Research ArticleGUT MICROBIOTA

Transplantation of fecal microbiota from patients with irritable bowel syndrome alters gut function and behavior in recipient mice

See allHide authors and affiliations

Science Translational Medicine  01 Mar 2017:
Vol. 9, Issue 379, eaaf6397
DOI: 10.1126/scitranslmed.aaf6397

Connecting the gut-brain axis

Irritable bowel syndrome (IBS), the most common gastrointestinal disorder worldwide, is characterized by abdominal pain and altered gut function and often is accompanied by anxiety. An association between intestinal dysbiosis and IBS has been reported, but the functional relevance remains unknown. De Palma and colleagues colonized germ-free mice with fecal microbiota from healthy controls or IBS patients with diarrhea (IBS-D) who did or did not have anxiety. They demonstrated that transplantation of fecal microbiota from patients with IBS-D and anxiety resulted in altered gut function and behavior in mouse recipients, including faster gastrointestinal transit, low-grade inflammation, and anxiety-like behavior.

Abstract

Irritable bowel syndrome (IBS) is a common disorder characterized by altered gut function and often is accompanied by comorbid anxiety. Although changes in the gut microbiota have been documented, their relevance to the clinical expression of IBS is unknown. To evaluate a functional role for commensal gut bacteria in IBS, we colonized germ-free mice with the fecal microbiota from healthy control individuals or IBS patients with diarrhea (IBS-D), with or without anxiety, and monitored gut function and behavior in the transplanted mice. Microbiota profiles in recipient mice clustered according to the microbiota profiles of the human donors. Mice receiving the IBS-D fecal microbiota showed a taxonomically similar microbial composition to that of mice receiving the healthy control fecal microbiota. However, IBS-D mice showed different serum metabolomic profiles. Mice receiving the IBS-D fecal microbiota, but not the healthy control fecal microbiota, exhibited faster gastrointestinal transit, intestinal barrier dysfunction, innate immune activation, and anxiety-like behavior. These results indicate the potential of the gut microbiota to contribute to both intestinal and behavioral manifestations of IBS-D and suggest the potential value of microbiota-directed therapies in IBS patients.

INTRODUCTION

The intestinal microbiota, composed of a wide diversity of microbial species (1), plays a crucial role in shaping innate and adaptive immune responses of the host (2). The microbiota also influences host physiology by modulating gut motility and intestinal barrier homeostasis (3), as well as host energy metabolism (4). Disruption of the normal intestinal microbiota can lead to gut inflammation and changes in visceral sensitivity (5). The influence of the intestinal microbiota extends well beyond the gut because it has been associated with disorders of host metabolism (6), the liver (7), and somatic sensory nerves (8). Recent demonstrations of the ability of the intestinal microbiota in mice to influence brain development (9), neurochemistry (10), and cognitive (11) and emotive (12) functions have prompted consideration of a microbiota gut-brain axis (13). This axis has been implicated in the pathogenesis of a spectrum of disorders including intestinal (14), neurological (15), and psychiatric (16) diseases. However, in many instances, the putative role of the microbiota is based on associations between disease phenotype and compositional changes in the intestinal microbiota, prompting recent skepticism of the field (17).

Irritable bowel syndrome (IBS) is a common intestinal disorder and is characterized by chronic abdominal pain accompanied by altered bowel habits in the absence of an underlying structural abnormality (18). Symptom cluster analysis has generated a classification of IBS, of which IBS with diarrhea (IBS-D) is a recognized subgroup (18). IBS is frequently accompanied by psychiatric comorbidities such as anxiety and has been considered a disorder of gut-brain communication (19, 20). IBS is likely to have a number of underlying etiologies that include stress and behavioral factors (18), enteric infection, as well as changes in the intestinal microbiota (21). An increasing number of studies have demonstrated changes in the composition, temporal stability, and metabolic activity of the gut microbiota in at least a subset of IBS patients compared to healthy control individuals (22, 23). Although no microbial signature has yet been identified for any of the subtypes of IBS, a reduced abundance of Bacteroidetes, Actinobacteria, and Faecalibacterium spp., with a corresponding increase in Firmicutes, Lachnospiraceae, Enterobacteriaceae, or Gammaproteobacteria, is a recurring finding, especially in IBS patients with predominant diarrhea (2426). Although factors known to trigger the onset or relapse of IBS, such as stress, infection, antibiotic usage, or dietary changes, have also been shown to perturb the resident microbial composition of the gut (2729), the case for implicating the gut microbiota in IBS is based almost exclusively on associations of IBS with altered microbiota profiles. As an initial step in exploring whether the intestinal microbiota might contribute to the expression of IBS, we colonized germ-free mice with fecal microbiota from healthy control individuals or patients with IBS-D, with or without comorbid anxiety. We characterized the microbiota and metabolic profiles of colonized mice and examined the effects of colonization on intestinal motility, gut permeability, immune activation, and behavior. We elected to focus on the microbiota from patients with IBS-D not only because of reported changes in the microbiota but also because this subgroup of patients exhibits increased intestinal permeability (30) that may predispose them to immune activation, a putative mechanism of gut dysfunction in at least a subset of IBS patients (31, 32).

