An Endogenously Anti-Inflammatory Role for Methylation in Mucosal Inflammation Identified through Metabolite Profiling

The inflammatory bowel diseases (IBD), including both Crohn’s disease and ulcerative colitis, are debilitating diseases of unknown etiology (1), but a combination of genetic and environmental factors are thought to be involved in disease pathology. Recent evidence suggests that IBD likely results from dysregulated immune responses to luminal triggers. Intestinal epithelial cells are a crucial cell type in the maintenance of colonic tissue homeostasis, and IBD is characterized by a breakdown of the intestinal epithelial barrier leading to increased exposure of the mucosal immune system to antigenic luminal material. This exposure leads to inflammation and increased breakdown of the epithelial barrier (1–3). It is now appreciated that ongoing inflammation leads to significant shifts in tissue metabolism. These shifts result, at least in part, from perturbation of the vasculature (4, 5) leading to the decreased supply of oxygen and nutrients (6–8). The expression of increased levels of cytokines and chemokines initiates the migration of innate immune cells such as neutrophils, macrophages, and dendritic cells into the mucosal tissues (9). The recruitment of inflammatory cells increases the local immune response as well as the oxygen and nutrient demand of the tissue, exacerbating the metabolic changes associated with IBD. IBD almost certainly elicits changes in a number of cellular metabolic pathways; however, the metabolic shifts induced in the colonic epithelia during the inflammatory process have not been elucidated in a systematic manner.

Studies to date indicate that alterations of cellular methylation have important implications in inflammation and immune responses (10). These studies have shown that methylation, through epigenetic mechanisms and protein modification, has important consequences for these processes, yet the molecular mechanisms of such changes remain poorly understood. Cellular methylation reactions include modification of DNA, RNA, proteins, and lipids. These reactions all require a methyl donor for the modification of the target. The methyl donor for the majority of these reactions is S-adenosylmethionine (SAM) (11). SAM is distributed throughout the cell to act as donor for the various methyltransferases. The donation of its methyl group produces S-adenosylhomocysteine (SAH) from SAM. SAH is a potent inhibitor of methyltransferase enzymes because these enzymes have a higher affinity for SAH (12–14), and SAH is rapidly converted to homocysteine and adenosine by SAH hydrolase. Therefore, inhibition of SAH hydrolase represents a powerful means of inhibiting cellular methylation reactions (15). It has been known for a number of years that inhibition of methylation had immunomodulatory properties (16), but mechanisms remain poorly understood. This led to the development of more specific and less toxic SAH inhibitors for use in animal models. The SAH hydrolase inhibitor methyl 4-(adenin-9-yl)-2-hydroxybutanoate (DZ2002) has been found to have immunosuppressive actions and ameliorate disease in a number of animal models (17–21). One mechanism for the immunosuppressive action of DZ2002 was shown to be the cell type-specific inhibition of NF-κB activation (19). These studies demonstrate that cellular methylation reactions play an important role in chronic inflammatory disease, such as IBD.

The hypothesis of the current study was that ongoing mucosal inflammatory responses are characterized by specific shifts in metabolic profiles. To define these principles, a nuclear magnetic resonance (NMR)-based metabolomics approach was used to elucidate the metabolic changes associated with murine models of IBD and in cellular models of mucosal inflammation. This analysis revealed that a number of metabolites associated with cellular methylation reactions are significantly impacted by ongoing inflammation. Dovetailing these findings with microarray analysis of epithelial inflammation models, we found that the expression of enzymes central to cellular methylation is significantly altered during mucosal inflammation. We confirmed these findings at the molecular level and extended these studies to a murine model of colitis. These results indicate that methylation reactions play an important role in mucosal inflammation.

Human intestinal epithelial cells (T84) and HeLa epithelial cells were grown and maintained in T175 cell culture flasks (Costar, Cambridge, MA) as described previously (22). DZ2002 (Diazyme, San Diego, CA) was prepared as described previously (19). Where indicated, DZ2002 was added at the indicated concentrations by preincubation for1 h prior to the indicated treatments.

