Microbes, oxytocin and stress: Converging players regulating eating behavior

Oxytocin is a peptide-hormone extensively studied for its multifaceted biological functions and has recently gained attention for its role in eating behavior, through its action as an anorexigenic neuropeptide. Moreover, the gut microbiota is involved in oxytocinergic signaling through the brain-gut axis, specifically in the regulation of social behavior. The gut microbiota is also implicated in appetite regulation and is postulated to play a role in central regulation of hedonic eating. In this review, we provide an overview on oxytocin and its individual links with the microbiome, the homeostatic and non-homeostatic regulation of eating behavior as well as social behavior and stress.


| INTRODUCTION
2][3] This peptide hormone has been implicated in a broad range of biological functions that are mediated through either central (sociability, bonding, stress response) or peripheral (contraction of the uterus, intestinal contractility and secretion, visceral pain) Oxt receptors (OXTRs). 4,55][6] More recently, increasing evidence suggests that Oxt also is a player in the central processing of reward. 4,7Oxt administration has been shown to modulate the preference of food types, and food pleasure in preclinical and clinical studies. 4,8,9t has also been investigated for its potential interaction with the human microbiota. 102][13] For instance, Oxt administration (10 IU SC/animal/day, 14 days), induced changes in the gut microbiota composition in rats and was associated with overall improvement in stress, and anxiety/depression-like behaviors. 11Noteworthy, the probiotic Limosilactobacillus reuteri (formerly known as Lactobacillus reuteri) has been shown to not only increase circulating Oxt levels 12 but also Oxt expression in the brain 13 in rodents, effects which have been linked with metabolism, 14,15 wound healing, 12 sociability and behavior. 14,16,17Whilst the effects of Limosilactobacillus reuteri have been reviewed elsewhere, 16 recently other bacterial strains such as Blautia stercoris MRx006 and bacteria-derived metabolites, such as 4-ethylphenylsulfate (4EPS), have also demonstrated ability to modify social and sensorimotor behavior and hypothalamic Oxt expression in rodents. 18,19Furthermore, a recent randomized controlled trial performed by Narmaki et al. 20 showed that probiotic administration (multistrain probiotic ≈10 10 CFU per day, 12 weeks) increased Oxt serum levels in obese women with food addiction. 20e supplementation also resulted in a significant improvement in eating behavior, increasing cognitive resistance towards food intake.
Thus, the gut microbiota and Oxt signaling have independently been linked with food intake and eating behavior, which begs the question whether the microbiota can modulate homeostatic and/or hedonic aspects of food intake (e.g., food reward and food motivated behavior) via the Oxt system.
In this review, we give an overview of the Oxt neuropeptide and its physiological functions, with a particular focus on social behavior, stress and eating behavior.We highlight novel findings relating Oxt, the microbiome and eating behavior suggesting that microbiometargeting interventions might modulate Oxt signaling and functionality, thereby affecting both homeostatic and non-homeostatic food intake.We also discuss the influence of the gut microbiome and Oxt signaling on eating behavior through their converging effect on social behavior and stress.

| OXYTOCIN AND OXYTOCIN RECEPTOR 2.1 | Oxytocin expression
The neuropeptide Oxt is encoded by the gene Oxt oxytocin/ neurophysin I prepropeptide (Gene ID: 5020) which is located in the same chromosomal locus as the vasopressin (Avp) gene (dAVPneurophysin II (NPII)) across all the species, both being adjacent paralogous genes in the chromosome 20p13 distanced by 12 kb of intergenic region and transcribed from opposite strands. 21The Oxt gene comprises two introns and three exons, of which exon 1 contains the sequence for the final OXT nonapeptide (further described in Figure 1). 21,26The rest of the gene encodes for a signal peptide and neurophysin I (also Oxt prepropeptide; 10-kDa and 125 amino acids length peptide) which serves as a carrier protein and is secreted along with Oxt.In the brain, Oxt synthesis occurs predominantly in the paraventricular nucleus (PVN) and the supraoptic nucleus (SON) of the hypothalamus as well as the medial preoptic area, and substructures of dorsomedial hypothalamic nucleus. 2,3,27As part of the hypothalamo-neurohypophysial system, Oxt vesicles produced in magnocellular oxytocinergic neurons are released into neurohypophysial capillaries from where it enters the bloodstream.
Oxytocinergic neurons have been found in the forebrain and other brain regions 2,3,28 but it is not clear yet whether Oxt is released in these areas and if so, what the release mechanisms are (via dendrites, synapses or in a non-synaptic mode 3,28,29 ).
In addition to its expression in the brain, there is evidence for expression of Oxt in peripheral organs, in particular the uterus, corpus luteum, placenta, amnion, testes (interstitial cells), heart, adrenal glands, dermis (fibroblasts and keratinocytes), thymus, and the gastrointestinal tract. 30The presence of these peripheral Oxt-producing cells, along with Oxtr (see Section 2.2), suggests that there are local circuits that induce local Oxt secretion independent of the hypothalamus.In contrast to central Oxt release which is induced by physiological stimuli such as suckling, parturition, or stress, 1,2 it is currently unknown how Oxt secretion from peripheral oxytocinergic cells is triggered.

