Antidepressive Mechanisms of Probiotics and Their Therapeutic Potential

The accumulating knowledge of the host-microbiota interplay gives rise to the microbiota-gut-brain (MGB) axis. The MGB axis depicts the interkingdom communication between the gut microbiota and the brain. This communication process involves the endocrine, immune and neurotransmitters systems. Dysfunction of these systems, along with the presence of gut dysbiosis, have been detected among clinically depressed patients. This implicates the involvement of a maladaptive MGB axis in the pathophysiology of depression. Depression refers to symptoms that characterize major depressive disorder (MDD), a mood disorder with a disease burden that rivals that of heart diseases. The use of probiotics to treat depression has gained attention in recent years, as evidenced by increasing numbers of animal and human studies that have supported the antidepressive efficacy of probiotics. Physiological changes observed in these studies allow for the elucidation of probiotics antidepressive mechanisms, which ultimately aim to restore proper functioning of the MGB axis. However, the understanding of mechanisms does not yet complete the endeavor in applying probiotics to treat MDD. Other challenges remain which include the heterogeneous nature of both the gut microbiota composition and depressive symptoms in the clinical setting. Nevertheless, probiotics offer some advantages over standard pharmaceutical antidepressants, in terms of residual symptoms, side effects and stigma involved. This review outlines antidepressive mechanisms of probiotics based on the currently available literature and discusses therapeutic potentials of probiotics for depression.


INTRODUCTION
Approximately 10 14 microbes, also known as gut microbiota, reside in the human gastrointestinal tract. The majority of these microbes belong to the Firmicutes, Bacteroidetes, Actinobacteria and Proteobacteria phyla. The gut microbiota flourishes in a symbiotic alliance with the host and, as such, has eminent regulatory effects on the host physiology. The gut microbiota actively engages with the proper development and functioning of both the immune system and brain. This is mediated by the microbiota-gut-brain (MGB) axis that lays the foundation for the intricate communicative pathways between gut microbiota and the nervous, immune and endocrine systems. However, the diversity and richness of gut microbiota are susceptible to change based on the host's lifestyle. An adverse change induces a gut dysbiosis which disrupts the symbiosis maintained by the MGB axis. Indeed, a gut dysbiosis has been linked to various health conditions, such as obesity, IBS, schizophrenia, Parkinson's disease and MDD (Sherwin et al., 2016;Thursby and Juge, 2017;van de Guchte et al., 2018).
Major depressive disorder is currently the leading cause of disability worldwide and is expected to outrank heart diseases as the number one disease burden by 2030 (Reddy, 2010;Tucci and Moukaddam, 2017). According to the Diagnostic and Statistical Manual of Mental Disorders-5, MDD is diagnosed when a person experiences most of the following symptoms for at least 2 weeks: depressed mood, anhedonia, excessive guilt, suicidal ideation, changes in appetite and sleep, psychomotor retardation, poor concentration and fatigue. Among these criteria, either depressed mood or anhedonia (or both) must be present for a diagnosis of MDD (American Psychiatric Association, 2013). In this review, the term "depression" would be used to refer to symptoms that characterize MDD.
A causal relationship potentially exists between the gut microbiota and MDD. Germ-free (GF) rodents developed depressive-like behaviors following fecal microbiota transplantation from MDD patients, but not from healthy people (Kelly et al., 2016;Zheng et al., 2016). As compared to healthy individuals, MDD patients have a different gut microbiota profile. The decrease in Faecalibacterium, Bifidobacterium, Lactobacillus (Aizawa et al., 2016), and Dialister (Kelly et al., 2016), and increase in Clostridium, Streptococcus, Klebsiella, Oscillibacter, Allistipes (Naseribafrouei et al., 2014;Jiang et al., 2015;Lin et al., 2017;Rong et al., 2019), Eggerthella, Holdemania, Gelria, Turicibacter, Paraprevotella, and Anaerofilum (Kelly et al., 2016) genera have been found among MDD patients. This shift in the gut microbiota composition may contribute to a shift in the regulation of the host physiology (Luan et al., 2017). It is, thus, worthwhile to tackle MDD from the MGB axis standpoint, with an emphasis on the gut microbiota.
Probiotics are microbes (usually lactic acid bacteria such as Lactobacilli and Bifidobacteria) that benefit the host physiology upon ingestion. Probiotics are marketed in the form of capsules, powder or fermented products. The global market size of probiotics amount to billions and is increasing annually due to consumers' interest in optimizing their health with functional foods (Di Cerbo and Palmieri, 2015). Probiotics have been utilized to modulate the MGB axis in an attempt to treat diseases, including MDD. Meta-analyses and systematic reviews have already supported the efficacy of probiotics in reducing clinical depression and depressive-like symptoms in MDD patients and healthy individuals, respectively Pirbaglou et al., 2016;Wang et al., 2016;McKean et al., 2017;Wallace and Milev, 2017).
To what extent are probiotics viable tools to treat MDD/depression? This review addresses this question by first outlining the workings of MGB axis and process by which this axis becomes maladaptive, leading to the development of depression. Antidepressive mechanisms of probiotics are further elucidated by drawing parallels between the physiological outcomes that accompanied the behavioral changes to the MGB axis from animal and human research. Lastly, in light of the heterogeneous nature of both the gut microbiota composition and depression subtypes in the clinical setting, challenges and potentials in translating probiotics for clinical use are discussed.

THE MGB AXIS AND DEPRESSION
Signaling Pathways of the MGB Axis: Neural and Humoral Routes The first point of contact between the gut microbiota and host nervous system is likely via the enteric nervous system (ENS). The ENS has been described as "the second brain" due to its neuronal complexity on par with the brain and its ability to function as an independent, discrete unit to regulate gutrelated activities and the immune system (Furness, 2012;Breit et al., 2018). Without gut microbiota, the excitability of enteric neurons would likely be attenuated, based on data observed in GF mice . Through the ENS, gut microbiota and the brain communicate bidirectionally through neural and humoral (systemic circulation) pathways (Luan et al., 2017). Parasympathetic vagus afferents carry neural information from internal organs, including the gut, to the brain (Breit et al., 2018). The vagus nerve also consists of motor neurons that innervate nearly all enteric neurons (Powley, 2000). This enables the brain to influence the activity of ENS to some extent, particularly the state of intestinal permeability and gut inflammation. Sympathetic spinal nerves also connect enteric neurons to the brain, albeit to a lesser extent than vagal nerves (Lomax et al., 2010;Breit et al., 2018). Additionally, the humoral route allows microbial metabolites to enter the systemic circulation and exert its effects elsewhere, including the brain. Likewise, the brain also sends chemical messengers, such as cytokines and glucocorticoids, via the humoral route to regulate the gut physiology (Luan et al., 2017). that respond to foreign antigens, such as LPS, as they penetrate the intestinal mucosal barrier. This promptly triggers production of inflammatory cytokines, mainly ILs, tumor necrosis factor (TNF)-α and IFN-γ (Sherwin et al., 2016). These cytokines enter the brain through various pathways. The humoral pathway enables cytokines to enter circumventricular organs or permeable regions of the BBB or bind to carrier proteins that cross the BBB. The neural pathway allows gut cytokines to stimulate specific brain areas such as the brainstem, hypothalamus and limbic structures via vagus and spinal afferents. The cellular pathway allows cytokines to be transported into the brain by the action of monocytes or macrophages. These cytokines could also bind to receptors on astrocytes and microglia, and subsequently trigger cytokine production within the brain (Schiepers et al., 2005;Miller and Raison, 2016).
When proinflammatory signals reach the brain, the hypothalamic-pituitary-adrenal (HPA) axis, a sympatheticneuroendocrine system, is activated to restore homeostasis. In response to stress, the hypothalamic paraventricular nucleus (PVN) synthesizes and releases corticotropin-releasing factor (CRF) to stimulate the anterior pituitary gland to release adrenocorticotropic hormone (ACTH) into the systemic circulation. ACTH stimulates the adrenal cortex to release glucocorticoids (cortisol in humans and corticosterone in rodents) which inhibit the release of CRF, establishing a negative feedback loop. Glucocorticoids are core effectors of the HPA axis that travel by the humoral route to exert its adaptive effects elsewhere; for instance, to reduce gut inflammation (Tsigos and Chrousos, 2002;Schiepers et al., 2005).

