Research Article

Microbiota-activated PPAR-γ signaling inhibits dysbiotic Enterobacteriaceae expansion

See allHide authors and affiliations

Science  11 Aug 2017:
Vol. 357, Issue 6351, pp. 570-575
DOI: 10.1126/science.aam9949

Healthy guts exclude oxygen

Normally, the lumen of the colon lacks oxygen. Fastidiously anaerobic butyrate-producing bacteria thrive in the colon; by ablating these organisms, antibiotic treatment removes butyrate. Byndloss et al. discovered that loss of butyrate deranges metabolic signaling in gut cells (see the Perspective by Cani). This induces nitric oxidase to generate nitrate in the lumen and disables β-oxidation in epithelial cells that would otherwise mop up stray oxygen before it enters the colon. Simultaneously, regulatory T cells retreat, and inflammation is unchecked, which contributes yet more oxygen species to the colon. Then, facultative aerobic pathogens, such as Escherichia coli and Salmonella enterica, can take advantage of the altered environment and outgrow any antibiotic-crippled and benign anaerobes.

Science, this issue p. 570; see also p. 548

Abstract

Perturbation of the gut-associated microbial community may underlie many human illnesses, but the mechanisms that maintain homeostasis are poorly understood. We found that the depletion of butyrate-producing microbes by antibiotic treatment reduced epithelial signaling through the intracellular butyrate sensor peroxisome proliferator–activated receptor γ (PPAR-γ). Nitrate levels increased in the colonic lumen because epithelial expression of Nos2, the gene encoding inducible nitric oxide synthase, was elevated in the absence of PPAR-γ signaling. Microbiota-induced PPAR-γ signaling also limits the luminal bioavailability of oxygen by driving the energy metabolism of colonic epithelial cells (colonocytes) toward β-oxidation. Therefore, microbiota-activated PPAR-γ signaling is a homeostatic pathway that prevents a dysbiotic expansion of potentially pathogenic Escherichia and Salmonella by reducing the bioavailability of respiratory electron acceptors to Enterobacteriaceae in the lumen of the colon.

A balanced gut microbiota is characterized by the dominance of obligate anaerobic members of the phyla Firmicutes and Bacteroidetes, whereas an expansion of facultative anaerobic Enterobacteriaceae (phylum Proteobacteria) is a common marker of gut dysbiosis (1) (fig. S1). Obligate anaerobic bacteria prevent dysbiotic expansion of facultative anaerobic Enterobacteriaceae, in part by limiting the generation of host-derived nitrate and oxygen (2, 3). It is not known which host-signaling pathways are triggered by the gut microbiota to limit the availability of these respiratory electron acceptors. We found that disruption of the gut microbiota by streptomycin treatment increased the bioavailability of host-derived nitrate in the lumen of the large intestine. Increased recovery of a wild-type Escherichia coli strain by comparison with an isogenic derivative deficient for nitrate respiration (napA narG narZ mutant) was observed in mice (C57BL/6 from The Jackson Laboratory) infected with a 1:1 mixture of both strains (Fig. 1A). Supplementation of streptomycin-treated mice with the inducible nitric oxide synthase (iNOS) inhibitor aminoguanidine hydrochloride (AG) abrogated the growth advantage conferred upon E. coli by nitrate respiration (Fig. 1A), supporting the notion that luminal nitrate was host-derived (2, 4).

Fig. 1 The PPAR-γ agonist butyrate limits the availability of nitrate by repressing Nos2 expression.