Cells were harvested by scraping, pelleted by centrifugation, washed with ice-cold PBS, and lysed by sonication in Tris lysis buffer (150 mM NaCl, 20 mM Tris [pH 5.5], 1 mM EDTA, 1 mM EGTA, and 1% Triton X-100). Similarly, snap-frozen colon tissues samples were sonicated in Tris buffer. Protein concentration was assessed by the bicinchoninic acid assay following the manufacturer’s instructions (Thermo Scientific) to ensure equal protein loading of each preparation. Proteins were separated by SDS-PAGE electrophoresis and transferred to polyvinylidene difluoride membrane (Bio-Rad Laboratories, Hercules, CA) for immunoblotting. Abs used for this study were as follows: anti-SAM synthetase (1/1,000; Abcam), anti-SAH hydrolase (1/1,000; gift from Dr. D. Kloor, University of Tuebingen, Tuebingen, Germany), anti-IκBα (1/2,000; Cell Signaling Technologies), anti–β-actin (1/10,000; Abcam), anti–β-tubulin (1/10,000; Abcam). Proteins were visualized using the SuperSignal detection substrate (Thermo Scientific).

RNA was isolated using TRIzol (Invitrogen) and cDNA synthesized as described previously (10). Potential contaminating genomic DNA was digested using Turbo DNA-free (Ambion, Austin, TX). The mRNA profile of intestinal epithelial cells (T84) subjected to modeled inflammation (IFN-γ at 10 ng/ml for 6 or 18 h) was compared with that of control cells exposed to vehicle only using Genechip expression arrays (Affymetrix) as described previously (23). Semiquantitative and real-time PCRs were performed using increasing numbers of cycles of 95°C for 45 s, 59°C for 35 s, and 72°C for 45 s and a final extension time of 7 min. The following primers were used to quantify expression in intestinal epithelial cells: MAT2A (SAM synthetase), 5′-ATACAATCTACCACCTACAGCC-3′ (forward) and 5′-CATAAGAGACCTGAACAAGAACC-3′ (reverse); AHCY (SAH hydrolase), 5′-GGTATCGGTTGAAGAATGG-3′ (forward) and 5′-GGTACTTGTCTGGATGGGTC-3′(reverse); DNMT3B, 5′-ATAAGACACCCCCTCAAACC-3′ (forward) and 5′-TTCCCGTTCTCCCTAAAAAC-3′ (reverse); IL-8: 5′-ATGACTTCCAAGCTGGCCGTGG-3′ (forward) and 5′-CATAATTTCTGTGTTGGCGCAGTGTGG-3′ (reverse); TNF-α, 5′-TCCTTCAGACACCCTCAACC-3′ (forward) and 5′-AGGCCCCAGTTTGAATTCTT-3′ (reverse); and β-actin, 5′-GCACTCTTCCAGCCTTCCTTCC-3′ (forward) and 5′-CAGGTCTTTGCGGATGTCCACG-3′ (reverse). Transcript levels and fold change in mRNA were determined as described previously (24). Intron-spanning primer pairs for were designed by using Primer3 software (http://frodo.wi.mit.edu/). Primer properties and secondary structures including hairpins, self-dimers, and cross-dimers were evaluated in a second step using Netprimer software (http://www.premierbiosoft.com/netprimer). Samples were controlled for β-actin.

HeLa cells were passaged into 24-well plates and allowed to attach for 24 h. Transient transfection of Hela cells and assessment of luciferase activity was carried out as described previously (22).

Dextran sodium sulfate (DSS) colitis was induced with a modification of the technique of Okayasu et al. (25). Colitis was induced on day 0 by the addition of 3% DSS (m.w. = 36,000–50,000; MP Biomedicals, Illkirch, France) solution in drinking water. Control animals received water alone. DZ2002 was delivered i.p. at a concentration of 50 mg/kg/day (19) beginning 1 d prior (day −1) to DSS administration. Folic acid (Sigma-Aldrich) was delivered i.p. at a concentration of 50 mg/kg beginning 1 d prior (day −1) to DSS administration and then every other day for the course of the experiment. BAY 11-7082 (Cayman Chemical) was administered i.p. at a concentration of 5 mg/kg beginning 1 d prior (day −1) to DSS administration and then every other day for the course of the experiment, as described previously (26).

