Report

Disease tolerance mediated by microbiome E. coli involves inflammasome and IGF-1 signaling

See allHide authors and affiliations

Science  30 Oct 2015:
Vol. 350, Issue 6260, pp. 558-563
DOI: 10.1126/science.aac6468

The benefits of Escherichia coli

Infection and intestinal damage can trigger severe muscle wasting and loss of fat in mice. How this happens is poorly understood. Palaferri Schieber et al. discovered a protective Escherichia coli strain in their mouse colony. Mice intestinally colonized with the E. coli and infected with the food-poisoning bug Salmonella or with the lung pathogen Burkholderia did not waste away. Without the E. coli, similarly infected mice became fatally ill. The protective E. coli stimulated an innate immune mechanism that ensured that muscle-signaling pathways were not damaged by infection. Thus, the friendly E. coli allowed its host to tolerate and survive the pathogens.

Science, this issue p. 558

Abstract

Infections and inflammation can lead to cachexia and wasting of skeletal muscle and fat tissue by as yet poorly understood mechanisms. We observed that gut colonization of mice by a strain of Escherichia coli prevents wasting triggered by infections or physical damage to the intestine. During intestinal infection with the pathogen Salmonella Typhimurium or pneumonic infection with Burkholderia thailandensis, the presence of this E. coli did not alter changes in host metabolism, caloric uptake, or inflammation but instead sustained signaling of the insulin-like growth factor 1/phosphatidylinositol 3-kinase/AKT pathway in skeletal muscle, which is required for prevention of muscle wasting. This effect was dependent on engagement of the NLRC4 inflammasome. Therefore, this commensal promotes tolerance to diverse diseases.

Infections and inflammation lead to profound metabolic alterations that are primarily driven by responses of muscle, fat, and liver tissues (1, 2). Coupled with loss of appetite, dysregulated metabolism can lead to the severe metabolic pathology called wasting syndrome (cachexia). Such wasting constitutes loss of skeletal muscle (with and without adipose tissue depletion), resulting in weight loss (2). Tuberculosis, sepsis, and HIV, as well as inflammatory diseases including colitis (1, 3, 4), can trigger wasting. Wasting can also interfere with therapeutic interventions and cause untreatable morbidity and mortality (5).

Mammals have coevolved with a complex gut microbial community, the gut microbiota, and depend on the metabolic benefits that it confers on the host (6, 7). Here, we have investigated whether constituents of the microbiota have any protective effect during metabolic dysregulation caused by gut trauma and/or infection.

Intestinal injury and inflammation can cause muscle wasting and inflammatory bowel disease (IBD), and Crohn’s disease (CD) patients suffer from muscle wasting (8). Antibiotics cause ecological perturbations in the microbiota (9), and coupling antibiotics with disease models can reveal how specific constituents of the microbiota impact disease. The dextran sulfate sodium (DSS) intestinal injury model is one of the best-studied models of IBD/CD–associated pathologies, including intestinal inflammation, mucus erosion, and microbiota decompartmentalization (10). C57Bl/6 mice treated with DSS exhibited muscle and fat wasting that was associated with anorexia, colonic inflammation, and bloody diarrhea (Fig. 1, A and B, and fig. S1). Administration of the broad-spectrum antibiotic cocktail ampicillin, vancomycin, neomycin, and metronidazole (AVNM) had no significant impact on the severity of DSS-induced wasting in C57Bl/6 mice obtained from Jackson Laboratories (Jax mice) (fig. S2). Surprisingly, we found that C57Bl/6 from the University of California, Berkeley, colony (CB mice) showed significantly less wasting when DSS was coupled with AVNM treatment (Fig. 1C and fig. S2).

Fig. 1 Effects of E. coli O21:H+ on DSS-induced skeletal muscle atrophy.

