Hypoxia-Inducible Factor Augments Experimental Colitis Through a MIF-Dependent Inflammatory Signaling Cascade

Abstract

Background & Aims

Colon epithelial cells are critical for barrier function and contain a highly developed immune response. A previous study has shown hypoxia-inducible factor (HIF) as a critical regulator of barrier protection during colon epithelial injury. However, the role of HIF signaling in colon mucosal immunity is not known.

Methods

With the use of cre/loxP technology, intestinal specific disruption of Vhl, Hif-1α, and Arnt were generated. Colon inflammation was induced using a dextran sulfate sodium (DSS)-induced colitis model and the mice analyzed by histology, western blot analysis, and quantitative polymerase chain reactions.

Results

In mice, colonic epithelium disruption of Vhl resulted in constitutively expression of HIF which initiated an increase in inflammatory infiltrates and edema in the colon. These effects were ameliorated in mice by disruption of both Vhl and Arnt/Hif1β (which inactivates HIF). In a DSS-induced colitis model, increased HIF expression correlated with more severe clinical symptoms and an increase in histological damage, while disruption of both Vhl and Arnt in the colon epithelium inhibited these effects. Furthermore, colons with constitutive activation of HIF displayed increased expression of pro-inflammatory mediators which were synergistically potentiated following DSS administration and reduced by inhibition of the pro-inflammatory and direct HIF-target gene macrophage migration inhibitory factor (MIF).

Conclusion

The present study demonstrates that a chronic increase in HIF signaling in the colon epithelial cells initiates a hyper-inflammatory reaction that may have important implications in developing therapeutic strategies for inflammatory bowel disease.

Introduction

Hypoxia, a deficiency in oxygen availability, was shown to regulate a large subset of genes critical in both oxygen delivery and adaptation to oxygen deprivation ^{1, 2}. Regulation of hypoxia-mediated genes are dependent on the heterodimeric nuclear transcription factor, hypoxia inducible factor (HIF) consisting of an oxygen sensitive alpha subunit, where three isoforms have been identified HIF-1α ^{3, 4}, HIF-2α ⁵ and HIF-3α ⁶, and a ubiquitously expressed beta subunit, also referred to as aryl hydrocarbon nuclear translocator (ARNT) ². In the presence of adequate oxygen levels (normoxia), HIF alpha subunits are rapidly degraded via post-translational hydroxylation and ubiquitination. Oxygen-dependent prolyl-hydroxylation is necessary for binding to the von Hippel-Lindau tumor suppressor protein (VHL) and consequently to the E3 ubiquitin ligase complex ^{7, 8}. Thus the absence of a functional VHL results in constitutively active HIF in vivo ⁹. HIF signaling was shown to activate transcription of genes critical in cell survival, angiogenesis, glycolysis and iron homeostasis ^10–13. The central role of HIF signaling in normal development and physiology is underscored by the embryonic lethality observed in mice lacking HIF-1α, HIF-2α, ARNT and VHL due to various vascular abnormalities ^14–17.

Recently, using a two-step 2,4,6-trinitrobenzene sulphonic acid (TNBS) or oxazolone-induced inflammatory bowel disease (IBD) model, it was shown that HIF-1α and VHL are critical factors in maintaining intestinal epithelial integrity during increased local inflammation ¹⁸. The two-step model initiates a delayed hypersensitive reaction. First, an epicutaneous treatment with TNBS primes T-cells. A subsequent inter-rectal instillation of TNBS results in a haptenization of the epithelial mucosa leading to a massive Th1 driven immune response to self cells ^{19, 20}. Mice containing an epithelial specific disruption of HIF-1α demonstrated an increase in the intestinal permeability and clinically more severe colitis as compared to their wild-type counterparts, whereas conditional targeting of Vhl in epithelial cells was protective. The mechanism by which HIF-1α maintains colonic mucosal integrity was shown to be through the induction of a number of barrier-protective genes ¹⁸. However, IBD is thought to be a combination of a disturbance in function of the intestinal epithelial barrier and a dysregulation of the mucosal immune system ^{21, 22}. Intestinal epithelial cells that are critical in mucosal immunity by expressing several immunomodulatory genes, act in concert with other immune mediators to elicit a pro-inflammatory signal ²³. Using the TNBS or oxazolone-induced colitis model, it is difficult to assess the immunomodulatory role of HIF and VHL in mucosal immunity due to a direct robust immune response caused by primed T-cells. Therefore, the present study used a DSS-induced acute colitis model where the immune response is secondary to disruption of the epithelial barrier ²⁰. In addition, to gain a better insight into HIF signaling in mucosal immunity, the present study used intestinal epithelial cell conditional knockouts of HIF-1α, ARNT and VHL by use of the cre/loxP technology where the Cre transgene is under the control of the murine villin promoter. The villin promoter was shown to target expression of transgenes to the small and large intestine in both differentiated and undifferentiated cells of the crypt ²⁴.