Dysregulated MGB Axis in Depression: Chronic Stress Response Loop
Acute psychological stress increases the release of ACh from cholinergic nerves (Saunders et al., 1997;Kiliaan et al., 1998) and glucocorticoids from the HPA axis (Alonso et al., 2012;Zheng et al., 2013;Vanuytsel et al., 2014), both of which loosen tight junctions of the intestinal barrier (Figure 1). Other stressors such as poor diet, sleep deprivation, antibiotics, environmental pollutants and excessive exercise also increase the intestinal permeability (Karl et al., 2018). Additionally, exposure to stress stimulates sympathetic spinal nerves to release NE into the gut which expedites quorum sensing systems and iron uptake of bacteria, leading to increased virulence and growth of pathogenic bacteria (e.g., Escherichia coli, Salmonella, Campylobacter, etc.) (Lomax et al., 2010;Freestone, 2013). These factors facilitate penetration of bacteria and their toxins, such as LPS, through the weakened intestinal barrier. Administration of LPS increased proinflammatory cytokines and caused anxiety and depression in healthy males in a dose-dependent manner (Grigoleit et al., 2011). This phenomenon is only transient due to the adaptive response of the immune system and HPA axis. However, chronic stress prevents this homeostatic restoration and causes prolonged inflammation and HPA axis overactivity, both of which aggravate the disrupted intestinal barrier. During this process, chronic inflammation renders the immune system insensitive to inhibitory signals from glucocorticoids (de Punder and Pruimboom, 2015). Excess proinflammatory cytokines, in turn, disrupt the negative feedback inhibition of circulating glucocorticoids of the HPA axis (Schiepers et al., 2005;Miller et al., 2009). Indeed, MDD patients often show increased intestinal barrier permeability (Stevens et al., 2018;Calarge et al., 2019;Ohlsson et al., 2019) and elevated serum antibodies against LPS (Maes et al., 2008).
Excessive glucocorticoids hyperactivate monoamine oxidases (MAOs; enzymes that degrade 5-HT, NE, and DA) (Grunewald et al., 2012). An overactive HPA axis can also induce gut dysbiosis (Murakami et al., 2017) and impairment of brain neurotransmitter systems (Pacak et al., 1993;Smith et al., 1995;Lopez et al., 1998;Hewitt et al., 2009). Higher baseline levels of cortisol, an indicator of an overactive HPA axis, were detected in more than 70% of MDD patients (Vreeburg et al., 2009;Lok et al., 2012). Proinflammatory cytokines and glucocorticoids upregulate indoleamine 2,3-dioxygenase (IDO) and tryptophan-2,3-dioxygenase (TDO) enzymes, respectively (Schimke et al., 1965;Young, 1981). Both enzymes metabolize TRP into KYN and quinolinic acid, which reduce the bioavailability of TRP to cross the BBB, thereby lowering 5-HT synthesis (Reus et al., 2015). This is evidenced by low plasma TRP levels that were also correlated to a heightened proinflammatory state found in MDD patients (Maes et al., 1993(Maes et al., , 1994. Furthermore, proinflammatory cytokines can decrease levels of DA, 5-HT and NE in the brain by upregulating their reuptake via presynaptic transporters and downregulating enzymatic cofactors required for their synthesis (Miller and Raison, 2016). Indeed, administration of cytokines consistently induced neurotransmitter imbalances in the brain and behavioral changes that are reminiscent of depression in animals and humans (Miller et al., 2009). Similarly, higher levels of proinflammatory cytokines were observed in depressed individuals as reported using meta-analyses of the data available in the literature (Howren et al., 2009;Dowlati et al., 2010).
A stress-induced inflamed gut adversely alters the relative abundances of preexisting bacteria in the gut (Figure 1). Acute psychological stress stimulated the release of inflammatory  Rowatt (1948), Girvin and Stevenson (1954), Pytka et al. (2016) mediators that were correlated with the lowered abundance of Coprococcus, Pseudobutyrivibrio, Dorea, and Lactobacillus in mice. This, in turn, allowed the proliferation of Clostridium species in the gut (Bailey et al., 2011). The gut microbiota of chronic-stressed mice also deviated from the baseline, whereby an increase in proinflammatory bacteria, such as Helicobacter and Streptococcus, and a decrease in butyrateproducing bacteria, such as Roseburia and Lachnospiraceae species, were observed (Gao et al., 2018). Altered gut microbiota composition consequently exacerbates gut inflammation and further increases intestinal permeability and production of proinflammatory cytokines (van de Guchte et al., 2018). The precise mechanism underlying vulnerability of certain bacteria to inflammation remains poorly understood. It is hypothesized that inflammation disrupts β-oxidation of intestinal epithelial cells (IECs, both enterocytes and colonocytes) to increase oxygen content in the gut lumen. This promotes formate oxidation that favors the growth of facultative anaerobes, such as E. coli, that are pathogenic and inflammatory at the cost of obligate anaerobes, such as Bacteroides and Firmicutes (Hughes et al., 2017). A dysregulated gut microbiota translates to a shift in the production of neuroactive metabolites and alters host neurotransmitter circuitry. This corresponds with disrupted levels of neurotransmitters in the brain of GF mice (Diaz Heijtz et al., 2011;Neufeld et al., 2011;Clarke et al., 2013;Pan et al., 2019). Altered neurotransmitter profile (e.g., GABA, Glu, 5-HT, FIGURE 1 | The maladaptive microbiota-gut-brain (MGB) axis in the pathophysiology of depression. Chronic exposure to stressors (e.g., psychological, poor nutrition) triggers prolonged release of (1) norepinephrine that alters gut microbiota composition by shifting to one that is enriched with pathogenic bacteria, and (2) acetylcholine and glucocorticoids that increase intestinal barrier permeability. The increased intestinal permeability allows bacteria and their toxins to enter systemic circulation, triggering stress responses from the HPA axis and immune system that, when excessive; (3) leads to chronic inflammation and HPA axis overactivity; (4) aggravate intestinal permeability; (5) alter composition of gut microbiota; and (6) disrupt neurotransmitter systems. Altered gut microbiota also results in an inflamed gut and (7) a shift in the production of bioactive molecules that regulate host neurotransmitter systems and gut motor functions. As a proof of concept, these five factors (in the circle) that depict the maladaptive MGB axis are often detected in MDD patients. Lastly, the constant negative emotions displayed by depressed patients further trigger a stronger reaction or sensitivity to various stressors.
DA, and NE) has been associated with the pathophysiology of depression. Therefore, pharmaceutical antidepressants function to restore synaptic levels of neurotransmitters (Harald and Gordon, 2012). In addition, impaired neurotransmitter systems within the ENS may alter gut motor function. This has direct implications as gut motility is a determining factor in the size and diversity of gut microbiota (Quigley, 2011). Therefore, chronic stress sets up a vicious cycle of increased intestinal permeability, chronic inflammation, hyperactive HPA axis, altered gut microbiota profile and neurotransmitter imbalancesforming a maladaptive MGB axis (Figure 1). Furthermore, MDD patients perceive stress as more threatening and challenging to cope with compared to healthy individuals (Farabaugh et al., 2004;Salomon et al., 2009). These negative emotions can increase their sensitivity to stressors, such as an elevated cortisol response (Mendonca-de-Souza et al., 2007). To restore this malfunctioned axis, probiotics have been demonstrated by meta-analyses and systematic reviews as a potential treatment for MDD/depression Pirbaglou et al., 2016;Wang et al., 2016;McKean et al., 2017;Wallace and Milev, 2017). Potential antidepressive mechanisms of probiotics are elucidated in the following section.

DELINEATING THE ANTIDEPRESSIVE MECHANISMS OF PROBIOTICS
Probiotics secrete a wide range of signaling molecules that operate via distinct pathways to exert their effects, be it antidepressive, immunomodulatory or modulation of neurotransmission (Luan et al., 2017). This review classifies probiotic-associated signaling molecules into four types: neurotransmitters, bacterial secreted proteins, butyrate and other bioactive molecules (Figure 2). Some probiotics can secrete signaling molecules of different types. In this regard, the mechanisms of individual probiotics will be presented in the order of pertinence and similarity to each other.