| Oxytocin receptor
Central and peripheral effects of Oxt are mediated by the OXTR.The OXTR is encoded by the Oxt receptor gene (Oxtr) (Gene ID: 5021), which is expressed both peripherally 2 and centrally.
The OXTR belongs to the Class I G-protein-coupled receptor (GPCRs) and as such is comprised of seven hydrophobic transmembrane helices.Agonists, such as Oxt, bind a cavity of the OXTR between the α-helical segments of the transmembrane domains 3 and 6 also with involvement of residues within extracellular domains. 22fferent binding sites have been shown between endogenous and synthetic Oxt agonists and antagonists. 22Studies using chimeric OXTR receptors suggested that Oxt antagonists may bind to transmembrane helices 1, 2, and 7, regions that are not involved in the Oxt binding. 31Moreover, it has been shown that Oxt binds and activates different OXTR GPCR subtypes depending on the neuropeptide concentration, for instance G αq/11 -signallng is activated at low Oxt concentrations whereas G oB requires higher concentrations (Figure 2, top section). 32Depending on the receptor subtype, activation of the OXTR leads to the initiation of one of three different signaling pathways: 2,28,[32][33][34] the G αq/11 /PLC/InsP3 pathway, the G βγ /Raf/MEK/ ERK1/2 and the G αi -G αs /AC/PKA (Figure 2).Although the precise OXTR signaling pathways in subpopulations of hypothalamic neurons, especially those regulating eating behavior, are still unknown, it is suggested that OXTR activation in the PVN involves the ERK and MAP kinase pathways (Figure 2). 35Inhibition of this pathway prior to injection of Oxt abolished its anxiolytic effect in male rats. 35 addition to OXTRs, Oxt also binds to the other GPCRs, including the Avp receptors V1aR, V1bR, and V2R, though with lower affinity than to OXTR. 36ntral Oxtr expression levels vary according to the organism, 37 the developmental stage 38 and sex. 39The hypothalamus is a brain region with high Oxtr expression in both humans and mice (2-monthold) whereas high levels of hippocampal and cortical Oxtr expression have only been demonstrated in mice. 37In mice, hippocampal OXTR protein expression is higher in embryonic and juvenile stages whereas F I G U R E 1 Representative scheme of the oxytocin (Oxt) and vasopressin (Avp) genes on the human chromosome 20 (chromosomal location 20p13) in addition to the Oxt biosynthesis and further modifications.Oxt and Avp genes are transcribed from opposite strands, and both are based on two introns and three exons.In the case of Avp, the first exon encodes different peptides (the AVP peptide, a signal peptide (SP), and the first residues of neurophysin (NP) peptide), the second encodes the central part of the NP peptide and finally, the third encodes a glycopeptide (GP) and the C-terminal region of NP.In the case of Oxt, the first exon encodes different peptides (the Oxt hormone, SP, and the first 9 residues of NP), the second exon encodes the central part of the NP and finally, the third one encodes the C-terminal region of NP.After the nuclear transcription of the genes, both Avp and Oxt prepropeptides are translated in the ribosomes of the rough endoplasmic reticulum (RER).The generated protein, we focus on Oxt, is transferred into the Golgi apparatus where concentration, packaging into neurosecretory vesicles and post-translational processing take place, which includes the action of a signal peptidase to release SP, and prohormone convertases that breaks down the peptide into NP-I and Oxt-Gly-Lys-Arg which generates Oxt after the action of a Carboxypeptidase E (CPE), α-amdating monooxygenase (PAM) and placental leucine aminopeptidase (PLAP) that cleave Arginine (Arg), Lysine (Lys) and Glycine, respectively.After the intravesicular post-translational process, Oxt is generated and the vesicles containing the neuropeptide are secreted from neuronal terminals into neurohypophysial capillaries and into either the peripheral bloodstream or projecting to other brain regions.In the brain, Oxt synthesis occurs in the magnocellular hypothalamo-neurohypophysial system, which involves mainly the hypothalamic supraoptic nucleus (SON) and paraventricular nucleus (PVN).][24][25] cortical OXTR expression increases during adulthood possibly relating to the relative importance of social and spatial memory and social decision making, respectively. 38In addition, sex differences regarding OXTR in the hypothalamus and the posterior bed nucleus of the stria terminalis (integration of limbic and autonomic system) among others have been reported in both mice and rats. 38,39ilst there are two predominant OXTR mRNAs present in the brain (6.7 and 4.8 kb), 40 truncated versions have been found in peripheral tissues such as the uterus (2.9 kb) and ovaries (1.2 kp). 41ether and how these genetic polymorphisms and epigenetic mechanisms affect the downstream signaling is incompletely understood. 42,43However, in uterine myocytes, Oxt activates Gα q/11 and CAMK/MLCK downstream signaling to promote uterine contraction 44 whereas OXTR activation in endothelial cells induces Gα q/11 /phosphoinositide 3-kinase (PI3K)/Protein kinase B (AKT)/endothelial nitric oxide synthase (eNOS) signaling to enhance cardiomyogenesis (Figure 2). 44,45Peripheral expression of OXTR is highest in female reproductive organs, 37 and it is also expressed in the enteric nervous system of the gastrointestinal (GI) tract.Specifically, OXTR is expressed in neurons of the intrinsic nervous system of the GI tract, and in sensory dorsal root ganglia and nodose ganglia neurons, where it has important gut-brain axis functions for gastrointestinal motility, inflammation, and visceral sensitivity. 46,47Whilst the existence of oxytocinergic neurons in the gastrointestinal tract has been appreciated for a while, further investigation is now warranted, particularly in the context of OXTR modulation by the microbiome.