(A and B) Jax mice were treated with 5% DSS. (A) Percent original weight. (B) Leg muscle mass at 7 days. Representative experiment (n = 4 to 6 independent experiments) using five animals per group, mean ± S.D. Unpaired t test. (C) Percent original weight of colony-born C57Bl/6 mice (University of California, Berkeley) treated with ampicillin, vancomycin, neomycin, metronidazole, and 5% DSS (methods). Representative experiment (n = 3 independent experiments) using four animals per group, mean ± SD. Unpaired t test. (D) Percent original weight of 5% DSS–treated Jax mice gavaged with 5 × 108 colony-forming units (CFU) E. coli O21:H+ or vehicle alone (methods). Representative experiment (n = 6 independent experiments) using four animals per group, mean ± SD. Unpaired t test. Untreated control mice ± E. coli O21:H+ are shown in fig. S3. (E and F) Germ-free Swiss Webster (SW) mice gavaged with 5 × 108 E. coli O21:H+ or E. coli MG1655 (supplementary materials, materials and methods) and treated with 5% DSS. (E) Percent original weight. (F) Leg muscle mass at 5 days from a subset of mice in (E) that were weight matched at day 0. Representative experiment (n = 4 independent experiments) using five to seven animals per group, mean ± SD. Unpaired t test. Monocolonized SW controls are shown in fig. S3. (G) Percent survival of Jax mice administered E. coli O21:H+ or heat-killed E. coli O21:H+ (supplementary materials, materials and methods) and treated with 5% DSS for 7 days, after which DSS was no longer administered. Results are from two independent experiments using four to seven mice per group per experiment. Log rank analysis. Survival analyses of untreated control ± E. coli O21:H+ mice have been carried out for 1 year with no deaths from either group. Vehicle/DSS–treated mice exhibit similar death kinetics as heat-killed/DSS treated mice. ****P < 0.0001, ***P < 0.0005, **P < 0.005, *P < 0.05. EDL, extensor digitorum longus; TA, tibialis anterior. Experiments terminated for postmortem analyses at indicated day when one or more mice lost [(A), (B), (D), (E), and (F)] >15% body weight in accordance with our Salk animal protocol or (C) >20% weight loss in accordance with our animal protocol (University of California, Berkeley).

We hypothesized that differences in the microbiota composition between Jax and CB mice may account for the differences we observed in wasting pathogenesis in response to the AVNM cocktail plus DSS. We cultured an AVNM-resistant (AVNMR) population of bacteria from the ceca of CB mice that was not present in Jax mice (fig. S2). On cohousing with AVNM-treated CB mice, AVNM-treated Jax mice became colonized with AVNMR bacteria and were now protected from DSS-induced wasting. Hence, protection from wasting and CB-associated AVNMR bacteria can be horizontally transferred to otherwise wasting-susceptible animals (fig. S2).

We performed culture-independent analysis of amplicons generated by primers specific to the V3 and V4 variable regions of bacterial 16S ribosomal RNA genes of cecal content samples from CB and Jax mice treated with AVNM for 5 days (11). The bacterial communities detected were compared phylogentically (11). We found that AVNM-treated CB mice hosted large numbers of Escherichia spp. compared with AVNM-treated Jax mice (fig. S2). Using culturing techniques, serotyping, and genetic and molecular characterization (11), we isolated a single AVNMR Escherichia from the ceca of CB mice that we classified as an E. coli O21:H+ strain that was absent from Jax mice (fig. S2) (11). We performed multilocus sequence typing (MLST) by analyzing seven loci of E. coli O21:H+ (tables S1 and S2) (11) and found this E. coli to be a perfect match at these loci for E. coli of the strain ST101.

