Clostridium difficile Toxin A Regulates Inducible Cyclooxygenase-2 and Prostaglandin E2 Synthesis in Colonocytes via Reactive Oxygen Species and Activation of p38 MAPK*

Clostridium difficile toxin A induces acute colitis with neutrophil infiltration and up-regulation of numerous pro-inflammatory mediators, but the contribution of cyclooxygenase-2 (COX-2) induction in this infection is unknown. We report here that toxin A induces expression of COX-2 and secretion of prostaglandin E2 (PGE2) in a dose- and time-dependent manner in cultured NCM460 human colonocytes and in human intestinal xenografts. This induction was blocked by SB203580, a p38 MAPK inhibitor, which also decreased the phosphorylation of MSK-1, CREB/ATF-1, and COX-2 promoter activity following toxin A stimulation. Gel shift assays indicated that CREB/ATF-1 was the major proteins binding to the COX-2-CRE. Moreover, colonocytes exposed to toxin A produced reactive oxygen species (ROS), which activated p38 MAPK, MSK-1, and CREB/ATF-1, leading to subsequent COX-2 induction and PGE2 secretion. In intact mice, blockage of p38 MAPK inhibited toxin A-mediated induction of COX-2 in enterocytes as well as lamina propria cells, and significantly blocked the toxin A-induced ileal secretion of fluid and PGE2. Furthermore, a selective COX-2 inhibitor also diminished toxin A-associated ileal fluid and PGE2 secretion. The main signaling pathway for toxin A induction of human COX-2 involves ROS-mediated activation of p38 MAPK, MSK-1, CREB, and ATF-1. Toxin A triggers ileal inflammation and secretion of fluid via COX-2 induction and release of PGE2.

tory cytokines that mediate a profound and rapid inflammatory response (5)(6)(7). The induction of fluid secretion and inflammation by toxin A involves extensive signaling cross-talk between epithelial cells, mast cells, sensory neurons, and inflammatory cells of the intestinal lamina propria (4). The cellular mechanism of toxin A involves glucosylation of a threonine residue at position 37 on Rho, Rac, and cdc42 (8), small GTP-binding proteins that regulate cell shape through modulation of the actin cytoskeleton. Monoglucosylation and inactivation of Rho proteins by the toxin causes severe cytoskeletal abnormalities in cultured and intact human colonocytes (9,10). However, the signal transduction pathways by which toxin A induces intestinal inflammation are not entirely known.
Toxin A binds to a G protein-coupled receptor (11) on the luminal aspect of the apical intestinal epithelial cell membrane (12) and is then internalized where it activates MAPKs 1 (13), intracellular calcium release (14,15), release of reactive oxygen species (ROS) (12,16), and secretion of pro-inflammatory mediators (16,17). We previously reported that toxin A releases prostaglandin E 2 (PGE 2 ) into the ileal lumen of intact rats (5) and Alcantara et al. (18) reported that toxin A-induced water and electrolyte secretion in vivo was significantly blocked by a COX-2 inhibitor. COX-2 is induced by pro-inflammatory cytokines, lipopolysaccharide, growth factors, and infectious agents in a variety of cell types (19 -23). PGE 2 is a potent stimulator of intestinal chloride and water secretion in mammalian gut (24), and PGE 2 is released during various forms of intestinal inflammation and infection. For example, infection of cultured enterocytes with Salmonella organisms induces COX-2 expression, followed by PGE 2 -induced apical chloride secretion (25).
In view of the potential importance of COX-2 in toxin A-induced enteritis, we studied the signaling pathway of toxin A induction of COX-2 in colonocytes, the natural target of toxin A, using the non-transformed human colonic epithelial line, NCM460. Because toxin B also elicits inflammatory and cytotoxic responses in human colon in vitro (10) and in vivo (26), we also examined whether it is also able to stimulate COX-2 expression in these cells. We found that toxin A, but not toxin B, induced COX-2 expression at both mRNA and protein levels through ROS-mediated activation of p38 MAPK, mitogen-and stress-activated protein kinase-1 (MSK-1), ATF-1, and CREB. We also demonstrate that a p38 inhibitor and a selective COX-2 inhibitor reduced toxin A-induced fluid and PGE 2 secretion in vivo.

C. difficile Toxin A-Toxin
A and toxin B were purified from culture supernatants of C. difficile strain VPI 10463 (American Type Culture Collection, Rockville, MD) using anion exchange chromatography and fast protein liquid chromatography as previously described (13,26).
Immunoblot Analysis-Cells were harvested and washed once with ice-cold 1ϫ PBS (pH 7.5), and then lysed for 60 min on ice in lysis buffer (150 mM NaCl, 50 mM Tris-Cl (pH 8.0), 5 mM EDTA, 1% Nonidet P-40) with a protease inhibitor mixture (Roche Applied Science) and a phosphatase inhibitor mixture (Sigma). Protein concentrations of the lysates were measured by the Bradford method (Bio-Rad, Hercules, CA), and equal amounts of total protein were fractionated on 10% SDS-polyacrylamide gels. Membranes were first incubated overnight at 4°C with the primary antibodies (1:1000 in 5% bovine serum albumin/in Tris-buffered saline with Tween 20 (150 mM NaCl, 10 nM Tris at pH 8.0, 0.005% Tween 20 [TBST]), then incubated with the corresponding horseradish peroxidase-conjugated secondary antibodies (in 5% nonfat dry milk/ TBST) at room temperature for 1 h and then washed three times with TBST. Antigen-antibody complexes were detected with LumiGlo reagent (New England Biolabs).
