|Home | About | Journals | Submit | Contact Us | Français|
The mucosa-associated microflora is increasingly considered to play a pivotal role in the pathogenesis of inflammatory bowel disease. This study explored the possibility that an abnormal mucosal flora is involved in the etiopathogenesis of granulomatous colitis of Boxer dogs (GCB). Colonic biopsy samples from affected dogs (n = 13) and controls (n = 38) were examined by fluorescent in situ hybridization (FISH) with a eubacterial 16S rRNA probe. Culture, 16S ribosomal DNA sequencing, and histochemistry were used to guide subsequent FISH. GCB-associated Escherichia coli isolates were evaluated for their ability to invade and persist in cultured epithelial cells and macrophages as well as for serotype, phylogenetic group, genome size, overall genotype, and presence of virulence genes. Intramucosal gram-negative coccobacilli were present in 100% of GCB samples but not controls. Invasive bacteria hybridized with FISH probes to E. coli. Three of four GCB-associated E. coli isolates adhered to, invaded, and replicated within cultured epithelial cells. Invasion triggered a “splash”-type response, was decreased by cytochalasin D, genistein, colchicine, and wortmannin, and paralleled the behavior of the Crohn's disease-associated strain E. coli LF 82. GCB E. coli and LF 82 were diverse in serotype and overall genotype but similar in phylogeny (B2 and D), in virulence gene profiles (fyuA, irp1, irp2, chuA, fepC, ibeA, kpsMII, iss), in having a larger genome size than commensal E. coli, and in the presence of novel multilocus sequence types. We conclude that GCB is associated with selective intramucosal colonization by E. coli. E. coli strains associated with GCB and Crohn's disease have an adherent and invasive phenotype and novel multilocus sequence types and resemble E. coli associated with extraintestinal disease in phylogeny and virulence gene profile.
There is mounting evidence that inflammatory bowel disease (IBD) is a consequence of an overly aggressive immune response to luminal commensal bacteria in genetically susceptible individuals (63, 70). Mechanistic studies of the interplay between the intestinal mucosa and bacteria in rodents with engineered susceptibility, e.g., interleukin 10-negative (IL-10−/−) and IL-2−/− mice and HLA-B27 β2-microglobulin transgenic rats, indicate that the enteric microflora is required for inflammation, with inflammation being observed in conventional housing but not under germfree conditions (59, 72, 73). In these models, bacterial species regarded as part of the normal intestinal flora, such as Bacteroides and members of the family Enterobacteriaceae, including E. coli, have been associated with inflammation (39, 59, 70, 72).
Evidence implicating the resident enteric microflora in the pathogenesis of spontaneous IBD in people is provided by the increased immune responses to enteric commensal bacteria observed in both Crohn's disease (CD) and ulcerative colitis (UC) (1, 24, 45, 52, 54), the clinical responses of CD and UC to antimicrobials (9, 10, 48, 58, 70) and the response of CD to fecal stream diversion (68). The discovery of genetic defects in the microflora-sensing ability of patients with CD and UC, such as NOD2/CARD15 and TLR-4, respectively, provides mechanisms to explain individual susceptibility to the resident microflora (33, 46). Insights into the bacterial factors that may drive the inflammatory response are provided by 16S ribosomal DNA (rDNA) sequencing and fluorescent in situ hybridization (FISH), which reveal increased numbers, but decreased diversity, of the mucosa-associated bacteria in patients with IBD (40, 60, 76). Disease-specific increases in colonization and invasion by the gamma subdivision of Proteobacteria, Enterobacteriaceae, and Bacteroides/Prevotella were found in CD, and these bacteria plus Clostridium spp., high-G+C gram-positive bacteria, and sulfate-reducing bacteria were found in UC (40). These observations are complemented by culture-based reports documenting a higher prevalence of mucosa-associated E. coli, which adheres to and invades cultured cells but lacks the invasive virulence genes of diarrheagenic strains, in CD than in UC and normal tissue (17, 18, 49). These findings help to link experimental and clinical observations and support a broad hypothesis that idiopathic inflammatory bowel disease is caused by an overly aggressive immune response to luminal commensal bacteria in genetically susceptible individuals (63, 70). However, this hypothesis has to be reconciled with studies reporting the association of a diverse spectrum of pathogenic bacteria, including Mycobacterium avium subsp. paratuberculosis, Listeria, Streptococcus, Enterococcus, Enterobacteriaceae, Bacteroides, Clostridium, and Yersinia spp., with IBD and should be rigorously tested in spontaneous idiopathic IBD (8, 11, 16, 43, 44, 48).
It is against this background that we sought to examine the role of the mucosa-associated flora in granulomatous colitis of Boxer dogs (GCB: also known as histiocytic ulcerative colitis), a severe disease of unknown etiology that typically affects Boxer dogs under four years of age and is characterized by frequent bloody mucoid stools, thickening and ulceration of the colon, anemia, hypoalbuminemia, and weight loss (80). The dominant histological features are loss of colonic epithelium and goblet cells and the accumulation of large numbers of periodic acid-Schiff stain (PAS)-positive macrophages (20, 25, 77, 80). Immunopathological studies describe an increase in immunoglobulin G3 (IgG3) and IgG4 plasma cells, CD3 T cells, and L1- and major histocompatibility complex II-positive cells (25). Several studies have described bacteria within the mucosa of affected dogs, but known enteropathogens such as Salmonella, Campylobacter Yersinia, and Shigella have not been isolated (28, 32, 78, 80). Ultrastructural studies suggest active phagocytosis of bacteria that in some instances resemble Chlamydia (78). An attempt to reproduce colitis in Boxer dogs with Mycoplasma isolated from the colon and regional lymph nodes of four affected dogs was unsuccessful (6). The predilection for Boxer dogs, with only sporadic cases of this type of colitis reported in non-Boxer dogs, and the absence of a causal infectious agent have led to GCB being considered a breed-specific, immune-mediated disease of unknown etiology (25, 75). However, a favorable outcome has been described in dogs receiving antibiotics such as chloramphenicol (80), and recent reports describe clinical responses to antibiotic regimens containing fluoroquinolones (19, 32) that have also been effective in some patients with Crohn's colitis and ileocolitis (10, 70). Thus, GCB represents a spontaneous idiopathic inflammatory bowel disease with features in common with ulcerative colitis (macroscopic appearance, regional distribution, immunopathology), Crohn's disease (granulomatous inflammation, bacteria within macrophages, response to fluoroquinolones) and Whipple's disease (PAS-positive macrophages, bacteria within macrophages), but it is not identical to any one of these diseases (20, 25, 77, 80).
