Commensal Bacteroides species induce colitis in host-genotype-specific fashion in a mouse model of inflammatory bowel disease

Abstract

The intestinal microbiota is important for induction of inflammatory bowel disease (IBD). IBD is associated with complex shifts in microbiota composition, but it is unclear whether specific bacterial subsets induce IBD and, if so, whether their proportions in the microbiota are altered during disease. Here, we fulfilled Koch's postulates in host-genotype-specific fashion using a mouse model of IBD with human-relevant disease-susceptibility mutations. From screening experiments we isolated common commensal Bacteroides species, introduced them into antibiotic-pretreated mice, and quantitatively reisolated them in culture. The bacteria colonized IBD-susceptible and -nonsusceptible mice equivalently, but induced disease exclusively in susceptible animals. Conversely, commensal Enterobacteriaceae were >100-fold enriched during spontaneous disease, but an Enterobacteriaceae isolate failed to induce disease in antibiotic-pretreated mice despite robust colonization. We thus demonstrate that IBD-associated microbiota alterations do not necessarily reflect underlying disease etiology. These findings establish important experimental criteria and a conceptual framework for understanding microbial contributions to IBD.

Figure 1. Antibiotic treatment quantitatively prevents colitis development in dnKO mice

(A) Representative images of H&E stained rectal histology of 4-week-old untreated and antibiotic-treated Il10r2^+/− and dnKO mice (distal 0.5 cm of colon). Abx = antibiotics (metronidazole + ciprofloxacin in drinking water). Scale bar = 100 μm. (B and C) Cecum and transverse colon pathology scores of untreated and antibiotic-treated Il10r2^+/− and dnKO mice as described in Figure 1A. Intestinal whole mounts were scored for gross pathology in a blinded fashion by an anatomic pathologist (T.S.S.) according to a validated system: ranging from 0 (no pathology) to 3 (severe pathology; see Figure S1 for details). Individual (squares) and median (bars) pathology scores are displayed. Kruskal-Wallis test: (B) H₃ = 13.75, p = 0.0033; (C) H₃ = 13.75, p = 0.0033. (D to G) Transverse colon and rectum crypt heights and crypt widths of mice in described in Figures 1B and 1C displayed as mean +/− SEM. 1-way ANOVA with post-hoc Tukey’s test: (D) F_3,11 = 40.87, p < 0.0001; (E) F_3,11 = 10.71, p = 0.0014; (F) F_3,11 = 33.25, p < 0.0001; (G) F_3,11 = 9.501, p = 0.0022. All statistically significant pairwise comparisons are displayed: **, p < 0.01, ***, p < 0.005. See also Figure S1

Figure 2. A non-germ-free system for colitis induction in antibiotic-pretreated dnKO mice

(A) Experimental timeline. Antibiotic treatment began at weaning (age 3 weeks); mice were antibiotic-treated for ≥3 weeks and gavaged 2 days after treatment cessation (Abx = antibiotics). (B) Representative images of H&E stained rectal histology of antibiotic-pretreated dnKO mice orogastrically gavaged with sterile PBS or with a standardized frozen stock of intestinal contents from untreated animals. Gavage dose = 2×10⁷ total cfu/mouse on Anaerobic Reducible Blood (ANB) agar (non-selective anaerobic culture media; cfu = colony-forming unit). Scale bar = 100 μm. (C and D) Cecum and transverse colon pathology scores of antibiotic-pretreated dnKO mice gavaged with sterile PBS or intestinal contents from untreated donors as described in Figure 2B displayed as individual (symbols) and median (bars) scores. Mann-Whitney U-test. (E to F) Transverse colon and rectum crypt heights and crypt widths of mice described in Figure 2C and 2D displayed as mean +/− SEM. Unpaired t-test. See also Figure S2

Figure 3. Mixed cultures of intestinal contents induce disease in antibiotic-pretreated dnKO mice

