Microbiota-dependent indole production stimulates the development of collagen-induced arthritis in mice

Abstract

Altered tryptophan catabolism has been identified in inflammatory diseases like rheumatoid arthritis (RA) and spondyloarthritis (SpA), but the causal mechanisms linking tryptophan metabolites to disease are unknown. Using the collagen-induced arthritis (CIA) model, we identified alterations in tryptophan metabolism, and specifically indole, that correlated with disease. We demonstrated that both bacteria and dietary tryptophan were required for disease and that indole supplementation was sufficient to induce disease in their absence. When mice with CIA on a low-tryptophan diet were supplemented with indole, we observed significant increases in serum IL-6, TNF, and IL-1β; splenic RORγt+CD4+ T cells and ex vivo collagen-stimulated IL-17 production; and a pattern of anti-collagen antibody isotype switching and glycosylation that corresponded with increased complement fixation. IL-23 neutralization reduced disease severity in indole-induced CIA. Finally, exposure of human colonic lymphocytes to indole increased the expression of genes involved in IL-17 signaling and plasma cell activation. Altogether, we propose a mechanism by which intestinal dysbiosis during inflammatory arthritis results in altered tryptophan catabolism, leading to indole stimulation of arthritis development. Blockade of indole generation may present a unique therapeutic pathway for RA and SpA.

Keywords: Amino acid metabolism; Autoimmunity; Bacterial infections; Mouse models.

Figure 1. Intestinal metabolomics profiling identifies microbiome-mediated alterations in the tryptophan pathway in mice with CIA.

CIA was induced in male 6-week-old DBA/1 mice, and cecal contents were harvested at day 35 from mice with CIA, mice with CIA that were depleted of microbiota by antibiotic administration after day 21, or untreated DBA/1 mice (Untx DBA). (A and B) LC-MS/MS was used to screen 244 metabolites from cecal contents of mice with CIA (n = 5), mice with CIA+Abx (n = 7), or untreated DBA/1 mice (n = 6). (A) PLSDA plot of CIA mice versus CIA+Abx mice versus untreated mice. (B) Volcano plot of CIA+Abx (left) versus CIA (right). (C and D) HPLC was used to quantify Trp pathway metabolites (indicated on the y axis) from cecal contents of mice with CIA (n = 3), mice with CIA+Abx (n = 7), or untreated DBA/1 mice (n = 6). All data are reported as individual mice (symbols) and the mean ± SEM (bars) after normalization to the weight (mg) of the cecal contents. *P < 0.05, **P < 0.01, and ****P < 0.0001, by 1-way ANOVA with Bonferroni’s correction for multiple comparisons. (E) Graphical representation of Trp metabolism pathways showing Trp metabolites identified in the LC-MS/MS analysis (A and B) and HPLC analysis (C and D). The log₂(fold change [FC]) was calculated for CIA versus CIA+Abx and is represented by a color gradient from yellow (greater increase in CIA) to blue (greater increase in CIA+Abx). The size of each circle represents the –log₁₀(P value) of an unpaired Student’s t test between CIA versus CIA+Abx. Kyn, kynurenine; 2-OA, 2-oxoadipate. The pound signs in E denote trends in metabolites that were also observed in Isolate 7-colonized mice. Lines denote pathways: black, indole; dashed line, serotonin; gray, kynurenine.

Figure 2. Indole is required for CIA.

(A) CIA was induced in 6-week-old male DBA/1 mice. On days 21–35, mice were treated with antibiotics with or without 0.1 mg/mL indole in the drinking water, and arthritis scores were assessed every other day. n = 10 (Abx+Indole); n = 7 (CIA+Abx). (B) Six-week-old male DBA/1 mice were fed a TL or TS diet starting on day –1 and through the duration of the experiment. Following induction of CIA, mice were treated with indole (200 μL of a 10 mM solution) or vehicle control (0.33% methanol) by oral gavage every other day starting on day 0. Arthritis scores were assessed every other day from days 21–35. n = 39 (TL+Vehicle), n = 24 (TL+Indole), and n = 25 (TS+Vehicle), pooled from 5 independent experiments. Red asterisks: TL+Indole versus TL+Vehicle; black asterisks: TS+Vehicle versus TL+Vehicle. TS+Vehicle versus TL+Indole was not statistically significant. (C) The sum of the inflammation, pannus, and bone erosion score of H&E-stained paws is plotted as the total histology score (maximum score of 15). n = 10–20, pooled from 2 independent experiments. (D) Schematic of Trp breakdown into indole by bacterial Tryptophanase A and Trp synthesis from indole by bacterial Tryptophan synthase. (E) HPLC analysis of Trp in cecal contents from mice with CIA at day 35, plotted as the AUC, normalized to the weight (mg) of the cecal contents. n = 5–10, from 1 experiment. (F) Six-week-old male germ-free DBA/1 mice were colonized with E. coli BW25113 mutants (ΔtnaA or ΔBcsQ) with 10⁸ CFU by oral gavage on day –7 before CIA induction. n = 7 per group. (G) HPLC analysis of indole in cecal contents from CIA mice colonized with either ΔtnaA or ΔBcsQ at CIA day 35. Indole levels were plotted a the AUC per milligram of cecal content weight. Data are reported as the mean ± SEM. *P < 0.05, **P < <0.01, ***P < 0.001, and ****P < 0.0001, by 2-way ANOVA with Bonferroni’s correction for multiple comparisons (A, B, and F), 1-way ANOVA with Bonferroni’s correction for multiple comparisons (C and E), and unpaired 2-tailed Student’s t test (G).