RESULTS

Selection of IBS-D patient donors with and without anxiety for fecal microbiota transplantation to germ-free recipient mice

Stool samples were obtained from five healthy control individuals (two males and three females; mean age, 42 years) and eight patients with IBS-D (three males and five females; mean age, 40 years). The control individuals had no history of gastrointestinal or psychiatric diseases. The patients were diagnosed with IBS-D according to Rome III criteria and Bristol stool form scale (≥6), with more than three bowel movements per day, and symptoms for at least 2 years. Postinfectious IBS was ruled out on the basis of history in all patients. None had elevated serum C-reactive protein (>5 mg/liter) or any common bacterial/viral pathogen in the stool, and bile acid malabsorption was not assessed. Four IBS patients (P1, P4, P5, and P6) displayed moderate anxiety as assessed by the Hospital Anxiety and Depression Scale (HADS) questionnaire (HAD-A score, ≥11); four IBS patients and all controls had normal HAD-A scores (<7). None of the patients or controls used psychotropic medications, antibiotics, or probiotics for at least 3 months before fecal sample collection. Stool samples were transported anaerobically to the laboratory and frozen at −80°C. Diluted stool was gavaged once into 8- to 10-week-old germ-free National Institutes of Health (NIH) Swiss mice (200 μl per mouse, at least 10 mice per donor) of either gender. After 3 weeks, mouse behavior and gastrointestinal transit were assessed. Mice were sacrificed thereafter, and tissue samples were collected. No gender effect was observed for any of the independent variables studied.

Bacterial profiles in recipient mice cluster according to the human donor profiles

A total of 9,112,614 16S ribosomal RNA (rRNA) gene sequences were assembled from 150 samples distributed over four separate sequencing runs, corresponding to fecal samples from 13 human subjects and cecal samples isolated from 137 gnotobiotic mice. Sequence counts ranged from 13,225 to 278,833 per sample, with an average of 39,420 sequences spanning the V3-V4 region of the 16S rRNA gene. Taxonomic profiles were consistent among samples from the same donor. Operational taxonomic units (OTUs) at 97% identity were distributed across 12 phyla, with Firmicutes (mean, 66%) and Bacteroidetes (mean, 25%) being predominant (Fig. 1A).

Fig. 1. Composition of the fecal microbiota of IBS-D and healthy human and mouse cohorts.

The composition profiles for the fecal microbiota of IBS-D patients and healthy controls before and after transplant to recipient germ-free mice are shown. (A) Distribution of taxonomic assignments for 97% identity OTUs (excluding singletons) in fecal samples from healthy controls (H1 to H5) and IBS-D patients (P1 to P8) and corresponding taxonomic assignments in cecal samples from recipient mice. Taxonomic assignments were made at a confidence threshold of ≥0.75 using the ribosomal database project (RDP) classifier trained against Greengenes (85, 86). Color corresponds to phylum; splits within a color correspond to genera within that phylum. (B) Nonmetric multidimensional scaling (NMDS) analysis of Bray-Curtis dissimilarity for human fecal microbiota from IBS-D patients (P1 to P8) and healthy controls (H1 to H5) and corresponding mouse recipient cecal samples.

Overall, there was no strong taxonomic trend distinguishing the microbiota of IBS patients from the microbiota of healthy controls, in either human donors or the transplanted mouse recipients [Fig. 1B; multiresponse permutation procedure (MRPP) A = 0.031, δ(obs) = 0.80, and δ(expected) = 0.83]. The microbiota of individual human donors and the mice that received them differed in ordinations of both OTU-based (Bray-Curtis) and phylogenetic (unweighted UniFrac; fig. S1A) metrics. In general, mouse cecal samples were consistent within groups, and nine mouse subpopulations were tightly grouped (five IBS and four controls); subpopulations from patients P1, P4, P7, and healthy volunteer H5 had a more peripheral grouping (Fig. 1B). Although the microbiota profiles were similar between mice colonized with healthy control H4 fecal microbiota and the microbiota of the H4 human donor, the microbiota profiles for the other mouse recipients were shifted with respect to their human donors.

Despite the lack of distinct taxonomic patterns discriminating IBS patients from healthy controls, we identified indicator species for disease state (IBS patients versus healthy controls; Table 1) and organism (IBS mouse and human versus healthy mouse and human; Table 1). Among these, the strongest IBS-associated indicator species were associated with the genera of the Lachnospiraceae and Bacteroidaceae families in both humans and mice, whereas healthy humans and mice were associated primarily with Desulfovibrionaceae and Rikenellaceae families.

Table 1. Composition of gut microbiota from IBS-D and healthy human and mouse cohorts.

Indicator OTUs are ranked by indicator value, and only the top 10 are reported for cases with >10. A higher indicator value for a certain species signifies a higher representation of a given group (IBS, healthy, etc.).

View this table:

An additional analysis explored the effect of sampling site (cecal and fecal) on profile similarity in five sets of mouse and human samples (H1, H5, P1, P7, and P8), generating 4,707,811 sequences over 77 samples (19,472 to 182,919 sequences per sample). Taxonomic profiles differed by donor [MRPP A = 0.3143, δ(obs) = 0.5206, and δ(expected) = 0.7592], but not sampling site [MRPP A = 0.001535, δ(obs) = 0.7581, and δ(expected) = 0.7592]. Thus, the microbiota profiles obtained from mouse feces were similar to corresponding mouse cecal profiles (fig. S1A), and they differed from the microbiota of conventional NIH Swiss mice (fig. S1B).

The fecal microbiota of IBS-D patients accelerates gastrointestinal transit in recipient mice

Diarrhea was the dominant gastrointestinal symptom in the IBS-D patients included in this study. Gastrointestinal transit, as an overall measure of gastrointestinal motility, was therefore assessed in colonized mice using videofluoroscopy and metallic beads (33), which were gavaged together with barium 3 hours before imaging (Fig. 2A). Overall, mice receiving the IBS-D fecal microbiota displayed a higher transit score, indicative of a faster gastrointestinal transit, compared to mice receiving healthy human microbiota (Fig. 2B). There was more intragroup variation in gastrointestinal transit among mice with IBS-D microbiota compared to mice with microbiota from healthy controls.

Fig. 2. Altered gastrointestinal transit and intestinal barrier function in mice colonized with IBS-D patient versus healthy control fecal microbiota.