(A) Diagram of preparation of mixed cultures of intestinal contents. An aliquot of frozen intestinal contents (see Figure 1C) was thawed, serially diluted, and cultured on the indicated media types. 3-day mixed cultures were harvested from the indicated culture plates (in squares) and frozen in 20% glycerol. Aliquots of each culture were subsequently thawed and adjusted to standardized doses for gavage. (B and C) Cecum and transverse colon pathology scores of antibiotic-pretreated dnKO mice gavaged with sterile PBS or with the indicated anaerobic mixed cultures of intestinal contents mixed 1:1 with aerobic cultures grown on chocolate agar at 0 dilution. Individual (symbols) and median (bars) pathology scores are displayed. Gavage doses: ANB = 6.4×10⁷ total cfu/mouse; LKV = 7.3×10⁷ total cfu/mouse; CNA = 5.4×10⁷ total cfu/mouse. Statistical significance relative to PBS determined by Dunn’s multiple comparison test: n.s., p > 0.05; *, p < 0.05; **, p < 0.01. (D and E) Transverse colon and rectum crypt heights of antibiotic-pretreated dnKO mice described in Figure 3B and 3C, displayed as mean +/− SEM. Statistical significance relative to PBS determined by Dunnett’s multiple comparison test. The screen was unrepeated. See also Table S1 and S2

Figure 4. Commensal Bacteroides induce disease in antibiotic-pretreated dnKO but not Il10r2^+/− mice

(A) Screen for colitis induction by Bacteroidetes isolates. Cecum gross pathology scores of antibiotic-pretreated dnKO mice gavaged with 1×10⁸ cfu/mouse of pure cultures of the indicated primary bacterial isolates (see Table S3). Not repeated for B. uniformis and P. goldsteinii. (B) Representative H&E stained cecal histology of mice gavaged with pure cultures of the indicated bacterial isolate as described in Figure 4A. Scale bar = 200 μm. Cecal histology of a non-colitic Il10r2^+/− mouse gavaged with B. thetaiotaomicron is shown for comparison. (C and D) Cecum and transverse colon pathology scores of antibiotic-pretreated mice of the indicated genotypes gavaged with PBS or a pure culture of the Bacteroides thetaiotaomicron isolate (7×10⁷ cfu/mouse). Kruskal-Wallis test with post-hoc Dunn’s test. (C) H₂ = 12.26, p = 0.0022; (D) H₂ = 11.36, p = 0.0034. All significant pairwise comparisons are displayed: *, p < 0.05. (E to H) Transverse colon and rectum crypt heights and crypt widths of mice described in Figure 4C and 4D displayed as mean +/− SEM. 1-way ANOVA with post-hoc Tukey’s test: (E) F_2,10 = 8.596, p = 0.0067; (F) F_2,10 = 12.55, p = 0.0019; (G) F_2,10 = 2.057, p = 0.1786; (H) F_2,10 = 12.05, p = 0.0022. All significant pairwise comparisons are displayed if the omnibus p-value by ANOVA was significant: *, p < 0.05; **, p < 0.01. (I) Graph of the number of neutrophils per 50 well-oriented crypts located within the intestinal epithelium/crypt lumen in colons of the mice described in 4C and 4D. Kruskal-Wallis test: H₂ = 6.529, p = 0.0382. (J) Percent of CD4+ T-cells expressing Ifnγ, determined using validated procedures (Kang et al., 2008), in the colonic lamina propria of antibiotic-treated Il10r2^+/− (n=2) and dnKO (n=4) mice or antibiotic-pretreated Il10r2^+/− (n=3) and dnKO (n=3) mice gavaged with 10⁸ cfu/mouse of B. thetaiotaomicron and sacrificed 3 weeks post-gavage. 1-way ANOVA with post-hoc Tukey’s test: F_3,8 = 27.93, p = 0.0001. All significant pairwise comparisons are displayed: **, p < 0.01; ***, p < 0.005. (K) Immunoblots of B. thetaiotaomicron lysate probed with a 1:200 dilution of pooled serum collected from antibiotic-pretreated Il10r2^+/− (n=3) or dnKO (n=8) mice immediately prior to gavage with 10⁸ cfu/mouse of B. thetaiotaomicron (pre-gavage) or at sacrifice 3 weeks later (post-gavage). Secondary antibody is α-mouse IgG. Equal amounts of bacterial lysate were separated by SDS-PAGE, transferred to nitrocellulose membranes, blotted, and exposed in parallel for all groups See also Table S3 and Figure S3

Figure 5. Intestinal Bacteroides are eliminated by antibiotics and stably re-colonize gavaged, antibiotic-pretreated mice independently of host genotype

