Carbon monoxide ameliorates chronic murine colitis through a heme oxygenase 1–dependent pathway

The immunomodulatory effects of CO were studied in a murine model of Th1-mediated chronic intestinal inflammation, IL-10^−/− mice. Mice were exposed to 250 ppm CO from 8 to 12 wk of age (n = 12) and compared with a control group exposed to ambient air (n = 8). Because hypoxia can become a confounding factor at carboxyhemoglobin levels of ≥20% (13, 21), carboxyhemoglobin levels of CO-exposed mice were determined and were uniformly <14% (as well as in all subsequent in vivo experiments). Assessment of histologic improvement in IL-10^−/− mice can be problematic because of incomplete penetrance and the segmental, patchy pattern of colitis. Therefore, histological results are depicted in two different ways. First, 20 random fields distributed over the entire length of colon for each mouse were graded by a pathologist blinded to the treatment group using a standard scoring system (22). Results are presented as the sum total of four averaged scores from five regions of the large intestine. CO-exposed mice demonstrate a 60% improvement in histological scores compared with air-exposed mice (Fig. 1 B, left). To represent the spectrum of histological changes in individual mice, results are also presented as the percentage of 20 fields per mouse that demonstrate no histological inflammation (colitis score of 0), mild to moderate inflammatory changes (colitis score of 1 and 2), and severe inflammation (colitis score of 3 and 4). In the CO-exposed group, compared with air-exposed IL-10^−/− mice, significantly more colonic fields demonstrate no evidence of histologic inflammation, and likewise significantly fewer fields demonstrated inflammatory changes (Fig. 1 B, right). Furthermore, CO-exposed IL-10^−/− mice show less weight loss compared with air-exposed controls (1.48 ± 1.27 vs. 0.22 ± 1.35 grams; P < 0.05).

Importantly, these results suggest that CO exposure ameliorates preexisting colitis. To conclusively demonstrate this point, we tested the therapeutic effects of CO administration in a model of severe, chronic intestinal inflammation with 100% penetrance. Severe colitis was induced in 8-wk-old IL-10^−/− mice by feeding 200 ppm piroxicam-supplemented chow for 5 d. In the original description of this model (23), piroxicam administration for 10 d to 4–6-wk-old IL-10^−/− mice resulted in severe colitis in 100% of mice that persisted upon withdrawal of piroxicam from the diet. Due to 100% mortality within 7 d of piroxicam feeding in our facility (not depicted), the protocol was accordingly modified. After removal of piroxicam from the diet, mice were exposed to either CO or ambient air (n = 5 mice each) for 2 wk. CO exposure significantly ameliorated established, severe colitis as determined by improved histologic parameters (Fig. 1 C). In this model, piroxicam represents an initiating factor for chronic inflammation rather than the cause per se. Severe inflammation persists after withdrawal of piroxicam, and although piroxicam may cause acute gastrointestinal injury, 8-wk-old wild-type C57Bl/6 mice treated with piroxicam for 5 d do not demonstrate intestinal inflammation 2 wk after withdrawal (not depicted; reference 23).

To determine whether CO exposure affects mucosal inflammatory cytokine production in IL-10^−/− mice, spontaneous IL-12 p40 and TNF levels were measured in supernatants from colonic mucosal tissue. Intestinal explants of IL-10^−/− mice exposed to CO in vivo for 4 wk express significantly less IL-12 p40 and TNF compared with explants from air-exposed mice (Fig. 1 D). Kinetics of inhibition of intestinal IL-12 p40 by CO was next determined. IL-10^−/− mice were exposed to 250 ppm CO for 1, 4, and 7 d. Intestinal secretion of IL-12 p40 was measured in intestinal explant culture and compared with values obtained from air-exposed IL-10^−/− mice. Intestinal IL-12 p40 is inhibited after 4 and 7 d of CO exposure, but not after 1 d (Fig. 1 E). Importantly, at 4 and 7 d, significant histological improvement was not demonstrated between the groups (not depicted), suggesting that CO inhibits IL-12 p40 expression before resolution of intestinal inflammatory infiltrates. IL-12 p40 heterodimerizes with p35 or p19 to produce biologically active IL-12 p70 or IL-23, respectively. Intestinal secretion of IL-12 p70 and IL-23 was measured in CO-treated and air-exposed IL-10^−/− mice. CO strongly inhibits IL-12 p70 at days 4 and 7, whereas a less significant inhibitory effect on intestinal IL-23 secretion is demonstrated (Fig. 1 E).