(A) Representative videofluoroscopy images of mice colonized with fecal microbiota from a healthy control and an IBS-D patient 3 hours after five steel beads and diluted barium were gavaged into the mice. In the mice colonized with the fecal microbiota from a healthy control individual, there were four beads (thin arrows) in the cecum (outlined) and one bead in the distal small bowel (thick arrow). In the mice colonized with fecal microbiota from an IBS-D patient, there was only one remaining bead (arrow) in the cecum (outlined in yellow); fecal pellets in the colon are outlined in white. (B) Left: Gastrointestinal transit in mice colonized with fecal microbiota from IBS-D patients (n = 88) or healthy controls (n = 53). Right: Gastrointestinal transit in individual groups of mice colonized with fecal microbiota from healthy controls (H1 to H5) or IBS-D patients (P1 to P8). *P < 0.01 versus control (pooled data), Mann-Whitney U test. (C) Paracellular permeability in jejunal and colonic tissues in mice colonized with fecal microbiota from IBS-D patients (n = 10) or healthy controls (n = 10). Mucosal-to-serosal transport of macromolecules was assessed by adding a radioactive probe 51Cr-EDTA on the luminal side of the Ussing chamber and calculating the percentage detected on the serosal side. Colonic paracellular permeability in IBS-D microbiota–colonized mice was higher (P = 0.03, Mann-Whitney U test) compared to healthy control–colonized mice. (D) Ion transport in mice colonized with fecal microbiota from IBS-D patients (n = 10) or healthy controls (n = 10). Net active ion transport across the epithelium was measured via short-circuit current under voltage-clamp conditions. Colonic ion transport was higher in IBS-D microbiota–colonized mice (P = 0.02, Mann-Whitney U test) compared to healthy control–colonized mice.

The fecal microbiota of IBS-D patients alters intestinal barrier function in recipient mice

Altered intestinal barrier function with associated low-grade inflammation has been postulated as an early pathophysiological event in IBS (34). We studied intestinal permeability in vitro in a random subset of mice including two groups with microbiota from healthy controls (H1 and H5) and two groups with microbiota from IBS-D patients (P1 and P6) using the Ussing chamber technique. Paracellular permeability in the colon, assessed by 51Cr-EDTA flux, was increased in mice colonized with IBS-D patient fecal microbiota compared to mice colonized with healthy control microbiota (Fig. 2C). A similar, although nonstatistically significant trend was observed in jejunal tissues (P = 0.09). Ion transport, assessed by a short-circuit current, in colonic (but not jejunal) mouse tissue was increased in mice with IBS-D patient fecal microbiota compared to mice with microbiota from healthy controls (Fig. 2D). This increase may contribute to water movement across the intestinal epithelium and the faster gastrointestinal transit observed in mice with IBS-D patient fecal microbiota.

Fecal microbiota from IBS-D patients with moderate anxiety induces anxiety-like behavior in recipient mice

Psychiatric comorbidities such as anxiety and depression affect 50 to 90% of patients with all subtypes of IBS (35), including IBS-D (20). We segregated IBS-D donors into those with anxiety (HAD-A score, 11 to 14) and those without anxiety (HAD-A score, 5 to 7). Behavior of colonized mice was evaluated using step-down and light preference tests. Mice colonized with fecal microbiota from patients with IBS-D and anxiety (P1 and P4 to P6) stepped down from the elevated platform with longer latency and spent less time in the illuminated compartment compared to mice colonized with a healthy control microbiota (Fig. 3, A and B). Behavior of mice colonized with fecal microbiota from IBS-D patients without comorbid anxiety (P2, P3, P7, and P8) was similar to that of mice colonized with microbiota from healthy controls or mice colonized with specific pathogen–free (SPF) NIH Swiss mouse microbiota (fig. S2). The overall locomotor activity (total distance and velocity) was similar in the three groups (Fig. 3C), which suggested that anxiety-like behavior in IBS-D microbiota–colonized mice was not due to an inflammation-induced sickness behavior.

Fig. 3. Anxiety-like behavior in mouse recipients of fecal microbiota from IBS-D patients or healthy controls.

(A) Anxiety-like behavior was assessed by the step-down test in mice colonized with fecal microbiota from healthy controls (Healthy; n = 53), IBS-D patients with no anxiety (IBS; n = 48), and IBS-D patients with anxiety (IBS + A; n = 40). Each mouse was placed in the center of an elevated platform, and latency to step down from the pedestal was measured. (B) Anxiety-like behavior was assessed by the light preference test in mice colonized with fecal microbiota from healthy controls (n = 53), IBS-D patients with no anxiety (n = 48), and IBS-D patients with anxiety (n = 40). Each mouse was placed in the center of an illuminated box connected to a smaller dark box, and behavior was recorded for 10 min. Total time spent in the illuminated area (light preference) was assessed. (C) Total locomotor activity was assessed during the light preference test by measuring the total distance traveled by each mouse during the 10 min of the test. Differences between groups (Healthy, IBS, and IBS + A) were assessed by Kruskal-Wallis test, with Dunn’s multiple comparisons. *P < 0.05 versus Healthy (pooled data), Mann-Whitney U test.

Fecal microbiota from IBS-D patients induces immune activation in the recipient mouse colon

Low-grade intestinal inflammation with a modest increase in proinflammatory cytokines has been reported in patients with IBS (31). We found no differences in colonic tumor necrosis factor–α (TNF-α), interferon-γ (IFN-γ), interleukin-1β (IL-1β), IL-2, IL-6, IL-8, IL-10, or monocyte chemoattractant protein–1 (MCP-1) between the two groups using a multiplex enzyme-linked immunosorbent assay (fig. S3). However, CD3+ T lymphocyte counts were higher in the colon, but not in the jejunum, of mouse recipients of fecal microbiota from IBS-D patients compared to healthy controls (Fig. 4A). The increase in infiltration by CD3+ T lymphocytes was observed in mice with anxiety-like behavior (P1 and P4 to P6) but not in mice without anxiety-like behavior, which had similar CD3+ T lymphocyte numbers to mouse recipients of healthy control fecal microbiota (Fig. 4A). β-Defensins, proteins with broad-spectrum antimicrobial activity, are known to be up-regulated in response to bacterial challenge (36, 37). An increase in fecal β-defensin 2 has also been reported in patients with IBS (38). We measured the mouse equivalent, β-defensin 3, in mouse colon tissues using quantitative polymerase chain reaction (qPCR) and found a 2.5-fold increase in β-defensin 3 in mouse recipients of IBS-D patient microbiota compared to mouse recipients of a healthy control microbiota (Fig. 4B). Although the amount of β-defensin 3 was similar within the control mouse group, there was a marked variability among mice colonized with IBS-D patient fecal microbiota (Fig. 4B), with a maximum 10-fold increase observed in mice colonized with fecal microbiota from an IBS-D patient (P5). Similar to CD3+ T lymphocyte infiltration, β-defensin 3 was higher in mice with anxiety-like behavior than in those without anxiety-like behavior.

