Microbiota-derived indoles alleviate intestinal inflammation and modulate microbiome by microbial cross-feeding

C57BL/6 mice (6–8 weeks old) were obtained from Beijing HFK Bioscience, while IL-10^−/− mice (6–8 weeks old) were purchased from Guangzhou Cyagen Biosciences. All mice were housed under a 12-h light/dark cycle and fed a standard chow diet and water ad libitum.

L. reuteri I5007 was isolated from the colonic mucosa of healthy weaning piglets and further confirmed through whole-genome sequencing (NCBI ID: 1340495); this strain was found to be chloramphenicol, kanamycin, and streptomycin-resistant. L. reuteri I5007, mutant L. reuteri I5007, L. johnsonii, L. plantarum, L. delbrueckii, and L. rhamnosus GG were cultured in MRS medium (Solarbio Science and Technology Co., Beijing, China) at 37 ℃ for 20 h under anaerobic conditions. Bacterial cells were collected by centrifugation at 12,000 rpm for 15 min. Heat-killed L. reuteri I5007 was prepared by heating at 95 °C for 20 min. No colonies were detected after 48 h of culture at 37 °C after heat treatment. Clostridium sporogenes (ATCC 15579; American Type Culture Collection) was cultured in fluid thioglycolate medium in an anaerobic atmosphere at 37 °C for 24 h. C. rodentium (ATCC 51459; American Type Culture Collection) was inoculated into LB broth and cultured in a shaker overnight at 37 °C. Overnight culture was then centrifuged at 12,000 rpm for 15 min, washed once, and resuspended at 10⁸ CFU/mL in PBS. L. reuteri I5007 was cultured anaerobically with or without tryptophan at 0.01, 0.1, or 1 mM at 37 °C for 20 h prior to measurement of indole derivatives in the supernatant.

ArAT-deficient strain of L. reuteri I5007 was constructed by homologous recombination as described previously [16]. Briefly, 871-bp and 850-bp DNA fragments were separately cloned from the upstream and downstream sequences of the ArAT gene of L. reuteri I5007 and then fused with an erythromycin resistance gene fragment amplified from the pMG36e plasmid. Recombinant DNA was cloned into the pLCNICK-2179 plasmid using TA cloning. The resulting plasmid was inserted into competent L. reuteri I5007 through electroporation, followed by plating onto the MRS medium agar plates containing erythromycin. An ArAT-knockout strain was obtained and confirmed for the absence of the ArAT gene by PCR of bacterial genomic DNA.

Fresh donor colonic contents were pooled and homogenized 5 times and diluted with sterile saline containing 20% (v/v) glycerol. The pooled samples were centrifuged at 500 × g for 3 min at 4 °C, and the supernatant was collected and stored at − 80 °C. Prior to colonic microbiota transplantation (CMT), glycerol in each sample was replaced with PBS by centrifugation and the number of live microbial cells was counted using methylene blue staining [16].

C57BL/6 mice were given 3% (wt/vol) DSS (36,000–50,000 Da, MP Biomedicals) in drinking water for 7 days. Water was changed every 2 days. The animals were weighed daily for the entire duration. The DAI was evaluated to assess the severity of colitis as previously described [17].

Mice were given an antibiotic cocktail of ciprofloxacin (0.2 g/L), vancomycin (0.5 g/L), and metronidazole (1 g/L) [18] in drinking water for 2 weeks prior to other treatments. Water was replenished every week.

Fluorescein isothiocyanate (FITC, 4 kDa, Sigma)-dextran was dissolved in PBS at 100 mg/ml. After 6 h of fasting, mice were orally administered with 600 mg/kg FITC-dextran. Blood was collected after 4 h and serum was prepared. The fluorescence of each serum sample was measured at an excitation wavelength of 490 nm and an emission wavelength of 530 nm.

The colon tissue was rolled, fixed with 4% paraformaldehyde for 24 h, embedded in paraffin wax, sliced (5 μm), and stained with hematoxylin and eosin (H&E) or alcian blue. The pathological analysis of H&E-stained sections was scored blindly by two independent investigators. The scoring system was as follows: ulcer size (0 = none, 1 = less than 3 mm, 2 = more than 3 mm), severity of inflammation (0 = none, 1 = mild, 2 = moderate, and 3 = severe), fibrotic degree (0 = none, 1 = mild, 2 = moderate, and 3 = severe), extent of epithelial/crypt damage (0 = none, 1 = basal 1/3, 2 = basal 2/3, 3 = crypt loss, 4 = crypt and surface epithelial destruction), and injury depth (0 = none, 1 = mucous layer, 2 = submucosa, 3 = muscle layer, 4 = serous layer).