Transfection and Luciferase Reporter Assays-NCM460 cells were plated in 6-well plates (0.4 ϫ 10 6 cells/well) and transfected with 1 g of the human COX-2 promoter (27). Cells were cotransfected with 0.1 g of the pRL-CMV plasmid containing the Renilla gene (Promega, Madison WI). One day after transfection, cells were incubated with toxin A (3 nM) and toxin B (20 nM) for 12 h, and firefly luciferase values were standardized to the Renilla values. The total amount of plasmid DNA was kept constant by adding the empty vector for each transfection. All assays were performed in triplicate, and data are expressed as mean values Ϯ S.E.
Measurement of ROS Generation-DCFH-DA is a non-polar compound, which enters the cell and is cleaved to form DCFH. Trapped DCFH is oxidized by oxygen free radicals to produce fluorescent DCF. NCM460 cells were preincubated on a 96-well microplate (2 ϫ 10 3 cells per well) in M3D medium lacking serum for 1 h at 37°C in the presence of 10 M DCFH-DA. Cells were washed three times in pre-warmed PBS and then exposed to toxin A (3 nM). Fluorescence intensity was analyzed by fluorospectroscan (Fluoroscan Ascent FL, Labsystems) using 485 nm excitation and 538 nm emission filters. DCF fluorescence was also measured by fluorescence microscopy. To avoid photo-oxidation of DCFH, fluorescent images were collected by a rapid scan (total scan time, 3 s). Five groups of 10 -20 subconfluent cells were randomly selected for analysis from each sample.
Electrophoretic Mobility Shift Assays-The following sequences derived from the human COX-2 promoter containing the CRE was used: CRE-sense, 5Ј-AAACAGTCATTTCGTCACATGGGCTTG-3Ј; CRE-antisense, 5Ј-CAAGCCCATGTGACGAAATGACTGTTT-3Ј. Probes were annealed and 5Ј-overhangs were labeled by incorporation of [ 32 P]dATP (PerkinElmer Life Sciences) with T4 polynucleotide kinase. For gel shift assays, 2 g of nuclear protein extracts were incubated at room temperature for 5 min with a mixture containing 6 mM HEPES (pH 7.9), 0.4 mM EDTA, 125 mM KCl, 10% glycerol, 0.05 mg/ml poly(dIdC), 1 mM dithiothreitol, 2.5 mM sodium pyrophosphate, 1 mM sodium orthovanadate, 10 mM NaF, 50 mg/ml aprotinin, 50 g/ml leupeptin. Approximately 1 ng of labeled probe was added, and the reactions were incubated at room temperature for another 20 min. In antibody supershift experiments, mixtures were preincubated with various amounts of antibodies on ice for 1 h before the addition of probe. DNA⅐protein complexes were separated on 5% polyacrylamide gel in low ionic strength Tris borate buffer, dried under vacuum, and visualized with a Phosphor-Imager (Amersham Biosciences).
RNA Isolation and Semi-quantitative RT-PCR-Total cellular RNA was isolated and RT-PCR was conducted according to Nakayama et al. (29). Two micrograms of total RNA was reverse transcribed at 42°C for 1 h in 20 l of the reaction mixture containing mouse Moloney leukemia virus reverse transcriptase with oligo(dT) primers. Thereafter, the resultant cDNA was amplified together with Taq polymerase (PerkinElmer Life Sciences) using the specific sets of primers. The primers used were: human COX-2 (300 bp), 5Ј-ATGAGATTGTGG-GAAAATTGCT-3Ј (sense), 5Ј-GATCATCTCTGCCTGAGTATC-3Ј (antisense), and mouse COX-2 (333 bp), 5Ј-GCAAATCCTTGCTGTTC-CAATC-3Ј (sense) 5Ј-GGAGAAGGCTTCCCAGCTTTTG-3Ј (antisense). ␤-Actin amplification was used as the control, respectively. PCR of each molecule was conducted with the optimal numbers of cycles consisting of 94°C for 1 min, optimal annealing temperature for 1 min, and 72°C for 1 min, followed by incubation at 72°C for 3 min. The generated PCR products did not reach a saturable level with the determined optimal cycle numbers. The amplified PCR products were fractionated on a 2% agarose gel and visualized by ethidium bromide staining.