The present study directly explored the possibility that an uncharacterized infectious agent such as Trophyrema whipplei (20, 66) or an abnormal mucosa-associated flora is involved in the etiopathogenesis of granulomatous colitis of Boxer dogs (GCB). A combination of 16S rDNA sequencing and fluorescence in situ hybridization revealed selective intramucosal colonization of GCB biopsy samples by Escherichia coli. E. coli isolated from the colonic mucosa of affected dogs adhered to, invaded, and persisted in cultured epithelial cells to the same degree as the Crohn's disease-associated strain E. coli LF 82. Invasion of cultured epithelial cells by GCB isolates and LF 82 is consistent with triggered endocytosis and involves the host cytoskeleton and signaling pathways. We determined that GCB-associated E. coli and LF 82 are more similar in phylogeny and virulence gene profiles to extraintestinal pathogenic E. coli than diarrheagenic E. coli. Our findings support the thesis that IBD is a consequence of mucosal colonization by a restricted subset of the luminal microflora in a susceptible individual and point to the association of E. coli that resembles extraintestinal pathogenic strains in genotype with chronic intestinal inflammation.
Formalin-fixed paraffin-embedded colonic biopsy samples from 13 Boxer dogs with a diagnosis of granulomatous colitis (age, 1 to 5 years [median, 1.5 years]; 10 females, 3 males) and control tissues from 27 dogs with other forms of colitis (non-GCB; 7 months to 13 years [median, 5 years]; 10 females, 17 males; 5 Boxers, 4 mongrels, 2 German Shepherd Dogs, 2 Border Collies, 2 Doberman Pinschers, and 12 other breeds), and 11 dogs without colonic inflammation (6 months to 11 years [median, approximately 5 years]; 3 females, 8 males: 5 mongrels, 2 German Shepherd Dogs, and 4 other breeds) were sectioned (5 μm), stained with hematoxylin and eosin (H&E) and PAS, and examined by a pathologist (S.P.M.) who was blinded to the origin of the section.
All GCB sections were characterized by the loss of colonic glands, mucosal erosion and ulceration, and the presence of PAS+ macrophages (Fig. (Fig.1).1). In the non-GCB group, colitis was predominantly lymphoplasmacytic (26/27 dogs; 1 Irish Setter had eosinophilic colitis), and nonulcerative (24/27 dogs), with PAS-positive macrophages absent or rarely detected. The severity of colitis in the non-GCB group, determined by the presence of ulcers or erosions, architectural changes such as glandular atrophy or dysplasia, alterations in goblet cells, and increases in cellularity, ranged from severe (n = 2) through moderate (n = 13) to mild (n = 12) (67). Aphthous ulcers were observed in two dogs with mild colitis (one roughCollie, one German Shepherd Dog). Mucosal ulceration was only observed in biopsy samples of a 13-year-old Cocker Spaniel with severe colitis.
Fresh colonic tissue was obtained from an additional two Boxer dogs (12- to 13-month-old spayed females with a 9- to 10-month history of bloody diarrhea and tenesmus) during diagnostic endoscopy. Fecal analysis was negative for endoparasites, Giardia, Campylobacter, Salmonella, Yersinia, and Shigella. The diarrhea was unresponsive to dietary modification and metronidazole in both dogs. Additional treatment with sulfasalazine and tylosin in one dog and enrofloxacin in the other produced no response. Both dogs had mild microcytic anemia (hematocrit, 36 and 37%; normal, 40 to 48%) and a mean corpuscular volume of 57 and 61fl (normal, 63 to 73 fl), and one dog had a low serum iron and iron saturation suggestive of iron deficiency anemia. Abdominal ultrasound revealed mesenteric lymphadenopathy in one dog and a thickened colon in the other. Colonoscopy (Fig. (Fig.1A)1A) showed irregular thickened ulcerated mucosa in both dogs. Colonic biopsy samples were collected into sterile tubes for microbial culture and DNA extraction and into formal saline for histopathology and in situ hybridization. Histopathological findings in both dogs were consistent with GCB.
One dog showed a dramatic clinical response to enrofloxacin (9 mg/kg of body weight orally [p.o.] once daily [SID] for 30 days and then 6 mg/kg p.o. SID for 14 days and 3 mg/kg p.o. SID for 7 days), with resolution of diarrhea and clinicopathological abnormalities and with an increase in body weight (+2 kg). A follow-up colonoscopy was performed 5 months after the initial procedure, 3 months after the end of antibiotic treatment.
Formalin-fixed paraffin-embedded histological sections (4 μm) (15 GCB, 27 non-GCB colitis, and 11 normal samples) were mounted on Probe-On Plus slides (Fisher Scientific, Pittsburgh, Pa.) and evaluated by FISH with a eubacterial probe (EUB-338; GCTGCCTCCCGTAGGAGT) as previously described (65). Briefly, paraffin-embedded biopsy specimens were deparaffinized by passage through xylene (three times, 20 min each), 100% alcohol (20 min), 95% ethanol (20 min) and finally 70% ethanol (20 min). The slides were air-dried. FISH probes 5′ labeled with Cy3 or 6-FAM (Integrated DNA Technologies, Coralville, IA) were reconstituted with sterile water and then diluted to a working concentration of 5 ng μl−1 with hybridization buffer (20 mM Tris-HCl, 0.1% sodium dodecyl sulfate [SDS], 0.9% NaCl [pH 7.2]). The sections were allowed to hybridize with 30 μl of DNA probe mix in a hybridization chamber at 46°C for 4 h. Slides were washed in wash buffer (hybridization buffer without SDS) at 48°C for 30 min. Hybridized samples were washed in phosphate-buffered saline (PBS), allowed to air dry, and mounted with a ProLong antifade kit (Molecular Probes Inc., Eugene, OR). Sections were examined on an Axioskop 2 (Carl Zeiss Inc., Thornwood, NY) or a BX51 (Olympus America, Melville, NY) epifluorescence microscope, and images were captured with a Zeiss Axiocam or Olympus DP-7 camera, respectively.
Slides spotted with suspensions of cultured E. coli DH5α, Shigella sonnei (ATCC 25931), Salmonella enterica serovar Typhimurium (ATCC 14028), and clinical isolates of Yersinia enterocolitica, Proteus vulgaris, Klebsiella pneumoniae, Bacteroides fragilis, Pseudomonas aeruginosa, Enterococcus faecium, Streptococcus equi, Streptococcus bovis, Clostridium perfringens, Clostridium difficile, Lactobacillus plantarum, Listeria monocytogenes, and Helicobacter pylori were used to control probe specificity. Probe specificity was additionally evaluated using tissue sections and bacteria treated with RNase and using the irrelevant probe non-EUB-338 (ACTCCTACGGGAGGCAGC).
Colonic sections with evidence of bacterial invasion on eubacterial FISH were subsequently evaluated by Gram, Ziehl-Nielsen, and Steiner stains and FISH probes directed against a subset of the Enterobacteriaceae (E. coli, Shigella, Salmonella, and Klebsiella; E. coli 1531 23S rRNA, CATGAATCACAAAGTGGTAAGCGCC) (64) and E. coli/Shigella (E. coli 16S rRNA, GCAAAGGTATTAACTTTACTCCC) (34).
