Microbiota-Derived Indole Metabolites Promote Human and Murine Intestinal Homeostasis through Regulation of Interleukin-10 Receptor

Animal Studies

For metabolomics analysis, C57BL/6-129 mice were administered 3% (wt/vol) dextran sodium sulfate (DSS) (molecular weight 36,000 to 50,000; MP Biochemicals, Burlingame, CA) in drinking water for 5 days, followed by a 2-day recovery period. Normal tap water was returned during these 2 days before tissue collection. Control mice were maintained on tap water for 7 days. In subsequent animal experiments, 8– to 12-week–old C57BL/6 mice were administered water or 2.5% DSS ad libitum for 9 days. This administration approach varied slightly to account for a change in death susceptibility to DSS. DSS was then removed and mice were allowed to recover for 2 days before euthanasia. For IPA treatment experiments, the addition of IPA at 0.1 mg/mL was administered to water- and DSS-treated animals. Mice were housed in accordance with guidelines from the American Association for Laboratory Animal Care and Research Protocols, and all animal work was approved by the Institutional Animal Care and Use Committee of the University of Colorado.

Cell Lines and shRNA Knockdown

Human T84 IEC were cultured in 1:1 Dulbecco's modified Eagle's medium–Ham's F12 with 2.5 mmol/L l-glutamine and 10% fetal bovine serum. Cells were maintained at 37°C with 5% CO₂. Transepithelial electrical resistances were monitored using an EVOM2 Voltohmmeter (World Precision Instruments, Sarasota, FL). Cytokines were purchased from R&D Systems (Minneapolis, MN). IAld, IPA, and AHR inhibitor, CH-223191, were obtained from Sigma-Aldrich (St. Louis, MO). 6-Formylindole[3,2-b] carbazole was obtained from Tocris Bioscience (Bristol, UK). All compounds were used at indicated concentrations. Human intestinal organoids (HIOs) were derived following previously described methods¹⁹ and kindly provided by Dr. Jason Spence (University of Michigan, Ann Arbor, MI). HIOs were maintained by embedding in Matrigel (BD Biosciences, San Jose, CA) and applying Advanced DMEM-F12 medium (Invitrogen, Carlsbad, CA) containing 1X B27 supplement (Invitrogen), 1X GlutaMAX (Life Technologies, Carlsbad, CA), 10 μmol/L HEPES, 10% penicillin/streptomycin, 100 ng/mL rhNoggin (R&D Systems), 100 ng/mL epidermal growth factor (R&D Systems), and approximately 500 ng/mL R-Spondin1 (RSPO1). RSPO1 was obtained from conditioned media collected from a HEK293 cell line that was stably transfected and zeocin-selected for the RSPO1 expression vector. The medium was changed every 2 to 4 days, and HIOs were transferred to fresh Matrigel once a week until they reached approximately 0.5 mm to 1 mm in size for experiments and RNA isolation. Lentiviral particles encoding shRNA directed against aryl hydrocarbon receptor nuclear translocator (ARNT; TRC MISSION shRNA, University of Colorado Functional Genomics Facility) were used to transduce T84 cells using standard protocols. Stable integration was achieved by puromycin selection at 6 μg/mL. Knockdown was confirmed by real-time quantitative PCR analysis, indicating 80% to 85% depletion of ARNT levels.

RNA Isolation and Real-Time PCR

Total RNA was extracted from cells using TRIzol (Invitrogen) and from tissue using RNeasy Mini Kit (Qiagen, Hilden, Germany). cDNA was prepared using iScript cDNA synthesis kit (Bio-Rad, Hercules, CA). Real-time PCR to measure transcripts was performed in 1 × Power SYBR Green Master Mix (Applied Biosystems, Foster City, CA) using an ABI 7300 thermocycler (Applied Biosystems). Fold-change in expression of target mRNA relative to β-actin mRNA was calculated as previously described.²⁰ Total RNA from HIOs was isolated using Direct-zol RNA Miniprep Kit (Zymo Research Corp, Irvine, CA). Fold-change in expression of mRNA transcript from HIO experiments was calculated relative to HSPCB (heat shock protein 90 α family class B member 1). Human and mouse primer sequences are listed in Table 1.