The present study demonstrates that a chronic increase in HIF signaling in colon epithelial cells triggers inflammatory response as assessed by an increase in pro-inflammatory mediators and colon histology that were dramatically potentiated by administration of DSS in the drinking water. Disruption of both VHL and ARNT in intestinal epithelial cells prevented development of intestinal inflammation indicating a HIF-dependent mechanism. Moreover, the inhibition of MIF activity, a direct HIF target ²⁵, ameliorated the increase in pro-inflammatory mediators demonstrating MIF as a critical factor in the HIF-induced pro-inflammatory cascade.

Methods

Animals

Vhl-floxed (Vhl^F/F) ⁹, Hif-1α-floxed (Hif-1α^F/F) ²⁶ and Arnt-floxed (Arnt^F/F) ²⁷ mice containing loxP sites flanking exons 1, 13–15, and 6 respectively, were crossed with mice harboring the Cre recombinase under control of the villin promoter (villin-cre mice) ²⁴. The intestine specific knockout mice for Vhl, Hif-1α, and Arnt were designated Vhl^ΔIE Hif-1α^ΔIE Arnt^ΔIE mice, respectively. Mice were housed in temperature and light controlled rooms, were given water and pelleted NIH-31 chow ad libitum. All animal studies were carried out in accordance with Institute of Laboratory Animal Resources (ILAR) guidelines and approved by the National Cancer Institute Animal Care and Use Committee.

Recombination efficiency and genotype determination

To assess recombination efficiency in the Vhl^F/F and Vhl^ΔIE mice, intestine epithelium and all other tissues were harvested and kept in liquid nitrogen. DNA was isolated using the DNeasy kit (Qiagen, Valencia, CA) and the following primers were used to assess recombination Vhl-FWD1 PRIMER 5′-CTGGTA CCCACGAAACTGTC-3′ Vhl-FWD2 PRIMER 5′ CTAGGCACCGAGCTTAGAGGTTTGCG-3′ Vhl-RVS PRIMER 5′-CTGACTTCCACTGATGCTTGTCACAG-3′. The same primers were used for routine genotyping of the mice. Routine genotyping for the Arnt and Hif-1α allele were previously described ²⁸.

Induction and assessment of colitis

Mice (6- to 8-weeks-old) were administered 2.5% or 5% (wt/vol) DSS (MW, 35,000–44,000) (MP Biomedicals, Aurora, OH) in the drinking water for five days. In experiments using ISO-1 (Calbiochem, San Diego, CA), the inhibitor was resuspended in 5% DMSO and delivered to the mouse daily via intraperitoneal injection at 20mg/kg and 5% DMSO was used as vehicle. Daily changes in body weight, diarrhea, bleeding, and histological damage were assessed and reported as previously described ²⁹.

Intracellular MIF tautomerase activity assay

Colon extracts were prepared by homogenizing colon epithelial cells from Vhl^F/F and Vhl^ΔIE mice treated with vehicle or ISO-1 in non-denaturing tris buffer. L-Dopachrome methyl ester was generated by mixing equal volumes of L-dopa methyl ester (4mM) and sodium periodate (8mM) (Sigma). Colon extracts (0.7mL) were mixed with freshly prepared L-dopachrome methyl ester (0.3mL) and the decay in absorbance was measured at 475nm.

RNA analysis

RNA was extracted from colon epithelium and qPCR was performed as previously described ²⁹. All primers sequences are available upon request.

Western blot analysis

Colon epithelium or HCT116 cells were lysed using NE-PER nuclear extraction kit for nuclear extract(Pierce, Rockford, IL) or RIPA buffer for whole cell extract. The membranes were incubated with an antibody against Hif-1α, Hif-2α (Novus Biologicals, Littleton, CO), and Arnt and MIF (Santa Cruz Biotechnology Inc, Santa Cruz, CA) the signals obtained were normalized to HNF4α (Santa Cruz) or GAPDH (Chemicon International, Temecula, CA).

Data analysis

Results are expressed as mean ± S.D. P values were calculated using multifactorial Anova test and Independent T Test. p < 0.01 was considered significant.

