The gut–brain axis defines a bidirectional communication system that connects the GI tract with the CNS through multiple pathways involving neural, endocrine, and immune cells. The gut brain axis allows the CNS to regulate GI functions, including motility and secretion, and the GI tract to signal sensations such as hunger, pain, or discomfort to the CNS (Julio-Pieper et al., 2013). However, new complex emotional (affective mood), cognitive (memory formation) and behavioral functions (food intake) are beginning to be discovered (Berntson et al., 2003), and bacteria could play a role in this scenario (Cani and Knauf, 2016). Direct evidence that gut stimuli can affect emotional states includes the fact that the intragastric infusion of fatty acids in humans has been shown to reduce the brain response to experimentally induced sad emotions (Van Oudenhove et al., 2011).

The reciprocal influence between the GI tract and the CNS is sustained by the enteric nervous system (ENS) (Furness, 2006; Cryan and Dinan, 2012), which is considered as the third branch of the autonomic nervous system and consists of about 200–600 million neurons. Owing to its high number of neurons, complexity, and similarity in signaling molecules with the brain, ENS has also been referred to as the “second brain” (Gershon, 1998). ENS interfaces with the gut-associated lymphoid tissue (GALT, which contains more than two-thirds of the body’s immune cells) and with 1000s of entero-endocrine cells (EECs, which contain more than 20 identified hormones). Furthermore, the GI tract is directly innervated by the CNS through spinal and vagal afferents (Mayer, 2011). For this reason, gut-to-brain communications can occur through three main routes: (i) direct perception of stimuli by primary afferent neurons (belonging to the ENS or the CNS); (ii) immune system mediated connections; (iii) EEC mediated connections.

Studies in germ-free animals have underlined the importance of gut microbiota in the early-life development of the gut–brain axis, with particular reference to the hypothalamic-pituitary-adrenal axis (Sudo et al., 2004). Pattern recognition receptors such as TLRs, which are transmembrane proteins that are able to recognize bacterial envelope components, such as lipopolysaccharide (LPS) and lipoteichoic acids (Medzhitov, 2007), have been detected in a number of neuronal cells which could potentially mediate the direct neuronal sensing of microbial antigens (Lim et al., 2016). TLR4, which can be activated by bacterial LPS, has been detected in the nodose ganglion of the vagus nerve of rats (Hosoi et al., 2005). TLRs 3, 7, and 4 are expressed in the myenteric and submucosal plexus of the murine intestine, the human ileum and the lower dorsal root ganglia (Barajon et al., 2009). Similar results were also obtained by Qi et al. (2011), who detected the expression of TLRs 3, 7, and 9 in both human and murine dorsal root ganglion neurons. High mRNA levels of TLRs 1, 4, 5, and 6 have been found in the colonic dorsal root ganglion neurons of mice (Ochoa-Cortes et al., 2010). Furthermore, a number of studies have reported the direct excitation of neurons by LPS (Lim et al., 2016). However, in yet other studies the stimulation of pattern recognition receptors of neurons did not cause any direct excitation, but instead sensitized them and potentiated their excitability (Lim et al., 2016). Nonetheless, these researches have highlighted the existence of other possible pathways for bacteria sensing and communication pathways between the nervous and the immune systems.

Microbial exocellular polysaccharides (EPSs) are essential for protecting bacteria from the host immune response, but also for interacting with intestinal mucosal cells, including epithelial cells and EECs, thus inducing the release of molecular messengers that are able to modulate neural signaling or directly act on primary afferent axons (Forsythe and Kunze, 2013).