For cytokine analysis, colonic tissue was extracted in Tris lysis buffer by sonication, and protein homogenates were stored at −80°C until use. Tissue concentrations of IFN-γ were measured in colonic protein extracts using a proinflammatory cytokine screen (Meso Scale Discovery). Assays were performed per manufacturer’s instructions and analyzed using a Sector Imager 2400 (Meso Scale Discovery). IFN-γ concentrations were normalized to total protein concentration.

5-Methyl cytidine staining was quantified using flow cytometry essentially as described by Watson et al. (27). Briefly, cells were trypsinized and fixed in 70% MeOH. Following fixation, cells were incubated in 1 M HCl for 1 h at 37°C. Cells were incubated with anti–5-methyl cytidine Ab (Abcam) or IgG control at identical concentrations, followed by application of Alexa Fluor 555 (Invitrogen) secondary Ab. Cells were analyzed using the FACSCanto system (BD Immunocytometry Systems, San Jose, CA). Postanalyses were performed using FlowJo software (Tree Star, Ashland, OR).

T84 human epithelial cells were cultured on glass coverslips and fixed in ice-cold methanol, followed by permeablization in 0.2% Triton X-100. Cells were incubated in 1 M HCl for 1 h at 37°C. Cells were incubated with blocking buffer (1% BSA) and then incubated with anti–5-methyl cytidine Ab (Abcam) or IgG-negative control, followed by Alexa Fluor 555 secondary Ab. Coverslips were mounted using ProLong Gold antifade (Invitrogen). Staining was visualized using Axioplan Zeiss microscope equipped with an AxioCam MRc5 system. Imaging software for data capture was Axiovision 4.6.

Blank sections were cut from formalin-fixed, paraffin-embedded samples. Following deparaffinization, slides were subjected to Ag retrieval by using a sodium citrate buffer. Sections were permeabilized using 2% Triton X-100, followed by blocking in 1% BSA. Sections were incubated with anti–5-methyl cytidine Ab, followed by incubation with Alexa Fluor 555 secondary Ab. Sections were counterstained with DAPI (Invitrogen). Images were visualized and captured as described above.

Collected cell pellets or frozen tumor specimens were homogenated in ice-cold 8% perchloric acid as described previously (28, 29). Briefly, after centrifugation, the supernatants (containing hydrophilic metabolites) were collected, and pH was adjusted to pH 7 by using KOH. The potassium perchlorate was removed by centrifugation, and the hydrophilic fraction was lyophilized overnight. The cell and tissue pellets (after the first centrifugation), which contained the lipophilic metabolites, were redissolved in water, and the pH was adjusted (7.0). The lipophilic fraction was lyophilized overnight. For the cell experiments only, previously collected media were lyophilized overnight after adjusting pH. The dried hydrophilic cell and tissue extracts were redissolved in 0.5 ml deuterium oxide, transferred into 5-mm NMR tubes, and used for [¹H]-, [³¹P]-, and [¹³C]-NMR analysis. The medium extracts were redissolved in 1 ml deuterium oxide, transferred to 5-mm NMR tubes, and underwent [¹H]- and [¹³C]-NMR analysis. The cell and tissue lipid extracts were redissolved in 1.2 ml deuterated chloroform/deuterated methanol mixture (2:1 v/v).

All [¹H]- and [¹³C]-NMR spectra were obtained by the Bruker 500 MHz DRX NMR spectrometer using an inverse Bruker 5-mm TXI probe. To assist [¹H]-NMR peak assignment and metabolite identification in cell, media and biopsy extracts, two-dimensional–hydrogen/carbon-heteronuclear single quantum correlation NMR techniques were used. All spectra were Fourier transformed, and lactate (Lac3, CH₃) was used as an internal chemical reference for both carbon (21 ppm) and proton (1.32 ppm) axes. For metabolite quantification, one-dimensional [¹H]-NMR spectra were obtained from each sample, with a standard water presaturation pulse program “zgpr”. A thin sealed glass capillary, containing trimethylsilyl proprionate, was placed in each 5-mm tube prior to [¹H]-NMR experiments. The total number of acquisitions varied from 40 to 128. Conventional [¹H] acquisition parameters were power level pl1 = 20 dB, power angle p1 = 6.3 ms (90° pulse), power level for water presaturation pl9 = 77 dB, water suppression at O1P = 4.76 ppm, and spectral width SW = 5000 MHz, and the pulse delay of 12.75 s (calculated as 5*T1) was applied between acquisitions. The trimethylsilyl proprionate from the reference capillary served as a chemical shift (0 ppm) and proton metabolite concentration reference. The cell and media extracts subsequently underwent [¹³C]-NMR analysis (with proton decoupling). The total number of scans was 24,000 and 800 for each cell and medium extract, respectively. The concentration of [3-¹³C] lactate at 21 ppm (calculated from two satellite peaks at the [¹H]-NMR spectra) served as an internal carbon concentration reference.

