Mucosal Protection by Hypoxia-Inducible Factor (HIF) Prolyl Hydroxylase Inhibition

Abstract

A number of recent studies have implicated tissue hypoxia in both acute and chronic inflammatory diseases, particularly as they relate to mucosal surfaces involving epithelial cells. In this context, a protective role for the transcriptional regulator hypoxia-inducible factor (HIF) was demonstrated through conditional deletion of epithelial HIF-1α in a murine model of colitis (J. Clin. Invest. 2004; 114:1098−1106). Here, we hypothesized that pharmacologic activation of HIF would similarly provide a protective adaptation to murine colitic disease. For these purposes, we used a novel prolyl hydroxylase (PHD) inhibitor (FG-4497) which readily stabilizes HIF-1α and subsequently drives the expression downstream HIF target genes (e.g. erythropoietin). Our results show that the FG-4497-mediated induction of HIF-1α provides an overall beneficial influence on clinical symptoms (weight loss, colon length, tissue TNFα) in murine TNBS colitis, most likely due to their barrier protective function and wound healing during severe tissue hypoxia at the site of inflammation. Taken together these findings emphasize the role of epithelial HIF-1α during inflammatory diseases in the colon and may provide the basis for a therapeutic use of PHD inhibitors in inflammatory mucosal disease.

Introduction

Mucosal surfaces, such as the lung and colon, are lined by a monolayer of epithelia which provides tissue barrier and transport function. As such, epithelial cells are uniquely equipped with barrier forming features, such as apically positioned tight junctions, microvili, the secretion of mucus, and the underlying supportive tissue is highly vascularized to, at least in part, counteract the steep oxygen gradient from luminal anaerobia to oxygenated tissue. In acute and chronic gastrointestinal inflammatory diseases like Crohn's and ulcerative colitis clinical and experimental studies have shown that the morphological abnormalities include amongst others changes in the microvasculature concurrent with tissue hypoxia. From this standpoint, it is now appreciated that a common feature in a variety of disease processes is diminished oxygen delivery and/or oxygen availability (hypoxia). Given the importance of HIF-1α as mediator of adaptation to low oxygen levels throughout the body¹ and recently shown to be protective in murine experimental colitis² the elevation of HIF-1α-levels as a therapeutic approach to treat colonic inflammatory diseases become a viable option. In particular, HIF-1α is involved in the regulation of a number of barrier-protective genes^3-5.

Diminished oxygen levels trigger an adaptive response which is centered around the controlled expression of hypoxia-inducible factor and facilitates oxygen delivery and energy conservation^{1, 6}. HIF-1α is an 827 amino acid long protein and as part of a transcription factor consist of a basic helix-loop-helix domain, two PAS-domains (member of the Per-ARNT-Sim domain family), an ‘N- and ‘C-terminal transactivation domain (TAD) and an oxygen-dependent degradation domain (ODD). Under normoxic conditions HIF-1α is hydroxylated at two specific proline residues in the ODD by HIF prolyl hydroxylases (PHD's)^7-9, which enables the binding of the von Hippel-Lindau (VHL) tumor suppressor protein (pVHL) coupled to the E3 ubiquitin ligase complex, which ubiquinates HIF-1α and thus targets it to proteasomal degradation^10-12. To become transcriptionally active HIF-1α stabilization, nuclear translocation and subsequent binding to constitutively expressed HIF1-β has to take place concurrent with binding of cofactors like p300/CBP at the carboxy-terminal TAD^{13, 14}. Stabilization and transactivation are regulated by a class of three HIF hydroxylases, which commonly depend on oxygen, Fe(II) and 2-oxoglutarate as substrates to hydroxylate HIF-1α. The functionality of the carboxy-terminal TAD is controlled by an asparaginyl hydroxylase (FIH)¹⁴, which exclusively hydroxylates asparagine residue N803 and thereby inhibits binding of transcriptional cofactors. Inactivation of hydroxylase activity therefore results in stabilization of HIF-1α and thus in the stabilizations of transcriptionally active HIF-1. In this study we use novel PHD inhibitors to increase HIF-1 levels and investigate their influences on whole body hypoxia and in model colitis. Our findings demonstrate the importance of HIF-1 in protecting barrier function in the context of tissue hypoxia associated with inflammatory lesions in experimental colitis and provide the basis for employing HIF PHD inhibitors as therapeutic agents.