Oral administration of E. coli O21:H+ resulted in colonization of the intestine of Jax mice but had no significant effect on weight, muscle mass, fat mass, or food consumption under homeostatic conditions (fig. S3). However, on DSS treatment, E. coli O21:H+ Jax mice showed significantly less wasting than did DSS-treated vehicle mice (Fig. 1D and fig. S4). Jax mice that were administered heat-killed E. coli O21:H+ or live doses of the commensal strain E. coli MG1655 exhibited similar wasting as that of vehicle-control mice (fig. S4). In accordance with our animal protocol, we used >15% weight loss as a clinical end point. We terminated all experiments on the day after challenge in which one or more animals exhibited >15% weight loss, then harvested tissues for postmortem analyses. This time point is consistent for a particular disease model but varies among models owing to different disease kinetics. In all experiments, the number of animals and replication numbers were chosen as statistically necessary and are reported in the figure legends (11).

We monocolonized germ-free Swiss Webster (SW) mice by means of oral gavage with E. coli O21:H+ or E. coli MG1655, treated them with DSS, and monitored wasting. These E. coli strains colonized the intestines of SW mice equivalently (fig. S4). Compared with E. coli MG1655–colonized Jax mice, E. coli MG1655–monocolonized SW mice were more susceptible to DSS, with some animals exhibiting >15% weight loss 5 days after treatment owing to muscle and fat loss (Fig. 1, E and F, and fig. S4). In contrast, E. coli O21:H+–monocolonized mice did not lose weight (Fig. 1, E and F, ands fig. S4). Thus, E. coli O21:H+–mediated protection does not require other microbiota constituents.

E. coli O21:H+ and E. coli MG1655 LPS were equivalent in their ability to stimulate the innate immune mediator Toll-like receptor 4 (TLR4) (fig. S4). E. coli O21:H+–monocolonized mice were more susceptible to an intraperitoneal challenge of E. coli O21:H+ compared with E. coli MG1655-monocolonized mice that received an intraperitoneal challenge of E. coli MG1655 (fig. S4). The canonical proinflammatory cytokines associated with wasting—tumor necrosis factor–α (TNF-α), interleukin-1β (IL-1β), and IL-6—were not down-regulated in E. coli O21:H+ mice, nor was there any difference in intestinal inflammation between DSS-treated vehicle and E. coli O21:H+ mice (fig. S5). Thus, differences in LPS levels, stimulatory capacity, or responsiveness to E. coli O21:H+ were not responsible for the reduced wasting observed in E. coli O21:H+ mice on DSS challenge. Likewise, we found no difference in intestinal tissue damage in wasting-susceptible and wasting-protected animals, as indicated by the lack of difference between the mice in epithelial cell loss, hyperplasia, edema, or fibrosis (fig. S5).

Metabolic changes observed during wasting are distinct from those observed during food deprivation, and nutritional interventions cannot reverse wasting (1214). The DSS-induced anorexic response was equivalent in vehicle-control and E. coli O21:H+ animals (fig. S4). The comprehensive laboratory animal monitoring system (CLAMS) revealed no differences in the rates of oxygen consumption or metabolic rate, carbon dioxide production, respiratory exchange ratio, activity, or heat production of DSS-treated animals with and without E. coli O21:H+ (fig. S6).

In untreated mice, E. coli O21:H+ had no effect on host survival (carried out to 1 year). On DSS treatment, 70% of Jax mice given E. coli O21:H+ survived, whereas all of the control Jax mice that were wasting-susceptible died within 12 days of treatment initiation (Fig. 1G).

We previously identified an E. coli O21:H+ strain in a sepsis model caused by intestinal insult (15); one consequence of sepsis is muscle wasting. Mice in this model and in a model of sepsis induced by intravenous injection with this microbe were protected from wasting (15). We therefore asked whether E. coli O21:H+ ST101 identified in the current study prevents wasting induced by other microbes.

The oral pathogen Salmonella Typhimurium induces wasting of muscle and adipose tissue in Jax mice (Fig. 2, A and B, and fig. S7). E. coli O21:H+ Jax mice did not show weight loss, skeletal muscle, or adipose tissue–wasting when infected orally with S. Typhimurium compared with vehicle-control infected Jax mice (Fig. 2, C and D, and fig. S7).