To aid localization of bacteria within the mucosa, representative sections were stained with a monoclonal antibody to vimentin (clone-V9 [Sigma], 1:50 with 2.5% bovine serum albumin [BSA] in PBS, 1 h) after the fluorescent in situ hybridization procedure. Incubation with the primary antibody was followed by washing (three times, 10 min each, in PBS), incubation with secondary antibodies (chicken anti-mouse immunoglobulin, Alexafluor488 [Molecular Probes]; 1:50 in 2.5% BSA in PBS for 1 h), washing (three times, 10 min each, in PBS), and mounting with antifade.
DNA was extracted from mucosal biopsy samples of two Boxer dogs with GC and from one of these dogs after remission induced by enrofloxacin (see patient details above) using a QIAamp DNA minikit (QIAGEN Inc., Valencia, CA) according to the manufacturer's instructions. A 320-bp fragment of the bacterial 16S gene was amplified from 5 μl of this extract using the primers 16SFa (5′ GCTCAGATTGAACGCTGG), 16SFb (5′ CTCAGGAYGAACGCTGG), and 16SR (5′ TACTGCTGCCTCCCGTA) (30). Cycling parameters were 94°C for 3 min followed by 26 cycles of 94°C for 30s, 63°C for 1 min, and 72°C for 1 min. A final extension was carried out at 72°C for 5 min. To remove potential contaminating DNA, sterile distilled PCR water was filtered through Micron YM-100 filters (Millipore, Bedford, MA) and exposed to UV irradiation for 10 min. Gel-purified PCR amplicons were extracted using a Perfectprep gel cleanup kit (Eppendorf, Westbury, NY) according to the manufacturer's instructions. Representative 16S rDNA libraries were established on the basis of the 16S rDNA fragments amplified from nucleic acid extracts. PCR products were cloned into a TA cloning vector (pGEM-T Easy; Promega Corp., Madison, WI) according to the manufacturer's instructions. Candidate clones were screened by restriction digestion, with clones containing the correct-sized fragments further screened by PCR. Clones with positive PCR results were sequenced at the Cornell University BioResource Center using M13 primers and an ABI 3700 automated DNA sequencer and ABI Prism BigDye terminator sequencing kits with AmpliTaq DNA polymerase (Applied Biosystems, Foster City, CA). DNA sequences obtained with both forward and reverse primers were proofread and then assembled in Editseq (DNAStar, Madison, WI). Partial 16S rDNA sequences were compared to the sequence databases at NCBI (BLAST-n) and the Ribosomal Database Project (RDPII: http://rdp.cme.msu.edu/), with representative sequences of high homology imported into our database. Sequences were aligned using the Clustal-W algorithm in MegAlign (DNAStar, Madison, WI), and data for each biopsy were summarized as a phylogenetic tree.
Colonic biopsy samples from 2 Boxer dogs with GC were ground in sterile saline with a sterile pestle. Half of the homogenate was incubated in GN broth (BBL, Becton Dickinson, Franklin Lakes, NJ) at 35°C overnight and then subcultured onto Levine EMB (BBL), while the remainder was directly plated onto Trypticase soy agar (5% sheep blood) (BBL), Levine EMB agar (BBL), and Columbia CNA agar (5% sheep blood) (BBL). All plates were incubated at 35°C in 6% CO2 for 18 to 24 h. Suspect bacterial colonies were then identified using a Sensititre system (Trek, Westlake, OH). Bacterial isolates were stored in glycerol broth at −70°C. DNA was extracted from all isolates, PCR amplified with 16S primers, cloned, and sequenced, with sequences incorporated into the phylogenetic analysis of each biopsy.
E. coli strains isolated from GCB mucosa were evaluated for their ability to adhere to, invade, and persist within cultured epithelial cells. E. coli strain LF 82, isolated from a chronic ileal lesion of a patient with CD (kindly provided by A. Darfeuille-Michaud) (17), and Salmonella enterica serovar Typhimurium (ATCC 14028) were used as positive controls. E. coli DH5α, a nonpathogenic strain, was used as a negative control. Bacterial isolates were stored at −80°C, and fresh nonpassaged bacteria were used for all investigations. E. coli and S. enterica serovar Typhimurium were streaked on Luria-Bertani (LB) agar, and a single colony was inoculated into LB broth. Cells were grown overnight at 37°C without shaking. Type 1 pilus expression was confirmed by mannose-sensitive agglutination of 1% commercial baker's yeast (Saccharomyces cerevisiae) suspended in phosphate-buffered saline (PBS).
It has been shown previously that human CD-associated E. coli strains are able to efficiently invade a wide variety of epithelial cell lines, including Caco-2, Intestine-407, HCT-8, and HEp-2 (5, 16). Preliminary studies in our laboratory with strain LF 82 and Salmonella enterica serovar Typhimurium confirmed adhesion and invasion of cultured Caco-2 (kindly provided by Andrea Quaroni) and Hep2 (ATCC CCL-23) cells. Strain LF 82 and Salmonella enterica serovar Typhimurium also displayed adherent and invasive behavior with cells of the bovine mammary epithelial cell line MAC-T (Nexia Biotechnologies Inc., Montreal, QC, Canada).
Monolayers of all cell lines were kept at 37°C in 5% CO2-95% air (vol/vol) using Dulbecco's modified Eagle's medium (DMEM; Sigma-Aldrich, St. Louis, MO) supplemented with 10% fetal bovine serum (FBS; Gemini Bio Products, Woodland, CA) for MAC-T and Caco-2 cells. HEp-2 cells were grown in RPMI 1640 (Gibco-Invitrogen, Grand Island, NY) supplemented with 10% FBS. The FBS concentration was dropped to 5% before infection assays.
Stationary-phase bacteria were pelleted, washed with excess phosphate-buffered saline (PBS, pH 7.4), and resuspended in PBS. Initial bacterial numbers were determined simultaneously by plate count. Confluent monolayers of cells grown in 24-well plates (3 × 105 cells/well) were washed twice with 1 ml/well of growth medium, and then 1 ml of fresh medium was added to each well. Epithelial cells were infected with a multiplicity of infection (MOI) of 10 bacteria per epithelial cell. After 1 h of incubation at 37οC with 5% CO2, cells were washed three times in PBS and lysed in 0.1% Triton X-100 in PBS for 10 min. Lysates were serially diluted and plated on LB agar plates, and colonies were enumerated following overnight incubation. The levels of adhesion were expressed as the total number of CFU recovered per well. Each assay was run in duplicate and repeated at least three times.
Invasion of MAC-T cells was evaluated by the gentamicin protection assay, differential immunostaining of intra- and extra-cellular bacteria, and electron microscopy.
