODD-luciferase mice following C. rod or DSS treatment. EPAS1 expression was further assessed in colon samples from both murine experimental colitis and IBD patients by immunohistochemical staining ( Figure 1B ). In untreated mice and normal control human intestines there was a diffuse background staining by the EPAS1 antibody. However, it is clear that in inflamed mouse colon following C. rod or 3% DSS treatment or in UC or CD samples there was a robust increase in nuclear EPAS1 staining, which co-localizes to an epithelial specific marker, E-Cadherin. Western blot analysis further confirmed an increase in EPAS1 expression in individual UC and CD patients ( Figure 1C ). These results indicate that EPAS1 may play a critical role in the progression of IBD. Since EPAS1 expression was rapidly activated in epithelial cells in acute models of colitis and predominantly observed in epithelium in UC and CD, mice with an intestinal epithelial-specific deletion of Epas1 ( Epas1 ΔIE ) were assessed following DSS treatment ( Figure 2 , Supplemental Figure 1 ). Following 7-days of 3% or 5% DSS treatment no difference in body weight was observed between Epas1 ΔIE and Epas1 F/F mice ( Supplemental Figure 1, A-C ). Western blot analysis from scraped mucosal samples demonstrated that the expression of EPAS1 in the colon was induced in wild-type littermate mice ( Epas1 F/F ) with 3% or 5% DSS treatment for 3 days, while this induction was inhibited in Epas1 ΔIE mice ( Figure 2A , Supplemental Figure 1D ). This data provides further evidence that an increase of EPAS1 following acute insult is primarily due to EPAS1 activation in intestinal epithelial cells. A treatment of 5% DSS for 7-days did not lead to a difference in the severity of colitis between Epas1 ΔIE and Epas1 F/F mice ( Supplemental Figure 1, E and F ). However, in a moderate model of colitis induced by 3% DSS the Epas1 ΔIE mice were protected from DSS-induced injury as assessed by colon length ( Figure 2B ). The disease activity index (DAI) based on rectal bleeding, diarrhea and histology were significantly improved in the Epas1 ΔIE mice compared to Epas1 F/F mice ( Figure 2C ). Colons from the Epas1 ΔIE mice had a decrease in inflammatory infiltrates, and damage to crypts and surface epithelium compared to Epas1 F/F mice treated with 3% DSS for 7-days ( Figure 2D ). Together this data suggest that intestinal epithelial EPAS1 is an important transcription factor in the pathogenesis of colitis. To assess if activation of EPAS1 potentiates colitis, mice with an intestinal epithelium-specific disruption of Vhl or double disruption of Vhl and Epas1 were assessed. Disruption of Vhl activates HIF signaling by stabilizing both HIF-1α and EPAS1, whereas the double disruption of Vhl and Epas1 prevents the formation of functional EPAS1 with a slight compensatory upregulation of HIF-1α ( Supplemental Figure 2 ) 13 , 17 - 19 . Intestinal epithelium-specific deletion of Vhl ( Vhl ΔIE ) increased the susceptibility of mice to C. rod - and 3% DSS-induced colitis ( Figure 3 ). The Vhl ΔIE mice demonstrated significant decrease in body weight and the experiment was stopped following 11 days of C. rod infection and 5 days of 3% DSS treatment ( Figure 3A ). A decrease in the colon length in the Vhl ΔIE mice compared to wild-type littermates ( Vhl F/F ) was observed ( Figure 3B ). Intestinal epithelial-specific deletion of both Vhl and Epas1 ( Vhl ΔIE / Epas1 ΔIE ) did not demonstrate any overt intestinal phenotype, and completely ameliorated the decrease in body weight and colon length in acute models of colitis ( Figure 3, A and B ). Accordingly, histological assessment confirmed severe inflammation in Vhl ΔIE , while only very limited inflammation could be observed in Vhl ΔIE / Epas1 ΔIE mice ( Figure 3, C and D ). Vhl ΔIE / Hif-1 α ΔIE mice were indistinguishable from Vhl ΔIE mice following DSS treatment ( Supplemental Figure 3 ). These results provide evidence that activation of EPAS1 potentiates inflammation and suggest the increase in EPAS1 expression that is observed in UC and CD exacerbates disease progression. VHL disruption is an indirect mouse model to assess HIF signaling. To directly understand the role of HIF in the intestinal epithelium, mouse models with intestinal epithelial-specific overexpression of HIF-1α or EPAS1 were generated. An oxygen stable Hif-1α and Epas1 cDNA was expressed downstream of a loxP-stop-loxP cassette 20 . These mice were crossed to villin-Cre transgenic mice 21 to overexpress HIF-1α ( Hif-1α LSL/LSL ) or EPAS1 ( Epas1 LSL/LSL ). Western blot analysis from scraped intestinal epithelial cells confirmed specific overexpression of HIF-1α (not EPAS1) in the Hif-1α LSL/LSL or EPAS1 (not HIF-1α) in the Epas1 LSL/LSL mice ( Figure 4A ). The expression of HIF-1α target gene Pdk1 22 was only induced in intestinal tissues from Hif-1α LSL/LSL mice, whereas the expression of EPAS1 target gene Dmt1a 17 was only induced in Epas1 LSL/LSL mice ( Figure 4B ). The Epas1 LSL/LSL mice were smaller and their body weights were significantly lower compared to wild-type littermates (WT) ( Figure 4, C and D ). No differences were observed in Hif-1 α LSL/LSL mice compared to their littermate controls (data not shown). In addition, Epas1 LSL/LSL mice did not survive past 40 days after birth, whereas the WT and Hif-1α LSL/LSL mice were healthy at the observation time of 3 months after birth ( Figure 4E ). H & E staining showed no histological signs of inflammation in control littermates and Hif-1α LSL/LSL mice in the colon and small intestine, but severe spontaneous inflammation was found in Epas1 LSL/LSL mice in both the small intestine and colon ( Figure 4F and data not shown). Bromodeoxyuridine (BrdU) incorporation and terminal deoxynucleotidyl transferase-mediated dUTP nick end labeling (TUNEL) assay revealed no significant difference in cell proliferation and cell apoptosis between the colon tissues of WT and Epas1 LSL/LSL mice ( Supplemental Figure 4 ). However, immunohistochemical staining revealed an increase of the macrophage marker F4/80 in the colon tissues of Epas1 LSL/LSL mice compared to WT mice. These data demonstrate that epithelial activation of EPAS1 induces overt inflammation in the intestine. The Epas1 LSL/LSL mice demonstrated severe overt inflammation and decreased survival. However Epas1 LSL/+ mice overexpressing EPAS1 from one allele, did not display any overt phenotypes as assessed up to 3 months of age (data not shown). Therefore, these mice were further assessed following C. rod and DSS treatment ( Figure 5 ). The Epas1 LSL/+ mice demonstrated significant decrease in body weight and the experiment was stopped following 7 days of C. rod infection and 5 days following 3% DSS treatment ( Figure 5A ). A decrease in the colon length in the Epas1 LSL/+ mice compared to WT was observed ( Figure 5B ). Histological scoring confirmed severe inflammatory cells infiltration and tissue damage in Epas1 LSL/+ mice, while only moderate inflammation could be observed in WT ( Figure 5, C and D ). These data demonstrate that EPAS1augments the development of colitis. Interestingly, double overexpression of HIF-1α and EPAS1 ( Hif-1α LSL/+ / Epas1 LSL/+ ) lead to a phenotype very similar to the Epas1 LSL/+ mice ( Supplemental Figure 5 ). The Hif-1α LSL/+ / Epas1 LSL/+ mice had no overt phenotype at the age assessed. However, the Hif-1α LSL/+ / Epas1 LSL/+ mice had an increased susceptibility to an acute colitis model, similar to that observed in the Epas1 LSL/+ mice. This data suggests that when HIF-1α and EPAS1 are expressed in the same cell the pro-inflammatory role of EPAS1 is the dominant response in the intestinal epithelial cells. To investigate the detailed molecular mechanisms for the regulatory role of EPAS1 in