| Modulation of oxytocin by the microbiome
The gut microbiota includes trillions of microbes including bacteria, archaea, viruses, and fungi that inhabit the gastrointestinal (GI) tract. 48[51][52] Microbiota status and probiotics have been shown to affect Oxt levels, Oxt and Oxtr expression (Table 1).For example, mice treated with antibiotics for >1 month displayed decreased Oxtr expression levels in the frontal cortex and a trend for increased Oxt expression in the hypothalamus.In contrast, there were no significant differences between conventional (CONV) and germ-free (GF, animals raised without a microbiota) animals. 54Another study, however, reported an increase of OXTR protein levels in specific cortical regions from male, but not female, GF rats. 55To date, to the best of our knowledge, it is not known how microbial depletion affects blood Oxt levels.However, treating mice with the probiotic L. reuteri MM4-1A or humans with a multispecies-probiotic (3 Bifidobacteria and 3 Lactobacilli) increased plasma Oxt levels. 20 with many microbiome-targeting interventions, the mechanism by which the microbiota affects Oxt expression in the brain is currently unclear.It is likely that it involves either bacterial metabolites in the circulation or vagal sensing of bacteria/bacterial products in the periphery leading to activation of oxytocinergic neurons in the brain, particularly the hypothalamus. 17These bacterial metabolites could also modulate the systemic immune system either directly or via gut-brain mediated pathways (i.e., neuroendocrine, vagal, immune) that affect central immunity and subsequent Oxt signaling. 16Alternatively, bacteria can modulate the secretion of neurotransmitters and peptides from hormone-releasing cells in the intestinal epithelium (serotonin, cholecystokinin, glucagon-like peptide, ghrelin) and the enteric nervous system that regulate gut-brain signaling 60 as well as intestinal physiology. 52,61This, in turn, affects the availability and absorption of nutrients, microbiota composition and thus, gut-brain communication. 62In addition to affecting central oxytocinergic neurons, the microbiota may also modulate enteric neurons that express Oxt or intestinal Oxtr. 63These neurons are strategically located in the submucous and myenteric plexus of the enteric nervous system and administration of Oxt indeed modulates intestinal ion transport and motility. 63,64Additionally, Oxtr knock-out mice develop more severe colitis compared to control animals and Oxt can reduce intestinal inflammation. 63,65Because gut-innervating sensory neurons also express Oxtr, 47 Oxt derived from enteric neurons is likely to play a role in regulating central Oxt release.The modulation of peripheral (gastrointestinal) Oxt and OXTR by the microbiota may directly contribute changes in central Oxt signaling or alternative changes in the brain via modulated gut-brain axis signaling, which warrants further investigation.
F I G U R E 2 Scheme with G-protein coupled receptors (GPCRs) subtypes to which Oxt binds according to its concentration and signaling pathways of oxytocin (Oxt) and the respective effector actions (green box) via gene protein expression.Oxt binding to OXTR Gαq/11 subtype activates phospholipase C (PLC), which cleaves phosphatidylinositol 4,5-bisphosphate (PIP2) into diacylglycerol (DAG) and inositoltrisphosphate (InsP3).Ca2+ is then released from the Ca2+ stores in rough endoplasmic reticulum (RER) via the binding of InsP3 to its receptor (InsP3R) or ryanodine receptor (RyR) in certain cells.DAG activates protein kinase C (PKC), which can activate in turn either Rho-mediated Rho kinase.PKC can also activate eukaryotic elongation factor (eEF2), which triggers multiple cellular responses such as protein synthesis in myometrial cells.Gαq/11 activation can cause membrane depolarization as well via activation for instance of voltage-gated Ca2+ channels (VGCCs), which mediate the Ca2+ signaling.When the intracellular Ca2+ ([Ca2+]i) is increased, Ca2+ binds to the protein Calmodulin, which then induces the stimulation of calmodulin kinase (CaMK), activating finally diverse enzymes such as nitric oxide synthase (NOS) or phosphoinositide 3-kinase (PI3K) linked with cellular responses with peripheral Oxt actions, including cardiovascular, bone or muscle trophic effects.Additionally, Gαqmediated OXTR activation can induce the transactivation of the epidermal growth factor receptor (EGFR; a receptor tyrosine kinase or RTK) which activates then the mitogen-activated protein kinase (MAPK) pathway mediated by MEK1/2.The complex Ca2+-CaM can also activate myosin light-chain kinase (MLCK), which induces myometrial contraction.Finally, Gαq-mediated OXTR activation can also activate PI3K/protein kinase B (AKT)/endothelial nitric oxide synthase (eNOS)/mammalian target of rapamycin (mTOR) pathway in endothelial cells to enhance cardiomyogenesis.Another pathway that is activated after ligand-binding to OXTR-mediated opening of Ca2+ channels involve free Gβγ subunits.This can activate afterwards Ras, Raf-1, MEK1/2 or ERK1/2, which then can activate certain transcription factors, such as COX-2, or cfos, linked with Oxt biological functions including regulation of prostaglandin secretion, labor and parturition, social behavior, and vascular function, among others.Gαi and Gαs can activate both the adenylate cyclase (AC) pathway.AC when activated transforms adenosine triphosphate (ATP) to cyclic adenosine monophosphate (cAMP), second messenger which is converted to protein kinase A (PKA), inducing then cellular responses including cardiovascular and reproductive function.Finally, Oxt can also bind and act through transient receptor potential vanilloid (TRPV), which for instance can activate MAPK1/2 signaling, inducing in last term actions such as anxiolytic effects.Overall, these cascades transduce in downstream signaling of diverse transcription factors, such as cAMP response element-binding protein (CREB), myocyte enhancer factor-2 (MEF2), transcriptional coactivator CREB-regulated transcription coactivator (CRTC) to induce Oxt-mediated actions, such as anxiolytic effects.The question mark represents the knowledge gap on the Oxt signaling involved in the regulation of eating behavior.][34] T A B L E 1 Studies investigating the effects of microbiota on central and peripheral oxytocin neuropeptide levels.No significant changes in serum Oxt levels in the 3 treatments compared to placebo groups.
Improvements in ASDlike behaviors.

| Challenges with measuring oxytocin
As has been introduced, Oxt is produced both in the brain and the periphery.It is currently unknown whether plasma Oxt levels correlate with central Oxt release and whether peripheral administration can increase central Oxt levels because Oxt cannot readily cross the blood-brain barrier (BBB). 66Recently, it was suggested that Oxt may reach the brain through transportation via the receptor for advanced glycation end products (RAGE) which is expressed in endothelial cells at the BBB. 67,68RAGE receptors have been shown to be crucial for the effect of exogenous Oxt on social behavior, such as bonding and parenting. 67,680][71][72] Furthermore, Oxt has a short half-life (around 4-6 min), 73,74 is broken down by tissue proteases and binds to plasma proteins which makes the quantification of Oxt levels, particularly in plasma, challenging.Thus, the field would benefit from novel detection methods enabling accurately determine how (microbiome-targeting) interventions modify Oxt levels across the gut-brain axis.