(A) Streptomycin (Strep)–treated mice (N = 8 animals) were inoculated with a 1:1 mixture of E. coli wild type (wt) and napA narG narZ mutant and received rosiglitazone (Rosi) or aminoguanidine (AG) supplementation. The competitive index [(CI), the ratio of wt and napA narG narZ mutant recovered from colon contents] was determined 3 days after inoculation. (B) Polarized Caco-2 cells (N = 4) grown in a tissue culture medium received IFN-γ/IL-22, Rosi, or AG treatment. Nitrate produced in the apical compartment was determined by a modified Griess assay. (C to F) Mice (N = 8) were mock-treated (inoculated with vehicle control) or treated with Strep, and organs were collected 3 days later. (C) Abundance of Clostridia in colon contents was determined by quantitative real-time polymerase chain reaction (PCR) using class-specific primers for Clostridia 16S ribosomal RNA genes. (D) Relative abundance of families belonging to the class Clostridia determined by 16S profiling of DNA isolated from colon contents. (E) Butyrate concentration was determined in cecal contents by gas chromatography. (F) Transcript level of Nos2 in colonocyte preparations was determined by real-time PCR. (G and H) Colonocytes were isolated 3 days after treatment with streptomycin from mice (N = 8) receiving the indicated supplementation and transcript levels of Angptl4 (G) and Nos2 (H), as determined by quantitative real-time PCR. (I and J) Mice (N = 6) were mock-treated or received the PPAR-γ antagonist GW9662 and were inoculated with E. coli indicator strains. Numbers of E. coli (I) and the CI of indicator strains (J) were determined 3 days after inoculation. (C and E to I) Bars represent geometric means ± SE. (A, B, and J) Circles represent measurements from individual animals [(A) and (J)] or wells (B), and bars represent geometric means. *P < 0.05; **P < 0.01.

To model nitrate production by the colonic epithelium, we induced NOS2 expression in human colonic epithelial cancer (Caco2) cells by stimulation with interferon-γ (IFN-γ) and interleukin (IL)–22 (model epithelia). We exposed the model epithelia to butyrate, a fermentation product of the gut microbiota that serves as the main carbon source of colonic epithelial cells (colonocytes) (5). Butyrate significantly reduced NOS2 expression (P < 0.05) (fig. S2A), lowered iNOS synthesis (P < 0.05) (6) (fig. S2B), and diminished epithelial generation of nitrate (P < 0.05) (Fig. 1B), a product of nitric oxide decomposition in the intestinal lumen (4). The host can sense butyrate by using the nuclear receptor PPAR-γ (peroxisome proliferator–activated receptor γ), which is synthesized at high levels in colonocytes (7) and does not respond to other short-chain fatty acids, such as acetate or propionate (8). To determine whether PPAR-γ repressed iNOS synthesis, we stimulated model epithelia with the PPAR-γ agonist rosiglitazone. Rosiglitazone treatment significantly blunted NOS2 expression (P < 0.05) (fig. S2A), reduced iNOS synthesis (P < 0.05) (fig. S2B), lowered nitrate production (P < 0.01) (Fig. 1B), and induced synthesis and nuclear localization of PPAR-γ in model epithelia (fig. S2C). On the basis of these data, we hypothesized that microbiota-derived butyrate suppresses iNOS synthesis in the gut by stimulating PPAR-γ signaling in colonocytes (fig. S1).

Streptomycin-mediated depletion of a PPAR-γ agonist drives growth by nitrate respiration

To test our hypothesis, we used mice to investigate whether streptomycin treatment would deplete butyrate-producing bacteria, thereby increasing Nos2 expression in colonocytes. Streptomycin treatment reduced bacterial numbers in colon contents (fig. S3A) and significantly (P < 0.01) reduced the abundance of Clostridia (phylum Firmicutes) (Fig. 1C and fig. S3B), which are obligate anaerobes that include abundant butyrate producers (9) (fig. S3C), specifically Lachnospiraceae and Ruminococcaceae (Fig. 1D and fig. S3D). The changes in the microbiota composition correlated with a significant (P < 0.01) drop in the cecal butyrate concentration (Fig. 1E) and significantly (P < 0.05) elevated Nos2 expression in murine colonocyte preparations (Fig. 1F).

We next investigated the role of PPAR-γ in altering epithelial gene expression. Streptomycin treatment reduced epithelial expression of Angptl4, a gene positively regulated by PPAR-γ (8), and expression was restored in streptomycin-treated mice that received the PPAR-γ agonist rosiglitazone (Fig. 1G). Treatment of mice with the PPAR-γ antagonist 2-chloro-5-nitrobenzanilide (GW9662) mimicked the reduction (P < 0.05) in Angptl4 transcript levels observed after streptomycin treatment (Fig. 1G). The effects of each treatment on epithelial Nos2 expression (Fig. 1H) were opposite of those observed for expression of Angptl4 (Fig. 1G), which supported the idea that PPAR-γ negatively regulates Nos2 (fig. S1).