Lactobacillus rhamnosus
Lactobacillus rhamnosus JB-1, the typical experimental strain of L. rhamnosus, was formerly referred to as Lactobacillus reuteri. Orally administered L. rhamnosus reduced depressivelike behaviors in normal, healthy mice (Bravo et al., 2011) and chronic-stressed mice (McVey Neufeld et al., 2018). Postpartum women (Slykerman et al., 2017) and obese individuals ) that were supplemented with L. rhamnosus reported lower depressive thoughts compared to the control group. In vagotomized rats, behavioral and physiological benefits of L. rhamnosus were abolished (Bravo et al., 2011). This substantiates the vagus nerve as an essential conduit in the signaling pathway of L. rhamnosus. Introduction of L. rhamnosus into the gut lumen heightened the firing rate of vagus nerve and enteric neurons in mice (Perez-Burgos et al., 2013. These findings suggest that L. rhamnosus signals to the brain via the neural route, which may influence the central GABAergic system and HPA axis to manifest an antidepressive effect (Figure 2A). However, it is unclear whether neurotransmitters, cytokines or other molecules are involved in the neural signaling of L. rhamnosus. FIGURE 2 | Signaling mechanisms underlying antidepressive effects of probiotics mediated through secretion of (A) Neurotransmitters: L. rhamnosus and L. casei secrete GABA that may signal central GABAergic system and HPA axis via the neural route. L. brevis secretes GABA that enhances sleep. L. helveticus secrete 5-HT that may signal the central 5-HT system via the neural route. L helveticus also secretes NE that may affect the central NE system. L. reuteri secretes histamine that decreases secretion of proinflammatory cytokines by IECs. This may reduce circulating inflammatory markers, such as LPS, IL-6 and corticosterone, and subsequently prevent the inflammation-induced decrease in hippocampal BDNF. (B) Butyrate: L. plantarum produces butyrate that strengthens intestinal barrier and diffuses through the circulation to regulate BDNF expression and reduce inflammation in the brain. The latter consequently regulates the HPA axis and its regulator, the DA system. C. butyricum produces butyrate that influences central 5-HT and BDNF systems and stimulates L cell to secrete GLP-1 into the bloodstream which increases expression of GLP-1 receptors. F. prausnitzii produces butyrate that strengthens the intestinal barrier. B. infantis and L. paracasei promote growth of butyrate-producing bacteria. Through butyrate, B. infantis upregulates Tph1 activity of EC which increases circulating 5-HT and strengthens intestinal barrier to lower IDO activity and increase circulating TRP, both of which affect the central 5-HT system and BDNF expression. Through butyrate, L. paracasei may influence the central 5-HT system and BDNF expression. (C) Bacterial secreted proteins: L. gasseri secretes gassericins that increase parasympathetic activity to facilitate sleep and improves gut microbiota composition. B. longum secretes serpins that alter neural activities in the brain via the neural route. L. paracasei secretes lactocepins that decrease proinflammatory chemokines in IECs. This lowers IDO activity which, in turn, affects the central 5-HT system and BDNF expression. (D) Other bioactive molecules: B. infantis secretes bioactive factors (likely polysaccharides) that decrease circulating IL-6 which affects the central NE system. L. reuteri secretes H 2 O 2 that decreases IDO activity and circulating KYN, and dgk that inhibits the initiation of proinflammatory pathways. B. breve converts albiflorin into BZA which affects the Glu system via the humoral route. L. kefiranofaciens secretes exopolysaccharides that have immunomodulatory and antibacterial properties, which may potentially prevent HPA axis overactivity. 5-HT, 5-hydroxytryptamine or serotonin; BDNF, brain-derived neurotrophic factor; DA, dopamine; BZA, benzoic acids; dgk, diacylglycerol kinase; ECs, enterochromaffin cells; EPS, exopolysaccharide; GABA, gamma-Aminobutyric acid; GLP-1, glucagon-like peptide-1; Glu, glutamate or glutaminergic; H 2 O 2 , hydrogen peroxide; HPA, hypothalamic-pituitary-adrenal; IECs, intestinal epithelial cells; IDO, indoleamine 2,3-dioxygenase; IL-6, interleukin-6; KYN, kynurenine; NE, norepinephrine; LPS, lipopolysaccharides; Tph1, tryptophan hydroxylase 1; TRP, tryptophan.

Microbial GABA, Central GABAergic System, and HPA Axis
Glutamine is a precursor to Glu while Glu is a precursor to GABA. Reduced levels of GABA and Glx (Glu + glutamine) have been consistently reported in cortical regions of MDD patients (Sanacora et al., 1999;Hasler et al., 2007;Bhagwagar et al., 2008;Moriguchi et al., 2018;Godlewska et al., 2019). A dysfunctional glutaminergic system, that is partly responsible by a decreased GABAergic tone, is also implicated in MDD (Murrough et al., 2017). N-acetyl aspartate (NAA) is regarded as a marker for neuronal vitality. In MDD patients, decreased NAA levels in the PFC and hippocampus have been detected (Gonul et al., 2006;Olvera et al., 2010;Lefebvre et al., 2017). These neurochemical (i.e., Glx, NAA, and GABA) levels in the PFC and hippocampus of mice increased when administered with L. rhamnosus (Janik et al., 2016), implicating its antidepressive potential.
Intake of L. rhamnosus altered the central mRNA expression of GABA A and GABA B receptors while reducing depressiveand anxiety-like behaviors in mice. These effects were also dependent on an intact vagus nerve (Bravo et al., 2011). With prebiotics, L. rhamnosus intake decreased hippocampal GABA Aα2 mRNA expression in stressed mice (McVey Neufeld et al., 2017). L. rhamnosus produced GABA and Glu efficiently from microbial glutamate decarboxylase and glutaminase, respectively, in vitro (Stromeck et al., 2011;Liao et al., 2013;Lin, 2013). These biosynthetic machineries utilized by microbes to synthesize Glu and GABA are mutual in neurons (Mathews and Diamond, 2003), which support the interkingdom communication of microbial GABA (Lyte, 2011). It was demonstrated in vitro that gut microbial GABA can cross the intestinal barrier via H + /GABA symporter (Thwaites et al., 2000;Nielsen et al., 2012). The microbial GABA may subsequently interact with GABA receptors and transporters that are widely expressed on enteric neurons and vagus afferents (Hyland and Cryan, 2010).
Administration of L. rhamnosus reduced stress-induced plasma corticosterone levels in mice that averted depression (Bravo et al., 2011;McVey Neufeld et al., 2018). This could be due to the innervation of PVN neurons by GABAergic synapses that can be desensitized by acute stress (Hewitt et al., 2009). Inhibited GABA signals allow continuous release of CRF by PVN neurons, which ultimately leads to cortisol overproduction and HPA axis overactivity (Cullinan et al., 2008). Impairment of GABA receptors also inhibits hippocampal neurogenesis, which has been shown to activate the HPA axis and induce depression in mice (Earnheart et al., 2007;Schloesser et al., 2009). Such effects may be possibly prevented by the production of GABA by L. rhamnosus.

Lactobacillus casei Strain Shirota
Individuals with low mood reported feeling happier after consuming milk containing L. casei, but not the placebo (Benton et al., 2007). Intake of mixed-species probiotics that included L. casei also reduced clinical depression and depressive-like symptoms in MDD patients (Akkasheh et al., 2016) and healthy individuals (Steenbergen et al., 2015;Mohammadi et al., 2016), respectively. Similar to L. rhamnosus, evidence suggests that L. casei may also regulate the HPA axis via the neural route (Figure 2A).

Microbial GABA and HPA Axis
Intake of L. casei stimulated vagus afferents and decreased both the activity and quantity of CRF-expressing cells in PVN of rats . Intragastric injection of L. casei downregulated the activity of sympathetic efferents to adrenal glands and liver, and this effect ceased upon vagotomy (Tanida et al., 2014). In clinical trials, L. casei supplementation lowered salivary cortisol levels, feelings of stress and frequency of abdominal-and flu-related symptoms in stressed individuals Takada et al., 2016). These studies imply that L. casei prevents HPA axis overactivity via the vagus nerve, which may consequently lower stress-related feelings and illnesses. L. casei produced GABA in vitro (Oleskin et al., 2014), indicating a possibility that it may share an antidepressive mechanism of L. rhamnosus. Stressed individuals that consumed L. casei showed improvements in mental health and gut microbiota composition, characterized by increased Lactobacillus and Bifidobacterium populations (Rao et al., 2009;Kato-Kataoka et al., 2016). As most of the antidepressive probiotics belong to Lactobacillus and Bifidobacterium genera, the potential antidepressive capacity of L. casei is highly supported.

Lactobacillus brevis
Similar to L. rhamnosus and L. casei, L. brevis produces GABA via glutamate decarboxylase in substantial amounts (Yokoyama et al., 2002;Siragusa et al., 2007;Barrett et al., 2012;Ko et al., 2013;Yunes et al., 2016). This indicates that L. brevis may share a mutual mechanism of action with L. rhamnosus and L. casei (Figure 2A). Although L. brevis has been shown to influence neither the central GABAergic system nor the HPA axis, L. brevis appears to promote sleep.

Microbial GABA and Sleep
Milk fermented with L. brevis had increased GABA content. This L. brevis-fermented milk demonstrated an antidepressive potency on par with fluoxetine, a SSRI, in depressed rats (Ko et al., 2013). Intriguingly, intake of L. brevis-produced GABA improved sleep duration in mice (Han et al., 2017). Another study also showed that dietary L. brevis enhanced sleep quality and voluntary physical activity in mice (Miyazaki et al., 2014). GABA is the main inhibitory neurotransmitter that is widely associated with sleep, and GABA receptors are frequent targets for pharmaceutical drugs, such as benzodiazepine, to treat insomnia (Gottesmann, 2002). GABA-enriched foods and GABA extract have also been shown to improve sleep quality in insomniacs (Byun et al., 2018) and healthy individuals (Yamatsu et al., 2015). Therefore, L. brevis has therapeutic value for insomnia, which reflects one of the diagnostic criteria for MDD (American Psychiatric Association, 2013).

Lactobacillus reuteri
Treatment of L. reuteri ameliorated depressive-like behaviors in chronic-stressed (Marin et al., 2017) and immobilization-stressed mice (Jang et al., 2019). The former study further elucidated the mechanism of L. reuteri which involves regulation of IDO, a ratelimiting enzyme of immune cells that catabolizes TRP to KYN (Reus et al., 2015). It is also well documented that L. reuteri exhibits anti-inflammatory activities (Thomas et al., 2012;Gao et al., 2015;Ganesh et al., 2018). It is, thus, conceivable that L. reuteri may also prevent activation of IDO by proinflammatory cytokines (Reus et al., 2015).

