INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Escherichia coli is a normal inhabitant of the intestinal tract but certain strains cause disease. Pathogenic E. coli belong to a restricted number of pathotypes defined by the presence of virulence factors which determine the host specificity and type of disease produced by these pathotypes (43, 62). The virulence mechanisms of E. coli strains are complex and only partially understood. They include the ability to colonize mucosal surfaces, invade extraintestinal tissues, survive and multiply in body fluids with low concentrations of available iron (58), and escape phagocytosis and intracellular killing by phagocytes (46). E. coli strains and/or their products modulate host cytokine responses (67). These cytokines, together with other inflammatory mediators are involved in the induction, persistence, or elimination of microbial infection (29, 70).

The production of cytokines during bacterial infection has been extensively studied in human septic shock (50). In this model, the release of endotoxin-lipopolysaccharide (LPS) triggers the synthesis of inflammatory cytokines such as tumor necrosis factor (TNF), interleukin-1 (IL-1), and IL-6. These cytokines induce many changes which result in the failure of the major organs and rapid death of the patient (50). In addition to LPS, other bacterial components have the capacity to induce cytokine production (for a review, see reference 72). Specific examples of pathogenic E. coli virulence factors that influence cytokine production include alpha-hemolysin, at nontoxic concentrations, which inhibits the production of TNF, IL-6, and IL-1 by human peripheral blood cells (38); an as-yet-unknown protein from enteropathogenic E. coli (EPEC) that inhibits IL-2, IL-4, IL-5, and gamma interferon (IFN-) expression by peripheral and mucosal mononuclear cells (37, 40); and Shiga-like toxin, which induces inflammatory cytokine production by murine macrophages (66). Adhesion to or invasion of epithelial cell monolayers by uropathogenic E. coli or EPEC also leads to the production of cytokines (19, 30, 59). Indeed, P fimbriae, which mediate attachment of uropathogenic E. coli to epithelial cells, enhance the host inflammatory response to infection and increase virulence (10, 31). Similarly, EPEC stimulate intestinal epithelial cell lines to produce IL-8 through the activation of NF-B (55).

Among the putative virulence factors produced by E. coli, cytotoxic necrotizing factors (CNFs) are produced by strains involved in diarrhea and septicemia in humans and in domestic animals (4, 6, 12). Necrotizing E. coli producing CNF1 have also been isolated from piglets with diarrhea and with edema disease (27) and have been associated clinically with lesions of polyserositis and septicemia in young pigs (22). CNF toxins are lethal when administrated intravenously to mice or sheep and are dermatonecrotic when inoculated into the rabbit skin (13-15). In addition, experimental oral inoculation of neonatal calves and pigs has shown that CNF-positive E. coli causes septicemia and enteritis (57, 73). S. Clément, B. Martineau-Doizé, I. P. Oswald, E. Oswald, M. Odin, and J. M. Fairbrother (submitted for publication) have also examined the dynamics of infection of CNF1-producing E. coli in experimentally inoculated conventional piglets of various ages and immune or weaning states. They demonstrated that CNF1-producing E. coli colonizes predominantly the large intestine and disseminates to mesenteric lymph nodes and internal organs, particularly in colostrum-deprived piglets. CNF1 and CNF2 are 110- to 115-kDa monomeric toxins that covalently interact with Rho (24, 48), resulting in its activation through the deamidation of a glutamine residue (25, 56). This activation of Rho GTPases results in polymerization of F actin, increased formation of stress fibers and multinucleation of cells (6, 23, 48). In addition to being implicated in the regulation of cytoskeletal structure, the Rho family of small GTP-binding proteins is also involved in the gene transcription and activation of the NF-B (41). Since this nuclear transcription factor plays a major role in the transcriptional regulation of many acute phase proteins and inflammatory cytokines (1, 2), we could anticipate that CNF induces cytokine synthesis.

Even if several lines of evidence implicate CNF1 and CNF2 in the pathogenesis of colibacillosis, their exact role still needs to be determined. Indeed, CNF1-producing E. coli strains often express other virulence factors (17, 34) and have also been found in the intestine of healthy piglets (27). Moreover, inactivation of the cnf1 gene in diarrhea-associated E. coli did not lead to a decrease in diarrhea and inflammation in a rabbit intestinal ligated loop model (20).

