|Home | About | Journals | Submit | Contact Us | Français|
We recently showed that Bifidobacterium animalis is more prevalent within the colons of IL-10 deficient (−/−) mice than in wild type (WT) animals colonized with the same specific pathogen free (SPF) fecal contents. Here we tested the ability of this organism to cause T cell-mediated intestinal inflammation by introducing it into germ-free (GF) IL-10−/− mice.
GF IL-10−/− or WT mice were monoassociated with Bifidobacterium animalis subsp. animalis ATCC 25527T or with Bifidobacterium infantis ATCC 15697T. Inflammation was measured by blinded histologic scores of the duodenum, cecum and colon and by spontaneous secretion of IL-12/IL-23 p40 from colonic explants. Bacterial antigen-specific CD4+ mesenteric lymph node (MLN) T cell recall responses were measured in response to antigen presenting cells (APC) pulsed with bacterial lysates.
B. animalis caused marked duodenal inflammation and mild colitis in monoassociated IL-10−/− mice, whereas the intestinal tracts of WT animals remained free of inflammation. B. infantis colonization resulted in mild inflammation in the duodena of IL-10−/− mice. CD4+ MLN T cells from B. animalis monoassociated IL-10−/− mice secreted high levels of IFN-γ and IL-17 in response to B. animalis lysate. B. animalis equally colonized the different intestinal regions of WT and IL-10−/− mice.
B. animalis, a traditional probiotic species that is expanded in experimental colitis in this model, induces marked duodenal and mild colonic inflammation and TH1/TH17 immune responses when introduced alone into GF IL-10−/− mice. This suggests a potential pathogenic role for this commensal bacterial species in a susceptible host.
Human inflammatory bowel diseases (IBD), including ulcerative colitis and Crohn’s disease, are believed to be caused by an inappropriate cell-mediated (T cell) immune response to commensal enteric bacteria in genetically susceptible individuals 1–2 This is supported by clinical observations linking increased concentrations of luminal and adherent bacteria to inflamed regions of the intestinal tract 3–5 and by studies of experimental colitis in genetically susceptible animals, including IL-10 deficient (−/−) mice, where colitis and immune activation fail to develop in the absence of commensal bacteria 1, 6. Colonization of susceptible gnotobiotic rodents has demonstrated that some commensal bacterial species cause inflammation, some have no effect, and others (probiotics) provide protection from the inflammation caused by detrimental species 3. For example, gnotobiotic IL-10−/− mice develop bacterial species-antigen specific T cell-mediated intestinal inflammation when monoassociated with Enterococcus faecalis or Escherichia coli 7, but remain healthy when colonized with Bacteroides vulgatus, which has been shown to induce colitis in HLA-B27/β2 microglobulin transgenic rats 8. IL-10−/− mice housed under specific pathogen free (SPF) conditions and fed Lactobacillus salivarius subsp. salivarius or Bifidobacterium infantis displayed reduced colonic inflammation and attenuated Peyer’s patch IFN-γ responses following in vitro stimulation with the enteric pathogen Salmonella typhimurium, compared to placebo controls 9. These observations highlight the importance of understanding how different intestinal inhabitants selectively impact the host’s mucosal immune system, and how this immune activation or inhibition influences health in hosts with distinct genetic backgrounds.
We recently reported significant alterations in the composition of the enteric microbiota in colitic IL-10−/− mice compared to healthy WT animals after germ-free (GF) mice were colonized with fecal microbiota obtained from SPF wild type (WT) mice 10. Changes in microbiota composition with progression of colitis were analyzed using PCR and denaturing gradient gel electrophoresis (PCR/DGGE) and specific DNA:RNA dot blot analysis. Interestingly, Bifidobacterium animalis was one of four bacterial species that increased in abundance as colitis developed. Bifidobacteria are widely considered to be beneficial commensal organisms that are “generally regarded as safe” (GRAS) and have been associated with probiotic activity 3, 11–13. Therefore, the increased prevalence of B. animalis in the colons of colitic IL-10−/− mice was unexpected and the inflammatory potential of this species in IL-10−/− mice warranted further investigation.