Cell-free supernatant (CFS) of L. reuteri was obtained by centrifugation of the bacterial culture at 12,000 rpm for 15 min, followed by filtration through a 0.22-µm filter and dilution to 10% with the corresponding culture. For in vitro anaerobic culture of colonic microbiota, approximately 100 mg of the colonic content was vortexed in 5 ml BHI medium under anaerobic conditions. MRS medium (200 µl), CFS (10%), or ILA (1 mM) was then added to the bacterial culture and anaerobically incubated for 12 h at 37 °C in a GasPak EZ Anaerobe Pouch system. C. sporogenes was co-cultured with or without an addition of tryptophan (0.3g/L), the strain and CFS of L. reuteri I5007 or ArAT-deficient mutant L. reuteri I5007 (△L.R), ILA for 24 h at 37 °C in an anaerobic atmosphere.

The mesenteric lymph nodes (MLNs) were collected and prepared as single-cell suspensions. The MLNs were then passed through a 70-μm filter and resuspended in FACS buffer for flow cytometric analyses. Common lymphoid progenitor (CLP) cells were isolated as previously described [19]. Briefly, the colon was opened up longitudinally and washed to remove the contents by scraping gently with PBS. The tissue were then diced into small pieces and incubated in prewarmed sterile HBSS (without Ca²⁺ and Mg²⁺) containing 3% fetal bovine serum (FBS), 10 mM EDTA, and 5 mM dithiothreitol (DTT) for 30 min at 37°C in a shaking water bath. The cells, including immune cells and enterocytes, were recovered by filtration through a 100-μm filter. Cells were then dissociated from each other in prewarmed HBSS (with Ca²⁺ and Mg²⁺) containing 3% FBS, 1% L-glutamine, 1% penicillin–streptomycin, 10 mM HEPES, 0.5 mg/mL collagenase D, 0.5 mg/mL Dispase, and 0.04 mg/mL DNase I for 45 min at 37 °C while shaking. Cell suspensions were filtered through a 70-µm cell strainer into precooled sterile PBS, followed by centrifugation in a 40/80 Percoll gradient for 20 min at 1000 × g at 37°C. CLPs were obtained by collecting the cells at the interface and washing twice in FACS buffer for flow cytometric analysis.

For detection of Th1, Th2, Th17F, and Th17A cells, lymphocytes were stimulated for 4 h with 50 ng/mL PMA (Sigma) and 1 µg/mL ionomycin (Sigma) in the presence of GolgiStop (BD Biosciences). Cells were first stained for surface CD4 and then fixed and permeabilized with a BD Cytofix/Cytoperm Kit, followed by intracellular staining with the following antibodies: PerCP/Cy5.5-IL-4, FITC-IFN-γ, AF647-IL-17F, and PE/Cy7-IL-17A. To quantify Treg cells, a True-Nuclear Transcription Factor Buffer Set (BioLegend) was used to label cells with anti-CD4-PE, anti-CD25-FITC, and anti-FoxP3-Alexa Fluor. The percentage of each T cell type was measured with BD FACS Celesta and FlowJo analysis software (BD Biosciences). Compensation was set using UltraComp eBeads.

The colonic tissues were lysed with RIPA extraction buffer (Solarbio, China) containing a 1% protease inhibitor cocktail. Following centrifugation at 12,000 rpm for 15 min at 4°C, the supernatant was collected and the total protein concentration was quantified using a bicinchoninic acid (BCA) protein assay kit (Solarbio, China). Proteins were separated by SDS–polyacrylamide gel electrophoresis and transferred onto polyvinylidene difluoride (PVDF) membranes (Millipore, USA). The membranes were blocked with 5% skim milk for 1.5 h and incubated with primary antibodies against PXR (Abcam, USA), E-cadherin (Cell Signaling Technology, USA), Cyp1a1 (Cell Signaling Technology, USA), and occludin (Cell Signaling Technology, USA) overnight at 4°C. The membranes were washed and incubated with HRP-labeled secondary antibodies at room temperature for 2 h, followed by signal detection with an ECL Kit (Solarbio, China).