In the present study, we orally inoculated germfree colostrum-deprived newborn piglets with either E. coli M623, a wild-type CNF1-producing strain, an isogenic cnf1 derivative of M623, or a nonpathogenic E. coli strain. Our aim was to investigate the pathogenicity of a CNF1-producing strain in this model and to elucidate the role of CNF1 in bacterial pathogenicity and host cytokine response. In addition, we analyzed in vivo cytokine levels by reverse transcriptase PCR (RT-PCR) in the tissues of piglets. Herein, we demonstrated that M623 is pathogenic in germfree piglets and induces an inflammatory cytokine response in intestinal organs. Nevertheless, there were few differences at the level of pathogenicity, colonization, and cytokine levels elicited by M623 or its isogenic cnf1 mutant M623CNF1. Overall, our results suggest that CNF1 is not essential for pathogenicity, nor does it greatly influence the induction of host inflammatory cytokines by strain M623 in experimentally inoculated germfree piglets.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacterial strains, plasmids, and media. Bacterial strains and plasmids used throughout this study are summarized in Table 1. Wild-type E. coli 862B (O115:K) is nonpathogenic and serum sensitive and was isolated from the intestinal contents of a pig (21, 45). E. coli M623 (O2:K+), isolated from the intestine of a pig with enteritis (C. Wray, Central Veterinary Laboratory, Surrey, England) was used for inoculation studies. This strain is serum resistant and produces cytotoxin CNF1, alpha-hemolysin, and P and S fimbriae (17) but does not produce cytolethal distending toxin. Cosmid 2CO2 (also called cosmid 10) contains the cnf1 and hly gene clusters from uropathogenic strain E. coli J96 and was kindly supplied by D. E. Berg (64). Cloning vector pBluescript KSII⁺ was obtained from Stratagene (La Jolla, Calif.) and cloning vector pILL570 was kindly supplied by A. Labigne (39). Plasmid pKNG101 is a positive selection suicide vector containing strAB, sacBR and a pir-dependent R6K replicon (35). Plasmids were maintained in laboratory strain DH5 (54), except for suicide plasmids (pKNG101 and derivatives), which were maintained in SM10pir (32). Bacteria were isolated on Luria-Bertani (LB) agar and cultured in LB broth. Prior to piglet inoculation, strains were grown in tryptic soy broth (TSB). Media were supplemented with appropriate antibiotics at the following concentrations: ampicillin, 100 µg/ml; kanamycin, 50 µg/ml; streptomycin, 50 µg/ml; and spectinomycin, 100 µg/ml.

Recombinant DNA, genetic techniques, and nonpolar mutations in cnf1. Routine recombinant DNA techniques were performed by using standard procedures (54). A 13-kb EcoRI fragment from 2CO2 was cloned into pILL570. A 4.1-kb EcoRI/BamHI fragment bearing only cnf1 was subcloned into pBluescript KSII⁺. The aphT gene devoid of its transcription terminator was retrieved from pSB315 (26) by BamHI restriction and cloned at the BglII site within cnf1. The ApaI fragment bearing cnf1 disrupted by aphT was cloned into pKNG101 vector, resulting in pKNGcnf::aphT. Suicide plasmid pKNGcnf::aphT was introduced into strain M623 by conjugation. Mutants that had undergone allelic exchange leading to the replacement of the wild-type locus with the locus disrupted by aphT were selected on LB plates without NaCl and containing 5% sucrose and kanamycin, as previously described (35). Mutations were confirmed by Southern blots and cytotoxic assay as previously described (15).

Experimental inoculation of piglets. Piglets were delivered from four specific-pathogen-free Yorkshire hybrid gilts by Caesarian delivery. Piglets were immediately passed through an iodine bath, placed in germfree isolators, and fed condensed milk ad libitum, as previously described (45). We confirmed the absence of bacteria in the feces of piglets in isolators prior to inoculation. At 2 days of age, piglets received 10 ml of 1.2% NaHCO₃ through an intragastric tube to neutralize gastric acid. Piglets, chosen at random, were then similarly intubated with 1 ml of 0.9 to 2.1 × 10⁹ CFU in 19 ml of TSB of the wild-type parent strain M623 (n = 12) or with its isogenic cnf1 derivative M623CNF1 (n = 14) or with the nonpathogenic strain 862B (n = 2). Piglets were examined for mortality for up to 7 days postinoculation. An additional group of three uninoculated piglets served as controls for the determination of baseline cytokine mRNA levels.

