key: cord-0007280-4rn9534g
authors: Subramanian, Sreedhar; Rhodes, Jonathan M.; Hart, Anthony C.; Tam, Ben; Roberts, Carol L.; Smith, Shirley L.; Corkill, John E.; Winstanley, Craig; Virji, Mumtaz; Campbell, Barry J.
title: Characterization of epithelial IL-8 response to inflammatory bowel disease mucosal E. coli and its inhibition by mesalamine
date: 2008-02-01
journal: Inflamm Bowel Dis
DOI: 10.1002/ibd.20296
sha: c3d7c160a58080b25da048ab095cd453e02f5527
doc_id: 7280
cord_uid: 4rn9534g

BACKGROUND: Mucosally adherent E. coli are found in inflammatory bowel disease (IBD) and colon cancer. They promote release of the proinflammatory cytokine interleukin-8 (IL-8). We explored mechanisms for this release and its inhibition by drugs. METHODS: IL-8 release from colon epithelial cells in response to mucosal E. coli isolates from IBD, colon cancer, and controls was characterized at the cellular and molecular level. RESULTS: IL-8 response of HT29 cells was greater with Crohn's disease (689 ± 298 [mean ± SD] pg IL-8/mL at 4 hours, n = 7) and colon cancer isolates (532 ± 415 pg/mL, n = 14) than with ulcerative colitis (236 ± 58 pg/mL, n = 6) or control isolates (236 ± 100 pg/mL, n = 6, P < 0.0001). Bacterial supernatants contained shed flagellin that triggered IL-8 release. For whole bacteria the IL-8 response to E. coli that agglutinate red blood cells (548 ± 428 pg IL-8/mL, n = 16), a function that correlates with epithelial invasion, was greater than for nonhemagglutinators (281 ± 253 pg/mL, n = 17; P < 0.0001). This was particularly marked among E. coli that, although flagellate, could not release IL-8 from TLR5-transfected HEK293 cells. IL-8 release was mediated by extracellular-regulated kinase (ERK) and p38 mitogen-activated protein kinase (MAPK) and inhibited by mesalamine, but not hydrocortisone, at therapeutic concentrations. CONCLUSIONS: Mucosa-associated E. coli shed flagellin that elicits epithelial IL-8 release but this may only become relevant when the mucosal barrier is weakened to expose basolateral TLR5. Adherent and invasive IBD and colon cancer E. coli isolates also elicit a flagellin-independent IL-8 response that may be relevant when the mucosal barrier is intact. The IL-8 release is MAPK-dependent and inhibited by mesalamine.

I nflammatory bowel disease (IBD) is thought to result from an aberrant mucosal immune response to the gut microbiota. This is supported by evidence that various animal models of IBD all require the presence of commensal bacteria to develop inflammation. 1 At least 7 independent studies have now reported an increased presence of adherent and invasive E. coli (AIEC) strains in Crohn's disease (CD) 2-8 and they have also been found in ulcerative colitis (UC) 3,6,9 and colon cancer. 4, 6 Similar E. coli have recently been shown to be associated with granulomatous colitis in boxer dogs. 10 They are present in both inflamed and uninflamed mucosa, 4, 6 suggesting that they are not just recruited as a consequence of inflammation and it is looking increasingly likely that they play an important role in IBD pathogenesis.

Although there is evidence of tissue invasion by E. coli in CD, where they have been identified within macrophages 11 and inside granulomata, 12 there is no evidence that they invade the mucosa in UC and their ability to elicit an IL-8 response from intestinal epithelial cells, 4 and thus to promote the superficial neutrophil recruitment 13 that typifies UC, is a likely alternative mechanism for a pathogenic effect. Moreover, the demonstration that unequivocal pathogens such as Salmonella spp, Shigella spp, Yersinia spp, and Vibrio chol-erae can only invade the gut mucosa via the specialized M cells that overlay Peyer's patches in the small intestine and lymphoid follicles in the colon and cannot invade the mucosa via normal colon epithelial cells 14 makes it unlikely that the AIEC found in IBD and colon cancer invade normal colon epithelial cells, although it is plausible that these bacteria might invade M cells and thus provoke the initial lesions of CD which probably occur at Peyer's patches or lymphoid follicles. 15 In this study we set out to characterize the mechanisms involved in the epithelial IL-8 response to IBD and colon cancer mucosal E. coli isolates with particular emphasis on assessing the relative importance of the AIEC phenotype in mediating this response, the downstream signaling mechanisms involved, and the effects of drugs used to treat IBD.

Escherichia coli were previously isolated from colonic mucosal biopsies of patients with CD, colon cancer, UC, and a control population (irritable bowel syndrome [IBS]/sporadic polyposis). 4 Thirty-three isolates were studied including 7 from CD, 6 from UC, 14 from colon cancer, and 6 control patient (irritable bowel or sporadic polyp) E. coli. The CD E. coli were isolated from 7 CD patients (2 were receiving corticosteroids and azathioprine, respectively, and the remainder were on 5-aminosalicylic acid ); UC E. coli were isolated from 5 patients (all 5 on 5-ASA and 2 were receiving corticosteroids); control E. coli were isolated from 2 patients (IBS and sporadic polyps), and colon cancer E. coli were isolated from 11 patients (2 patients undergoing resection received intravenous cephalosporin and metronidazole). The phylogenetic group of each E. coli (A, B1, B2, or D) was defined by triplex polymerase chain reaction (PCR) as described previously. 16 The characteristics of the bacterial strains used in this study are summarized in Table 1 .

Isolates were cultured on Columbia agar with overnight incubation in air at 37°C. Prior to infection of cultured epithelial cells, bacteria were washed in sterile PBS and adjusted to an OD 550nm of 0.125 (equivalent to 150 ϫ 10 6 CFU/mL). In some studies E. coli were heat-inactivated at 60°C for 30 minutes. To demonstrate whether E. coli possessed flagellae, isolates were stab-inoculated into 0.3% motility agar. To assess for the presence of curli fimbriae, E. coli isolates were grown on nutrient agar supplemented with 30 g/mL Congo red for 48 hours at 26°C.