Mucosal immune cells are able to distinguish commensal from pathogenic bacteria (Artis, 2008) through TLRs, which are able to detect microbial antigens. Similarly, dendritic immune cells (extending their dendrites between epithelial cell tight junctions so as to directly sense the luminal environment) and some EECs can respond to the presence and activity of intraluminal microbial organisms through TLRs (Mayer, 2011). The intestinal infusion of Escherichia coli proteins has been reported to increase the secretion of anorexigenic GLP-1 and PYY, thus providing direct evidence of the ability of the gut microbiota to control appetite (Tennoune et al., 2014; Breton et al., 2016). In addition, EECs possess G-protein-coupled receptors that are able to sense microbial metabolites, such as amino acids (for example glutamate) (Bezençon et al., 2007), protein hydrolysates (Choi et al., 2007) and both long and short fatty acids (Tanaka et al., 2008; Nøhr et al., 2013). The expression of GPR41 receptor of SCFAs has been detected in several types of EECs (e.g., those secreting CCK, ghrelin, gastrin, GLP-1, PYY, neurotensin, secretin) throughout the GI tract of mice, from the stomach to the colon (Nøhr et al., 2013). Both the GPR41 and 43 receptors of SCFAs are expressed by enteroendocrine L-cells (Nøhr et al., 2013). Their stimulation promotes the secretion of GLP-1 and PYY (Samuel et al., 2008; Tolhurst et al., 2012; Nøhr et al., 2013). Furthermore, gut microbiota indirectly regulates the release of GLP-1, GLP-2 and PYY by modulating the differentiation of stem cells into EECs, and thus modifying the number of GLPs and PYY secreting L-cells (Everard et al., 2011). Another group of gut microbiota metabolites, namely indole and tryptamine derived from tryptophan, are known to modulate gut hormone release. Indole triggers GLP-1 secretion by enteroendocrine L-cells (Chimerel et al., 2014). Tryptamine induces ECs to secrete 5-HT (Takaki et al., 1985). Gut microbiota also contributes to regulating sensitivity to leptin, the so-called “satiety hormone” (Schéle et al., 2013).

Several microbial metabolites that display neuroactive properties have been described: (i) gaseous molecules, such as carbon monoxide, hydrogen sulfide and nitric oxide (Bienenstock and Collins, 2010); (ii) SCFAs such as n-butyrate, propionate, and acetate (Engelstoft et al., 2008; Nicholson et al., 2012); (iii) amines, such as putrescine, spermidine, spermine and cadaverine, which have been shown to be involved in CNS responses to stress (Bienenstock and Collins, 2010).

Curiously, bacteria can produce a wide range of molecules which mimic human hormones. For instance, muramyl dipeptide (similar to serotonin) (Masek and Kadlec, 1983) and indole (similar to melatonin) (Norris et al., 2013) can both cause sleep and drowsiness, while LPS from Gram-negative bacteria can directly act on thyroid cells, via type 4 TLRs, and up-regulate thyroglobulin gene expression (Nicola et al., 2009).

On the other hand, the direct response of bacteria to some of the regulatory peptides/neuroactive molecules that are secreted by EECs and/or the human nervous system has been demonstrated, which indicates the presence of neuromodulator/neurotrasmitter receptors in the bacterial envelope (Lyte, 2011). For example, the stress hormones epinephrine and norepinephrine increase the in vitro growth of E. coli by more than four orders of magnitude (Freestone et al., 2002) and the Clostridium/Bacteroides ratio in the human gut (Bailey et al., 2011). A recent report has demonstrated that Vibrio cholerae can respond to epinephrine and norepinephrine (enhancing the growth rate, swimming motility, and production of virulence factors such as iron sequestrating phenotype) by means of specific sensor proteins (Halang et al., 2015).

Actually, the most forefront hypotheses consider that brain and gut commensal bacteria communicate with each other through shared chemical mediators, and this is part of the homeostasis mechanisms that help to maintain gut microbiota stability and possibly brain functions and behavior (Bienenstock and Collins, 2010; Lyte, 2011). This neurochemical-mediated “two-way street” (Figure 1) is one of the principles that supports the microbial endocrinology construct (Lyte, 2011).

L-Glutamate (Glu) and γ-aminobutyric acid (GABA) are mainly known for their role as the main neuro-transmitters in the mammalian CNS, with excitatory and inhibitory roles, respectively (Hyland and Cryan, 2010; Julio-Pieper et al., 2013). However, these molecules are widespread in nature and have multiple functions, including plant signaling and communication between bacteria (Dagorn et al., 2013).