Necropsy procedure. Piglets were killed by an intracardiac injection of Euthanyl Forte (sodium pentobarbital at 540 mg/ml diluted in 0.20 ml of propylene glycol; Pharmacie, Faculté de Médecine Vétérinaire, Saint-Hyacinthe, Québec, Canada) at 1 or 7 days postinoculation or if moribund.

Bacteriological counts. Tissues were evaluated quantitatively for the presence of E. coli. Samples were weighed and suspended in 2 ml of phosphate-buffered saline (PBS), homogenized at 5,000 rpm by using a Cat homogenizer x120 (PolyScience, Niles, Ill.), and 10-fold serially diluted in sterile PBS. Dilutions were plated on tryptic soy agar for the parental strain and on the same medium with kanamycin (50 µg/ml) for the mutant strain by using a Spiral Plater System Model C (Meyer Service and Supply Ltd., Long Sault, Ontario, Canada) as recommended by the manufacturer. After overnight incubation at 37°C, bacterial counts were determined. Several colonies from each individual were confirmed as being the infecting strain by PCR and agglutination tests.

Histopathology. Tissue samples were fixed in 10% neutral buffered formalin, embedded in paraffin, sectioned at 6 µm, and stained with hematoxylin and phloxin saffron for examination by light microscopy. Bacterial localization in intestinal and extraintestinal tissues was determined by immunocytochemistry. Sections were stained by Vector red (Vector Laboratories, Burlington, Ontario, Canada) as already described (52) by using rabbit polyclonal anti-O2 serogroup serum which was prepared as previously described (47).

RNA extraction. Samples from each organ, maintained in Trizol at 80°C, were homogenized by using a Cat homogenizer. Total RNA was extracted as recommended by the manufacturer. The RNA was resuspended in 50 to 500 µl of ultrapure water containing 0.02% (wt/vol) diethyl pyrocarbonate (Sigma, St. Quentin Fallavier, France) and 1 mM EDTA. Total RNA was quantified by using a spectrophotometer at an optical density of 260 nm (OD₂₆₀), and the purity was assessed by determining the OD₂₆₀/OD₂₈₀ ratio. All of the samples had an OD₂₆₀/OD₂₈₀ ratio above 1.8.

RT-PCR detection of cytokine mRNA and densitometric quantification of PCR products. An RT-PCR procedure was performed as previously described (18). Briefly, 1 µg of RNA was reverse transcribed (Superscript II RNase H; Life Technologies, Eragny, France) and then amplified (Taq DNA polymerase; Promega, Charbonnières, France). Primer sequences and the number of PCR cycles chosen for each cytokine are listed in Table 2. Amplified DNA was analyzed after electrophoresis on 1.2% TBE (Tris-borate-EDTA) agarose gels, which were stained with ethidium bromide and photographed with Polaroid 665 film. The level of each PCR product was quantified by densitometry by using Image Aquisition and Whole Band Analyzer software (Bioimage) on a Sun Sparc Station 5 (Cadrus, Ramonville St-Agne, France) as previously described (18). To compare the relative cytokine mRNA expression levels among samples, the values are presented as the ratio of the band intensity of the cytokine-specific RT-PCR product over that of the corresponding constitutively expressed "housekeeping" gene, cyclophilin.