Material and Methods

Cell culture

HeLa cells were grown and maintained as confluent monolayers on 24-well plates (Costar Corp., Cambridge MA). Cell cultures were exposed to hypoxia as described previously¹⁵.

FG-4442 and FG-4497 were synthesized at FibroGen, Inc. (San Francisco, CA). These small molecule inhibitors of prolyl hydroxylase enzymes have been previously disclosed in patent filings US 006093730A and US20040254215A1, respectively. Where indicated, cells were exposed to the PHD inhibitors FG-4497 or FG-4442 at indicated concentrations for indicated periods of time. Control conditions included exact exposure to equivalent concentrations of vehicle (DMSO).

Plasmid transfection and luciferase reporter assay

HeLa cells were passaged into 24 well plates and allowed to attach for 24 hours. Transient transfection of HeLa cells and assessment of luciferase activity was carried out as described previously¹⁶.

Immunoblotting

HeLa cells were grown to confluency on 60 mm plastic petri dishes and exposed to indicated concentrations of FG-4497 for 6 hr. Cells were lysed and probed for HIF-1α by western blot as described previously¹⁷.

PHD isoforms determinations

In subsets of experiments, we examined PHD-1, –2, and –3 mRNA expression levels in murine colonic mucosal scrapings (enriched in epithelial cells)¹⁸ derived from BalbC mice. Amplification of cDNA was performed using the following gene specific primers: Phd-1, Forward: 5'-ACC GCG CAG CAT TCG TG-3' and Reverse: 5'-GGG GCT GGC CAT TAG GTA GGT GTA-3'; Phd-2, 5'-GCG GGA AGC TGG GCA ACT AC-3' and 5'-CAA CCC TCA CAC CTT TCT CAC C-3'; Phd-3, 5'-CTG CGT GCT GGA GCG AGT CAA-3' and 5'-TCA TGT GGA TTC CTG CGG TCT G-3'; and β-Actin: 5'-CTA GGC ACC AGG GTG TGA T-3' and 5'-TGC CAG ATC TTC TCC ATG TC-3'; Cycle parameters were 3 minutes at 95°C, then 30 cycles with 45 sec at 95°C, 30 seconds at 60°C plus 30 seconds at 72°C.

Determination of erythropoietin (EPO) induction in vivo in normal mice

Ten week old male Swiss Webster mice (Simonson Labs, Gilroy, CA) were dosed either orally (PO) or intravenously (IV) with FG-4497 at 60 mg/kg. Blood samples were taken under general anesthesia 4 hours post-IV and 6 hours PO dosing and heparin plasma was collected. Samples were analyzed for EPO by ELISA. Circulating levels of plasma EPO were quantitated with a human EPO ELISA kit (R&D Systems) according to manufacturer's instructions, which exhibits some cross-reactivity with murine EPO but thus, underestimates the levels of circulating murine EPO. EPO values represent the mean + SEM (n=3 mice/cohort).

TNBS-colitis model

TNBS colitis was induced with a modification¹⁸ of that by Morris et al. and Boirivant et al.^{19, 20}, respectively. Control animals received a corresponding volume of 50% ethanol alone. In subsets of experiments, we examined tissue hypoxia using nitroimidazole reduction, as described previously^{18, 21}

The PHD inhibitor FG-4497 was administered as a solution (100 mg FG-4497 to 9.675 ml of 5% Dextrose to 325 μl of 1N NaOH) intraperitoneally using a 30 G syringe (BD Bioscience) following a dosing scheme of 40mg/kg/day or 20mg/kg/day on day −1, 0, +1 or exclusively on day 0 and +1 by injection of a volume of 100 or 50 μl, respectively. On day 0 TNBS treatment was performed 4 hours after injection of FG-4497. Only animals that showed an initial response to TNBS-treatment were included in the weight loss analysis; this response was defined as a 5% loss of weight after induction of colitis. As a further parameter, colon length was determined by measurement of the distance from the most distal aspect of the cecum to the most terminal aspect of the rectum.

Hypoxia and intestinal permeability in vivo

Intestinal permeability in vivo was examined using a FITC-labeled-dextran method, as described previously^{3, 22}. Administration and analysis of Texas Red-conjugated E. coli was performed similar to FITC dextran; briefly, mice were gavaged with 0.0125 mg/g body weight of Texas Red-conjugated E. coli Bioparticles (used at a concentration of 2.5 mg/ml; Molecular Probes, Eugene, OR). The hydroxylase inhibitor FG-4442 was administered as a suspension (60 mg/kg body weight in 0.1% carboxymethylcellulose in 1× PBS) intrarectally after sedation using 14 μl/g body weight of 2.5% Tribromoethanol solution in 1× PBS.