Total RNA from cells and tissue was extracted using TRIzol reagent (Takara, Japan). The concentration of RNA was determined with NanoDrop 2000 (Thermo Fisher Scientific, Wilmington, DE, USA). Quantitative PCR was performed as previously described [20] using gene-specific primers and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as the reference gene (Additional file 2: Table S2). The relative mRNA expression level of each target gene was determined using the 2^−△△Ct method.

Inflammatory cytokines in the serum and colonic tissue samples were measured with enzyme-linked immunosorbent assay (ELISA) kits (Cusabio, Baltimore, MD, USA) according to the manufacturer’s instructions.

The colonic contents (20 mg) were homogenized in 200 µl water and mixed with 600 µl of precooled methanol containing internal standard (1-methyl-tryptophan, 2.0 µg/ml). Serum and CFS samples (30 µL) were mixed with 120 µl methanol containing 1-methyl-tryptophan. The mixture was then vortexed for 5 min and centrifuged at 16,000 rpm for 10 min at 4℃. The supernatant was vacuum-dried and reconstituted with 100 µl of 10% methanol and centrifuged at 16,000 rpm for 15 min at 4℃ prior to the UHPLC–MS/MS analysis. An Agilent 1290 Infinity II UHPLC coupled to an Agilent 6470A triple quadrupole (QQQ) mass spectrometer was used for tryptophan metabolite analysis. The mobile phase consisted of 0.1% formic acid in water (A) and methanol (B). Chromatographic separation was conducted using a gradient elution program for the detection of tryptophan metabolites as follows: 0 min, 95% B; 2 min, 95% B; 8 min, 50% B; 10 min, 5.0% B; 12 min, 5.0% B; 13 min, 95% B; 16 min, 95% B. The gradient elution program for the detection of AAAs was as follows: 0 min, 95% B; 2 min, 95% B; 6 min, 50% B; 7.5 min, 5.0% B; 9 min, 5.0% B; 10 min, 95% B; and 15 min, 95% B. The column temperature was 35 °C, and the flow rate was 0.3 ml/min. For analysis, the eluate from the column was ionized by an electrospray ionization source in positive mode (ESI +) under the following settings: gas temperature: 300°C, gas flow: 5 L/min, nebulizer: 45 psi, ShenthGasHeater: 300 °C, SheathGasFlow: 10 L/min, and capillary voltage: 4000 V. Dynamic multiple reaction monitoring (dMRM) was used for quantitative analysis. Data were acquired and processed in Agilent MassHunter software (version B.04.00).

Total microbial DNA was extracted from the colonic contents using the QIAamp DNA Stool Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer’s instructions. Extracted DNA was quantified using a NanoDrop 2000 (Thermo Fisher Scientific, Wilmington, DE, USA). The V3–V4 hypervariable region of the 16S rRNA gene was amplified using universal primers 338F (5′-GTGCCAGCMGCCGCGG-3′) and 806R (5′-CCGTCAATTCMTTTRAGTTT-3′). Agarose gel electrophoresis was performed to confirm the size of PCR products. Amplicons were quantified, pooled, and sequenced on the Illumina MiSeq system (Illumina, San Diego, CA, USA) for paired-end reads. Sequence analysis was performed using the UNOISE pipeline through USEARCH v10.0 [21] and VSEARCH v2.15 [22]. Briefly, paired-end sequences were merged, quality filtered, and decomplexed using VSEARCH. Denoising and chimera removal were conducted using UNOISE3 to correct for sequencing errors. Assembled sequences were mapped back to the denoised chimera-free sequences as OTUs with 97% identity. The taxonomy of the features was classified using the USEARCH SINTAX algorithm in RDP training version 16 [23]. The sequencing reads of all samples were rarefied to the sample with the lowest reads to compensate for the difference in the sequencing depth. The α-diversity was determined from the rarefied OTU table based on the richness, Simpson, and Shannon indices. The β-diversity was estimated from the Bray–Curtis distances, which were applied to build distance matrices and reported according to the PCoA. PICRUSt2 was used to predict altered KEGG pathways based on 16S sequencing data [24]. For differential abundance analysis and association analysis, all bacterial taxa with a relative abundance of < 0.01% and present in < 20% of samples were removed.