Results

Intestine specific disruption of Vhl, Arnt and Hif-1α genes via Cre-loxP–mediated recombination in mice

To estimate the extent of cell-specific disruption of the Vhl locus, PCR analysis was used. The Vhl null allele amplifies as a 260 bp product, and was detected in genomic DNA isolated from intestinal epithelium cells of Vhl^ΔIE mice and was not detected in intestinal epithelium DNA isolated from Vhl^F/F mice (Fig 1A). In contrast, the intact floxed allele was the only band evident in the intestinal epithelium from Vhl^F/F mice and from all non-gut derived tissues in Vhl^ΔIE mice (Fig 1A). The expression of the Vhl mRNA was markedly decreased in the colon and throughout the small intestine of Vhl^ΔIE mice (Fig 1B), whereas, Vhl mRNA levels were unchanged in non-gut-derived tissues (data not shown). Next, western blot analysis was performed using nuclear protein extracts isolated from colonic epithelium from Vhl^F/F and Vhl^ΔIE mice. Interestingly, while no specific signal was observed for HIF-1α from colonic epithelium extracts, a robust HIF-1α expression was detected in the HCT116 colon cancer-derived cell line following a 24 hour incubation in 1% O₂. However, induction of HIF-2α was observed in Vhl^ΔIE compared to their wild-type littermate Vhl^F/F mice, hepatic nuclear factor 4 alpha (HNF4α) used as a loading control for the nuclear fraction (Fig 1C). qPCR analysis demonstrated an increase in mRNAs encoding well-characterized HIF target genes (Fig 1D). As previously shown ²⁸, Hif-1α and Arnt expression was significantly impaired in the colon epithelium from Hif-1α^ΔIE and Arnt^ΔIE mice when compared to their wild-type counterparts Hif-1α^F/F and Arnt^F/F, and no decrease was observed in non-gut-derived tissues (Fig 1E and F). Together these data demonstrate that inactivation of the Vhl gene leads to an increase in HIF signaling. Interestingly, no HIF-1α protein expression was observed suggesting the increase in HIF gene expression may be due solely to HIF-2α.

Figure 1. Colon Specific disruption of Vhl.

(A) PCR diagnostic for Cre-mediated recombination of the Vhl allele in genomic DNA isolated from Vhl^ΔIE or Vhl^F/F mice. (B) qPCR analysis measuring Vhl mRNA expression in intestinal epithelium from Vhl^ΔIE or Vhl^F/F mice. (C) Western blot analysis measuring Hif-1α or Hif-2α expression in colon epithelial cells isolated from Vhl^ΔIE or Vhl^F/F mice. Expression was normalized to HNF4α protein expression, and HCT116 cells treated with 1% O₂ for 24 hours served as positive control. (D) qPCR analysis of HIF target genes. (E) qPCR analysis measuring Hif-1α mRNA expression in total RNA from colon epithelium from Hif1α^ΔIE or Hif1α^F/F mice. (F) qPCR analysis measuring Arnt expression in total RNA from colon epithelium from Arnt^ΔIE or Arnt^F/F mice. For qPCR analysis the expression was normalized to β-actin and each bar represents the mean value ± S.D. (†)= P<.01 compared to wild-type littermates.

Colon inflammation as assessed by histology and pro-inflammatory gene expression was increased in Vhl^ΔIE mice

Macroscopic examination of the colons in Hif-1α^F/F, Hif-1α^ΔIE, Arnt^F/F and Arnt^ΔIE mice demonstrated no apparent abnormalities (Fig 2A). However, histological examination of colons from Vhl^ΔIE mice, demonstrated edema in the submucosa layer and marked increase in inflammatory infiltrates (Fig 2A). In addition, 6-month-old Vhl^ΔIE mice demonstrated inflammatory polyps in the colon. Of the 72 Vhl^ΔIE mice assessed at 6-months of age, 27% contained polyps consisted of regenerative epithelium and were accompanied by ulceration and covered by exudates including cell debris, and fibrin. Infiltration of inflammatory cells and micro hemorrhage were also noted. The incidence increased to over 50% in 1-year-old mice, whereas no colon abnormalities were observed in Vhl^F/F wild-type littermate mice (Fig 2B). The inflammatory polyps were similar to those described in patients with Crohn's Disease and other IBDs such as ischemic colitis and ulcerative colitis. Consistent with the increase in the histological signs of inflammation, all pro-inflammatory mediators assessed by qPCR were increased in the Vhl^ΔIE compared to Vhl^F/F (Fig 2C). Interestingly, no changes were observed in Hif-1α^ΔIE or Arnt^ΔIE mice compared to their wild-type littermates (Fig 2C). These results clearly indicate that disruption of Vhl in the colon epithelium results in a marked increase in inflammation.