In vivo assessment of cytokines

RNA was isolated from colonic mucosal scrapings from 8−10 week old C57BL/6 mice as described before². Amplification of cDNA was performed on an i-Cycler IQ real-time PCR detection system (BioRad Laboratories, Hercules, California) using the following gene specific primers: TNFα : Forward-5'-CCA CCA CGC TCT TCT GTC TAC-3', Reverse-5'-TGG GCT ACA GGC TTG TCA CT-3'; IFN-γ: Forward-5'-TCA AGT GGC ATA GAT GTG GAA GAA-3', Reverse-5'-TGG CTC TGC AGG ATT TTC ATG-3'; β-Actin: Forward-5'-CTA GGC ACC AGG GTG TGA T-3', Reverse-5'-TGC CAG ATC TTC TCC ATG TC-3'; Cycle parameters were 3 minutes at 95°C, then 40 cycles with 45 sec at 95°C, 30 seconds at 58°C plus 30 seconds at 72°C followed by repetitive melting cycles to establish product specificity. Comparison of gene expression in a semi-quantitative manner was performed based on the mathematical model of Pfaffl²³. All procedures involving animals were performed according to National Institute of Health guidelines for use of live animals and were approved by the Institutional Animal Care and Use Committee at the University of Colorado Health Sciences Center.

Collagen gel contraction assay

NIH 3T3 fibroblast stably expressing luciferase-linked to multiple copies of HRE (NIH3T3/HIF-luc cells purchased from Panomics, Redwood City, CA) were lifted from culture plates with trypsin, washed with PBS, and resuspended in complete medium at 500,000 cells/ml. Collagen gels were made as previously described²⁴. All gels contained a final concentration of 150,000 cells/ml and 1.0 mg/ml collagen I with or without indicated concentrations of FG-4497. Gels were digitally imaged after release (t = 0) and at various time points thereafter. Gel surface area was quantified in terms of pixel number using ImageJ (http://rsb.info.nih.gov/ij/). Relative changes in surface area are reported as a percent of the original surface area.

Data Analysis

Weight loss, colon length, mRNA-levels and luciferase reporter data were compared by 2-factor ANOVA or Student's t test, where appropriate. Values are expressed as means ± SEM from separate experiments. P values less than 0.05 were considered significant.

Results

FG-4442 and -4497 increase HIF-1 activity in vitro and in vivo

Initially we sought to characterize the ability of two new HIF PHD inhibitors (FG-4442 and FG-4497) to activate HIF in vitro and examine expression of HIF target genes in vivo. As shown in Figure 1, utilizing a HIF reporter plasmid transiently transfected into HeLa cells, we showed that FG-4442 increased HIF activation in a concentration-dependent fashion (p<0.025), with maximal HIF activity increases of 5.1±0.8-fold (p<0.01). In this regard, FG-4497 was more potent that FG-4442. As shown in Figure 1B, FG-4497 increased HIF activity in a concentration-dependent fashion (p<0.01) with a maximal induction of 10.2±1.6 fold at 5 μM (p<0.001 compared to vehicle treated). Similar analysis of cells subjected to a combination of FG-4497 and hypoxia (1% O₂) revealed that FG-4497 (5 μM) enhanced hypoxia-induced HIF activation (p<0.05), to levels of 13.4±1.1-fold over normoxic controls. Such findings identify FG-4442 and FG-4497 as activators of HIF in vitro, and demonstrate higher cell-based potency for FG-4497. Verification of FG-4497-mediated HIF-1α stabilization by western blot revealed prominent HIF-1α activation at concentrations as low as 500 nM (Figure 1C).

Figure 1. Influence of PHD inhibitors FG-4442 and FG-4497 on HIF activation.

HeLa cells were transfected with HRE-reporter-gene construct, exposed with FG-4442 (A) or FG-4497 (B) for 24 hours in normoxic and hypoxia (pO2 20 torr) and then assayed for luciferase activity. Data are expressed as mean ± SD luciferase/protein, and are pooled from 3 independent experiments with 3 samples per condition, where * p<0.025 and # is p<0.05. (C) HeLa cells were exposed to indicated concentrations of FG-4497 for 6hr and probed for HIF-1α stabilization by western blot. Representative blot from n = 3.