Before [³¹P]-NMR analysis, 100 mmol/l EDTA was added to the extracts to complex divalent cations (for ATP and ADP quantification). All [³¹P]-NMR spectra (with proton decoupling) on cell and tissue extracts were obtained by the Bruker 300 MHz Avance NMR spectrometer using a Bruker QNP probe. The total number of scans was 8,000–16,000 per extract. A thin capillary glass containing 2.3 mmol/l methyl-diphosphoric acid served as a chemical shift (18.6 ppm) and phosphor metabolite concentration reference.

Histological examination was performed on three samples of the distal colon. Samples were fixed in 4% formalin before staining with H&E. All histological quantitation was performed in a blinded fashion, using a previously described scoring system (30). Severity of inflammation: 0, rare inflammatory cells in the lamina propria; 1, increased numbers of inflammatory cells; 2, confluence of inflammatory cells extending into the submucosa; 3, transmural extension of the inflammatory cell infiltrate. Extent of injury: 0, nil mucosal damage; 1, discrete lymphoepithelial lesions; 2, surface mucosal erosion; 3, widespread mucosal ulceration and extension through deeper bowel wall structures. The scores of the two parameters were added and the mean was calculated. Maximum possible score was 3.

Unpaired t test and/or one-way ANOVA test were used to determine differences between groups, as indicated. The significance level was set at p < 0.05 for all tests (Prism 4; GraphPad Software, San Diego, CA).

Previous studies have shown that inflammatory lesions associated with IBD are the site of significant metabolic changes (31–33). These changes in metabolism, however, have not been studied in a systematic manner. In this study, we sought to define the metabolic changes associated with modeled inflammation. Because previous reports strongly implicated intestinal epithelial cells (23, 32), we used T84 cells, a well characterized human intestinal epithelial cell line. Initial experiments involved subjecting T84 human colonic epithelial cells to IFN-γ for 48 h. In addition, cells were cultured in the presence of [1-¹³C]glucose for the final 4 h of culture to obtain data on the fate of glucose during inflammation. Following the incubation period, cells were immediately harvested and prepared for magnetic resonance spectroscopy (MRS) analysis. The media were also preserved and subjected to metabolite quantification. It has been previously shown that intestinal inflammation results in a state termed “inflammatory hypoxia” (32). In addition, it has been previously shown that treatment of human epithelial cells with proinflammatory cytokines leads to an increase in glycolytic rate (34, 35). The present study corroborates these studies with the finding that treatment of T84 cells with IFN-γ results in an increase in glucose uptake and lactate production as well as intracellular adenosine levels, all of which are indicative of hypoxia (Fig. 1A, Supplemental Fig. 1, Supplemental Table I). Importantly, we also found changes in a number of metabolites that participate in (betaine and cholines) or are products of (creatine, taurine, and adenosine) cellular methylation reactions (Fig. 1A, Supplemental Fig. 1, Supplemental Table I). These results strongly implicate methylation reactions as a contributing factor to mucosal inflammatory responses. Moving forward, we compared these results with those obtained from microarray analysis (Fig. 1B).