Figure 3. Indole minimally affects bacterial dysbiosis imparted by a TL diet during CIA.

Six-week-old male DBA/1 mice were fed a TL or TS diet starting on day –1 and through the duration of the experiment. CIA was induced, and indole (200 μL of a 10 mM solution) or vehicle control (0.33% methanol) was added by oral gavage every other day starting on day 0. On day 35, fecal pellets were harvested, genomic DNA was isolated, and 16S rRNA-Seq was performed to assess microbial diversity for TS+Vehicle (TS, n = 5), TL+Vehicle (TL, n = 10), and TL+Indole (TLI, n = 10). (A–C) α-Diversity indices are shown for each group. *P < 0.05 and **P < 0.01, by 1-way ANOVA with pairwise P values determined by Tukey’s honest significance difference tests for differences between groups. Richness: measured by the Chao1 index. Shannon Diversity: measured by Shannon’s diversity index, H. Evenness: measured by H/Hmax where Hmax = the maximum H for a subject. (D) PCA, in which smaller, lighter symbols represent individual mice and large, darker symbols represent group means plus 95% CIs for PC1 and PC2. (E) Bar charts showing mean distributions of taxa for each group. Taxa with relative abundances of less than 1.0% were collapsed into the “Other” category to simplify the figure. OTU, operational taxonomic units. Differences in β-diversity between groups were assessed using PERMANOVA tests with the weighted Aitchison dissimilarity index: **P < 0.01 and ***P < 0.001. (F–I) Volcano and effect size plots generated by ANOVA-like differential expression (ALDEx2) analysis indicate taxa that were significantly enriched or depleted (FDR-corrected P value < 0.05) in mice with CIA on 1 diet compared with another: TS+Vehicle versus TL+Vehicle (F and G), TL+Vehicle versus TL+Indole (H and I).

Figure 4. Indole alters the cytokine profile in CIA.

Six-week-old male DBA/1 mice were fed a TL or TS diet starting on day –1 and through the duration of the experiment. Following induction of CIA, mice were treated with indole (200 μL of a 10 mM solution) or vehicle control (0.33% methanol) by oral gavage every other day starting on day 0. (A–F) Terminal serum was collected at days 14 and 21 and at the plateau of disease (days 35–50) from mice with CIA fed a TL or TS diet and treated with indole (200 μL of a 10 mM solution) or vehicle control (0.33% methanol). Sera were analyzed by an 8-plex immunoassay (Meso Scale). To account for experiment-to-experiment variability, serum cytokine concentrations (as denoted on the y axis) were normalized to the mean of the TS+Vehicle group for each experiment and time point. n = 4–8 per group (day 14), 10 per group (day 21), and 16–28 per group (days 35–50), pooled from 9 independent experiments and plotted as individual mice (symbols) and the mean ± SEM (bars). *P < 0.05 and ***P < 0.001, by 1-way ANOVA with Bonferroni’s correction for multiple comparisons.

Figure 5. Indole alters complement activation, IgG subclass, and glycosylation.