Western Blot Analysis

Whole-cell and tissue lysates were extracted in Tris-Lysis buffer on ice and disrupted by sonication. Protein content was quantified using BCA protein assay reagent (Thermo Fisher Scientific, Waltham, MA) and 30 μg of whole-cell or tissue extract was boiled in Laemmli buffer in reducing conditions subjected to SDS-PAGE. The SDS-PAGE was transferred onto a polyvinylidene difluoride membrane and probed for IL-10R1 using a rabbit polyclonal IL-10RA antibody (dilution 1:1000; Thermo Fisher Scientific) and β-actin (dilution 1:10,000; Abcam, Cambridge, UK).

Metabolomic Analysis

Distal colon tissue (1 cm) was collected from control and DSS-treated mice. Tissues were flash-frozen and placed at −80°C. Global metabolomics was performed by Metabolon, Inc. (Durham, NC). Briefly, tissue samples were thawed and weighed in tared cryovials, and 80% ice-cold methanol was added at a ratio of 75 μL of solvent per mg of sample, then incubated overnight at 4°C to extract biochemicals. Internal standards were included to control for extraction efficiency. Following methanol extraction, colon samples were processed and analyzed as previously described.²¹

HPLC Metabolite Analysis

Indole derivatives were quantified in mouse and human samples using reversed-phase high-performance liquid chromatography with electrochemical coulometric array detection (EC-HPLC) (CoulArray; Thermo Fisher Scientific). Archived human serum samples from patients diagnosed with IBD or from healthy individuals were obtained under research protocols approved by the Colorado Multi-Institutional Review Board. Study participant demographics are listed in Table 2. Tissue and serum samples were extracted in 80% methanol, and protein precipitate was removed by centrifugation at 15,000 × g. Separation was achieved using an Acclaim Polar Advantage II C18 column (Thermo Fisher Scientific) at a flow rate of 1 mL/minute on a gradient of 10% to 55% acetonitrile in 50 mmol/L sodium phosphate buffer (pH 3), containing 0.42 mmol/L octanesulphonic acid as an ion-pairing agent. Calibration curves were composed by performing linear regression analysis of the peak area versus the analyte concentration. The data were quantified using the peak area in comparison to standards.

Bacterial Strains and Gnotobiotic Colonization Experiments

Escherichia coli (E. coli) strains were obtained from GE Dharmacon (Lafayette, CO). An E. coli K12 parent strain [E. coli BW25113; wild type (WT)], along with strains containing a deletion of tnaA (E. coli JW3686; ΔtnaA) and tnaB (E. coli JW5619; ΔtnaB) were used for experiments. All strains were grown in Lysogeny broth (10 g/L tryptone, 10 g/L NaCl, 5 g/L yeast extract) or on Lysogeny broth–agar plates supplemented with appropriate antibiotics (kanamycin, 50 μg/mL), if needed. Cultures were grown overnight at 37°C to stationary phase and tested for the presence of indole by the addition of 200 μL of Kovac's reagent (Sigma-Aldrich) to 2 mL bacterial culture. For T84 cell experiments, cultures were grown overnight at 37°C to stationary phase. Briefly, cultures were spun, supernatants collected, and afterward serially diluted and placed onto cells for 24 hours. Bacterial supernatants used for EC-HPLC analysis were placed through a 0.22-μm filter before the run. For gnotobiotic colonization experiments, 5- to 8-week–old germ-free C57BL/6 mice were gavaged with a 100-μL bacterial suspension for mono-association (10⁹ colony forming units of each bacterial strain collected from a stationary-phase culture and re-suspended in phosphate-buffered saline). Mice were colonized for 2 weeks before euthanasia. Fresh fecal pellets were collected periodically, weighed, homogenized, and then serially diluted in phosphate-buffered saline phosphate-buffered saline for plating to determine bacterial colony forming units per g of feces.