For the gentamicin protection assay, cells were grown in six-well plates for 2 days (~1.5 × 106) and infected with E. coli strains at an MOI of 10 for 1 h. After the initial 1-h infection period, cells were washed three times in PBS and then incubated for another 2 h in medium containing 100 μg/ml gentamicin to kill extracellular bacteria. The MIC of gentamicin for all strains was 0.5 μg/ml. Cells were then washed three times with PBS, lysed, and plated as described above. The levels of invasion are expressed as the total number of CFU recovered per well. The viability of epithelial cells before and after infection was assessed by trypan blue exclusion.
Invasion of epithelial cells was also examined using the Caco-2 and HEp-2 cell lines. Caco-2 and HEp-2 cells were grown in six-well plates for 2 days (~1.2 × 106) and infected with E. coli strains at an MOI of 100 for 3 h. Cells were then washed three times with PBS and incubated for an additional 1 h in medium containing 100 μg/ml gentamicin to kill extracellular bacteria. Numbers of intracellular bacteria and the levels of invasion were determined as described above.
The involvement of host cytoskeletal and signaling molecules in the invasion process was examined using the microfilament inhibitor cytochalasin D, the microtubule inhibitor colchicine, the tyrosine protein kinase inhibitor genistein, and phosphoinositide 3-kinase (PI 3-kinase) inhibitor wortmannin. Stock solutions were prepared in dimethyl sulfoxide (DMSO), except for colchicine, which was prepared in distilled H2O: cytochalasin D, 1 mg/ml; colchicine, 1 mg/ml; genistein, 100 mM, and wortmannin, 0.1 mM. Prior to use, the inhibitors were diluted in DMEM supplemented with 10% fetal bovine serum. Cytochalasin D (1 μg/ml) and colchicine (5 μg/ml) were added 1 h prior to addition of bacteria. Genistein (100 μM) and wortmannin (100 nM) were added 30 and 10 min before infection, respectively. DMSO (0.1% final concentration) was used as a control. Cells were infected at an MOI of 100. The inhibitors and DMSO were maintained throughout the 1-h invasion period. After the 1-h infection and a 2-h gentamicin killing period, the numbers of intracellular bacteria were determined as described above. Results are reported as the percentage of the number of bacteria that were internalized in control cells with no inhibitor. The viability of epithelial cells was determined in uninfected and infected epithelial cells in the presence and absence of inhibitors.
MAC-T cells seeded and grown overnight on glass eight-well chamber slides were infected with E. coli at an MOI of 100. After 3 h of incubation, the cells were washed three times in PBS and fixed with 3.7% formaldehyde for 15 min. PBS containing 10% fetal bovine serum (FBS) was used as blocking buffer and for dilution of immunoreagents. Polyclonal rabbit anti-E. coli antibody (B65001R [Biodesign, Saco, ME]; 1:50 in PBS with 10% FBS) was used as the primary antibody. All incubations were carried out at room temperature. For differential immunostaining of extra- and intracellular E. coli, the fixed cells were incubated with blocking buffer for 15 min to reduce nonspecific binding of antibodies. Thereafter, the monolayer was incubated with primary antibody for 1 h at room temperature. After washing three times for 10 min each in PBS, monolayers were incubated in the dark with fluorescein isothiocyanate (FITC)-conjugated goat anti-rabbit IgG antibodies (F0382 [Sigma-Aldrich, St. Louis, MO]; 1:100 in PBS with 10% FBS) for 45 min. Unbound secondary antibody was removed by washing three times for 10 min each in PBS. Cells were permeabilized with PBS containing 0.1% Triton X-100 for 15 min. After permeabilization, the monolayers were incubated with blocking buffer for 15 min and then with primary antibody for 1 h. After washing three times for 10 min each in PBS, treated monolayers were incubated in the dark with tetramethyl rhodamine isothiocyanate (TRITC)-conjugated goat anti-rabbit IgG antibodies (T677 [Sigma-Aldrich]; 1:100 in PBS with 10% FBS) for 45 min. Unbound secondary antibody was removed by washing three times for 10 min each in PBS. Slides were mounted with ProLong antifade (Molecular Probes Inc., Eugene, OR) mixed with DAPI (4′,6′-diamidino-2-phenylindole) and examined with an Olympus BX51 epifluorescence microscope.
MAC-T cells grown on 24-mm collagen Transwell inserts (Costar, Corning Inc., Corning, NY) and infected at an MOI of 100 were maintained in culture for 3 h. Culture supernatants were removed, and cells were washed three times in culture media before being fixed in 2% glutaraldehyde in PBS, postfixed in 1% osmium tetroxide, dehydrated through ethanol, and embedded in Spurr's resin. Thin sections were cut, stained with uranyl acetate and Reynolds' lead, and examined in a Technai electron microscope.
The invasion assay was modified by incubating infected monolayers up to 48 h. After the 1-h invasion and 2-h incubation with 100 μg/ml of gentamicin, cells were washed once in PBS, and fresh medium containing 15 μg of gentamicin/ml was added to the cells. This concentration of gentamicin was used to prevent extracellular bacterial growth while reducing the chances of gentamicin leaching into the host cells during longer incubation times. Numbers ofintracellular bacteria were determined at 2, 24, and 48 h as described above.
Macrophages derived from the bone marrow of C57BL/6 mice were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% FCS, 5% horse serum, 2 mM l-glutamine, 1 mM sodium pyruvate, and 20% L-cell conditioned media. Fully differentiated macrophages were transferred to coverslips in 24-well plates and incubated for 12 h to allow a confluent monolayer to establish (cell density of 4.2 × 105, in 1 ml of medium). Before infection, the cell monolayers were washed twice with PBS, and the medium was replaced with 1 ml of infection media. Macrophages were infected at a multiplicity of infection (MOI) of 10 bacteria per macrophage, and the plates were centrifuged at 500 × g for 5 min. After a 30-min incubation at 37°C with 5% CO2, infected macrophages were washed twice with PBS, and fresh cell culture medium containing 100 μg of gentamicin/ml was added to kill extracellular bacteria. After incubation for an additional 2 h, the medium was removed and fresh medium containing 15 μg of gentamicin/ml was added for longer postinfection periods. To measure intracellular survival beyond 48 h postinfection, fresh cell culture medium containing gentamicin (15 μg/ml) was added to the infected cells. At specified time points, coverslips were washed three times with PBS, and cells were lysed in 2 ml of 0.1% Triton X-100 (Sigma-Aldrich) in PBS. Samples were removed, diluted, and plated onto LB agar plates to determine the number of CFU/well. The number of bacteria surviving gentamicin was determined at 2, 4, 8, 24, 48, and 72 h. Survival was expressed as CFU per well, and the mean percentage of the number of bacteria recovered after 2 h postinfection was defined as 100%. Each experiment was performed in triplicate and repeated at least three times.
