The gut microbiome regulates the clinical efficacy of sulfasalazine therapy for IBD-associated spondyloarthritis

Abstract

Sulfasalazine is a prodrug known to be effective for the treatment of inflammatory bowel disease (IBD)-associated peripheral spondyloarthritis (pSpA), but the mechanistic role for the gut microbiome in regulating its clinical efficacy is not well understood. Here, treatment of 22 IBD-pSpA subjects with sulfasalazine identifies clinical responders with a gut microbiome enriched in Faecalibacterium prausnitzii and the capacity for butyrate production. Sulfapyridine promotes butyrate production and transcription of the butyrate synthesis gene but in F. prausnitzii in vitro, which is suppressed by excess folate. Sulfasalazine therapy enhances fecal butyrate production and limits colitis in wild-type and gnotobiotic mice colonized with responder, but not non-responder, microbiomes. F. prausnitzii is sufficient to restore sulfasalazine protection from colitis in gnotobiotic mice colonized with non-responder microbiomes. These findings reveal a mechanistic link between the efficacy of sulfasalazine therapy and the gut microbiome with the potential to guide diagnostic and therapeutic approaches for IBD-pSpA.

Keywords: Faecalibacterium prausnitzii; butyrate; folate; inflammatory bowel disease; microbiome; spondyloarthritis; sulfasalazine.

Graphical abstract

Figure 1

The gut microbiome stratifies sulfasalazine (SAS) BASDAI responders from non-responders with IBD-pSpA (A and B) Mean Δ BASDAI between baseline and week 12 of controls compared to SAS treatment (A) and SAS treatment stratified by IBD diagnosis of Crohn's disease (CD) or ulcerative colitis (UC). (B) Red line represents the cutoff defining clinical response. Error bars represent the SEM. ∗∗p ≤ 0.01, t test. (C) Linear correlation between fecal abundance of sulfapyridine (SP) and Δ BASDAI. (D) Boxplots comparing alpha diversity, based on Shannon index, between baseline and week 12. Wilcoxon matched-pairs signed-rank test was not significant (ns). (E) Principal-coordinate analysis plot is shown using Bray-Curtis and stratified by BASDAI clinical response and time of sample collection. Monte Carlo, permutational multivarate analysis of variance (PERMANOVA) p values are shown. (F) Bar plot displays the differentially abundant microbial species between responders and non-responders identified by LEfSe at baseline (p < 0.05, Mann-Whitney). Modules with a linear discriminant analysis (LDA) score >2 are plotted. (G) ROC curves demonstrating the ability of the indicated bacterial taxa in discriminating responders from non-responders in a separate validation cohort (n = 16). AUC is indicated. ∗p ≤ 0.05 and ∗∗p ≤ 0.01.

Figure 2

Baseline differences in butyrate synthesis pathway stratify SAS clinical response (A) Principal-coordinate analysis plot is shown using Bray-Curtis and stratified by BASDAI clinical response and time of sample collection. Monte Carlo, PERMANOVA p values are shown. (B) Boxplots comparing median carbohydrate degradation sub-pathway abundance (copies per million) between responder and non-responder at baseline (see Figure S3A, marked in bold). 1: dTDP L-rhamnose biosynthesis I; 2: superpathway of hexuronide and hexuronate degradation; 3: D-galacturonate degradation I; 4: superpathway of β-D-glucuronide and D-glucuronate degradation; 5: D-galactose degradation V (Leloir pathway); 6: sucrose degradation III (sucrose invertase); 7: glutaryl-CoA degradation; 8: galactose degradation I (Leloir pathway); 9: four-deoxy-L-threo-hex-4-enopyranuronate degradation; 10: starch degradation V; and 11: D-fructuronate degradation. Boxplots present the median, 25th, and 75th percentiles, and Mann-Whitney p < 0.05 is shown. (C) Schematic of butyrate synthesis pathway from glutaryl-CoA degradation. ccr, crotonyl-CoA reductase; gcd, glutaryl-CoA dehydrogenase; but, butyryl-CoA:acetate CoA transferase; buk, butyrate kinase. (D) Relative abundance of fecal short-chain fatty acid was determined by mass spectrometry and normalized per mg fecal sample used. t test, ∗p < 0.05 and ∗∗p < 0.01 is shown. (E) Baseline but and buk mean abundances (copies per million) between responders and non-responders. Error bars represent SEM. ∗p < 0.05, t test.