SP600125 (an inhibitor of c-Jun N-terminal kinase [JNK]), SB203580 (a pyridinyl imidazole inhibitor of p38-MAPK pathway), and Bay11-7082 (an irreversible inhib-itor of NF-B activation, through inhibition of IB-␣ phosphorylation) were all obtained from Calbiochem (Nottingham, UK). U0126 (a MEK inhibitor that inhibits the kinase activity of ERK1/2) was obtained from Promega (Madison, WI). Monoclonal antibodies to JNK1 (SC1648), phospho-JNK1/2 (SC6254), and polyclonal antibodies to ERK2 (SC154G) and TLR5 (SC10742) were all from Santa Cruz Biotechnology (San Francisco, CA). Other monoclonals included anti-phosphoERK1/2 (Cell Signaling Technology, Beverley, MA), pan-actin antibody (Neomarkers, Fremont, CA) and anti-Ras (Calbiochem). Polyclonal rabbit anti-flagellar serum was a kind gift from Dr. J. Giron (Microbiology & Immunology, University of Arizona, Tucson). An NF-B p65 subunit siRNA kit (Cell Signaling Technology) included anti p65 and ERK2 antibodies.

Recombinant proteins included human IL-8 (Peprotech EC, London, UK), TNF-␣ (R&D Systems, Abingdon, UK), and E. coli flagellin (Inotek, Beverley, MA). E. coli O127:B8 lipopolysaccharide (LPS), polymixin B, hydrocortisone, methyl ␣-D-mannopyranoside, and protease type XIV were all from Sigma (Poole, UK). Human leukocyte elastase was from Elastin Products (Owensville, MO). 5-ASA (mesalamine) was a kind gift from Dr. S. Andersson (Pharmacia, Uppsala, Sweden). The carcinoembryonic antigen-related cell adhesion molecule (CEACAM)-binding polypeptide rD-7 was produced using the pQE30 expression system as described previously. 17 The rD-7 polypeptide had been shown to inhibit the binding of respiratory and urogenital pathogens (M. catarrhalis, H. influenzae, N. gonorrhoeae, and N. meningitidis) via inhibition of CEACAM-dependent bacterial adhesion. 17

Human colon cancer cell lines HT29 and T84 were obtained from the European Collection of Animal Cell Culture (Public Health Laboratory Service, Wiltshire, UK). HT29 cells were cultured in DMEM containing 10% FCS, 100 U/mL penicillin, 100 g/mL streptomycin, and 4 mM glutamine. HEK293 uroepithelial cells, stably transfected with pUNO-hTLR5 (encoding human toll-like receptor 5), pUNO-LacZ (encoding ␤-galactosidase) or vector alone were obtained from Invitrogen (San Diego, CA). HEK293 were cultured in complete DMEM supplemented with 10 g/mL blasticidin. Polarized T84 model epithelia were cultured in Ham's F12 medium (10% FCS, penicillin and streptomycin) on 0.45 m collagen-coated permeable supports (Millipore, Watford, UK). Monolayers with a transepithelial electrical resistance (TEER) of Ͼ800°/cm 2 were used for assessment of flagellin transcytosis. All cell lines were maintained at 37°C in 5% CO 2 , 95% air.

HT29 cells (8 ϫ 10 5 per well) were infected, in triplicate, for up to 8 hours, with apical addition of whole E. coli isolates from each disease group (7 CD, 14 colon cancer, 6 UC, and 6 controls) at a multiplicity of infection (MOI) of 10 (Ϸ7 ϫ 10 6 CFU/mL) in keeping with previous studies of bacterial adherence to intestinal cell lines. 2,4 IL-8 released to the medium was measured by enzyme-linked immunosorbent assay (ELISA). TNF-␣ (2 ng/mL, 4 hours) was used as a control stimulus.

Pooled bacteria-free E. coli culture supernatants were generated from E. coli isolates, 6 each pooled separately from 

Crohn's disease -n/a n/a ϩ ϩ n/a n/a B2 ϩ

HM250 a Ulcerative colitis ϩ ϩ n/a n/a n/a n/a B2 n/a HM295 a Ulcerative colitis ---n/a n/a n/a n/a B2 n/a HM378 a Ulcerative colitis ---n/a n/a n/a n/a B2 n/a HM380 a Ulcerative colitis --ϩ n/a n/a n/a n/a B2 n/a HM387 a Ulcerative colitis --ϩ n/a n/a n/a n/a B2 n/a HM394 a Ulcerative colitis --ϩ n/a n/a n/a n/a B2 n/a HM452 a Control ---n/a n/a n/a n/a B2 n/a HM454 a Control ϩ ϩ n/a n/a n/a n/a B2 ϩ HM455 a Control ---n/a n/a n/a n/a B2 n/a HM456 a Control ϩ ϩ n/a n/a n/a n/a B2 ϩ HM488 a Control ---n/a n/a n/a n/a B2 ϩ HM489 a Control --ϩ n/a n/a n/a n/a B2 n/a a Denotes strains used for generation of pooled bacteria-free E. coli culture supernatant. n/a indicates data not available.

4 disease groups (UC, CD, colon cancer, and control patients). Bacterial suspensions (OD 550nm 0.125) were 0.2 mfiltered and supernatants confirmed bacteria-free by overnight culture on Columbia agar. Epithelial cells were treated with 50 L supernatant/mL culture medium and IL-8 release compared to supernatants predigested (3 h, 37°C) with either 0.25 M leukocyte elastase or 20 g/mL Streptococcal protease XIV to remove flagellin, as described previously. 18 Heatinactivated elastase and protease (90°C, 15 min) were used as controls.