In the current study, we evaluated B. animalis for its ability to induce intestinal inflammation in gnotobiotic WT and IL-10−/− mice using the type strain Bifidobacterium animalis subsp. animalis ATCC 25527T for colonization. For comparison, we also colonized gnotobiotic WT and IL-10−/− mice with Bifidobacterium infantis ATCC 15697T. We further investigated the antigen specificity of the T cell-mediated immune response of B. animalis colonized mice and assessed possible cross-reactivity of the T cell response to other bacteria. Here we report that IL-10−/− mice monoassociated with B. animalis developed extensive duodenal and mild colitic inflammation with CD4+ T cell responses directed against both unique and shared bacterial antigens expressed by B. animalis.
Bifidobacterium animalis subsp. animalis ATCC 25527T, Bifidobacterium infantis ATCC 15697T and Bifidobacterium bifidum ATCC 11863T were obtained from the American Type Culture Collection (ATCC) and propagated anaerobically at 37°C in Lactobacilli MRS medium (BD Biosciences, Sparks, MD). A murine Escherichia coli strain designated NC1017 and a human oral isolate of Enterococcus faecalis (strain OG1RF provided by Mark Huycke M.D.) 14 have been previously described. Bacterial lysates were prepared as previously described 7.
GF IL-10−/− mice on the 129S6/SvEv background and WT control (inbred 129S6/SvEv) mice were from breeding colonies maintained at the National Gnotobiotic Rodent Resource Center (NGRRC), University of North Carolina, Chapel Hill or the Center for Gastrointestinal Biology and Disease (CGIBD) Gnotobiotic Animal Core at North Carolina State University. These mouse colonies were originally derived by hysterectomy at the Gnotobiotic Laboratory (University of Wisconsin, Madison) 6. Mice were monoassociated at 10–26 weeks of age with B. animalis or B. infantis by gavage feeding and rectal swabbing with viable cultured bacteria. Monoassociated mice were maintained in either the NGRRC or the CGIBD gnotobiotic facilities. Mice colonized with B. animalis or B. infantis were housed in separate isolators. Bacterial monoassociation and absence of contamination by other bacterial species were confirmed by periodic aerobic and anaerobic culture of stool samples. Mice were killed 11, 17 or 20–24 weeks after colonization with B. animalis and 22–23 weeks after colonization with B. infantis. WT 129S6/SvEv mice (Taconic Laboratories, Germantown, NY) maintained under SPF conditions free of Helicobacter species were used to prepare antigen-presenting cells (APC). Animal use protocols were approved by the Institutional Animal Care and Use Committees of North Carolina State University and the University of North Carolina at Chapel Hill.
At necropsy after 24 weeks of colonization with B. animalis, a portion of the cecal contents from two 129 WT and two IL-10−/− mice was used to inoculate sterile Lactobacilli MRS liquid media and cultures were incubated in sealed glass culture tubes overnight at 37°C. Fecal samples were collected from two 129 WT and two IL-10−/− B. infantis monoassociated mice after 8 weeks of colonization and were similarly inoculated into sterile Lactobacilli MRS liquid media. An aliquot from each culture was streaked onto Lactobacilli MRS agar supplemented with 0.05% L-cysteine and plates were incubated for 72 hrs under anaerobic conditions. Cell and colony morphology of these isolates were identical to the B. animalis and B. infantis strains initially used to colonize the animals (data not shown). Genomic DNA was isolated from randomly selected individual colonies of B. animalis and B. infantis stock cultures and from each ex vivo isolate using the DNeasy Blood & Tissue Kit (Quiagen, Valencia, CA). 16S ribosomal RNA gene sequence was amplified by PCR using the universal primers 27F (AGAGTTTGATCCTGGCTCAG) and 1491R (GGTTACCTTGTTACGACTT). The following reagents were included in each PCR tube: 5 μL 10 × buffer, 1.5 μL 50 mM MgCl2 (Invitrogen, Carlsbad, CA), 0.5 μL 5 mM dNTP (GE Healthcare, Piscataway, NJ), 1 μL each of primers 27F and 1491R (5 pmol/μL), 30–40 ng DNA template, 1.25 U Taq DNA Polymerase (Invitrogen, Carlsbad, CA) and deionized water to a final volume of 50 μL. The PCR cycling was performed in a GeneAmp PCR System 9700 (Applied Biosystems, Foster City, CA), under the following conditions: 95°C for 2 min and 30 sec, then 35 cycles consisting of 95°C for 60 s, 51°C for 60 s and 72°C for 120 s. The PCR products were incubated at 4°C until used. Sequencing of PCR products was carried out using 27F and 1491R as sequencing primers. DNA was sequenced at the UNC-CH Genome Analysis Facility on a 3730 DNA Analyzer (Applied Biosystems, Foster City, CA). Sequence analysis using the BLASTN algorithm (http://www.ncbi.nlm.nih.gov/BLAST/) confirmed that all ex vivo isolates from mice monoassociated with B. animalis or B. infantis were B. animalis and B. infantis, respectively.