Fig. 2 Effects of E. coli O21:H+ colonization on muscle wasting induced by infections.

(A) Percent original weight of Jax mice orally infected with 8.5 × 106 S. Typhimurium. Representative experiment (n = 3 independent experiments) using five animals per group, mean ± SD. Unpaired t test. (B) Leg muscle mass from S. Typhimurium–infected mice in (A) at day 5 (ST) or B. thailandensis–infected mice in (E) at day 2 (Bt). Representative experiment (n = 3 independent experiments) using five animals per group, shown as mean ± SD. Unpaired t test. (C) Percent original weight of Jax mice gavaged with 5 × 108 CFU E. coli O21:H+ or vehicle alone and orally infected with 2 × 107 S. Typhimurium. Representative experiment (n = 7 independent experiments) using four to five animals per group, mean ± S.D. Unpaired t test. Untreated control mice ± E. coli O21:H+ are shown in fig. S3. (D) Leg muscle mass from infected E. coli O21:H+ or vehicle-control mice from (C) at day 3 (ST) and from (F) at day 2 (Bt). Representative experiment (n = 3 to 7 independent experiments) using four to five animals per group, mean ± SD. Unpaired t test. Control mice ± E. coli O21:H+ are shown in fig. S3. (E and F) Jax mice were (E) left untreated or (F) gavaged with 5 × 108 CFU E. coli O21:H+ or vehicle alone (supplementary materials, materials and methods) and intranasally infected with 2500 CFU B. thailandensis. Representative experiment (n = 3 to 7 independent experiments) using four to five animals per group, shown as mean ± SD. Unpaired t test. (G) S. Typhimurium (2 × 107 CFU inoculum) or B. thailandensis (2500 CFU inoculum) burdens in target tissues from infected mice ± E. coli O21:H+ at 3 days (ST) or 2 days (Bt). B. thailandensis was not detected in the liver and spleen at this dose. Representative experiment (n = 3 to 7 independent experiments) using three to seven animals per group, mean ± SD. Unpaired t test. Dotted line indicates limit of detection. (H) Percent survival of Jax mice treated with 5 × 108 CFU E. coli O21:H+ or vehicle-control orally infected with 2 × 107 S. Typhimurium. Results are from two independent experiments with 10 to 12 animals per group per experiment. Log rank analysis. ****P < 0.0001 ***P < 0.0005, **P < 0.005, *P < 0.05. Postmortem analyses at day when one or more mice lost >15% body weight. EDL, extensor digitorum longus; TA, tibialis anterior.

We tested whether E. coli O21:H+ could antagonize wasting in the Burkholderia thailandensis pneumonia model in which there is no compromise of the gut barrier. Consistent with critically ill patients with pulmonary dysfunction (16), intranasally infected Jax mice exhibited rapid wasting (~15% weight loss by 2 days after infection) characterized by a depletion of both muscle and fat (Fig. 2, B and E, and fig. S7). Histological analysis showed that the intestinal barrier remained intact during infection in contrast to DSS-treated animals (figs. S5 and S7). To assess gut barrier function, B. thailandensis–infected mice were gavaged with fluorescein isothiocyanate (FITC)–dextran, and serum FITC remained negative, confirming that in contrast to FITC-positive serum obtained from DSS-treated mice, the gut barrier was not breached (fig. S7). B. thailandensis–infected/E. coli O21:H+–colonized Jax mice showed a ~2% decrease in body weight with significantly less muscle and fat wasting than infected mice given vehicle alone, which lost ~15% weight (Fig. 2, D and F, and fig. S7). Consistent with our findings with DSS, E. coli O21:H+ mitigation of S. Typhimurium– and B. thailandensis–induced wasting was independent of changes in the infection-induced anorexic response, caloric absorption by the intestine, or alterations in oxygen consumption, carbon dioxide production, activity levels, or heat production (figs. S7 and S8). E. coli O21:H+ colonization did not result in down-regulation of systemic TNF-α, IL-1β, or IL-6 or any reduction in tissue damage (S. Typhimurium, liver and spleen; B. thailandensis, lung, liver, and spleen) (figs. S9 to S11). Consistent with previous reports (17), intestinal pathology in S. Typhimurium–infected animals was minimal and indistinguishable in animals with and without E. coli O21:H+ (fig. S10). The pathogen burdens in E. coli O21:H+ animals were not reduced compared with infected animals treated with vehicle alone (B. thailandensis did not disseminate to the spleen and liver at our doses) (Fig. 2G). Thus, the presence of E. coli O21:H+ in the intestinal microbiota appears to reduce/inhibit lean body–wasting induced by pathogens or intestinal injury. It is this protective effect, rather than pathogen killing, that significantly promoted survival of S. Typhiumurium–infected animals given E. coli O21:H+ (Fig. 2H and fig. S7).