Following treatment, culture medium and cell monolayers were harvested and processed as previously described. 19 IL-8 was measured in triplicate using a solid-phase sandwich ELISA with antihuman IL-8 capture and detection antibodies (790A 28G2 and 893C 4G2; Biosource, Camarillo, CA). Within and between assay coefficients of variation were 10.1% (n ϭ 3) and 13% (n ϭ 7), respectively.

Flagella from CD (HM427) and colon cancer (HM288) E. coli isolates were purified by ammonium sulfate precipitation. 20 Purity was monitored by SDS-PAGE, immunoblotting, and transmission electron microscopy (TEM). OMVs were isolated by ultracentrifugation from a 0.2 m-filtered E. coli culture supernatant, pooled from 6 mucosal E. coli (5 hemagglutination-positive and 1 negative isolate, from 3 UC patients) or E. coli HM427 supernatant, as described previously. 21 In separate experiments, pretreatment of bacterial suspensions with polymixin B (PMB; 10 g/mL, 1 h) was used to block any action of LPS. IL-8 release evoked by LPS derived from E. coli serotype O127:B8 (1 g/mL) was used as a control for the inhibitory action of polymixin B.

Polarized T84 model epithelia were apically colonized with 10 8 CFU/mL CD E. coli HM427. After 4 hours, flagellin present within the apical and basolateral media was analyzed by SDS-PAGE as per Lyons et al 22 followed by immunoblotting with anti-flagellar serum.

To assess the role of TLR5, HT29, and HEK293 cells (expressing TLR5, LacZ, or vector alone) were apically infected (MOI 10), for 4 hours with each of 33 E. coli and also with pooled bacteria-free E. coli culture supernatants, purified flagella (1 g/mL), recombinant flagellin (1 g/mL), or OMVs (from pooled UC isolates). TNF-␣ (2 ng/mL) was used as a positive control.

To investigate the role of CEACAMs and Type 1 (mannose-dependent) fimbriae, HT29 cell monolayers were pretreated with anti-CEACAM peptide rD-7 (1 g/mL) or methyl ␣-D-mannopyranoside (10 mM) for 60 -90 minutes. Subsequently, cells were incubated for 4 hours with apical addition of whole E. coli isolates. Inhibitors were maintained in the medium throughout the period of infection. IL-8 released into the medium was measured by ELISA and compared to untreated controls.

Amplification of fliC was performed by PCR in a total reaction volume of 50 L containing 1 L of bacterial supernatant, 100 nM each of fliC oligonucleotide primers (sense 5Ј ATGGCACAAGTCATTAAT 3Ј and anti-sense 5Ј TTAACCCTGCAGTAGAGA 3Ј) and 1.1x PCR Master Mix (ABgene, Surrey, UK). Thermal cycling conditions were 95°C for 5 minutes, followed by 30 cycles of 95°C for 1 minute, 50°C for 1 minute, 72°C for 2 minutes, with a final extension cycle of 72°C for 10 minutes. Amplification was verified using agarose gel electrophoresis (10% reaction volume) and the remaining 90% purified using MicroSpin S-400 HR columns (Amersham Biosciences, Little Chalfont, UK). DNA sequencing was performed by Lark Technologies (Essex, UK) and deduced amino-acid residues from fliC amplicon sequences were analyzed using CLUSTAL W (1.83) multiple sequence alignment software. E. coli flagellin (FliC) sequences were compared with respect to those regions previously determined to be important for TLR5 recognition. [23] [24] [25] 

Following apical infection with each of the 33 E. coli isolates, IL-8 release was assessed in HT29 cells pretreated (all 10 M) with U0126 (to block ERK1/2 activity), SB203580 (to inhibit p38 MAPK pathway), or SP600125 (to inhibit JNK activity) for 60 -90 minutes. Inhibitors were present throughout the 4-hour infection. In separate experiments, MAPK activation by CD and colon cancer isolates, for up to 60 minutes, was confirmed by SDS-PAGE and immunoblotting as described previously. 26 Phorbol 12-myristate 13-acetate (PMA) (0.25 M, 20 min) was used as a positive control for MAPK activation.

Activation of Ras, the upstream guanine-nucleotide binding protein that stimulates the Raf-MEK-MAPK pathway, 27 was assessed over a 30-minute infection. Glutathione S transferase (GST)-K85A RBD (Ras-binding domain of Raf1) fusion protein was used to pull down active (GTP-bound) Ras, as described elsewhere. 28 Total lysates and pulldown samples were analyzed by SDS-PAGE and immunoblot using antibody against Ras. Equal loading was confirmed using ERK2 antibody.

To assess the role of NF-B, IL-8 release was assessed in response to each of the 33 E. coli isolates from HT29 cells pretreated with IB␣ phosphorylation inhibitor Bay11-7082 (10 M, 60 min) to block NF-B activation. Inhibitor was present throughout the 4-hour infection. Downregulation of NF-B was also performed using p65 subunit siRNA duplexes introduced to 50% confluent HT29 cells in the presence of 8 L/mL TransIT-TKO (Cambridge Bioscience) 24 hours before treatment with flagella or apical addition of whole E. coli. Suppression of the p65 subunit was confirmed by immunoblotting. 

Protein and double-stranded DNA concentrations of cell lysates were determined using the bicinchoninic acid assay (Sigma) and PicoGreen (Molecular Probes, Paisley, UK) as per the manufacturer's instructions.

5-ASA and hydrocortisone were assessed for cytotoxicity toward the cultured epithelial cells using the adenylate kinase assay (ToxiLight Non-Destructive Cytotoxicity Bio-Assay Kit, Cambrex Bio, Cambridge, MA) as per the manufacturer's instructions.

Independent sample groups were assessed for both normality and for equality of variances. As appropriate, treatment groups were analyzed using Student's t-test or Mann-Whitney U-test. For multiple treatment groups, 1-way or Kruskall-Wallis analysis of variance (ANOVA) was employed, followed by pairwise comparisons of treatment means (StatsDirect v2.3.1; Sale, UK). Differences were considered significant when P Ͻ 0.05.