Histology

Colon samples were fixed in 10% neutral buffered formalin and paraffin embedded before staining with hematoxylin and eosin. All histologic quantitation was performed blinded by the same individual (S.P.C.) using a scoring system previously described.²² Briefly, the three independent parameters measured were severity of inflammation (0 to 3: none, slight, moderate, severe), extent of injury (0 to 3: none, mucosal, mucosal and submucosal, transmural), and crypt damage (0 to 4: none, basal 1/3 damaged, basal 2/3 damaged, only surface epithelium intact, entire crypt and epithelium lost). The score of each parameter was multiplied by a factor reflecting the percentage of tissue involvement (×1: 0% to 25%, ×2: 26% to 50%, ×3: 51% to 75%, ×4: 76% to 100%), and all numbers were summed. Maximum possible score was 40.

Quantification of Cytokines in Colon Tissue

For cytokine analysis, colon tissue was extracted in Tris-Lysis buffer by sonication and protein homogenates were quantified using BCA protein assay reagent (Thermo Fisher Scientific). Tissue concentrations of cytokines were measured using a proinflammatory cytokine screen (Meso Scale Discovery, Rockville, MD). Assays were performed according to the manufacturer's instructions. Cytokine concentrations were normalized to total protein concentration.

Quantification and Statistical Analysis

Data are expressed as means ± SEM. Statistical analyses were performed with GraphPad Prism software version 7.0 (GraphPad Software, La Jolla, CA) using a two-tailed unpaired t-test for direct comparisons, and one-way or two-way analysis of variance with Tukey's test for multiple comparisons. Statistical differences are reported as significant when P < 0.05.

Tryptophan-Indole Metabolism Is Altered in Murine and Human Colitis

To better understand microbe-derived factors that contribute to intestinal homeostasis, a comprehensive, unbiased screen of serum and colonic tissue metabolites was performed from healthy and DSS colitic mice. Mice were administered DSS via drinking water, and serum and colonic tissue were collected during peak disease (day 7). Of the microbe-derived metabolites, most notable were decreases in tryptophan-indole metabolites in both serum and colons of colitic mice compared with mice administered water alone (Figure 1A). Tryptophan is metabolized through various pathways, including the indole pathway, resulting in derivatives that are produced from metabolism by gut microbiota (Figure 1B). Colitis profoundly altered tryptophan metabolism, specifically revealing a selective decrease in indole metabolites. To validate this mass spectrometry–based metabolite screen, an HPLC-based protocol was developed using electrochemical detection methods (Figure 1C). Eight- to 10-week–old C57BL/6 mice were administered water or 2.5% DSS ad libitum for 9 days. DSS was then removed and mice were allowed to recover for 2 days before euthanasia. Serum indole metabolites were profiled by EC-HPLC (Figure 1D). Serum indole and IPA levels were significantly decreased in actively colitic animals (P < 0.05). A marked decrease in serum IAld was also observed (P = 0.09). It is notable that overall, tryptophan levels in DSS colitis actually increase (Figure 1A), suggesting that the findings of indole depletion are not a result of diminished tryptophan absorption.

Guided by results of indole depletion in active murine colitis, we attempted to translate our results to human patients. Here, the EC-HPLC method was used to quantify various indole metabolites in serum samples from patients with UC. For these purposes, serum samples from healthy controls (n = 20), subjects with active UC (n = 15), and subjects with UC in remission (n = 20) were profiled (Table 2). This analysis revealed that serum IPA was deceased by nearly 60% in subjects with active UC compared with healthy controls (P < 0.05) (Figure 1E). Notably, this IPA deficiency normalized in UC patients in remission, implicating IPA as both a biomarker for active UC as well as an indicator of disease remission in human UC. Based on these findings, the role of indole and metabolites IPA and IAld on intestinal epithelial function was further studied.