| PHYSIOLOGICAL FUNCTIONS OF OXYTOCIN AND EFFECTS OF THE MICROBIOTA
Oxt was originally discovered for its effects on peripheral tissues, most importantly female reproductive organs, where it induces uterus contractions during labor, and mammary glands, where it contributes to lactation.Importantly, Oxt also affects the function of adipocytes, 6,75 pancreatic β cells 76 and the GI tract. 63,77These all produce hormones and other nutrient-related signals that influence food intake.
Since the late 1970s, Oxt has also been investigated as a neurotransmitter or neuromodulator with important effects on reproductive behavior. 78For example, Oxt has been shown to play a role in maternal attachment, 79 the onset of maternal behavior, 80 sexual responsiveness 81,82 and mating behaviors. 82,83Further, Oxt modulates memory functions, 84,85 and olfactory sensory perception. 86The mechanism of these physiological actions is thought to involve interactions between the central oxytocinergic, dopaminergic, serotonergic [87][88][89] and the endogenous opioid system. 90However, further research is needed to further understand the precise underlying signaling pathways.
The focus of the review is to discuss the potential impact of the microbiota on Oxt's role in the central regulation of food intake behavior (Section 4), but microbiota-mediated changes in Oxt in the context of sociability (Section 3.1) and stress (Section 3.2) also need to be considered, as they can subsequently lead to altered homeostatic and hedonic eating.For example, stress increases food intake particularly of sweet carbohydrate-rich foods [91][92][93] and food intake increases when consumed in a group environment. 94,95The role of the microbiome in the neurobiology of social behavior and social stress has been reviewed in detail elsewhere. 96We focus on the potential link with Oxt and the relevance of the gut microbiota-oxytocin-behavior triad in modulating eating behavior.

| Oxytocin, the microbiome, and social behavior
1][102] For instance, intranasal Oxt is associated with antiaggressive behavior in male rats after acute and repeated administration.These animals displayed increased sociability toward an unfamiliar intruder and reduced offensive aggression.In addition to which they had strengthened social bonding towards a female partner. 103Moreover, Oxt also influences reward processes such as social reward and food reward mechanisms. 4,5,104,105Recent preclinical studies demonstrated that Oxt interacts with numerous brain regions involved in sociability, including the orbital frontal cortex, anterior cingulate cortex, and prefrontal cortex (PFC) to exert prosocial effects, and regions involved in the mesocorticolimbic dopamine system (i.e., the ventral tegmental area (VTA), substantia nigra, striatum) (see Figure 3). 7,106,107Specifically, Oxt has been shown to activate VTA neurons leading to increased dopaminergic activity in the mesocorticolimbic system. 7Furthermore, Oxt release in VTA appears to be necessary in order to evoke social reward. 7Oxt also displays potent anxiolytic effects. 107,108It is suggested that this is mediated by modification of serotonergic signaling. 108However, despite the overwhelmingly prosocial effects documented in these studies, other reports suggest that the action of Oxt is dependent upon a prosocial environment during administration and there are a lot of unknowns regarding factors that affect an individual's response. 109e microbiota has been shown to affect social behaviors and overall mental health in preclinical and clinical studies in rodents and humans which has led to the term "psychobiotics". 110,111Although many of these studies have not specifically investigated the involvement of Oxt (see all the studies found linking microbiota with Oxt in Table 1), one study suggests that antibiotic-treated mice show reduced anxiety-like behavior and a concomitant increased OXT protein expression levels in the brain 56 (further details in Table 1).Additional studies have investigated the possible contribution of Oxt to the beneficial effects of microbiome-targeting interventions in the context of social behavior, most of which were performed in disease models (Table 1).Recent studies highlight that a dysregulation of Oxt and single nucleotide polymorphisms (SNPs) or epigenetic modification of the Oxtr gene contribute to the neurobiology and pathophysiology of certain neuropsychiatric disorders, including autism spectrum disorder (ASD), 17,112,113 schizophrenia, 114,115 anxiety disorders, 116 depression, 117 obsessive compulsive disorder 118 and eating disorders. 119Moreover, microbiome-targeting interventions may be able to correct these deficits. 120For instance, certain studies suggested that patients with ASD have significantly lower Oxt circulating levels compared to a control population, 113,121 even though opposite results have also been described. 122,123Oxt administration, in turn, has been shown to reduce autistic-like behaviors 124 ; although the data are not consistent in human trials. 125At the level of the brain, Oxt administration enhances functional connectivity between the ventral PFC and striatum, 126 which suggests that the mesocorticolimbic system is potentially involved in this Oxt-mediated effects on ASD.Microbiome-targeting strategies such as administration of Blautia stercoris MRx0006 ameliorate autistic-like behavior and increase OXT protein levels and Oxt expression in the hypothalamus, in genetic mouse models of ASD (Shank3B À/À ) 19 (see Table 1).Administration of L. reuteri MM4-1A to Shank3B À/À mice restored social interaction and VTA plasticity, effects which were prevented in mice lacking the Oxtr in dopaminergic neurons. 17In humans, administration of probiotic Lactiplantibacillus plantarum PS128 and Oxt had a synergistic positive effect on social and behavioral scores in ASD patients which was associated with changes in the gut microbiota composition 59 (further details in Table 1).A microbiota-Oxt link has also been suggested in schizophrenia, which is another disorder where sociability is affected and where microbiome-targeting interventions have shown benefits.
It has been shown that Oxt levels in the cerebrospinal fluid are higher in patients with schizophrenia compared to controls. 127In contrast, serum Oxt levels appear to be lower, indicating that peripheral and central levels of Oxt may be differentially regulated and have divergent effects which need to be further investigated. 127Increased Oxt levels, however, are thought to improve schizophrenia behaviors and symptoms. 128Preliminary data suggests that probiotic administration may improve symptomology and comorbidity in patients with schizophrenia, however, the extent to which such microbiota-targeted strategies involve Oxt remains to be investigated. 129,130cordingly, microbiota seems to impact host Oxt and Oxt signaling, but recent evidence points towards a modulation of the gut microbiota by this neuropeptide.A recent study shows alterations in gut microbiota composition after Oxt administration.In the study by Dangoor et al. 11 Oxt administration led to altered gut microbiota diversity, which was linked with a decrease in anxiety-and depression-like behaviors in a corticosterone (CORT)-induced stress rat model. 112][133][134] Stress not only impacts social or anxietylike behaviors, which is relevant in the aforementioned disorders, 135,136 but also plays a key role in the regulation of food intake behaviors 91,[137][138][139] (see studies on stress, food intake and their link to Oxt in Section 4.3).
F I G U R E 3 Overview of the brain regions involved in oxytocin's effects on eating behavior, social behavior and stress and potential interaction of the gut microbiota as a key player.Oxytocin (Oxt) is a neuropeptide that is produced in magnocellular neurons (MCN) located in the paraventricular nucleus (PVN) and supraoptic nucleus (SON) of the hypothalamus (HYP) as well as in parvocellular neurons (PCN) located in the PVN (blue box).After Oxt synthesis in MCN within the PVN and PON, the neuropeptide is transported along axons to the posterior pituitary gland (brown, blue arrows), from where is released into the bloodstream.The brain regions with oxytocinergic signaling involved in regulating eating behavior are shown in the yellow circle and blue box, social behavior in the red circle and stress in the green circle.In the HYP (blue box), there is a complex interaction of Oxt-producing and Oxt receptors (OXTR) expressing cells in different nuclei.Glutamatergic neurons (GLU) in the arcuate nucleus (ARC, blue region) express OXTR and project to neurons in the PVN (orange arrow) to regulate Oxt secretion and enhance satiation through modulation of proopiomelanocortin (POMC) neurons.Oxt from PVN and SON also modulate the activity of these OXTRexpressing POMC neurons in the ARC (green arrows), which also induces satiation.Oxt released from MCN in the SON can also regulate Oxt secretion in the SON itself and in the PVN in an autocrine manner.Oxytocinergic neurons from the PVN (pink lines) project to many other brain regions (pink arrows) that are involved in the control of eating behavior including the nucleus tractus solitarius (NTS), the ventral tegmental area (VTA), the hippocampus (HPC), the amygdala (AMY), the nucleus accumbens (NAcc), and the prefrontal cortex (PFC).These brain regions with Oxt neurocircuitry have been also shown to be involved in the control of stress (green circle) and social behaviors (pink/red circle).As discussed in the review, the gut microbiota (purple circle) has been linked (blue arrows) with stress, eating behavior and Oxt independently, but also with Oxt particularly in social behavior context (the three effects are represented within a pyramid).The vagus nerve (in orange) connects the gut and the brain via gut-brain axis and is one mechanism by which gut microbiota is suggested to regulate appetite, eating and social behavior and stress.The question mark shows the brain regions suggested to be involved in oxytocinergic neurocircuitry-linked to stress regulation but that are not demonstrated yet.
Therefore, it is likely that Oxt-mediated effects on stress responses will have converging effects on sociability, anxiety as well as eating behavior.