All mucosal E. coli isolates from CD (n ϭ 7) and colon cancer (n ϭ 14) were shown to stimulate the release of IL-8 from HT29 cells in a time-dependent manner (Fig. 1A) . A lower IL-8 release was seen in response to E. coli isolates from control patients (IBS/sporadic polyposis) (236 Ϯ 100 pg IL-8/mL at 4 hours [all mean Ϯ SD unless indicated otherwise], n ϭ 6) and those from UC (236 Ϯ 58 pg/mL, n ϭ 6) compared with isolates from CD (689 Ϯ 298 pg/mL, n ϭ 7) and colon cancer (532 Ϯ 415 pg/mL, n ϭ 14; P Ͻ 0.0001, ANOVA) (Fig. 1B) .

Intracellular IL-8 content was also increased in response to treatment with CD and colon cancer E. coli, rising from a median of 2.4 pg/mL (range 1.2-51 pg/mL) at 0.5 hours to a median of 261 pg/mL (range 108 -601 pg/mL) at 8 hours as compared to uninfected control cells: 2 pg/mL (1.3-2.2) and 60 (59.5-69.5) pg/mL at 0.5 and 8 hours, respectively (P Ͻ 0.01 ANOVA; n ϭ 21, i.e., 7 CD and 14 colon cancer isolates). No significant changes were seen in cellular double-stranded DNA levels in the HT29 cells when infected with mucosal E. coli isolates over the 8-hour incubation period (201 Ϯ 73 ng/mL DNA) at the end of the treatment period compared to 207 Ϯ 85 ng/mL in untreated controls (n ϭ 21).

Our previous work had shown a strong correlation between the ability of mucosal E. coli to hemagglutinate human red blood cells and their ability to adhere to intestinal epithelial cells in culture. 4 E. coli that cause hemagglutination (n ϭ 16) induced a greater IL-8 response from HT29 cells (548 Ϯ 428 pg IL-8/mL) than nonhemagglutinators (281 Ϯ 253 pg/mL, n ϭ 17; P Ͻ 0.0001 Mann-Whitney U) (Fig.  1C) ; heat-inactivation of CD E. coli isolate HM427 only caused a modest reduction in its ability to induce an IL-8 response: 203 Ϯ 53 pg/mL at 4 hours compared with 323 Ϯ 84 pg/mL for live E. coli HM427 (n ϭ 3, P ϭ 0.01, Mann-Whitney U).

Twenty-four E. coli isolates, 6 each from control patients (including 1 hemagglutinator), CD (including 4 hemagglutinators), UC (including 1 hemagglutinator), and colon cancer (including 3 hemagglutinators) were pooled to form a bacterial suspension and a bacteria-free supernatant was derived by passage through a 0.2-m filter. Filtered supernatants caused an IL-8 response, (911 Ϯ 292 pg/mL) that was 51.3 Ϯ 10.9% of that produced by an equivalent preparation of whole bacteria (1763 Ϯ 314 pg/mL) (n ϭ 24, P Ͻ 0.0001, Mann-Whitney U).

Pretreatment of bacteria-free E. coli supernatants with either protease XIV (20 g/mL for 3 h at 37°C) or leukocyte (Fig. 2) . These findings would be compatible with flagellin as the stimulus for IL-8 release.

Flagella, isolated from 7 ϫ 10 6 E. coli, either colon cancer isolate HM288 or the CD isolate HM427 (Fig. 3A) , caused an IL-8 release of 178 Ϯ 40 pg/mL and 148 Ϯ 38 FIGURE 1. Stimulation of IL-8 release from HT29 colon epithelial cells by mucosal E. coli. A: CD mucosal E. coli (n ϭ 7) stimulate IL-8 release from confluent HT29 cells in a time-dependent manner. IL-8 release was measured by ELISA and data expressed as pg/mL (mean Ϯ SEM). Solid bar represents IL-8 release at 8 hours induced by 2 ng/mL TNF-␣ (n ϭ 6). Significant differences from 8 hours basal release, ***P Ͻ 0.0001 (ANOVA). B: A significantly greater IL-8 response was seen following treatment using mucosal E. coli isolated from patients with CD (n ϭ 7) and colon cancer (n ϭ 14) compared to E. coli isolated from UC (n ϭ 6) or control patients (n ϭ 6). Significant differences from basal IL-8 response, **P Ͻ 0.01, ***P Ͻ 0.001 and significant differences from control and UC patient responses, (***)P Ͻ 0.001 (Kruskal-Wallis ANOVA). C: Hemagglutinating E. coli isolates (n ϭ 16) elicited a greater IL-8 response from HT29 cells than nonagglutinating E. coli strains (n ϭ 17), ***P Ͻ 0.0001 (Mann-Whitney U). pg/mL, respectively, as compared to 220 Ϯ 34 pg/mL and 201 Ϯ 29 pg/mL in response to equivalent amounts of whole bacteria (Fig. 3B ). Almost all (99 Ϯ 5%, n ϭ 3) of the IL-8-stimulatory activity of the bacterial supernatant was shown to be attributable to an OMV preparation (Fig. 4A) .

OMV are known to contain proteins relevant to pathogenicity 29 but immunoblotting showed flagellin to be present within the purified OMV preparations (Fig. 4B) and TEM confirmed the presence of flagella associated with OMVs ( Fig. 4C) , again compatible with flagellin as the major stimulant for IL-8 response to the bacterial supernatants.

Inhibition of LPS by preincubation of bacterial suspension with 10 g/mL polymyxin B for 1 hour only caused a reduction of 7 Ϯ 4% in IL-8 response to 2 mucosal E. coli isolates from CD.