Several bacterial strains are able to produce Glu. Coryneform bacteria, such as Corynebacterium glutamicum, Brevibacterium lactofermentum and Brevibacterium flavum, have been used extensively for the industrial fermentative production of Glu (Sano, 2009). LAB strains belonging to Lactobacillus plantarum, Lactobacillus paracasei, and Lactococcus lactis are also able to synthesize Glu (Tanous et al., 2005). A recent study has revealed that about 15% of the LAB strains isolated from Asian fermented foods are Glu producers (Zareian et al., 2012).

Both prokaryotes and eukarya synthesize GABA through the decarboxylation of Glu by glutamate decarboxylase (GAD). GAD has been found in both Gram-positive and Gram-negative bacteria, where it is associated with systems that are involved in pH homeostasis and the generation of metabolic energy (i.e., proton motive force) (Pessione, 2012; Tsai et al., 2013). Marine microorganisms (Morse et al., 1980), E. coli (Richard and Foster, 2003) and Pseudomonas (Chou et al., 2008) can synthesize GABA. Among those microorganisms that are generally recognized as safe or health-promoting, several LAB (e.g., strains belonging to Lactobacillus, Lactococcus, and Streptococcus genera) and Bifidobacterium strains have been reported to biosynthesize GABA (Siragusa et al., 2007; Li and Cao, 2010; Mazzoli et al., 2010; Lyte, 2011). Recently, one Lactobacillus strain and four strains of Bifidobacterium isolated from the human intestine have been reported to being able to produce GABA (Barrett et al., 2012). Furthermore, a very recent analysis on metagenomic data from the Human Microbiome Project suggests that genes encoding GAD could be present in a significant proportion of human gut microbiota (Pokusaeva et al., 2016). Lactobacilli include the strains with the highest GABA production, although this metabolic ability is more likely strain- rather than genus-related (Li and Cao, 2010).

The role of GABA in the communication between bacteria is in line with the identification of GABA-binding proteins in different prokaryotes. These proteins consist not only of transporters, such as GabP in E. coli or Bacillus subtilis) (Brechtel and King, 1998; Hu and King, 1998) and Bra in Agrobacterium tumefaciens (Chevrot et al., 2006), but also of possible specific receptors (Guthrie and Nicholson-Guthrie, 1989). It has been reported that metabotropic Glu and GABA_B receptors, as well as bacterial periplasmic amino acid binding proteins, may have evolved from a common ancestor (Cao et al., 2009). In addition, as reported in the previous section, many bacteria (including those colonizing the human gut) can synthesize GABA, thus suggesting that, as in the case of eukaryotes, GABA might be a conserved and ubiquitous communication molecule.

The fact that GABA has an effect on the physiology of Pseudomonas can be supported by several observations. A periplasmic protein showing high affinity for GABA and displaying similar biochemical features to a subunit of the mammalian GABA_A ionotropic receptor was identified in an environmental strain of P. fluorescens (Guthrie and Nicholson-Guthrie, 1989; Guthrie et al., 2000). Specific receptors for sensing GABA have been found in other Pseudomonads including: the PctC of P. aeruginosa and the McpG of P. putida, which are involved in GABA chemoreception (Reyes-Darias et al., 2015). Another study has demonstrated that GABA increases the cytotoxicity and virulence of P. aeruginosa through a sophisticated modulation of the protein expression (Dagorn et al., 2013). A functional proteomic analysis revealed that six proteins were differentially expressed in P. aeruginosa when exposed to GABA (Lyte, 2011). Hence, some bacteria are able to respond to GABA, of both prokaryote and mammalian origin.

A structurally similar ion channel to eukaryotic ionotropic GABA_A receptors was identified in the plant pathogen Erwinia chrysanthemi (Zimmermann and Dutzler, 2011). The prokaryotic protein is activated by different amines, including GABA, and is modulated by benzodiazepines in a similar manner to its eukaryotic homolog (Zimmermann and Dutzler, 2011; Spurny et al., 2012).