Fig. 4. Increased immune activation in mouse recipients of fecal microbiota from IBS-D patients or healthy controls.

(A) CD3+ T lymphocyte numbers were assessed by immunohistochemistry in the colon of mice colonized with fecal microbiota from healthy controls, IBS-D patients with no anxiety, and IBS-D patients with anxiety. (B) β-Defensin 3 was assessed by qPCR in the colon of mice colonized with fecal microbiota from healthy controls, IBS-D patients with no anxiety, and IBS-D patients with anxiety. (A and B) *P < 0.05 versus Healthy control (pooled data), Mann-Whitney U test. (C) Heat map showing expression of genes associated with innate immunity in the colon of mice colonized with fecal microbiota from healthy controls (H1 to H5) and IBS-D patients (P1 to P6) assessed by NanoString nCounter Gene Expression CodeSets. All genes in the heat map were differentially expressed in IBS-D microbiota–colonized mice compared to control microbiota–colonized mice as assessed by the log2 ratios with nSolver 2.5 analysis software. (D) Differential expression of CXCR3, CXCR4, C3, IL-22ra2, NF-κB, and CCR2 in the colon of mice colonized with fecal microbiota from healthy controls (Healthy), IBS-D patients with no anxiety (IBS), and IBS patients with anxiety (IBS + A). Differences between groups (Healthy, IBS, and IBS + A) were assessed by Kruskal-Wallis test, with Dunn’s multiple comparisons test. FOV, field of view.

We next assessed the expression of a wide array of inflammation-related genes in colonic tissue. In total, 22 genes were up-regulated in mice colonized with IBS-D patient fecal microbiota, with most genes associated with innate immunity and the biggest increase in gene expression in mice with anxiety-like behavior (Fig. 4, C and D). Several genes related to the maintenance and repair of gut barrier function, such as ATF-2, Mylk3, MKNK1, Pk2, RAPGEF2, and CXCR4, were up-regulated in mice colonized with IBS-D patient fecal microbiota (Fig. 4C), in agreement with the functional changes observed in vitro.

We analyzed the expression of CXCR3, a gene associated with host defense against pathogenic organisms (39) and dysregulated immune responses in inflammatory bowel diseases (40), which is elevated in the colonic mucosa of IBS patients (41). We found that mice colonized with IBS-D patient fecal microbiota that also had anxiety-like behavior showed increased expression of CXCR3 compared to mice colonized with a healthy microbiota or with an IBS-D microbiota but without anxiety-like behavior (Fig. 4D).

Ingenuity pathway analysis revealed three biological networks related to cell-to-cell signaling and interaction, cellular movement, and tissue morphology that were up-regulated in the colonic tissue of mouse recipients of IBS-D patient microbiota (table S2 and fig. S4). Among the main canonical signaling pathways were p38 mitogen-activated protein kinase (MAPK) (P = 1.45 × 10−9), pattern recognition receptors (P = 2.38 × 10−9), and glucocorticoid receptors (P = 3.22 × 10−2) (figs. S5 and S6). When stratifying the mice according to their behavior, ingenuity pathway analysis revealed four biological networks in mice with anxiety-like behavior (table S2) related to altered T and B cell signaling (P = 1.19 × 10−15), macrophages (P = 3.33 × 10−13), pattern recognition receptors (P = 2.45 × 10−12), and corticotropin-releasing hormone signaling (P = 2.65 × 10−6). Many of the genes associated with anxiety were involved in lipid metabolism, cell-to-cell signaling and interactions, and production of reactive oxygen species (table S2).

Serum metabolomic profiles are different in mice colonized with IBS-D patient fecal microbiota compared to a healthy control microbiota

Sera from 22 mice colonized with fecal microbiota from three healthy donors (H1 to H3) and 30 mice colonized with fecal microbiota from five IBS-D patient donors (P1 to P5) were analyzed by liquid chromatography–mass spectrometry using two orthogonal columns in tandem. Data preprocessing of serum samples resulted in 3518 putative metabolites under both positive and negative ion modes after the removal of redundant ions and unreliable molecular features. An ordination procedure (two-dimensional principal components analysis scores) demonstrated that quality control samples clustered tightly, indicating good instrumental reproducibility without system drift (fig. S7). The metabolomic profiles of mice colonized with IBS-D fecal microbiota were different from those of mice colonized with a healthy microbiota, as shown by the supervised multiple regression orthogonal partial least squares–discriminant analysis (OPLS-DA) plot, with good model robustness after cross-validation (Q2 = 0.88; Fig. 5A). The metabolomic profiles of mice colonized with healthy control fecal microbiota clustered more tightly together, whereas those of mice colonized with IBS-D patient fecal microbiota showed greater variability. The latter could be differentiated from each other and could be divided into three distinct subgroups: P1-P2, P3, and P4-P5 (Fig. 5, B to D). When analyzing differences between the three IBS-D patient subgroups, we found altered amounts of palmitic, oleic, and stearic acids; glycerophosphocholine; lysophosphatidylcholine; and phosphatidylserine (Fig. 5E). The P4-P5 subgroup had the most distinct metabolomic profile relative to the other two IBS-D patient subgroups and healthy controls.

