The Brain-Gut-Microbiome Axis

Several experimental approaches in animal models have been used to study the influence of gut microbiota on the BGM axis, including manipulation with antibiotics, fecal microbial transplantation,^,  colonization with synthetic or human^,  microbiota, cultured gut organ systems,^,  probiotic administration, and germ-free (GF) animal models. Despite limitations to these approaches, significant progress has been made since Sudo et al first showed that the absence of normal gut microbiota early in life has significant effects on stress responsiveness in the adult and that these changes can be partially reversed by early colonization of the gut with conventional microbiota, even a single species. Subsequent investigations have further characterized the associated neurochemical changes, including altered cortical and hippocampal brain-derived neurotrophic factor levels,^,  reduced hippocampal serotonin (5-HT) receptor 1A expression, increased striatal monoamine turnover, and reduced synaptic plasticity gene expression, showing the microbiome’s diverse and potent influence over central nervous system (CNS) phenotypes. Accordingly, in addition to stress responsiveness,^,  the gut microbiota are implicated in relation to anxiety-like^{,  ,  ,  ,  ,  ,  ,  ,  ,  ,  ,}  and depression-like behavior,^{,  ,  ,  ,  ,  ,  ,}  nociceptive response,^{,  ,  ,  ,  ,}  feeding behavior, taste preference, metabolic consequences,^{,  ,}  and their respective underlying physiologies. These preclinical studies have been reviewed extensively in the literature.^{,  ,  ,  ,}  Despite acknowledged limitations to the GF model,^{,  ,}  phenotype reversal by recolonization with specific-pathogen-free (SPF), human-derived, and synthetic microbiota validates some of the conclusions. Still, the well-characterized role of microbiota in neurogenesis and neurodevelopment moderates the translational relevance of these findings for adult conditions.

An alternative to the GF model approach is the use of broad-spectrum antibiotics to induce transient changes on the composition and diversity of fecal microbiota, although the effects on mucosa-associated microbial communities are incompletely understood. Antimicrobials also may interact directly with host physiology in mechanisms independent of the microbiome, such as their well-documented neurotoxic effects (reviewed by Grill and Maganti). Nonetheless, broad-spectrum antibiotic treatment remains a powerful tool to identify gut microbial influence on the CNS. In mice with SPF microbiota, oral antibiotic administration increased exploratory behavior and hippocampal expression of brain-derived neurotrophic factor, which was associated with changes in the microbial profile. The failure to replicate these antibiotic-induced effects in GF mice suggests the CNS changes are not the result of off-target antibiotic interactions, however, developmental alterations in the GF model make this finding inconclusive. Long-term antibiotic treatment in adult mice reduced hippocampal neurogenesis and lead to deficits in novel object recognition tasks through a mechanism dependent on circulating monocytes. Adoptive transfer of Ly-6c^hi monocytes or voluntary exercise and probiotic treatment rescued these phenotypes.

In contrast to the complete or partial depletion of gut microbiota as an experimental approach to identify and characterize microbial influence on the host CNS, the introduction of known microorganisms (usually as probiotics) to conventional models allows for normal development and risks fewer off-target effects. However, it is critical to acknowledge the possibility that transient exposure induces host responses inaccessible to resident communities. Orally administered probiotics have been shown to reduce basal or induced anxiety-like behavior,^{,  ,  ,  ,  ,}  attenuate induced obsessive-compulsive–like behavior, improve inflammation-associated sickness behavior, and even normalize developmental trajectories of emotion-related behavior after early life stress. Although infrequently used as an intervention directed specifically at the gut microbiota, diet can have profound, rapid, and reproducible effects on the structure of the gut microbiota in human beings and mice.^{,  ,}  Alterations in the gut microbial community through dietary change also have been shown to influence memory and learning,^,  while probiotic administration rescued diet-induced memory deficits in rats. In summary, preclinical studies have identified unequivocally the potent influence of gut microbiota on the CNS, but issues of reproducibility and off-target intervention effects demand continued improvement of experimental approaches.

