Report

The Microbial Metabolites, Short-Chain Fatty Acids, Regulate Colonic Treg Cell Homeostasis

See allHide authors and affiliations

Science  02 Aug 2013:
Vol. 341, Issue 6145, pp. 569-573
DOI: 10.1126/science.1241165

Protecting the Guts

Regulatory T cells (Tregs) in the gut are important sentinels in maintaining the peace between our gut and its trillions of resident bacteria and have been shown to be regulated by specific strains of bacteria in mouse models. Smith et al. (p. 569, published online 4 July; see the Perspective by Bollrath and Powrie) asked whether metabolite(s) generated by resident bacterial species may regulate Tregs in the gut. Indeed, short-chain fatty acids (SCFAs), bacterial fermentation products of dietary fibers produced by a range of bacteria, restored colonic Treg numbers in mice devoid of a gut microbiota and increased Treg numbers in colonized mice. The effects of SCFAs on Tregs were mediated through GPCR43, a receptor for SCFAs, which is expressed on colonic Tregs. Mice fed SCFAs were protected against experimentally induced colitis in a manner that was dependent on GPR43-expressing Tregs.

Abstract

Regulatory T cells (Tregs) that express the transcription factor Foxp3 are critical for regulating intestinal inflammation. Candidate microbe approaches have identified bacterial species and strain-specific molecules that can affect intestinal immune responses, including species that modulate Treg responses. Because neither all humans nor mice harbor the same bacterial strains, we posited that more prevalent factors exist that regulate the number and function of colonic Tregs. We determined that short-chain fatty acids, gut microbiota–derived bacterial fermentation products, regulate the size and function of the colonic Treg pool and protect against colitis in a Ffar2-dependent manner in mice. Our study reveals that a class of abundant microbial metabolites underlies adaptive immune microbiota coadaptation and promotes colonic homeostasis and health.

The intestinal immune system has coevolved with the gut microbiota for the maintenance of intestinal health (1). Disruption of this homeostasis leads to intestinal inflammation and disease (2, 3). Colonic regulatory T cells (cTregs) expressing the transcription factor Foxp3 are critical for limiting intestinal inflammation and depend on microbiota-derived signals for proper development and function (47). Bacteroides fragilis and clostridial species induce Treg responses (6, 7); however, how the gut microbiota affect cTreg responses across mammalian hosts remains unclear. Although polysaccharide A from B. fragilis modulates Treg responses (6), such effects are also likely mediated through more common factor(s) produced by many bacterial genera.

Humans and mice rely on bacteria to break down undigestible dietary components, such as fibers (8). Short-chain fatty acids (SCFAs) are bacterial fermentation products and range in concentration from 50 to 100 mM in the colonic lumen (9). We examined SCFA concentrations in specific pathogen–free (SPF) mice, gnotobiotic altered Schaedler flora (ASF)–colonized mice, and germ-free (GF) mice and found that GF mice had reduced concentrations of the three most abundant luminal SCFAs—acetic acid, propionic acid, and butyric acid (table S1)—as previously reported (10) (see also supplementary materials and methods). The decrease of these SCFAs in GF mice suggests that SCFAs may contribute to their immune defects, specifically reduced cTreg numbers. We provided SCFAs in the drinking water (150 mM) to GF mice for 3 weeks and found that SCFAs individually or in combination (SCFA mix) increased cTreg frequency and number (Fig. 1A) but did not increase the number or frequency of splenic, mesenteric lymph node (MLN) cells or thymic Tregs (fig. S1). These effects coincided with increased luminal SCFAs (table S1). SCFAs increased CD4+ T cell frequency and number (fig. S2) but did not alter colonic T helper 1 (TH1) or TH17 cell numbers significantly (fig. S3).

Fig. 1 SCFAs restore colonic Treg populations and function in germ-free mice.