Figure 3

SP-induced folate stress regulates Faecalibacterium prausnitzii expression of butyrate synthesis pathway genes (A) Average percentage of but reads from the IBD-pSpA cohort at baseline. Metagenomic sequences aligned by bacterial species based on 90% homology. (B and C) F. prausnitzii grown anaerobically with sub-inhibitory dose of SP (1 mM) or sodium hydroxide vehicle (control [C]) over 46 h. (C) Principal-component plot from RNA-seq analysis performed at h 46 depicted in (B). On average, four million reads mapped to F. prausnitzii (∼10× genome coverage). (D) Schematic highlights pathway and enzymes regulating butyrate synthesis. Normalized counts (log₂ copies per millions) of genes related to the F. prausnitzii butyrate synthesis pathway genes. Error bars represent the SEM. ∗p < 0.05 and ∗∗∗∗p < 0.0001, DESeq (p adjusted, false discovery rate [FDR]). (E) Butyrate abundance measured by mass spectrometry following in vitro culture of for 22 or 46 h with either vehicle C or SP treatment as indicated. Technical replicates are shown, ∗∗p < 0.01, t test. (F) Fold change of but transcription in either F. prausnitzii or E. rectale at 22 h with either vehicle C or SP treatment as indicated. Technical replicates are shown, ∗p < 0.05, t test. (G) In vitro F. prausnitzii but expression was measured at 30 h post-bacteria exposure to SP, folate, and/or sodium hydroxide vehicle (C) by qPCR. Scatterplot error bars represent the SEM. ∗∗∗∗p < 0.0001, ANOVA, Tukey’s multiple comparison test. (H) Boxplot comparing the median deoxyuridine and thymine fecal abundance within BASDAI clinical response groups (R, responders; NR, non-responders). Boxplots present the median, 25th, and 75th percentiles; ∗p < 0.05; Wilcoxon matched-pairs signed-rank test.

Figure 4

SAS promotes butyrate production and limits colitis (A) Fecal acetate, butyrate, and propionate levels of germ-free (GF) or Jackson specific-pathogen-free (SPF) mice at day 14 post-treatment initiation with SAS, balsalazide (BSD), or water (H₂O) C. Graphs show data from 2 independent experiments. Error bars represent SEM. ∗∗p < 0.01; ANOVA with Tukey’s multiple comparison test. (B and C) Jackson SPF mice were treated with SAS or H₂O vehicle for 7 days and then exposed to 2% DSS ad libitum for 7 days. Treatments were maintained throughout the experiment. Percentage of survival (B) and lipocalin (C) are shown. Graph shows data from 2 independent experiments (H₂O, n = 9; SAS, n = 8; BSD, n = 9). (D and E) SPF Gpr109a^−/− and heterozygous (Het) littermate controls were treated with SAS or H₂O vehicle for 7 days and then exposed to 2% DSS ad libitum for 7 days. SAS treatment was maintained throughout the experiment. Percentage of survival (D) and cecal lipocalin (E) are shown. Graph shows data from 2 independent experiments (Het-H₂O, n = 7; Het-SAS, n = 6; Gpr109a^−/−-H₂O, n = 10; Gpr109a^−/−-SAS, n = 9). Survival analysis; ∗p < 0.05 and ∗∗p < 0.01; log-rank (Mantel-Cox) test. Boxplots present the median, 25th, and 75th percentiles; ANOVA with Kruskal-Wallis multiple comparison test; p values are shown.

Figure 5

SAS reduces colitis in gnotobiotic mice colonized with responder microbiomes or non-responder microbiomes with F. prausnitzii (A) Schematic of gnotobiotic mouse colonization. (B and C) Fecal butyrate per mg feces at 7 days after colonization prior to initiation of DSS. One representative of three responders (B) or non-responders (C) is shown as indicated. Error bars represent SEM; ∗p < 0.05, t test. (D–F) GF mice received fecal microbial transplants (FMTs) from three responders (D) or three non-responder subjects (E and F). F. prausnitzii group was orally gavaged with F. prausniztii prior to DSS (E and F). SAS treatment was initiated 14 days post-FMT and maintained throughout the experiment. Mice were exposed to 2% DSS ad libitum for 7 days starting 7 days post-SAS initiation. Weight loss and levels of lipocalin in cecal contents are shown (E and F). Graphs show average data (thick line) of three individual donors (thin line). Each thin line is average of n = 3–5 mice/donor. Dark lines are the average and SEM of pooled data from all three donors. Total responder-H₂O, n = 12; responder-SAS, n = 14; non-responder-H₂O, n = 10; non-responder-SAS, n = 13; non-responder + F. prausnitzii, n = 9; non-responder-SAS + F. prausnitzii, n = 9). Mixed-effects model (weight loss) or Kruskal-Wallis test with multiple comparison (lipocalin). p values are shown. Error bars represent SEM; ∗p < 0.05, ∗∗p < 0.01, and ∗∗∗∗p < 0.0001.