Experimental approaches to study the role of gut microbes to brain signaling have been restricted mostly to small clinical studies showing the association of gut microbial community structure with brain parameters and subjective outcomes of interventions with probiotics and prebiotics. Although no high-quality, controlled studies in human beings have reported the effects of interventions such as antibiotics or fecal microbial transplants on the brain or behavior, studies of probiotic interventions are increasing rapidly in number and gradually in scale and quality. A double-blind, placebo-controlled, pilot study of the probiotic Bifidobacterium longum NCC3001 in 44 adults with IBS and diarrhea was shown to reduce responses in the amygdala and frontolimbic regions to negative emotional stimuli as measured by functional magnetic resonance imaging. Although depression scores were lower with the intervention, anxiety and IBS symptoms were not affected. In healthy female control subjects, consumption of a fermented milk product with probiotics over 4 weeks was associated with significant changes in the functional connectivity between brain regions during an emotion recognition task, notably without concomitant detectable changes in gut microbial composition. Probiotic consumption also has been reported to reduce self-reported feelings of sadness and aggressive thoughts. A probiotic cocktail used to achieve reduction of anxiety- and depression-related behaviors in mice also was administered to healthy human beings to a similar effect.

The translation of promising findings obtained in rodent studies has been limited. In a clinical trial with Lactobacillus rhamnosus (JB-1), the effects of which were seminally shown on mice by Bravo et al, performed no better than placebo on stress-related measures, hypothalamic pituitary adrenal axis response, inflammation, or cognitive performance in an 8-week trial with healthy males. Moreover, the pilot trial of Bifidobacterium longum NCC3001 described earlier did not recapitulate the effects observed in mice by the same research group and has been criticized for its fragility. This translational disconnect, or inconsistency, highlights the likelihood of host-specific microbiota interactions and underscores the importance of cautious extrapolation of preclinical findings. Furthermore, as shown by several studies, probiotic supplementation in human beings does not appear to change the gut’s microbiota composition but induces its effect on behavior via transient modification of the collective microbiome transcriptional state, as shown in GF mice and confirmed in monozygotic twins. This finding demands measurement of probiotic intervention on gut microbial profiles with technologies integrating metatranscriptomics and metabolomics and fundamental reconsideration of the functional equivalence of transient vs resident microorganisms. Better characterization of microbial community dynamics and metabolism coupled with improved models of their community ecology will help refine the mechanisms responsible for these effects and identify putative targets for therapeutic intervention.

An important pathway by which gut microbes and their metabolites communicate with the CNS involves the cells making up the endocrine system of the gut. There are at least 12 different types of these cells with several subtypes (in particular A, K, and L cells) present as subgroups along the intestine that contain different combinations of molecules. EECs are interspersed between gut epithelial cells throughout the length of the gut and contain more than 20 different types of signaling molecules, which often are colocalized and co-released. Released in response to chemical and or mechanical stimuli, these molecules can enter the systemic circulation and reach centers in the CNS involved in ingestive behavior (including the nucleus tractus solitarius and the hypothalamus) or act locally and activate closely adjacent afferent vagal terminals in the gut or liver to generate brain signals. A series of receptors involved in the regulation of satiety and hunger have been identified on these cells, which are activated by microbial metabolites including bile acids and SCFAs.

Although bile acids are endogenous molecules synthesized from cholesterol in the liver, the size and composition of the host’s pool of these molecules is heavily influenced by dietary intake, especially of fat, and the downstream metabolism by the gut microbiota (reviewed by de Aguiar et al and Wahlstrom et al). Ileal expression of farnesoid X receptor (FXR), a nuclear receptor, is activated by bile acids leading to production of fibroblast growth factor 19 (FGF19), or its similarly functioning ortholog FGF15 in mice, which can enter the systemic circulation and cross the blood-brain barrier. Activation of the arcuate nucleus of the hypothalamus by the FXR/FGF19 action on agouti-related peptide/neuropeptide Y has been implicated in improved central regulation of energy and glucose metabolism^{,  ,}  and suppression of hypothalamic pituitary adrenal axis activity. Some intestinal L cells express surface receptor G protein-coupled bile acid receptor (TGR5), which is activated mostly by secondary bile acids, which are strongly influenced by microbial activity.^,  TGR5 signaling controls glucose homeostasis by mechanisms including increased glucagon-like peptide-1 (GLP-1) release by these L cells. Interestingly, these L cells also express FXR, which can regulate GLP-1 synthesis.