(A) cLP lymphocytes were isolated and stained for CD4 and Foxp3. (Left) Representative dot plots and percentage of CD4+Foxp3+ within the CD45+CD3+ population from SPF and GF mice and GF mice treated with propionate (P), acetate (A), butyrate (B), or the SCFA mix in the drinking water. -, no SCFA in pH- and sodium-matched drinking water. (Right) Numbers of Foxp3+ Tregs for the left panel. (B) cTregs were isolated from in vivo propionate-treated GF mice; sorted by CD4, CD127, and CD25; and examined ex vivo for expression of Foxp3 and IL-10 by reverse transcriptase quantitative polymerase chain reaction (RTqPCR). (C) cTregs were isolated from GF mice and purified as in (B); cultured for 24 hours in the presence or absence of 0.1 mM propionate; and examined for expression of Foxp3, TGFβ, and IL-10 by RTqPCR and IL-10 protein production by enzyme-linked immunosorbent assay. In (A), symbols represent data from individual mice. Horizontal lines show the mean; error bars denote the SD. In (B) and (C), each symbol or bar represents pooled cTregs from three to five mice. All data shown are representative of at least three independent experiments. (A) Kruskal-Wallis test with a Dunn’s post hoc test: ***P < 0.001; *P < 0.05. (B and C) Mann-Whitney U test. ns, not significant.

Microbiota-induced cTreg development is associated with increases in de novo generation of inducible Tregs (iTregs), not Tregs of thymic origin (nTregs) (7). These populations can be distinguished by expression of Helios, which, in vivo, is restricted to nTregs (11). We found that Helios+ Treg frequency was similar between GF and SCFA-treated GF mice (fig. S4) but was lower in SPF mice. SCFA-treatment increased Helios+ Treg numbers in GF mice, indicating that Tregs already present in the colonic lamina propria (cLP) were expanding.

To test if SCFAs could affect cTregs in a GF setting, we isolated cTregs from GF mice treated with propionate in vivo for 3 weeks and examined expression of Foxp3 and interleukin-10 (IL-10), a key cytokine in Treg-mediated suppression. We also isolated cTregs from GF mice and stimulated them in vitro with propionate. Both treatments increased Foxp3 and IL-10 expression significantly (Fig. 1, B and C). In vitro treatment increased IL-10 production but not that of transforming growth factor–β (TGFβ), a Treg-mediated suppression factor, suggesting that SCFAs specifically induce Foxp3+IL-10–producing Tregs (Fig. 1, B and C).

The antibiotic vancomycin, which targets Gram-positive bacteria and disrupts the gut microbiota (12), reduces cTregs to similar levels as observed in GF mice (7) (fig. S5). However, when SPF mice were treated with a combination of vancomycin and SCFAs, the reduction in cTregs was completely restored (fig. S5). Collectively, these results suggest that SCFAs play a role in cTreg homeostasis.

We questioned whether SCFAs would augment cTregs in SPF mice. SCFA treatment of SPF mice increased Foxp3+ and Foxp3+IL-10+ cTreg frequency and number (Fig. 2, A to C), but not Foxp3+TGFβ+ cTregs (fig. S6). We did not observe changes with SCFA treatment in small intestinal Treg numbers (fig. S7). Neither colonic TH17 and TH1 nor MLN cell and splenic Treg frequency and number from SPF mice were affected by SCFAs (fig. S8 to S11). These results may explain the benefits of dietary fibers and bacteria, such as clostridia and bifidobacteria (13), that can increase colonic luminal SCFA production and modulate inflammation in mice and humans. We measured SCFA production of species belonging to Clostridium clusters XI (Clostridium bifermentans), XIV (ASF 356 and 492) and XVII (C. ramosum) and the bacteroides species B. fragilis, as Clostridium cluster XIV members and B. fragilis affect cTregs (6, 7). ASF 356 and 492 and C. ramosum generated more propionate [14 to 62 versus 0.05 to 1.1 μmol/105 colony-forming units (CFU)] and acetate (118 to 220 versus 0.1 to 2 μmol/105 CFU) compared with the other strains (table S1).

Fig. 2 SCFAs augment colonic Treg population size and function in SPF mice.