Transcriptional induction of the E3 ubiquitin ligases Atrogin-1 and Murf1 is crucial for muscle atrophy (1820) and is induced upon challenge with B. thailandensis, S. Typhimurium, or DSS (Fig. 3A and fig. S12). Induction of Atrogin-1 and Murf1 did not occur in the muscle of E. coli O21:H+ Jax mice challenged with pathogen or DSS and was at equivalent levels found in untreated mice with and without E. coli O21:H+ (Fig. 3A and fig. S12).

Fig. 3 Gut colonization by E. coli O21:H+ associated with down-regulation of muscle atrophy programs and maintenance of IGF-1 signaling during infection.

(A and B) Jax mice gavaged with 5 × 108 CFU E. coli O21:H+ or vehicle alone and intranasally infected with 1000 CFU B. thailandensis. (A) Murf1 and Atrogin-1 expression in leg muscle at 2 days after infection normalized to ribosomal protein S17 (RPS17) expression. Representative experiment (n = 3 independent experiments) using four to nine animals per group, mean ± SD [analysis of variance (ANOVA)]. (B) Serum IGF-1 at 2 days. Representative experiment (n = 3 independent experiments) using seven to nine animals per group, shown as mean ± SD (ANOVA). Additional controls are provided in fig. S14. (C and D) Jax mice were given 5 × 108 CFU E. coli O21:H+ or vehicle alone and intranasally infected with 2500 CFU B. thailandensis. (C) IGF-1 levels in WAT and (D) E. coli O21:H+ levels in the WAT at day 2. Representative experiment (n = 4 independent experiments) using six to ten animals per group, mean ± SD (ANOVA). Red line indicates the limit of detection. (E to H) Mice given 5 × 108 CFU E. coli O21:H+ or vehicle alone were injected intraperitoneally with antibody to IGF-1 or IgG isotype control antibody and intranasally infected with 1000 CFU B. thailandensis. (E) Percent original weight. Results are from two independent experiments using six to nine mice per group per experiment (12 to 17 mice total), mean ± SD (ANOVA). Additional controls are provided in fig. S14. (F) Leg muscle mass and (G) adipose tissue mass at 2 days. EDL, extensor digitorum longus; TA, tibialis anterior; GWAT, gonadal white adipose tissue; RPWAT, retroperitoneal WAT; IWAT, inguinal WAT; BAT, brown adipose tissue. Results are from the two independent experiments described in (E) using five to nine mice weight-matched at day 0 (~18 g) for each group for each experiment (10 to 14 mice total), mean ± SD (ANOVA). (H) Murf1 and Atrogin-1 expression in leg muscle at 2 days and normalized to RPS17. Representative experiment (n = 2 independent experiments) using five to nine animals per group, mean ± SD (ANOVA). ****P < 0.0001, ***P = 0.0006, **P < 0.005, *P < 0.05. Postmortem analyses were performed at day when one or more mice lost >15% body weight.