SCFAs have been implicated as major signaling molecules mediating host-microbe communication via EECs and ECCs. These molecules are generated by microbial fermentation of host dietary-resistant starch and nonstarch polysaccharides and serve an important part in the host’s energy harvest while also stimulating colonic blood flow, fluid and electrolyte uptake, and mucosal proliferation. Dietary fiber intake is a major regulator of SCFA concentrations. In instances in which the host diet is low in fermentable fibers, microbes feed on mucus glycans and use alternative, less energetically favorable sources, resulting in reduced fermentative activity and SCFA production. Both preclinical and clinical data have shown that microbial activity, in particular the production of SCFAs, stimulate L cells located at the distal ileum to secrete peptide YY and GLP-1, which induce satiety and behavioral changes.^{,  ,}  Acetate, butyrate, and propionate modulate the expression and secretion of GLP-1 via free fatty acid receptor 2 (FFAR2)/G-protein–coupled receptor 43 (GPR43) and FFAR3/GPR41 on L cells, and GPR41s have been identified on different EECs in the gut. This widespread distribution of these receptors on and in EECs and ECCs involved in the regulation of food intake and digestion is consistent with the important role of the gut microbes in these processes and the expectations of holobiont co-evolution.

One of the best characterized examples of these microbial host interactions is the bidirectional interaction between microbes, ECCs, and the central nervous system (Figure 1). 5-HT is produced by the ECCs of the gastrointestinal (GI) tract, with 95% of the body’s 5-HT stored in ECCs and enteric neurons, and only 5% stored in the CNS. Considering 5-HT’s central role in regulating GI motility and secretion, there is likely immense selective pressure on the gut microorganisms to act on the serotonergic system to modulate their environment effectively (eg, by influencing regional transit times and fluid secretions). An analysis of the plasma metabolite profile of germ-free mice shows a more than 2-fold decrease in 5-HT levels relative to conventionally colonized mice. SCFAs and 2BAs derived from spore-forming bacteria of the gut regulate a significant percentage of ECC 5-HT synthesis and release. The essential amino acid tryptophan (Trp) is a key molecule in the BGM axis because it is the precursor to the neurotransmitter 5-HT and a number of other metabolites that contribute to the neuroendocrine signaling within the BGM. Because the host is unable to produce tryptophan, dietary intake of proteins that contain it serve as the primary regulator of its availability. Gut microbiota contribute to the peripheral availability of Trp, which is imperative to the CNS synthesis of 5-HT. GF mice show increased levels of plasma Trp and hippocampal 5-HT, and, interestingly, colonization with bacteria normalizes plasma Trp but not hippocampal 5-HT. Although the exact mechanisms of peripheral Trp regulation are unknown, the same study suggests that the microbiota modulate the degradation of Trp down the kynurenine pathway. In a separate study, this pathway interaction was observed and linked to behavioral phenotypes. Administration of Lactobacillus reuteri normalized stress-induced behavioral changes and was associated with decreased circulating kynurenine levels resulting from microbially derived H₂O₂ inhibition of ido1 messenger RNA expression.

The wide-ranging interaction of commensal bacteria with the gut-associated immune system and consequently the CNS is beyond the scope of this review and has been reviewed extensively elsewhere. Mouse models of multiple sclerosis and stroke have identified substantial roles for gut microbial regulation of autoimmunity, inflammation, and immune cell trafficking.^{,  ,  ,  ,}  It is important to highlight that the gut microbiota influence the development and function of the CNS resident immune cells, especially microglia. Relative to SPF mice, GF mice have compromised microglia maturation and morphology, leading to weakened early responses to pathogen exposure. This phenotype can be normalized by postnatal SCFA supplementation or colonization with a complex microbial community. Remarkably, antibiotic treatment to eradicate bacteria in SPF adult mice leads microglia to regain immature status, which then can be normalized by recolonization with complex microbiota, suggesting that active microbial signaling is required throughout adulthood to preserve microglial maturation.

Most evidence to date relies on vagal receptors that sense regulatory gut peptides, inflammatory molecules, dietary components, and bacterial metabolites to relay signals to the CNS, but there is also some evidence for direct activation of neurons by gut microbiota. Toll-like receptors 3 and 7, which recognize viral RNA, and Toll-like receptors 2 and 4, which recognize peptidoglycan and lipopolysaccharide, are expressed in the murine and human enteric nervous systems.^,  L rhamnosus (JB-1), B fragilis, and isolated polysaccharide A of B fragilis all have been shown to activate intestinal afferent neurons ex vivo. However, it remains unclear to what degree luminal microbial antigens make direct physical contact with neurons in vivo.