Next, we used E. coli indicator strains to investigate whether silencing PPAR-γ signaling would increase the bioavailability of nitrate in the colon. To this end, mice were inoculated with a 1:1 mixture of a nitrate respiration–proficient indicator strain (E. coli wild type) and an isogenic nitrate respiration–deficient indicator strain (napA narG narZ mutant). Treatment with the PPAR-γ agonist rosiglitazone abrogated the fitness advantage conferred to wild-type E. coli by nitrate respiration in streptomycin-treated mice (Fig. 1A). To investigate whether inhibition of PPAR-γ signaling would support a nitrate respiration–dependent expansion of E. coli without antibiotic treatment, mice were mock-treated (inoculated with sterile phosphate-buffered saline) or treated with the PPAR-γ antagonist GW9662 and then infected with E. coli indicator strains. Treatment with GW9662 significantly increased the overall number of E. coli recovered from mouse colons (P < 0.05) (Fig. 1I) by driving nitrate respiration–dependent E. coli expansion, as shown by increased recovery of the wild type as compared with a nitrate respiration–deficient mutant (P < 0.05) (Fig. 1J). Next, we wanted to determine whether treatment with a PPAR-γ antagonist would increase the abundance of endogenous Enterobacteriaceae. Whereas endogenous Enterobacteriaceae were not detected in C57BL/6 mice from Jackson, C57BL/6 mice from Charles River Laboratories carried endogenous E. coli strains producing nitrate reductase activity (fig. S4A). Treatment of Charles River mice with GW9662 significantly (P < 0.05) increased the abundance of endogenous E. coli, which could be abrogated by supplementation with the iNOS inhibitor AG (fig. S4B).

Epithelial PPAR-γ signaling limits luminal nitrate availability

To exclude the possibility that our results were due to off-target effects of chemical agonists or antagonists, we generated mice lacking PPAR-γ in the intestinal epithelium (Ppargfl/flVillincre/– mice) and compared them with wild-type littermate control animals (Ppargfl/flVillin/ mice). Mice lacking epithelial PPAR-γ signaling exhibited significantly elevated transcript levels of Nos2 in the colonic epithelium (P < 0.01) (Fig. 2A), which resulted neither from a reduced abundance of butyrate-producing bacteria in their gut microbiota (Fig. 2B and fig. S5) nor from lower butyrate levels in their cecal contents (Fig. 2C). Inoculation with E. coli indicator strains revealed that epithelial PPAR-γ deficiency increased the bioavailability of nitrate through a mechanism that required iNOS activity, because treatment with the iNOS inhibitor AG abrogated the nitrate respiration–dependent growth advantage (P < 0.05) (Fig. 2D). Similar results were obtained when mice were infected with the murine E. coli isolate JB2 (fig. S4C), which produced nitrate reductase activity (fig. S4A). To test directly whether genetic ablation of epithelial PPAR-γ signaling increased the concentration of nitrate in the intestinal lumen, we measured the concentration of this electron acceptor in colonic mucus scrapings, which revealed a significant increase (P < 0.01) in mice lacking epithelial PPAR-γ signaling compared with littermate controls (Fig. 2E).

Fig. 2 Microbiota-induced epithelial PPAR-γ signaling limits nitrate availability in the colon.

(A) Nos2 expression in the colonic epithelium of mice (N = 6) was determined by real-time PCR in Ppargfl/flVillincre/– mice (Pparg), which lack PPAR-γ in epithelial cells, and in littermate control Ppargfl/flVillin/ mice (WT). (B) Relative abundance of families belonging to the class Clostridia in colon contents of mice (N = 6) was determined by 16S profiling. (C) Butyrate concentration was determined in cecal contents of mice (N = 6) by gas chromatography. (D) Mice (N = 6) were inoculated with a 1:1 mixture of E. coli wild type (wt) and napA narG narZ mutant and received aminoguanidine (AG) supplementation or vehicle control. The competitive index (CI) was determined 3 days after inoculation. (E) Concentration of nitrate in the colonic mucus layer was determined in groups of animals (N = 9) by a modified Griess assay. (F and G) Streptomycin-treated mice (N = 6) were inoculated with E. coli indicator strains and received supplementation with tributyrin or a community of 17 human Clostridia isolates (C17). The butyrate concentration in cecal contents (F) and the CI in colon contents (G) were determined 3 days after inoculation. (A, C, and F) Bars represent geometric means ± SE. (D, E, and G) Each circle represents data from an individual animal, and black bars represent geometric means. *P < 0.05; **P < 0.01; ns, not statistically significantly different.