(A) Plot of fecal titers from untreated and antibiotic-treated dnKO mice grown in parallel anaerobically on non-selective (ANB) agar and Bacteroides bile esculin (BBE) agar. Titers from individual mice (symbols) and means of log₁₀-transformed titers (bars) are displayed. Unpaired t-test (log₁₀-transformed titers). (B) Plot of serial fecal titers on BBE agar from antibiotic-pretreated mice gavaged on Day 0 with sterile PBS or B. thetaiotaomicron (7×10⁷ cfu/mouse). Individual titers (symbols) and means of log₁₀-transformed titers (lines) are displayed. Unpaired t-tests comparing log₁₀-transformed fecal titers from B. thetaiotaomicron-gavaged dnKO mice to Il10r2^+/− mice. (C) Frequency histogram of all 92 B. thetaiotaomicron fecal titer measurements (log₁₀-transformed) we have performed on individual, single-isolate-gavaged, antibiotic-pretreated mice regardless of mouse genotype. Fecal titers were performed between 6 and 20 days post-gavage; data are compiled from multiple independent experiments with 4–9 mice per experiment. Gavage doses ranged from 6.6×10⁷ to 8.6×10⁸ cfu/mouse. Excluding the single outlier value (8.78), the titers are normally distributed as assessed by the D’Agostino & Pearson K² omnibus normality test: p = 0.5894, K² = 1.057, skewness = −0.153, kurtosis = 0.3345. Statistical analysis performed using GraphPad Prism v5.01 (GraphPad Software). See also Figure S4

Figure 6. Commensal Enterobacteriaceae are enriched in spontaneous colitis

(A to C) Plot of fecal titers from untreated or antibiotic-treated mice of the indicated genotypes grown in parallel anaerobically on ANB and BBE agar and aerobically on MacConkey agar (selective for Enterobacteriaceae). Il10r2^+/− and dnKO mice were co-housed; data compiled from ≥2 cages/group. 1-way ANOVA (ANB titers: F_3,13 = 1.344, p = 0.3031) or unpaired t-test (log₁₀-transformed titers). (D and E) Proportions of total cultivable bacteria in Figure 6A to 6C that grew on BBE agar and MacConkey agar, calculated by dividing BBE and MacConkey titers by ANB titers. Data is displayed as individual (symbols) and means of log₁₀-transformed proportions (bars). Unpaired t-test (log₁₀-transformed proportions). (F) Plots of fecal samples from untreated and antibiotic-treated dnKO and Il10r2^+/− mice (n = 6 per group) analyzed by quantitative PCR using primers specific for the B. thetaiotaomicron 16S rRNA gene. Il10r2^+/− and dnKO mice within each treatment group were co-housed; data are compiled from ≥3 cages/group. Results of qPCR assays were normalized to sample weight. Mann-Whitney U-test (log₁₀-transformed values). See also Figure S5

Figure 7. A colitis-enriched commensal Enterobacteriaceae isolate robustly colonizes antibiotic-pretreated dnKO mice without disease induction

(A and B) Plot of serial fecal titers grown on BBE and MacConkey agar from antibiotic-pretreated dnKO mice (5–6 per group) gavaged with PBS, B. thetaiotaomicron (1×10⁸ cfu/mouse), or Escherichia coli (2×10⁸ cfu/mouse). Bacterial identity was determined by colony characteristics, confirmed by sequencing the 16S rRNA gene of representative colonies. (C and D) Cecum and transverse colon pathology scores of mice described in Figures 7A and 7B displayed as individual (symbols) and median (bars) scores. Kruskall-Wallis test with post-hoc Dunn’s tests: (C) H₂ = 11.49, p = 0.0032; (D) H₂ = 12.02, p = 0.0025. All significant pairwise comparisons are displayed: *, p < 0.05; **, p < 0.01. (E to H) Transverse colon and rectum crypt heights and crypt widths from mice described in Figures 7A and 7B displayed as mean +/− SEM. 1-way ANOVA with post-hoc Tukey’s test: (E) F_2,14 = 2.580, p = 0.1112; (F) F_2,14 = 5.434; p = 0.0179; (G) F_2,14 = 6.438, p = 0.0104; (H) F_2,14 = 7.268, p = 0.0068. All significant pairwise comparisons are displayed if omnibus p-value by ANOVA was significant: *, p < 0.05. See also Figure S6.