Total microbial RNA was extracted from the colonic contents using the E.Z.N.A.® Soil RNA Midi Kit (Omega Bio-Tek, USA) following the manufacturer’s instructions. The RNA concentration and purity were quantified using a NanoDrop 2000 (Thermo Fisher Scientific, USA). Only samples with an RNA integrity number (RIN) > 7 were used to generate libraries. Total RNA was subjected to ribosomal RNA removal using the Ribo-Zero rRNA Removal Kit (Illumina, San Diego, USA). TruSeq™ RNA Sample Prep Kit (Illumina) was used to construct cDNA libraries. Paired-end sequencing was performed with the barcoded libraries on the Illumina HiSeq 2500 platform at Majorbio Bio-Pharm Technology (Shanghai, China) using the HiSeq 4000 PE Cluster Kit and HiSeq 4000 SBS Kits according to the manufacturer’s instructions.

Raw sequencing reads first went through quality control using Trimmomatic [25], and clean reads were then aligned to the mouse genome and the Silva database with Bowtie2 [26] to exclude host and ribosomal RNA contamination. Quality-filtered metatranscriptomic reads were de novo assembled for each sample with Trinity (v2.2.0) [27]. All contigs were clustered using CD-HIT (v 4.5.8) [28] to obtain unigenes. The abundance of unigenes in each sample was estimated by calculating the transcripts per million (TPM) value based on the number of aligned reads using Salmon (v 0.9.1) [29]. Protein coding sequences were predicted from contigs using MetaGeneMark [30]. For functional annotation, protein sequences were annotated against those of evolutionarily conserved genes in the nonsupervised orthologous groups (eggNOG) using DIAMOND (v.0.8.23) [31] under default settings. Differential enrichment of genes and functions were determined by DESeq2 [32]. Taxonomic classifications were aligned against the NCBI taxonomy database using DIAMOND (v.0.8.23) [31] with default parameters. Then, the blastx results were analyzed using the longReads LCA algorithm of MEGAN [33].

Total RNA from the colonic tissue was extracted and purified using TRIzol reagent (Invitrogen). Only those samples with RIN > 7 were used for RNA-seq. TruSeq™ RNA Sample Prep kit (Illumina, USA) was used to construct cDNA libraries. The barcoded libraries were sequenced on the Illumina HiSeq 2500 platform at Majorbio Bio-Pharm Technology (Shanghai, China). Paired-end raw sequencing reads were first quality-controlled by FastQC and mapped to the mouse genome using hisat2 (v 2.1.0) [34]. The low-expression transcripts with an average count < 20 were removed. Analysis of differential gene expression was performed using the DESeq2 package [32]. Differentially expressed genes (DEGs) were defined as genes with a minimum twofold change and a false discovery rate < 0.05. KEGG pathway functional enrichment analysis of DEGs and gene set enrichment analysis (GSEA) of gene ranks were performed using clusterProfiler [35].

To identify modules of coabundant host genes and bacterial taxa in RNA-seq and 16S rDNA sequencing data, we applied the R package WGCNA [36]. The correlation matrix quantified the interconnectedness between genes or taxa and then assigned them to coabundance modules. The genes or taxa that did not show high enough coexpression metrics with any module were removed. The signed network was derived based on a biweight midcorrelation (bicor) method. A scale-free topology criterion was used to choose soft thresholds of 12 and 5 for gene and taxon correlations, respectively. To identify the modules of genes or taxa that were most relevant in the context of colonic inflammation, correlations between serum tryptophan metabolites and each module were tested with Spearman rank.

Publicly available IBD metatranscriptomic datasets were used to determine the differences in microbial genes in tryptophan metabolic pathways. The datasets were obtained from the Sequence Read Archive (BioProject: PRJNA389280), which consists of the human HMP2 metatranscriptomic sequencing reads in the Inflammatory Bowel Disease Multiomics Database [37]. The reads used in the present study were from 46 patients with Crohn’s disease (CD), 21 patients with ulcerative colitis (UC), and 11 healthy controls. The data was processed and analyzed similarly to the mouse metatranscriptomic data as described above. For the IBD cohort, linear mixed effect models were further applied to determine the abundance of diagnosed microbial DEGs as follows: gene ~ (intercept) + diagnosis + antibiotics + chemotherapy + immunosuppressant. Diagnosis was coded as a categorical variable (CD, UC, non-IBD) with non-IBD as the reference state. The three medical treatments were coded as binary covariates with nonuse as the reference state. Gene abundance values were log-transformed and zero-smoothed prior to linear modeling. All P values of multiple comparisons in the analysis above were adjusted to FDR. The threshold of significance was set at P < 0.05 (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001); ns, not significant; P < 0.1 was considered a trend.