Mice lacking epithelial PPAR-γ signaling were treated with streptomycin, infected the next day with E. coli indicator strains, and inoculated 1 day later with a community of 17 human Clostridia isolates (10). Inoculation with the Clostridia isolates restored cecal butyrate concentrations (P < 0.01) (Fig. 2F) and suppressed nitrate respiration–dependent growth of E. coli in streptomycin-treated littermate control mice. However, nitrate respiration–dependent growth of E. coli was not suppressed in streptomycin-treated mice lacking epithelial PPAR-γ signaling (P < 0.05) (Fig. 2G). To directly test whether butyrate was responsible for inhibiting nitrate respiration of E. coli in littermate control animals, mice were treated with streptomycin, infected the next day with E. coli indicator strains, and inoculated 1 day later with 1,2,3-tributyrylglycerol (tributyrin). Tributyrin, a natural ingredient of butter, exhibits delayed absorption in the small intestine compared with butyrate, and its degradation in the large intestine increases luminal butyrate concentrations (11). Tributyrin supplementation restored cecal butyrate concentrations (P < 0.01) (Fig. 2F), which abrogated nitrate respiration–dependent growth of E. coli in streptomycin-treated littermate control mice but not in streptomycin-treated mice lacking epithelial PPAR-γ signaling (P < 0.05) (Fig. 2G). Collectively, these data support the idea that microbiota-derived butyrate maintains gut homeostasis by inducing epithelial PPAR-γ signaling, which in turn limits nitrate respiration–dependent dysbiotic E. coli expansion (fig. S1).

Lack of epithelial PPAR-γ signaling increases colonocyte oxygenation during colitis

Colonocytes obtain energy through β-oxidation of microbiota-derived butyrate, which consumes a considerable amount of oxygen, thereby rendering the epithelium hypoxic (12). However, after a streptomycin-mediated depletion of butyrate (Fig. 1E), colonocytes switch their energy metabolism to converting glucose into lactate (anaerobic glycolysis) (11). Consistent with this metabolic reprogramming, streptomycin treatment increased the concentration of lactate (Fig. 3A) and reduced adenosine triphosphate (ATP) levels (Fig. 3B) in primary murine colonocyte preparations. Anaerobic glycolysis does not consume oxygen, which then permeates through the epithelium into the gut lumen (3, 11). This scenario was supported by increased recovery of aerobic respiration–proficient (Nissle 1917 wild type) E. coli compared with an E. coli strain that is impaired for aerobic respiration under microaerophilic conditions (cydAB mutant) from the colon of streptomycin-treated mice (Fig. 3C). Similar results were obtained when the result was repeated with a different E. coli strain (MG1655) (fig. S6A). Increasing the concentration of the PPAR-γ antagonist butyrate in streptomycin-treated mice, either by inoculation with a community of 17 human Clostridia isolates or by supplementation with tributyrin (fig. S6B), appeared to reduce the bioavailability of oxygen, as E. coli indicator strains that were proficient (wild type) or deficient (cydAB mutant) for aerobic respiration under microaerophilic conditions were recovered equally in the gut (Fig. 3C). Furthermore, the aerobic growth benefit observed in streptomycin-treated mice inoculated with E. coli indicator strains was abrogated by treatment with the PPAR-γ agonist rosiglitazone (Fig. 3C). Surprisingly, E. coli indicator strains were recovered at the same ratio from mice lacking epithelial PPAR-γ signaling and from their littermate controls (Fig. 3D), suggesting that reducing PPAR-γ signaling alone was not sufficient for increasing the bioavailability of oxygen.

Fig. 3 Lack of epithelial PPAR-γ signaling increases colonocyte oxygenation during colitis.