As IL-12 p40 protein secretion was reduced in intestinal explants of CO-exposed IL-10^−/− mice, the molecular effects of CO on IL-12 p40 expression were investigated in murine BM-derived macrophages. Because IFN-γ is important for optimal expression of LPS-induced IL-12 p40 secretion, macrophages were stimulated with LPS or LPS/IFN-γ. Exposure of cells to CO at 250 ppm inhibits synergistic activation of IL-12 p40 protein production (Fig. 2 A) and mRNA expression (Fig. 2 B) by LPS/IFN-γ. A dose-dependent inhibition of LPS/IFN-γ–induced IL-12 p40 mRNA and protein at CO concentrations of 0, 50, 100, and 250 ppm was observed (not depicted). Furthermore, CO inhibits LPS/IFN-γ–induced IL-12 p40 protein and mRNA expression across a range of LPS concentrations (10–1,000 ng/ml; not depicted). Next, the effects of CO on biologically active IL-12 p70 and IL-23 were studied. Again, the main effect of CO was inhibition of the synergistic induction of IL-12 p70 by LPS/IFN-γ. IFN-γ does not augment LPS-induced IL-23 secretion, and CO did not inhibit IL-23 activation (Fig. 2 C). Similar results were obtained using murine splenocytes and peritoneal macrophages (not depicted).

IL-12 p40 is transcriptionally regulated through a number of cis-acting elements that are responsive to bacterial products and IFN-γ (15). To study inhibition of LPS/IFN-γ–induced IL-12 p40 transcription by CO, IL-12 p40 promoter activity was determined in the J774 murine macrophage cell line. J774 cells were transiently transfected with a murine IL-12 p40 promoter fused to a luciferase reporter gene and then activated with IFN-γ plus LPS in the presence or absence of CO at 250 ppm for 24 h. CO selectively inhibits IL-12 p40 promoter activity synergistically induced by LPS/IFN-γ (Fig. 2 D).

IFN-γ signaling induces gene transcription through DNA motifs, designated ISRE, found in promoters of IFN-γ–inducible genes that bind the IRF family of transcription factors. Therefore, we determined whether CO modulates LPS/IFN-γ activation of a promoter that contains a consensus ISRE. J774 cells were transiently transfected with a plasmid containing a multimerized consensus ISRE fused to a luciferase reporter gene. Cells exposed to 250 ppm CO and activated with LPS/IFN-γ showed significantly augmented ISRE reporter activity (Fig. 2 E).

To summarize, for IL-12 p40, CO antagonizes IFN-γ responses, whereas for an ISRE, CO augments IFN-γ–induced activity. Accordingly, we hypothesized that CO affects gene transcription through IRF family members. A logical target is IRF-8. IRF-8 is an important transcription factor for IL-12 p40 regulation and promoter activation (16). Moreover, IRF-8 is a transcriptional repressor through an ISRE. J774 cells were stimulated with LPS/IFN-γ and cultured either in ambient air or 250 ppm CO for 4, 6, and 12 h. As determined by Western immunoblot in nuclear extracts, CO inhibits LPS/IFN-γ–activated IRF-8 protein expression compared with cells exposed to air (Fig. 3 A). Similar results are observed in murine splenocytes (Fig. 3 B). To determine whether CO inhibition of IRF-8 was specific, other transcription factors activated by IFN-γ were evaluated. CO exposure did not inhibit IRF-1, IRF-4, and phosphorylated STAT3 in LPS/IFN-γ–activated J774 cells (Fig. 3 A) or murine splenocytes (Fig. 3 B). Similarly, activated STAT1 (phosphorylated STAT1) and IRF-2 expression, induced by LPS/IFN-γ in murine macrophages, were not affected by CO (not depicted).