Based on higher potencies in vitro (Figure 1), we focused on the ability of FG-4497 to activate HIF target genes in vivo. Administration of FG-4497 in mice was well tolerated with no adverse side effects in any animal studied (dose range 6 − 100mg/kg for as long as 1 week, data not show). Erythropoietin (EPO) is one of the earliest described and most sensitive HIF target genes; being positively regulated at the transcription level²⁵. Here, we examined the ability of FG-4497 to regulated EPO levels in vivo. As shown in Figure 2A, in vivo administration of FG-4497 by IV and oral routes in mice (60mg/kg/day) strongly increased circulating plasma EPO levels (both p<0.001 compared to vehicle controls) with the orally dosed mice displaying a distinctly less increase in EPO than animals dosed IV (p<0.025, Fig. 2A), due to lower bioavailability of FG-4497 with oral dosing (data not shown). Nonetheless, upregulation of EPO and its influence on hemoglobin (Figure 2B) and hematocrit (Figure 2C) with oral dosing were found to be dose-dependent in the range of 6 − 100mg/kg (for both, p<0.025 by ANOVA). These latter findings provide in vivo correlates to our in vitro findings of HIF activation by FG-4497.

Figure 2. Influence of PHD inhibitor FG-4497 on EPO and EPO endpoints in vivo.

In panel A, Swiss Webster mice were administered FG4497 (one dose 60mg/kg) IV (black bars) or PO (open bars). Animals were sacrificed after 4 and 6 hours, respectively, and plasma EPO was measured by human ELISA kit. Data are expressed as mean ± SD EPO (mIU/ml) and are pooled from 3 animals in each condition where * p<0.025. In panels B and C, animals were administered FG-4497 orally at the indicated doses on Monday, Wednesday and Friday (day 1, 3 and 5) and, hemoglobin (panel B) and hematocrit (panel C) were determined on day 7. Data are expressed as mean ± SD are pooled from 6 animals in each condition where * p<0.025.

Confirmation of HIF-1α target by FG-4497 in vivo

In these studies, we sought to better define the potential therapeutic benefits of HIF activation on mucosal inflammation, particularly in murine colitis models. As a starting point, we screened the expression of PHD isoforms in mouse colon. As shown in Figure 3A, mucosal scrapings (enriched in epithelial cells) derived from wild-type mice expressed mRNA for all three PHD isoforms, where PHD-1 was expressed at the lowest level and PHD-2 and PHD-3 were expressed at nearly equivalent levels.

Figure 3. Screen of PHD isoform expression and influence of FG-4497 on intestinal permeability: role of HIF-1α.

A: Screen of murine PHD isoform expression in wild-type mice. Total RNA was obtained from mucosal scrapings (enriched in epithelial cells) and assessed for expression of PHD-1, -2 and-3 by RT-PCR (30 cycles), relative to actin controls. B: Hif1a WT and conditional hif1a-null mice were administered FG-4497 (60mg/kg IP) or PBS (vehicle) for 20 h, gavaged with FITC-dextran for an additional 4h and and intestinal permeability was quantified as serum FITC-dextran. Data are expressed as mean ± SD relative FITC-dextran (relative to WT control) and are pooled from 4 animals in each condition. C: Quantitation of serum Texas Red-conjugated E. coli as a measure of intestinal permeability. Hif1a WT and conditional hif1a-null mice were administered FG-4497 (60mg/kg IP) or PBS (vehicle) for 20 h, gavaged with Texas Red-conjugated E. coli and subjected to room air (Nmx) or hypoxia (Hpx, 4h at 8% O₂, 92% N₂) as indicated, harvested and serum extracted to measure LPS levels. Data are expressed as mean ± SD serum LPS (ng/ml) and are pooled from 3−5 animals in each condition where * p<0.025 compared to vehicle controls and # p<0.05 compared to wild-type animals. Hif1a WT and conditional hif1a-null mice were subjected to induction of TNBS colitis at day 0. Controls received TNBS vehicle alone.