RNA sequencing (RNA-seq) analysis of leg muscles from S. Typhimurium–infected mice revealed that components of the insulin/insulin-like growth factor-1 (IGF-1) signaling pathway, muscle physiology, and metabolism were up-regulated in E. coli O21:H+ mice compared with infected vehicle-control mice (fig. S13). The IGF-1/phosphatidylinositol 3-kinase (PI3K)/AKT pathway is a master regulator of muscle size that activates factors for protein synthesis and regeneration as well as the down-regulation of Atrogin-1 and Murf1 expression in atrophying muscle (2123). Untreated Jax mice with or without E. coli O21:H+ had comparable levels of serum IGF-1 as measured by means of enzyme-linked immunosorbent assay (fig. S13) (11) and showed no muscle hypertrophy (fig. S3); likewise when infected with S. Typhimurium or B. thailandensis or given DSS, serum levels of IGF-1 and downstream signaling components in muscle [IGF-1 receptor (IGFR) and AKT] in E. coli O21:H+ mice showed sustained activation levels comparable with those of untreated animals, whereas activation levels in challenged animals given vehicle alone decreased significantly (Fig. 3B and figs. S13 and S14).

IGF-1 is produced primarily by the liver in response to growth hormone (GH) (24). We found that serum levels of GH, liver Igf-1 expression, and IGF-1 protein levels in the liver, muscle, and intestine were unchanged in challenged E. coli O21:H+ animals compared with challenged animals treated with vehicle alone or unchallenged animals with and without E. coli O21:H+ (fig. S13). Instead, during challenge with pathogens or DSS, white adipose tissue (WAT) from challenged E. coli O21:H+ mice showed higher levels of IGF-1 as compared with WAT from challenged vehicle-control animals and unchallenged animals with and without E. coli O21:H+ (Fig. 3C and fig. S15). This was associated with E. coli O21:H+ colonization of the WAT, but not the muscle or WAT-associated lymph nodes (Fig. 3D and fig. S15); neither B. thailandensis nor S. Typhimurium infected the WAT. Thus, sustained systemic levels of IGF-1 by E. coli O21:H+ are associated with increased WAT IGF-1 levels and colonization of the WAT by this commensal. A role for WAT-derived IGF—eliciting systemic effects—is supported by previous studies (25).

Systemic administration of an IGF-1–neutralizing antibody to E. coli O21:H+ mice in the absence of pathogen infection had no effect on mouse weight (fig. S15). When given the neutralizing antibody intraperitoneally, E. coli O21:H+ mice infected with B. thailandensis were no longer protected from wasting (Fig. 3, E and F, and fig. S15). These animals showed significant weight loss, muscle depletion (but not fat loss), a decrease in muscle IGF-1 signaling, and an increase in muscle expression of Atrogin-1 and Murf1 compared with infected E. coli O21:H+ Jax mice given an immunoglobulin G (IgG) isotype control antibody (Fig. 3, E to H, and figs. S15 and S16). There was no significant difference in the ceca levels of E. coli O21:H+ or of B. thailandensis lung infection levels in E. coli O21:H+ mice that were treated with antibody to IGF-1 or the IgG isotype control (fig. S15).

The inflammasome—a cytoplasmic multiprotein complex required for activation of the caspase-1 (CASP1) protease (26)—has recently emerged as a critical regulator of host-microbiota interactions and metabolism (26, 27). We therefore examined Casp1−/− mice, which are also deficient for CASP11 (28). In contrast to wild-type mice, oral administration of E. coli O21:H+ to B. thailandensis–infected Casp1−/−11−/− mice did not protect against muscle wasting, the up-regulation of Murf1 and Atrogin-1 in muscle, or an infection-induced drop in systemic IGF-1 (Fig. 4, A to C, and fig. S17). Similarly, IGF-1 levels in the WAT were less in B. thailandensis–infected Casp1−/−11−/− E. coli O21:H+ mice than in infected wild-type E. coli O21:H+ mice (fig. S17). Lung infection levels of B. thailandensis and intestinal levels of E. coli O21:H+ were not significantly different between infected Casp1−/−11−/− and wild-type mice with or without E. coli O21:H+ (fig. S17).