To provide evidence that IL-12 p40 inhibition by CO is directly mediated through IRF-8, IRF-8 was expressed in BM-derived macrophages by retroviral transduction using a derivative of the Moloney murine leukemia virus vector, pMMP412. 48 h after retroviral transduction, cells were exposed to 250 ppm CO or ambient air and activated with LPS with or without IFN-γ for 24 h. As expected, IRF-8–transduced macrophages produce higher levels of LPS-stimulated IL-12 p40 protein (Fig. 3 C). Importantly, in IRF-8–transduced BM-derived macrophages, the inhibitory effects of CO on LPS/IFN-γ–stimulated IL-12 p40 is abrogated, compared with control virus–transduced cells (Fig. 3 C). Collectively, these data demonstrate that CO inhibits LPS/IFN-γ–induced IL-12 p40 expression through IRF-8.

The effects of CO on inflammatory cytokine signaling may encompass multiple pathways (13, 24). IL-10 has numerous antiinflammatory activities, perhaps the most important of which is its ability to strongly inhibit IL-12 p40 (25). Therefore, increases in LPS-induced IL-10 expression in CO-exposed macrophages may explain inhibition of IL-12 p40 production. To test this possibility, IL-10^−/− BM-derived macrophages were activated with LPS/IFN-γ in the presence or absence of CO. CO exposure inhibits IL-12 p40 protein production (Fig. 4 A) and mRNA expression (Fig. 4 B) induced by LPS and LPS/IFN-γ. Similar results were found in IL-10^−/− splenocytes, and CO also inhibits bioactive IL-12 p70 secretion (not depicted). CO also inhibits IRF-8 protein expression in IL-10^−/− macrophages (Fig. 4 C).

In a sepsis model, induction of IL-10 by CO was mediated by the MAPK signaling pathway MKK3/p38. p38 activation has been correlated with increased IL-12 p40 expression (13). Therefore, we investigated whether the inhibitory effects of CO on IL-12 is mediated through the MKK3/p38 axis. In BM-derived macrophages from MKK3^−/− mice, CO inhibits LPS/IFN-γ–activated IL-12 p40 and p70 secretion (not depicted). Therefore, CO inhibition of LPS/IFN-γ–stimulated IL-12 p40 expression is independent of IL-10 and the p38 MAPK pathway.

We have previously shown that NO inhibits IL-12 p40 transcription (26). Furthermore, CO induces NO in hepatocytes (27). To test whether NO mediates inhibition of IL-12 p40 by CO, BM-derived macrophages from inos^−/− mice were incubated in 250 ppm CO and stimulated with LPS or LPS/IFN-γ. CO inhibits LPS/IFN-γ–stimulated IL-12 p40 and p70 secretion in inos^−/− macrophages (not depicted), suggesting that the inhibition of IL-12 expression by CO is inducible NO synthase (iNOS) independent.

In hepatocytes, CO up-regulates HO-1 activity, and inhibition of HO-1 abrogates the protective effects of CO in a model of acute liver failure (27). To determine whether HO-1 is required for downstream effects of CO on IFN-γ signaling, BM-derived macrophages were obtained from hmox^−/− mice. Interestingly, in hmox^−/− macrophages, inhibition of LPS/IFN-γ–induced IL-12 p40 production by exogenous CO exposure is completely abrogated (Fig. 5 A), fostering the hypothesis that the immunomodulatory effects of CO are mediated by HO-1 or one of its other metabolic products. In support of this hypothesis, in BM-derived macrophages, the selective HO-1 inhibitor tin protoporphyrin (SnPP) abrogates the inhibitory effect of exogenous CO on LPS/IFN-γ–stimulated IL-12 p40 secretion in a concentration-dependent manner (Fig. 5 B).