(A and B) Groups of mice (N = 5) were mock-treated or treated with streptomycin (Strep), and colonocytes were isolated 1 day later to measure intracellular concentrations of lactate (A) or ATP (B). (C) Groups (N = 6) of streptomycin-treated or mock-treated mice were inoculated with a 1:1 mixture of E. coli wild type (wt) and cydAB mutant and received supplementation with rosiglitazone (Rosi), tributyrin, or a community of 17 human Clostridia isolates (C17). (D) Groups of mice (N = 6) receiving no supplementation or water supplemented with 1% dextran sulfate sodium (DSS) were inoculated with a 1:1 mixture of E. coli wild type (wt) and cydAB mutant. (C and D) The competitive index (CI) was determined 3 days after inoculation. (E and F) Groups of mice (N = 6) were inoculated with the indicated Salmonella strain mixtures. (F) Mice received no supplementation or water supplemented with 1% DSS. (G and H) Mice (N = 6) were treated as indicated and were injected intraperitoneally with pimonidazole 1 hour before euthanasia. Binding of pimonidazole was detected using hypoxyprobe-1 primary antibody and a Cy-3 conjugated goat anti-mouse secondary antibody (red fluorescence) in sections of the colon that were counterstained with DAPI (4′,6-diamidino-2-phenylindole) nuclear stain (blue fluorescence). (G) Representative images. Scale bars, 50 μm. (H) A veterinary pathologist scored blinded sections for hypoxia staining. Each circle represents data from one animal. (A and B) Bars represent geometric means ± SE. (C to F) Each circle represents data from an individual animal, and black bars represent geometric means. *P < 0.05; **P < 0.01; ***P < 0.001; ns, not statistically significantly different.

PPAR-γ signaling activates mitochondrial β-oxidation in alternatively activated (M2) macrophages (13). Because IFN-γ signaling drives the energy metabolism of macrophages toward anaerobic glycolysis (13), we hypothesized that in addition to silencing PPAR-γ signaling, metabolic reprogramming of colonocytes might also require an inflammatory signal (fig. S1). To test this idea, we turned to S. enterica serovar Typhimurium (S. Typhimurium), a pathogen that employs two type III secretion systems to trigger intestinal inflammation and uses the cyxAB genes, encoding cytochrome bd-II oxidase, for its subsequent aerobic expansion in the intestinal lumen (3). When mice were infected with S. Typhimurium strains that were proficient (wild type) or deficient (cyxA mutant) for aerobic respiration under microaerophilic conditions, a benefit provided by aerobic respiration was observed in mice lacking epithelial PPAR-γ signaling but not in littermate control animals (Fig. 3E). To investigate whether inflammatory responses elicited by Salmonella virulence factors were required to increase the bioavailability of oxygen, we inactivated the two type III secretion systems essential for Salmonella enteropathogenicity through mutations in invA and spiB (3). Consistent with our hypothesis, aerobic respiration no longer provided a benefit to the pathogen in mice lacking epithelial PPAR-γ signaling when mice were infected with avirulent S. Typhimurium strains that were either proficient (invA spiB mutant) or deficient (invA spiB cyxA mutant) for aerobic respiration under microaerophilic conditions (Fig. 3E).

To test whether, in addition to genetic ablation of PPAR-γ signaling, an inflammatory signal was needed to increase luminal oxygen bioavailability, mice received low-dose (1% in drinking water) dextran sodium sulfate (DSS) treatment, which elicited inflammatory changes as indicated by a reduction in colon length (fig. S6C). Inoculation with E. coli indicator strains revealed that aerobic respiration provided a larger growth benefit in DSS-treated mice lacking epithelial PPAR-γ signaling compared with their DSS-treated littermate controls (Fig. 3D). Similarly, inoculation with avirulent S. Typhimurium strains that were either proficient (invA spiB mutant) or deficient (invA spiB cyxA mutant) for aerobic respiration under microaerophilic conditions provided evidence for increased oxygen bioavailability only in DSS-treated mice that lacked epithelial PPAR-γ signaling (Fig. 3F).

Next, we investigated whether either DSS treatment or infection with wild-type S. Typhimurium would increase epithelial oxygenation in mice lacking epithelial PPAR-γ signaling. To this end, we visualized the hypoxia of surface colonocytes by using the exogenous hypoxic marker pimonidazole, which is reduced under hypoxic conditions to hydroxylamine intermediates that irreversibly bind to nucleophilic groups in proteins or DNA (14, 15). Genetic ablation of PPAR-γ signaling was not sufficient to reduce epithelial hypoxia. However, DSS treatment or infection with wild-type S. Typhimurium increased epithelial oxygenation in mice lacking epithelial PPAR-γ signaling, whereas hypoxia staining remained unchanged in littermate control animals (Fig. 3, G and H).