(A) SPF Foxp3YFP-Cre mice (YFP, yellow fluorescent protein) were treated with water alone (-) or P, A, B, or the mix. Colonic LP lymphocytes were isolated and stained for CD4 and IL-10. (Top) Representative dot plots and percentage of CD4+ Foxp3-YFP+ within the CD45+CD3+ population. (Bottom) Representative dot plots and percentage of the CD4+Foxp3+IL-10+population. (B) Cell numbers for the data in the top panel of (A). (C) Cell numbers for the data in the bottom panel of (A). Symbols in (B) and (C) represent data from individual mice and four independent experiments. (D) cTregs were cocultured with splenic Teff and P, A, B, or media (sodium- and pH-matched) for 96 hours. Percent suppression is shown on the y axis; the x axis denotes Treg:Teff ratios. Symbols represent mean values; error bars denote SD for four independent experiments. *P < 0.01. (E) cTregs were isolated from the LP of in vivo propionate-treated SPF Foxp3YFP-Cre mice, sorted for CD4 and YFP, and examined for ex vivo expression of Foxp3 and IL-10. Each symbol represents pooled cTregs from three to five mice, horizontal lines show the mean, and error bars denote SD. Four independent experiments were performed. (B, C, and D) Kruskal-Wallis test with a Dunn’s post hoc test: ***P < 0.0001; **P < 0.01; *P < 0.05. (E) Mann-Whitney U test: P values are shown where significant.

Colonic regulatory T cells regulate intestinal homeostasis and control inflammation by limiting proliferation of effector CD4+ T cells (Teff). Addition of SCFAs to cTreg and Teff cocultures increased cTreg suppressive capacity (Fig. 2D and fig. S12). In SPF mice, SCFAs are taken up by colonic epithelial cells but also diffuse through the epithelium into the lamina propria, where they can mediate their effects directly (9, 14). To determine if SCFAs directly affect cTregs, we isolated cTregs from SCFA-treated SPF mice. In vivo treatment increased cTreg Foxp3 and IL-10 expression (Fig. 2E). We also isolated cTregs from SPF mice and incubated them with SCFAs in vitro. Foxp3 expression, as well as IL-10 expression and protein production, increased (fig. S13), whereas TGFβ levels remained unchanged (fig. S14). As enhanced suppressive activity (Fig. 2D) could be attributed not only to higher IL-10 levels per cTreg but also to increased cTreg proliferation, we examined cTreg proliferation. cTregs exhibited enhanced proliferation when cultured in the presence of propionate (fig. S15).

In addition, we analyzed the expression patterns of Treg trafficking molecules. Although levels of the chemokine receptor CCR9 or α4β7 integrin were not altered in propionate-treated GF and SPF mice, levels of the cTreg homing receptor, G protein–coupled receptor 15 (GPCR15) (15), did increase (fig. S16). Taken together, these data indicate that SCFAs may have a beneficial effect in SPF mice through their ability to increase Foxp3+IL-10–producing cTregs and cTreg proliferative capacity, as well as to alter cTreg GPCR15 expression.

Considering that SCFAs can influence cTregs directly, we asked if this was a receptor-mediated process. GPCR43 (Ffar2 is the gene that encodes GPCR43) binds SCFAs and, through its expression on innate immune cells, mediates resolution of inflammatory responses (3, 16). We found that cTregs had significantly higher levels of Ffar2 than Tregs isolated from the spleen or MLN (Fig. 3, A and B), and Ffar2 levels were largely dependent on microbiota-derived signals. As a reference, we compared cTreg Ffar2 expression levels to colonic myeloid (CD11b+) cells, which are known to express Ffar2 (3), and found that, on average, CD11b+ cells expressed 1.6-fold more Ffar2 than cTregs (fig. S17).

Fig. 3 Ffar2 mediates SCFA effects on cTregs.