| Oxytocin, the microbiome, and stress
Stress is defined as physiological response to a potentially harmful stimulus and leads to the activation of the hypothalamus-pituitaryadrenal axis (HPA) and the sympathetic nervous system.High Oxt levels are associated with an attenuated stress response, [140][141][142] which may protect against the detrimental effects of stress and signify an improved stress-resilience.It is suggested that Oxt release within the hypothalamus directly modulates corticotropin-releasing hormone (CRH) expression leading to reduced cortisol levels and the activity of presynaptic neurons that project to sympathetic efferents in the medulla. 143,144These regulate sympathetic output 143,144 and Oxt has been shown to reduce heart rate and blood pressure. 145,146In addition, Oxt might also reduce the sensitivity to stressful stimuli by reducing amygdala activity. 147Moreover, chronic ICV Oxt injection at low dose (1 ng/h for 15 days) exerted protective effects against stress and anxiety, whereas high dose (10 ng/h) was detrimental with anxiogenic effects in mice. 148e composition of the microbiome modulates the response to acute 149,150 and chronic stress. 151GF mice showed hyperresponsiveness in the HPA axis after restraint stress, which was accompanied elevated serum levels of corticosterone. 152,153On the contrary, microbiome-targeting interventions have been shown to improve the stress response.For example, administration of the probiotic L. plantarum 299v 154 decreases stress-induced cortisol levels in humans.In addition, a reduction in cortisol awakening responses was observed in healthy obese adults following administration of the probiotic B. longum APC1472. 155Furthermore, B. longum APC1472 also reduced the plasma corticosterone levels in high-fat diet (HFD) mice. 155The contribution of Oxt signaling has been studied in a limited number some of these studies.Desbonnet et al. (2015) found that although stress-induced cortisol levels were not different between antibiotic-treated and control animals, Oxt levels were higher in the stressed control group compared to stressed antibiotic-treated animals suggesting that an intact microbiota is required to trigger stressinduced Oxt expression. 53It is tempting to speculate that modulation of Oxt levels also contributes to the beneficial effects of probiotics on stress responses (Figure 3, green circle).

| EATING BEHAVIOR, OXYTOCIN AND THE MICROBIOME
7][158] However, the complex neurobiology and neurocircuitry that is involved in the full control of eating behavior can be classified according to the homeostatic and non-homeostatic regulation. 159,160meostatic regulation comprises the control of metabolic and energy status, mainly via the hypothalamus and brainstem nuclei (Section 4.1) whereas non-homeostatic regulation involves motivation, hedonic eating learning, memory, and cognitive aspects of feeding, which are controlled by high-order brain regions, including the hippocampus, the mesolimbic system, and the frontal cortex (Section 4.2). 156,158Even though this classification is made, the same peripheral signals, such as gut hormones and microbiota-derived metabolites, can converge on both systems, for instance via the vagus nerve signaling, blood circulation or other direct and indirect brain signaling mechanism. 157,1612][93] For example, consumption of highly palatable food increases CORT levels and stress increases food cravings. 138,162Thus, in addition to affecting food intake by itself, Oxt and the microbiota may also contribute in part by modifying social behavior and stress levels (see Section 3.1 and 3.2).