Expression of iNOS is up-regulated in macrophages by LPS/IFN-γ (28). Importantly, similar to IL-12 p40, the murine iNOS promoter is regulated by IRF-8 (28). Therefore, we studied whether CO would alter IFN-γ–induced accumulation of the end product of iNOS activity, NO. In murine BM-derived macrophages, CO inhibits LPS/IFN-γ–stimulated NO production (measured by nitrite production; not depicted). CO also inhibits LPS/IFN-γ–stimulated iNOS mRNA expression (not depicted). These results lend further credence to the observation that CO alters IFN-γ signaling through specific inhibition of IRF-8. Furthermore, pharmacologic inhibition of HO-1 by SnPP abrogates the inhibitory effect of CO on LPS/IFN-γ–stimulated NO production (not depicted), suggesting that, similar to IL-12 p40, CO inhibits iNOS activity and/or expression through an HO-1–dependent pathway.

To demonstrate that IRF-8 is downstream of HO-1 in this CO inhibitory pathway, IRF-8 expression was evaluated in the presence of an HO-1 inhibitor and in hmox^−/− macrophages. Pharmacologic inhibition of HO-1 with SnPP (Fig. 5 C) abrogates the inhibitory effect of CO on LPS/IFN-γ–stimulated IRF-8 protein expression in BM-derived macrophages. Likewise, in BM-derived hmox^−/− macrophages, CO does not inhibit LPS/IFN-γ–stimulated IRF-8 expression (Fig. 5 D). These results further substantiate that CO alters IFN-γ signaling and abrogates IL-12 p40 expression through IRF-8, and that these effects require HO-1.

To begin to determine mechanistically how HO-1 mediates the described actions of CO, we tested whether CO alters HO-1 expression. HO-1 protein expression was determined in cellular extracts from murine BM-derived macrophages stimulated with LPS, IFN-γ, or both in the presence or absence of CO. IFN-γ inhibits basal and LPS-induced HO-1 protein expression in murine macrophages, and CO antagonizes the inhibitory effects of IFN-γ on HO-1 expression (Fig. 6 A). Likewise, HO-1 mRNA expression is inhibited by IFN-γ, and the inhibitory effect is abrogated by CO (not depicted). To determine whether CO alters HO-1 expression in vivo, intestinal sections from CO-exposed IL-10^−/− mice and air-exposed IL-10^−/− mice, matched for histologic colitis severity, were evaluated for HO-1 expression. Increased immunoreactivity for HO-1 in the intestinal epithelia and increased numbers of lamina propria cells with HO-1 immunoreactivity are observed in CO-exposed mice (Fig. 6 B). Thus, CO abrogates the inhibitory effect of IFN-γ on HO-1 expression in macrophages and induces intestinal HO-1 in vivo in IL-10^−/− mice.

Therefore, genetic and pharmacologic experiments demonstrate that CO acts through an HO-1–dependent pathway, and we hypothesized that another metabolic product of HO-1 may mediate the immunomodulatory effects of CO. Biliverdin has been recently described to alter T cell signaling and decrease IFN-γ production (29). BM-derived murine macrophages were treated with biliverdin at concentrations of 10 to 100 μM for 2 h before activation with LPS or LPS/IFN-γ. Biliverdin does not inhibit IL-12 p40 mRNA expression or protein production (not depicted). Therefore, the effects of CO and HO-1 on IL-12 expression are not mediated by biliverdin in this experimental system.

BM-derived macrophages were treated with the HO-1 inducer hemin and then stimulated with LPS or LPS/IFN-γ. Pretreatment with hemin inhibits LPS/IFN-γ–stimulated IL-12 p40 secretion in a concentration-dependent manner (Fig. 7 A). Furthermore, hemin dose dependently induces HO-1 and inhibits IRF-8 protein expression (Fig. 7 B). To study HO-1 induction in vivo, hemin was administered by i.p. injection (25 mg/kg) twice weekly for 2 wk to 10-wk-old IL-10^−/− mice. In vivo administration of hemin increased intestinal HO-1 protein expression (Fig. 8 A). Histologic evaluation of colitis was compared with a control group treated i.p. with vehicle alone (n = 7 per group). Hemin-treated IL-10^−/− mice demonstrate improved histology (Fig. 8 B) and less weight loss (not depicted) than vehicle-treated controls. Similar histological improvement was observed in IL-10^−/− mice treated with another HO-1 inducer, cobalt protoporphyrin (CoPP; 5 mg/kg i.p. twice weekly for 2 wk; n = 7 per group; Fig. 8 C). Furthermore, decreased IL-12 p40 expression is demonstrated in intestinal explant cultures from hemin- and CoPP-treated IL-10^−/− mice compared with vehicle treated mice (Fig. 8 D). These results suggest a novel pathway where HO-1 mediates the antiinflammatory properties of CO. Collectively, cell culture and in vivo data demonstrate a potent immunomodulatory role for CO and HO-1 in chronic mucosal inflammation.