(A) Tregs were isolated from the colon, small intestine (SI), spleen, and MLN of SPF and GF BALB/c mice and purified as in Fig. 1B, and Ffar2 expression was examined by qPCR. Each symbol represents data from three to five individual mice, horizontal lines show the mean, and error bars indicate SD. Data reflect three to seven independent experiments. (B) Lymphocytes were isolated from the colon, small intestine, spleen, and MLN of SPF Ffar2−/− and littermate Ffar2+/+ mice. Cells were stained for CD4, Foxp3, and Ffar2. (Left) A representative flow cytometry histogram comparing colonic Ffar2 expression in Ffar2−/− versus littermate Ffar2+/+ mice. (Right) mean fluorescence intensity (MFI) for Ffar2 for Tregs from the indicated sites. Bars show the mean, error bars denote SD, and data reflect four independent experiments. (C) Colonic LP lymphocytes were isolated from Ffar2−/−and littermate Ffar2+/+ mice exposed to propionate (P) or water alone and stained for CD4 and Foxp3. (Left) Representative dot plots with percentage of CD4+Foxp3+ within the CD45+CD3+ population. (Right) Foxp3+ Treg number for the left panel. Each symbol represents data from individual mice, horizontal lines show the mean, and error bars indicate SD. (D) Ffar2−/− and littermate Ffar2+/+ cTregs were cocultured with splenic Teff cells in media with or without propionate for 96 hours. Percent suppression, y axis; Treg:Teff ratios, x axis. Symbols represent the mean of three independent experiments; error bars show the SD. (E) cTregs were isolated from the LP of Ffar2−/− and littermate Ffar2+/+ mice; purified as in Fig. 1B; cultured in the presence of 0.1 mM propionate or media (pH- and sodium-matched) for 24 hours; and examined for expression of HDAC1, -2, -6, -7, and -9 by RTqPCR. Bars show the mean and error bars the SD of three independent experiments. (F) Whole-cell extracts were generated from cTregs isolated from the LP of Ffar2−/− and littermate Ffar2+/+ mice, purified as in Fig. 1B, and cultured in the presence of 0.1 mM propionate or media (pH- and sodium-matched) for 24 hours. Samples were analyzed by Western blotting for histone acetylation by examining levels of acetylated histone (H3K9); total histone levels were used as a loading control. The Western blot shown is representative of two independent experiments with cTregs cell lysates pooled from 10 to 12 mice per group. A bar graph of densitometry ratios of acetylated histone H3:total histone H3 is shown. Bars represent the mean; error bars denote SD. (A and D) Kruskal-Wallis test with a Dunn’s post hoc test: ***P < 0.001. (C and E) Mann-Whitney U test. (B and F) Student’s t test.

To determine if Ffar2 contributes to cTreg homeostasis, we treated Ffar2−/− mice and Ffar2+/+ littermates with propionate, which has the highest affinity for Ffar2 (17), and measured cTreg numbers. Propionate enhanced cTreg frequency and number in Ffar2+/+, but not Ffar2−/− mice (Fig. 3C). SCFA-mediated, enhanced cTreg suppressive capacity was also dependent on Ffar2 (Fig. 3D). In addition, propionate enhanced Foxp3 and IL-10 expression and IL-10 protein levels in Ffar2+/+ cTregs, but not in Ffar2−/− cTregs (fig. S18). Furthermore, we examined whether propionate could restore cTreg populations and numbers in the setting of vancomycin treatment and found that Ffar2 was necessary (fig. S19).

One mechanism by which SCFAs mediate their effects is histone deacetylase (HDAC) inhibition (18). The HDAC inhibitor trichostatin A increases Treg gene expression and suppressive capacity (19), and HDAC6 and -9 down-regulate nTreg function (20). Given that SCFAs promote cTreg homeostasis in our studies, we hypothesized that SCFAs mediate their effect through HDAC inhibition. We treated Ffar2+/+ and Ffar2−/− mice with propionate and measured HDAC expression. Propionate treatment of Ffar2+/+ mice reduced cTreg HDAC6 and HDAC9 (class IIb and IIa, respectively) expression but did not reduce HDAC1 and HDAC2 (class I) or HDAC7 (class IIa) expression (Fig. 3E). Western blot analysis demonstrated that propionate treatment of cTregs enhanced histone acetylation, which was dependent on Ffar2 expression (Fig. 3F). These results suggest that SCFAs via Ffar2 may affect cTregs through HDAC inhibition.

To further evaluate the effects of SCFAs on cTregs, we used the T cell–transfer model of colitis (21). In this model, lymphopenic mice (e.g., Rag2−/−) are injected with either naïve T cells, which results in severe colitis, or naïve T cells in combination with Tregs, which reduces colitis severity. Propionate or SCFA mix–treated mice injected with naïve T cells and Tregs had less severe weight loss, reduced disease score, and lower levels of colonic proinflammatory cytokines than mice that received water alone (Fig. 4A and fig. S20). SFCA mix or propionate did not affect colitis severity in mice that received only naïve T cells (fig. S20). The frequency and number of cLP Foxp3+ Tregs in mice receiving propionate and SCFA mix increased (Fig. 4, B and C). SCFAs, however, did not result in conversion of naïve T cells to cTregs (Fig. 4, B and C). To evaluate whether these effects were cTreg-intrinsic and dependent on Ffar2, we performed the T cell–transfer colitis model using Rag2−/− recipients, wild-type naïve T cells, and Ffar2+/+ or Ffar2−/− Tregs with or without propionate in the drinking water (Fig. 4D and fig. S21). The effect of propionate on intestinal inflammation was dependent on Ffar2 expression in Tregs, as indicated by the colitis scores (Fig. 4D). Treg cell populations and numbers further substantiated our finding that the propionate effects on cTregs were dependent on Ffar2 (Fig. 4, E and F).