Indole Metabolites Induce IL-10R1 Expression on Intestinal Epithelia and Improve Barrier Formation

We have previously shown that tryptophan metabolites are important regulators of IL-10R1 in IEC.²³ Guided by our unbiased metabolomic profile of serum and colon tissue from healthy and DSS-colitic mice, and our recent work identifying the epithelial interleukin-10 receptor as the dominant signature for resolution of inflammation in DSS colitis,6, 7 it was hypothesized that indole-containing tryptophan metabolites regulate intestinal homeostasis via regulation of epithelial IL-10R1 expression. To examine the induction of IL-10R1 in response to indole metabolites, T84 IEC were exposed to IPA or IAld for 6 hours; real-time quantitative PCR of IL-10R1 transcript levels showed a concentration-dependent induction of IL-10R1 (Figure 2A). These findings were not limited to colonic cancer cell lines. HIOs are complex, three-dimensional spheroid tissues derived from human pluripotent stems cells¹⁹ and contain the majority of functional epithelial cell types (ie, enterocytes, goblet, Paneth, enteroendocrine, intestinal stem cells) and structures (ie, brush borders of microvilli, crypt-like structures, and a mesoderm layer) comprising the human small intestine.²⁴ HIOs treated with IPA exhibited increasing levels of IL-10R1 transcript, with significant induction at 24 hours (P < 0.001) (Figure 2B). Further analysis revealed a prominent induction of IL-10R1 protein expression in T84 IEC exposed to IPA and IAld (Figure 2C).

The influence of the metabolite IAld on barrier formation in T84 cell monolayers was further examined, as our previous work has demonstrated that IL-10 signaling is important in IEC barrier development and maintenance.7, 23 Cells exposed to both IAld and IL-10 exhibited significantly increased barrier formation at 72 hours compared with untreated cells as measured by transepithelial electrical resistances (P < 0.01) (Figure 2D). Further, IAld in combination with IL-10 (10 ng/mL) significantly induced suppressor of cytokine signaling 3 (SOCS3), an IL-10–responsive gene that has been found to be protective in intestinal inflammation²⁵ (P < 0.05) (Figure 2E). From this perspective, it is notable that SOCS3 was not induced by IL-10 alone as a result of nearly undetectable levels of IL-10R1 at baseline.6, 7 Overall, these results identify indole metabolite–dependent induction of the epithelial IL-10R1 as a target pathway for mucosal homeostasis.

Loss of Indole-Dependent IL-10R1 Induction in Cells Lacking ARNT

The mechanism of indole signaling in intestinal epithelia was studied next. Others have shown that tryptophan metabolites function as endogenous ligands for AHR.²⁶ To determine the relative contribution of AHR to indole signaling, shRNA knockdown was used to deplete the AHR dimeric partner ARNT, and IL-10R1 induction was examined as an endpoint. Lentiviral shRNA-mediated knockdown of ARNT (shARNT) in T84 IEC resulted in significantly reduced ARNT mRNA expression (81% ± 5% decrease; P < 0.01) relative to cells containing a nontemplate control (shNTC) (Figure 3A). A broader examination of classical AHR-ARNT target genes26, 27, 28 revealed that transcript levels of cytochrome P450 family 1 subfamily A member 1 (CYP1A1) and cytochrome P450 family 1 subfamily B member 1 (CYP1B1) were significantly increased in shNTC cells treated with IAld compared with untreated shNTC cells. Both targets were also significantly increased in the IAld-treated shNTC compared with shARNT T84 IEC following IAld treatment (P < 0.01, P < 0.05, respectively) (Figure 3B). Analysis of these ARNT-deficient cells revealed a prominent decrease in IAld-dependent induction of IL-10R1 transcript and protein (P < 0.05) (Figure 3, C–E). The AHR small-molecule inhibitor, CH-223191, serves as a specific AHR antagonist. shNTC and shARNT T84 IEC were exposed to IAld (1 mmol/L), CH-223191 (10 μmol/L), or in combination for 24 hours. IAld treatment alone significantly increased IL-10R1 expression at 24 hours. CH-223191 significantly reduced IAld-mediated IL-10R1 induction, supporting the hypothesis that IAld modulates IL-10R1 expression via an AHR-mediated mechanism (P < 0.05) (Figure 3F). These results suggest that the AHR pathway is involved in the indole metabolite–dependent expression of IL-10R1 in IEC.