The genetic diversity of E. coli isolated from GCB mucosa and the Crohn's disease isolate LF82 was evaluated by randomly amplified polymorphic DNA (RAPD)-PCR with informative primers 1254, 1281, and 1283 as previously described (82).
The triplex PCR described by Clermont et al. (12) was performed to determine the phylogenetic group (A, B1, B2, or D) of these strains, with E. coli strains ECOR-03, -25, -26, -34, -48, -50, -62 and -64 (ECOR collection: http:foodsafe.msu.edu/whittam/ecor/index.html) used as controls.
Multilocus sequence typing (MLST) for seven loci (aspC, clpX, fadD, icdA, lysP, mdh, and uidA) was performed according to the protocol established by Whittam et al. (http://www.shigatox.net/stec/mlst-new/mlst_pcr.html). Gel-purified PCR amplicons were sequenced at the Cornell University BioResource Center using M13 primers and an ABI 3700 automated DNA sequencer and ABI PRISM BigDye terminator sequencing kits with AmpliTaq DNA polymerase (Applied Biosystems, Foster City, CA). DNA sequences obtained with both forward and reverse primers were proofread and then assembled in Editseq (DNAStar, Madison, WI). Sequences were aligned using the Clustal-W algorithm in MegAlign (DNAStar, Madison WI), and the allele and the st7 and clonal groups were determined using web-based software (http://www.shigatox.net/cgi-bin/mlst7/dbquery).
Serotyping and a PCR-based screen for virulence genes of diarrheagenic E. coli (heat-labile toxin [LT], heat stable toxins [STa and STb], Shiga-like toxins [SLT-I and SLT-II], cytotoxic necrotizing factors [CNF-1 and CNF-2], and the gamma variant of the intimin-encoding gene [eae]) were performed at the Gastroenteric Disease Center, Pennsylvania State University (College Station, PA). The presence of the ipaH gene (a marker for Shigella and enteroinvasive E. coli) in cultured isolates and colonic DNA was also determined by PCR as previously described (49).
The possibility that GCB- and CD-associated E. coli contain common and potentially pathoadaptive alleles of the polymorphic adhesin FimH, which is associated with invasion in uropathogenic E. coli (50, 74), was determined by PCR and sequencing. Briefly, bacterial lysates were prepared with a DNeasy tissue kit according to the manufacturer's instructions. Oligonucleotide primers (fimH-F [5′-CAG GGA ACC ATT CAG GCA GTG ATT AGC ATC-3′] and fimH-R [5′-AAT ATT GCG TAC CAG CAT TAG C-3′]) were designed from the published sequence for the fimH gene in E. coli K-12 strain MG1655 (GenBank accession number NC_000913) using the Primer Select software program (DNAStar). PCR was performed in a Bio-Rad MyCycler automatic thermal cycler with denaturation at 96°C for 1 min, annealing at 55°C for 1 min, and extension at 72°C for 2 min, for a total of 40 cycles. PCR products were cloned, sequenced, and aligned as described above.
The genome sizes of E. coli isolated from GCB mucosa and the Crohn's disease isolate LF 82 were determined by pulsed-field gel electrophoresis (PFGE). Genomic DNA for PFGE was prepared in agarose plugs and cleaved with the restriction enzyme I-CeuI. PFGE was performed with a Chef Mapper (Bio-Rad Laboratories, Hercules, CA) in 0.5× Tris-borate-EDTA buffer at 6 V/cm and 14°C. Electrophoresis was carried out with switch times of 2.16 s to 54.17 s for 20 h. Images were acquired with a Bio-Rad Gel Doc (Bio-Rad Laboratories), and chromosomal sizes were determined with KODAK 1D image analysis software.
A comprehensive screening of a complete set of virulence genes of diarrheagenic and extraintestinal pathogenic E. coli strains was performed with the DNA microarray recently developed by Bruant et al. (7a). This microarray was modified from a previous virulence microarray (2) and contains 251 oligonucleotide probes specific for 183 virulence genes or markers representative of all known E. coli pathotypes. These probes are specific for various virulence genes, including genes encoding adhesins, toxins, hemolysins, invasins, autotransporters, capsular, flagellar, and somatic antigens, iron acquisition system or transport proteins, and outer membrane proteins, as well as genes recently shown to be associated with virulence in E. coli. Microarray experiments were performed in the laboratory of Josée Harel at the University of Montreal, as described previously (7a). Briefly, DNA from E. coli isolated from GCB mucosa and E. coli strain LF-82 were fluorescently labeled with a Cy5 dye and with a random-priming protocol derived from the Invitrogen Bioprime DNA labeling system (Invitrogen Life Technologies). Microarrays were then hybridized overnight with 500 ng of labeled DNA, in a slide hybridization chamber (Corning Canada) and in the dark. After hybridization, the microarrays were scanned with a ScanArray Lite fluorescent microarray analysis system (Canberra-Packard Canada). All microarray hybridizations were performed in duplicate, with DNA obtained from two separate bacterial cultures.
The mean of duplicate wells from at least three separate experiments was used as the input. Data from adhesion, invasion, and persistence assays were initially examined for equal variance by Bartlett's test. Adhesion and invasion assays with equal variance were analyzed by one-way analysis of variance (one-way ANOVA), where the strain type was used as the ‘treatment’ variable. Multiple pairwise comparisons were performed with Tukey's honestly significant difference test. Data for persistence in epithelial cells was logarithmically transformed prior to evaluation by a one-way ANOVA for analysis of strain. Statistical significance was defined at a P value of <0.05. All analyses were done with the statistical program Statistix (Analytical Software, Tallahassee, FL).
The EUB-338 FISH probe hybridized to all reference strains of bacteria. Sections of histologically normal colon had a readily staining bacterial flora that was restricted to the surface mucus layer and colonic glands (Fig. (Fig.2A).2A). Bacteria were not detected within the mucosa. Control slides spotted with bacteria and histological sections treated with RNase and the probe non-EUB-338 were negative (data not shown).
In contrast to normal colonic tissue, bacteria in GCB samples were scattered throughout the upper third of the mucosa (Fig. (Fig.2B).2B). Clumps of small coccobacilli (1 to 3 μm) were most commonly observed in areas where goblet cells and glands were replaced by a cellular infiltrate. In many sections bacteria appeared to be localized within cells (Fig. (Fig.2C).2C). Histochemical analysis of GCB sections revealed intramucosal gram-negative, non-acid-fast, argyrophilic coccobacilli, mainly within histiocytes (Fig. (Fig.2D),2D), concurring with mucosal localization by FISH. Invasive bacteria were not visible with eubacterial FISH or Gram staining in biopsy samples taken from a GCB-affected dog 3 months after the completion of antibiotic therapy.