QIIME v.2.0, GraphPad Prism (v 6.0), and the R program (v 3.6.1) were used for statistical analysis. The difference in bacterial abundance was evaluated using one-way ANOVA and Tukey’s post hoc test for data with a normal distribution or the Kruskal–Wallis test for data with a nonnormal or nonparametric distribution. P < 0.05 was considered statistically significant. When applicable, P values were corrected for multiple testing with the Benjamini–Hochberg false discovery rate (FDR) with a cutoff value of 0.1.

We previously showed L. reuteri I5007 to be capable of ameliorating colitis and correcting microbiota in mice and sows [14, 38]. To investigate whether the intestinal microbiota is involved in L. reuteri-mediated alleviation of colitis, a colitis-recovery-colitis-colonic microbiota transplantation (CMT) model was employed (Additional file 1: Fig. S1A). Briefly, mice were orally administered with or without L. reuteri I5007 for 3 weeks, followed by induction of colitis with DSS for another week. Mice were then allowed to fully recover from colitis for 12 days before a second round of 7-day DSS treatment and reciprocal transplantation of the colonic microbiota of DSS-treated mice administered with and without L. reuteri I5007. As expected, L. reuteri I5007 significantly alleviated the disease activity index (DAI) and colonic pathology in the first round of DSS treatment, while recovery for 12 days completely restored intestinal health in both groups of DSS-treated mice (Additional file 1: Fig. S1B-D). Interestingly, the colonic microbiota of DSS-treated mice was largely restored to normal, while that of L. reuteri I5007-supplemented mice remained significantly divergent (p < 0.05) on day 40 even after 12 days of recovery (Additional file 1: Fig. S1E). The α-diversity indices of the colonic microbiota as indicated by richness and Shannon Index remained drastically reduced in DSS-treated mice administered with L. reuteri on day 40 (Additional file 1: Fig. S1F).

Transplantation of day-28 colonic microbiota from L. reuteri-administered mice significantly increased the survival and reversed the weight loss of DSS-treated mice (p < 0.05) (Additional file 1: Fig. S2A and B). Similarly, DSS-treated mice receiving L. reuteri-supplemented microbiota showed decreased pathology scores and increased colon length (Additional file 1: Fig. S2C-E). Moreover, the colonic microbiota from L. reuteri-supplemented mice significantly decreased Th17A cells and increased regulatory T cells in the colonic lamina propria and mesenteric lymph nodes (p < 0.05) (Additional file 1: Fig. S2F-I). Collectively, the results suggested that the intestinal microbiota is likely to be involved in L. reuteri-mediated protection against intestinal inflammation.

Linear discriminant analysis effect size (LEfSe) analysis was further performed to identify the major colonic bacteria enriched in DSS-treated mice with or without L. reuteri I5007 administration. The results showed that 30 bacterial taxa were significantly enriched by L. reuteri I5007 (LDA > 3.0, p < 0.05) (Fig. 1A). Interestingly, those enriched bacterial taxa were mostly Clostridium (Clostridium XlVa, Clostridium XlVb, and Clostridium IV), which are known to be capable of metabolizing tryptophan [39]. Consistently, DSS significantly impaired microbial metabolism of tryptophan as evidenced by reduced levels of indole derivatives such as indole-3-carbaldehyde (ICA), IPA, and IAA in the serum (p < 0.01). Conversely, the host metabolism of tryptophan appeared to be enhanced, showing a significant increase in the activity of indoleamine 2,3-dioxygenases (IDO) (p < 0.001) (Fig. 1B), key enzymes involved in the kynurenine pathway [40]. On the other hand, administration of L. reuteri largely restored the serum concentrations of ICA, IAA, and IPA as well as the IDO activity in DSS-treated mice to healthy levels (Fig. 1B). Interestingly, L. reuteri also restored the serum levels of 5-HT and tryptophan, but with no impact on kynurenine or ILA (Additional file 1: Fig. S3A).

RNA sequencing of the colonic tissue further identified 4680 differentially expressed genes in response to L. reuteri I5007 in DSS-treated mice (Fig. 1C). KEGG pathway enrichment and gene set enrichment analysis (GSEA) showed that L. reuteri administration downregulated inflammation-related pathways such as TNF and NF-κB signaling (Additional file 1: Fig. S3B and C). The correlation analysis of distance dissimilarities further revealed IPA to be in the strongest correlation with both the microbiota and host gene expression, while IAA was significantly correlated only with the microbiota composition (Fig. 1D).