Statistical analysis. Student's t test and/or analysis of variance were used to analyze bacterial counts and cytokine production. P values of <0.05 were considered significant. Macroscopic lesions were analyzed by use of the ² test. ² values of <3.84 were not considered significant.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Pathogenicity of E. coli M623 in piglets. We first investigated the pathogenicity of E. coli M623 in germfree, colostrum-deprived piglets inoculated by the oral route. Of 12 piglets, 2 demonstrated respiratory distress and died suddenly at less than 12 h postinoculation. At 24 h postinoculation, six inoculated piglets were randomly chosen and euthanized for necropsy; of the four remaining piglets, only one survived up to 7 days postinoculation. Table 3 summarizes the macroscopic lesions noted in the 10 piglets necropsied after death or after euthanasia. Most of the piglets inoculated with M623 demonstrated congestion of the lung, and half of these piglets presented intestinal congestion. Fluid and fibrin were observed in the thoracic, pericardial, and/or peritoneal cavities of all piglets examined at more than 24 h postinoculation but rarely in piglets examined at 24 h postinoculation. Microscopically, changes in piglets examined at 24 h postinoculation were minimal, the most important being inflammatory changes in the lung. These changes were characterized by a multifocal septal leukocytic infiltration composed of a mixed population of neutrophils and mononuclear cells and by an occasional fibrinous to leukocytic alveolitis. Bronchioalveolar necrosis was observed in one piglet (Table 4). Of note, no significant lesions were observed in uninoculated controls or piglets inoculated with strain 862B (data not shown).

View this table: [in this window] [in a new window]   TABLE 5.   Quantitative bacterial counts in tissues of piglets necropsied at 1 day postinoculation with E. coli M623 and M623CNF1

View larger version (53K): [in this window] [in a new window]   FIG. 1.   In situ visualization of bacteria in tissues by immunohistochemistry with an anti-O2 serum by Vector red staining. Piglets were inoculated with M623 and were euthanized 24 h postinoculation. (A) Clusters of bacteria are found extracellularly close to the hilus of mesenteric lymph node (arrow). Bar, 25 µm. (B) Bacteria are observed along the pulmonary epithelium (arrow facing right) and are occasionally found in the alveolar lumen, associated or not with alveolar macrophages (arrow facing left). A focal septal leukocytic infiltration composed of a mixed population of neutrophils and mononuclear cells can be seen (arrowhead). Bar, 10 µm.

Role of cnf1 on these effects. In order to understand the role of CNF1 in pig colibacillosis, a derivative of strain M623 unable to produce CNF1 (M623CNF1) was constructed by allelic exchange. As expected, bacterial lysates from M623CNF1 did not induce stress fibers and multinucleation in cultured HeLa cells, in contrast to lysates obtained from the wild-type M623 strain (data not shown).

E. coli M623 and M623CNF1 induce an enhanced production of inflammatory cytokines in the intestine. The ability of E. coli M623 and M623CNF1 to induce inflammatory cytokines such as IL-1, IL-6, IL-8, IL-12p40, and TNF- at the transcriptional level was then compared in piglets necropsied at 24 h postinoculation. Cytokine mRNA expression was measured by semiquantitative RT-PCR in samples from various organs. Since piglets were inoculated by the oral route, we first investigated the mRNA expression of inflammatory cytokines in the small intestine (ileum) and in the large intestine (colon). In the ileum, IL-1, TNF-, and IL-12p40 mRNA levels were significantly higher in piglets inoculated with M623 than in control piglets (Fig. 2). On the other hand, IL-8 mRNA production was only slightly higher in the ileum of M623-inoculated piglets than in uninoculated controls. In the colon, IL-1 and TNF- mRNA levels for M623-inoculated piglets were significantly greater than for control piglets, whereas IL-12p40 mRNA production was only weakly enhanced. IL-8 mRNA was not detected in the colon of any of the piglets. The expression of IL-6 was not detected in the two intestinal organs investigated. Expression of the different cytokines by a piglet inoculated with the strain 862B was similar to that observed in uninoculated piglets, except in the case of IL-1 in the colon (Fig. 2).

View larger version (22K): [in this window] [in a new window]   FIG. 2.   Inflammatory cytokine production in intestinal samples from inoculated piglets. Piglets were inoculated with M623 (wild type [WT]) or with its isogenic cnf derivative (M623CNF1) (CNF) and were euthanized 1 day postinoculation. Untreated piglets were used as controls (Cont.). Tissues from the ileum and colon were sampled and homogenized. Total RNA was then isolated and assayed for expression of inflammatory cytokine (IL-1, TNF-, IL-12p40, and IL-8) and the cyclophilin housekeeping genes by RT-PCR. Quantification of these cytokine mRNAs from the ileum and the colon in different animals (closed triangles) and the mean (± the standard error of the mean [SEM]) of these results for each group are shown (vertical bars). Opened triangles in the controls represent the values of piglets inoculated with nonpathogenic strain 862B. Student's t tests were performed to compare M623-inoculated piglets with control ones or M623CNF1-inoculated animals. *, P < 0.05; **, P < 0.01.