Effects of SB203580 and NS-398 on Toxin A-induced Secretion of PGE 2 and Fluid and Immunohistochemistry for COX-2 in Mouse Intestine in Vivo-This study was approved by the Beth Israel Deaconess Medical Center Animal Care and Use Committee (Boston, MA). CD1 male mice (Charles River Laboratories, Wilmington, MA) weighing 30 -35 g had free access to food and water in a 12-h light/dark cycle. Mice were acclimated to these conditions at least 7 days before the experiment. Mice were anesthetized by intraperitoneal injection of sodium pentobarbital (50 mg/kg) and ileal loops (3-4 cm) were prepared and injected with buffer alone or with the specific p38 inhibitor SB203580 hydrochloride (100 g in water) and the selective COX-2 inhibitor, NS-398 (1 mg/kg), in a volume of 200 l. After 30 min, toxin A (10 g in PBS) or PBS alone was injected intraluminally, and animals were sacrificed 2 h later by CO 2 . Ileal loop fluid was collected and centrifuged at 50,000 ϫ g for 15 min. Secretion of PGE 2 was assayed in the supernatant with specific ELISA kits according to the manufacturer's instructions. Ileal loops were excised and weighed, and length was measured. Fluid secretion was expressed as the loop weight-to-length ratio (mg/cm). Ileal tissue samples were quick frozen for immunohistochemical analysis and for protein determination. Frozen intestinal sections were placed on Fisher Superfrost slides, fixed in 80% acetone at room temperature for 60 s, and air-dried. Before labeling, sections were rehydrated in PBS and incubated with 10% normal goat serum for 20 min. Sections were incubated with a fluorescein isothiocyanate conjugated goat anti-COX-2 antibody (Santa Cruz, CA, sc-1745-fluorescein isothiocyanate) (1:100) for 16 h. The slide was then counter-stained for 30 min in propidium iodide staining solution at a concentration of 5 g/ml in PBS. After final washing, slides were coverslipped with Vectashield (Vector Laboratories) and viewed on a fluorescence microscopy (MRC 1024, Bio-Rad). To confirm expression of COX-2 mRNA, tissue samples (100 mg) were homogenized in 1 ml of TRIzol reagent (Invitrogen) using a power homogenizer, total RNA was isolated and RT-PCR performed using a specific mouse COX-2 primer.
Human Intestinal Xenografts-Human fetal intestine was obtained from Brigham and Woman's Hospital (Boston, MA) (mean age, 14.2 Ϯ 1.6 weeks) after therapeutic abortion as previously described (30). Procurement and procedures involving xenografting of human fetal tissues into C.B.-17 SCID/SCID mice were performed with full approval from the Institutional Review Board. Fetal small intestine and colon were washed in Dulbecco's modified Eagle's medium and xeno-transplanted subcutaneously into SCID mice. Twelve weeks after xenografting, the graft lumen was injected subcutaneously with 400 l of serum-free Dulbecco's modified Eagle's medium containing 40 nM purified toxin A (n ϭ 4) or medium control (n ϭ 2). Mice were killed after 6 h, and the xenografts were excised. We isolated total proteins from these tissue and performed immunoassays with specific COX-2 antibody and loading control, ␤-actin antibody (Sigma).

Induction of COX-2 by C. difficile Toxin A in Colonocytes-
Because COX-2 has been associated with intestinal inflammation, monocyte infiltration, and prostaglandin production, we examined COX-2 expression levels and PGE 2 synthesis in colonocytes exposed to purified toxin A (3 nM). Toxin A induced the expression of COX-2, but had no effect on the expression of COX-1 (Fig. 1A). This response was evident at 1-3 h and peaked at 6 h. Under basal conditions, we observed weak expression of COX-2 mRNA in cultured colonocytes. Toxin A also increased COX-2 mRNA in a time pattern that paralleled the COX-2 protein expression (Fig. 1B). The effect of toxin A on COX-2 protein levels was concentration-dependent between doses of 0.8 and 3 nM (Fig. 1C). Despite the close structural relationship between toxins A and B and their common enzymatic activity against rho proteins, toxin B at 20 nM showed no effect on COX-2 expression (Fig. 1D). Colonocytes were transiently transfected with a construct containing the nucleotide sequence from Ϫ1432 to ϩ59 of the human COX-2 promoter and then treated with 1.5 and 3 nM of toxin A or with 10 and 20 nM of toxin B. Toxin A, but not toxin B, increased COX-2 promoter activity in a dose-dependent manner (Fig. 1E). Next, we examined whether toxin A induces prostaglandin E 2 secretion in colonocytes and determined the COX-2 dependence of this response using the selective COX-2 inhibitor NS-398. The basal concentration of secreted prostaglandin E 2 was ϳ17 pg/ ml/2 ϫ 10 6 cells (Fig. 1F). Toxin A increased PGE 2 secretion by 2-to 3-fold at 6 -12 h (Fig. 1F), which was completely blocked by the selective COX-2 inhibitor NS-398. As expected, toxin B did not induce PGE 2 secretion up to 12 h (Fig. 1F). These results indicate that toxin A-mediated secretion of PGE 2 in colonocytes required COX-2 induction.