To further establish that FG-4497 targets HIF in vivo, we utilized two established murine mucosal inflammation models, namely normobaric hypoxia^{17, 26} and TNBS colitis². Thus, we verified that FG-4497 activates HIF-1α by utilizing functional endpoints (intestinal permeability) in animals bearing conditional inactivation of epithelial hif1a². For these purposes, we used two intestinal permeability tracers, namely FITC-labelled dextran (4kDa) and Texas Red-conjugated E. coli. Consistent with our previous studies², loss of intestinal epithelial hif1a resulted in increased intestinal permeability to FITC-dextran (Figure 3B, p<0.025) and Texas Red-conjugated E. coli (Figure 3C, p<0.05). Administration of FG-4497 to normoxic hif1a wild-type and conditional hif1a-null animals, as indicated, resulted in decreased intestinal permeability to FITC-dextran (Figure 3B) and Texas Red-conjugated E. coli (Figure 3C) in wild-type (both p<0.05), but not hif1a-null animals (p = not significant). Similarly, subjection of wild-type animals to hypoxia (Figure 3C) resulted in increased permeability to Texas Red-conjugated E. coli (p<0.05), which was prevented with FG-4497 (p<0.025). These findings provide functional evidence that FG-4497 activates HIF-1α in intestinal epithelium in vivo.

Influence of FG-4497 on outcomes in murine colitis

We next tested the hypothesis that activation of HIF in vivo is protective for murine colitis. Colitis models were selected based on previous studies implicating HIF as a protective element in this mucosal inflammation model¹⁸. Initially, we confirmed that TNBS-induced colitis results in tissue hypoxia. For these purposes, we utilized the characteristic reduction and binding of the nitroimidazole compound EF5 to cellular macromolecules in absence of adequate oxygen levels²⁷. As shown in Figure 4A and 4B, colonic samples from vehicle-treated control animals revealed no detectable retention of EF5 under these conditions. By contrast, animals exposed to the colitis-inducing hapten TNBS revealed a profound retention of EF5 within colonic epithelia (Fig. 4C). Moreover, such EF5 retention was prominently associated with colitic lesions both in superficial and in deeper submucosal regions of the mucosa (compare histology in Fig. 4D). Such findings indicate that the TNBS-induced changes result in significant tissue hypoxia.

Figure 4. Documentation of tissue hypoxia in colitis and influence of conditional hif1a mutation on colitis outcomes.

Panel A shows localization of EF5 staining in colonic section from vehicle control animal at day 3. Panel B shows corresponding H&E stain from the same block. Panel C reflects EF5 localization in sections from the distal colon in TNBS exposed animals demonstrating intense EF5 immunofluorescence overlying the area of inflammation. Panel C demonstrates H&E staining of from the same block displaying intense inflammatory infiltration. In panel E, body weight was monitored following induction of TNBS colitis (* p<0.025 by ANOVA). In panel F, colon length was measured at the time of sacrifice. Conditional hif1a-null mice displayed significant colon shortening compared to their WT littermates after TNBS administration (*p<0.025 mutant vs. littermate TNBS mice). Data are expressed as mean ± SD percent colon length change and are pooled from 3−5 animals in each condition.WT (n=4) and mutant (n=5).

We next confirmed our previous colitis findings in animals bearing conditional inactivation of epithelial hif1a² and focused on earlier time points (<7 days) of colitis. As shown in Figure 4E and 4F, and as we have previously reported², TNBS-induced colitis in conditional hif1a mutant animals was clinically more severe. Weight loss is a reliable method to assess TNBS colitis severity²⁸, and compared to littermate controls, conditional hif1a mutant animals lost weight more rapidly, to a greater extent, and failed to gain weight during the course of a 4 day experiment (p<0.025 by ANOVA, Figure 4E). Moreover, colon length (reflected as colon shortening related to more severe inflammation) was significantly different in conditional hif1a mutant mice (p< 0.025 compared to littermate controls, Figure 4F). These findings implicate HIF-1α as an endogenous protective mechanism in murine colitis, and provide a strong rationale for the use of PHD inhibitors as therapeutic modalities for colitis.