Bacterial numbers of B. animalis in the gut of 129S6/SvEv WT and IL-10−/− mice (six mice each) were measured in the duodenum, cecum, proximal colon and distal colon by quantitative plating of serial dilutions of homogenized intestinal segments on Lactobacilli MRS agar supplemented with 0.05% L-cysteine (Sigma, St. Lois, MO). Plates were incubated for 72 hrs under anaerobic conditions at 37°C and colonies were enumerated.
Sections of colon (proximal and distal), cecum and duodenum were fixed in 10% neutral buffered formalin. The fixed tissue was embedded in paraffin and stained with H&E. Sections were scored blindly for severity of inflammation by a single individual using a well-validated scale 6, 7, 15, 16. Histological scores (0 to 4) were based on the degree of lamina propria and submucosal mononuclear cellular infiltration, crypt hyperplasia, goblet cell depletion, and architectural distortion. These scoring criteria that were developed previously to evaluate colitis were modified for use with the duodenal tissue sections. In the duodenum a score of 0 represents no inflammation and normal villus architecture. A score of 1 represents mild focal cellular infiltration and normal villous architecture. A score of 2 represents mild lamina propria cellular infiltration and early crypt epithelial hyperplasia with normal villus architecture. A score of 3 represents more pronounced cellular infiltration, thickened mucosa, marked epithelial hyperplasia and moderate distortion of villus architecture. A score of 4 represents extensive cellular infiltration throughout the section and severe architectural distortion.
Colonic tissue fragment cultures were prepared from the large intestine as previously described 6, 7. Colonic tissue was thoroughly irrigated with phosphate-buffered saline (PBS), shaken at room temperature in RPMI containing 50 μg/mL gentamicin for 30 minutes at 280 rpm, cut into 1-cm fragments, blotted to remove excess media, and weighed. Colonic tissue fragments were distributed (0.05 g per well) into 24-well plates (Costar 3524) and incubated in 1 mL of RPMI 1640 medium supplemented with 5% fetal bovine serum, 50 μg/mL gentamicin, and 1% antibiotic/antimycotic (penicillin/streptomycin/amphotericin B; Invitrogen, Carlsbad, CA) for 20 hours at 37°C. Supernatants were collected and stored at −20°C prior to use for IL-12/IL-23 p40 quantification by ELISA.
CD4+ T cells were enriched by negative selection from mesenteric lymph node (MLN) cells harvested from IL-10−/− or WT mice using magnetic beads coated with B cell specific (anti-CD45R) antibody plus beads coated with CD8-specific (anti-CD8α) antibody according to the manufacturer’s instructions (Miltenyi - Biotec, Auburn, CA) as previously described 7, 16. Over 95% of the enriched MLN cell populations expressed CD4 as determined by flow cytometry.
Splenic antigen presenting cells (APC) were prepared from SPF 129S6/SvEv WT mice and pulsed overnight with 10 μg/mL of B. animalis, B. bifidum, B. infantis, Enterococcus faecalis, or Escherichia coli lysate or keyhole limpet hemocyanin (KLH; Pierce, Rockford, IL) as an unrelated antigen control, as described previously7, 16. CD4+ T cells (2 × 105 CD4+ T cells/well) were co-cultured for 72 hours with antigen-pulsed APC (3 × 105 APC cell/well) in flat bottom 96-well cell culture plates (Costar 3595), 0.2 mL per culture. Supernatants were collected after 3 days and stored at −20°C.