Figure 2. Pro-inflammatory gene expression in the Hif1α, Arnt or Vhl disrupted colon epithelium.

(A) Representative H & E stained colon sections from wild-type littermates and Hif1α^ΔIE, Arnt^ΔIE, and Vhl^ΔIE mice. (B) Representative H & E stained pro-inflammatory polyp from two individual Vhl^ΔIE mice (inset indicates higher magnification). (C) qPCR analysis of pro-inflammatory mediators in the colon epithelium from wild type littermates and Hif1α^ΔIE, Arnt^ΔIE, and Vhl^ΔIE mice.± S.D. (†)= P<.01 compared to vehicle treated wild-type mice.

Exacerbation of colitis in Vhl^ΔIE mice treated with low-dose (2.5%) DSS

Vhl^ΔIE mice showed an increased susceptibility to DSS-induced colitis and the study was stopped following five days of 2.5% DSS administration due to marked increase in the pathological phenotype of Vhl^ΔIE mice. Vhl^ΔIE mice had significantly more severe disease as assessed by rectal bleeding, diarrhea, body weight, and colon length as compared to Vhl^F/F mice (Fig 3A and B). Furthermore, histological analysis revealed increased inflammation in the mucosa, thickening and edema in the submucosa, and muscularis propria with complete loss of the crypts and surface epithelia, and increased recruitment of inflammatory infiltrates in Vhl^ΔIE mice when compared to the Vhl^F/F mice (Fig 3C). Consistent with the increase in pathological phenotype and histological score following 5 days of DSS administration in Vhl^ΔIE mice, several pro-inflammatory mediators demonstrated robust potentiation in expression following DSS administration (Fig 3D). These data suggest that Vhl is critical in maintaining the mucosal immune response homeostasis following an inflammatory insult.

Figure 3. Clinical assessment of DSS-induced colitis in Hif1α^ΔIE, Arnt^ΔIE, and Vhl^ΔIE mice.

(A) Rectal bleeding, diarrhea, and body weight changes following DSS-induction of colitis, (B) colon length, (C) representative H & E stained colon sections and histology score. (D) qPCR analysis of pro-inflammatory mediators in the colon epithelium from Vhl^F/F and Vhl^ΔIE mice treated with control H₂O or 2.5% DSS H₂O. Data represent the mean value ± S.D, (†)= p < 0.01 compared to Vhl^F/F DSS treated mice.

Colon specific double disruption of Vhl and Arnt genes via Cre-loxP–mediated recombination in mice

To assess the influence of the HIF-dependent pathways on the hyper-inflammatory response in the colons of Vhl^ΔIE mice, mice with a disruption of both Vhl and Arnt were made. To use background matched littermate control mice, Vhl^F/FArnt^F/+ mice hemizygous for villin-cre were mated to each other and four genetic littermate strains were used for comparison; 1) Vhl^F/FArnt^+/+ 2) Vhl^ΔIEArnt^+/+ 3) Vhl^F/FArnt^F/F 4) Vhl^ΔIEArnt^ΔIE. The expression level of Arnt and Vhl mRNAs were assessed in the four-littermate genetic strains. As expected, no difference in Arnt mRNA expression levels were seen in the Vhl^F/FArnt^+/+, Vhl^ΔIEArnt^+/+ and Vhl^F/FArnt^F/F mice, however in the Vhl^ΔIEArnt^ΔIE mice, Arnt expression was abolished. No change in Vhl expression was observed in the Vhl^F/FArnt^+/+ and Vhl^F/FArnt^F/F, but the mRNA expression was completely reduced in Vhl^ΔIEArnt^+/+ and Vhl^ΔIEArnt^ΔIE mice (Fig 4A). Next, western blot analysis was performed on nuclear extracts for HIF-1α, HIF-2α, ARNT, and HNF4α as a loading control. No HIF-1α expression was observed with Vhl^F/FArnt^+/+ (lanes 1 and 2) and Vhl^F/FArnt^F/F mice (lanes 5 and 6) and consistent with what was observed n the Vhl^ΔIE mice (Fig 1C), no induction of HIF-1α was observed in Vhl^ΔIEArnt^+/+ mice (lanes 3 and 4). Interestingly, HIF-1α expression was increased in the Vhl^ΔIEArnt^ΔIE mice (lanes 7 and 8) (Fig 4B). As expected, HIF-2α expression was induced in both the Vhl^ΔIEArnt^+/+ and Vhl^ΔIEArnt^ΔIE mice, whereas ARNT protein expression was abolished only in the Vhl^ΔIEArnt^ΔIE mice (Fig 4B). Interestingly, Hif-1α mRNA levels were reduced in the colons of Vhl^ΔIEArnt^+/+ mice, whereas no change in expression was observed in Vhl^F/FArnt^+/+, Vhl^F/FArnt^F/F and Vhl^ΔIEArnt^ΔIE mice (Fig 4C). The expression of Hif-1α mRNA in Vhl^ΔIEArnt^+/+ mice was similar to that observed in Hif-1α^ΔIE mice (Fig 1E), suggesting a complete disruption of expression. In addition, no change in Hif-2α expression was seen in any genotype (Fig 4C). The data suggest a HIF-dependent negative feedback, which will be further discussed below. As observed with the Vhl^ΔIE mice (Fig 1D), Vhl^ΔIEArnt^+/+ mice demonstrated an increase in several HIF target genes, which were inhibited in the Vhl^ΔIEArnt^ΔIE mice (Fig 4D), indicating that HIF alpha subunit function is dependent on binding to its obligate heterodimer partner ARNT.