Six-week-old male DBA/1 mice were fed a TL or TS diet starting on day –1 and through the duration of the experiment. Following induction of CIA, mice were treated with indole (200 μL of a 10 mM solution) or vehicle control (0.33% methanol) by oral gavage every other day starting on day 0. (A) Day 35 serum was evaluated by ELISA for C3 binding to anti–CII IgG. n = 5–10 per group from 1 independent experiment. (B) Formalin-fixed, paraffin-embedded (FFPE) joints were stained for complement C3 by immunohistochemistry, and staining intensity was scored. Each data point represents the average complement deposition score of all 4 paws for 1 mouse (maximum score = 3 per paw). n = 10–15 per group, pooled from 2 independent experiments. (C and D) Day 35 serum was evaluated by ELISA for anti–CII IgG2a (C) and anti–CII IgG2b (D). n = 5–10 per group from 1 independent experiment. (E) Diagram of possible glycosylation patterns on N297 of the IgG Fc domain. Blue squares denote N-acetylglucosamine; green circles denote mannose; yellow circles denote galactose; purple diamonds denote sialic acid. (F and G) Total IgG was purified from serum, and IgG glycosylation patterns were assessed by LC-MS/MS. The percentage of galactosylation (Gal) and the percentage of sialylation (Sia) are plotted, respectively. Galactosylation and sialylation were calculated as a percentage of all glycoforms (G0, G1, G2, S1, and S2). n = 8 per group from 1 independent experiment. (H and I) In a separate experiment, anti–CII IgG was purified using CII-linked CNBr Sepharose 4B beads. IgG glycosylation patterns were assessed by LC-MS/MS. Galactosylation and sialylation are plotted as the percentage of G1, G2, S1, and S2 glycoforms only. n = 5–10 per group from 1 representative experiment. For all panels, values are plotted as individual mice (symbols) and the mean ± SEM (bars). *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001, by 1-way ANOVA with Bonferroni’s correction for multiple comparisons (A–D, H, and I) and unpaired Student’s t test (F and G). rel., relative.

Figure 6. Indole skews toward Th17 cells.

Six-week-old male DBA/1 mice were fed a TL or TS diet starting on day –1 and through the duration of the experiment. Following induction of CIA, mice were treated with indole (200 μL of a 10 mM solution) or vehicle control (0.33% methanol) by oral gavage every other day starting on day 0. Spleens were harvested on day 35 for analysis by flow cytometry. (A) Splenic naive (CD44^–CD62L⁺) T cells as a percentage of CD4⁺ T cells. (B) Splenic effector T cells (CD44⁺CD62L^–) as a percentage of CD4⁺ T cells. (C) Splenic central memory T cells (CD44⁺CD62L⁺) as a percentage of CD4⁺ T cells. n = 5–10 per group from 1 independent experiment for A–C. (D) Splenic FoxP3⁺RORγt^–CD25⁺ Tregs are plotted as a percentage of total CD4⁺ T cells. (E) Splenic CD3⁺CD4⁺FoxP3^–RORγt⁺ Th17 cells are plotted as a percentage of total CD4⁺ T cells. (F) Ratio of splenic Th17 to Tregs at CIA days 21 and 35. n = 10–20 per group, pooled from 2 independent experiments (day 35 [d35]) (D–F) and n = 10 per group (day 21 [d21]) from 1 independent experiment (F). (G and H) Total splenocytes from CIA day 35 were harvested and restimulated (stim) with bovine type II collagen (G) or CD3/CD28 Dynabeads (H); supernatant was saved and IL-17A/F was measured by Meso Scale Discovery (MSD). One statistical outlier was identified in the TS+Vehicle group using Grubb’s test and was excluded from analysis. n = 4 (TS+Vehicle) and n = 5 (TL+Indole, TL+Vehicle) per group from 1 independent experiment. (I) TL+Indole-treated mice received i.p. injections of 100 μg anti–IL-23p19 or isotype (anti-HRP) on CIA days 0, 7, 14, and 21, and CIA severity was monitored. n = 10 (TL+Indole+anti–IL-23), n = 10 (TL+Indole+Isotype), and n = 4 (TS+Vehicle). Asterisks show the comparison between TL+Indole⁺Isotype versus TL+Indole+anti–IL-23. For all panels, values are plotted as individual mice (symbols) and the mean ± SEM (bars). *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001, by 1-way ANOVA with Bonferroni’s correction for multiple comparisons (A–H) or 2-way ANOVA with Bonferroni’s correction for multiple comparisons (I).

Figure 7. Indole-stimulated human intestinal cells also highlight changes in function.

LPMCs isolated from healthy human colon tissue were stimulated with 1 mM indole or vehicle for 4 hours. CD19⁺ B cells and CD3⁺ T cells were flow sorted, and RNA was isolated for RNA-Seq. Differentially expressed pathways (indole versus vehicle) were identified with Ingenuity Pathway Analysis for (A) CD19⁺ B cells and (B) CD3⁺ T cells. n = 5 paired samples.