Bone marrow derived dendritic cells (BMDC) were isolated and cultured from femora and tibiae of SPF 129S6/SvEv mice, as previously described 17. BMDC (2 × 104) were seeded in triplicate wells of flat bottom 96-well cell culture plates (Costar 3595) in the presence of media alone, LPS (1 μg/ml) or various bacterial lysates (10 μg/ml) at 0.2 mL per culture. Supernatants were collected after 3 days and stored at −20°C.
We used commercially available monoclonal anti-mouse IFN-γ, IL-12/IL-23 p40 (BD Biosciences Pharmingen, San Diego, CA), and IL-17 (e-Bioscience, San Diego, CA) specific capture and detection reagents to measure amounts of secreted cytokines by ELISA. For detection of IL-12/IL-23 p40, plates were coated with anti-mouse IL-12 p40/p70 (clone C15.6) and bound ligand was detected with biotin anti-mouse IL-12 p40/p70 (clone C17.8). To quantify IFN-γ, plates were coated with anti-mouse IFN-γ (clone R4-6A2) and bound ligand was detected with biotin anti-mouse IFN-γ (clone XMG1.2). For detection of IL-17, plates were coated with anti-mouse IL-17A, clone eBioTC11-18H10.1 and IL-17 was detected using biotin anti-mouse IL-17A, clone eBioTC11-8H4. Cytokine levels were measured in triplicate supernatants and compared to standard curves generated using recombinant murine cytokines.
MLN cell subpopulations (before and after negative selection) were evaluated as previously described 7 on the FACScan (BD Biosciences, Mountain View, CA) using FITC anti-CD4, PE anti-CD8, or FITC anti-B220 (all from Invitrogen).
Nonparametric histologic scores were analyzed with the Mann Whitney test (SAS; Cary, NC) by statisticians in the Biostatistics Core of the Center for Gastrointestinal Biology and Disease. The paired Student’s T-test was used to compare all other data. Statistical significance was defined as p < 0.05 for comparisons indicated. Data are reported as mean values ± standard error of the mean (SE), unless otherwise indicated.
To test whether B. animalis could elicit intestinal inflammation in IL-10−/− mice, we colonized inbred GF IL-10−/− (129S6/SvEv background) and GF 129S6/SvEv WT mice with B. animalis ATCC 25527T for 20–24 weeks. We confirmed monoassociation by analyzing the 16S ribosomal RNA sequence of several ex vivo bacterial isolates from cecal contents at necropsy (data not shown). Two of nineteen monoassociated IL-10−/− mice studied developed rectal prolapse, however most of the mice did not display outward signs of distress beyond lethargy and soft stools. Histological evidence of mild inflammation was apparent throughout the colons of IL-10−/− mice, characterized by moderate crypt hyperplasia, infiltration of predominantly mononuclear cells into the lamina propria, but not the submucosa, and moderate goblet cell depletion (Figure 1D, E, and F). WT 129S6/SvEv mice, similarly colonized for 20–24 weeks had no histological evidence of colitis (Figure 1A, B, and C). Importantly, there was considerable grossly evident thickening of the duodena in IL-10−/− mice with evidence of partial obstruction, demonstrated by gastric and proximal duodenal distention with fluid and luminal content retention (Figure 2B ). However, there were no significant differences in the weights of age matched WT and IL-10−/− mice at 20–24 weeks after colonization, suggesting the absence of high grade duodenal obstruction (data not shown). Histological examination of duodenal tissue revealed massive cellular infiltration of mononuclear cells into the lamina propria, including the villi and significant crypt hyperplasia, compared to healthy duodenal tissue from monoassociated WT animals (compare IL-10−/−, Figure 2D–E with WT Figure 2C). The typical architecture of the mucosa was altered by the formation of abnormal crypt and villus structures consisting of branched and fused villi, but no mucosal ulceration (Figure 2D–E). This inflammation was localized to the duodenum and did not extend into the jejunum or ileum (data not shown). Histological scores for the colons of IL-10−/− mice reflected a mild pancolitis in these animals (distal colon: 1.9 ± 0.9; proximal colon: 1.6 ± 0.8; cecum: 2.3 ± 0.6) and were significantly higher than the scores for WT colons (distal colon: 0.7 ± 0.2; proximal colon: 1.0 ± 0.5; cecum: 1.2 ± 0.4) (Figure 3A). This histologic evidence of colitis in B. animalis-monoassociated IL-10−/− mice was further supported by the spontaneous release of higher amounts of IL-12/IL-23 p40 from colonic explants from IL-10−/− mice compared to WT mice (Figure 3B). Scores for histologic detection of inflammation in the duodenum reflected the extensive abnormalities within this intestinal segment in IL-10−/− animals (3.1 ± 0.3) compared to WT duodenal tissue (0.6 ± 0.2) (Figure 3A). In preliminary experiments we evaluated WT and IL-10−/− mice monoassociated with B. animalis for 11 and 17 weeks. Low levels of inflammation were detected in the distal colons and ceca of IL-10−/− animals by histological analysis and by spontaneous IL-12/IL-23 p40 release from colonic explants (data not shown). Duodena were not evaluated at these earlier time points. Together, these results show that IL-10−/− mice monoassociated with B. animalis developed a mild pancolitis and extensive duodenal inflammation with proximal duodenal and gastric dilation, whereas similarly colonized WT mice remained healthy.