Microbial metabolites also are likely candidates mediating direct activation of neurons. The receptors FXR and TGR5 are expressed in brain neurons, but bile acid concentrations are low or undetectable in these tissues of healthy subjects.^,  Several studies have localized GPR41/FFAR3 receptors to the superior cervical ganglion, prevertebral ganglia, submucosal and myenteric ganglia neurons, sympathetic ganglia of the thoracic and lumbar sympathetic trunks, and vagal ganglion, suggesting neuronal activation by microbially derived SCFAs. Upon GPR41 activation, primary-cultured superior cervical neurons release norepinephrine, establishing this as a direct functional interface for microbial derivatives and the sympathetic nervous system.

The intestinal barrier is characterized by 2 layers: a basal monolayer of epithelial cells interconnected by tight junctions, and a mucus layer whose thickness and composition changes over time and that contains secretory IgA and antimicrobial peptides. Upon detection of specific microbial products, pattern recognition receptors located throughout GI mucosa can mediate the induction of enhanced antimicrobial defense, intestinal inflammation, and even immunologic tolerance.^,  Under healthy homeostatic conditions, many microorganisms and macromolecules gain entry through microfold cells (M cells), found in gut- and mucosa-associated lymphoid tissue, which enables constant sampling by immune cells. Paneth cells autonomously sense bacteria though MyD88-dependent Toll-like receptor activation, which triggers antimicrobial factors and ultimately limits bacterial penetration of host tissue. Microbes and microbe-derived ligands help maintain the cell-cell junctions critical to integrity.^,  Probiotic treatment can help normalize stress-induced barrier defects (discussed later) via unknown mechanisms.

The intestinal mucus layer is the second component of intestinal barrier function. Colonic mucus is organized into 2 layers: a thicker loose outer layer, and an inner layer attached firmly to the epithelium. Commensal microbes inhabit the outer layer, a critical habitat for biofilm formation, and a reliable energy source rich in glycoproteins that the microbiota degrade when deprived of dietary fiber, subsequently increasing pathogen susceptibility. The inner layer usually is bacteria-free and serves to protect epithelial cells from microbial contact through physical separation, innate immune mechanisms including antimicrobial peptides, and adaptive immune mechanisms including secretory IgA.

The blood-brain barrier (BBB) regulates molecular traffic between the circulatory system and the cerebrospinal fluid of the CNS. Gut microbiota can up-regulate the expression of tight junction proteins, including occludin and claudin-5, therefore decreasing BBB permeability. From intrauterine life through adulthood, GF mice have a more permeable barrier compared with controls, but introduction of normal gut microbiota to GF adults partially restores function. Permeability is decreased by monocolonization with SCFA-producing bacteria and oral gavage with sodium butyrate. SCFAs may serve as the primary signaling metabolite in BBB development and maintenance likely via entering cells and working as histone deacetylase inhibitors to epigenetically modulate the phenotype or via binding to GPR41 and/or GPR43.^, 

Systemic immune activation may cause disruptive BBB changes and often is modeled using LPS. But in a systematic review, studies evaluating in vivo LPS effects on BBB function only showed disruption 60% of the time, a figure potentially subject to publication bias. Interestingly, the host species is the only significant predictor explaining variance: mice are 4 times more likely than rats to show BBB change. Dose-dependent effects were not observed across all studies, although the levels used were mostly equivalent to septic doses. The variability of the BBB response in this model of systemic immune activation limits the generalizability of most preclinical findings to human microbiome interactions, especially in nonpathologic conditions.

Both branches of the autonomic nervous system (ANS) regulate gut functions including regional motility, secretion of gastric acid, mucus, bicarbonate, gut peptides, antimicrobial peptides, epithelial fluid maintenance, intestinal permeability, and mucosal immune response (reviewed by Mayer). These ANS-induced changes in gut physiology affect the microbial habitat, thereby modulating microbiota composition and activity.

Regional intestinal transit times affect water content, nutrient availability, and bacterial clearance rates. Relatively rapid flow in the small intestine inhibits permanent colonization of the upper gut, in particular in the proximal small intestine. The frequency of migrating motor complexes, which play a crucial role in intestinal transit during the fasting state, is influenced by food intake patterns, sleep quality, and stress. Impaired migrating motor complex regularity can reduce the flow rate, leading to small intestine bacterial overgrowth. Intestinal transit time assessed by the Bristol Stool Scale^,  strongly correlates with microbial richness and composition. In fact, a microbiome population level analysis identified such transit ratings as the top nonredundant covariate. A study using radiopaque markers for transit corroborated its association to microbial composition and additionally showed association with diversity and metabolism. In vitro simulation with Environmental Controls Systems for Intestinal Microbiota showed that increased transit time causally reduced bacterial biomass and diversity in distal gut regions.