Toxin A Induction of COX-2 Expression in Human Intestinal Xenografts-To confirm the induction of COX-2 by toxin A in human intestine, we injected toxin A into the lumen of human intestinal xenografts growing subcutaneously in SCID mice. We previously reported (26) that epithelial cells in the grafted intestine are exclusively of human origin and that injection of toxin A induces an acute inflammatory response, similar to that seen in animal models and in the colon of patients with C. difficile-associated colitis. We found no expression of COX-2 in control xenografts, but robust increases of COX-2 expression 6 h after toxin A exposure (Fig. 2), similar to the results observed in cultured colonocytes (Fig. 1, A and B).

FIG. 1. C. difficile toxin A, but not toxin B, induces COX-2 expression and PGE 2 secretion in human colonocytes. Colonocytes were serum-starved for 24 h and incubated with toxin A (3 nM).
A, cells were lysed and total protein was fractionated on 10% SDS-polyacrylamide gels and probed with antibodies against COX-2, COX-1, or ␤-actin. B, total RNA was isolated, and expression of COX-2 and ␤-actin mRNA was determined by RT-PCR. C, colonocytes were exposed to various concentrations of toxin A for 6 h (h). D, colonocytes were incubated with toxin B (20 nM) for the indicated times and probed with antibodies against COX-2 or ␤-actin. A-D are representative of three independent experiments. E, colonocytes were transfected with a COX-2 promoter construct (Ϫ1432 to ϩ59) for 12 h. Cells were placed in serum-free medium, treated with various concentrations of toxin A or toxin B for 12 h, and cell extracts were prepared to measure COX-2 promoter activity expressed as relative luciferase activity. Bars represent mean Ϯ S.D. of three separate experiments performed in triplicate; *, p Ͻ 0.005 versus non-stimulated cells. F, colonocytes were exposed to toxin A with or without the selective COX-2 inhibitor NS-398, or toxin B for the indicated times, and PGE 2 levels were measured in the conditioned medium by ELISA. Bars represent the mean Ϯ S.D. of three independent experiments, each with quadruple determinations; *, p Ͻ 0.001 versus non-stimulated cells.
Phosphorylation of CREB, ATF-1, ATF-2, MSK-1, and p38 by Toxin A in Colonocytes-We next investigated toxin A-associated phosphorylation of p38 MAPK and its downstream transcription factors ATF-1, ATF-2, and CREB. Toxin A strongly increased the phosphorylation of p38, CREB, ATF-1, and ATF-2 after 1 and 3 h of exposure with return to basal levels at 6 h (Fig. 3A). Mitogen-and stress-activated protein kinase-1 (MSK-1) is directly activated by p38 MAPK and may mediate activation of CREB, as shown in embryonal human kidney cells activated by growth factors and oxidative stress (31). We therefore determined phosphorylation of threonine 581 of MSK-1 following toxin A stimulation. We observed phosphorylation of MSK-1 at 1 and 3 h after toxin A exposure, similar to the time course of CREB/ATF-1 phosphorylation.
Because NF-B, and ERK1/2 are known regulators of COX-2 expression (32, 33), we examined phosphorylation of IB, and ERK1/2 in colonocytes exposed to toxin A. As shown in Fig. 3B, we did not observe phosphorylation of IB and only slight phosphorylation of ERK1/2.
Activation Pathways for COX-2 in Colonocytes Exposed to Toxin A-To define specific induction pathways of COX-2 by toxin A, we treated colonocytes with the p38 MAPK inhibitor SB203580, the JAK1 (Janus tyrosine kinase 1) inhibitor AG490, the NF-B inhibitor 6-amino-4-quinazoline, the ERK1/2 kinase inhibitor PD98059, and toxin A for 6 h. We found that inhibition of p38 by SB203580 completely prevented the expression of COX-2 by toxin A (Fig. 4A). None of the other inhibitors had any significant effect on toxin A-induced COX-2 expression, apart from the JAK1 inhibitor AG490, which slightly inhibited COX-2 protein expression (Fig. 4A). Similarly, only the p38 MAPK inhibitor, SB203580 reduced the secretion of PGE 2 in supernatants (from 49 pg/ml/2 ϫ 10 6 cells to 21 pg/ml/2 ϫ 10 6 cells, p Ͻ 0.001) (Fig. 4B), whereas, none of the other inhibitors exerted any significant effect at the doses used. Blockage of JAK1 by AG490 weekly inhibited toxin A-induced COX-2 protein (Fig. 4A), but failed to inhibit toxin A-mediated PGE 2 secretion (Fig. 4B). In view of the lack of inhibition of COX-2 expression by the ERK1/2 kinase inhibitor PD98059 (Fig. 4A) as well as by the weak phosphorylation of ERK1/2 induced by toxin A (Fig. 3B), we conclude that the ERK1/2 and JAK1 pathways are not significantly involved in the regulation of COX-2 expression by toxin A.