As expected, the expression of Pxr (also known as Nr1i2), a receptor for IPA [41], and PXR target genes Ugt1a1 and Cyp3a11 increased markedly in the colon of DSS-treated mice in response to L. reuteri (Fig. 1E). IAA is an AhR ligand that promotes IL-22 production [2, 42]. However, RNA-seq analysis and RT–PCR assays failed to reveal conspicuous differences in the colonic expression of Ahr and Cyp1a1 with and without L. reuteri (Fig. 1C, Additional file 1: Fig. S3D, Additional file 2: Table S1). The serum level of IL-22 was unaffected by L. reuteri as well (Additional file 1: Fig. S3D). Weighted gene coexpression network analysis showed that there were common host genes and bacterial taxa modules that were significantly correlated with IPA and IAA (Additional file 1: Fig. S3E and F).

To directly evaluate the protective effect of major microbial tryptophan metabolites on intestinal inflammation and mucosal barrier function, 20 and 40 mg/kg IAA or IPA were separately administered to mice with or without DSS treatment for 7 days (Additional file 1: Fig. S4A). Both IPA and IAA significantly reversed DSS-induced weight loss, increased colon length, reduced colonic pathology, increased goblet cells, and decreased IL-1β and TNF-α gene expression in the colon, while augmenting the protein expression of E-cadherin and occludin (Additional file 1: Fig. S4B-G), suggesting that both IAA and IPA are capable of alleviating intestinal inflammation and enhancing mucosal barrier function. Similar to L. reuteri, IAA failed to enhance the expression of Ahr and Cyp1a1, while IPA significantly increased the expression of Pxr (Additional file 1: Fig. S4H).

Colonic microbiome analysis further revealed that IAA significantly altered both α- and β-diversities of the colonic microbiota of DSS-treated mice, while IPA only had a marginal effect (Additional file 1: Fig. S5A and B), suggesting that the microbiota is differentially regulated by IPA and IAA. To further assess whether IPA- and IAA-mediated protection of DSS-treated mice is dependent upon the intestinal microbiota, mice were treated with antibiotics for 2 weeks to deplete the intestinal microbiota, followed by oral administration of IPA, IAA, or a mixture of IPA and IAA with or without DSS treatment (Fig. 2A). Clearly, IPA or the IPA/IAA mixture was capable of protecting antibiotic-treated mice from DSS-induced colitis, while the protective effect of IAA, as indicated by weight loss, the colon length, intestine barrier, and inflammation was abrogated (Fig. 2B–G), suggestive of a differential involvement of the intestinal microbiota in the protection of mice from intestinal inflammation by IPA and IAA.

Although L. reuteri significantly increased the serum concentrations of IPA and IAA in DSS-treated mice (Fig. 1B), ILA, but not ICA, IPA, or IAA, was found to be synthesized by L. reuteri and several other Lactobacillus species (Additional file 1: Fig. S6A), consistent with earlier observations with Lactobacillus and Bifidobacterium [13]. Tryptophan supplementation of the L. reuteri culture also significantly increased ILA synthesis (Additional file 1: Fig. S6B). Microbial tryptophan metabolism involves multiple enzymatic reactions, and ILA is a critical intermediate (Fig. 3A). Metatranscriptomic analysis of the colonic microbiota from DSS-treated mice supplemented with or without L. reuteri I5007 revealed that, while DSS suppressed the expression of most bacterial genes involved in tryptophan metabolism, L. reuteri significantly increased the mRNA expression of acyl-CoA dehydrogenase (ACD), indolelactate dehydrogenase (ILD), and indoleacetamide hydrolase (IAAH), with reduced expression of indolepyruvate decarboxylase (ID) (Fig. 3B, C), suggesting that L. reuteri prefers the synthesis of ILA, IPA, and IAA. Additionally, the relative abundance of Clostridium, which encodes both ACD and IAAH, was increased upon L. reuteri administration (Fig. 3D, E).