Fig. 4 The Nlrc4 inflammasome is required for sustaining IGF-1 levels and mediates the E. coli O21:H+–associated protection against muscle wasting promoted by pathogenic infection.

(A to C) Wild-type and Casp1−/−11−/− mice were administered 5 × 108 CFU E. coli O21:H+ or vehicle alone and then intranasally infected with 2500 CFU B. thailandensis as described in the supplementary materials, materials and methods. (A) Percent original weight at day 2. Additional controls are provided in fig. S14. (B) Expression of Murf1 and Atrogin-1 was determined in leg muscle at 2 days after infection normalized to expression levels of RPS17. (C) Serum IGF-1 quantified at day 2. Representative experiment (n = 3 independent experiments) using 10 animals per wild-type group and five to seven mice per Casp1−/−11−/− group, mean ± S.D. (ANOVA). (D) Wild-type and Nlrc4−/− mice were given 5 × 108 CFU E. coli O21:H+ or vehicle alone and infected intranasally with 2500 CFU B. thailandensis. Serum levels of IGF-1 at 2 days after infection were measured. Representative experiment (n = 2 independent experiments) using 10 animals per wild-type group and five mice per Nlrc4−/− group, shown as mean ± SD (ANOVA). (E) Jax mice were given 5 × 108 CFU E. coli O21:H+ or vehicle alone and then intranasally infected with 2500 CFU B. thailandensis as described in the supplementary materials, materials and methods, and serum IL-18 levels were measured at 2 days after infection. Representative experiment (n = 2 independent experiments) using 10 animals per group, mean ± SD (ANOVA). (F) Wild-type and Il1β−/−Il18−/− mice were administered 5 × 108 CFU E. coli O21:H+ or vehicle alone and then intranasally infected with 2500 CFU B. thailandensis. Serum levels of IGF-1 at 2 days after infection were measured. Representative experiment (n = 3 independent experiments) using 10 animals per wild-type group and eight mice per Il1β−/−Il18−/− group, mean ± SD (ANOVA). ***P < 0.0003, **P < 0.006, *P < 0.05. Experiment was terminated at 2 days for postmortem analyses when one or more mice lost >15% body weight in accordance with our Salk animal protocol. Additional controls for (C), (D), and (F) are provided in fig. S14.

Several distinct inflammasomes have been described, each of which contains a protein of the nucleotide-binding domain–containing (NLR) protein superfamily. The NLRC4 inflammasome is activated by the inner rod protein of type three secretion systems (TTSSs) and by flagellin of Gram-negative bacteria (26), both of which are encoded by E. coli O21:H+ (table S2). Similarly to Casp1−/−11−/− mice, the presence of E. coli O21:H+ in B. thailandensis–infected Nlrc4−/− mice did not antagonize wasting or prevent an infection-induced drop in systemic IGF-1 levels (Fig. 4D and fig. S18). Colonization levels of E. coli O21:H+ in the WAT deposits of B. thailandensis–infected Casp1−/−11−/− were equivalent to those in the WAT of infected wild-type mice (fig. S17). Therefore, the inflammasome is not necessary for translocation of E. coli O21:H+ from the intestine to the WAT.