experimental colitis, gene expression profiles following DSS treatment between Epas1 ΔIE and Epas1 F/F mice were compared by microarray (Gene Expression Omnibus database accession number: GSE43416 http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE43416 ). In contrast to what has been shown with HIF-1α 6 - 8 , 23 , EPAS1 did not regulate genes important in the intestinal epithelial-barrier as demonstrated by the microarray and qPCR analysis (data not shown). Functional annotation analysis suggested a significant decrease in the pro-inflammatory response in Epas1 ΔIE mice compared to Epas1 F/F mice following 3% DSS treatment for 3 days. qPCR analysis confirmed marked reductions of mRNAs encoding IL-6, IL1RL1, IL-1β, TNF-α and MIF in Epas1 ΔIE mice ( Supplemental Figure 6A ). Whereas overexpression of EPAS1 increased expression of pro-inflammatory meditators, which was further potentiated in acute colitis models ( Supplemental Figure 6 and 7 ). The gene expression analysis demonstrates an increase in TNF-α specifically by EPAS1 (but not HIF-1α) in epithelial cells, which precedes overt colonic inflammation ( Supplemental Figure 8A and B ). Activation of Tnf-α expression by EPAS1 is similar to that of direct targets such as Ptgs2 and Ptges 24 . To assess direct regulation of TNF-α by hypoxic signaling, human embryonic stem cell (hESC)-derived intestinal organoids (HIOs) were assessed 25 - 28 . The intestinal organoid cultures provide an ideal model system to assess hypoxic activation of TNF-α in the context of normal human intestinal epithelial cells. HIOs were incubated in 1% O 2 (hypoxia) or 21% O 2 (normoxia) with or without EPAS1 inhibitor piplartine (PIP) or acriflavine (ACF) for 24-hours 29 , 30 . qPCR analysis demonstrated that the expression of DMT-1a and TNF-α was significantly induced by hypoxia and inhibited by PIP and ACF ( Figure 6A and Supplemental Figure 8C ). Furthermore, EPAS1 but not HIF-1α activated the proximal promoter of Tnf-α in HCT116 cells, whereas both EPASI and HIF-1α activated the canonical hypoxia response element (HRE)-luciferase construct p2.1 ( Figure 6B ). These data demonstrate the specificity of HIF-2α for the regulation of TNFα maintained in an in vitro promoter assay, and in vivo in intestinal epithelial cells. The nuclear factor-kappaB (NF-κB) p65 transcription factor is a master regulator of Tnf-α expression 31 . Repressing the nuclear translocation of NF-κB with a super-repressor form of IκBα (SR-IκBα) could suppress TNF-α mediated activation of the NF-κB promoter but did not reverse EPAS1-mediated activation of the Tnf-α promoter 32 ( Supplemental Figure 9, A and B ). Moreover, the NF-κB pathway was not activated in mice overexpressing intestinal EPAS1 ( Supplemental Figure 9C ). Taken together, these results provide evidence that EPAS1 is a critical transcription factor regulating intestinal epithelial inflammatory response independent of the NF-κB pathway. EPAS1 responsive element was narrowed down to 50 bp region by 5’ promoter deletion constructs ( Figure 6B ). This region however contains no HREs, suggesting a novel EPAS1-dependent mechanism. By analyzing the sequence of this Tnf-α promoter region with MatInspector ( www.genomatix.de ), several putative transcription factor-binding sites were identified. Deletion of the myc-associated zinc finger protein (MAZ) binding site, but not myelin transcription factor (MYT) or myeloid zinc finger protein (MZF) sites abolished EPAS1-mediated activation of TNF-α ( Figure 6C ). Overexpression of MAZ together with EPAS1 potentiated the transcriptional activity of TNF-α in a dose dependent manner ( Figure 6D ). Moreover, in cells in which MAZ was reduced by shRNAs, the EPAS1-induced TNF-α promoter activity was attenuated ( Figure 6E ). To assess MAZ DNA-binding activity, we performed DNA affinity precipitation assays (DAPA) using nuclear extracts