Less information is available on prokaryotic Glu receptors. As far as we know, only a single case of a Glu-activated potassium channel in a bacterium (i.e., GluR0 from Synechocystis PCC 6803) has been reported to date (Chen et al., 1999). However, 100 prokaryotic channel proteins with putative Glu binding domains have recently been identified through a bioinformatic study (Ger et al., 2010). Among them, 22 proteins have been found to be homologs of vertebrate ionotropic Glu receptors (Ger et al., 2010).

Based on these findings, it is possible to hypothesize that members of the human gut microbiota can also sense and respond to GABA/Glu, although this has yet to be confirmed through dedicated studies. Furthermore, human-derived or dietary Glu can affect the balance of gut microbial populations by stimulating the growth of certain bacterial species at the expense of others. It is known, for instance, that LAB decarboxylate Glu to GABA in order to create a protonic gradient and enhance metabolic energy. Hence, Glu can constitute a privileged energy source that can be used to expand LAB populations.

Glu and GABA are the major excitatory and inhibitory neurotransmitters found in the human CNS, respectively (Julio-Pieper et al., 2013). These compounds are the molecular effectors of a homeostatic neuronal circuit that is characterized by extreme complexity and sophistication at both a physiological and a biochemical level. This system also includes glutamine (Gln), which plays a key role as a non-neuroactive intermediate in the recycling of neurotrasmitters in the brain, mainly Glu and GABA (Soeiro-de-Souza et al., 2015).

At a physiological level, this network consists of feed forward (FF) and feedback (FB) neuronal connections in which the main role of inhibitory (i.e., GABAergic) neurons is currently considered as that of balancing the activity of excitatory (i.e., Glutamatergic) neurons (Roth and Draguhn, 2012; Figure 2A). Both Glutamatergic and GABAergic neurotransmissions are complex systems in which glial cells (mainly astrocytes), pre- and post-synaptic neurons and a large spectrum of different receptors, transporters and enzymes are involved (Cherlyn et al., 2010; Femenía et al., 2012; Figure 2B).

The biochemical relationship between Glu and GABA goes far beyond this reciprocally modulating activity, since Glu and GABA are easily interconverted. Actually, the biochemical reactions involved in GABA-to-Glu-to-Gln conversion are integral features of such a regulatory network, which also involves the central energy metabolism of neurons (Soeiro-de-Souza et al., 2015; Figure 2B).

Glu is synthesized in situ in the CNS since it cannot pass the blood–brain barrier (BBB) (Hawkins, 2009). One of the main Glu biosynthetic pathways consists of the conversion of glucose (through the Embden-Meyerhof-Parnas pathway and the Krebs cycle) to α-ketoglutarate (α-KG), which is subsequently transaminated (Cherlyn et al., 2010). Glu is accumulated in vesicles and, after suitable stimulation, released to the synaptic space, where it can bind to a panel of Glu receptors (GluRs) that may be found in post-synaptic neurons. GluRs differ on the basis of their mechanism of action (e.g., ionotropic, iGluR, or metabotropic, mGluR, Glu receptors) and pharmacological responsiveness [e.g., N-methyl D-aspartate, NMDA, or α-amino-3-hydroxy-5-methylisoxazole-4-proprinoic acid (AMPA) or kainate sensitive receptors]. Ligand-gated ionotropic receptors mediate fast synaptic responses by directly regulating the ion influx. Metabotropic receptors are coupled to G-proteins and modulate slower signal transduction cascades (Hyland and Cryan, 2010; Julio-Pieper et al., 2013). Any excess Glu present in the synaptic cleft is taken up by glial cells or neurons to prevent the prolonged excitation of post-synaptic neurons, which can lead to cell death (Cherlyn et al., 2010). Glu in the glia is either oxidatively deaminated to α-KG [by Glu dehydrogenase (GluDH)], or converted to Gln by Gln synthetase. Gln is released from glial cells, uptaken by pre-synaptic neurons [via Gln transporters (GlnT)], and is then converted back to Glu by glutaminase. Apart from its role in neurotransmission, Glu is also involved in other neural processes, such as neuronal development and synaptic plasticity (Cherlyn et al., 2010).