| Oxytocin and homeostatic appetite regulation: A role for the microbiome?
The hypothalamus is the main brain region involved in the homeostatic control of food intake.The activity of orexigenic AgRP/NPY and anorexigenic POMC neurons is regulated by signals from the periphery.This includes signals from the brainstem, specifically the nucleus tractus solitarius (NTS), which receives information about the status of the periphery via the vagus nerve and spinal afferents, as well as circulating metabolites and hormones (leptin, Glp-1, ghrelin, CCK). 5,156,157,161,163Importantly, there is a bidirectional communication between hypothalamic centers regulating food intake (ARC) and Oxt-producing parvo-and magnocellular cells in the PVN (Figure 3, blue box).Not only will Oxt secretion modulate the activity of ARC neurons, but POMC and AgRP/NPY neurons also increase and decrease Oxt secretion, respectively. 164,165t's role in regulating food intake is incompletely understood.Food intake stimulates oxytocinergic neurons, which would suggest an anorexigenic mode of action for Oxt.Indeed, many preclinical studies have demonstrated that ICV as well as IP injection of Oxt reduced food intake in different animal models, such as rats, mice, and monkeys (Supporting Information, Table 1). 166,167This is associated with an increase in c-Fosexpression in cells in the NTS, 168 and a vagus-dependent activation of c-Fos in hypothalamic Oxt neurons at a molecular level. 169In contrast, Oxt-KO animals display an increased preference for palatable and sugary food. 9,170However, in humans, high Oxt levels are observed in obese individuals 6,171 and intranasal Oxt administration has shown conflicting results with regard to food intake 4,8,105,[172][173][174] (see further details in Supporting Information, Table 1).An additional mechanism by which Oxt might regulate food intake behavior, includes the modification of signaling pathways from other food intake-regulating hormones.Our laboratory has shown that an OXTR-ghrelin receptor (GHSR) heterocomplex reduces GHSR receptor signaling, but it is currently unknown whether this modulates food intake. 1753][184] GF mice consume more food than their CONV counterparts which is associated with higher expression of orexigenic Npy and Agrp and lower expression of anorexigenic Pomc in the hypothalamus. 181In the brainstem, expression of Bdnf and Glp-1 (both considered as anorexigenic neuropeptides) are higher in GF compared to CONV mice. 181The effects of antibiotic treatment on food intake are inconsistent between studies and may be dependent upon the antibiotic regime, the complete or incomplete depletion of the microbiota, as well as other factors such as sex and age.
It is suggested that a direct impact of bacteria-derived metabolites, such as the short-chain-fatty acids (SCFAs), amino acids or bile acids 48 and hormones produced by the microbiota (GABA, serotonin, 5-HT) may act on central neurocircuitry through the vagus nerve, as one mechanism by which the gut microbiota regulates appetite and eating behavior.Administration of the SCFA butyrate for example, decreases food intake and reduces c-FOS expression in orexigenic NPY neurons in a vagus nerve-dependent manner.Serotonin (5-HT) derived dietary tryptophan and GABA derived from glutamate are both generated by the microbiota and have been shown to reduce appetite regulation via gut-brain signaling and secretion of appetitelinked hormones, such as cholecystokinin (CCK). 185In addition to such mechanistic studies, there is ample evidence that microbiometargeting interventions (prebiotics, probiotics) modulate food intake and hypothalamic gene expression (see Supporting Information, Table 2).For instance, administration of prebiotics, such as oligofructose was shown to prevent the increase in food intake, and bodyweight associated with HFD 186 (see Supporting Information, Table 2).
On the contrary, supplementation with Lacticaseibacillus rhamnosus, Lactobacillus acidophilus, and Bifidobacterium bifidum was associated with a reduction in food intake and bodyweight as well as improvements in leptin and insulin resistance in the hypothalamus of mice 187 (see Supporting Information, Table 2).
As previously described, the gut microbiota has been demonstrated to modulate Oxt expression (Section 2.3), but it is currently unknown whether this relates to homeostatic control of food intake and eating behavior.Given the evidence presented above, it seems likely that Oxt is involved in the effects of at least some microbiometargeting interventions.Indeed, a recent study by Narmaki et al. 20 demonstrated that supplementation with a multistrain probiotic increased circulating Oxt levels in obese women with food addiction and decreased in serum leptin and NPY levels. 20Consequently, this improved eating behavior and reduced food intake. 71Further research is needed to better understand this potential crosstalk between microbiota, Oxt and food intake, both centrally and peripherally.