from the colons of Vhl ΔIE and Vhl F/F mice and biotinylated TNF-α oligos containing the putative MAZ binding site to precipitate proteins that bind. Immunoblotting of the precipitates confirmed that HIF-2α strongly induced MAZ binding which was absent when using a TNF-α oligo in which the MAZ site was mutated ( Figure 6F ). Together, our data suggest that EPAS1 induces the promoter activity of TNF-α through MAZ binding. TNF-α antagonist treatment is an approved therapy for UC patients 33 . To test whether TNF-α is a critical downstream mediator for EPAS1-mediated colitis, Vhl ΔIE mice were treated with TNF-α antagonists Enbrel (soluble TNF-α decoy receptor) or Remicade (monoclonal neutralizing antibody) in C. rod -induced colitis model. As expected, following C. rod treatment the Vhl ΔIE mice demonstrated a significant decrease in body weight compared to Vhl F/F mice ( Figure 7A , Supplemental Figure 10A ). The TNF-α inhibitors Enbrel or Remicade, completely abolished the decrease in body weight following C.rod treatment in Vhl ΔIE mice ( Figure 7A , Supplemental Figure 10A ). The Vhl ΔIE mice had more severe colonic hemorrhage and decreased colon length compared to Vhl F/F mice, which were attenuated by Enbrel or Remicade treatment ( Figure 7, B and C ; Supplemental Figure 10, B and C ). Histological scoring confirmed severe tissue damage and massive inflammatory cells infiltration in Vhl ΔIE mice following C.rod treatment, while Enbrel or Remicade treatment ameliorated these histological changes ( Figure 7, D and E ; Supplemental Figure 10D ). Moreover, the expressions of Tnf-α, Il-6 and Il-1β were significantly increased in the colons of Vhl ΔIE mice compared with Vhl F/F mice following C.rod treatment ( Figure 7F , Supplemental Figure 10E ). Enbrel or Remicade treatment did not reduce the expression of Tnf-α in Vhl ΔIE mice, however it significantly reduced the expressions of Il-6 and Il-1β ( Figure 7F , Supplemental Figure 10E ). This is consistent with the data that TNF-α is directly downstream of EPAS1, whereas activation of other pro-inflammatory cytokines by EPAS1 is dependent on TNF-α expression. In addition, Remicade also protected Epas1 LSL/+ mice from C. rod or DSS-induced colitis ( Supplemental Figure 11 ). These data demonstrate that EPAS1 regulates the epithelial inflammatory response through TNF-α.

Intestinal epithelial cells play a major role in IBD as a barrier to the luminal mircroflora. However, recent data demonstrates that epithelial cells are also critical in regulating the mucosal immune response 34 - 38 . Through studying the role of EPAS1 in IBD the present manuscript demonstrates that EPAS1 is a critical transcription factor regulating the epithelial inflammatory response. It is difficult to discern the relative importance of the epithelial inflammatory response compared to the innate and adaptive immune cells. However, this work clearly demonstrates that attenuation of the epithelial elicited inflammation through disruption of EPAS1 limits colonic inflammation. Moreover, overexpression of epithelial EPAS1 leads to increased expression of pro-inflammatory cytokines and potentiates colitis. In addition to activating the epithelial-elicited inflammatory response EPAS1 activates iron transporters 17 , 18 , 39 , which may play a role in epithelial inflammation and contribute to dysregulation of iron transporters in IBD 40 . Although TNF-α neutralization reverses EPAS1-induced inflammation it will be interesting to understand if EPAS1-medited intestinal iron homeostasis plays any contributory role in epithelial inflammation or changes in the gut microflora that are associated with IBD. The data in the Vhl ΔIE and Epas1 LSL/+ mice clearly demonstrates that activation of epithelial EPAS1 signaling potentiates inflammation. This data is contrary to the well-characterized function of PHD inhibitors or genetic