Expression of Th1 and Th2 cytokines by M623 and M623CNF1. In order to determine whether the immune response elicited by M623 and M623CNF1 was of the Th1 or Th2 type, we examined the production of a Th1 (IFN-) and a Th2 (IL-4) cytokine, by RT-PCR, in two immune organs (the spleen and the intestinal lymph nodes). As we only observed a weak production of these cytokines, probably due to the immaturity of the immune system of the piglets, we determined the frequency of detection of these two cytokines from controls and from piglets inoculated with M623 or M623CNF1 (Fig. 3). Control animals did not express these two cytokines in their intestinal lymph nodes. In inoculated animals, IFN- was not expressed in the spleen and was expressed in only a low proportion of the lymph nodes. By contrast, IL-4 was expressed in a much higher proportion (up to 100%) of spleen and intestinal lymph node samples from inoculated animals, suggesting that these piglets display a Th2 response.

View larger version (25K): [in this window] [in a new window]   FIG. 3.   Regulatory cytokine gene expression in immune organs from inoculated piglets. IL-4 and IFN- mRNA production by spleen and intestinal lymph nodes from uninoculated controls (Cont., ) or from M623 (M623, )- and M623CNF1 (CNF, )-inoculated piglets. Frequencies of detection of IL-4 and IFN- mRNAs are expressed as the percentage of positive samples observed by RT-PCR assays. Student's t tests were performed to compare M623-inoculated piglets with control ones or with M623CNF1-inoculated animals. *, P < 0.05; ND, not done.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

In this study, we demonstrated the pathogenicity of the CNF1-producing E. coli strain M623 in colostrum-deprived germfree piglets. This strain induced pulmonary interstitial and exudative inflammation and subsequently typical lesions of polyserositis which are observed in natural cases of E. coli septicemia in the pig (22, 46). E. coli isolates from such cases are often CNF1 positive. Rapid death (>24 h postinoculation) was observed in two piglets inoculated with the CNF1-producing E. coli strain and three piglets inoculated with the CNF1 mutant. These piglets demonstrated nonspecific lesions of pulmonary and intestinal congestion macroscopically, which may have been due to endotoxic shock rather than to trauma or injury due to the inoculation, since no evidence of inoculum aspiration was observed during the removal of the stomach tube and this phenomenon has not been observed in piglets similarly inoculated with other septicemic or nonpathogenic E. coli (45). Only one piglet examined at 24 h postinoculation with M623 demonstrated a pulmonary lesion, i.e., bronchoalveolar necrosis, which could have been related to the inoculation procedure with secondary aspiration of remaining bacteria during extubation. The bacteria not present at the site of inflammation may have already been eliminated by the inflammation. Hence we feel that, in general, this route was not a contributing factor in bacterial colonization of the lung.

Strain M623 colonizes intestinal and extraintestinal organs from at least 1 day postinoculation and persists up to 7 days postinoculation (Table 5). The colonization is predominantly in the intestine and particularly in the cecum and colon. These results are in agreement with the fact that the gastrointestinal tract would act as a reservoir for bacteria that can cause extraintestinal infections (63). Bacteria probably pass through the epithelial cells lining the intestines and are carried in the lymph to the mesenteric lymph node and possibly to the systemic complex (3), allowing bacterial establishment in extraintestinal organs. Of note, as already demonstrated (45), the nonpathogenic strain 862B was also able to colonize the intestine and to translocate into the mesenteric lymph nodes in colostrum-deprived germfree piglets. However, this strain did not persist in other extraintestinal organs nor did it induce any lesions or mortality. Hence, strain M623 possesses additional virulence attributes which may not be required for translocation from the intestine but which permit bacteria to persist and induce lesions in extraintestinal sites, at least in this model. In conventional, 2-day-old colostrum-deprived piglets (S. Clément, submitted), we also found that M623 was able to translocate to the draining lymph nodes and to the extraintestinal organs. However, in this model the strain did not induce any lesions or mortality. The difference in the pathogenesis observed in the two cases is probably due to the absence of intestinal barrier in germfree piglets. Indeed, newborn germfree piglets lack the ability to transfer maternal regulatory factors, both antigens and immunoglobulins, which have been shown to inhibit bacterial adherence to receptors on the intestinal epithelial cells and to neutralize the activity of the cytotoxins produced by E. coli (68). Using other CNF1-producing strains and higher doses of bacteria, Wray et al. (73) observed bacterial colonization of the intestine and extraintestinal organs, diarrhea, and respiratory signs in inoculated piglets, although clinical signs appeared to vary widely depending on the bacterial strain used and even within a group of animals given the same strain.