| Beyond homeostatic appetite regulation: A microbiota-gut-oxytocin pathway?
Non-homeostatic regulation of eating behavior and food reward involves frontal cortical regions (PFC, orbitofrontal cortex (OFC), and insula), the mesolimbic system (VTA and nucleus accumbens [NAcc]) and the hippocampus.These brain regions are involved in regulate impulsivity, reward, memory, and sociability.
The PFC is thought to inhibit impulsive eating behavior except for a certain subset of neurons in ventral medial PFC (vmPFC). 158,188Petrovich et al. ( 2007) demonstrated that an intact vmPFC is required to induce eating in sated rats whereas sated rats with a lesion of the vmPFC consume less food than control rats suggesting that this region is necessary for controlling eating responses. 188Both the OFC and insula are involved in the coordination of perception of palatable foods, which increases motivation towards food. 156,158It is suggested that these effects demonstrated in rodents might translate to humans since hunger activates similar cortical regions across species. 189The mesolimbic dopamine pathways (comprising the VTA and the NAcc) encode food reward processes which enhance food consumption 158,159,190 and thus, can override signals from homeostatic systems, such as the hypothalamus 158,159 but also gut hormones signals. 158,190Lastly, the hippocampus is a brain region that contributes to the regulation of food intake by integrating memory cues, learning, visual, spatial environment information, and energy status. 158,159,190,191The ventral hippocampus in particular has been shown to be involved in the regulation of palatability and food preference through receiving endocrine signals and transferring these signals to higher-brain regions such as the NAcc. 192ytocinergic neurons project into many of the brain regions that were described above (Figure 3) 30,193 and thus, may regulate food intake through modulation of these areas.However, to date, our understanding of the underlying neurocircuitry is limited.Oxt neurons in the PVN have been shown to activate dopaminergic neurons in the NAcc and VTA, 5,194,195 which is essential for reward-driven social behaviors. 196Exogenous administration of Oxt into the CNS of different animal models, specifically injection into the NAcc 94 and VTA, 194 decreases sucrose intake after food deprivation 197 and the consumption of fructose-sweetened beverages 166 (see Supporting Information, Table 1).Another study confirmed this finding on food intake in a nonsocial context, but also that found that Oxt injection was not able to inhibit food intake when rats were presented food in social setting. 94e same research group demonstrated that administration of the OXTR antagonist L-368,899 increases sugar intake independently of social status as it increased sugar intake in both subordinate and dominant mice 198 (see Supporting Information, Table 1).These effects are ascribed to changes in the perceived reward and food palatability 5,104,194 and have been replicated in human trials.Intranasal Oxt administration, for instance, was found to reduce reward-driven eating of palatable food such as chocolate cookies. 4Furthermore, Oxt has been linked with food choice, 199 cognitive control of food cravings 174 and the processing of food-reward value 105 (see Supporting Information, Table 1).In this line, Olszewski et al. 199 demonstrated that the OXTR antagonist L-368,899 induced a preference towards sugar intake compared to either sucrose alone or in combination with lipids in mice, which was accompanied by increased c-Fos immunoreactivity in brain regions linked with reward, satiety and aversion 199 (see Supporting Information, Table 1).This research group also demonstrated that L-368,899 also increases carbohydrate (sucrose, glucose, fructose, polycose, or corn starch) compared to noncarbohydrate intake.This was accompanied by an upregulation of the expression of Oxt gene in hypothalamus of mice with sucrose and corn starch intake. 200though the gut microbiota is known to modulate PFC-regulated behaviors (including social and anxiety-linked behaviors), 201,202 reward signaling 203 and hippocampal function, 204 our understanding of the subsequent effects on food intake and eating behavior is currently limited.These regions, and particularly the PFC, however, are closely interlinked with brain regions that regulate food intake (Figure 3) which suggest that microbiome-targeting interventions may affect food intake through modulating these regions. 158,205For instance, L. rhamnosus supplementation to CONV-R mice downregulated GABA A -α2 receptor gene expression in PFC, which was linked with a reduction in depression-and anxiety-like behaviors mediated by the vagus nerve 206 (see Supporting Information, Table 2).Fecal microbiota transplant (FMT) from obese mice modulated gene expression and learning in PFC of recipient suggesting that the microbiota affects feeding-linked cognition in the PFC 207 (see Supporting Information, Table 2).The gut microbiota has been shown to modulate reward signaling. 203It has been shown that GF compared to CONV animals have increased synthetizing dopaminergic turnover (DOPAC/dopamine ratio). 203,208,209Moreover, GF mice display an increase in dopamine D1 receptor expression in the hippocampus, which was decreased in striatum and NAcc.No significant differences in D2 receptor expression in these brain regions were detected when GF mice were compared to specific pathogen free (SPF) mice. 210us, highlighting an implication of the microbiome on brain regions involved in reward processes and DA signaling, which is critical on cognitive and reward functions, as well.Control mice consuming the prebiotic fructo-oligosaccharides (FOS) have increased food intake and preference towards palatable food.However, FOS administration reduced overall food intake and preference for palatable food when given together with control chow diet to mice previously exposed to high-fat high-sugar (HFHS) diet. 179This was associated with a trend for increased expression of dopamine-associated genes in the NAcc and, increased expression of the orexigenic Agrp and Npy in the hypothalamus. 179Additionally, mice receiving FMT from obese animals displayed increased sensitivity to the effects of a HFHS diet compared to those receiving FMT from lean mice.This included a higher food reward and changes on the expression of dopaminergic genes in the striatum. 178In humans, administration of the SCFA propionate, via colonic delivery, reduced the appeal of high energy food pictures suggesting that colonic propionate might play a role in the attenuation of reward-context eating behavior. 211Effects of the microbiota on the hippocampus have previously been investigated because of its involvement in memory function.
It has been shown that the microbiota affects neurogenesis, 212 structure, 213 physiology, 214 gene expression 215 and neurotransmitter levels. 216Antibiotic-treated, GF and probiotic-supplemented mice have shown alterations in memory function and related behavior, but there is currently no evidence that this in turn affects eating behavior. 191A recent preclinical study has shown however that a high-dose administration of the probiotic-mix VSL#3 prevented diet-induced memory deficits on a place task in a dependent manner of hippocampus, even though this exposure did not alter energy intake or bodyweight. 217cordingly, the hippocampus is linked with episodic memory, which has been shown to modulate food intake in humans. 158,218This overall suggests that a potential effect of gut microbiota on hippocampaldriven regulation of food intake warrants investigation.

| Eating behavior, stress, and oxytocin
Stress alters metabolic needs and the function of many higher brain regions involved in the regulation of food intake.Further, brain regions involved in the regulation of stress responses are in close anatomical proximity to those regulating food intake.For example, in the hypothalamus which ensures that metabolic demands and energy status are adjusted during stressful events. 2190][221] On the other hand, stress modulates the amount and composition of food that is consumed.Acute stress, for example, suppresses food intake and increases energy expenditure, 219,222 whereas perceived stress has been shown to increase intake of palatable high-fat/high-carbohydrate food. 219,220,223e exact mechanisms that link food intake and stress response are not completely understood yet, but it is thought to involve a direct crosstalk between orexigenic AgRP/NPY neurons in the ARC and CRH-releasing neurons in the PVN.In addition, appetite-linked hormones ghrelin and leptin might be involved.In mice, activation of CRH neurons enhances food preference towards a high-carbohydrate diet. 224In line with this finding, glucocorticoid intake in humans leads to an increase in food intake and a higher preference for palatable carbohydrate-rich food 225 suggesting a modulation of the orexigenic AgRP/NPY neurons and reward-driven processes.Inhibition of the glucocorticoid receptor, CRHR1 was able to reduce stress-induced food consumption and food cravings in humans. 226Additionally, the orexigenic hormone ghrelin has been shown to enhance stress responses in rodent, whereas the anorexigenic hormone leptin, inhibits the release of noradrenaline and adrenocorticotrophic from PVN under stressful stimulus. 227 previously mentioned, Oxt is also produced in the hypothalamus 3 and can affect the bidirectional communication between stress and food intake centers. 5,156,167,219,220Both food intake and stressful stimuli activate Oxt neurons 169,219 which is linked to a modulation of downstream responses in brain regions involved in the regulation of reward, cognition, emotion, and homeostasis. 71,219In addition, manipulation of Oxt in the brain has been shown to modulate both stress responses and food intake (SI, Table 3).In a preclinical setting, Barrett et al. 228 showed that intraparaventricular thalamus (PVT) Oxt injection in female and male mice attenuated stress-and anxiety-induced hypophagia but did not alter food intake under normal conditions. 228In addition, Oxt administration increased the number of active lever presses suggesting increased motivation and learnt food reward behavior which were inhibited when animals were treated with the OXTR antagonist, atosiban. 228In humans (both obese and normal weight men), Oxt administration was able to reduce stress-induced HPA activation and inhibits hunger-induced food intake, but only in obese males 229 (see all the studies found linking Oxt and stress in context of food intake/eating behavior in Supporting Information, Table 3).Further, genetic polymorphisms of OXTR have been linked with higher vulnerability to stressful stimuli and an aggravation of symptomatology, including behavioral, 230 of eating disorders, such as AN 119 or overeating 231 in humans (also see section 4.4).
Whilst it has been shown individually that the microbiota affects Oxt, stress and food intake, it is currently unknown how the microbiota affects this triad.Evidence suggest that Oxt might be an important player especially because, in addition to the brain, Oxt is expressed in enteric neurons (see section 2) which a more closely located to the intestinal microbiota and may regulate food intake and stress responses in the hypothalamus via vagal afferents.