In summary, CO exposure at low concentrations ameliorates Th1-mediated chronic murine colitis. Mechanistically, the effects of CO upon macrophage activation were studied, focusing on the regulation of the IL-12 p40 gene, important in the pathogenesis of Th1-mediated colitis. In murine macrophages, CO inhibits the synergistic activation of IL-12 p40 and iNOS by LPS/IFN-γ through selective inhibition of IRF-8. Furthermore, the immunologic effects of CO require HO-1. CO induces HO-1 expression, again, through antagonism of IFN-γ signaling. In the absence of functional HO-1, the inhibitory action of CO on IRF-8 and IL-12 p40 expression is lost. As proof of this concept, pharmacologic HO-1 induction with hemin and CoPP recapitulates the effects of CO in cells and in vivo. In macrophages, HO-1 inducers inhibit LPS/IFN-γ–induced IRF-8 and IL-12 p40 expression, and in IL-10^−/− mice, HO-1 inducers ameliorate colitis.

Despite the classification of CO as a toxic entity, lethal to aerobic life and one of the primary pollutants in industrial society, CO has recently emerged as a potent immunomodulatory entity. In the gastrointestinal tract, CO has been studied as a gaseous enteric neurotransmitter and as such is critically involved in smooth muscle membrane potential and ileal contractility (30). Recently, in a murine model of acute mucosal injury (dextran sodium sulfate–induced colitis), the HO-1 inducer CoPP modestly ameliorated inflammatory changes, but CO had no effect (29). In addition, pharmacologic inhibition of HO-1 exacerbated TNBS-induced acute colitis in rats (31). This study is the first to suggest an antiinflammatory role for CO in chronic intestinal inflammation and the first to investigate the HO-1 pathway in established chronic intestinal inflammation.

We directed our mechanistic analyses toward the observation that CO ameliorated Th1-mediated chronic intestinal inflammation in IL-10^−/− mice. In murine macrophages, CO affects IFN-γ signaling and IL-12 p40 expression. The IRF family member IRF-8 is specifically expressed in hematopoietic cells and is induced by IFN-γ (as well as LPS) at the protein and mRNA level in macrophages (32). IRF-8^−/− macrophages display selective impairment of IL-12 p40 expression. We have recently demonstrated a functional composite DNA binding element in the murine IL-12 p40 promoter that binds IRF family members, most notably IRF-8 (16). In different biologic situations, IRFs may serve as transcriptional activators or repressors (33). IRF-8 and IRF-2 have been demonstrated to inhibit IRF-1–mediated activation of promoters with a consensus ISRE, but IRF-8 is clearly involved in activation of the IL-12 p40 promoter. Therefore, our results suggest that CO exerts immunomodulatory effects through inhibition of IRF-8, thereby altering the ratio of IRF family members induced by IFN-γ in macrophages. Our finding that CO inhibits the synergistic activation of NO production by IFN-γ and LPS supports this hypothesis. The iNOS promoter contains a novel ISRE where IRF-8 plays an activating role (28).