We wondered whether B. animalis was unique among bifidobacteria in its ability to induce intestinal inflammation in IL-10−/− mice, or whether other Bifidobacterium species shared this capacity. Therefore, we colonized GF WT and IL-10−/− mice with B. infantis for 22–23 weeks. All WT and IL-10−/− mice appeared healthy throughout the colonization period and lacked macroscopic evidence of intestinal inflammation or gastric or duodenal dilation. Histological scores for different regions of the intestinal tract and spontaneous IL-12/IL-23 p40 secretion from colonic explants reflected a low level inflammatory response in IL-10−/− animals (Table 1). Only the duodenal histologic scores showed significant differences between IL-10−/− and WT mice, with no differences in the colon. Importantly, these measurements of inflammation in the intestines of B. infantis monoassociated IL-10−/− animals were lower than those obtained from B. animalis monoassociated IL-10−/− animals shown in Figure 3.
We hypothesized that the severe inflammation seen within the duodena of IL-10−/− mice might be due to selective colonization of this compartment by B. animalis. To test this we performed quantitative bacterial cultures on homogenized intestinal segments from the distal colon, proximal colon, cecum and duodenum of WT and IL-10−/− mice. This technique measures both luminal and adherent bacteria colonizing each compartment 18. Comparing WT and IL-10−/− mice, there were no statistically significant differences in the numbers of viable bacteria colonizing any of these compartments (Table 2). Also, bacterial concentrations were 2–3 orders of magnitude higher in the colons than in the duodena of these animals. These data suggest that the severe inflammation observed within the duodena of IL-10−/− mice compared to healthy WT mice was not due to differential colonization of this compartment by B. animalis.
We have previously shown that IL-10−/− mice monoassociated or dual-associated with bacterial species that cause intestinal inflammation display CD4+ MLN T cell responses to lysates of the species that induces inflammation with bacterial species-antigen specificity 7, 19. Here we asked whether B. animalis monoassociated IL-10−/− mice developed similar T cell responses directed against B. animalis bacterial antigens and whether T cells would cross-react with antigens expressed by other bacterial species. MLN-derived CD4+ T cells (isolated from IL-10−/− and WT monoassociated mice) were cultured with WT APC that had been pulsed in vitro with either KLH (as an unrelated antigen control) or with B. animalis, B. infantis, B. bifidum, E. coli or E. faecalis lysates. CD4+ MLN T-cells from IL-10−/− mice, but not from WT mice, produced high levels of IFN- γ (Figure 4A) and IL-17 (Figure 4B) when stimulated with B. animalis lysate-pulsed APC, indicating a strong recall T cell response against bacterial antigens expressed by this organism. Interestingly, these IL-10−/− T cells also secreted moderate levels of IFN-γ and IL-17 in response to APC pulsed with lysates of B. infantis, B. bifidum and E. coli, but not to E. faecalis lysate-pulsed APC. Cytokine measurements from CD4+ T cells from monoassociated WT mice cultured with the same lysates were consistently at or below the detection limits for these assays.