We next examined the involvement of p38 MAPK on toxin A-induction of COX-2 mRNA. The p38 MAPK inhibitor, SB203580, at 10 M completely blocked the toxin A effect on COX-2 mRNA in colonocytes (Fig. 4C). Because weak expression of COX-2 mRNA was detected in basal condition of cultured colonocytes and previous studies report that p38 activation stabilizes COX-2 messenger RNA (34 -36), we investigated the role of toxin A on the stability of COX-2 transcripts after treatment with toxin A and/or actinomycin D for 30, 60, and 90 min. Both the toxin A-induced COX-2 transcripts 6 h following toxin A exposure and the basal expression of COX-2 transcripts disappeared at the same rate in the presence or absence of toxin A exposure (data not shown), indicating that toxin A-mediated p38 MAPK activation does not affect the stability of COX-2 mRNA.
Inhibition of p38 MAPK with SB203580 also blocked the toxin A-induced activation of the human COX-2 promoter in a dose-dependent manner (Fig. 4D), consistent with the results of COX-2 mRNA inhibition by these antagonists (Fig. 4C). COX-2 reporter activity was decreased to basal levels by 10 M SB203580 (Fig. 4D).
p38 MAPK and Phosphorylation of CREB/ATF-1 and MSK-1-Because our results indicated that toxin A activates CREB/ ATF-1 and MSK-1 (Fig. 3A), and p38 is phosphorylated by this toxin (Fig. 3A), we next explored the involvement of the p38 pathway on CREB/ATF-1 and MSK-1 activation in response to toxin A. Incubation of colonocytes with the p38 inhibitor SB203580 (10 M) diminished toxin A-induced phosphorylation of CREB and ATF-1, whereas PD98059 (20 M), a selective inhibitor of the ERK1/2 pathway, had no detectable effect (Fig.  5A). In addition, neither an NF-B inhibitor nor AG490, a JAK inhibitor blocked phosphorylation of CREB and ATF-1. MSK-1, an upstream kinase of ATF-1 and CREB, was also completely blocked by SB203580, whereas toxin A-induced phosphorylation of ATF-2 was not affected by any of the inhibitors used (Fig. 5A). These data suggest that phosphorylation of ATF-2 associated with toxin A is not regulated by p38, but probably by other kinase(s), such as JNK. In addition, these results also reflect that toxin A-induced ATF-2 activation is not involved in COX-2 induction in colonocytes following toxin A exposure. Next, we examined whether a p38 MAPK dominant negative plasmid (p38-DN) prevents induction of COX-2 promoter activity following toxin A exposure. Overexpression of p38 dominant negative at 500 g and 1000 g, but not 100 g, markedly diminished the COX-2 promoter response (Fig. 5B). These re-

FIG. 2. Toxin A induces COX-2 expression in human intestinal xenografts.
Human colonic xenografts transplanted in SCID mice were injected with toxin A (40 nM) or medium (control). After 6 h mice were killed, grafts were excised, and graft tissue was washed and homogenized in buffer. Colonic proteins were fractionated on 10% SDSpolyacrylamide gels and probed with antibodies against COX-2 or ␤-actin, used as internal control. Each lane represents blots from two control (C) and four toxin A-injected (T) xenografts.

FIG. 3. Toxin A induces CREB, ATF-1, ATF-2, MSK-1, and p38 activation.
A and B, colonocytes were incubated with 3 nM toxin A for the indicated times. Cells were lysed in SDS-sample buffer, and cell lysates were resolved on 10% polyacrylamide gels and probed with antibodies against phosphorylated p38, MSK-1, CREB/ATF-1, ATF-2, ERK1/2, and IB or ␤-actin. The figures are representative of three independent experiments. sults indicated that the p38 MAPK pathway is the major regulator of the COX-2 response to toxin A.
CREB and ATF-1 Bind to the COX-2 CRE-We performed electrophoretic mobility shift assays to identify the transcriptional factors that bind to the COX-2 cAMP-responsive element (CRE) following toxin A treatment. Incubation of colonocytes nuclear extracts with the COX-2 CRE probe produced two major bands and the maximum complex formation was detected 1 h after toxin stimulation (Fig. 6A). The upper complex was more strongly competed than the lower complex by excess of cold COX-2 CRE. The upper complex contained a CREbinding protein (CREB), as shown by supershift experiments using an anti-phospho CREB antibody (Fig. 6B). Because this antibody reacts with both phospho-CREB and phospho-ATF-1, the CREB⅐ATF-1⅐DNA complex shown in Fig. 6B may also contain ATF-1. However, we were unable to supershift or clear the complex using antibodies against ATF-2 and c-Jun (Fig.  6B). Taken together, these results suggest that CREB and ATF-1 are important transcription factors in COX-2 gene expression in response to toxin A.
Toxin A Induction of COX-2 Involves ROS-We recently reported that toxin A exposure of Chinese hamster ovary cells generates reactive oxygen species (ROS) reflecting mitochondrial damage (12,16). It is also known that CREB is regulated by cAMP (37,38). Based on these considerations, we examined whether colonocytes exposed to toxin A generated ROS or cAMP. In line with our previous results (12), ROS levels measured either by fluorospectroscan (Fig. 7A, upper panel) or fluorescence microscopy (Fig. 7A, lower panel) were increased by ϳ4to 5-fold following toxin A exposure (Fig. 7A). In contrast, intracellular cAMP levels were not significantly increased by toxin A (Fig. 7A), indicating that toxin A-induced activation of CREB in colonocytes is not cAMP-dependent.