Figure 4. Colon-specific double-disruption of Vhl and Arnt.

(A) qPCR analysis measuring Arnt and Vhl expression in total RNA from colon epithelium. (B) Western blot analysis measuring Hif-1α, Hif-2α and Arnt expression in colon epithelial cells. Expression was normalized to HNF4α protein expression. (C) qPCR analysis of Hif-1α and Hif-2α in colon epithelium. (D) qPCR analysis of HIF target genes in colon epithelium. For qPCR analysis the expression was normalized to β-actin and each bar represents the mean value ± S.D.(†)= P<.01 compared to wild-type littermates.

Colon inflammation induced in Vhl^ΔIE mice is dependent on HIF signaling

Examination of the colon in Vhl^ΔIEArnt^+/+ mice demonstrated an increase in the histological signs of inflammation (Fig 5A). In addition, of the 54 Vhl^ΔIEArnt^+/+ mice assessed at 6-months of age, 25% displayed inflammatory polyps in the colons similar to what was observed in Vhl^ΔIE (data not shown). These effects were ameliorated in the Vhl^ΔIEArnt^ΔIE mice (Fig 5A and data not shown). The induction of pro-inflammatory mediators observed in Vhl^ΔIEArnt^+/+ mice were inhibited in the Vhl^ΔIEArnt^ΔIE confirming that HIF signaling is required for the increase in inflammation following Vhl disruption (Fig 5B). To clarify whether the increase in susceptibility to DSS-induced colitis in Vhl^ΔIE mice was dependent on intact HIF signaling, Vhl^F/FArnt^+/+, Vhl^ΔIEArnt^+/+, Vhl^F/FArnt^F/F, and Vhl^ΔIEArnt^ΔIE mice were subjected to 2.5% DSS treatment. Vhl^ΔIEArnt^+/+ mice developed severe bloody diarrhea and body weight loss, while the body weight loss, diarrhea, and rectal bleeding were only marginally changed in Vhl^F/FArnt^+/+, Vhl^F/FArnt^F/F and Vhl^ΔIEArnt^ΔIE mice (Fig 6A). The colon length of Vhl^ΔIEArnt^+/+ mice treated with 2.5% DSS was dramatically decreased compared to 2.5% DSS-treated colons from Vhl^F/FArnt^+/+, Vhl^F/FArnt^F/F and Vhl^ΔIEArnt^ΔIE mice (Fig 6B). The histological injury score obtained from Vhl^ΔIEArnt^+/+ mice was significantly higher than that from Vhl^F/FArnt^+/+, Vhl^F/FArnt^F/F and Vhl^ΔIEArnt^ΔIE mice (Fig 6C). In addition, using a 5% DSS dose demonstrated that Vhl^ΔIEArnt^+/+ mice had a decreased survival when compared to Vhl^F/F/Arnt^+/+, Vhl^F/FArnt^F/F and Vhl^ΔIEArnt^ΔIE mice demonstrating 100% survival following seven days of 5% DSS treatment (Fig 6D). These findings indicate that the disruption of Vhl in the colon epithelium exacerbates colitis, due to an increase in HIF signaling.

Figure 5. Pro-inflammatory gene expression in the double Vhl and Arnt disrupted colon epithelium.