Having shown that flagellin was a major determinant of the IL-8 response to the mucosal E. coli, we investigated the role of the known flagellin-receptor, TLR5, by testing the IL-8 response in HEK293 cells with and without transfected TLR5. Nineteen of the 33 E. coli isolates tested (8/14 colon cancer isolates and 5/7 CD and 2/6 each of UC and 4/6 control E. coli) elicited IL-8 release in a TLR5-dependent fashion, defined as a response to TLR5-expressing HEK293 cells Ն20% of that seen with HT29 cells (Fig. 5A ). For these E. coli the median IL-8 response was 394 Ϯ 253 pg/mL in HEK-hTLR5 cells compared to 32 Ϯ 42 pg/mL in HEK293-LacZ cells. For E. coli where the IL-8 release was not mediated by flagellin-TLR5 interaction, the median IL-8 response was only 23 Ϯ 18 and 38 Ϯ 37 pg/mL, respectively, from HEK293-hTLR5 and HEK293-LacZ cells (n ϭ 14). TNF-␣ (2 ng/mL, for 4 h) evoked a similar IL-8 response from TLR5-and ␤-galactosidase-expressing HEK293 cells. All 33 E. coli isolates were able to induce an IL-8 release from HT29 cells, with a median IL-8 release of 405 Ϯ 346 pg/mL. HT29 cells were also shown to express TLR5 (Fig.  5B) .

Amplification of fliC from representative mucosa-associated E. coli isolates, able (n ϭ 4) or unable (n ϭ 5) to induce TLR5-dependent IL-8 production in HEK293-hTLR5 cells, showed some variation in amplicon length but no correlation with IL-8 response in HEK293-hTLR5 cells (data not shown). In addition, deduced amino-acid sequences (aligned using the CLUSTAL W algorithm) demonstrated no significant differences in flagellin residues within those regions previously determined to be important for TLR5 recognition, [23] [24] [25] indicating that all our mucosa-associated E. coli isolates are likely to make flagellin molecules that can be recognized by TLR5 (see Table 2 ). All 33 isolates were shown to possess flagella and to be motile on 0.3% agar plates.

Among the 14 E. coli isolates that did not mediate IL-8 release via TLR5, the IL-8 response from HT29 cells was considerably greater for hemagglutinators (median 513 pg/ mL, range 277-813) than for nonhemagglutinating strains (median 175 pg/mL, range 48 -295; P ϭ 0.004 Mann-Whitney U). Study of pooled bacterial supernatants from each disease group and isolated OMV confirmed the TLR5-mediated IL-8 response (see Fig. 5C ).

Curli fimbriae, an extracellular matrix component of Enterobacteriaceae (including commensal E. coli), have recently been shown to be involved in IL-8 release from HT29 cells via interaction with flagellin. 30, 31 We therefore assessed all 33 E. coli isolates in this study for the presence of curli fimbriae and its association with TLR5-dependent epithelial cell IL-8 release. Overall, 17/33 E. coli isolates possessed curli fimbriae. There was a good association between presence of curli and the magnitude of IL-8 response from HEK293-hTLR5 cells (P ϭ 0.05, unpaired t-test). However, there was no relationship seen between curli and hemagglutination status of the E. coli, nor any association with isolates from specific patient groups.

It is noteworthy that supernatants from CD, colon cancer, and UC E. coli isolates, unlike the whole bacteria, all cause similar IL-8 responses, suggesting that the enhanced IL-8 responses to whole bacteria from CD and colon cancer are not explained by flagellin-TLR5 interaction.

Following a 4-hour incubation of polarized T84 epithelium with CD mucosa-associated E. coli HM427, an AIEC, 4 applied to the apical aspect, flagellin was undetectable in the basolateral medium as determined by immunoblot analysis (n 

CEACAM6 was recently shown to act as a receptor for AIEC adhesion and was found to be upregulated in the ileal mucosa of CD patients. 32 Pretreatment of HT29 cells with 1 g/mL recombinant polypeptide rD-7, an antimicrobial agent that blocks the interactions of a number of mucosal pathogens with CEACAM-expressing epithelial cells, 17 failed to reduce the overall IL-8 response to all 33 E. coli isolates (IL-8 release in presence of rD-7: 298 Ϯ 66 pg IL-8/mL, compared to untreated controls 323 Ϯ 72 pg/mL). Only 3/33 E. coli isolates showed a reduction in IL-8 response of Ͼ20% (range 21%-32%) with rD-7 but there was neither any disease group association, relation to hemagglutination status, nor correlation with possession of Afa/Dr. We also assessed the possible inhibitory effect of addition of 10 mM ␣-methyl mannopyrannoside; however, this showed a similar IL-8 response (301 Ϯ 68 pg/mL, n ϭ all 33 E. coli isolates) compared with untreated controls. These studies suggest little or no role for Afa/CEACAM interaction in the IL-8 response.

We next investigated the intracellular signaling pathways responsible for the IL-8 response. Inhibition of ERK1/2 activation by pretreatment with 10 M U0126 for 1 hour reduced IL-8 release in response to whole bacteria by 77 Ϯ 14%, from 610 Ϯ 355 to 194 Ϯ 107 pg IL-8/mL across all 33 E. coli isolates (P Ͻ 0.0001, ANOVA) (Fig. 6A) . Similarly, inhibition of p38 MAPK activation by SB203580 (10 M) reduced IL-8 release by 84 Ϯ 18%, from 509 Ϯ 322 to 149 Ϯ 81 pg IL-8/mL (P Ͻ 0.0001, ANOVA; n ϭ 33) (Fig.  6A ). By contrast, inhibition of c-Jun N-terminal kinase (JNK) activation by pretreatment of HT29 cells with 10 M SP600125 for 90 minutes caused a more modest reduction in (Fig. 6A) . Treatment of HT29 cells with mucosal CD E. coli HM427 resulted in the activation of Ras between 1 and 5 minutes, returning to basal levels 30 minutes after infection. Peak activation occurred at 2 minutes, with a 1.8 Ϯ 0.62 (mean Ϯ SEM; n ϭ 3) fold increase in active (GTP-bound) Ras (Fig. 6B,C) . This was followed by subsequent activation of ERK1/2 within 5 minutes of infection (Fig. 6D,E) . Peak activation was seen at 15 minutes, with 1.15 Ϯ 0.05 and 1.94 Ϯ 0.25-fold increase in phosphorylation of ERK1 (p44) and ERK2 (p42), respectively (mean Ϯ SEM; n ϭ 3). Phosphorylation of ERK1/2 was sustained for at least 60 minutes (Fig. 6D,E) . Blockade of NF-B activation using the IB␣ phos-phorylation inhibitor Bay 11-7082 (10 M for 1 h) reduced E. coli-induced IL-8 release by 29 Ϯ 23%, from 349 Ϯ 108 to 255 Ϯ 129 pg/mL across all 33 E. coli isolates (P Ͻ 0.0001, ANOVA). A greater than 50% reduction in IL-8 release by NF-B inhibition was only observed in response to 11/33 E. coli isolates. For 2 of these isolates, colon cancer E. coli HM288 and CD E. coli HM427, both hemagglutinators, coincubation with Bay 11-7082 caused Ͼ80% inhibition of IL-8 response. Pretreatment of HT29 cells for 24 hours with siRNA to down-regulate the p65 subunit of NF-B resulted in a 48 Ϯ 8% and 45 Ϯ 19% reduction in IL-8 release in response to HM288 and HM427, respectively (P Ͻ 0.001, ANOVA; n ϭ 3) (Fig. 7A,B) . These studies show that the IL-8 response is largely ERK/MAPK-dependent, with relatively little input from JNK or NF-B activation. 