Next, we tested whether NAC, a hydrogen peroxide scavenger (39), or sodium formate, a hydroxyl radical scavenger (40), could attenuate toxin A-mediated induction of COX-2. Serum-starved colonocytes were preincubated with NAC and sodium formate for 30 min and then exposed to 3 nM toxin A for 6 h. We found that 20 mM NAC preincubation markedly reduced toxin A induction of COX-2, whereas sodium formate had no effect (Fig. 7B). Moreover, colonocytes exposed to 100 and 500 M H 2 O 2 showed increased expression of COX-2 (Fig. 7C,  left panel), and activation of p38 in a time-dependent manner (Fig. 7C, right panel). In addition, NAC, but not sodium formate, completely blocked toxin A-induced MSK-1 activation and partially blocked CREB and ATF1 phosphorylation (Fig.  7D, left panel). Phosphorylation of p38 following toxin A exposure was slightly inhibited by NAC (Fig. 7D, left panel). Colonocytes exposed to 100 and 500 M of H 2 O 2 also showed increased activation of p38, MSK-1, CREB, and ATF-1 (Fig. 7D, right  panel). In addition, toxin A-induced PGE 2 secretion after 36 h was substantially inhibited by 10 mM, and 20 mM of NAC, but not sodium formate (Fig. 7F), whereas hydrogen peroxide exposure for 36 h also increased PGE 2 secretion in a dose-dependent manner (Fig. 7E) but to a lesser degree than toxin A (Fig.  7F). Thus, toxin A rapidly stimulates (within 10 min) release of ROS that mediates activation of p38 MAPK, MSK-1, CREB, and ATF-1, leading to COX-2 induction and PGE 2 secretion. FIG. 4. Activation pathways for COX-2 in colonocytes exposed to toxin A. A, serum-starved colonocytes were treated for 6 h with either medium (control) or toxin A alone, or toxin A together with either Me 2 SO (DMSO) or the p38 inhibitor SB203580 (10 M), the JAK inhibitor AG490 (50 M), the NFB inhibitor 6-amino-4-quinazoline (50 nM), or the ERK1/2 kinase inhibitor PD98059 (10 M). Cells were then lysed, and Western blot analysis was performed using an antibody against COX-2. B, colonocytes were treated with either medium (control) or toxin A alone, or toxin A together with the chemical inhibitors indicated above. After 12 h conditioned medium was collected and PGE 2 levels were determined by ELISA. Bars, mean Ϯ S.D. of three independent experiments, each with quadruple determinations; *, p Ͻ 0.0001 versus non-stimulated cells. C, colonocytes were incubated with 3 nM toxin A either alone or with SB203580. After 6 h RNA was isolated and COX-2 and ␤-actin mRNA were measured by RT-PCR. Results are representative of three independent experiments. D, colonocytes were transfected with a COX-2 promoter construct (Ϫ1432 to ϩ59) or an internal control plasmid for 12 h. Cells were placed in serum-free medium, treated with either toxin A or medium alone, or toxin A plus the p38 inhibitor. Cell extracts were prepared to measure COX-2 promoter activity expressed as relative luciferase activity. Bars represent mean Ϯ S.D. of three experiments performed in triplicate. *, p Ͻ 0.001 versus non-stimulated cells.

Effect of p38 MAPK and COX-2 Inhibitors in Toxin A-induced COX-2 Expression and Activity and Fluid and PGE 2 Secretion in Mouse Ileum in Vivo-
Our results indicate that C. difficile toxin A-induced p38 MAPK activation is required for the expression of COX-2 in cultured human colonocytes (Fig. 4). To confirm the pathophysiological relevance of this in vitro model, we investigated whether the p38 MAPK inhibitor SB203580 might prevent toxin A-induced COX-2 expression, the ileal secretion of fluid, and PGE 2 in intact mice. We also examined whether the selective COX-2 inhibitor, NS-398 abolishes toxin A-induced ileal fluid and PGE 2 secretion. Mice were pretreated with SB203580 (100 g) and NS-398 (100 M) for 30 min before intraluminal toxin A administration. As shown in Fig. 8A, mice injected with toxin A alone had increased expression of COX-2 compared with mice injected with buffer alone, and this was strongly inhibited by the p38 inhibitor SB203580 (Fig. 8, A and B). The p38 inhibitor also significantly reduced toxin A-associated fluid secretion by 53% (p Ͻ 0.001; n ϭ 3) and PGE 2 secretion by 87% (p Ͻ 0.001; n ϭ 3) (Fig. 8, C and D). The selective COX-2 inhibitor also inhibited toxin A-induced fluid secretion by 39% (p Ͻ 0.001; n ϭ 3) and PGE2 secretion by 42% (p Ͻ 0.001; n ϭ 3) (Fig. 8, C and D). Severe epithelial cell destruction following toxin A injection was also substantially blocked by SB203580 (Fig. 8E). Immunohistochemistry confirmed that toxin A exposure strongly induced COX-2 in both epithelial and lamina propria cells of mouse ileal mucosa (Fig.  8F), and SB203580 significantly reduced toxin A-induced COX-2 expression. These results indicate that enteritis and diarrhea following toxin A exposure is dependent for full expression on COX-2 induction via a p38 MAPK pathway. DISCUSSION We report here that C. difficile toxin A strongly induces colonocyte COX-2 expression and PGE 2 secretion both in vitro and in vivo in human colon and mouse ileum, and this contributes to toxin A-induced enteritis and fluid secretion. The signal transduction cascade following toxin A-receptor binding involves release of hydrogen peroxide, which mediates activation of p38 MAPK and the downstream effectors, MSK-1, CREB, and ATF-1, leading to subsequent COX-2 induction and PGE 2 secretion. In contrast, toxin B did not have an effect on COX-2 induction and PGE 2 secretion. Different expression of receptors for these two distinct toxins in cultured colonocytes probably accounts for these observed responses.