(A) Representative H & E stained colon sections. (B) qPCR analysis of pro-inflammatory mediators in the colon epithelium. Expression was normalized to β-actin and each bar represents the mean value ± S.D. (†)= P<.01 compared to wild-type littermates.

Figure 6. Clinical assessment of DSS-induced colitis in the double Vhl and Arnt disrupted colon epithelium.

(A) Rectal bleeding, diarrhea, and body weight changes following DSS-induction of colitis, (B) colon length, (C) representative H & E stained colon sections and histology score. (D) Survival of Vhl^F/FArnt^+/+, Vhl^ΔIEArnt^+/+, Vhl^F/FArnt^F/F, and Vhl^ΔIEArnt^ΔIE mice. Data represent the mean value ± S.D, (†)= p < 0.01 compared to Vhl^F/F DSS treated mice.

MIF inhibition decreased the expression of pro-inflammatory mediators in Vhl^ΔIE mice

MIF was shown to be a direct target of HIF signaling ²⁵. To assess whether MIF is critical in HIF-induced colonic inflammation, the protein expression and enzyme activity were measured. Western analysis demonstrated an increase in MIF protein expression from colonic extracts (Fig 7A). In addition, MIF activity was measured by utilizing the unique catalytic domain capable of tautermirizing L-dopachrome to indole derivates ³⁰. Colon extracts from Vhl^ΔIE mice demonstrated an 8-fold increase in MIF tautomerase, and the specific MIF tautomerase inhibitor ISO-1 completely inhibited MIF activity thus demonstrating the specificity of the assay (Fig 7B). Vhl^ΔIE mice treated with ISO-1 exhibited a significant decrease in pro-inflammatory mediator expression (iNOS, IFNγ, TNFα, IL-6, IL-10 COX-2, ICAM-1, TLR2 and IL-1β) when compared to vehicle treated Vhl^ΔIE mice. As expected, ISO-1 had no effect on MIF expression as it functions as an enzymatic inhibitor. Moreover, MIF antagonism was shown to be specific for HIF-induced pro-inflammatory mediators; no decrease in mRNA expression was observed for any pro-inflammatory mediators in Vhl^F/F mice treated with ISO-1 compared to vehicle treated Vhl^F/F mice (Fig 7C). The enzymatic activity of COX-2, a direct pro-inflammatory target of HIF ³¹, was also shown to be increased in Vhl^ΔIE mice (Supplemental Fig 1A and B). However when Vhl^F/F and Vhl^ΔIE mice were treated with nimelsulide, a COX-2 specific inhibitor, no decrease in any pro-inflammatory expression was observed when compared to vehicle treated Vhl^F/F and Vhl^ΔIE mice (Supplemental Fig 1A and B). Interestingly IL-1β levels were significantly increased in nimesulide treated Vhl^ΔIE mice, possibly due to intestinal side effects known to be associated with COX-2 inhibitors ³². In addition, Vhl^F/F and Vhl^ΔIE mice treated with ISO-1 were protected from DSS-induced colitis. Due to robust response observed in both Vhl^F/F and Vhl^ΔIE mice on 5% DSS, the study was stopped following five days DSS administration. Vhl^ΔIE mice had a significantly more severe response as assessed by rectal bleeding, diarrhea, body weight, and colon length as compared to Vhl^F/F mice. However ISO-1 administration protected both Vhl^F/F and Vhl^ΔIE mice (Fig 7D and E). Mortality was also slightly higher (2/10) in Vhl^ΔIE mice, when compared to Vhl^ΔIE mice administered ISO-1 (0/10) following DSS treatment (data not shown). The pathological findings were consistent with histological analysis revealing an improvement in mucosal inflammation, thickening and edema in the submucosa, and muscularis propria loss of the crypts and surface epithelia, following ISO-1 treatment (data not shown). These data demonstrates a critical role for MIF in the HIF-induced pro-inflammatory cascade and colitis.

Figure 7. Pro-inflammatory gene expression in Vhl^F/F and Vhl^ΔIE mice following nimesulide or ISO-1 administration.