After finding that key metabolites of methylation reactions were changed in our in vitro model of mucosal inflammation, we compared these metabolomic end points with microarray data generated in T84 intestinal epithelial cells following exposure to IFN-γ for either 6 or 18 h. As shown in Fig. 1B, a number of genes strongly implicated in IFN-γ signaling (e.g., CXL11, CXCL10, and IDO1) were highly induced by IFN-γ and thus served as positive controls for this microarray. More relevant for the current study, this microarray also revealed strong induction of several methylation-related gene transcripts, including SAM synthetase (MAT2A), SAH hydrolase (AHCY), O-6-methylguanine-DNA methyltransferase, and guanidinoacetate N-methyltransferase. MAT2A, the enzyme responsible for SAM production, was among the most highly upregulated genes in the array (Fig. 1B). Given its central role in all methylation reactions (14), we pursued the induction of MAT2A by IFN-γ. To confirm this microarray result, we performed real-time PCR using primers specific for the MAT2A and examined SAM synthetase protein levels by immunoblot analysis. These results indicate that both MAT2A expression and SAM synthetase protein are increased in epithelial cells following exposure to IFN-γ (Fig. 1C, for both p < 0.05 by ANOVA). In a similar manner, we also examined the expression of SAH hydrolase. This enzyme plays a key role in methylation reactions by catabolizing SAH, a potent methyltransferase inhibitor, to homocysteine and adenosine (14). These studies revealed that SAH hydrolase expression and protein levels are also upregulated during inflammation (Fig. 1D). Taken together, these results and that of metabolomic analysis provide strong evidence that cellular methylation pathways are important in the inflammatory process.

Having determined that key enzymes of cellular methylation are upregulated during inflammation, we next examined functional end points of these reactions, namely DNA methylation. For these studies, T84 epithelial cells were exposed to IFN-γ (10 ng/ml) for the indicated times and analyzed levels of 5′-methylcytidine, which is indicative of DNA methylation, using fluorescent microscopy and flow cytometry. As depicted in Fig. 2A, untreated T84 cells exhibit a small amount of nuclear 5′-methylcytidine staining. In stark contrast, as shown in Fig. 2B, nuclear staining for 5′-methylcytidine is substantially increased following exposure to IFN-γ for 24 h. This result was corroborated using flow cytometry. As depicted in Fig. 2C, incubation of T84 cells for 24 h with IFN-γ resulted in a 2.23 ± 0.02 increase in 5′-methylcytidine (p < 0.01). An insight into a potential mechanism for the increase in DNA methylation came when the expression of DNMT3b, a DNA methyltransferase responsible for de novo DNA methylation (14), was examined in cells exposed to IFN-γ. As shown in Fig. 2D, DNMT3b expression is significantly induced in T84 cells in an inflammatory setting (p < 0.01 by ANOVA). These data indicate that DNA methylation is increased in an in vitro model of inflammation.

NF-κB is one of the master regulators of proinflammatory gene expression (for review, see Ref. 36). NF-κB upregulates a number of genes including proinflammatory cytokines, chemokines, and adhesion molecules and can induce specific sets of genes in response to a particular triggers (37). NF-κB has been shown by EMSA and IHC studies to be highly activated at the site of inflammation in a number of diseases, including IBD (36), as wells as activated under conditions of hypoxia (38, 39). Additionally, NF-κB has been found to be protective of the mucosal epithelium in murine disease models (40, 41). Interestingly, NF-κB has been shown to elicit epigenetic modifications of its inducible genes (42). Based on these studies and our preliminary data suggesting a link between methylation and the inflammatory response in epithelial cell models, we examined NF-κB activity using a luciferase reporter assay following inhibition of methylation by DZ2002, a potent reversible SAH hydrolase inhibitor (21). As shown in Fig. 3A, IFN-γ elicited a >8-fold increase in NF-κB activity at the highest concentrations (p < 0.01). Pretreatment with DZ2002 effectively inhibited NF-κB activity in a concentration-dependent manner (Fig. 3A). As IFN-γ does not represent a canonical activator of NF-κB, we also studied the impact of DZ2002 on NF-κB following treatment with TNF-α. As shown in Fig. 3B, although TNF-α is a much more potent NF-κB activator (maximal 58.02 ± 9.66; p < 0.001), inhibition of methylation significantly repressed such induction, likely indicating that this inhibition of NF-κB by DZ2002 is not stimulus dependent.