Fig. 5. Metabolomic profiles in mouse recipients of fecal microbiota from IBS-D patients or healthy controls.

(A) OPLS-DA score plot of serum samples from mice colonized with fecal microbiota from healthy controls (n = 22) or patients with IBS-D (n = 30), highlighting group classifications based on their characteristic metabolomic profiles. PC1, predictive principal component (separation component); PC2, orthogonal principal component (intraclass variability). (B) OPLS-DA score plot of serum samples from three groups of mice colonized with fecal microbiota from healthy controls (H1 to H3) and five groups of mice with fecal microbiota from IBS-D patients (P1 to P5), showing heterogeneity in the metabolomic profiles. (C) S-plot from OPLS-DA scores of serum metabolomic profiles of mice colonized with fecal microbiota from healthy controls or IBS-D patients. Red dots indicate identified metabolites significantly associated with microbiome status among more than 3500 total molecular features detected (x variable, magnitude; y variable, reliability). (D) Supervised analysis of serum metabolomic profiles (OPLS-DA score) from mice colonized with fecal microbiota from IBS-D patients showing three different subgroups (P1-P2, P3, and P4-P5). (E) Differences in ion responses measured for serum metabolites among three subgroups of mice colonized with fecal microbiota from IBS-D patients P1-P2, P3, and P4-P5. Metabolites of palmitic acid (16:0), glycerophosphocholine, oleic acid (18:1), stearic acid (18:0), lysophosphatidylcholine (LPC; 18:0), phosphatidylserine (PS; 39:3), and PS (41:5) were differentially abundant across the three subgroups of mice colonized with fecal microbiota from IBS-D patients [P < 0.01, P < 0.001, P < 0.001, P < 0.0001, P < 0.001, P < 0.01, and P < 0.01, respectively, one-way analysis of variance (ANOVA) followed by post hoc Fisher’s least significant difference test]. Post hoc analysis P values are shown on individual graphs.

We identified seven metabolites that differed between mice colonized with IBS-D patient fecal microbiota and mice colonized with healthy control fecal microbiota based on their S-plot scores and P values (Fig. 5C and table S1). O-Acetyl-l-carnitine (C2) and several lysophosphatidylcholine species were increased, whereas phosphatidylserine metabolites were decreased in mice colonized with IBS-D patient fecal microbiota compared to those colonized with healthy control microbiota.

DISCUSSION

The results of this study demonstrate that the fecal microbiota from IBS-D patients had an impact on gut function, immune activation, and behavior in mouse recipients, and support the notion that resident bacteria may contribute to at least a subset of IBS. We chose IBS-D patients because this subgroup is more closely associated with compositional changes in the microbiota (42), as well as predictable changes in intestinal transit times, increased intestinal permeability (43), and evidence of immune activation (32). Mice colonized with fecal microbiota from IBS-D patients, but not from healthy controls, exhibited changes in each of these characteristics as well as in behavior.

An intriguing finding in this study was the apparent transfer of the behavioral phenotype of our IBS-D patients with anxiety to mice through fecal microbiota transplantation. Although the ability to transfer components of behavioral phenotypes has been demonstrated between mice (10), the transfer from human to mouse suggests that the microbiota may contribute to the comorbid anxiety observed in many patients with IBS-D. It also raises the possibility that microbiota-directed therapies, including pre- or probiotic treatment, may be beneficial in treating not only intestinal symptoms but also components of the behavioral manifestations of IBS. This is supported by recent studies in animals showing that administration of the probiotic Bifidobacterium longum NC3001 normalized anxiety-like behavior and hippocampal expression of brain-derived neurotrophic factor in mice (44) and that Lactobacillus rhamnosus (JB-1) reduced stress-induced corticosterone production and anxiety- and depression-related behaviors in mice (12). In humans, a recent clinical study in healthy women demonstrated that administration of probiotics in fermented milk altered brain responses to emotional attention tasks (45). Furthermore, in a pilot trial, we have demonstrated that treatment with B. longum NCC3001 improves depression scores, as well as overall gastrointestinal symptoms and quality of life in patients with IBS (46).

The pathophysiology of IBS-D includes rapid intestinal transit and increased luminal fluid secretion (47), and these changes were induced in recipient mice after transplantation of fecal microbiota from IBS-D patients but not from healthy controls. Increased intestinal permeability and immune activation, which have been reported in a subset of IBS patients, were evident in mice colonized with fecal microbiota from some, but not all, IBS-D patient donors. Those mice with stronger immune activation (P1 and P4 to P6) exhibited anxiety-like behavior. Anxiety and depression are common in patients with IBS and have been associated with immune activation (32).

Low-grade gut inflammation, commonly documented in IBS patients by an increase in CD3 T lymphocytes, is a putative cause of altered gut physiology (32). This is supported by a previous study in which T cell activation in mice, induced by anti-CD3 antibodies, resulted in rapid intestinal transit and increased epithelial ion transport (48). We found differences in the numbers of CD3+ T cells between groups of mice colonized with IBS-D patient fecal microbiota from different donors, reflecting the heterogeneity observed in IBS patients. Although production of β-defensin 3 was similar among recipient mice colonized with healthy control microbiota, there was a marked heterogeneity in β-defensin 3 production in mice colonized with IBS-D patient fecal microbiota. In some IBS-D recipient mice, β-defensin production was comparable to that of control mice, whereas in other IBS-D mouse groups (P1, P4, and P5) β-defensin production was several-fold higher. We found similar patterns in the expression of multiple innate immunity–related genes, which were up-regulated in IBS-D microbiota–colonized mice compared to controls (Fig. 4, C and D). Ingenuity pathway analysis suggested that these genes were part of several complex signaling networks, which included those of pattern recognition receptors, lymphocytes and macrophages, and pathways regulating production of reactive oxygen species. Furthermore, we identified changes in gene expression involved in glucocorticoid receptor pathway signaling, which were previously linked to changes in peripheral and central nociceptive signaling (49, 50).