The CRE element in the COX-2 promoter is necessary for the induction of COX-2 transcription mediated by platelet derived growth factor, and lipopolysaccharide (41)(42)(43). NF-B mediates COX-2 induction by lipopolysaccharide and other stimuli (42), but this transcription factor is not involved in the toxin A-mediated COX-2 transcription. We found that toxin A did not A, serumstarved colonocytes were incubated with 3 nM toxin A for the indicated times, and crude nuclear extracts were prepared. Nuclear extracts (2 g) were incubated with 1 ng of radiolabeled doublestranded COX-2-CRE at room temperature for 30 min. DNA⅐protein complexes were separated on 5% polyacrylamide gel. "Probe" indicates that no nuclear extracts were added. B, nuclear extracts from cells treated with toxin A (TxA) for 1 h were incubated with radiolabeled double-stranded COX-2-CRE for 30 min and rabbit control IgG, CREB, ATF-2, or c-Jun antibodies were added to the reaction mixture before the addition of labeled probe. DNA⅐protein complexes were separated on 5% polyacrylamide gel. Results in all panels are representative of three independent experiments. induce NFB activation in NCM460 colonocytes, in contrast to earlier studies indicating a rapid toxin A-induced NFB activation in colon adenocarcinoma HT29 cells (16). Non-transformed (NCM460) versus transformed (HT-29) cell phenotypes may account for these different toxin A-associated responses. Our results show that the CRE element in the COX-2 promoter plays a major role in toxin A-mediated induction of intestinal COX-2 transcription. Yamaoka et al. (44) reported that lipopolysaccharide activated STAT5, which induced COX-2 through granulocyte macrophage-colony stimulating factor production in human monocytes. These authors also showed a ␥-interferon-activated sequence in the promoter region of human, mouse, and rat COX-2 genes, and reported binding of activated STAT5 to this element. Although our results indicated that toxin Ainduced phosphorylation of STAT5 was regulated by JAK1 in colonocytes (data not shown), we found that inhibition of JAK1 by AG490 had no effect on the PGE 2 secretion following toxin A exposure (Fig. 4B). Moreover, activation of ERK1/2, a known regulator of COX-2 expression, was not clearly evident in colonocytes following exposure to toxin A. These data suggest that p38 MAPK, but not JAK/STAT5, ERK1/2, and NFB is the major pathway regulating the COX-2 response to toxin A.
Gel shift assays using a CRE probe from the COX-2 promoter also indicate that CREB and ATF-1, but not c-Jun and ATF-2, bind to the COX-2 CRE. Toxin A induces the phosphorylation of MSK-1, CREB, and ATF-1, which are downstream of p38 MAPK. As expected, we found that the p38 inhibitor SB203580 blocks toxin A-induced COX-2 promoter activity and phosphorylation of MSK-1 and CREB/ATF-1.