To determine if bacterial colonization is a general feature of canine colitis, we examined sections from 27 dogs with non-GCB colitis. Bacterial invasion of the mucosa was observed in only two non-GCB samples, and in these sections the flora was pleomorphic, mixed gram positive and negative, and not cell associated.
Culture of GCB mucosa from two affected dogs yielded Klebsiella pneumoniae, Proteus mirabilis, and E. coli. Two E. coli strains were isolated from each dog and were identified as KD-1, KD-2, KD-3 and KD-4. Streptococcus faecalis and Chryseobacterium meningosepticum were isolated from separate individuals. Shigella, Salmonella, Campylobacter, and Yersinia were not isolated.
PCR amplification of DNA from GCB mucosa with 16S primers yielded amplicons of the expected size (320 bp) from pre- and posttreatment samples. Analysis of bacterial sequences yielded a predominance of Enterobacteriaceae, particularly E. coli/Shigella, in both affected dogs (Fig. (Fig.3).3). The posttreatment 16S rDNA library, 3 months after antibiotic-induced remission, contained a much more diverse flora, with only 4/34 clones belonging to Enterobacteriaceae (Fig. (Fig.33).
On the basis of the results of culture, 16S sequencing, and presence of intramucosal gram-negative bacteria, we employed FISH probes designed to hybridize to a subset of the Enterobacteriaceae (E. coli 1531, 23S rRNA) and E. coli/Shigella (16S rRNA). The specificity of these probes was confirmed with cultured bacteria, with the E. coli 1531 probe hybridizing to E. coli, Shigella, Klebsiella and Salmonella, and with the E. coli/Shigella probe to only these species. Both probes hybridized with the invasive flora of GCB specimens (Fig. (Fig.4).4). Culture of feces and mucosa for Shigella and PCR of colonic mucosal DNA with primers against the ipaH gene found in Shigella and enteroinvasive E. coli were negative.
Four E. coli strains (KD1 to -4) were isolated from GCB mucosa from two affected dogs. All of these strains adhered to cultured epithelial cells (Fig. (Fig.5A).5A). Strains KD2 to -4 displayed levels of adherence similar to those of the Crohn's disease isolate LF 82. KD-1, DH5α, and Salmonella were less adherent than LF 82 (P < 0.05).
In gentamicin protection assays, GCB-associated E. coli strains KD1 to -4 and LF-82 were significantly more invasive than DH5α but less invasive than Salmonella enterica serovar Typhimurium (P < 0.05) on both Caco-2 and MAC-T cells (Fig. 5B and C). GCB strains differed in their invasiveness: strain KD-2 invaded Caco-2 and MAC-T cells to a greater extent than KD-3 and KD-4 and similarly to LF-82. All E. coli strains were more invasive on Caco-2 than MAC-T cells, but the differences between strains were greater on MAC-T than Caco-2 cells. For example, KD-2 invaded Caco-2 cells 25 times more than DH5α and MAC-T cells 100 times more than DH5α.
The invasive behavior of GCB-associated E. coli was independently confirmed using differential immunostaining of extra- and intracellular bacteria and transmission electron microscopy (Fig. (Fig.6).6). Electron microscopy also revealed that GCB-associated E. coli induces alterations in the cell membrane characterized by elongations and actin condensation that are consistent with induced endocytosis by a “trigger” process (14). Internalized E. coli appeared to be in a close vacuole, but its exact location within the endosomal lysosomal continuum remains to be determined.
The invasion of epithelial cells by GCB-associated E. coli KD-1 and KD-2 was markedly reduced (P < 0.05) by cytochalasin D, colchicine, genistein and wortmannin, indicating involvement of microtubules, microfilaments, tyrosine kinase, and PI 3-kinase, respectively (Fig. (Fig.7).7). This pattern of inhibition was similar to that of E. coli LF 82. In contrast, invasion by the Salmonella enterica serovar Typhimurium control was inhibited by cytochalasin D and genistein, but not colchicine or wortmannin.
GCB strains KD1 to -3 and LF-82 were able to persist and replicate in epithelial cells over a 48-h period more effectively than strain KD4 and commensal DH5α (P < 0.05) (Fig. (Fig.8A8A).
In contrast to their behavior in epithelial cells, GCB-associated E. coli and LF-82 were unable to survive longer than DH5α in cultured primary bone marrow-derived macrophages (P > 0.05) (Fig. (Fig.8B8B).
Serotyping revealed a different O:H serotype for each GCB isolate and LF-82 (Table (Table1).1). Evaluation of genotype by RAPD-PCR (Fig. (Fig.9)9) showed a unique banding pattern for each strain, indicative of diversity in overall genotype. Phylogenetic grouping of E. coli strains by triplex PCR (Table (Table1)1) indicated that GCB strains KD-1 and KD-3 and CD strain LF-82 are part of the B2 phylogenetic group. The other two canine strains belonged to groups D and A. Phylogenetic triplex PCR performed on mucosal DNA samples from two affected Boxers gave a triple band pattern consistent with type B2. No bands were detected in the posttreatment mucosal DNA sample.
Multilocus sequence typing revealed that GCB strains KD1 to -3 and LF82 have novel st7 sequence types and clonal groups (Table (Table1).1). In contrast, GCB strain KD4 belonged to clonal group 23 that harbors predominantly E. coli isolated from healthy people (Selander strains) and a healthy dog (http://www.shigatox.net/cgi-bin/mlst7/strainquery?).
The genome sizes of LF-82 and KD1 to -3, strains that invade and persist within cultured cells, were between 188 and 276 kbp larger than that of MG1655, whereas the genome size of KD-4 (4.37 Mbp), a strain that invades but does not persist, was similar to that of MG1655 (4.47 Mbp) (Table (Table11).
PCR-based screening for virulence genes of diarrheagenic E. coli (LT, STa, STb, SLT-I, SLT-II, CNF-1, CNF-2, and the gamma variant of eae) and an invasion plasmid (ipaH) was negative for all GCB-associated strains (Table (Table1).1). All strains, including DH5α, were positive for fimH, the adhesin-encoding gene associated with epithelial cell invasion in uropathogenic E. coli (50). Sequencing to detect potentially pathoadaptive fimH alleles yielded polymorphisms at residues 27, 70, 78, 163, and 184 (Table (Table1).1). The translated sequences of the GCB isolate KD-1 and the Crohn's disease isolate LF-82 were identical. Strains KD1, KD-3, and LF82 had the N70S or S78N mutation, reported to be lineage specific for the B2 phylogenetic group (31, 74).