Bacterial cell lysates contain many components that could influence T cell secretion of inflammatory cytokines through their interaction with pattern recognition receptors on APC in an antigen non-specific manner 20. In order to evaluate the adjuvant-mediated stimulating capacity of these lysates we cultured WT BMDC in the presence of media alone, LPS (1 μg/ml) or each bacterial lysate at the same concentration used to pulse APC prior to co-culture with CD4+ T cells (10 μg/ml) and measured the secretion of IL-12/IL-23 p40 (Table 3). LPS and E. coli lysate stimulated the highest levels of IL-12/IL-23 p40 from BMDC (13,307.7 ± 1,975.9 and 15,307.7 ± 4,137.4 pg/ml, respectively); the three Bifidobacterium lysates stimulated approximately 15 fold lower levels and E. faecalis stimulated only slightly more than media alone. These results indicate that the three different Bifidobacterium lysates share similar potential for adjuvant activity. Therefore, IL-10−/− mice monoassociated with B. animalis likely develop CD4+ T cell immune responses against antigens expressed by B. animalis that are either not shared or are expressed at lower levels by the other species of bifidobacteria that we evaluated (Figure 4).
Our study demonstrates that Bifidobacterium animalis subsp. animalis ATCC 25527T causes mild colitis and marked duodenal inflammation in monoassociated gnotobiotic IL-10−/− mice. These results provide potential pathogenetic significance for our earlier study that showed increased abundance of B. animalis in the colons of SPF IL-10−/− mice as cecal and colonic inflammation progressed 10. We demonstrate that a bacterial species that is selectively expanded in SPF IL-10−/− mice with experimental colitis 10 is capable of inducing colitis and TH1 and TH17 CD4+ T cell responses in monoassociated gnotobiotic IL-10−/− mice.
B. animalis now joins E. faecalis, 7, 14 E. coli (mouse isolate), 7 and Enterobacter cloacae 21 in a growing list of commensal organisms shown to initiate intestinal inflammation in IL-10−/− mice. Importantly, many commensal isolates are not able to cause intestinal inflammation in monoassociated IL-10−/− mice, including: Pseudomonas fluorescens 7, Bacteroides vulgatus (guinea pig), Streptococcus faecium [Group D], Escherichia coli, Peptostreptococcus productus, Eubacterium contortum, and Streptococcus avium (isolates from Crohn’s patients)6, Helicobacter hepaticus22, Candida albicans, Lactococcus lactis, a Bifidobacterium species, a Bacillus species, several species of Lactobacillus,14 Viridans group streptococci, and Clostridium sordellii 23. In our study, GF IL-10−/− mice monoassociated with B. infantis developed mild inflammation in the duodenum after 22–23 weeks, but there were no statistically significant differences in the histological scores from the ceca and colons of WT B. infantis-colonized animals. The studies available to date demonstrate considerable heterogeneity within bacterial genera and even within individual species in the capacity to induce intestinal inflammation in GF IL-10−/− mice. For example, a commensal strain of E. coli (designated NC 101) isolated from a SPF-housed 129S6/SvEv WT mouse caused moderate colitis in GF IL-10−/− mice 7, whereas an E. coli strain isolated from a patient with Crohn’s disease 6 and a laboratory strain, E. coli K12 24 did not. Similar diversity appears to exist among various isolates/species of Bifidobacterium, since the B. animalis ATCC 25527T strain used in our study induced marked duodenal and mild colonic inflammation in GF IL-10−/− mice, B. infantis 15697T induced a low-grade response and an isolate used by Balish et. al. did not cause intestinal inflammation 14. Indeed, significant genetic and phenotypic heterogeneity has been reported for different species and strains within the genus Bifidobacterium 25, 26.
Our observation that B. animalis causes intestinal inflammation in this genetically susceptible model is of considerable interest, since this species is widely utilized as a probiotic organism. Much of the clinical research has been done using the strain Bifidobacterium animalis DN-173 010 12. This strain has a high postgastric survival rate that is greater than another commercially available Bifidobacterium strain 27, 28. B. animalis DN-173 010 also reduced colonic transit time when consumed in Bifidobacterium-fermented milk 29. Van der Meulen et al, reported B. animalis DN-173 010 to be unique among B. animalis isolates in its preferential metabolism of certain short-chain fructans, suggesting it may have a competitive advantage over other microorganisms in the human gut, where oligo- and polysaccharides are the main sugar substrates 30. Recently, B. animalis has been reclassified to include the two subspecies Bifidobacterium animalis subsp. animalis and Bifidobacterium animalis subsp. lactis based on molecular and phenotypic characteristics 26. Strains of the lactis subspecies including BB12 and HN019 have variable probiotic effects in animals and humans 31–35. Following our previous observation that B. animalis forms an increased proportion of the gut microbiota of colitic IL-10−/− mice compared with WT mice 10, we selected strain ATCC 25527T to colonize GF animals because of its designation as the type strain for B. animalis. Our results suggest that the strain ATCC 25527T possesses unique phenotypic traits that allow it to initiate an inflammatory response under certain circumstances, including monoassociation of IL-10 deficient rodents.