Activation of any inflammasome leads to maturation of the proinflammatory cytokines IL-1β and IL-18. Systemic levels of IL-18 in B. thailandensis–infected E. coli O21:H+ wild-type mice were increased compared with infected vehicle-control mice, with no differences in IL-1β levels (Fig. 4E and fig. S5). The protective effect of E. coli O21:H+ against B. thailandensis–induced muscle- and fat-wasting occurred in Il1β−/−mice but not in Il1β−/−Il18−/− mice (fig. S18). B. thailandensis–infection levels in the lung and E. coli O21:H+ levels in the intestine and WAT were similar between infected wild-type and Il1β−/−Il18−/− mice given E. coli O21:H+ (fig. S18). When Il1β−/−Il18−/− mice were challenged with B. thailandensis, IGF-1 loss was not prevented by E. coli O21:H+ (Fig. 4F). We conclude that the ability of E. coli O21:H+ to maintain levels of IGF-1 during challenge is mediated by the Nlrc4 inflammasome via IL-18 (with a possible synergistic role for IL-1β) and thereby prevents muscle-wasting and promotes disease tolerance.

The intestinal microbiota performs essential functions for its host to achieve metabolic homeostasis. Our results strongly suggest that a specific constituent of the mouse intestinal microbiota, E. coli O21:H+, antagonizes wasting triggered by infections and intestinal injury. During homeostasis, E. coli O21:H+ remains in the intestine, has no effect on weight, body composition, or IGF-1 levels. During challenge, E. coli O21:H+ sustains systemic IGF-1 and signaling of IGF-1 in muscle to prevent wasting and promote disease tolerance (2931) independent of metabolic changes, food consumption, caloric uptake, inflammation, or tissue damage. This protection is associated with challenge-induced translocation of E. coli O21:H+ to WAT and activation of the NLRC4 inflammasome that correlates with increased production of IGF-1 by WAT. The ability to translocate to and colonize the WAT distinguishes E. coli O21:H+ from other microbes (including B. thailandensis and S. Typhimurium) that can activate inflammasomes where its presence antagonizes wasting in an inflammasome-dependent manner, possibly induced by its flagellin or the inner rod protein of its TTSS, eprJ. (fig. S19).

The microbiota is crucial for the development and maintenance of a robust immune system and resistance defenses (32, 33). Discovery of a molecular mechanism by which the microbiota can promote tolerance to infection provides a perspective into the evolutionary forces that have driven the coevolution of host-microbiota interactions.

Supplementary Materials

www.sciencemag.org/content/350/6260/558/suppl/DC1

Materials and Methods

Figs. S1 to S19

Tables S1 and S2

References (3445)

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. Acknowledgments: We thank R. Vance for mice and reagents; G. Barton and M. Koch for TLR4-ELAM NFkB reagents; M. Montminy and E. Wiater for use of the EchoMRI and MiSeq sequencer and technical assistance; S. Mazmanian, S. McBride, J. Sonnenburg, and S. Higginbottom for germ-free mice and advice; D. Monack and F. Re for Salmonella and Burkholderia strains; W. Fan and S. Bapat for reagents and help with muscle and fat anatomy and physiology; I. Verma for use of the BioRad Bio-Plex MAGPIX multiplex reader; D. Cook for technical assistance with Luminex assays; M. Ku and S. Heinz for advice and technical assistance with RNA-seq and metagenomics analyses; R. Shaw and P. Hollstein for Western blot reagents and advice; G. Van Gerpen, K. Kujawa, and M. Bouchard for administrative assistance with establishing our gnotobiotic facility; and D. Schneider and D. Green for helpful discussions and thoughtful comments on the manuscript. A special thanks to R. Lamberton for continuous support. This work was supported by NIH grant R01AI114929 (J.S.A), the NOMIS Foundation, the Searle Scholar Foundation (J.S.A), the Ray Thomas Edward Foundation (J.S.A.), NIH grants DK0577978 (R.M.E.) and CA014195, the Leona M. and Harry B. Helmsley Charitable Trust (2012-PG-MED002), the Samuel Waxman Cancer Research Foundation, and Ipsen/Biomeasure. R.M.E. is an investigator of the Howard Hughes Medical Institute at the Salk Institute and March of Dimes Chair in Molecular and Developmental Biology. Data described in this paper can be found in the supplementary materials. The authors declare no conflicts of interest.
View Abstract

Navigate This Article