Based on pilot experiments, we elected to dose mice with FG-4497 at 20 or 40 mg/kg IP on the day before TNBS, 4 hrs prior to instillation of TNBS, and universally on the day after TNBS administration (i.e. 3 doses at day −1, 0 and +1). As shown in Figures 5A and 5B, dosing of 20 mg/kg and 40 mg/kg FG-4497, respectively, on day −1, day 0, and day +1 attenuated initial weight loss (p<0.05 for both), and more significantly protected weight loss with the ongoing course of recovery (p<0025 for both). Body weight results are summarized for day 3 measurements in Figure 5C and indicate that FG-4497 is generally beneficial for the weight loss outcome of colitis. No differences were observed between the 20 and 40mg/kg doses (p = not significant). In data not shown here, dosing of animals with FG-4497 at 40mg/kg only at 4 h prior to instillation of TNBS was not effective at ameliorating weight loss resulting from colitis. Histologic examination of colonic sections revealed a significant degree of protection from TNBS colitis in animals administered FG-4497. As shown in Figure 6A, colitic animals receiving drug vehicle showed tissue thickening, a loss of epithelia, prominent inflammatory infiltrate and an overall loss of architecture. By contrast, animals receiving 40 mg/kg FG-4497 showed less inflammatory infiltrate, a more intact epithelium and an overall preservation of tissue architecture.

Figure 5. Influence of FG-4497 on changes of body weight following induction of TNBS colitis.

Wild-type mice received vehicle or FG-4497 at indicated concentrations at day −1, day 0 and day +1. Mice were subjected to induction of TNBS colitis at day 0. Controls received TNBS vehicle. Body weight was monitored following induction of TNBS colitis (* p<0.025 by ANOVA) in animals receiving FG-4497 at 40mg/kg (panel A) or 20mg/kg (panel B). Panel C represents a weight loss summary on day 3. Data are expressed as mean ± SD percent change in initial body weight and are pooled from 4−6 animals in each condition, where * p<0.025 in animals receiving vehicle as compared to FG-4497.

Figure 6. Influence of FG-4497 on changes in colon length and tissue TNFα and IFNγ.

Wild-type mice received vehicle or FG-4497 at indicated concentrations at day −1, day 0 and day +1. Mice were subjected to induction of TNBS colitis at day 0. Controls received TNBS vehicle. Panel A depicts histological sections on day 3 post-TNBS from animals receiving vehicle or FG-4497 (40mg/kg). In panel B, colon length was measured at sacrifice and compared between animals receiving FG-4497 at 40mg/kg or 20mg/kg. Panel B represents TNFα analysis by real-time PCR in tissue derived from animals exposed to a combination of TNFα and FG-4497. Data are expressed as mean ± SD and are pooled from 4−6 animals in each condition, where * p<0.025 and **p<0.01.

Colon length is a marker of colonic inflammation and is reflected as increased shortening due to increased fibrosis associated with inflammation². As shown in Figure 6B, FG-4497 significantly attenuated colon shortening associated with TNBS colitis (p<0.05 by ANOVA). Moreover, changes in colon length were dose-dependent, with the 40mg/kg dose more effectively influencing colon length than 20mg/kg (p<0.05).

As an additional inflammatory endpoint, we examined changes in TNFα and IFNγ transcript levels. TNFα and IFNγ are inflammatory cytokines which have been shown to correlate with the severity of TNBS-induced colitis²⁹. In these studies, we obtained mRNA from epithelial-enriched colonic scrapings² and performed real-time PCR analysis for TNFα and IFNγ mRNA. Mice subjected to TNBS-induced colitis showed a 3.8±0.4 and 4.9±0.6-fold increase in TNFα and IFNγ, respectively, compared to control animals (Figure 6C and 6D). Correlating to a better outcome in weight loss and colon length, mice administered 40 mg/kg FG4497 on day −1, 0, 1 showed a significant decrease in TNFα (p<0.025) and IFNγ (p<0.01) mRNA, representing a 65±7% and 75±8% decrease in TNFα and IFNγ mRNA, respectively, when compared to TNBS alone. As an aside, dosing animals only on day 0 and on day 1 with 40 mg/kg also significantly (p<0.01) prevented excessive TNBS colitis mediated increase of TNFα mRNA levels (2.3 ± 0.6 fold increase compared to TNBS alone). Overall, these results suggest a beneficial influence of HIF activation on colitis outcomes, including weight loss, colon fibrosis and pro-inflammatory mRNA expression.