Microbial Hydrogen Peroxide and Kynurenine Pathway
The etiology of depression is partly attributed to a dysregulated KYN/TRP pathway (Reus et al., 2015). An elevated ratio of plasma KYN/TRP often correlates positively with the depression severity in human (Maes et al., 2002;Gabbay et al., 2010;Baranyi et al., 2013;Zhou et al., 2019). It was demonstrated that L. reuteri intake improved behaviors of depressed mice by reversing the stress-induced (1) decrease in fecal H 2 O 2 levels and Lactobacillus populations, and (2) increase in intestinal IDO1 expression and plasma KYN levels (Marin et al., 2017). KYN administration attenuated this antidepressive effect, which indicates that L. reuteri ameliorates depression by reducing plasma KYN levels. This study also showed that L. reuteri generated high amounts of H 2 O 2 in vitro, and the author proposed that H 2 O 2 is the key metabolite in mediating antidepressive effect of L. reuteri (Marin et al., 2017). This is because H 2 O 2 catalyzes peroxidasemediated reactions that inhibit IDO activity (Freewan et al., 2013). H 2 O 2 is transported by aquaporin-3 transporters that are expressed on IECs (Thiagarajah et al., 2017) and immune cells (Moon et al., 2004). These findings suggest that microbial H 2 O 2 can potentially cross the intestinal barrier to suppress IDO activity in immune cells, which would lower circulating KYN levels ( Figure 2D).

Microbial Histamine, Diacylglycerol Kinase, and Brain-Derived Neurotrophic Factor (BDNF) Expression
Lactobacillus reuteri possesses histidine decarboxylase that converts dietary L-histidine to histamine, which inhibits the production of TNF-α in vitro (Thomas et al., 2012;Hemarajata et al., 2013). The microbial histamine suppressed proinflammatory cytokine activities in IECs via the histamine-2 receptor signaling pathway in mice. This effect disappeared when the histidine decarboxylase gene of L. reuteri was inactivated by mutagenesis (Gao et al., 2015). Intriguingly, microbial histamine also activated histamine-1 receptors to initiate downstream proinflammatory pathways in mice (Ganesh et al., 2018). However, the substrate for this pathway, diacylglycerol, is metabolized to phosphatidic acid by diacylglycerol kinase produced by L. reuteri. Thus, L. reuteri secretes both histamine and diacylglycerol kinase that act on histamine receptors to produce an anti-inflammatory effect (Ganesh et al., 2018). Orally administered L. reuteri simultaneously alleviated colitis and behaviors indicative of anxiety and depression in stressed mice. These effects were also accompanied by a decrease in colon inflammation and blood levels of LPS, interleukin-6 (IL-6) and corticosterone. In the same study, this reduction in peripheral inflammation prevented the infiltration of activated microglia into the hippocampus and increased hippocampal BDNF expression (Jang et al., 2019; Figure 2A). BDNF has been extensively studied for its vital role in neuronal function and its causal link to depression. Antidepressants such as SSRI and ketamine also increase hippocampal BDNF expression as part of their mechanism of action (Bjorkholm and Monteggia, 2016). Furthermore, this anti-inflammatory effect of L. reuteri may prevent IDO activation by proinflammatory cytokines (Reus et al., 2015).

Lactobacillus plantarum
Lactobacillus plantarum supplementation decreased depressivelike symptoms in chronic-stressed mice (Liu Y.W. et al., 2016;Dhaliwal et al., 2018) and stressed adults with mild depression (Lew et al., 2018), though the latter study did not reach statistical significance. Following L. plantarum intake, reduction in plasma corticosterone levels and inflammation were seen in mice with reduced depressive-like behaviors (Liu Y.W. et al., 2016). Another study reported that mice fed with L. plantarum displayed an increase in cecum SCFAs levels (acetic and butyric), and a decrease in intestinal permeability and level of MAOs in the brain (Dhaliwal et al., 2018). These physiological changes can be unified into a mutual mechanism that L. plantarum likely mitigates systemic inflammation ( Figure 2B).

Butyrate, Intestinal Barrier, and BDNF Expression
Chronic-stressed mice fed with L. plantarum exhibited reduced depressive-like behaviors, coupled with an increase in butyrate and butyrate-producing bacteria, such as Lactobacillus, Bacteroidetes, and Roseburia (Dhaliwal et al., 2018). L. plantarum synthesizes butyrate via fatty acid synthase II-thioesterase, a glutamine-mediated butyrogenic pathway (Botta et al., 2017). Butyrate can enter IECs through cholesterol-rich microdomains and/or monocarboxylate transporter 1 protein Goncalves et al., 2011;Nedjadi et al., 2014), and promote synthesis and assembly of tight junction proteins of IECs (Bordin et al., 2004;Ohata et al., 2005;Peng et al., 2009;Yan and Ajuwon, 2017). Butyrate also has anti-inflammatory properties; for instance, butyrate inhibited proinflammatory activities of IECs in vitro (Elce et al., 2017) and interacted with IECs to regulate host T cell responses (Lew et al., 2018;Xu et al., 2018). Butyrate may also diffuse into the systemic circulation to exert anti-inflammatory effects on various organs and tissues, including the brain (McNabney and Henagan, 2017; Matt et al., 2018). Indeed, butyrate has been shown to normalize behavior of depressed rodents through epigenetic regulations of hippocampal BDNF expression (Han et al., 2014;Wei et al., 2014;Sun et al., 2016). These outcomes are consistent with the finding that L. plantarum intake increased hippocampal BDNF expression and cecum butyrate levels in chronic stress-induced depressed mice (Dhaliwal et al., 2018).

HPA Axis and Central DA System
Lactobacillus plantarum supplementation decreased MAOs levels in brain tissues of mice with reduced depression (Dhaliwal et al., 2018). This is in line with another finding that L. plantarum intake in mice increased levels of DA and its metabolites (HVA and 3,4-dihydroxyphenylacetic acid, DOPAC) in the PFC, along with reduced depressive-like behaviors (Liu Y.W. et al., 2016). However, another study showed that L. plantarum increased DA levels in the striatum of mice while alleviating anxiety-like behaviors (Liu W.H. et al., 2016). These studies suggest that L. plantarum likely affects the central DA system in a context-dependent manner. It was also proposed that L. plantarum increases DA levels in the PFC to prevent HPA axis overactivation (Liu Y.W. et al., 2016). DA neurons in the PFC and ventral tegmental area (VTA) form the mesocortical pathway which regulates reward-seeking behaviors (Pariyadath et al., 2016) and the HPA axis (Sullivan and Dufresne, 2006). Glucocorticoids from the HPA axis can also influence the DA system either directly or indirectly, via epigenetic control and MAOs inhibition, respectively (Feenstra et al., 1992;Grunewald et al., 2012;Butts and Phillips, 2013). Taken together, L. plantarum may regulate both the DA system and HPA axis by attenuating glucocorticoid-induced MAOs activity.

Faecalibacterium prausnitzii (Previously Known as Fusobacterium prausnitzii)
Recently, it was discovered that oral gavage of F. prausnitzii exerted antidepressive and anxiolytic effects in chronic-stressed mice (Hao et al., 2019). F. prausnitzii, as the sole species of Faecalibacterium genera (Duncan, 2002), represents around 5% of the total human gut microbiota (Hold et al., 2003). Low populations of F. prausnitzii correlated with the disease severity of those with MDD (Jiang et al., 2015) and bipolar depression (Evans et al., 2017). In a recent large cohort study, fecal levels of F. prausnitzii correlated negatively with depressed mood and positively with quality of life (Valles-Colomer et al., 2019). Therefore, F. prausnitzii seems to have pertinent contributions to mental health.

Butyrate, Microbial Anti-inflammatory Molecules, and Peripheral Inflammation
Faecalibacterium prausnitzii produces butyrate in large quantities from fermenting glucose and fiber (Duncan, 2002;Hold et al., 2003). F. prausnitzii also secretes microbial anti-inflammatory molecules that suppress the proinflammatory nuclear factor (NF)-κB pathway in IECs (Sokol et al., 2008;Quevrain et al., 2016a,b). These immunomodulatory effects are consistent with neurochemical changes observed in F. prausnitzii-treated depressed mice, whereby cecum SCFAs and plasma IL-10 levels increased, while corticosterone and IL-6 levels decreased (Hao et al., 2019). Moreover, intragastric administration of F. prausnitzii decreased colonic cytokine levels and intestinal permeability in mice with colitis (Laval et al., 2015;Martin et al., 2015). Thus, butyrate produced by F. prausnitzii potentially strengthens the intestinal barrier (similar to L. plantarum; Figure 2B). However, whether local immunomodulatory effects of F. prausnitzii extend to the brain remains unknown.
Nevertheless, the ability of F. prausnitzii to attenuate gut inflammation is sufficient to reduce depressive-and anxiety-like behaviors in mice (Hao et al., 2019).

Lactobacillus helveticus
Lactobacillus helveticus intake enabled the recovery of chronicand subchronic-stressed rodents from their state of depression (Liang et al., 2015;Maehata et al., 2019). Probiotic sticks containing L. helveticus, in addition to Bifidobacterium longum, reduced clinical depression and depressive-like symptoms in MDD patients (Kazemi et al., 2019) and healthy individuals (Messaoudi et al., 2011), respectively. Most of the animal and human studies also showed that L. helveticus intake enhanced memory and, sometimes, attention and learning (Ohland et al., 2013;Chung et al., 2014;Luo et al., 2014;Liang et al., 2015;Ohsawa et al., 2018). Cognitive impairments, such as poor memory and concentration, represent one major cluster of MDD symptoms (Sharpley and Bitsika, 2014). Evidence suggests that L. helveticus may modulate the central NE system and HPA axis to improve cognition, and the central 5-HT system and BDNF expression to reduce depression (Liang et al., 2015) (Figure 2A).