A more comprehensive screening of virulence factors from diarrheagenic and extraintestinal pathogenic E. coli using microarray analysis showed that GCB-associated and LF 82 E. coli strains lacked invasion plasmids, the attaching and effacing E. coli type III secretion system-related genes, and toxin-encoding genes associated with virulence in intestinal or extraintestinal pathogenic E. coli (Table (Table2).2). GCB-associated and LF 82 strains shared a variety of genes implicated in iron acquisition and metabolism, notably irp1, irp2, fyuA (yersiniabactin), chuA (hemoglobin utilization), fepC (ferric enterobactin transport ATP-binding protein), and iroN (siderophore receptor), though aerobactin-related genes (iutA and iucD) were not detected. The malX gene, which is a marker for the pathogenicity island of UPEC CFT073, and the iss gene, encoding a protein responsible for serum resistance, were present in most strains. The overall virulence gene pattern was most similar for the GCB-associated isolate KD-1 and the CD isolate LF 82, and both of these strains contained the ibeA gene, encoding an invasin of meningitis-associated E. coli (3). Strain LF 82 also possessed the usp gene, which encodes a uropathogenesis-specific protein, but not any other UPEC-associated genes, such as papA, papC, papGII (pilus P-encoding genes), cnf1 (cytotoxic necrotizing factor 1), or afa genes (afimbrial adhesin-encoding genes).
This study directly explored the possibility that an uncharacterized infectious agent or an abnormal mucosal flora is involved in the etiopathogenesis of granulomatous colitis of Boxer dogs. Eubacterial FISH provided clear evidence that large numbers of coccobacilli are present within the colonic mucosa of Boxer dogs with granulomatous colitis, but not other types of canine colitis, and histologically normal tissues. Culture of the mucosas from two affected dogs yielded E. coli, Klebsiella Streptococcus, Proteus and Chryseobacterium, considered normal fecal flora in dogs (38). To create a more comprehensive inventory of the bacterial species inhabiting the colonic mucosa, including uncultivable bacteria, we generated 16S rDNA libraries from colonic biopsy samples of two GCB patients. These libraries were dominated by sequences for Enterobacteriaceae, predominantly E. coli and Shigella, concurred with the presence of intramucosal gram-negative coccobacilli, and guided the selection of 16 and 23S rRNA FISH probes that hybridized with the invasive flora. The 16S E. coli/Shigella probe employed in the present study is specific for these species but cannot distinguish between them, so PCR for ipaH, a marker for the invasion plasmid of Shigella and enteroinvasive E. coli, and culture were performed to detect Shigella. As both PCR and culture were negative for Shigella we concluded that the invasive flora was E. coli.
Interestingly, E. coli was isolated previously from the regional lymph nodes of two Boxer dogs with granulomatous colitis at necropsy, but an association with disease was not determined (80). Another study, published while the present article was in preparation, describes the immunolocalization of E. coli, Lawsonia intracellularis, Campylobacter, and Salmonella to macrophages in the colons of 10, 3, 2, and 1 of 10 Boxer dogs with granulomatous colitis, respectively (79). While these findings lend independent support to our observations, their specificity is questionable because the antibody they used to localize E. coli (Dako B0357) is polyclonal, recognizes at least 80 different E. coli antigens in crossed immunoelectrophoresis and a multitude of E. coli antigens in immunoblotting from SDS-polyacrylamide gel electrophoresis, and is dilution dependent in its specificity (44, 81).
The association of E. coli with the intestinal mucosa of Boxer dogs with granulomatous colitis is similar to findings in people with inflammatory bowel disease, particularly Crohn's disease, and rodents with engineered susceptibility to IBD (17, 49, 55, 72, 73, 76). However, the mucosal localization of E. coli in GCB, with multifocal clusters of E. coli consistently observed in the upper two-thirds of the mucosa, is more uniform than reported in human IBD, where some investigators describe intact bacteria, E. coli antigens, or DNA within granulomas and the lamina propria (11, 44, 55, 69) but others indicate that bacteria are restricted to the mucosal surface (40, 76). The variable localization of mucosa-associated E. coli in people with IBD may be attributable to differences in disease phenotyping between studies (15), biopsy site (ileal, colonic, or rectal), biopsy type (surgical versus endoscopic) (52), and method used, e.g., FISH (the 23S E. coli 1531 probe employed by previous studies [55, 72] is not E. coli specific). It could also reflect the presence of E. coli strains that differ in their ability to invade and persist within the mucosa. The latter possibility is supported by observations that strains with an ability to adhere to and invade cultured epithelial cells, hallmarks of pathogenic E. coli, are more commonly isolated from patients with Crohn's disease than those with ulcerative colitis or healthy controls (17, 49): i.e., such strains are found in 36% of early CD lesions in the ileum, compared with 3.7% of colonic CD, 0% of UC, and 1.9% of control samples. In the present study we found that GCB-associated E. coli was able effectively to adhere to and invade cultured epithelial cells in numbers similar to those of the well-characterized ileal CD-associated E. coli strain, LF 82 (16). GCB E. coli, like LF 82, was also able to invade a diverse spectrum of cultured epithelial cell lines (5). Transmission electron microscopy of canine GCB E. coli infecting an epithelial cell monolayer show a trigger, or “splash,” type of endocytosis, similar to that induced by pathogenic bacteria such as Salmonella and Shigella spp. (14). This type of behavior, also displayed by E. coli LF82 (5), usually requires the translocation of effector molecules from the bacterium into the host cell via a type III secretion system and frequently involves the activation of the Rho family of GTPases and cytoskeletal reorganization (14, 61). Invasion by canine GCB isolates and LF 82 was markedly decreased by cytochalasin D, colchicine, genistein, and wortmannin, indicating involvement of microfilaments, microtubules, tyrosine kinase, and PI 3-kinase, and is consistent with previous studies with LF-82 in Hep2 and Int 407 epithelial cells (5). Microtubules and microfilaments are also utilized by enteroinvasive and meningitis-associated E. coli and invasive Klebsiella and Campylobacter but not Salmonella (22, 53, 57, 61).
Invasive GCB-associated E. coli was able to persist in cultured epithelial cells for 48 h and appeared to reside in a tight vacuole within the cytoplasm, suggesting a location in the endosomal lysosomal continuum. E. coli LF 82 has also been shown to reside and replicate in the cytoplasm of epithelial cells (5). Intracellular persistence and replication suggest an ability to escape from the endosomal and lysosomal network and parallel the situation with the pathogens Shigella and Listeria (14). Intracellular persistence of IBD-associated E. coli may directly contribute to the proinflammatory mucosal environment in IBD, and this is supported by studies demonstrating translocation of NF-κB and release of IL-8 in epithelial cells infected with IBD-associated E. coli (27, 42).