The mechanisms underlying the ability of some commensal bacterial strains to cause inflammation in susceptible hosts, while others remain innocuous are currently unknown. One possibility is that detrimental species colonize genetically susceptible individuals to a greater extent than non-susceptible individuals, thereby providing increased stimulus for the innate and adaptive immune systems 2. However, quantitative bacterial culture from different intestinal segments revealed no difference in B. animalis colonization patterns between monoassociated WT and IL-10−/− mice. Furthermore, B. animalis colony counts from the duodena of both WT and IL-10−/− mice were 2–3 orders of magnitude lower than those obtained from the colon. We previously reported moderate duodenal inflammation with functional obstruction in IL-10−/− mice monoassociated with E. faecalis, but not with E. coli NC101 despite higher duodenal concentrations of E. coli vs. E. faecalis, thus demonstrating a similar disassociation of regional intestinal inflammation and luminal bacterial concentrations 7. The quantitative bacterial culture technique used in the current study measures both luminal and adherent bacteria 18 but does not distinguish the relative proportion of adherent organisms, which has been reported to correlate with regional intestinal inflammation in colonic biopsies 4.
The extensive inflammation and possible partial obstruction observed in the duodena of B. animalis- monoassociated IL-10−/− mice is noteworthy. IL-10−/− mice housed in our SPF facility lack duodenal inflammation 6 although Kuhn et al described extensive duodenal and small bowel intestinal inflammation in IL-10−/− mice housed under conventional, but not SPF conditions 36. The moderate duodenitis observed by Kim et al in E. faecalis-monoassociated IL-10−/− mice was not noted until after 30 weeks of colonization, whereas distal colitis was first observed at 10–12 weeks and was well advanced by 22 weeks 7. In contrast, B. animalis colonized IL-10−/− mice exhibit extensive duodenal inflammation after 20–24 weeks with only mild inflammation in the colon. GF HLA-B27 transgenic rats develop mild/moderate antral and proximal duodenal inflammation in addition to severe colitis one month after colonization with SPF fecal microbiota, but only mild/moderate colitis and no duodenal inflammation after monoassociation with Bacteroides vulgatus 15. Dohi et al described a model of chronic gastroduodenitis in TCR −/− mice housed under SPF conditions that were transfer recipients of IL-4−/− CD4+CD45RBhi T cells 37. These mice developed gastroduodenitis and colitis without obvious changes in the jejunum or ileum. Interestingly, the gastroduodenitis was only partially blocked with antibiotic treatment, whereas, colitis was completely inhibited, suggesting a role for nonviable antigens in the duodenal compartment 37. This is consistent with the observations that SAMP-1/Yit Fc mice continue to exhibit ileitis 38 and IL-2 deficient mice have gastroduodenitis 39 in GF conditions, although mice of both types have attenuated colonic inflammation in the absence of viable bacteria. The explanation for more active inflammation in the duodenum compared to other intestinal compartments that we evaluated remains unclear. One possibility is that incompletely degraded food components and fecal bacteria ingested through coprophagia could increase exposure of the duodenal mucosa to nonviable B. animalis antigens and adjuvants, thereby recruiting pathogenic immune cells to this region. As stated above, it is also possible that there is a difference in the extent of mucosal epithelial adherence of B. animalis in the duodenum compared to the colon, which warrants further study.