Microbial NE, Central NE System, and HPA Axis
Supplementation of L. helveticus improved memory and cognitive performance in chronic-stressed rats, comparable to the SSRI citalopram-treated rats. This memory improvement correlated with increased plasma IL-10 and hippocampal NE levels, and reduced plasma corticosterone and ACTH levels (Liang et al., 2015). A previous study also showed that ingestion of L. helveticus enhanced memory and mitigated gut inflammation in neuroinflammation-induced rats (Luo et al., 2014). However, another study reported that memory improvement in L. helveticus-treated mice did not correlate with the state of gut inflammation (Ohland et al., 2013). Despite this discrepancy, it is well established that the hippocampal NE system and HPA axis both interact to regulate hippocampal glucose metabolism for memory consolidation (Osborne et al., 2015). This mechanism may be affected by microbial NE as L. helveticus produced NE in vitro in amounts that exceed the human bloodstream (Oleskin et al., 2014). It was also shown in vivo that gut bacteria are responsible for converting conjugated NE into its biologically active form (Asano et al., 2012). This neuroactive NE likely influences the MGB axis, but the exact mechanism remains unknown (Lyte, 2011). Liang et al. (2015) showed that elevated hippocampal 5-HT levels correlated with reduced depression severity in L. helveticus-fed rats. The same study also demonstrated that treatment with SSRI citalopram alleviated depression and increased hippocampal BDNF expression and 5-HT levels (Liang et al., 2015). Hence, the antidepressive mechanism appears similar between L. helveticus and citalopram. Cultures of L. helveticus produced 5-HT at concentrations close to that in the human bloodstream (Oleskin et al., 2014). As shown in vivo, the gut microbiota has an indispensable function in deconjugating glucuronide-conjugated 5-HT to generate their free, biologically active counterparts in considerable amounts (Hata et al., 2017). It is hypothesized that gut luminal 5-HT may sensitize 5-HT 3A receptors of enteric neurons by stimulating the glial cell-derived neurotrophic factor of IECs (Hata et al., 2017). 5-HT3 receptors are also expressed on IECs (Hasler, 2009) and vagal afferents (Hillsley and Grundy, 1998). Therefore, it can be speculated that L. helveticus influences the central 5-HT circuitry via the neural route. This is supported by a recent study showing that L. helveticus intake increased expression of 5-HT 1A receptors in the nucleus accumbens while restoring behaviors of depressed mice (Maehata et al., 2019).

Microbial 5-HT and Central 5-HT-BDNF System
Chronic-stressed mice that ingested L. helveticus displayed an increase in hippocampal BDNF levels (Liang et al., 2015) and neurogenesis in the nucleus accumbens (Maehata et al., 2019). Nucleus accumbens is a brain region implicated in reward behavior. The central BDNF and 5-HT systems are synergistic, whereby 5-HT upregulates hippocampal BDNF-TrkB signaling to increase expression and synthesis of BDNF. The elevated BDNF, in turn, facilitates neurogenesis of 5-HT neurons (Martinowich and Lu, 2008;Bjorkholm and Monteggia, 2016). Therefore, L. helveticus likely increases hippocampal BDNF levels via modulation of 5-HT circuitry, in a similar manner to SSRIs (Liang et al., 2015).

Lactobacillus paracasei
Dietary intervention of heat-killed L. paracasei prevented mood deterioration in times of stress in healthy individuals (Murata et al., 2018). In corticosterone-induced depressed mice, oral gavage of either live or heat-killed L. paracasei exhibited antidepressive efficacy equivalent to or better than fluoxetine. The same study also showed that live and heat-killed L. paracasei operated via different mechanisms. Live L. paracasei increased 5-HT levels whereas heat-killed L. paracasei increased DA levels in the brain (Wei et al., 2019). The signaling mechanism of L. paracasei appears independent of the HPA axis (Wei et al., 2019) or vagus afferents (Tanida and Nagai, 2011). The remaining evidence suggests that L. paracasei potentially functions via an immune-mediated humoral pathway.

Lactocepin, Butyrate, and Central 5-HT-BDNF System
Lactobacillus paracasei secretes lactocepin, a PrtP-encoded serine protease, that selectively degrades proinflammatory chemokines in inflamed ileal tissue of mice (von Schillde et al., 2012). Lactocepin is most likely a heat-labile cell surface protein unique to L. paracasei (Hoermannsperger et al., 2009;von Schillde et al., 2012). Mice fed with live L. paracasei exhibited lower inflammatory markers in serum, such as increased IL-10 and glutathione peroxidase and decreased TNF-α and MCP-1 . Another study showed that oral gavage of live L. paracasei with its bacterial products prevented adverse effect of stress on intestinal permeability in rats (Eutamene et al., 2007). This can be linked to a suppressed IDO activity, resulting in higher TRP bioavailability for 5-HT synthesis in the brain (Reus et al., 2015). Following this, it was shown that live L. paracasei delivered via gavage increased 5-HT and 5-HIAA (the main metabolite of 5-HT) levels in the hippocampus and striatum of mice Wei et al., 2019). As 5-HT facilitates BDNF synthesis (Martinowich and Lu, 2008), the upregulated central 5-HT expression presumably explains the accompanying increase in hippocampal BDNF expression of mice alleviated of depression from L. paracasei intake (Wei et al., 2019). Therefore, L. paracasei may upregulate the central 5-HT-BDNF system (similar to L. helveticus; Figure 2C).
Treatment of live L. paracasei also increased fecal Bifidobacterium populations while normalizing behaviors of depressed mice (Wei et al., 2019). The gut microbiota profile, inflammatory markers and levels of acetate and butyrate were improved in IBS patients supplemented with live L. paracasei (Bertani et al., 2017;Cremon et al., 2018). Reduction in systemic inflammation, coupled with an improvement in hippocampal function, was also observed in obese rats fed with live L. paracasei (Chunchai et al., 2018). Thus, live L. paracasei may facilitate the colonization of butyrate-producing bacteria to reduce systemic inflammation (similar to L. plantarum) and increase 5-HT secretion from ECs (similar to Bifidobacterium infantis; Figure 2B).

Bifidobacterium infantis
In naïve rats, intake of B. infantis was shown to alter depressionrelated biomarkers (Desbonnet et al., 2008). The same group later showed that chronic-stressed mice no longer displayed depressive-like behaviors after B. infantis intake (Desbonnet et al., 2010). In flood victims with IBS, B. infantis consumption did not affect their IBS symptoms but improved their mental health instead (Murata et al., 2018). B. infantis did not influence corticosterone levels in mice (Desbonnet et al., 2008(Desbonnet et al., , 2010, implying that the effect of B. infantis is likely to be independent of the HPA axis. Evidence suggests that B. infantis has immunomodulatory effects that regulate the central NE system (Desbonnet et al., 2010). A recent study also provided support for the antidepressive mechanism of B. infantis that involves the hippocampal 5-HT system (Tian et al., 2019).

Bioactive Factors, IL-6, and Central NE System
Bifidobacterium infantis treatment manifested two physiological changes in vivo. First, B. infantis decreased plasma IL-6 levels in mice (Desbonnet et al., 2008(Desbonnet et al., , 2010 and patients with inflammatory conditions (Groeger et al., 2013). In depressed mice, the IL-6 release also correlated positively with the severity of depression (Desbonnet et al., 2010). Second, B. infantis increased NE levels in the murine brainstem (Desbonnet et al., 2010) containing the majority of NE neurons (Schwarz and Luo, 2015). Therefore, B. infantis likely regulates plasma IL-6 and central NE system to exert an antidepressive effect.
Bifidobacterium infantis secretes bioactive factors (probably polysaccharides) that enhance transepithelial resistance of IECs (Ewaschuk et al., 2008). Other studies involving rodents also showed that B. infantis treatment enhanced the intestinal barrier by strengthening the formation of tight junction proteins and anti-inflammatory activities of immune cells (Lomasney et al., 2014;Zuo et al., 2014;Javed et al., 2016). Indeed, bacterial DNA translocation from the gut lumen into the circulation was reduced in B. infantis-fed rodents (Osman et al., 2006;Gómez-Hurtado et al., 2012).
Bacterial DNA is a potent inducer of TLRs which facilitate the release of proinflammatory cytokines, including IL-6 (Gutierrez et al., 2016). Administration of IL-6 induced depression in mice, and this outcome was prevented by pharmaceutical blockage of NE neurons in the brainstem (Kurosawa et al., 2016). Hence, B. infantis potentially modulates the NE system via an immune-mediated humoral route to reduce depression ( Figure 2D). This mechanism appears to be independent of the vagus nerve as oral gavage of B. infantis also decreased proinflammatory cytokine (including IL-6) levels in vagotomized mice with an inflamed colon (van der Kleij et al., 2008).