(A) Western blot analysis for MIF expression in colon extracts. (B) MIF tautomerase activity measured in colon extracts of Vhl^F/F and Vhl^ΔIE mice treated with vehicle or 20mg/kg of ISO-1. Data represent the mean value ± S.D, (†)= p < 0.01 compared to wild-type littermates. (C) qPCR analysis of pro-inflammatory mediators in the colon epithelium from Vhl^F/F and Vhl^ΔIE mice treated with vehicle or 20mg/kg of ISO-1. Expression was normalized to β-actin and each bar represents the mean value ± S.D. (†)= P<.01 compared to vehicle treated Vhl^ΔIE mice. (D) Rectal bleeding, diarrhea, body weight changes, and (E) colon length following DSS-induction of colitis. Data represent the mean value ± S.D, (*)= p < 0.01 compared to Vhl^F/F MIF treated mice and (†)= p < 0.01 compared to Vhl^ΔIE MIF treated mice

Discussion

Individuals in a chronic hypoxic state, such as patients with chronic obstructive pulmonary disease demonstrate systemic inflammation ³³. Interestingly, when a focused case report review was performed on respiratory disorders leading to whole-body hypoxia, cardio-pulmonary disorders were more frequent among patients with IBD than previously considered ³⁴. In addition, hyperbaric oxygen treatment has been demonstrated to be effective in animal models of acute colitis and patients with severe IBD ³⁵, however the molecular mechanism leading to either observation is currently unclear. The present study provides evidence demonstrating a molecular link between dysregulation of oxygen signaling and an increase in inflammation. Utilizing conditional Vhl^ΔIE, Hif-1α^ΔIE, Arnt^ΔIE and Vhl^ΔIEArnt^ΔIE mutant mice, a chronic increase in colon epithelial HIF-2α signaling resulted in a hyper-inflammatory response with an increase in colon inflammation and pro-inflammatory mediators. Hif-1α^ΔIE and Arnt^ΔIE mice demonstrated no significant difference in response to the susceptibility of DSS-induced colitis compared to their wild-type littermates possibly due to a limited role of hypoxia during the initial stage of DSS-induced colitis (Supplemental Fig 2). An increase in HIF signaling was shown to be a late event following low dose DSS administration (Supplemental Fig 2), therefore suggesting a mechanism whereby increased physiological HIF signaling predisposes mice to an inflammatory insult such as DSS. This hypothesis is further supported by the finding that all basal level pro-inflammatory mediators measured are increased in Vhl^ΔIE mice, whereas no change in basal level expression was observed in Hif-1α^ΔIE and Arnt^ΔIE mice.

Previous work has revealed that HIF-1α signaling is critical in the intestinal epithelial barrier and several HIF-1α target genes are also barrier protective ¹⁸. Paradoxically, the loss of VHL in the epithelial cells demonstrated a protective role in TNBS-induced colitis. This observation is different from that of the present study, and several possibilities may account for the differences in phenotype. A protective role for HIF in colitis was shown by use of a Cre-recombinase under transcriptional control of the fatty acid–binding protein (Fabp) promoter demonstrating less than 60% cre-mediated recombination of the Vhl allele, whereas in the present study using the villin promoter to drive the Cre gene, almost complete recombination was observed and the residual floxed allele in total gut tissue was thought to be due to inflammatory cells within the colon epithelium. Interestingly, gene dosage effects of Hif-1α have indeed been noted; heterozygous null mice demonstrated an impaired response to chronic hypoxia ³⁶. Therefore, the increase in recombination efficiency of the villin-cre promoter may have uncovered the hyper-inflammatory response that was not seen utilizing the FABP promoter-driven Cre recombinase that yields incomplete gene disruption in the gut.

Interestingly, loss of VHL in myeloid cells and overexpression of a degradation resistant HIF-1α in the skin epithelium results in a hyper-inflammatory response in 12-O-tetradecanoyl- phorbol-13-acetate (TPA)-induced acute skin inflammation model ^{37 38}. The results of these studies are consistent with the present work and may suggest that the intrinsic differences between the inflammatory models may account for some of the differences in phenotype following initiation of colitis. In the TNBS-induced colitis model, disruption of Vhl was shown to be protective via HIF-1α-mediated induction of mucosal barrier genes. However, in the DSS model, the primary insult overrides the role of barrier protection, and thereby revealing the pro-inflammatory role of HIF signaling in colon epithelia.