To directly assess the efficacy of ILA in alleviating intestinal inflammation, DSS-treated mice were orally administered daily with or without ILA for a week. ILA significantly reversed DSS-induced weight loss (Additional file 1: Fig. S7A) and colon length shortening (Additional file 1: Fig. S7B). Consistently, ILA significantly suppressed DSS-induced production of TNF-α and IL-1β in the serum (Additional file 1: Fig. S7C). Colonic histological scores were also significantly reduced in ILA-treated mice (Additional file 1: Fig. S7D), as indicated by lessened ulceration, edema, crypt damage, and immune cell infiltration of the intestinal epithelial layer (Additional file 1: Fig. S7E). ILA also partially restored DSS-induced reduction of richness and Shannon index (Additional file 1: Fig. S7F) of the colonic microbiota. Notably, ILA further caused a clear segregation of the colonic microbiota in DSS-treated mice (Additional file 1: Fig. S7G). At the genus level, ILA restored relative abundance of Clostridium_XlVa (from 6.3 to 35.8%) and Lactobacillus (from 0.45 to 3.8%) in DSS-treated mice (Additional file 1: Fig. S7H).

To further examine whether the protective effect of ILA is dependent upon the intestinal microbiota, mice were treated with antibiotics for 2 weeks to deplete the intestinal microbiota, followed by DSS treatment with or without ILA supplementation (Additional file 1: Fig. S8A). The results showed that the protective effects of ILA on weight loss and colon shortening were significantly diminished in the microbiota-depleted mice, relative to the conventional mice (Additional file 1: Fig. S8B-E). ILA is known to exert its protective effect by activating AhR [11, 13]. To examine whether pharmacologic inhibition of AhR abrogates ILA-mediated amelioration of intestinal inflammation, mice were administrated with CH223191↗, an AhR inhibitor, together with ILA during DSS treatment (Additional file 1: Fig. S8F). To our surprise, ILA conferred significant protection even in the presence of CH223191↗, as indicated by weight loss, the colon length, and histopathology (Additional file 1: Fig. S8G-J). Therefore, we conclude that the microbiota modulation may be another pathway to AhR activation for IAA to exert the protective effect on intestinal inflammation.

ILA, IPA, and IAA alleviate intestinal inflammation in IL-10^−/− mice.

IL-10^−/− mice are known to spontaneously develop colitis [43]. To confirm whether ILA, IPA, and IAA are capable of protecting mice from intestinal inflammation in a different colitis model, ILA, IPA, or IAA was orally administered to IL-10^−/− mice for 4 weeks. Both ILA and IPA significantly suppressed the colonic expression of IL-1β and IL-6 mRNA, while IAA has a lesser effect (Fig. 4A). ILA and IPA were also capable of restoring TNF-α and LPS to healthy levels in the colon, with a marginal effect observed with IAA (Fig. 4B). Overall, these results suggested that ILA and IPA are more efficient than IAA in protecting mice from intestinal inflammation in both DSS and IL-10^−/− murine models of colitis. Additionally, ILA and IPA altered the colonic microbiota of IL-10^−/− mice more drastically than IAA (Fig. 4C). Importantly, ILA supplementation markedly elevated relative abundance of Lactobacillus (from 6.45 to 39.58%), while IPA enriched Bifidobacterium (from 0.01 to 0.81%) and Akkermansia (from 0.21 to 10.37%) (Fig. 4D). ILA and IPA also increased relative abundance of Blautia (from 0.06 to 1.25% and 1.22%, respectively). The co-occurrence networks revealed Blautia and Bifidobacterium to be key bacterial taxa in the microbial networks in response to ILA and IPA (Fig. 4E). Notably, IPA had the most complex network of the microbial features (Fig. 4E). Taken together, these results suggested that the intestinal microbiota is likely involved in ILA- and IPA-mediated alleviation of colitis.

Because only ILA was the only tryptophan metabolite synthesized by L. reuteri (Additional file 1: Fig. S6), enhanced synthesis of IPA and IAA in L. reuteri- supplemented DSS mice was likely a result of utilization of ILA as a substrate by other intestinal bacteria. To directly confirm whether ILA increases IPA and IAA production through microbial cross-feeding, we incubated ILA with the colonic microbiota of DSS-treated mice and found ILA significantly increased IPA and IAA levels (Fig. 5A). The microbiome analysis indicated that ILA markedly altered the composition of the microbiota (Fig. 5B). Functional metagenomic prediction by PICRUSt2 showed that the expression of ACD, ID, and ArAT increased in response to ILA (Fig. 5C). The coabundance network revealed that ACD had the highest degree of the interactions showing a strong positive correlation with several differentially enriched bacterial genera (Fig. 5D). Furthermore, oral administration of ILA to DSS-treated mice also significantly increased the ACD and IAAH expression in the colonic microbiota as predicted by PICRUSt2 (Fig. 5E, F). To further verify that ILA promotes the production of other indole derivatives, we co-cultured tryptophan, L. reuteri I5007, CFS and ILA with C. sporogenes, which is a specific IPA-producing bacterium. As expected, ILA and CFS of L. reuteri markedly increased the production of IPA, but not IAA (Fig. 5G). Collectively, these data suggested that ILA may be utilized by bacteria as a substrate to produce IPA and IAA.