| Considerations for weight management, energy balance and dysregulation of food intake
In this review, we have so far concentrated on evidence linking Oxt and eating behavior under normal conditions.Modulation of Oxt has, however, also been suggested as a novel treatment strategy in multiple eating disorders such as anorexia nervosa (AN), binge-eating, bulimia nervosa (BN) and obesity. 156,158dividuals with AN, but not BN, display lower plasma Oxt levels compared to healthy controls. 232In contrast, food-induced Oxt responses were higher in individuals with AN. 233 It is suggested that this increase is related to the food-related stress response in individuals with AN. 234 Intranasal Oxt administration has shown beneficial effects in individuals with AN and was associated with a decrease in attentional vigilance and eating-related concern as well as a reduction in salivary cortisol levels which overall points to a potential beneficial effect of Oxt in this disorder 219,235,236 which is likely because of its holistic effects on food intake regulation, mood, stress response, memory function and sociability.
With regard to obesity and metabolic syndrome, increased levels of plasma Oxt have been reported in both women and men. 171However, the opposite has also been found.Serum Oxt levels negatively correlated with body mass index (BMI), waist circumference and waist-tohip-ratio. 237The therapeutic effects of Oxt administration in obesity and metabolic syndrome are currently under investigation 70 to assess dosage, administration regime, efficacy, and safety of such treatments.
Given differences in the microbiome composition between lean and obese individuals and changes of microbiome composition associated with dietary interventions, 183 microbiome-targeting interventions have been discussed as an additive therapeutic approach for obesity. 183,238Indeed, preclinical and clinical studies suggest that probiotics can reduce bodyweight in obese rodents and humans, 238 but manipulation of the gut microbiota alone might not be sufficient or effective enough to alter bodyweight. 183To date, the precise mechanisms of the gut microbiota action are unknown and only few studies addressed effects on food intake and energy expenditure, particularly in humans.In mice, a reduction of bodyweight by administration of Hafnia alvei HA4597, was associated with reduced food intake in ob/ob mice (3 Â 10 8 CFU/day, 18 days), but not high-fat diet-induced obese mice (4 Â 10 7 CFU/day, 46 days). 239In humans, it was recently shown that a Hafnia alvei strain improved weight loss in overweight individuals in combination with a hypocaloric diet throughout the study. 240This suggests that probiotics might function to increase the efficacy of more conventional lifestyle interventions aiming to improve obesity and metabolic disorders. 183Further, it is also suggested that their beneficial effects may not be primarily mediated through a reduction of food intake but rather changes of metabolism and increased energy expenditure. 241To our knowledge, however, only Shirouchi et al. 242 have previously demonstrated that Lactobacillus gasseri SBT2055 fermented milk (6 Â 10 7 CFU/g diet, 4 weeks) increased energy expenditure in a high fat-high sugar-induced obesity rat model. 242Future studies should further explore how probiotic administration and microbiota changes through dietary approaches modifies both food intake and energy expenditure especially in humans. 243Given the complex nature of human studies and the promising findings regarding Oxt-based therapies, we suggest that combining probiotics with Oxt could provide an additional benefit in weight management, through better modulation of central appetite, reward and stress.
We hypothesize that Oxt is a major player involved in the modulation of food intake, and that the microbiome can also impact eating via Oxt.The current evidence linking both the microbiota and Oxt independently to food intake behavior, food reward and stress, warrants further investigation into the mechanisms and neurocircuitry involved in mediating their convergent effects.

| CONCLUSIONS
The Oxt neuropeptide has biological and behavioral functions that extent beyond parturition, lactation, and social behavior.More recently, it has been shown to also modulate eating behavior.Indeed, Oxt coordinates and integrates homeostatic signals derived mainly from the hypothalamus, but also the brainstem, with non-homeostatic signals from the mesocorticolimbic system and the hippocampus.
Thereby, it controls not only energy and glucose metabolism, but also food intake through affecting brain centers involved in decisionmaking processes.At the same time, growing evidence suggests that this regulation of both homeostatic and non-homeostatic eating behaviors is interconnected with the gut microbiota.This adds the microbiota-gut-brain-axis as a player in the oxytocinergic regulation of brain function and behavior.Recent studies have shown that Oxt and the gut microbiota are tightly interlinked with potential significant outcomes on stress and social behaviors.However, a gap in knowledge at this point is whether these two key players also converge their influence on eating behavior.For this reason, further research is needed to elucidate the precise mechanisms underlying the effects of Oxt and the gut microbiota on homeostatic and non-homeostatic regulation of eating behavior, but also their combined impact.Furthermore, stress is an additional factor that modulates each of these players independently, but it remains unknown how gut microbiota-and Oxt-mediated effects on eating behavior would be impacted by stress or in conditions of dysregulated eating.Future research in this area needs to step away from solely demonstrating a correlation between these puzzle pieces, but move towards understanding causality, the mechanisms of action and the involved neurocircuitry.
Understanding this knowledge gap should foster further studies to investigate microbiota interventions targeting the Oxt system as a potential treatment not only for neurodevelopmental disorders and social disorders but also in the context of stress-related disorders and eating disorders, in stress and non-stress contexts.