Among the up-regulated genes, there were several genes involved in the regulation of inflammatory responses, such as CCR2 (51), LT-a, IL-22ra2 (38), and the nuclear factor κB (NFB) transcription factor (52). C3, a crucial component in the activation of the complement system (39), can trigger degranulation of mast cells (53), which are thought to play an important role in the pathophysiology of IBS (54). Both IL-1α and CXCR4 have been shown to be up-regulated in both mice and patients with colitis (40, 41), with CXCR4 modulating intestinal barrier function (42) together with MKNK1, Ptk2, RAPGEF2, ATF-2, and Mylk3. Both ATF-2 and Mylk3 gene expression has been recently linked to intestinal tight junction regulation (55), supporting our in vitro permeability experiments showing altered barrier function in mice colonized with IBS-D patient fecal microbiota. DDIt3, a transcription factor activated by p38 MAPK and involved in the endoplasmic reticulum stress response (56), was also up-regulated, reinforcing our hypothesis of a disruptive immune response triggered and sustained by IBS-D patient fecal microbiota. Several genes with altered expression were also linked to nervous system function. CCR2 has been reported to be key for the expression of anxiety-like behaviors in murine models of social stress (57), whereas Limk1 has been shown to affect function of select G protein (guanine nucleotide–binding protein)–coupled receptors, such as ORL1 (a nociceptin receptor), which modulates numerous neural activities including emotional behavior (43).

Although the primary objective of our study was to examine whether IBS-D patient fecal microbiota has the ability to influence gut and brain function in recipient mice rather than to perform a comprehensive characterization of the composition of the donor microbiota, our sequencing data agree with published observations on the microbiota composition of IBS-D patients. Although microbiota profiles from IBS-D patients and healthy controls were taxonomically similar and often overlapped, as shown previously (24), in-depth statistical analysis revealed that Lachnospiraceae and Bacteroidaceae families were associated with IBS-D, whereas Desulfovibrionaceae and Rikenellaceae families were associated with healthy controls, in agreement with earlier studies (23, 24, 58, 59). Although Lachnospiraceae family was associated with both gut and brain dysfunction in mice, Gammaproteobacteria families Shewanellaceae and Halomonadaceae, and Verrucomicrobiae (Akkermansia) were linked to anxiety-like behavior in mice (Table 1). Moreover, the genus Blautia (Lachnospiraceae) was associated with both brain and gut dysfunction in humans and mice (Table 1).

Although there were relatively minor differences in the microbial composition between mice colonized with the IBS-D microbiota compared to a healthy control microbiota, their metabolomic profiles differed (Fig. 5B). When comparing mice colonized with IBS-D and healthy control microbiota, we identified several prominent metabolites with neuroactive and immunomodulatory properties that might explain findings in the recipient mice. Lysophosphatidylcholine species, lipid mediators with both pro- and anti-inflammatory properties (60) that can inhibit bacterial endotoxin–induced proinflammatory responses (61), were increased, whereas phosphatidylserine, a phospholipid component of cell membranes involved in apoptosis and immunomodulation (62), was decreased in mice colonized with IBS-D compared to control microbiota. Phosphatidylserine administration has been reported to improve endocrine and psychological responses to mental stress (63) and ameliorate anxiety in patients with Alzheimer’s disease (64) and has been shown to have antidepressive effects in rats (65) and patients (66). Also, acetyl-l-carnitine has multiple neurobiological properties including reduction of oxidative stress, regulation of brain energy and phospholipid metabolism, as well as modulation of neurotrophins and neurotransmitters (67, 68). Finally, we identified several metabolites differentially produced among mice colonized with IBS-D patient microbiota versus control fecal microbiota, such as palmitic acid that has been shown to affect the central control of insulin and leptin signaling in rats (69) and glycerophosphocholine, which is a parasympathomimetic acetylcholine precursor, with potential for the treatment of Alzheimer’s disease and dementia (70, 71).

Overall, our data suggest that fecal microbiota from IBS-D patients may induce gut and brain dysfunction in recipient mice through multiple mechanisms, including immune and metabolic pathways. Whereas rapid intestinal transit was seen in all mice receiving IBS-D patient fecal microbiota, anxiety-like behavior was observed only in mice with increased immune activation. These mechanisms will need to be studied in further detail, focusing on individual microbiotas and specific pathways. Our results are limited to IBS-D patients and cannot be directly extrapolated to other IBS subgroups. Further studies using microbiota from IBS patients with constipation and other symptoms will be required. Moreover, the precise mechanisms through which the microbiota from IBS-D patients may induce gut and brain dysfunction in recipient mice need to be investigated in future studies. The number of IBS-D patients in our study was small, and future studies with larger numbers of patients are required.

Our results support a role for the gut microbiota in the expression of IBS-D, although we cannot claim that this is the primary driving force. Changes in the gut microbiota and its functional impact may be secondary to alterations in gut physiology resulting from, for example, chronic stress (72, 73). We also cannot rule out the possibility that human host factors, transferred together with the IBS-D patient fecal microbiota, might have contributed to the functional changes observed in the gnotobiotic recipient mice.

The approach used in our study can be applied to other conditions in which the microbiota gut-brain axis has been implicated. Our findings may serve as a basis for developing markers to identify those patients who might benefit from emerging microbiota-directed therapies. Finally, this study raises an important health issue related to fecal transplantation, an increasingly popular treatment for multiple disorders, often performed without medical supervision. On the basis of our results, fecal transplant donors should be screened for functional gut disorders and other conditions.