To confirm that DZ2002 directly impacts NF-κB, we next examined the fate of the NF-κB inhibitory protein, IκBα, in response to DZ2002. As shown in Fig. 3C, IκBα levels are decreased in response to IFN-γ treatment. This response is abrogated with pretreatment with DZ2002 indicating that the inhibition of NF-κB is mediated, at least in part, through modulation of the inhibitory protein IκBα. To confirm these results at a functional level, we examined the expression of TNF-α and IL-8, two known NF-κB target genes. As depicted in Fig. 3D and 3E, expression of both TNF-α and IL-8 is increased in response to IFN-γ treatment. In support of our hypothesis, expression of both of these targets was diminished upon pretreatment with DZ2002 (p < 0.05). These results strongly implicate methylation in NF-κB activation, at least in part, through the regulation of IκBα.

Previous studies have shown that inflammatory lesions in mucosal inflammation (e.g., IBD) result in substantial metabolic changes and major shifts in metabolite supply and demand (10). In this study, we sought to define the metabolic changes induced during inflammation in an epithelial-driven in vivo model of experimental colitis. For these purposes, we selected the DSS model. C57BL/6 mice were administered DSS or vehicle (water) alone for 5 d, after which mice were euthanized, and colon tissue was immediately removed and frozen. Colons were extracted and prepared for MRS analysis. Fig. 4 displays representative proton spectra for both control (Fig. 4A) and DSS-treated (Fig. 4B) colonic tissue. These experiments revealed alterations in a number of metabolites associated with methylation reactions, including adenosine, betaine, choline, creatine, glutathione (p < 0.05; n = 6) (Fig. 4C, Supplemental Table II). In addition to an impact on metabolites involved in methylation, changes in several metabolites are indicative of hypoxia (glucose, lactate, and adenosine), consistent with previous work in experimental colitis models (23, 32, 33, 43). To confirm our in vitro experiments, we examined the parallel expression of SAM synthetase/SAH hydrolase and IFN-γ in colonic tissue extracts from control and DSS-treated mice harvested at 2 and 4 d of treatment. As shown in Fig. 4D, both SAM synthetase and SAH hydrolase expression were increased after 2 and 4 d of DSS administration. Likewise, tissue concentrations of IFN-γ were increased at day 4 of DSS treatment (Fig. 4E).

We next examined the impact of methylation-inhibition on DSS colitis outcomes. Given that the SAH hydrolase (AHCY)-null mouse is embryonic lethal (44) and that SAM synthetase is not easily inhibited pharmacologically (15), we reverted to a pharmacological approach using DZ2002. Our in vitro data revealed that modeled inflammation results in an increase in DNA methylation in epithelial models, as evidenced by an increase in 5′-methylcytidine staining detected both by immunocytochemistry and flow cytometry (Fig. 2). On the basis of these findings, we initially examined the extent of DNA methylation in response to DSS in the presence and absence of DZ2002 in vivo. As shown in Fig. 5, treatment with DSS resulted in a significant increase in colonic DNA methylation. Colonic tissue from vehicle-treated animals displayed low levels of basal 5-methylcytidine staining (Fig. 5, column 1). By contrast, DSS induction resulted in a substantial increase in DNA methylation as indicated by the increase in 5′-methylcytidine and DAPI/5′-methylcytidine costaining (see arrows in merged image, column 2). Treatment of animals with DZ2002 in combination with DSS significantly dampened DNA methylation as indicated by the decrease in both 5′-methylcytidine staining as well as DAPI/5′-methylcytidine colocalization (see arrows in merged image, column 3).

We next demonstrated that DZ2002, administered in combination with administration of DSS, exacerbated disease progression. As shown in Fig. 6A–C, inhibition of methylation in combination with DSS treatment resulted in significant decrease in weight (p < 0.05; Fig. 6A), increased colon shortening (p < 0.001; Fig. 6B), and an increase in disease activity as indicated by histological score (p < 0.001; Fig. 6C). Importantly, DZ2002 treatment alone resulted in no weight loss or colon shortening (Fig. 6A, 6B), indicating that this inhibitor is not toxic at the doses administered. In addition, these indicators of disease severity were confirmed by histology. As shown in Fig. 6D, vehicle-treated mice display normal tissue morphology, including intact epithelium and crypt structure. DSS-treated animals exhibited a deterioration in normal crypt architecture and epithelial cell depletion (Fig. 6E), and animals treated with DZ2002 and DSS display an almost complete lack of normal tissue architecture and loss of the epithelium (Fig. 6F). These results confirm that inhibition of methylation worsens disease progression in a murine model of colitis and strongly implicate that inflammation-associated methylation functions as an endogenous protective mechanism. Importantly, the finding that NF-κB inhibition exacerbates epithelial-driven colitis is consistent with recent literature (45, 46).