To further examine whether ILA promotes IPA and IAA production through microbial cross-feeding in vivo, we treated mice with antibiotics to deplete the intestinal microbiota, followed by supplementation of heat-killed or an ArAT-deficient L. reuteri mutant strain (ΔArAT) with or without oral administration of ILA (Fig. 6A). Both heat-killed and ΔArAT L. reuteri failed to protect mice against DSS-induced colitis (Fig. 6B, C). Notably, the heat-killed and ΔArAT L. reuteri cannot promote the production of IPA and IAA, while administration of ILA to mice receiving heat-killed or ΔArAT mutant strain led to increased syntheses of both metabolites in the colonic content (Fig. 6D, E). These results confirmed that ILA is the major tryptophan metabolite produced by L. reuteri to be responsible for microbial synthesis of IPA and IAA in the GI tract.

To further verify whether ILA enhances IPA and IAA production dependent upon the intestinal microbiota, we either disrupted the microbiota with antibiotics or induced dybiosis with Citrobacter rodentium or DSS with or without oral administration of ILA (Fig. 6F). The microbiome analysis showed that ILA had no impact on microbial α- and β-diversity of healthy, C. rodentium-challenged, or antibiotic-treated mice (Fig. 6G, H). Notably, ILA remained largely effective in both dysbiotic mouse models. ILA significantly enhanced IPA and IAA in DSS-treated mice (p < 0.05), with a strong tendency (p = 0.09) to increase IPA and IAA in C. rodentium-challenged mice; however, ILA failed to promote the synthesis of either IPA or IAA in the antibiotic-treated mice (Fig. 6G, H), suggesting clearly that the intestinal microbiota is required for ILA-induced IPA and IAA synthesis. Collectively, these results indicated that ILA enhances IPA and IAA production through microbial cross-feeding.

Similar to our observations in DSS-induced colitis (Fig. 1B), host tryptophan metabolism is enhanced in IBD patients [44]. However, it is unclear whether microbial metabolism of tryptophan is suppressed in IBD patients as in DSS-treated mice. To examine the expression levels of the major bacterial genes responsible for tryptophan metabolism in IBD patients, we retrieved the metatranscriptomic profiling datasets from the HMP2 IBDMDB cohort, which includes the stool samples from both ulcerative colitis (UC, n = 21) and Crohn’s disease (CD, n = 46) patients as well as healthy controls (n = 11) [45]. PCoA plot showed that the microbiota profile responsible for tryptophan metabolism are different from healthy controls, with no obvious difference between UC and CD patients (Additional file 1: Fig. S9A). Among the microbial metabolic genes, only ACD showed a significant reduction in individuals with UC (Additional file 1: Fig. S9B). ACD was mainly produced by Faecalibacterium (21.9%), Roseburia (20.7%), and Prevotella (7.5%) in the stool samples, and their abundances were substantially reduced in IBD patients (Figure S9C and 9E). At the species level, three most abundant bacteria including Roseburia sp. CAG:18_43_25 (13.2%), Faecalibacterium prausnitzii (9.6%), and Prevotella copri (8.8%), were markedly reduced in IBD patients (Additional file 1: Fig. S9D and F). Suppression of ACD and ACD-producing bacteria is consistent with an earlier observation on IPA reduction in IBD-induced dysbiosis [46]. Collectively, these results strongly suggested that tryptophan metabolism by the intestinal microbiota is very likely to be reduced in IBD patients.

All mouse experiments were approved by the Institutional Animal Care and Use Committee of China Agricultural University (Beijing, CAU20150925-2).

Not applicable.

The authors declare no competing interests.

16S rDNA and metatranscriptomic sequencing data in the present study have been deposited in the NCBI Sequence Read Archive database under accession numbers PRJNA782376 and PRJNA782681.