Several studies have investigated the action of CNF toxin in vitro. Based on these studies, the effect of CNF appears to be very different depending on the model used. For example, CNF1 increases intestinal permeability in Caco-2 cells (28), whereas it does not affect tight junction permeability in T84 monolayer (33). Similarly, this toxin induced a phagocytic behavior in human epithelial HEp-2 cells (23) but downmodulated integrin activation-dependent phagocytosis in human monocytes (5). In addition, CNF1 does not seem to have any effect on the ability of hemolytic E. coli to damage human bladder cell monolayers in vitro (34), but CNF1 effaced cell microvilli and decreased transepithelial migration of polymorphonuclear leukocytes on polarized T84 epithelial intestinal cell monolayers (33).

In the present study we investigated the in vivo role of CNF1 in the pathogenicity of colibacillosis. CNF1-positive E. coli are associated clinically with lesions of polyserositis and septicemia in young pigs (22). Initial experiments carried out in conventional piglets of different ages (10-day-old weaned or unweaned and 2-day-old colostrum fed or colostrum deprived) in an attempt to reproduce these lesions were not able to demonstrate any pathogenicity for E. coli M623 (Clément et al., submitted). Indeed, even in the presence of bacterial colonization of the intestine and dissemination to mesenteric lymph nodes and internal organs, we did not observe any mortality or significant lesions. Hence, we chose the more sensitive colostrum-deprived, germfree piglet, with which we have had considerable experience in the study of the pathogenesis of porcine E. coli septicemia (45), as our model for the study of the pathogenesis of M623 infection. We constructed a CNF1-isogenic mutant of a wild-type pathogenic E. coli strain, and we orally inoculated germfree neonatal piglets with either the M623 parental strain or the M623CNF1 mutant strain. We hypothesized that inactivation of cnf1 would decrease the ability of E. coli to colonize the intestine and/or to translocate to and cause lesions in internal organs in swine. However, inactivation of cnf1 did not reduce the pathogenicity of the wild-type M623 strain except for a weaker colonization of the ileum (Table 5), a slight decrease in mortality, and a slight decrease in inflammatory response in the lungs and serosal surfaces. Our results confirm and extend those of Elliot et al. (20), who used CNF1 mutants in a rabbit model of intestinal ligated loops and did not observe a significant effect on the onset, duration, or severity of diarrhea.

The inability to find significant differences in the pathogenicity of the M623 and the M623CNF1 can be interpreted in several different ways. First, CNF1 may play no role in bacterial pathogenicity and the fact that CNF-positive strains are pathogenic may result solely from a genetic linkage of the cnf1 gene with other virulence factors genes, such as those encoding alpha-hemolysin (hly) and P-related adhesin (prs) located on the same pathogenicity island on strain J96 (64) as well as strain M623 (E. Oswald, unpublished data). Second, our germfree piglet model may be insufficiently sensitive for the detection of the effect of CNF1. Moxley et al. found that inactivation of hemolysin did not reduce the incidence of septicemia in gnotobiotic piglets inoculated with isogenic enterotoxigenic E. coli strains after oral inoculation (42). Similarly, expression of heat-stable enterotoxin STb by adherent E. coli is not sufficient to cause severe diarrhea in neonatal pigs (7). Finally, other virulence factors may obscure the effects of CNF1. Indeed, strain M623 produces P and S fimbriae and hemolysin, which may contribute to the development of infection (17). In light of the results presented here, we tend to favor the last two hypotheses. Indeed, necrotoxigenic E. coli strains could be considered as pathogens of an opportunistic nature (Clément et al., submitted), and we have demonstrated that in vitro CNF potentiates the effect of another toxin, CDT (S. Pérès, F. Daigle, N. Ghichemerre, O. Marchés, F. Hérault, J. De Rycke, and E. Oswald, submitted for publication). Thus, using another infectious model, such as an immunocompromised piglet and/or E. coli strains that express other virulence factors, we would expect to increase the slight differences observed in the present study between isogenic strains expressing or not CNF1 toxin, in terms of mortality, inflammatory lesions, and cytokine response.