In an attempt to clarify the role of the NF-κB pathway in the context of our findings, we administered the NF-κB inhibitor BAY 11-7082 to mice, alone or in combination with DSS. This inhibitor selectively and irreversibly inhibits NF-κB activation by blocking cytokine-induced phosphorylation of IκB-α without influencing constitutive IκB-α phosphorylation (47). Although animals receiving BAY 11-7082 alone showed no symptomology compared with vehicle controls, mice receiving both BAY 11-7082 and DSS developed significantly more severe disease, to the extent that it was necessary that we humanely sacrifice animals prior to completion of the DSS time course (data not shown). These experiments do not provide direct evidence of a role for NF-κB activation in the protection of the mucosa in a DSS model of colitis. However, the results do corroborate our findings and those of others indicating that NF-κB is protective within the intestinal epithelium (40, 41, 48) and suggest that the endogenously protective effect of methylation may be due, at least in part, to epithelial NF-κB activation.

Having shown that inhibition of methylation using DZ2002 in combination with administration of DSS worsens colitis in mice, we next determined whether augmentation of methylation would be protective in the DSS model. For these experiments, we administered folate in the same manner as the methylation inhibitor (i.e., systemic via i.p. injection) to make the experiments as comparable as possible. As shown in Fig. 7A–C, administration of folate in combination with DSS treatment resulted in an increase in weight (p < 0.05; Fig. 7A), significantly decreased colon shortening (p < 0.05; Fig. 7B), and a significant decrease in disease activity as indicated by histological score (p < 0.01; Fig. 7C). Administration of folate alone had no effect on weight loss or colon length (Fig. 7A, 7B). Once again, these indicators of disease severity were confirmed by histology. As shown in Fig. 7D, vehicle-treated mice display normal tissue morphology, including intact epithelium and crypt structure. DSS-treated animals exhibited deterioration in normal crypt architecture and epithelial cell depletion (Fig. 7E). Importantly, animals administered folate in combination with DSS display improved tissue architecture, diminished loss of the epithelium, and less inflammatory cell migration than DSS alone (Fig. 6F). These results confirm that augmentation of methylation through the administration of folic acid ameliorates disease progression and provides further evidence that inflammation-associated methylation is protective in a murine colitic model. These data are in agreement with clinical studies demonstrating vitamin B12 and folate deficiencies are common in IBD patients (49–51) and suggest folate supplementation in IBD treatment. In addition, a recent study demonstrated that a methyl-deficient diet exacerbates disease in an animal model of colitis (52).

This study aimed to identify metabolic changes associated with modeled intestinal inflammation, particularly related to the epithelium. As previous work had demonstrated specific shifts in metabolism during inflammation, we reasoned that such changes could be reflected on a more global basis. An amalgamated approach using NMR-based metabolomics and transcriptional arrays identified shifts in methylation-dependent pathways as a major metabolic fingerprint. These studies significantly extend previous work related to inflammation-associated metabolism and identify changes in methylation as a target signature within the epithelium during mucosal inflammation. Evidence is provided that shifts in methylation associated with inflammation serve an endogenously protective role in murine colitis.