Lastly, the Vhl^ΔIE mouse model used in the present study was HIF-2α mediated; no protein expression of HIF-1α was detected in the colon epithelium. Therefore, the pro-inflammatory phenotype observed was mainly due to HIF-2α signaling, whereas Karhausen et al.¹⁸ demonstrated that the protective role of HIF signaling, was due to HIF-1α expression in the colon. Recently, two reports demonstrate a protective function in mouse colitis models using pharmacological inhibitors of prolyl hydroxylases, which activate HIF signaling by inhibiting its degradation ^{39, 40}. The HIF prolyl hydroxylase inhibitor FG-4497 was shown to decrease intestinal permeability thereby protecting the intestine. Interestingly, the pharmacological inhibitor displayed no effects on barrier function in the intestine specific Hif-1α-null mice, confirming that the protective role HIF signaling in colon homeostasis is primarily dependent on HIF-1α ³⁹. In addition, HIF-1α was shown to directly regulate several barrier protective genes, such as intestinal trefoil factor, CD73 and multidrug resistance gene 1 ^41–43, however in present study, HIF-2α does not increase their expression in the colon epithelium (data not shown). Due to high sequence similarity between HIF-1α and HIF-2α, these transcription factors share many similar functions, however it is becoming apparent that HIF-1α and HIF-2α can regulate unique sets of target genes with distinct functions. The roles of HIF-1α and HIF-2α have diverged in respect to cellular growth. In cell lines, HIF-2α can promote cycle progression, whereas HIF-1α inhibits cell proliferation ⁴⁴. Therefore the differences observed in present study and in Karhausen et al.¹⁸ may reflect the divergent roles of HIF-1α and HIF-2α in colon homeostasis.

Currently, the mechanism by which Hif-1α is downregulated is unclear. The present study suggests that Hif-1α gene expression is under a negative feedback regulation following prolonged HIF signaling mediated by HIF-2α or its respective target genes. Several in vitro studies have described this phenomenon ^45–47. In lung epithelial-derived cell lines, prolonged hypoxia decreased HIF-1α protein expression via a decrease in the Hif-1α mRNA, whereas no change in Hif-2α expression was observed ⁴⁶. Inhibiting HIF signaling using a dominant negative HIF-2α, prevented the down regulation of Hif-1α expression following a prolonged hypoxia treatment ⁴⁶. Currently this is major focus and the molecular mechanism will be assessed in future studies.

MIF was identified as a primary mediator downstream of HIF-signaling responsible for the increase in pro-inflammatory gene expression following induction of HIF-2α. MIF is a well-characterized pro-inflammatory mediator secreted by several different cell types including colon epithelium and has been demonstrated to be critical in the pathogenesis of colitis. MIF serum concentrations were shown to be elevated in Crohn's patients ⁴⁸, and a decrease in MIF activity by anti-MIF antibody administration or MIF-null mice were highly protective in DSS and TNBS-induced colitis models ^{48, 49 50}. Furthermore, mice transgenically overexpressing MIF demonstrated an increase in susceptibility to DSS-induced colitis ⁵¹. Consistent with the above reports inhibiting MIF activity decreased pro-inflammatory cytokines and protected Vhl^F/F and Vhl^ΔIE mice in DSS-induced colitis model. The importance of the catalytic domain in modulating MIF activity is well documented ⁵²; the present study provides further evidence demonstrating the utility in small-molecule inhibitors of MIF as a therapeutic modality and clearly demonstrates that the increase in MIF expression and activity are critical in the HIF-induced pro-inflammatory cascade in the colon.

In conclusion, the present study demonstrates an increase in colonic inflammation in murine models with constitutive epithelial HIF signaling, which is mediated by HIF-2α activation of the pro-inflammatory gene MIF. Taken together with the previous study ¹⁸, the present working model suggests that an increase in HIF signaling may be protective early in the pathogenesis of IBD via HIF-1α-mediated maintenance of the epithelial barrier. However, as the epithelial barrier breaks down, HIF-2α may potentiate the chronic inflammatory reactions, worsening disease progression. Currently several HIF modulators are either in pre-clinical or clinical trials for a variety ischemic diseases and cancers ⁵³. Therefore the present study provides important implications for using HIF modulators as therapeutic modalities and suggests that caution should be exercised in the long-term use of these compounds in patients.

Abbreviations

ARNT

aryl hydrocarbon nuclear translocator

COX-2

cyclooxygenase-2

DSS

dextran sulfate sodium

HIF

hypoxia inducible factor

HNF4α

hepatic nuclear factor 4 alpha

IBD

inflammatory bowel disease

iNOS

inducible nitric oxide synthase

MIF

macrophage migration inhibitory Factor

PDK1

pyruvate dehydrogenase kinase 1

PGK1

phosphoglycerate kinase 1

PGE₂

prostaglandin E₂

qPCR

quantitative RT-PCR

TNBS

2,4,6-trinitrobenzene sulphonic acid

VEGF

vascular endothelial growth factor

Vhl

von Hippel-Lindau tumor suppressor protein