PPAR-γ signaling and Tregs cooperate to maintain colonocyte hypoxia

Although streptomycin treatment reduced PPAR-γ signaling (Fig. 1G) by depleting microbiota-derived butyrate (Fig. 1E and fig. S6B), the findings shown above suggested that reducing PPAR-γ signaling was necessary but not sufficient for increasing oxygen bioavailability in the colon, because DSS-induced inflammation or Salmonella virulence factors were required for increasing colonocyte oxygenation (Fig. 3, D to H). Although streptomycin treatment does not lead to overt inflammation, disruption of the gut microbiota reduced concentrations of other microbiota-derived short-chain fatty acids (fig. S6D) [in addition to butyrate (fig. S6B)] that signal through G protein–coupled receptor (GPR) 43, GPR109A, and histone deacetylases expressed by T cells, dendritic cells, and macrophages, respectively, to reduce intestinal inflammation (1619). Engagement of these host cell receptors by short-chain fatty acids induces maturation and expansion of colonic regulatory T cells (Tregs), a cell type that limits proinflammatory responses (2024). Consistent with previous reports showing that antibiotic-mediated depletion of short-chain fatty acids leads to contraction of the Treg population in the colonic mucosa (20, 23), we observed that streptomycin treatment shrank the pool of colonic Tregs (CD3+-enriched CD4+ FOXP3+ cells) to one-third of its normal size (Fig. 4A and figs. S6E and S7). Thus, during antibiotic treatment, the second input that increases oxygen bioavailability in the colon might be provided by contraction of the colonic Treg population, which increases the inflammatory tone of the mucosa (fig. S1).

Fig. 4 Microbiota-induced PPAR-γ signaling and Tregs cooperate to limit the bioavailability of oxygen in the colon.

(A and B) Groups of mice (N = 4) were treated with streptomycin (A) or with anti-CD25 antibody (B), and CD3+-enriched live colonic cells were analyzed for expression of CD4 and FOXP3 by flow cytometry. (C) Groups of mice (N = 6) were treated with anti-CD25 antibody or isotype control and, 10 days later, were inoculated with a 1:1 mixture of an avirulent Salmonella strain (invA spiB mutant) and an avirulent Salmonella strain lacking cytochrome bdII oxidase (invA spiB cyxA mutant). (D) Groups (N = 6) of streptomycin-treated or mock-treated mice were inoculated with Salmonella indicator strains and received supplementation with tributyrin or a community of 17 human Clostridia isolates (C17). (C and D). The CI was determined 4 days after inoculation. (E to G) Groups of mice (N = 6) were treated with anti-CD25 antibody or isotype control antibody, and colonocytes were isolated to measure intracellular concentrations of lactate (E), ATP (F), or mitochondrial cytochrome c oxidase activity (G). (G) The width of the box shows the interquartile range, the horizontal line shows the median, and the top and bottom lines show the highest and lowest values, respectively. (H) Groups (N = 6) of anti-CD25–treated mice lacking epithelial PPAR-γ signaling or untreated wild-type mice (WT) were infected with a 1:1 mixture of the indicated E. coli strains. The CI was determined 4 days after inoculation. (A, B, E, and F) Bars represent geometric means ± SE. (C, D, and H) Each circle represents data from an individual animal, and black bars represent geometric means. *P < 0.05; **P < 0.01; ***P < 0.001; ns, not statistically significantly different.