Ongoing mucosal inflammatory responses are characterized by significant shifts in tissue metabolism (10). These changes include a shift toward a glycolytic phenotype and careful analysis has shown the development of significant hypoxia, termed inflammatory hypoxia, particularly prominent within the epithelium (53). Coinciding with inflammation-associated hypoxia is the stabilization of hypoxia inducible factor (32, 53), and given the central role of hypoxia inducible factor in most metabolic processes (54, 55), we reasoned that overall metabolism would be changed. In support of this hypothesis, global analysis of metabolism using MRS identified significant changes in a number of cellular metabolites. Many of these alterations are in agreement with results from our laboratory indicating the presence of inflammatory hypoxia. These include increased intracellular adenosine, increased glucose uptake and intracellular glucose, increased intracellular lactate and lactate export, and increased turnover of ATP (Figs. 1, 4, Supplemental Fig. 1, Supplemental Tables I, II). In addition, significant changes were observed in a number of lipid metabolites including monounsaturated fatty acids, triacylglycerol, polyunsaturated fatty acids, phosphatidyl choline, and phosphatidyl ethanolamine (Supplemental Tables I, II), which may be indicative of dynamic alterations in cellular membranes in response to inflammation. Importantly, this analysis revealed a distinct and specific methylation fingerprint. Cellular methylation reactions include modification of DNA, RNA, proteins, and lipids (56, 57). These reactions all require a methyl donor for the modification of the target. The methyl donor for the majority of these reactions is SAM (11). SAM is distributed within all cell and tissue compartments and functions as a methyl donor for a number of different methyltransferases. The donation of methyl groups results in the generation of SAH. Methyltransferase enzymes have a higher affinity for SAH than SAM, and thus, SAH functions as a potent feedback inhibitor (14). SAH is rapidly converted to homocysteine and adenosine by SAH hydrolase. On the basis of this methylation fingerprint, we performed parallel microarray analysis of modeled inflammation using a Th1 cytokine (IFN-γ) as a stimulus and discovered that SAM synthetase (MAT2A) and SAH hydrolase (AHCY) were among the top 20 highest IFN-γ–induced transcripts in T84 intestinal epithelia. The mammalian MAT2A gene promoter is relatively well-characterized. Previous studies have shown that the human MAT2A promoter is basally controlled by three tandem Sp1 sites (58). Under modeled inflammatory conditions (TNF-α activation), MAT2A is strongly induced through activation of AP-1 and NF-κB (59). These studies support our findings of MAT2A induction by IFN-γ, a known activator of both NF-κB and AP-1 (60). At present, essentially nothing is known about the regulation of the mammalian AHCY gene.

As a proof of principle, we extended these findings to a mucosal inflammation model. Given our understanding of inflammation-associated metabolic changes within the epithelium, we selected DSS colitis as an appropriate animal model to study colitis. DSS functions primarily as an epithelial irritant to drive permeability-induced colonic inflammation (61). Therefore, inhibition of SAH hydrolase represents a powerful means of inhibiting cellular methylation reactions (15). It has been known for several years that inhibition of methylation had immunosuppressive influences (16). This led to the development of more specific, reversible, and less toxic SAH inhibitors for use in animal models of inflammation. Using these compounds, it was demonstrated that SAH hydrolase inhibition particularly downregulates T cell activation and adaptive immune responses. One of these SAH hydrolase inhibitors, DZ2002, has been found to have potent immunosuppressive effects and ameliorates disease in a number of animal models including delayed-type hypersensitivity (20), arthritis (20), and experimental autoimmune encephalomyelitis (19). Although we do not know the exact mechanism by which DZ2002, we have shown that inhibition of methylation by DZ2002 inhibits NF-κB in vitro. NF-κB is a master regulator of the proinflammatory response, upregulating a number of genes including cytokines, chemokines, and adhesion molecules, as well as activated under conditions of hypoxia (62). A number of studies have shown that inhibition of NF-κB may be protective in animal models of IBD (63–65). Paradoxically, however, NF-κB activation has also been found to be protective in the context of the mucosal epithelium in murine disease models (40, 41). Our findings with DZ20002 support this hypothesis and demonstrate that parenteral administration of DZ2002 significantly inhibits tissue methylation and that such inhibition is associated with exacerbated DSS colitic responses. In addition, augmentation of methylation through the administration of folate ameliorates disease in this colitic model. Whether methylation inhibition would impact other models of intestinal inflammation in the same manner is not currently known.

Taken together, these studies provide a new and compelling role for methylation as an endogenously protective mechanism for mucosal inflammation. In addition, the findings in this study provide a tractable and potentially new therapeutic opportunities for methylation-dependent targets in mucosal diseases such as IBD.

The authors have no financial conflicts of interest.