Mesalamine, a drug effective in the treatment of UC, is known to inhibit MAPK. 33 We assessed the effect of mesalamine on the E. coli IL-8 response and also the effect of hydrocortisone, a drug that is also effective in UC but that probably has its main action via inhibition of NF-B activation. Pretreatment of HT29 cells with 5-ASA dose-dependently inhibited the IL-8 response to a bacteria-free E. coli culture supernatant pooled from 6 UC isolates (Fig. 8A) . Maximal inhibition of IL-8 release (97 Ϯ 1%) was seen with 20 mM 5-ASA and a plausible therapeutic concentration, 5 mM 5-ASA, 34 inhibited IL-8 release by 45 Ϯ 17%. A similar effect was also observed in response to whole E. coli isolates (data not shown). Hydrocortisone also reduced the E. coli supernatant-induced IL-8 response but only by 49 Ϯ 13% at 100 M, a concentration Ϸ20-fold greater than that likely to occur at usual therapeutic doses 35 (P Ͻ 0.05 ANOVA) (Fig.  8B) . At the maximal concentrations tested, 5-ASA (20 mM) and hydrocortisone (100 M), the adenylate kinase levels released in to the culture medium was within 90%-110% of vehicle-treated control cells.

In this study we demonstrate that mucosal E. coli isolates from IBD and colon cancer induce an IL-8 response from HT29 colon epithelial cells that is mediated by activation of the ERK1/2 and p38 MAPK pathways but not by activation of NF-B. This activity is inhibited by therapeutic concentrations of mesalamine but not hydrocortisone. For many of the E. coli isolates, similar IL-8 stimulatory activity is present in the bacteria-free culture supernatants. This is shown to be due to flagellin, shed along with OMVs, interacting with TLR5 on the epithelial cells. Some E. coli isolates, however, are unable to induce IL-8 release via flagellin-TLR5 interaction and for these bacteria the IL-8 response correlates with their ability to agglutinate red blood cells and to adhere to epithelial cells. The whole bacteria-mediated IL-8 response consequently tends to be greater with the more adherent E. coli isolated from CD and colon cancer than with the UC and control E. coli isolates.

This study has shown a substantial IL-8 response to E. coli isolated from the control groups used (IBS/sporadic polyps). It needs to be recognized that these are not completely healthy controls and that there is increasing evidence for subtle inflammatory changes in some patients with IBS. 36 There is also growing interest in the possibility that bacteria may have a role in the development of sporadic adenomas and colon cancer. 37 Flagellin initiates the inflammatory response to several pathogenic organisms including enteropathogenic E. coli 38 -40 The present study illustrates that intestinal bacteria need not possess classical pathogenicity genes in order to elicit this response and also shows that epithelial IL-8 release may be promoted by such bacteria without any need for cellular invasion. This finding is consistent with previous reports of nonpathogenic E. coli strains capable of inducing cytokine response from explanted colonic mucosa 41 and from intestinal epithelial cells in vitro. 42 The flagellin receptor, TLR5, is largely expressed on the basolateral aspect of intestinal epithelial cells, 43 so this raises the question of how this receptor will be accessed by the mucosal E. coli without cellular invasion. Although there is some evidence from human colonic epithelium 44 and murine ileal epithelium 42 of low-level expression of TLR5 on the apical surface, the healthy colonic epithelium is probably not responsive to luminal flagellar stimulation. 45 This situation changes in the presence of inflammation, however. Mice cotreated with dextran sulfate sodium (DSS) and flagellin FIGURE 6. The IL-8 response to mucosal E. coli is mediated via the Ras-ERK1/2 and p38 MAPK pathways. A: IL-8 release in response to all 33 mucosal E. coli isolates was markedly reduced in the presence of inhibitors of the ERK and p38 MAPK pathway (U0126 and SB203580, respectively). In contrast, inhibition of JNK activation with 10 M SP600125 only caused a modest reduction in IL-8 response. All P Ͻ 0.001 (ANOVA). B: Mucosal E. coli isolates cause rapid activation of Ras within 1 minute in HT29 cells, returning to basal levels 30 minutes after infection. C: Densitometric analysis of immunoblots for active (GTP-bound) Ras. Ras activation was expressed relative to control (n ϭ 3 experiments; mean Ϯ SEM). D: Subsequent phosphorylation of ERK1/2 (p44 and p42) occurred within 5 minutes of infection with E. coli. Peak activation was observed at 15 minutes and phosphorylation sustained for at least 60 minutes. E: Densitometric analysis of the p44/p42 immunoblots. ERK1/2 activity was expressed relative to control (n ϭ 3 experiments; mean Ϯ SEM).