Our findings that GCB E. coli isolates and LF 82 did not outlive commensal E. coli (DH5α) in primary bone marrow-derived macrophages contrast with previous reports describing survival and replication of LF 82 relative to harmless commensal E. coli in cultured J774-A1 macrophages and a survival advantage in human-derived mononuclear macrophages and mouse peritoneal cells (7, 27). Those observations appear contradictory, but they can be reconciled by considering the relatively weak killing ability of the J774-A1 cell line relative to that of primary macrophages (27). In the light of recent studies showing impaired innate immunity in the form of an abnormal acute inflammatory response to E. coli in people with Crohn's disease (47), different outcomes based on phagocytic ability may be analogous to differences between healthy and susceptible individuals, with host susceptibility (e.g., predisposition of Boxer breed for GCB, polymorphisms in NOD-2) and disease-associated luminal bacteria (e.g., IBD-associated E. coli) acting as joint determinants of disease.
The combination of phylogenetic analysis and virulence gene profiling provided insights into the lineage and genetic armory of GCB and Cohn's associated E. coli. The presence of chuA, determined by PCR and microarray analysis, placed three of four GCB strains and LF 82 into group B2 or D. The presence of yjaA indicated that E. coli KD-1, KD-3 and LF 82 are B2, while its absence placed KD-2 into group D (12). The detection of the fimH polymorphisms associated with B2 lineage, N70S and S78N, further supported these results (31, 74). The serotypes (O1, O8, and O83) and gene profiles (fyuA, irp1, irp2, malX, ompT, ibeA and kpsMII) of these B2 and D strains are consistent with those of extraintestinal pathogenic E. coli (ExPEC), which causes cystitis, pyelonephritis, prostatitis, sepsis, and meningitis, and avian pathogenic E. coli (4, 21, 26, 36, 37). In contrast, strain KD-4, our least invasive and persistent GCB isolate, belonged to group A, which contains most of the commensal strains of E. coli (62). The mucosal association of B2 E. coli in GCB was confirmed by phylogenetic triplex PCR of colonic DNA that yielded amplicons for chuA, yjaA, and TSPE4.C2 in samples from two GCB-affected dogs but not the postremission sample. These observations suggest that GCB- and CD-associated E. coli strains are genetically more similar to ExPEC than diarrheagenic E. coli strains.
Precise placement of GCB-associated E. coli strains and LF 82 within the ExPEC group is difficult. For example, the gene profiles of KD1 to -3 and LF 82 are broadly similar to that of UPEC CFT073: all contain malX, a marker of a pathogenicity island in CFT073, and strain LF 82 also hybridizes with a uropathogenesis-specific gene (usp), but these strains lack genes, such as papA, papC, papGII, cnf1, afa, hlyA, and iucD, that are commonly associated with UPEC and other ExPEC strains (3, 21, 35, 36). The presence of ibeA, the gene encoding invasin for brain endothelium, in KD-1 and LF 82 suggests these strains may belong to the meningitis-associated group (3). However, ibeA is not restricted to meningitis-causing strains, and KD-1 and LF 82 lack genes, such as sfa, cdtB, neuA, and neuC, that are present in many B2 meningitis strains (26, 36). Moreover, GCB-associated E. coli and LF 82 are able to invade and persist in epithelial cells, whereas invasion by meningitis-associated E. coli is restricted to endothelial cells (53).
The pathogen-like behavior displayed by GCB and CD E. coli isolates in cultured cells strongly suggests that these strains harbor genes encoding virulence, but an extensive PCR- and microarray-based screen of GCB-associated and LF 82 E. coli strains failed to detect genes involved with the pathogenic behavior of E. coli strains associated with intestinal disease, such as invasion plasmids, type III secretion systems, or toxins, and concurs with previous studies of CD-associated E. coli in people (16, 17, 42, 49). The few virulence genes we found in GCB-associated E. coli and LF 82 were largely part of a cluster of genes, irp1, irp2, fyuA (yersiniabactin), chuA (hemoglobin utilization), fepC (ferric enterobactin transport ATP-binding protein), and iroN (siderophore receptor), involved in iron acquisition and metabolism (13, 41). These genes are considered important for iron acquisition by ExPEC within an infected host, and bacterial siderophores may also impact the cellular immune response (23).
The paucity of virulence genes detected in GCB and CD E. coli strains suggests that these strains may harbor as-yet-uncharacterized genes to account for their disease association and pathogenic behavior in cultured cells. This seems feasible considering the high degree of diversity in E. coli as a species (only 39% of core proteins are conserved in UPEC, EHEC, and K-12) and its propensity for acquiring DNA from distantly related organisms (56, 83). The B2 lineage is also a particularly appropriate genetic background for acquiring virulence traits: B2 strains are associated with lethality in mice (21, 31, 62) and often have larger genome sizes than the commensals in group A, reflecting the presence of virulence-associated genes such as pathogenicity islands (29, 35, 62). Our findings that GCB strains KD1 to -3 and LF 82 have larger genomes than E. coli MG1655, belong to undefined MLST clonal groups, and contain genes that are thought to be acquired by horizontal transfer (e.g., the yersiniabactin gene and malX) (71) further support this possibility and the notion that this group of E. coli strains represents a new pathotype, adherent and invasive E. coli (AIEC), as proposed by Darfeuille-Michaud et al. (16) Strains with an AIEC pathotype could potentially belong to a clonal group associated with chronic intestinal inflammation, comparable to the association of O157:H7 with hemorrhagic gastroenteritis and hemolytic uremic syndrome. However, serotyping and genotyping (with random amplified polymorphic DNA PCR) showed marked heterogeneity between strains and does not support the presence of a unique E. coli strain associated with GCB and Crohn's disease. These observations are similar to the results of ribotype analysis of E. coli strains from patients with CD showing that no single strain is found in every patient (51).
In conclusion, we have determined that granulomatous colitis of Boxer dogs, a disease that has features in common with idiopathic IBD in people, is associated with selective intramucosal colonization by Escherichia coli. E. coli strains isolated from the mucosas of two affected dogs adhere to, invade, persist in, and replicate in cultured epithelial cells to the same degree as Crohn's disease-associated E. coli LF-82. The invasion process, which resembles triggered endocytosis, requires intact microtubules, microfilaments, PI 3-kinase, and tyrosine kinase. The similar phylogeny and virulence gene profiles of GCB strains and LF 82 hint at the possibility of lineage-specific pathoadaptation and point to the association of E. coli strains resembling extraintestinal pathogenic strains in genotype with chronic intestinal inflammation in dogs and people. The role of these E. coli strains in the etiopathogenesis of GCB and CD remains to be determined.
K. Simpson is supported by a grant from the U.S. Public Health Service (DK002938). This study was funded in part by the Indirect Vitamins Purchasers Antitrust Litigation Settlement administered by the New York State Attorney General.
We thank J. Chaitman, D. J. Chew, M. J. Day, E. J. Hall, R. A. Hostutler, and D. F. Kelly for providing patient information and tissue sections, A. Darfeuille-Michaud and T. S. Whittam for E. coli strains, and A. Quaroni for the Caco-2 cell line. We thank Francis Davis for technical support.
Editor: F. C. Fang