Treatment with anti-CD25 antibody reduced the pool of colonic Tregs (Fig. 4B and fig. S6F) by a magnitude similar to that observed after streptomycin treatment (Fig. 4A) and elicited inflammatory changes in mice lacking epithelial PPAR-γ signaling, as indicated by a reduction in colon length (fig. S6G). When anti-CD25–treated mice were infected with avirulent S. Typhimurium strains that were either proficient (invA spiB mutant) or deficient (invA spiB cyxA mutant) for aerobic respiration under microaerophilic conditions, there was no benefit provided by aerobic respiration to S. Typhimurium in wild-type mice. Hence, depletion of Tregs was not sufficient for increasing oxygen bioavailability. In contrast, depletion of Tregs increased oxygen bioavailability in mice lacking epithelial PPAR-γ signaling but not in wild-type littermate control mice (Fig. 4C and fig. S6H). Genetic ablation of PPAR-γ signaling combined with Treg depletion phenocopied the effects of streptomycin treatment on the recovery of avirulent S. Typhimurium strains proficient or deficient for aerobic respiration under microaerophilic conditions (Fig. 4D). Depletion of Tregs increased epithelial oxygenation in mice lacking epithelial PPAR-γ signaling but not in littermate control mice (Fig. 3, G and H). Consistent with metabolic reprogramming toward anaerobic glycolysis, Treg depletion increased intracellular lactate levels and lowered ATP concentrations in colonocyte preparations from mice lacking epithelial PPAR-γ signaling but not from littermate controls (Fig. 4, E and F). Measurement of mitochondrial cytochrome c oxidase activity revealed that Treg depletion caused a significant (P < 0.001) reduction in oxygen consumption in colonocyte preparations of mice lacking epithelial PPAR-γ signaling but not in littermate control animals (Fig. 4G).

To further study how colonic Tregs and PPAR-γ signaling cooperate to limit respiratory growth of facultative anaerobic bacteria, mice lacking epithelial PPAR-γ signaling were treated with anti-CD25 antibody and infected with a 1:1 mixture of wild-type E. coli and a mutant lacking cytochrome bd oxidase and nitrate reductases (cydAB napA narG narZ mutant). The competitive index was ~1000 times greater (P < 0.01) in anti-CD25–treated mice lacking epithelial PPAR-γ compared with wild-type littermate control animals (Fig. 4H). Similar results were obtained when mice were infected with individual bacterial strains (fig. S6I).

The emerging picture is that epithelial hypoxia maintains anaerobiosis in the colon to drive the microbial community toward a dominance of obligate anaerobes, which produce short-chain fatty acids. In turn, short-chain fatty acids sustain Tregs and epithelial PPAR-γ signaling, which cooperatively drives the energy metabolism of colonocytes toward β-oxidation of microbiota-derived butyrate to preserve epithelial hypoxia, thereby closing a virtuous cycle maintaining homeostasis of a healthy gut. PPAR-γ signaling also activates expression of β-defensins, which might contribute to shaping the intestinal environment (25). An antibiotic-induced lack of epithelial PPAR-γ signaling and a contraction of the Treg pool cooperatively drive a metabolic reorientation of colonocytes toward anaerobic glycolysis, thereby increasing epithelial oxygenation and consequently elevating oxygen bioavailability to promote an expansion of Enterobacteriaceae (fig. S1), a common marker of dysbiosis (1). Thus, an expansion of Enterobacteriaceae in the gut-associated microbial community is a microbial signature of epithelial dysfunction, which has important ramifications for targeting PPAR-γ signaling as a potential intervention strategy.

Supplementary Materials

www.sciencemag.org/content/357/6351/570/suppl/DC1

Materials and Methods

Figs. S1 to S7

References (2635)

References and Notes

Acknowledgments: We acknowledge the Host-Microbe Systems Biology Core (HMSB Core) at the University of California at Davis School of Medicine for expert technical assistance with microbiota sequence analysis. The data reported in the manuscript are tabulated in the main paper and the supplementary materials. M.X.B. and A.J.Bä. filed invention report number 0577501-16-0038 at iEdison.gov for a treatment to prevent postantibiotic expansion of Enterobacteriaceae. This work was supported by Public Health Service grants AI060555 (S.A.C.), TR001861 (E.E.O.), AI112241 (C.A.L.), DK087307 (C.G.), AI109799 (R.M.T.), AI112258 (R.M.T.), AI112949 (A.J.Bä. and R.M.T.), AI096528 (A.J.Bä.), AI112445 (A.J.Bä.), AI112949 (A.J.Bä.), and AI114922 (A.J.Bä.). K.L.L. was supported by an American Heart Association Predoctoral Fellowship (15PRE21420011).
View Abstract

Navigate This Article