Bacterial Indole Production Induces IL-10R1

Whether microbe-derived sources of indole could similarly regulate epithelial target genes was examined next. Indole production was targeted in E. coli. Indole is synthesized from tryptophan by microbial tryptophanase.²⁹ Using Kovac's reagent to test for the presence of indole, an E. coli K12 strain (E. coli BW25113; WT) clearly produced indole (Figure 4A), whereas deletion of the tryptophanase gene (ΔtnaA) revealed a lack of detectable indole. Conversely, deletion of one of the tryptophan transporters (ΔtnaB) did not compromise indole production (Figure 4A). To validate these observations, the presence of indole in cell-free supernatants of bacteria was examined by EC-HPLC. Although E. coli WT bacteria exhibited readily detectable levels of indole, metabolism of tryptophan to indole was completely abolished in E. coli ΔtnaA (Figure 4B). To test the activity of this microbial-derived indole, cell-free supernatants from E. coli WT and ΔtnaA mutant were serially diluted and exposed to T84 IEC for 24 hours. Supernatants from E. coli WT readily induced IL-10R1 protein expression, whereas minimal IL-10R1 induction was observed in cells exposed to E. coli ΔtnaA supernatant (Figure 4C).

These findings were extended to an in vivo model. The function of indole-producing bacteria was examined on epithelial IL-10R1 in germ-free mice. Germ-free mice were colonized with E. coli WT or E. coli ΔtnaA for 2 weeks followed by euthanasia. Fresh fecal pellets were collected periodically for plating to ensure equal colonization, quantified as bacterial colony forming units per gram of feces (data not shown). Indole concentrations in cecal contents were validated by HPLC in vehicle and colonized animals. Mice monocolonized with E. coli WT displayed significantly increased cecal indole, whereas negligible levels were measured in phosphate-buffered saline– and E. coli ΔtnaA–gavaged animals (P < 0.01) (Figure 4D). Further, colons were harvested and RNA extracted for real-time quantitative PCR analysis. This strategy revealed that mice monocolonized with E. coli WT displayed significantly increased colonic il10r1 expression, whereas no significant il10r1 induction was observed in colon tissue of mice colonized with E. coli ΔtnaA mutant (P < 0.05) (Figure 4E). Taken together, these in vitro and in vivo results strongly support the hypothesis that microbe-derived indoles promote epithelial homeostasis.

IPA Improves DSS Colitis Outcomes

Given the observation that indole metabolites are significantly decreased in active colitis, the therapeutic potential of IPA was tested in murine colitis outcomes. For these purposes, mice were administered either normal drinking water, DSS (2.5% wt/vol), or a combination of DSS and IPA (0.1 mg/mL) in drinking water for 9 days, followed by DSS removal and a 2-day recovery. To verify that oral administration normalized IPA levels, colonic IPA was quantified by EC-HPLC. This analysis revealed that although DSS colitic mice displayed significantly lower levels of IPA, oral administration of IPA normalized colonic levels of this metabolite (P < 0.01) (Figure 5A). Under these treatment conditions, DSS/IPA-treated animals displayed significantly less reduction in colon length, a marker of intestinal inflammation, compared with DSS-treated animals (P < 0.05) (Figure 5B). Histologic analysis revealed that DSS/IPA-treated mice displayed attenuated inflammatory infiltration and decreased loss of architecture in comparison with mice treated with DSS alone, which displayed pronounced loss of epithelium and tissue architecture (Figure 5C). These differences resulted in greater histopathologic severity scores (P < 0.001) (Figure 5D). Tissue cytokine levels were also examined in these mice. Proinflammatory cytokine levels were dramatically increased in DSS-treated mice, specifically IFN-γ (P < 0.05) (Figure 5E), tumor necrosis factor-α (TNF-α) (P < 0.01) (Figure 5F), and IL-1β (P < 0.05) (Figure 5G), whereas cytokine levels in DSS/IPA-treated mice were similar to mice administered water alone. Similarly, IL-10 levels were significantly increased in DSS-treated mice, though IPA administration did not alter levels of this cytokine (Supplemental Figure S1). Further, IPA administration increased transcript and protein levels of AHR target genes (Supplemental Figure S2). Taken together, these results indicate that therapeutic normalization of IPA during active colitis attenuates disease and promotes intestinal homeostasis.