Inhibitory pre-synaptic neurons convert Glu to GABA through GAD and pack it into vesicles via vesicular GABA transporters (vGATs). Again in this case, Gln is an alternative Glu source to α-KG. The multiple GABA receptors that are present in post-synaptic neurons include ionotropic (GABA_A, GABA_C) and metabotropic (GABA_B) receptors. Apart from GABA, these receptors are possibly modulated by a number of other compounds, e.g., barbiturates, baclofen, benzodiazepines and steroids (Simeone et al., 2003; Goetz et al., 2007). Any excess unbound GABA is cleared by glial cells through GABA transporters (GAT-1). GABA is then converted by GABA transaminase (GABA-T) into succinic semialdehyde, with concomitant Glu biosynthesis (Sivilotti and Nistri, 1991; Michels and Moss, 2007).

As described above, several bacteria, including probiotic bacteria and bacteria that colonize the human GI tract, are able to produce Glu or GABA (Barrett et al., 2012). As far as we know, only one study concerning the modulation of host Glutamate/GABAergic systems by GABA/Glu producing gut-colonizing bacteria is available in the literature (Bravo et al., 2011). A recent study has demonstrated that a GABA-producing Bifidobacterium dentium is able to attenuate sensitivity of dorsal root ganglia neurons in a rat model of visceral pain (Pokusaeva et al., 2016). It is worth noting that these few examples refer to the limited number of studies on interaction between gut microbiota and the host nervous system which is currently available. Experimental evidence is likely to grow as a higher number of reports become available. To date, most studies on the Glu/GABA effects on humans concern dietary Glu/GABA.

Glu and GABA can be found in food as natural components or as food supplements. Glu is among the most abundant amino acids (8–10%) found in dietary proteins. Furthermore, monosodium glutamate is employed extensively as a flavor additive in food (Brosnan et al., 2014). In Germany, Glu average intake (comprinsing both free Glu and Glu as protein constituent in foods) of about 10 g/day has been estimated (Beyreuther et al., 2007). In asian consumers this value is likely higher since free Glu intake can reach up to 4 g/day in these countries (Beyreuther et al., 2007). GABA is relatively abundant in plant food, such as brown rice germs and sprouts, spinach, barley and bean sprouts, where GABA concentrations of between 300 and 720 nmol/g dry weight have been detected (Oh et al., 2003), while higher levels are found in fermented foods (Abdou et al., 2006). In recent years, the utilization of GABA as a food supplement has progressively increased in both Asian and Western (USA and Europe) countries (Boonstra et al., 2015). GABA concentration of few micrograms per gram of cecal content has been detected in mice (Pokusaeva et al., 2016).

Hundreds of people have reported the benefits of using GABA-supplemented food, for instance in alleviating anxiety and/or improving sleep quality (Boonstra et al., 2015). Moreover, at least five different studies have reported that the oral administration of GABA or GABA-supplemented food/beverages (corresponding GABA amounts of about 50–100 mg) has had positive effects on human health. These effects include: (i) the reduction of psychological stress in people who performed arithmetic tasks (Nakamura et al., 2009; Kanehira et al., 2011; Yoto et al., 2012); (ii) the reduction of stress in acrophobic subjects exposed to heights (Abdou et al., 2006); (iii) an increased ability to perform prioritized planned actions (Steenbergen et al., 2015).

Multiple pathways by which gut luminal Glu/GABA may affect the CNS can be hypothesized, including their transport across intestinal and BBB and their sensing through afferent terminals that innervate the GI tract (Figure 3). In the next sections, evidence possibly supporting these mechanisms will be illustrated.