In hmox^−/− macrophages, CO does not inhibit IL-12 p40 and IRF-8 expression. In addition, pharmacologic induction of HO-1 inhibits LPS/IFN-γ–induced IL-12 p40 and IRF-8 in wild-type macrophages. Therefore, CO may act primarily through induction of HO-1. Our findings using pharmacologic inducers and inhibitors of HO-1 parallel observations in a model of liver failure where inhibition of HO-1 activity in hepatocytes resulted in a loss of the protective effects conferred by CO (27). The precise mechanism(s) by which HO-1 confers protection against inflammation remains poorly understood, but more recently its products have emerged as the potential mechanisms of action. CO and the bile pigments biliverdin and bilirubin when administered exogenously may exert potent cytoprotective effects. Moreover, they exert their effects via distinctly different modes of action. For example, CO has effects on vasomotor tone, increasing organ perfusion after transplantation, whereas biliverdin does not (34). Biliverdin prevented the up-regulation of adhesion molecules and infiltrating leukocytes, whereas CO did not (35). In two recent studies of acute intestinal inflammation induced by dextran sodium sulfate in mice, HO-1 induction reduced mucosal injury (29, 36). In contrast to our findings, biliverdin recapitulated the protective effect of HO-1 induction in vivo, whereas CO did not (29). In our study, although CO and pharmacologic HO-1 inducers alter IFN-γ signaling in murine macrophages, biliverdin does not (not depicted). These results suggest that the products of the HO-1 pathway impart antiinflammatory effects via different modes of action. Furthermore, HO-1 may subserve different functions in acute versus chronic models of inflammation. Finally, we hypothesize that HO-1 may have a cytoprotective role independent of its two major metabolic byproducts, CO and biliverdin. Other possibilities include that ferritin expression resulting from iron produced through HO-1 mediates protective effects (37). Alternatively, perhaps endogenously produced CO directs additional protective mechanisms over that administered exogenously.

Collectively, these results show that CO alters IFN-γ signaling, suggesting that CO and HO-1 deviate an immune response from a Th1 profile toward an antiinflammatory response (Fig. 9). In Th1-mediated chronic mucosal inflammation, IFN-γ is a central mediator through induction of IRF-8 and IL-12 p40 as well as through direct inhibition of HO-1. Exogenous CO or HO-1 induction, likely through inhibition of IRF-8, reverses the mucosal inflammatory process. This concept is supported by a recent study (38) in which hmox^−/− splenocytes were found to express increased amounts of Th1 cytokines. This Th1 phenotype shift in hmox^−/− mice may be attributable to lack of appropriate control of IFN-γ signaling through the HO-1 pathway. Another prediction of this model that requires testing is that IFN-γ blockade in IL-10^−/− mice should be associated with induction of HO-1. Not surprisingly, antibodies against IFN-γ ameliorate Th1-mediated murine colitis (39) and may be an effective therapy in human CD (40).

This study may have important implications in human IBD. First, therapeutic strategies designed to induce HO-1 may eliminate toxicity concerns associated with direct administration of CO. Future experiments will further characterize metabolic products and enzymes in the HO-1 pathway as well as HO-1–inducing agents for immunologic actions in IBD models. Furthermore, we had initially pursued this line of investigation to study mechanisms of an important epidemiological observation in human IBD: Cigarette smoking is protective in UC (41). The “active” immunomodulatory component(s) in cigarette smoke that may explain this phenomenon remains ill defined. Cigarette smoke is a complex mix of more than 500 characterized compounds, each of which may exert independent immunologic effects. Despite this complexity, immunologic and clinical studies to address this issue in IBD have virtually all focused on nicotine. Therapeutic trial experience in UC patients treated with nicotine gum and transdermal nicotine has been inconclusive (42, 43). Interestingly, two recent reports demonstrated that cigarette smoke suppresses Th1 cytokines and enhances the production of IL-4 by human dendritic cell–primed T cells (44, 45). Based on our current findings, the immunomodulatory effects of cigarette smoke may in part be explained by CO (46). Blood carboxyhemoglobin level, a measure of systemic exposure to CO, has been reported to range from 1 to 18% in active smokers (47). The association between cigarette smoking, carboxyhemoglobin levels, and disease activity in patients with IBD warrants further investigation.

In conclusion, we demonstrate that CO administration ameliorates established chronic mucosal inflammation in an immunologic model of IBD. In Th1-mediated inflammatory responses, CO alters IFN-γ signaling in macrophages, resulting in decreased IRF-8 and IL-12 p40 expression. Immunologic and therapeutic effects of CO are dependent on HO-1. We have focused this analysis on how CO may alter macrophage activation, but CO may affect chronic inflammation at many levels, including inhibiting T cell proliferation (48), decreasing epithelial cell apoptosis (36), altering cell adhesion and trafficking (49), and increasing intestinal blood flow (50). Elucidation of these phenomena is likely to clarify the pathogenesis of human IBD and may form the basis for new therapeutic approaches.