Butyrate, TRP, and Central 5-HT-BDNF System
Treatment of B. infantis upregulated mRNA expression of Tph1 in RIN14B cells, a cell line that mimics ECs (Tian et al., 2019). Tph1 converts TRP to 5-hydroxytryptophan (5-HTP) and aromatic amino acid decarboxylase subsequently converts 5-HTP to 5-HT. B. infantis-fed mice displayed reduced depressivelike behaviors, along with an increase in TRP biosynthesis and hippocampal 5-HT and 5-HTP levels. In the same study, B. infantis increased cecum butyrate levels and the abundance of butyrate-producing Bifidobacterium. The elevated butyrate levels also correlated with increased hippocampal 5-HTP and PFC BDNF levels (Tian et al., 2019). This could be due to the ability of butyrate and other SCFAs to increase Tph1 activity of ECs, thereby promoting 5-HTP and 5-HT secretions (Reigstad et al., 2015;Yano et al., 2015;Lund et al., 2018). This is consequential as ECs contribute about 95% of the bodily 5-HT (El-Merahbi et al., 2015), and that mice with a gut microbiota had 2.8-fold higher plasma 5-HT levels than GF mice (Wikoff et al., 2009). The evidence for the ability of 5-HT to cross the BBB is conflicting (Brust et al., 2000;Wakayama et al., 2002;Nakatani et al., 2008;El-Merahbi et al., 2015). In contrast, 5-HTP readily crosses the BBB and can be converted into 5-HT. Therapeutic 5-HTP has also been shown to treat clinical depression with a potency equivalent to or better than SSRIs (Birdsall, 1998;Jangid et al., 2013;Jacobsen et al., 2016).
Furthermore, B. infantis intake increased plasma TRP levels in healthy rats (Desbonnet et al., 2008), but another study with chronic-stressed rats reported otherwise (Desbonnet et al., 2010). The author then suggested that B. infantis regulates TRP metabolism differently, depending on the rat strain (Desbonnet et al., 2010). Therapeutic TRP can improve symptoms of mood, sleep and cognitive disorders as TRP readily passes through BBB to regulate numerous brain functions, such as 5-HT synthesis (Richard et al., 2009). The elevated plasma TRP levels from B. infantis intake is most likely a result of reduced proinflammatory cytokines (Desbonnet et al., 2008(Desbonnet et al., , 2010, which reduces IDO activity and prevents overcatabolism of TRP (Reus et al., 2015). Thus, B. infantis may upregulate the hippocampal 5-HT system via modulation of peripheral 5-HTP, 5-HT and/or TRP levels. As 5-HT promotes BDNF synthesis (Martinowich and Lu, 2008), this presumably explains the concomitant increase in BDNF levels in PFC of rats ameliorated of depression with B. infantis treatment (Tian et al., 2019). Taken together, L. helveticus, L. paracasei and B. infantis upregulate the central 5-HT-BDNF system as their mutual antidepressive mechanism, although via different pathways ( Figure 2B).

Clostridium butyricum
Treatment of C. butyricum improved depressive-like behaviors in chronic-stressed mice. These treated mice also showed upregulated central 5-HT, BDNF and GLP-1 receptors in the brain (Sun et al., 2018). Remarkably, the combination of C. butyricum with antidepressants reduced depression in about 70% of treatment-resistant MDD patients, of which 30% achieved remission (Miyaoka et al., 2018). These studies support the antidepressive efficacy of non-pathogenic C. butyricum. It should be noted that certain strains of C. butyricum are pathogenic which may cause botulism and necrotizing enterocolitis (Cassir et al., 2016).

Butyrate, Central 5-HT-BDNF System, and GLP-1
Clostridium butyricum, as a resident of healthy gut microbiota, produces butyrate from carbohydrate fermentation (Araki et al., 2002;He et al., 2005;Liu J. et al., 2015). Treatment of C. butyricum increased central 5-HT levels and BDNF expression in mice with reduced depression (Sun et al., 2018). Another study also reported that C. butyricum intake upregulated neurogenesis-related pathways, such as BDNF, via butyrate production in mice (Liu J. et al., 2015). Additionally, intragastric inoculation of C. butyricum increased intestinal secretion of GLP-1 and the central expression of GLP-1 receptors in mice alleviated from depression (Sun et al., 2018). This effect may also be mediated by butyrate as SCFAs can bind to receptors expressed on intestinal L cells to stimulate GLP-1 secretion into the bloodstream (Tolhurst et al., 2012). GLP-1 is known for appetite and glucose control, but the activation of central GLP-1 receptors has been shown to regulate the central 5-HT system and reduce anxiety-and depressivelike behaviors in rats (Anderberg et al., 2016). Therefore, antidepressive mechanism of C. butyricum potentially involves a butyrate-mediated upregulation of central BDNF-5-HT system (similar to L. paracasei and B. infantis) and GLP-1 receptor expression ( Figure 2B).

Lactobacillus kefiranofaciens
Lactobacillus kefiranofaciens is isolated from kefir, a type of fermented milk. Oral gavage of L. kefiranofaciens improved behaviors of chronic-stressed, depressed mice. These treated mice also showed several physiological alterations. Levels of circulating TRP, splenic IL-10 and beneficial gut bacteria (e.g., Lachnospiraceae, Bifidobacteriaceae, and Akkermansia) increased, and KYN/TRP ratio, splenic IL-6 and IFN-γ levels and Proteobacteria abundance decreased (Sun et al., 2019). What factors mediate such broad effects of L. kefiranofaciens on the TRP/KYN pathway, immune system, HPA axis and gut microbiota remain unclear, but exopolysaccharide is potentially a candidate ( Figure 2D).

Exopolysaccharide, Peripheral Inflammation, and Gut Microbiota
The only known metabolite of L. kefiranofaciens is an exopolysaccharide called kefiran (Maeda et al., 2004;Xing et al., 2017). The intake of kefiran modulated the gut mucosal immune system of mice (Vinderola et al., 2006), which could potentially account for changes in splenic cytokines seen in depressed mice (Sun et al., 2019). Kefiran was also shown to protect human enterocyte cell lines from adhesion and damage inflicted by toxins of pathogenic bacteria (Santos et al., 2003;Medrano et al., 2008). A further study discovered that L. kefiranofaciens produces a novel exopolysaccharide (not kefiran) that is bactericidal toward enteropathogens Listeria monocytogenes and Salmonella enteritidis (Jeong et al., 2017a). It may be possible that the antibacterial effects of this exopolysaccharide extend to other species in the gut microbiota. This supports the finding that L. kefiranofaciens supplementation ameliorated depressive-like behaviors in chronic-stressed mice by regulating gut microbiota content, which included the decreased abundance of Proteobacteria, a phylum that includes pathogens such as Salmonella (Sun et al., 2019). Other mice studies also supported the role of L. kefiranofaciens in modulating gut microbiota composition (Jeong et al., 2017b;Xing et al., 2018). Collectively, these changes in gut microbiota profile prevent gut dysbiosis that could lead to chronic inflammation, HPA axis overactivity and depression (Jeong et al., 2017b).

Bifidobacterium breve
Bifidobacterium breve treatment improved symptoms of depression in innately anxious mice (Savignac et al., 2014), chronic-stressed mice (Tian et al., 2019) and schizophrenic patients with depression (Okubo et al., 2019). B. breve supplementation also improved mood and cognition in elderly people with mild cognitive impairment . However, none of the accompanying physiological changes among these studies overlapped, making it difficult to identify an exact mechanism of B. breve. In spite of this, one study demonstrated that antidepressive mechanism of B. breve involves the generation of benzoic acid  Figure 2D).

Benzoic Acid and Central Glu System
Among the 18 bacterial strains isolated from gut microbiota, B. breve was the most efficient converter of albiflorin to benzoic acid via microbial carboxylesterase, at the rate of 75% as compared to L. casei, Lactobacillus acidophilus and B. longum at about 5%. The same study further showed that orally administered benzoic acid alleviated depression in mice . Benzoic acid readily crosses the intestinal barrier and BBB to inhibit D-amino acid oxidase that catabolizes D-serine, a co-agonist of N-methyl-D-aspartate receptor (NMDAR, a type of Glu receptor) . Both D-serine and NMDARs are therapeutic targets in neuropsychiatric disorders, such as depression, schizophrenia and cognitive impairment (Durrant and Heresco-Levy, 2014). Indeed, a dysfunctional Glu system is linked to the pathophysiology of depression (Pytka et al., 2016). In line with this, B. breve intake increased Glu synapses in chronic-stressed mice while treating its depressivelike behaviors (Tian et al., 2019).

Serpin, Central Neural Activity, and HPA Axis
Both in vitro and in vivo studies showed that B. longum weakened the excitability of murine myenteric neurons (Bercik et al., 2011;Khoshdel et al., 2013). Mice with inflamed intestines that were fed with B. longum demonstrated reduced anxiety-like behaviors, and this effect ceased upon vagotomy (Bercik et al., 2011). Intriguingly, B. longum intake also alleviated anxiety in coloninflamed mice that were vagotomized before treatment (Bercik et al., 2010). The author postulated that vagus afferents are an essential conduit when B. longum signals enterocytes, but not colonocytes (Bercik et al., 2011). The genome of B. longum encodes serpin, a serine protease inhibitor (Ivanov et al., 2006;Mkaouar et al., 2016). Serpin can inhibit the activation of enteric neurons by suppressing the secretion of elastase-like proteases from IECs (Ivanov et al., 2006;Buhner et al., 2018). These studies support the premise that B. longum interacts with the host via the neural pathway (similar to L. rhamnosus). Following this, the neural activity and HPA axis of the brain may be altered. Individuals consuming B. longum had increased neural activity in the PFC and decreased neural activity in the amygdala and fronto-limbic regions (Allen et al., 2016;Pinto-Sanchez et al., 2017). Anomalies in the anatomy and activity of the amygdala and PFC are also commonly observed among depressed patients (Liu W. et al., 2017). Furthermore, B. longum intake exerted simultaneous glucocorticoids-lowering and anxiolytic effects in humans and mice (Allen et al., 2016;Orikasa et al., 2016), suggesting that B. longum potentially modulates the HPA axis.