develop a much more florid colitis than those treated with DSS alone, implying that flagellin plays an important role in the development and progression of colitis once the mucosal barrier is impaired. 45 Thus, the impaired barrier that probably exists as a result of altered tight junction structure and function in UC 46 and CD 47 is likely to be sufficient for flagellin to access basolateral TLR5 in vivo. The weakened mucus barrier in ulcerative colitis would be an additional factor. 48 In mammals, 3 major MAPK pathways have been identified: MAPK-ERK1/2, stress-activated p38 MAPK (␣, ␤, ␥, and ␦), and JNK1/2/3. 49 In most cases activation of MAPK signaling networks occurs by small G proteins such as Ras, Rac, and Rap1, but also involves other enzymes. 27 Our finding that the epithelial IL-8 response to mucosal E. coli isolates is mediated via Ras-ERK1/2 and p38 MAPK is consistent with the reported involvement of MAPK in the host cell responses to infection with pathogenic E. coli such as enterohemorrhagic E. coli (EHEC), 50 EPEC, 51 and EAggEC. 38 Phosphorylation of MAPK leads to downstream activation of transcription factors including NF-B and AP-1, both of which upregulate IL-8 transcription. It has been shown that p38 MAPK, JNK1/2, and ERK1/2 are activated in the inflamed colonic mucosa of IBD patients, with the p38␣ isoform exhibiting the greatest increase in phosphorylation. 52 There is preliminary evidence that MAPK inhibitors may be therapeutically effective in IBD. 53 5-ASA (mesalamine)-based compounds are used widely in the therapy of IBD and have a variety of effects that include inhibition of MAPK. They act locally at the mucosal level and efficacy is related to mucosal concentrations of the drug. 54 In vitro effects of 5-ASA include inhibition of pros-taglandin and leukotriene biosynthesis, 55 proinflammatory cytokine production, and inhibition of both ERK and NF-B pathways. 33 The antiinflammatory effect of 5-ASA has also been shown to be dependent on its interaction with the peroxisome proliferator-activated receptor gamma (PPAR␥), a member of the nuclear receptor superfamily. 34 Some of these effects are complementary; inhibition of ERK phosphorylation, for example, will result in reduced phosphorylation and hence reduced degradation of PPAR␥. 56 Here, we demonstrate that clinically and biologically relevant concentrations of 5-ASA inhibit the IL-8 release triggered by mucosal E. coli isolates. In contrast to 5-ASA, hydrocortisone was only able to abolish epithelial IL-8 release at 20-fold higher concentrations than are relevant clinically. This is consistent with reports that hydrocortisone only reduces cytokine levels at supratherapeutic concentrations in the respiratory epithe-FIGURE 7. A: Downregulation of NF-B by p65 siRNA reduced IL-8 release in response to representative CD and colon cancer mucosa-associated E. coli isolates HM427 and HM288 (both inhibited by Ͼ80% using Bay 11-7082). All experiments were conducted in triplicate. Significant differences from basal IL-8 release, ***P Ͻ 0.001; and from E. coli-induced IL-8 release, (***)P Ͻ 0.001 (ANOVA). B: Immunoblot showing downregulation of NF-B p65 subunit by siRNA. Equal loading was confirmed using pan actin and ERK2 monoclonal antibodies. p65 downregulation was calculated to be 62 Ϯ 18% (n ϭ 3; P Ͻ 0.0001, Student's t-test). lium in response to coronoviral infection. 57 Thus, the epithelial cell is the main target cell for 5-ASA therapy, whereas hydrocortisone exerts its effect mainly by acting on other cells such as T cells, 58 monocytes, 59 macrophages, 60 and dendritic cells. 61 It is also consistent with the fact that inhibition of NF-B activation, which is a principal target for corticosteroids, only had a minor effect on epithelial IL-8 response to mucosal E. coli in the current study. The potential role of adhesive and invasive E. coli in colon cancer is intriguing. There are similarities between the biology of sporadic colon cancer and IBD-associated cancer. 37 Epithelial activation of NF-B is critical in the development of inflammation-associated cancer, probably acting via suppression of apoptosis. 62 The adherent and adhesive mucosal E. coli isolates are able to invade colon cancer cell lines in vitro 4 and the present study also suggests that at least some of these bacteria are able to activate NF-B in these colon cancer cells. It is therefore reasonable to speculate that adherence to or invasion of dysplastic epithelial cells by such E. coli could have an important role in determining progression from dysplasia to cancer.

Inhibition of the mechanisms that are essential for the epithelial IL-8 response to mucosa-associated E. coli represents a logical approach to therapy in IBD, perhaps particularly in UC, where the inflammation is more superficial and where there is little evidence of bacterial invasion.

Animal models of inflammatory bowel disease: an overview

Presence of adherent Escherichia coli strains in ileal mucosa of patients with Crohn's disease

Mucosal flora in inflammatory bowel disease

Enhanced Escherichia coli adherence and invasion in Crohn's disease and colon cancer

Invasive E. coli strains are increased in Crohn's disease

High prevalence of Escherichia coli belonging to the B2ϩD phylogenetic group in inflammatory bowel disease

Abnormal microbiota composition in the ileocolonic mucosa of Crohn's disease patients as revealed by polymerase chain reaction-denaturing gradient gel electrophoresis

Molecular characterization of rectal mucosa-associated bacterial flora in inflammatory bowel disease

Temperature gradient gel electrophoresis of fecal 16S rRNA reveals active Escherichia coli in the microbiota of patients with ulcerative colitis

Adherent and invasive Escherichia coli is associated with granulomatous colitis in boxer dogs

Immunocytochemical evidence of Listeria, Escherichia coli, and Streptococcus antigens in Crohn's disease

Bacterial DNA within granulomas of patients with Crohn's disease-detection by laser capture microdissection and PCR