Male C57/Bl6, IL-10^−/− and inos^−/− mice were obtained from The Jackson Laboratory, and mkk3^−/− and hmox^−/− mice were generated by R. Flavell (Yale University School of Medicine, New Haven, CT) and S. Yet (Harvard Medical School, Boston, MA), respectively. All wild-type and genetically deficient mice used in this study were on the C57/Bl6 background and matched for age and sex in all experiments. IL-10^−/− mice were fed a chow mixed with 200 ppm piroxicam (Sigma-Aldrich) for 5 d as indicated in Results. Pelleted diets were prepared by Dyets, Inc. All animals were housed in accordance with guidelines from the American Association for Laboratory Animal Care and Research Protocols and were approved by the Institutional Animal Care and Use Committee of the University of Pittsburgh. At the end of the study period, animals were killed using excess CO₂ inhalation. Immediately afterward, blood by cardiac puncture (carboxyhemoglobin determination), spleens, intestinal tissue samples, and femurs were collected. BM-derived macrophages and splenocytes were cultured as described previously (26).

Mice or macrophages were exposed to compressed air or CO at a concentration of 250 ppm as described previously (13). In brief, CO at a concentration of 1% (10,000 ppm) in compressed air was mixed with compressed air before being delivered into the exposure chamber. Flow into animal chamber was maintained at rate of 12 liters/min and into the cell culture chamber at a rate of 2 liters/min. The cell culture chamber was humidified and maintained at 37°C. A CO analyzer (Interscan) was used to measure CO levels continuously in the chambers. Cardiac blood samples (0.2 ml) were taken immediately after the mice were killed to measure carboxyhemoglobin using a hemoximeter (OSM3; Radiometer Copenhagen).

SnPP, hemin, and CoPP (Frontier Scientific Porphyrin Products) were dissolved in sodium hydroxide, and the final pH of 7.4 was achieved by adding hydrochloric acid and further dilution with PBS. Because of light sensitivity, SnPP, CoPP, and hemin were prepared in dim light, light protected, and freshly made before injection.

Murine IL-12 p40, IL-12 p70, IL-10, and TNF immunoassay kits (R&D Systems) and IL-23 (eBioscience) were used according to the manufacturers' instructions.

Western blot analyses were performed on nuclear or whole cell extracts as noted in Results (Figs. 3–8) and described previously (26). Anti–IRF-1, anti–IRF-2, anti–IRF-4, anti–IRF-8, anti–STAT-1, or anti–STAT-3 antibodies were from Santa Cruz Biotechnology, Inc., and anti–HO-1 was from StressGen Biotechnologies.

For retroviral transduction, we used a derivative of the Moloney murine leukemia virus vector pMMP412 (provided by H. Xiong, Mount Sinai School of Medicine, New York, NY; reference 28). To prepare pseudotyped virus, human 293 EbnaT cells were seeded at a density of 4 × 106 cells in a 10-cm dish. The next day, cells were transfected using Superfect (QIAGEN) with a mixture of 2.5 μg of plasmid pMD.G encoding vesicular stomatitis virus G protein, 7.5 μg of plasmid pMD.OGP encoding gag-pol, and 10 μg of the IRF-8 retroviral expression construct. 48 h after transfection, the viral supernatant was collected, centrifuged at 1,500 rpm, and used to infect adherent BM-derived macrophages in the presence of 4 μg/ml polybrene for 6 h. 48 h after infection, macrophages were seeded in 24-well plates at 10⁶ cells/ml, stimulated with LPS or LPS and IFN-γ, and incubated for 24 h in the absence or presence of 250 ppm CO.

Nitrite concentration, a measure of NO synthesis, was assayed by a standard Griess reaction adapted to a microplate system as described previously (51). NaNO₂ was used as a standard, and the data are presented as a concentration of nitrite.

Total RNA was isolated using TRIZOL reagent (Life Technologies). cDNA was derived from 1 μg total RNA by RT using an RT system (Promega). Real-time RT-PCR was performed on RT products with specific primers.