Lactobacillus gasseri
Supplementation of L. gasseri improved mood (Sashihara et al., 2013) and depressive-like symptoms  in stressed individuals. However, no studies have evaluated the effect of L. gasseri on clinically depressed individuals. Interestingly, L. gasseri is the only dietary probiotic which showed consistent sleep-enhancing effects in humans (Nishida et al., 2017a,b;Sawada et al., 2017). Irregular sleeping patterns are frequently associated with MDD (American Psychiatric Association, 2013; Wallace and Milev, 2017), supporting the use of L. gasseri as a potential treatment for MDD-related sleep disturbances.

Gassericins, Gut Microbiota, and Parasympathetic Activity in Sleep
Stressed individuals that were given probiotic-based milk containing either heat-killed or live L. paracasei showed alterations in the gut microbiota profile. Heat-killed L. gasseri decreased Bacteroides vulgatus and increased Dorea longicatena populations (Nishida et al., 2017a), whereas live L. gasseri decreased growth of inflammatory Enterobacteriaceae and Veillonella . Both studies also showed that L. gasseri enhanced sleep quality of participants. Another study reported that heat-killed L. gasseri (in milk) increased the population of Clostridium cluster IV group and SCFAs levels in individuals with altered bowel movements . Using a similar methodology, decreased Clostridium cluster IV and increased Bifidobacterium populations were found in another group of participants . Taken together, these results suggest that heat-killed L. gasseri does not have a specific microbial target, but rather modifies the preexisting gut microbiota that is unique to each individual. Nevertheless, these changes in the gut microbiota composition favor an anti-inflammatory state Sugawara et al., 2016;Nishida et al., 2017a). L. gasseri likely alters the gut microbiota profile through its unique, heat-resistant gassericins A and T with potent antibacterial properties against enteric pathogens (Pandey et al., 2013).
Heat-killed L. gasseri decreased expression of leukocytic stress-responsive microRNAs and salivary cortisol levels in stressed individuals (Nishida et al., 2017b). L. gasseri intake also prevented downregulation of EIF2-related genes in IBS patients (Nobutani et al., 2017). These studies suggest that L. gasseri confers protection against detrimental effects of stress. Moreover, heat-killed L. gasseri intake promoted parasympathetic nerve activity while improving sleep quality of stressed individuals (Nishida et al., 2017b). In healthy individuals, administration of either live or heat-killed L. gasseri increased their parasympathetic activity (Otomi et al., 2015;Sugawara et al., 2016). Therefore, L. gasseri may modify the gut microbiota profile in such a way that lowers gut inflammation and stress response, which may consequently promote parasympathetic activity to facilitate sleep ( Figure 2C).

CHALLENGES AND PERSPECTIVES FOR PROBIOTICS AS TREATMENT FOR DEPRESSION
The existence of different gut microbiota compositions, depression subtypes and probiotic formulations complicate treatment outcomes and necessitate an individualized approach when using probiotics to treat depression. Despite these challenges, probiotics confer some benefits over antidepressant drugs, and there are more promising candidate probiotics that can potentially treat depression.

Heterogeneity of Gut Microbiota Composition
Several factors are known to influence the gut microbiota composition, such as diet, medications, genetics, age, geographical location and smoking (Thursby and Juge, 2017). Recently, approximately 1000 gut-derived putative bacterial species that do not belong to any existing genus were discovered in humans (Almeida et al., 2019). Such tremendous diversity complicates the understanding of how introduced probiotics affect the overall gut microbiota. One study showed that tolerability of individuals' gut microbiota toward the colonization of probiotics ranges from permissive to resistant (Zmora et al., 2018). This appears to depend on the baseline abundance of probiotic species in the host gut microbiota. For instance, those who were permissive toward the colonization of Lactobacillus had prior low levels of Lactobacillus populations before treatment (Zmora et al., 2018). Similarly, B. longum colonized the gut for a longer period in 30% of users who initially had low levels of B. longum (Maldonado-Gomez et al., 2016). Another study showed that the antidepressive effect of multispecies probiotics (MSP) only manifests when the administered MSP successfully colonized the gut of rats (Abildgaard et al., 2019). This is consistent with the observation that lower levels of two main probiotic genera, Lactobacillus and Bifidobacterium, are commonly found in individuals with MDD (Aizawa et al., 2016).
Despite most studies supported the effectiveness of probiotic supplements in reducing depression, not all randomized controlled trials reported the same outcome ( Table 2). For instance, L. rhamnosus did not affect scores of anxiety, depressions, sleep, cognition, inflammatory and stress responses among healthy adults . L. rhamnosus also did not affect perceptions of wellbeing, anxiety and stress among healthy older adults (Ostlund-Lagerstrom et al., 2016). In healthy individuals, L. helveticus exhibited no antidepressive effect Ohsawa et al., 2018). These results imply that probiotics are less efficacious among the healthy population, which agree with a meta-analysis that reported an insignificant effect of probiotics on mood, particularly in healthy individuals (Ng et al., 2018). Therefore, probiotics could be generally more effective in colonizing gut microbiota of depressed individuals that are different from healthy people (Jiang et al., 2015;Zheng et al., 2016). In some cases, probiotic colonization may be optional for their effects to manifest. For instance, heat-killed L. paracasei benefited the human and animal host, in terms of neurochemical and behavioral changes (Corpuz et al., 2018;Murata et al., 2018;Wei et al., 2019). Some probiotics, such as L. reuteri, L. paracasei, L. plantarum, L. gasseri, L. kefiranofaciens, B. breve, and B. infantis, promoted the colonization of other beneficial microbes that contributed to the reduction of depressive-like symptoms in animals (Marin et al., 2017;Dhaliwal et al., 2018;Jang et al., 2019;Sun et al., 2019;Tian et al., 2019;Wei et al., 2019).

Heterogeneity of Depression
Major depressive disorder is characterized by depressed mood and/or anhedonia, in addition to excessive guilt, suicidal ideation, Frontiers in Neuroscience | www.frontiersin.org     (Ostergaard et al., 2011) and 19 depression subtypes (Harald and Gordon, 2012;Sharpley and Bitsika, 2014) can be derived. These subtypes of depression are often grouped as a single term, namely depression, which should not be the case when evaluating therapeutic potential of probiotics. Some associations can be drawn by matching behavioral benefits of probiotics to the characteristics of depression subtypes ( Table 2). For instance, the sucrose preference test in rodents reflects the anhedonia subtype (Dedic et al., 2011). Probiotics that have been shown to improve the outcome of this test include L. helveticus (Liang et al., 2015), L. plantarum (Liu Y.W. et al., 2016), L. paracasei (Wei et al., 2019), L. kefiranofaciens (Sun et al., 2019), B. infantis (Tian et al., 2019), and B. breve (Tian et al., 2019). Among these probiotics, L. plantarum (Liu Y.W. et al., 2016) and L. paracasei (Wei et al., 2019) also modulated the central DA system, whereas B. infantis and B. breve upregulated tyrosine (precursor to DA) biosynthesis (Tian et al., 2019). An impaired DA system represents the hallmark pathophysiology of anhedonia (Dunlop and Nemeroff, 2007). This provides a proof of concept that these probiotics may be effective in treating anhedonia.
Somatic depression subtype is characterized by psychomotor agitation/retardation (i.e., locomotion), changes in weight/appetite, insomnia/hypersomnia and fatigue without physical exertion (Sharpley and Bitsika, 2014). Probiotics that improved locomotor activity of rodents include L. plantarum (Liu W.H. et al., 2016;Dhaliwal et al., 2018), L. helveticus (Liang et al., 2015) and L. brevis (Miyazaki et al., 2014). Intake of L. brevis increased sleep duration in healthy mice (Miyazaki et al., 2014;Han et al., 2017), and L. gasseri enhanced sleep quality in medical students with mild depression (Nishida et al., 2017a,b). L. rhamnosus supplementation modulated appetite-associated genes and attenuated appetite in zebrafish (Falcinelli et al., 2016(Falcinelli et al., , 2017. In combination with prebiotics, L. rhamnosus exerted antidepressive effect and appetite control in obese individuals . Hence, symptoms of somatic depression are rather distinct and may be improved differently with different probiotics. Cognitive depression subtype is distinguished by poor concentration and memory function as well as indecisiveness (Sharpley and Bitsika, 2014). Behavioral assessments for memory function in mice include the Morris water maze, Barnes maze and other behavioral tests (Dedic et al., 2011). Administration of probiotics including L. helveticus (Ohland et al., 2013;Luo et al., 2014;Liang et al., 2015), L. plantarum (Dhaliwal et al., 2018), and L. paracasei (Corpuz et al., 2018;Huang et al., 2018) enabled animals to perform these memory test more effectively. Attention, memory and learning behaviors in humans are assessed by cognitive tests, such as the Stroop, verbal-learning and digit-symbol tests. Improvements in these tests have been shown with the intake of (1) L. helveticus Ohsawa et al., 2018) and B. longum (Allen et al., 2016) in healthy adults; (2) L. plantarum in MDD patients (Rudzki et al., 2019) and stressed adults with mild depression (Lew et al., 2018); and (3) B. breve in elderly with mild cognitive impairment . Thus, some probiotics appear to improve cognition regardless of depression.
Conventional SSRIs that target the 5-HT system often fail to treat anhedonic patients and, in some cases, worsen their symptoms (Dunlop and Nemeroff, 2007). Antidepressant drugs (e.g., SSRI and SNRI) are also ineffective against other depression subtypes, namely the somatic (Tylee and Gandhi, 2005), cognitive (Shilyansky et al., 2016) and anxious depression (Ionescu et al., 2014). Therefore, certain probiotics may serve as an adjuvant or alternative treatment for MDD and its subtypes. A pilot study showed that MSP, together with a magnesium supplement, decreased depression in SSRI treatment-resistant patients (Bambling et al., 2017). A clinical trial also reported that the combination of B. longum and L. helveticus decreased depression in MDD patients with prior use of standard antidepressants (Kazemi et al., 2019).

