Complexity of the Hypothalamic Oxytocin System and its Involvement in Brain Functions and Diseases

Oxytocin Neurons and Their Diversity

Oxytocin is mainly synthesized and released by oxytocin neurons residing in several hypothalamic nuclei, including paraventricular nucleus (PVN), supraoptic nucleus (SON), and accessory nucleus (AN) [12]. Recent studies reported the distribution of oxytocin neurons in the mouse tuberal nucleus [2] and possibly in the human prefrontal cortex [13]. Traditionally, hypothalamic oxytocin neurons are classified into magnocellular (Magno) and parvocellular (Parvo) subtypes based on their morphologies, electrophysiological properties, and neural circuitry projections (Fig. 1) [14, 15]. Magno neurons, but not Parvo neurons, exhibit a pronounced transient outward rectification and can be labeled by Fluorogold (FG) signal when FG is delivered by intraperitoneal or tail vein injection [14–17]. Microinjecting retrograde tracers into the pituitary has also been used to label Magno neurons [18]. Magno neurons are found in the PVN (anterior subregions), SON, and AN, but Parvo neurons mainly reside in PVN, especially its caudal region [1, 17].

Despite the fact that the canonical classification of oxytocin neurons is widely used [16, 18, 21–23], single-cell transcription, morpho-electric properties, or single-neuron whole-brain projectome studies, mainly in mice, suggest that PVN oxytocin neurons have more than two subtypes (Fig. 1) [1, 19, 20]. Although only 37 Oxt-positive neurons were included, single-cell RNA sequencing data demonstrated 4 types of oxytocin neurons with distinct genetic profiles [19]. Chen and colleagues profiled the morpho-electric properties of mouse PVN oxytocin neurons and found that they have 6 subtypes [1]. Unexpectedly, one subtype, mainly residing in the posterior part of PVN, is composed of both Magno and Parvo neurons, and they share most of their morpho-electric properties [1]. Neural circuitry studies based on single-neuron projections of PVN oxytocin neurons also suggested that these neurons are clustered into 6 subtypes [20]. Despite the 6 subtypes of oxytocin neurons being reported based on both morpho-electric properties and single-neuron projections [1, 20], the subtypes reported in these two studies are not well matched, which indicates that oxytocin neurons have more than 6 subtypes when considering morpho-electric properties, projections, and gene expression. In addition, fiber photometry-based experiments showed that oxytocin is released differentially in discrete brain regions during specific stages of mating behaviors [24], which implies that oxytocin neurons are functionally diverse. Although the oxytocin system is evolutionarily conserved in vertebrates [11], classification studies are mainly conducted in mice, and it is necessary to confirm the diversity of oxytocin neurons across species [25].

Input-Output Circuits of Oxytocin Neurons

No neuron is an island, and every neuron integrates synaptic inputs from many brain regions and transfers signals to multiple downstream targets [26]. With the development of novel neural circuitry tracing tools [27], the input-output circuit structures of many types of neurons, including hypothalamic oxytocin neurons, have been mapped.

Classical immunohistochemical studies have reported the distributions of oxytocinergic axons in some brain regions of rodents, monkeys, and humans, and different distribution patterns have been found across species [28–30]. In rodents, we can efficiently label oxytocin neurons and trace their axonal projections depending on adeno-associated viruses (AAVs) and transgenic mouse lines [2, 20, 31–33]. Grinevich and colleagues developed a series of oxytocin-promoter AAVs to target rodent oxytocin neurons and dissected the whole-brain distribution of oxytocinergic axons in rat brains. They reported that PVN oxytocin neurons have broad projections to the olfactory system, cortical areas, the nucleus accumbens (NAc), lateral septum (LS), the bed nucleus of the stria terminalis (BNST), hippocampus, amygdala, and paraventricular thalamus, but SON oxytocin neuronal axons have only been detected in five regions [31]. Studies with Oxt-Cre transgenic mice [33] showed that PVN oxytocin neurons also have broader projections than SON neurons, and target the cerebral cortex, cerebral nuclei, thalamus, hypothalamus, midbrain, pons, and medulla [2, 32]. Similar to the morpho-electric properties [1], the axonal projections of mouse PVN oxytocin neurons do not exhibit evident sex differences [2, 20, 32]. In addition to releasing oxytocin, PVN oxytocin neurons co-release glutamate and other bioactive molecules. Though PVN oxytocin axons are detectable in many brain regions, whether oxytocin is released to modulate these regions remains unclear, and this requires investigation by electron microscopic imaging, electrophysiological recording, or genetically encoded oxytocin sensor detection [22–24].