Fig. 4 SCFA exposure ameliorates T cell–transfer colitis in a Treg-intrinsic, Ffar2-dependent manner.

BALB/c Rag2−/− mice were injected with CD4+CD45RBhiCD25lo naive T cells alone or in combination with Tregs. After injection, mice received propionate, SCFA mix, or pH- and sodium-matched drinking water. (A) Weekly percentage of body weight change is shown across the experimental groups from experimental days 0 to 42 (D0 to D42). Symbols show the mean; error bars indicate SD. Data reflect three independent experiments. Colonic LP lymphocytes were isolated and stained for CD4 and Foxp3, and (B) percentage and (C) number of CD4+Foxp3+ within the CD45+CD3+ population are shown. Symbols represent data from individual mice, horizontal lines show the mean, and error bars denote SD. (D to F) C57BL/6 Rag2−/− mice were injected with CD4+CD45RBhiCD25lo naïve T cells alone or in combination with Ffar2+/+ or Ffar2−/− Tregs. After injection, mice received propionate or pH- and sodium-matched drinking water. (D) Histologic colitis score is shown along the y axis, the treatment group and experimental conditions are shown along the x axis. Colonic LP lymphocytes were isolated, and (E) percentage and (F) number of CD4+Foxp3+ within the CD45+CD3+ population are shown. Symbols represent data from individual mice. Horizontal lines show the mean, and error bars denote the SD. Data in (D) to (F) are from two independent experiments. (A to F) Kruskal-Wallis test with a Dunn’s post hoc test: ***P < 0.001; **P < 0.01; *P < 0.05. N/A, not applicable.

Gut microbiota–host immune misadaptation has been implicated in the rising incidence of inflammatory bowel disease, other inflammatory diseases, and obesity (22). The Western dietary pattern, specifically reduced ingestion of plant-based fibers, may be a critical factor that links the gut microbiome to disease (23). Although the gut microbiota composition is divergent across individuals, functional gene profiles are quite similar (24, 25), and alterations to common gut microbial metabolic pathways may affect the production of symbiotic factors, such as SCFAs, which regulate intestinal adaptive immune responses and promote health.

Supplementary Materials

www.sciencemag.org/cgi/content/full/science.1241165/DC1

Materials and Methods

Figs. S1 to S24

Table S1

References (2628)

References and Notes

  1. Acknowledgments: We thank members of the Garrett Lab for discussion, K. Sigrist for embryo rederivation, C. Gallini for genotyping, J. Ramirez for animal husbandry, and A. Rudensky (Memorial Sloan-Kettering Cancer Center) for Foxp3YFP-Cre mice. A material transfer agreement was required for obtaining Ffar2+/− embryos from AstraZeneca. The data presented in this study are tabulated in the main paper and the supplementary materials. This study was supported by National Research Service Award (NRSA) F32DK095506 to P.M.S.; NRSA F32DK098826 to M.R.H.; and grants R01CA154426 and K08AI078942, a Burroughs Wellcome Career in Medical Sciences Award, a Searle Scholars Award, and a Cancer Research Institute Investigator Award to W.S.G. A patent (U.S. 61/800,299) was filed with the U.S. Patent and Trademark Office on 15 March 2013 on short chain fatty acids and their regulation of colonic regulatory T cell populations and function by Harvard University, with W.S.G. and P.M.S. as co-inventors. P.M.S., M.R.H., and W.S.G. designed and carried out the experiments and wrote the manuscript; M.M., C.A.G., and N.P. carried out experiments; J.N.G. performed all histological assessment; and M.B.-Y provided Ffar2+/− embryos.
View Abstract

Navigate This Article