disruption of PHDs that mimic hypoxic stimulation, and protects mice in acute colitis models 11 , 12 , 41 . The PHD inhibitor dimethyloxaloylglycine (DMOG) at a dose that was protective in DSS-induced colitis was assessed for EPAS1 activation. DMOG stimulated colonic hypoxic signaling was confirmed in vivo and ex vivo with the ODD-luciferase mice ( Supplemental Figure 12A ). The expression of HIF-1α specific target gene pdk1 was increased but no increase in the expression of EPAS1 or its specific target genes such as Dmt1a and Dcytb was observed ( Supplemental Figure 12, B-D ) 17 , 39 , suggesting the DMOG treatment maybe a more potent activator of HIF-1α in the intestine for protection in acute colitis models. It is also possible that activation of EPAS1 by PHD inhibitors may require significantly longer treatment times. This is consistent with the data demonstrating a rapid activation of HIF-1α compared to EPAS1 in several cell lines 42 . Hypoxic activation of pro-inflammatory mediators has been described in innate immune cells 43 - 45 . However, the present study revealed that hypoxia is rapidly increased in epithelial cells of the inflammatory foci and regulates expression levels of a number of inflammatory mediators in an EPAS1-dependent manner. Our previous work had demonstrated the importance of macrophage migration inhibitory factor signaling to hypoxic inflammation 13 . However, the relative importance of these and likely other gene products to the increased colon inflammation has not been precisely defined. In mice overexpressing EPAS1 but not HIF-1α a significant increase in Tnf-α gene expression is observed. Moreover, wild-type mice demonstrated a robust increase in the expression of Tnf-α following DSS treatment, which was significantly attenuated in Epas1 ΔIE mice. To identify putative functional EPAS1 binding sites, promoter analysis of Tnf-α was performed. Interestingly, EPAS1 (but not HIF-1α) regulated the promoter activity of Tnf-α through a NF-κB independent mechanism. The data suggests that HIF-2α activates Tnf-α through a MAZ-dependent manner. Currently studies are being performed to assess if EPAS1 is in a complex with MAZ on the TNF-α promoter. A similar mechanism has been demonstrated for EPAS1 interaction with c-myc, which is critical in the regulation of cell proliferation 46 . The present study demonstrates that TNF-α is the major pro-inflammatory mediator required for induction of hypoxic inflammation. As anti-TNF-α therapy is expensive and has become the main costs driver for IBD patients 47 , alternative cost-effective medication is needed. Moreover, anti-TNF-α therapies also lead to several deleterious side effects 48 - 52 . Therefore, EPAS1 may represent a potential target for treatment of IBD since its expression is nearly undetectable in most normal tissues and is present in inflammatory foci. In conclusion we demonstrate that hypoxia signaling is robustly induced in intestinal epithelial cells. Using a powerful collection of mouse models it was demonstrated that hypoxia and EPAS1 is essential for the proper activation of the epithelial inflammatory response. Together with previous studies assessing the role of HIF-1α in the intestine 8 , 10 , the present study establishes a working model where HIF-1α tightens the epithelial barrier limiting luminal bacteria antigen, and EPAS1 activates the epithelial inflammatory response to eliminate bacteria. This study clearly shows that the EPAS1/TNF-α axis is a critical signaling pathway in activating hypoxic inflammation. Since the dominant role of HIF activation in the intestine is a pro-inflammatory response mediated by EPAS1, more work is required to determine the temporal regulation of HIF-1α and EPAS1 in colitis. It is possible that HIF-1α and EPAS1 are activated in colitis, but HIF-1α response persists for proper barrier repair and resolution of inflammation.