We demonstrate that CD4+ MLN T cells from B. animalis monoassociated IL-10−/− mice, but not WT mice, secrete IFN-γ and IL-17 when restimulated in vitro with APC previously incubated with B. animalis bacterial lysate. These T cells also secreted the same two cytokines, albeit to a lesser extent, in response to APC pulsed with lysates of B. infantis, B. bifidum and E. coli, but not E. faecalis bacterial lysate or an unrelated protein antigen, KLH. Therefore, there appears to be some immunologic cross-reactivity and/or shared bacterial antigens among the three Bifidobacterium species and E. coli. The higher recall response to B. animalis than to B. infantis or B. bifidum suggests that mucosal CD4+ T cells react to unique antigens expressed by B. animalis that are not shared by other Bifidobacterium species. Thus, antigenic heterogeneity among different bacterial isolates may influence their potential to initiate an inflammatory response. However, the presence of an immunogenic bacterial antigen may not be sufficient to cause inflammation. Sydora et al demonstrated that IL-10−/− mice monoassociated with Bacteroides vulgatus did not develop intestinal inflammation, but did generate a strong systemic T cell response to B. vulgatus antigens 23. The authors recently went on to show that disruption of the intestinal epithelial barrier of B. vulgatus monoassociated IL-10−/− mice with indomethacin allowed the bacteria to initiate and sustain an intense intestinal inflammatory response 40. These studies suggest that in order to cause inflammation, bacteria may require the capacity to breach the epithelial barrier to interact with lamina propria antigen presenting cells such as dendritic cells. Whether B. animalis ATCC 25527T is more invasive than other bifidobacteria remains to be determined. We evaluated the adjuvant stimulating capacity of our bacterial lysates and found no difference in the ability of the three Bifidobacterium lysates to stimulate IL-12/IL-23 p40 secretion by BMDC, further supporting our conclusion that the strong T cell recall response to B. animalis represents true bacterial antigen specificity.
The inflammation initiated by Bifidobacterium animalis subsp. animalis ATCC 25527T is characterized by a mild pancolitis and marked duodenitis. The fact that duodenal inflammation is not observed in IL-10−/− mice housed under SPF conditions 6 suggests that either the B. animalis that expand in these animals with disease progression are in some way different from strain ATCC 25527T, that the microbiota in the proximal gut of SPF mice prevent B. animalis from colonizing, attaching to or invading the duodenum, or that other members of the microbial community stimulate or inhibit the inflammatory response in a manner that is not reproduced by simple monoassociation. In conventional mice, the duodenum is populated by high numbers of lactobacilli 41 and these organisms can prevent inflammation in various animal models of IBD 11. Therefore, it is possible that lactobacilli prevent B. animalis from colonizing, attaching to the epithelium or invading the duodenum in SPF animals or may exert an anti-inflammatory effect at this site to prevent visible signs of duodenitis.
In summary, we have demonstrated that B. animalis, which has previously been shown to represent an increased proportion of the gut microbiota of colitic IL-10−/− mice, can induce intestinal inflammation and recall TH1 and TH17 immune responses after these animals have been monoassociated. This suggests that B. animalis may contribute to the colitis observed in SPF IL-10−/− mice. We do not suggest that our study implicates B. animalis as an unsafe probiotic organism for human consumption. Most strains utilized in this manner have undergone rigorous testing for safety and have no documented adverse effects 13, 42. However, our results indicate that even enteric bacterial species that are traditionally viewed to have probiotic activities also posess the potential to contribute to intestinal inflammation in a susceptible host.
NIH R01 DK53347 (RBS and SLT), National Institute of General Medical Sciences GM 00678 (JPM), the Center for Gastrointestinal Biology and Disease (CGIBD) NIH P30 DK 034987, and the National Gnotobiotic Rodent Resource Center NIH P40 R018603.
The authors thank Maureen Bower and Jerilyn Shaw of the National Gnotobiotic Rodent Resource Center (Department of Laboratory Medicine, UNC-CH) and Donna Kronstadt (Gnotobiotic Core, Center for Gastrointestinal Biology and Disease, College of Veterinary Medicine, NCSU) for maintaining the gnotobiotic mice; Joseph Galanko, Biostatistics Core, Center for Gastrointestinal Biology and Disease, UNC-CH; Bayley Crane (NCSU) for technical assistance; and the Center for Gastrointestinal Biology and Disease for providing administrative support. Portions of this work were presented in oral and poster presentations at the 2006 Digestive Diseases Week (American Gastroenterological Association) in Los Angeles, California that are published in abstract form (Gastroenterology 2006; 130: A6).