Molecular cloning of a human monocyte-derived neutrophil chemotactic factor (MDNCF) and the induction of MDNCF mRNA by interleukin 1 and tumor necrosis factor

M cells as ports of entry for enteroinvasive pathogens: mechanisms of interaction, consequences for the disease process

Pathogenesis of aphthoid ulcers in Crohn's disease: correlative findings by magnifying colonoscopy, electron microscopy, and immunohistochemistry

Rapid and simple determination of the Escherichia coli phylogenetic group

Carcinoembryonic antigenrelated cell adhesion molecule (CEACAM)-binding recombinant polypeptide confers protection against infection by respiratory and urogenital pathogens

Neutrophil serine proteinases cleave bacterial flagellin, abrogating its host response-inducing activity

Interaction between bacterial peptides, neutrophils and goblet cells: a possible mechanism for neutrophil recruitment and goblet cell depletion in colitis

Characterization of flagella produced by clinical strains of Stenotrophomonas maltophilia

The release of outer membrane vesicles from the strains of enterotoxigenic Escherichia coli

Salmonella typhimurium transcytoses flagellin via an SPI2-mediated vesicular transport pathway

Evasion of Toll-like receptor 5 by flagellated bacteria

Two nonadjacent regions in enteroaggregative Escherichia coli flagellin are required for activation of toll-like receptor 5

Toll-like receptor 5 recognizes a conserved site on flagellin required for protofilament formation and bacterial motility

Peanut lectin stimulates proliferation of colon cancer cells by interaction with glycosylated CD44v6 isoforms and consequential activation of c-Met and MAPK: functional implications for disease-associated glycosylation changes

Regulation of mitogen-activated protein kinase signaling networks by G protein-coupled receptors

Quantifying the activation of the small molecular weight G-protein Ras

Cytotoxic necrotizing factor type 1 delivered by outer membrane vesicles of uropathogenic Escherichia coli attenuates polymorphonuclear leukocyte antimicrobial activity and chemotaxis

Flagellin in combination with curli fimbriae elicits an immune response in the gastrointestinal epithelial cell line HT-29

Mechanisms of disease: defensins in gastrointestinal diseases

CEACAM6 acts as a receptor for adherent-invasive E. coli, supporting ileal mucosa colonization in Crohn disease

Mesalamine blocks tumor necrosis factor growth inhibition and nuclear factor kappaB activation in mouse colonocytes

Intestinal anti-inflammatory effect of 5-aminosalicylic acid is dependent on peroxisome proliferator-activated receptor-gamma

Prednisolone absorption in acute colitis

The fecal microbiota of irritable bowel syndrome patients differs significantly from that of healthy subjects

Inflammation and colorectal cancer: IBDassociated and sporadic cancer compared

Enteroaggregative Escherichia coli flagellin-induced interleukin-8 secretion requires Toll-like receptor 5-dependent p38 MAPkinase activation

Flagellin is the major proinflammatory determinant of enteropathogenic Salmonella

Flagellin of enteropathogenic Escherichia coli stimulates interleukin-8 production in T84 cells

Effects of nonpathogenic bacteria on cytokine secretion by human intestinal mucosa

In vitro and ex vivo activation of the TLR5 signaling pathway in intestinal epithelial cells by a commensal Escherichia coli strain

Cutting edge: bacterial flagellin activates basolaterally expressed TLR5 to induce epithelial proinflammatory gene expression

Differential alteration in intestinal epithelial cell expression of toll-like receptor 3 (TLR3) and TLR4 in inflammatory bowel disease

Pathophysiological role of Toll-like receptor 5 engagement by bacterial flagellin in colonic inflammation

Epithelial barrier defects in ulcerative colitis: characterization and quantification by electrophysiological imaging

Augmented increase in tight junction permeability by luminal stimuli in the non-inflamed ileum of Crohn's disease

Comparative study of the intestinal mucus barrier in normal and inflamed colon

Mitogen-activated protein (MAP) kinase signal transduction pathways and novel antiinflammatory targets

Role of EHEC O157:H7 virulence factors in the activation of intestinal epithelial cell NF-kappaB and MAPkinase pathways and the upregulated expression of interleukin 8

Implication of mitogen-activated protein kinases in T84 cell responses to enteropathogenic Escherichia coli infection

p38 mitogen-activated protein kinase is activated and linked to TNF-alpha signaling in inflammatory bowel disease

Inhibition of stressactivated MAPkinases induces clinical improvement in moderate to severe Crohn's disease

Mucosal 5-aminosalicylic acid concentration inversely correlates with severity of colonic inflammation in patients with ulcerative colitis

Role of prostaglandins in ulcerative colitis. Enhanced production during active disease and inhibition by sulfasalazine

Inhibition of HER-kinase activation prevents ERK-mediated degradation of PPARgamma

High-dose hydrocortisone reduces expression of the pro-inflammatory chemokines CXCL8 and CXCL10 in SARS coronavirus-infected intestinal cells

Novel mechanism for inhibition of human T cells by glucocorticoids. Glucocorticoids inhibit signal transduction through IL-2 receptor

Hydrocortisone regulation of interleukin-6 protein production by a purified population of human peripheral blood monocytes

Effects of a corticosteroid, budesonide, on alveolar macrophage and blood monocyte secretion of cytokines: differential sensitivity of GM-CSF, IL-1 beta, and IL-6

Corticosteroids inhibit the production of inflammatory mediators in immature monocyte-derived DC and induce the development of tolerogenic DC3

IKKbeta links inflammation and tumorigenesis in a mouse model of colitis-associated cancer

IL-8 Response to IBD and Colon Cancer E

The authors thank the staff of the University of Liverpool Physiological Laboratory, including Drs. I. Prior and C. Muncke (Biomedical Electron Microscopy Unit) and Dr. D. Hammond (who kindly provided GST-K85A RBD fusion protein). We also thank Dr. Lu-Gang Yu (School of Clinical Science) for invaluable advice on immunoblotting of phosphorylated proteins of the MAPK signal cascade.