Influence of FG-4497 on collagen gel contraction

In the course of these studies, we considered whether PHD inhibitors influence wound healing and tissue remodeling. In particular, administration of FG4497 was most effective at early stages of colitis Figure 5A and 5B) and significantly enhanced the re-epithelialization of colitic wounds (Figure 6B). To model this event, we adapted a previous model of wound healing²⁴ to assess the influence of FG4497 on murine fibroblast collagen gel contraction. Murine fibroblasts stably transfected with HRE showed a FG4497 concentration-dependent induction of HIF activity (Figure 7A, p<0.001 by ANOVA), with significant increases at concentrations as low as 1nM (p<0.05). Using these conditions, we determined the extent to which FG4497 induced collagen gel remodeling and contraction. As shown in Figure 7B and 7C, exposure of fibroblast embedded gels to FG4497 enhanced contraction of collagen gels in a concentration-dependent manner (p<0.001 by ANOVA), with highest concentrations of FG4497 increasing contraction by 7.5±1.2-fold (p<0.001). Moreover, HIF activity significantly correlated with collagen gel contraction at the various FG-4497 concentrations (in linear plot of HIF activity versus gel contraction, Figure 7C, R² = 0.94, p<0.01). These findings directly implicate HIF in tissue remodeling and wound contraction.

Figure 7. Influence of FG4497 on collagen gel contraction.

NIH 3T3 fibroblasts stably transfected with a luciferase-HRE (NIH3T3/HIF-luc) were incorporated into collagen gels with and without indicated concentrations of FG-4497. Panel A shows HRE-luciferase responses to PHD inhibitor FG-4497, where indicated * p<0.025 and ** p<0.01. Panel B shows images of contracted of gel matrices in response to FG4497 for 24h. Collagen gel matrices were incubated with (i) 0, (ii) 1 nM, (iii) 10 nM, (iv) 100 nM and (v) 1 μM FG4497. Gel surface area quantified in terms of total pixel number using ImageJ, where indicated * p<0.025 and ** p<0.01. Panel C depicts the correlation between FG-4497-induced HIF activity and collagen gel contraction.

Discussion

These studies provide new insight into the potential therapeutic use of PHD inhibitors on the acute phases of mucosal inflammation. Based on previous work implicating a protective role for HIF-1α in mucosal inflammatory disease, we focused on defining epithelial HIF activation and colitis endpoints. These studies extend previous work into the role of HIF in vivo and provide new insight into the potential use of PHD inhibitors as therapeutic modalities.

HIF protein expression and activity occur primarily through post-translational modification of the alpha subunit of the HIF heterodimer³⁰. The discovery of HIF PHD's identified targets for the possible pharmacological control of HIF expression, whereby inhibitors activate the stabilization of the alpha subunit of HIF and PHD activators promote the degradation of HIF³¹. Our previous work in vitro and in vivo utilizing mucosal inflammatory models have identified HIF-1 as an endogenously protective molecule². To extend these findings, we sought to define the functional role of systemic HIF activation. We demonstrate here a potent and specific upregulation of fully functional HIF-1α in vitro and in vivo by administration of a novel class of HIF prolyl hydroxylase inhibitors, FG-4442 and FG4497, confirming the unique regulation of HIF-1α half life via hydroxylation of conserved proline residues^7-9. The capacity of FG-4497 to activate HIF was demonstrated by several means, including the induction of HIF-responsive luciferase constructs, the induction of EPO and EPO-dependent endpoints in vivo, and the lack of HIF response in epithelial conditional hif1a-null animals. It is currently not known to what extent differences exist between functional PHD-1, PHD-2 or PHD-3 expression in various cells or tissues. The most recent work in this regard utilized monoclonal antibodies directed against the various PHD molecules, and revealed that the PHD-1, -2 and -3 are widely expressed in all tissues and cells examined, including the epithelium³². Likewise, it is currently not known to what extent differences exist between functional HIF subunits. While it is likely that the influence seen here with colitis is HIF-1α-mediated, our previous studies also suggested that HIF-2α is expressed in the colon². Further studies in intact tissue will be necessary to define these principles.