Single-Species and Multi-Species Probiotic
In studies that investigated behavioral effects of probiotics, about 60% of animal studies and 50% of human studies used single-species probiotics (SSPs) (Joseph and Law, 2019). Studies with SSPs promote a better understanding of the function and contribution of individual probiotic, which is difficult to measure in MSPs. However, MSPs may have higher potency in humans. In MDD patients, SSP (L. plantarum) did not reduce depression but improved cognition (Lew et al., 2018), whereas MSPs had repeatedly shown antidepressive efficacy (Akkasheh et al., 2016;Bambling et al., 2017;Ghorbani et al., 2018;Kazemi et al., 2019). MSPs often gave better therapeutic efficacy compared to that of SSPs in gut-related disorders and pathogen infections, which could be explained by an overall higher dosage (Chapman et al., 2011(Chapman et al., , 2012. Indeed, MSPs with a higher dosage improved symptoms of depression and anxiety in healthy individuals compared to that of a lower dosage (Tran et al., 2019). MSPs are also hypothesized to exhibit synergistic effects that would have an expanded effect on the host physiology (Chapman et al., 2012). In contrast, SSPs are speculated to promote better colonization as it does not have to compete for nutrient or adhesion sites in the host (Chapman et al., 2011). This highlights the need for more studies to understand how probiotics in MSPs interact with each other and with existing gut microbiota, and which probiotic(s) is suitable in formulation of MSPs for antidepressive efficacy.

Advantages of Probiotics as Antidepressive Treatment
Probiotics are generally safe for consumption, except for immune-compromised and critically sick individuals wherein probiotics may cause sepsis, pneumonia, endocarditis and allergies (Didari et al., 2014). Still, it has been viewed by some that more human trials are required to establish the dosage efficacy and long-term safety profile of probiotics (Kothari et al., 2018). For antidepressant drugs such as SSRIs, side effects occur in 40-60% of users which include sexual dysfunction, suicidality, emotional numbness and addiction (Read and Williams, 2018). A meta-analysis data showed that users of antidepressant drugs were associated with a 33% increased risk of mortality (Maslej et al., 2017). On the other hand, probiotics possess fewer side effects than antidepressant drugs. For instance, rats fed with L. brevis-fermented milk exhibited comparable antidepressive efficacy to fluoxetine-treated rats, but without side effects of fluoxetine (decreased appetite and weight loss) (Ko et al., 2013).
Antidepressant usage is also associated with stigma, such as being perceived as emotionally weak and dependent on drugs, which contributes to the disease severity and poor adherence to treatment (Castaldelli-Maia et al., 2011). In a survey study, 77% of depressed patients prefer to hide their use of antidepressant medication from others (Martinez et al., 2018). However, the prevalence of perceived stigma against antidepressants differs based on the population studied (Castaldelli-Maia et al., 2011). To this end, probiotics may help as an alternative treatment for depression, given that probiotics have not been associated with any perceived social stigma (Wallace and Milev, 2017).

Candidate Probiotics With Potential Antidepressive Effect
Bifidobacterium pseudocatenulatum is known for its regulation of obesity-related changes in metabolism and the immune system (Cano et al., 2013;Moya-Perez et al., 2014Sanchis-Chorda et al., 2018). B. pseudocatenulatum intake reversed diet-induced obesity, depression, high corticosterone and low hippocampal 5-HT levels in mice (Agusti et al., 2018). However, a high-fat diet model is meant to study the pathophysiology of obesity and type 2 diabetes (Winzell and Ahren, 2004;Wang and Liao, 2012). It is, thus, unclear if B. pseudocatenulatum would decrease depression in mice without obesity. Another study showed that anxiety-like behaviors diminished in chronic-stressed mice fed with B. pseudocatenulatum, but depressive-like behaviors were unevaluated (Moya-Perez et al., 2017). Therefore, further studies are required to determine whether B. pseudocatenulatum has an independent antidepressive effect.
Bacillus coagulans supplementation relieved symptoms of both IBS and depression in patients diagnosed with IBS and MDD. This clinical recovery is accompanied by a decrease in serum myeloperoxidase, an inflammatory marker (Majeed et al., 2018). However, patients might have experienced less depression as a result of reduced IBS symptoms. Interestingly, B. coagulans intake increased levels of circulating IL-10, fecal F. prausnitzii and SCFAs in older adults (Nyangale et al., 2014(Nyangale et al., , 2015. As F. prausnitzii and butyrate are associated with antidepressive properties (Hao et al., 2019), B. coagulans may also indirectly reduce depression and improve gut health.
Bifidobacterium bifidum and L. acidophilus were often included in the formulation of MSPs to treat depressive symptoms in patients with MDD (Akkasheh et al., 2016;Bambling et al., 2017;Ghorbani et al., 2018) and other health conditions, such as polycystic ovarian syndrome, multiple sclerosis and IBS (Kouchaki et al., 2017;Ostadmohammadi et al., 2019;Zhang et al., 2019). Surprisingly, B. bifidum and L. acidophilus have not been tested independently for its antidepressive effect. B. bifidum intake improved mood and reduced symptoms of abdominal pain, diarrhea and constipation in patients with gastrointestinal disorders (Urita et al., 2015). However, the mood elevation could be due to recovery of gastrointestinal symptoms rather than effect of probiotics solely. Both in vitro and in vivo models showed that L. acidophilus protects the intestinal barrier integrity by preventing pathogen adherence and release of proinflammatory cytokines (Chen et al., 2009;Justino et al., 2015;Alamdary et al., 2018;Lepine et al., 2018;Najarian et al., 2019). Taken together, B. bifidum and L. acidophilus potentially exhibit antidepressive effect and their direct influence on depression warrants further investigation.
Bacteroides fragilis has been proposed as a potential probiotic, although its pathogenicity needs to be taken into consideration. B. fragilis secretes polysaccharide A and expresses sphingolipids that benefit the host gut health and immune system (Troy and Kasper, 2010;Tan et al., 2019). Bacteroides genus is likely to be the largest GABA producer amongst human gut microbiota, with B. fragilis produces GABA at low pH. The same study also found that neural patterns of a typical MDD patient correlated with low fecal levels of Bacteroides (Strandwitz et al., 2019). Hence, antidepressive potential of B. pseudocatenulatum, B. coagulans, B. bifidum, L. acidophilus, and B. fragilis warrants further investigation. It is also worth noting that Bifidobacterium adolescentis's antidepressive capability may be a new probiotic candidate (Jang et al., 2019). Evidently, an increasing number of probiotics are being presented as a potential treatment for depression. This provides a wide repository of available probiotics, with different species combinations, that can be assessed for clinical efficacy against depression.

CONCLUSION
The MGB axis enables the bidirectional communication between the gut microbiota and the brain. When this axis becomes maladaptive, the host physiology is adversely affected which may lead to the development of depression. Probiotics have shown clinical efficacy in the treatment of depression by modulating the MGB axis. Yet, the complexity of gut microbiota and heterogeneity of depression presents a challenge to explain the underlying mechanisms that contribute to this clinical efficacy. Nonetheless, cumulative evidence suggests the therapeutic potential of probiotics for certain depression subtypes, with fewer side effects and less stigma compared to standard antidepressants.
Limitations of this review include: (1) inferences of probiotic mechanisms were derived from preclinical and in vitro data; (2) interactions of probiotics with other members of gut microbiota were unexplored, therefore the mechanisms of MSPs was unable to be explored; (3) strain-specific effects of bacterial species were neglected; (4) potential applications for probiotics for depression subtypes are hypothesized, however, clinical evidence is limited; (5) effect sizes of probiotics as antidepressants was not evaluated. Notwithstanding these caveats, this review adds further understanding to the potential antidepressive effects and therapeutic potentials of probiotics. Venema (2017) stated that it is imperative to grasp the underlying molecular mechanisms of the MGB axis, and which microbial populations are pertinent for this intervention, to advance the marketability of probiotics.

AUTHOR CONTRIBUTIONS
SY wrote and edited the manuscript. TT, JC, and WL conceptualized and edited the manuscript. All authors contributed to the intellectual input and critical revision of the manuscript.