We reported earlier that p38 MAPK regulates toxin Amediated induction of interleukin-8 expression in human monocytes (13). In addition, prostaglandin E 2 stimulates interleukin-8 gene expression in human colonocytes by a post-transcriptional mechanism (45). These observations suggest that toxin A activates p38 MAPK, which up-regulates COX-2 expression in intestinal epithelial cells and secretion of PGE 2 that in turn up-regulates interleukin-8 gene transcription. Thus, activation of colonocyte p38 MAPK by toxin A is a key modulator of several target genes that ultimately control infiltration of monocytic cells, acute intestinal inflam- FIG. 7. Toxin A-generated ROS regulate COX-2 expression and PGE 2 secretion. A, upper panel. Colonocytes were incubated with toxin A (3 nM), and intracellular ROS and cAMP levels were determined by DCFH-DA or ELISA, respectively. Bars and dots represent mean Ϯ S.D. of three to five independent experiments, each with triplicate determinations. Lower panel, fluorescence microscopy. NCM460 cells growing on slides were preincubated with DCFH-DA (green) and then exposed to toxin A (3 nM). B, colonocytes were pretreated with the hydrogen peroxide scavenger NAC, or the hydroxyl radical scavenger sodium formate for 30 min. Cells were then incubated with toxin A for 6 h and Western blot analysis was performed using antibodies against COX-2. C, colonocytes were treated with H 2 O 2 , and COX-2 expression and p38 phosphorylation were analyzed by Western blot analysis. D, colonocytes were first pretreated with NAC or sodium formate for 30 min and then exposed to toxin A for 1 h (left panel). Colonocytes were exposed for 1 h to H 2 O 2 (right panel). Cell lysates were resolved on 10% polyacrylamide gel and probed with antibodies directed against the indicated molecules; p indicates phosphorylated form. Cells were treated with either H 2 O 2 (E), or toxin A alone or together with NAC or sodium formate (F). After 12 h PGE 2 levels in the conditioned media were determined by ELISA. Bars represent mean Ϯ S.D. of three to five independent experiments, each with triplicate determinations; *, p Ͻ 0.0001 versus non-stimulated cells. mation, epithelial permeability, and intestinal electrolyte and water secretion. PGE 2 is a potent stimulator of acute fluid secretion by the intestine particularly during bacterial infection. PGE 2 and other prostanoids are elevated in luminal secretion from experimental animals and human with infectious diarrhea (46), and administration of PGE 2 analogues caused diarrhea in volunteers (47). Epidermal growth factor receptor activation induces basolateral release of PGE 2 by COX-2 in colon cancer cells (48). Infection of intestinal epithelial cell lines with Salmonella induces the expression of COX-2 and PGE 2 with subsequent apical chloride secretion (25). PGE 2 induces cAMP-mediated apical chloride secretion and inhibits electroneutral sodium chloride and water absorption in enterocytes (49,50). Cholera toxin also induces PGE 2 and fluid secretion that can be prevented by selective COX-2 inhibitors (51,52). C. difficile toxin A injected into the rabbit ileum also induces secretion of PGE 2 and massive fluid secretion (53), but the cell(s) of origin and the mechanism(s) of secretion of PGE 2 by toxin A are not clear. Increased COX-2 protein expression was reported in response to toxin A administration in rabbit ileum, particularly in lamina propria mononuclear cells, probably macrophages (18). Our results demonstrate that toxin A can induce COX-2 and secretion of PGE 2 in human intestinal epithelial cells. This process appears to require release of ROS by enterocytes within 10 min of toxin A exposure (Fig. 8A). Activation of the p38 pathway by ROS appears to be a critical early response to toxin A, because scavenging of hydrogen peroxide by NAC treatment suppresses COX-2 induction and secretion of PGE2 in cultured colonocytes, and blocking this pathway by the specific p38 inhibitor SB203580 markedly attenuates ileal mucosal damage and fluid secretion (Fig. 9) in intact mice. Moreover, the selective COX-2 inhibitor NS-398 markedly reduced the toxin A-mediated PGE 2 release and subsequent fluid secretion in mice, further supporting COX-2 as a major regulator of the intestinal responses to toxin A.
In summary, we demonstrate that toxin A directly induces COX-2 expression in human colonocytes through activation of p38, MSK-1, and CREB/ATF-1 pathways (Fig. 9). Activation of p38, MSK-1, and CREB/ATF-1 pathways appears to be impor- FIG. 8. Blocking of p38 MAPK inhibits toxin A-induced COX-2 expression in mouse ileum. Closed ileal loops were prepared in the distal ileum of anesthetized CD1 mice (n ϭ 3 mice per group) and injected with the p38 MAPK inhibitor SB203580 or the selective COX-2 inhibitor NS-398 30 min prior to toxin A (10 g) or buffer administration. After 2 h, ileal fluid was collected and ileal loops were removed and processed as described below. A, tissue lysates were resolved on 10% polyacrylamide gel and probed with a COX-2 antibody. Each lane represents blots from control (C), toxin A (T), and toxin A with SB203580-injected mice (S). B, total RNA was isolated and COX-2 mRNA expression was analyzed by RT-PCR. Results in A and B are representative of three separate experiments for each condition. C, fluid secretion expressed as the loop weight-to-length ratio (mg/cm). D, PGE 2 secretion. *, p Ͻ 0.001 versus non-stimulated cells; **, p Ͻ 0.005 versus toxin A-stimulated cells. E, frozen tissue slides stained with propidium iodide (red) to stain cellular nuclei. F, tissues were exposed to fluorescein (green) isothiocyanate-conjugated COX-2 antibody and then counterstained with propidium iodide (magnification, ϫ100). Toxin A-induced expression of COX-2 in both epithelial and lamina propria cells of mouse ileal mucosa (arrows) were markedly reduced in mice treated with SB203580. tant as determined by chemical inhibitor studies in cell culture and in intact intestine. Inducible COX-2 and PGE 2 release are critical mechanisms for the luminal fluid secretion and enteritis following toxin A exposure. Our findings suggest that COX-2 activation and PGE 2 release may represent a potential therapeutic target in C. difficile toxin A-mediated colitis.