Inflammatory diseases such as ulcerative colitis or Crohn's disease are heavily impacted by changes in energy demand and tissue oxygenation. It is recently appreciated that tissue hypoxia and inflammation occur coincidentally, a condition we have termed “inflammatory hypoxia”³³. Such findings are evidenced by increased localization of oxygen-sensitive (nitroimidazole) dyes within the inflamed tissue². Moreover, studies as background for the current work revealed that HIF activation accompanying inflammatory disease subserves the mucosa in a protective manner². These studies have resulted in the identification of several targets of inflammation transcriptionally controlled by HIF-1α. Most prominently implicated are genes which promote nonclassical epithelial barrier function and control leukocyte trafficking, including the ecto-5'-nucleotidase (CD73)¹⁷, P-glycoprotein¹⁶, the adenosine A2B receptor³⁴, intestinal trefoil factor²⁶ and CD55³⁵. These findings of HIF activation in murine colitis models are consistent with studies in human patients indicating that HIF-1α expression is consistently induced in both ulcerative colitis³⁶ and in Crohn's disease³⁷, as well as ischemic colitis³⁸. Here, we demonstrate both a barrier protective and an anti-inflammatory role for HIF PHD inhibitors in inflammatory mucosal disorders. Indeed, administration of FG-4497 24 h prior to TNBS challenge clearly reduces the initial drop in weight when compared to the respective controls, and thus may be attributable to the concerted upregulation of barrier protective and immunoregulatory genes mutually involved in the adaptive response to decreased oxygen. In support of this notion, we demonstrate a significant decrease (∼40%) in baseline permeability when HIF levels are stabilized, correlating with results using genetically engineered for constitutive over-expression of HIF-1alpha (via expression of mutant von Hippel-Lindau)². To rule out possible non-HIF mediated drug related effects on barrier, we used conditional hif1a-null mice, wherein no changes were observed in response to pharmacological PHD inhibition in vivo. Our studies also showed a beneficial influence of PHD inhibition on colon TNFα, a molecule with a clear pathophysiologic role in colitis³⁹. These studies revealed a dose-dependent decrease in tissue TNFα and IFNγ mRNA levels, with nearly 70% inhibition. In this context, TNFα and IFNγ levels provide a clear indicator of overall colonic inflammation and suggest an anti-inflammatory role for PHD inhibition in inflammatory disorders of the colon.

The intestine, particularly the colon, may be a unique tissue for which to study HIF function. Previous studies, in fact, have established a view of the intestine as relatively hypoxic under normal physiological conditions^{2, 40, 41}, with a steep oxygen gradient from anaerobic lumen across the highly metabolic epithelium. In this context it is not surprising that the intestinal epithelium is comparatively resistant to hypoxic challenge³, assuming that “physiologic hypoxia” potentially elevates baseline levels of HIF-regulated gene activity. To this end, it was shown that the expression of HIF-target genes (e.g. CD73) is highest in the colon, with other epithelial tissues (e.g. kidney) expressing nearly two-fold less^{4, 42}. Remarkable in this context, PHD inhibitors have shown efficacy in epithelial tissues which are considered highly oxygenated. For example, it was recently reported that the PHD inhibitor FG-4095, , promotes VEGF and angiogenesis in human microvascular endothelia and in baboon lung explants⁴³. Important in this regard, and likely explanatory, FG-4095 s was shown to be effective at even high oxygen concentrations (e.g. 95% O₂). Thus, as a class, the actions of PHD inhibitors are likely to be relatively insensitive to local oxygen concentrations.

As guided by findings of increased histologic re-epithelialization by FG-4497, we explored the possible interplay between HIF and wound healing. For these purposes, we used the established collagen gel model system to measure functional outputs associated with tissue remodeling²⁴ with NIH 3T3 fibroblasts stably expressing a luciferase-linked HRE. These studies revealed a remarkable increase in collagen gel contraction induced by FG-4497. Moreover, HIF activity correlated with the extent of contraction, and thereby linking HIF activity with collagen contraction. We do not currently know how HIF influences collagen contraction. While much is known about HIF and angiogenesis⁴⁴, far less is known about the role of HIF in wound healing. It is possible that the same mechanisms which drive angiogenesis also control wound healing, particularly since many of the cytokines and growth factors overlap in the GI tract⁴⁵. Alternatively, it was recently shown that HIF-mediated induction of the alpha subunit of heat shock protein 90 (Hsp90α) significantly enhances cell migration and wound closure. Whether these studies explain our findings with FG-4497-mediated collagen contraction await further experimentation. Nonetheless, these findings directly implicate HIF in tissue remodeling and suggest that PHD inhibitors such as FG4497 might significantly enhance mucosal restitution and wound healing.

Taken together, the combined inflammatory parameters of weight loss, colon length, and TNFα/IFNγ expression suggest a protective role for PHD-dependent induction of HIF in murine experimental colitis. These results demonstrate the beneficial impact and regulatory influence of HIF stabilization via PHD inhibition and hence propose a novel therapeutic approach to inflammatory intestinal diseases like ulcerative colitis and Crohn's disease using HIF prolyl hydroxylase inhibitors. Future work will be aimed at better defining these parameters, including gene targets and mechanisms of anti-inflammation, as well as tissue-specific influences of PHD inhibition.