MATERIALS AND METHODS

Study design

To explore whether the intestinal microbiota has the ability to affect gut and brain function, we colonized germ-free NIH Swiss mice (n = 141) with fecal microbiota from healthy controls (n = 5) or patients with IBS-D (n = 8), with or without comorbid anxiety. Three weeks later, we examined the effects of colonization on mouse intestinal motility, gut permeability, immune activation, and behavior and characterized the murine microbiota and metabolomic profiles. The minimum number of mice (n = 10) to be colonized with each stool donor was determined on the basis of previous behavioral and gastrointestinal motility experiments. Mice of both sexes were randomly selected for bacterial colonization; each group was evenly balanced for male/female ratio. The investigators were blinded to group allocation for most of the assessments, except for behavior and gastrointestinal motility tests.

Animals

Germ-free NIH Swiss mice (10 to 12 weeks old) were obtained from the Axenic Gnotobiotic Unit of McMaster University. The mice were gavaged with diluted human fecal samples, which were previously tested for bacterial and viral pathogens using bacterial cultures and the xTAG Gastrointestinal Pathogen Panel (Luminex), respectively. The mice were housed for 3 weeks in sterilized ventilated racks on a 12-hour light/12-hour dark cycle with free access to food and water and were handled in a level 2 hood by a dedicated technician to minimize bacterial colonization. All experiments were approved by the McMaster University Animal Care Committee.

Behavior assessment

Anxiety-like behavior was assessed with two well-validated tests: the light/dark preference test and the step-down test (7476). For the light preference test, each mouse was placed in the center of an illuminated box connected to a dark box, and its behavior was recorded for 10 min. Total time spent in the illuminated area, distance traveled, and average velocity were assessed using an automated system (Med Associates) as previously described (74). The step-down test was performed as described previously (76). Briefly, each mouse was placed in the center of an elevated platform, and latency to step down from the pedestal was measured (maximum, 5 min).

Gastrointestinal transit assessment

Mice had free access to water and food until beginning the experiment. Five small steel beads (0.79 mm diameter; Bal-tec) plus barium (0.1 ml, 40% dilution) were gavaged intragastrically to each mouse. A second intragastric gavage of barium (0.2 ml, 40% dilution) was performed after 170 min. After a further 10 min, the mouse was placed in a custom-built Plexiglas restrainer, and fluoroscopic video recording (duration, 10 to 20 s) was then obtained using digital videofluoroscope (Siemens Polystar TOP) (33). Video images were digitized and analyzed by using public domain NIH Image 1.62 software. A simple scoring system was used to measure a transit score. An individual bead was given a score that was dependent on its location within the gastrointestinal tract (stomach, 0; proximal small bowel, 1; distal small bowel, 2; cecum, 3; colon, 4; expelled, 5). The scores of each individual bead within a single animal were then summed together to give a total transit score.

Measurement of gut barrier function

Barrier function was assessed by the Ussing chamber technique as previously described (33, 77). Briefly, sections of colon from each mouse were mounted in an Ussing chamber with an opening of 0.6 or 0.7 cm2. Baseline conductance (G: millisiemens per square centimeter) was calculated according to Ohm’s law. Mucosal-to-serosal transport of macromolecules was assessed by adding a radioactive probe 51Cr-EDTA in the luminal side of the chamber. 51Cr-EDTA fluxes were calculated by measuring the proportion of radioactive 51Cr-EDTA detected in the serosal side of the chamber after 2 hours compared to the radioactive 51Cr-EDTA placed in the luminal side at the beginning of experiment. Net active transport across the epithelium was measured via a short-circuit current response (Isc; microampere) injected through the tissue under voltage-clamp conditions. Tissue conductance (the passive permeability to ions) was calculated using Ohm’s law. Baseline Isc (microampere per square centimeter) and conductance (millisiemens per square centimeter) were recorded at equilibrium 20 min after mounting jejunum sections.

Histology and immunohistochemistry

Colonic and jejunal tissue samples fixed in 10% buffered formalin were embedded in paraffin. Tissue sections were stained with hematoxylin and eosin and assessed for inflammation by a blinded observer. Additional colonic sections were stained with polyclonal rabbit anti-mouse CD3 (Dako) as previously described (77), and in each section, CD3+ lymphocytes/crypts were counted by a blinded observer.

Assessment of proinflammatory cytokines

Sections of colon were snap-frozen in liquid nitrogen. Tissue was stored at −80°C until homogenization in tris-HCl buffer containing protease inhibitors (Sigma-Aldrich). Cytokine levels in the supernatant were measured using a BD Cytometric Bead Array (mouse inflammation kit; BD Biosciences) according to the manufacturer’s instructions. Results were analyzed using a BD FACSArray Bioanalyzer System (BD Biosciences).

β-Defensin 3, CXCR3 expression, and NanoString gene expression assay in colonic tissues

The mRNA expression of mouse intestinal β-defensin 3 (homolog of human β-defensin 2) and CXCR3 (chemokine receptor CXCR3) was analyzed by reverse transcription PCR in colonic sections. Total RNA extractions were conducted with the RNeasy Mini Kit (Qiagen) according to the manufacturer’s instructions. Deoxyribonuclease (DNase) digestion during purification was carried out using the ribonuclease-free DNase (Qiagen). The RNA was reversely transcribed to complementary DNA with M-MLV Reverse Transcriptase (Invitrogen) according to the manufacturer’s instructions. Additional information on the primers used is available in Supplementary Materials and Methods. NanoString nCounter Gene Expression CodeSet for mouse inflammation genes and a custom NanoString Gene Expression CodeSet for selected genes were run according to the manufacturer’s instructions (NanoString Technologies Inc.). The results obtained were analyzed with the analysis software nSolver 2.5 (NanoString Technologies). The log2 ratios built from the data obtained were then uploaded into Ingenuity Pathway Analysis software (Qiagen) for further analysis. The network score is based on the hypergeometric distribution and is calculated with the right-tailed Fisher’s exact test.