Cytokines are important in the regulation of the immune response and in the control of inflammation, but they can also contribute to immunopathological changes in the host after bacterial infection. In the current study, we investigated the cytokine response in both the intestinal tract and the immune tissues. In the spleen and intestinal draining lymph nodes, we particularly investigated the Th1-Th2 balance by measuring IFN- and IL-4. These two cytokines were weakly expressed, probably due to the age of the animals whose lymphoid organs were not totally developed. However, in inoculated animals the higher frequency of IL-4 mRNA detection in the lymphoid organs compared to that of IFN- argue in favor of a Th2 response following E. coli inoculation. This is in agreement with the shift toward the Th2-cell-type response observed in volunteers given a low dose of E. coli endotoxin (74) and may reflect the fact that E. coli, as do other extracellular bacteria, stimulates a stronger humoral than cellular immune response.

We also determined the cytokine response in intestinal tissues from control animals and from piglets inoculated with M623, M623CNF1, or 862B strains. In contrast to control uninoculated or 862B inoculated animals, piglets inoculated with M623 or M623CNF1 strains produced increased levels of mRNA encoding for inflammatory cytokines in their intestinal tract. Bacterial LPS did not seem to trigger this local synthesis of cytokines since strain 862B, which colonizes the intestine to the same extent as the two other strains, did not induce any inflammatory response in the ileum or in the colon (Fig. 2). Inoculation with either of the pathogenic strains M623 and M623CNF1 induces an inflammatory response in the intestinal tract, as measured by the production of IL-1 and TNF- (Fig. 2). These cytokines play a major role in the course of bacterial infection and in sepsis (16), and local induction of these inflammatory cytokines has been detected in murine models of pyelonephritis (36, 53) and epididymitis (65) induced by E. coli. Recombinant IL-1 and TNF have also been shown to increase the colonization of EPEC in the rabbit small bowel (71). Thus, the local induction of these inflammatory cytokines during oral inoculation by pathogenic E. coli may create a microenvironment that facilitates their own colonization of the intestine. Surprisingly, we did not find any increase in IL-6 and IL-8 mRNA levels in the intestine of inoculated piglets, although these cytokines have been detected in clinical patients or in experimental animals inoculated with E. coli (11, 61).

Similar responses were observed in piglets inoculated with either M623 or M623CNF1 E. coli strains, except in the ileum where animals inoculated with the mutant strain displayed lower inflammatory cytokine transcript levels than animals inoculated with the parental strain (Fig. 2). In the ileum, the bacterial colonization also differed between the two groups of animals (Table 5). We believe that the lower cytokine production reflects the lower bacterial colonization of this organ and is not directly related to a putative effect of CNF1 on NF-B (1, 2). This hypothesis is substantiated by the observation of Capo and coworkers (5), who did not find any specific inflammatory cytokine production in macrophages stimulated by purified CNF1 toxin. However, we cannot exclude an alternative explanation, i.e., that higher levels of IL-1 and TNF- expression in M623-inoculated piglets compared to M623CNF1-inoculated ones could facilitate bacterial colonization of the wild-type E. coli strain, as has already been demonstrated for EPEC (71).

Possible sources of the inflammatory cytokines induced by M623 strains include macrophages (9, 44), as well as dendritic cells (8, 69), and epithelial cells (19, 30, 60). The latter cells which line the intestine are continuously in contact with bacteria and their products and are known to play an active role in the mucosal immune system (19, 30). Of note, most of the inflammatory cytokine mRNAs that we investigated were detected by RT-PCR, although at low levels, in samples from control piglets (Fig. 2). This has been described in other studies and may reflect the dynamic nature of immune regulation even in the absence of microbial invasion (49, 51).

In conclusion, our results showed that the CNF1-producing strain M623 is pathogenic in germfree piglets and specifically induces the production of inflammatory cytokines. The CNF1 toxin does not seem to be the essential factor of virulence since the isogenic mutant is also pathogenic.