Stress can cause epithelial barrier defects (leaky gut) by at least 2 mechanisms: direct modulation of epithelial permeability and alterations in the properties of the intestinal mucosal layer, ultimately leading to increased translocation of gut microbes or microbe-associated molecules. Rodent models have shown that jejunal and colonic permeability increases in response to both acute and chronic stress.^,  This increased leakiness facilitates the translocation of bacteria, such as Escherichia coli, and their products, such as lipopolysaccharide (LPS), leading to a proinflammatory environment in the gut, although there are conflicting reports describing stress-induced changes in expression of messenger RNA encoding tight junction proteins in the colon and jejunum.^,  Increased intestinal permeability and susceptibility to experimental inflammation observed in mouse models of depressive behavior induced by maternal separation is reversed by antidepressant therapy, highlighting brain-driven systemic and epithelial immune activity regulating the gut.

The ANS modulates the secretion of mucus by intestinal goblet cells, affecting the thickness and quality of the intestinal mucus layer. In addition to hypersensitivity, psychological stress leads to a less-protective mucus layer via catecholamine signaling, which alters the composition and size of secreted mucus.^{,  ,}  Changes in microbiota composition observed in a mouse model of brain injury are thought to result from altered mucoprotein production and goblet cell population size mediated by increased sympathetic nervous system signaling.^, 

In addition to CNS-induced changes in the intestinal microbial environment (eg, by influencing regional transit and secretions), the host neuroendocrine system can communicate with the microbiota more directly via intraluminal release of host signaling molecules, including but not limited to catecholamines, 5-HT, dynorphin, and cytokines, from neurons, immune cells, and ECCs.^,  The CNS likely modulates this process.^{,  ,}  Epinephrine and norepinephrine are shown to increase the virulence properties of several enteric pathogens as well as nonpathogenic microbes via stimulation of native quorum-sensing mechanisms.^{,  ,  ,  ,}  Other gut microbes contain sequences that share 24% to 42% identity to human genome sequences for the binding sites of melatonin, whose gut luminal concentrations have been reported at more than 10-fold serum concentrations in rats and pigs.^,  In vitro assays of one such microbe, Enterobacter aerogenes, show that melatonin not only induces swarming and motility behavior, but helps synchronize the circadian period and phase across culture plates. The gut microbiota show circadian rhythmicity in both abundance and expression in a manner dependent on the host and its behavior, especially feeding timing, and simulated jet-lag shifts composition, enhancing dysbiosis.

A number of studies (n = 22 in a total of 827 subjects) have reported significant microbial shifts in fecal microbial community composition between healthy controls and IBS patients, based on disease subtypes (diarrhea-predominant IBS, constipation-predominant IBS, and IBS mixed type), age (pediatric vs adult), and compartment (mucosa vs stool). Recent studies have suggested that there are at least 2 subgroups of patients who meet Rome criteria for IBS, based on gut microbial community structure, 1 subgroup not differing from healthy control subjects, despite similar GI symptoms.^,  In one of these studies, the dysbiotic IBS subgroup also differed in regional brain volumes from the eubiotic group, suggesting a relationship between microbial community structure and brain structure. Another recent study did not find a group difference in microbial composition between HCs and IBS, even though IBS symptom severity was correlated with dysbiosis. Despite a lack of consensus on the wide range of gut microbial differences between IBS subjects and healthy controls and the specific microbial changes that may be correlated to disease outcome, recent molecular-based methods of mucosal brushings or luminal aspirates have suggested decreased diversity in small-bowel microbiota with increased abundance of gram-negative organisms in IBS.^,  Based on analysis of fecal samples, regardless of the analytical methodology used, a number of studies reported a decreased relative abundance of the genera Bifidobacterium and Lactobacillus, and an increased Firmicutes:Bacteroidetes ratios at the phylum level. Because stress has been associated with a reduction in Lactobacilli in preclinical and clinical studies,^{,  ,}  one may speculate that the reported IBS-related changes in community structure and resulting metabolism are in part owing to altered ANS modulation of the gut as described earlier.