Illumina sequencing

Microbiota profiles were assessed in fecal samples from each human donor (five healthy controls and eight IBS-D patients) and in cecal and fecal samples from recipient mouse populations using V3 and V4 variable regions of the 16S rRNA gene by MiSeq 250-base paired-end multiplex sequencing (Illumina) (78, 79). All sequence analysis was managed using the AXIOME v.1.7 package (80), extending QIIME v.1.7 (81). Illumina reads were binned according to sample index, and contigs were assembled from paired-end reads using PANDAseq (82), UPARSE (83) at 97% identity, and the RDP classifier v.2.2 (84) trained against the Greengenes SSU database (October 2012) (85, 86). OTUs not assigned to bacteria or assigned to either mitochondria or chloroplast were removed from analysis, as were OTUs identified as chimeric by de novo chimera detection in UCHIME v.4.2 (87). All β-diversity analyses were performed on sequence subsets randomly rarefied to the size of the smallest library. Taxonomic assignments were used in NMDS using Bray-Curtis dissimilarity. A MRPP (88) was used to test for significant differences between sampling groups and the degree of within-group sample clustering. When a priori grouping was required, human donor and mouse subpopulations were kept separate. Additionally, NMDS scaling procedure was performed using unweighted UniFrac phylogenetic metric (89), derived from a FastTree (90, 91) phylogeny generated using the default search strategy from a PyNAST (92) sequence alignment. Indicator species analysis (93) was used to identify OTUs associated with groups defining IBS/healthy or human/mouse model populations. Identical sequence analysis protocols were followed for the human subject (fecal) versus mouse (cecal) and the subject (fecal) versus mouse (fecal and cecal) data sets (files S1 and S2). Illumina sequence data, together with the NanoString gene expression data, have been deposited at National Center for Biotechnology Information (NCBI), under the umbrella BioProject PRJNA 263919.

Serum sample preparation

Mouse blood was obtained by cardiac puncture under isoflurane anesthesia. Serum was collected by centrifugation and stored at −80°C. Ice-thawed serum samples were then extracted using an ice-cold mixture of 2:2:1 methanol/ethanol/water (v/v/v) containing 20 μM of the internal standard l-phenylalanine-d8 (Cambridge Isotope Laboratories). The ratio of extraction solvent to serum was 4:1 (v/v). The serum-solvent mixture was vortexed for 2 min before centrifugation at 4°C for 10 min. The pellet was discarded, and the supernatant was analyzed by liquid chromatography–time-of-flight–mass spectrometry. Pooled serum extracts were run after every fifth sample as a quality control measure to assess for system drift. Additional information is available in the Supplementary Materials and Methods.

Statistical analysis

Statistical analysis was performed using SPSS 20.0 software for Windows (SPSS Inc.). The data are presented as median (interquartile range) or mean ± SD. Statistical comparisons were performed using Mann-Whitney and Kruskal-Wallis tests, multiple t test, or ANOVA, as appropriate. Benjamini and Hochberg false discovery rate correction method was used when multiple comparisons were performed. P < 0.05 was considered statistically significant.

SUPPLEMENTARY MATERIALS

www.sciencetranslationalmedicine.org/cgi/content/full/9/379/eaaf6397/DC1

Materials and Methods

Fig. S1. Microbiota profiles in human donors and recipient gnotobiotic mice.

Fig. S2. Behavior of germ-free NIH Swiss mice and mice colonized with microbiota from healthy controls, IBS-D patients, and an SPF NIH Swiss mouse.

Fig. S3. Colon cytokine levels are similar in healthy control and IBS microbiota–colonized mice.

Fig. S4. Biological networks activated in mice with IBS-D microbiota.

Fig. S5. Activation of pattern recognition receptors in mice colonized with IBS-D microbiota.

Fig. S6. Activation of glucocorticoid receptor signaling in mice with IBS-D microbiota.

Fig. S7. Metabolomic profiles in mouse colonized with IBS and healthy microbiota.

Table S1. Seven metabolites identified in serum of gnotobiotic mice based on their S-plot scores and P values.

Table S2. Immune gene-related networks altered in mice colonized with fecal microbiota from IBS-D patients.

File S1. AXIOME script for analysis of Illumina sequence data for healthy and IBS patient fecal samples and corresponding mouse cohort cecal samples.

File S2. AXIOME script for analysis of Illumina sequence data for single healthy and IBS patient fecal samples and corresponding mouse cohort fecal and cecal samples.

References (9496)

REFERENCES AND NOTES

  1. Acknowledgments: We thank B. Berger and E. Rezzonico for critical comments. Funding: This study was supported by grants from Canadian Institutes of Health Research (CIHR) and Nestlé Switzerland. P.B. is a recipient of a Hamilton Health Sciences Early Career Research Award. G.D.P. is a recipient of a CIHR–Canadian Association of Gastroenterology Postdoctoral Fellowship. Author contributions: G.D.P., J.L., J.J., Y.D., G.U., M.P.P., P.M.M., S.S., and V.P. performed the mouse experiments and data analysis. M.D.J.L., M.-G.H., and J.D.N. performed 16S rRNA data analysis. V.T.D. and P.B.-M. performed the metabolomic analysis. M.G.S. provided the Illumina platform and support. E.F.V. provided germ-free mice and support for gnotobiotic experiments and reviewed the manuscript. M.I.P.S. recruited human IBS-D patient donors. P.B. and S.M.C. conceived and supervised the study. G.D.P., S.M.C., and P.B. wrote and reviewed the manuscript. G.E.B. and P.G.M. participated in the study design and reviewed the manuscript. Competing interests: G.E.B. is an employee of Nestlé Research Center, Lausanne, Switzerland. P.G.M. was affiliated with the Nestlé Research Center at the time of manuscript submission and is now at Takeda Pharmaceuticals, Geneva, Switzerland. The other authors declare that they have no competing interests. Data and materials availability: Illumina sequencing data, together with gene expression data, have been deposited at NCBI under the umbrella BioProject PRJNA 263919.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article