Fluorophore-labeled LUX primers (52) and their unlabeled counterparts were purchased from Invitrogen. LUX primers were designed using a software program (LUX Designer). LUX primers were designed to produce amplicons ranging in size between 69 and 145 basepairs. Primer pairs for IL-12 p40 and GAPDH were selected to span an intron. The primer sequences for IL-12 p40 are: LUX: 5′-GAACTTGTCAAAGGCTTCATCTGCAAG5TC and its unlabeled counterpart: 5′-GGAAGCACGGCAGCAGAATA; for GAPDH: LUX: 5′-GACATACAGGCCGGTGCTGAGTATG5C and its unlabeled counterpart: 5′-GCGGAGATGATGACCCGTTT. Plots of fluorescence versus PCR cycle were generated by the ABI PRISM 7700 SDS software. Each 25-μl PCR reaction mix contained 2 μl cDNA, 1 μl of each gene-specific primer, and 12.5 μl 2X Platinum Quantitative PCR SuperMix-UDG (Invitrogen), 1 μl 1X ROX reference dye, and 7.5 μl DEPC-treated water. Reactions were incubated at 50°C for 2 min, 95°C for 2 min, and then cycled (45 times) at 95°C for 15 s and 60°C for 30 s. Quantitation of IL-12 p40 mRNA was expressed as relative fold-increase in transcript level with respect to unstimulated cells. Real-time RT-PCR for iNOS and HO-1 were performed as described previously (53).

J774 murine macrophages were transiently transfected using SuperFect Transfection Reagent (QIAGEN) by the described protocol with modifications. For each transfection, 2.5 μg of plasmid were mixed with 100 μl DMEM (without serum and antibiotics) and 10 μl of SuperFect reagent. The mixture was incubated at room temperature for 10 min, and then 600 μl DMEM complete medium was added and immediately placed onto the cells in six-well plates. After incubation for 3–4 h at 37°C, the cells were washed with PBS and split equally into two plates. 12–16 h later, one six-well plate was treated with CO. Cells were either unactivated or activated with 1 μg/ml LPS, 10 ng/ml IFN-γ, or both. 16–18 h after activation, the cells were harvested by using 1X Reporter Lysis Buffer (Promega). Luciferase activity was determined from 20 μl of cell extract as described previously (15). Cells were cotransfected with a constitutively active HSP promoter that expresses β-galactosidase to monitor transfection efficiency.

Sections of the transverse colon from individual mice were washed with PBS to remove fecal contents, shaken at 280 rpm at room temperature for 30 min in RPMI 1640 plus 50 μg/ml gentamicin, and cut into small fragments. Tissue fragments (0.5 g dry weight) were incubated in 1.0 ml RPMI 1640 supplemented with 50 μg/ml gentamicin, 100 U/ml penicillin, 100 μg/ml streptomycin, 0.25 μg/ml fungizone (GIBCO BRL), and 5% heat-inactivated fetal calf serum. Tissue fragment supernatants were collected after 24 h for cytokine ELISAs.

Colons were removed after mice were killed and rinsed with cold PBS to remove fecal material. Tissue sections were fixed in 10% buffered formalin and embedded in paraffin. 5-μm-thick sections were stained with hematoxylin and eosin. Colitis scores (0–4) were determined by a staff pathologist who was blinded to the experimental protocol using the criteria reported by Berg et al. (22). 20 separate microscopic fields (magnification of 100) were evaluated for each mouse by a pathologist (A.R. Sepulveda) blinded to the treatment groups.

Statistical analysis was performed using Student's t test. A probability level of P < 0.05 was considered statistically significant.

The authors would like to thank Emeka Ifedigbo, Rasha Salem, Aaron May, and Fang Liu for technical assistance.

This study is supported by a Crohn's and Colitis Foundation of America (CCFA) Senior Research Award, National Institutes of Health grants R01 DK054452 to S.E. Plevy and RO1 HL071797 to L.E. Otterbein, and a CCFA Research Fellowship Award to R.A.F. Hegazi.

L.E. Otterbein is a paid consultant of Linde Gas Therapeutics. None of the other authors have any potential conflicts of interest.