The whole-brain projections of the classical subtypes of oxytocin neurons, Magno and Parvo neurons, have also been separately investigated. Zhang and colleagues reconstructed the projections of Magno neurons in rats by combining a subtle virus tracing strategy and fluorescent micro-optical sectioning tomography (fMOST) imaging and found that some Magno neurons projecting to the posterior pituitary send collaterals to the striatum, amygdala, and LS [18]. Another recent study based on fMOST single-cell imaging traced and reconstructed the single-neuron, whole-brain projections of 264 PVN oxytocin neurons in mice [20]. They found that one class of neurons, including all the Magno neurons, exclusively project to the hypothalamic regions, and the other class, including Parvo neurons, have extensive projections to the endbrain and brainstem [20]. About 30% of oxytocin neurons (more than > 600) in mouse PVN are the Parvo subtype [1, 17], but the rat has a small number of Parvo neurons, < 70 cells [11]. This difference may contribute to the broader projections of Magno neurons to extrahypothalamic regions in rats [18, 20]. As the subtle neural projections of PVN oxytocin neurons have been dissected [20], their functions await deeper exploration.

In addition to having broad axonal projections, oxytocin neurons integrate information from many brain regions. Utilizing the rabies virus (RV)-dependent retrograde tracing strategy [34], brain-wide synaptic inputs to hypothalamic oxytocin neurons in rodents have been investigated [2, 21, 32, 35–37]. PVN and SON oxytocin neurons receive similar inputs from hypothalamic nuclei, thalamic areas, and cerebral nuclei, but more lateral brain regions send inputs to SON oxytocin neurons [2]. Synaptic inputs to PVN oxytocin neurons are sex-invariant in adult mice [32]. In virgin males and fathers of mice, the input fraction from hypothalamic nuclei is reported to be largely similar, but the lateral hypothalamus (LH) and medial preoptic nucleus (MPO) exhibit more retrograde labeling by RV in the fathers than in the virgin males; this is transient and reversible [35]. In non-pregnant female mice and lactating mothers, there are no statistically significant differences in input neural circuits to PVN oxytocin neurons [37]. When separately considering Magno and Parvo neurons, Parvo neurons in rats receive more inputs from both hypothalamic and extrahypothalamic regions than Magno neurons, and some regions, such as the subfornical organ and substantial nigra, provide more inputs to Magno neurons [21]. As mice have more Parvo neurons than rats, it is worth comparing whether Magno and Parvo neurons in these two species have different information integration strategies.

Dissecting neuronal input-output relations is instrumental in understanding how the brain processes information [38]. The strategy combing canine adenovirus (CAV) or retrograde AAV with RV tracing, also called the “tracing the relationship between input and output” (TRIO) method, has been developed to dissect the input-output organization of specific neurons, such as norepinephrinergic and dopaminergic neurons [38–41]. Using a strategy similar to the TRIO method, Tang and colleagues investigated the inputs of PVN oxytocin neurons projecting to the SON and posterior pituitary in rats [21]. With the development of tracing tools and imaging techniques [41, 42], it is possible to anterogradely and retrogradely trace single neuronal whole-brain connections. Considering the functional diversity of oxytocin neurons, dissecting the subtle input-output circuits with spatial transcriptomics will advance our understanding of the neural mechanisms underlying oxytocin functions.

Widespread OxtR Expression in the CNS

Oxytocin mainly binds to OxtRs to modulate neuronal activity and intracellular signaling [3, 43]. Exploring OxtR expression is necessary to learn how oxytocin signals modulate cellular activity, circuitry dynamics, and animal behaviors. Receptor autoradiography was first used to detect oxytocin binding in prairie voles and rats, and these studies provided valuable information about the distribution and dynamic change of oxytocin binding sites in the brain during development [44, 45], but its resolution is much lower than immunohistochemistry. Commercially available OxtR antibodies have been reported to have specificity issues [46]. In recent years, Froemke and colleagues developed an efficient OxtR antibody, which identified the OxtR expression profile in the brains of mouse mothers, virgin females, and males [47]. OxtR-positive neurons are distributed in a myriad of brain regions; the piriform cortex and hippocampal CA2 regions have more OxtR-positive neurons in female mice than males, and OxtR-positive neurons are more dense in the left auditory cortex than in the right auditory cortex [47]. With the OxtR-Venus mice generated by Nishimori and colleagues, OxtR neurons labeled with Venus are more dense in the MPO of female mice, especially in the anteroventral periventricular nucleus [48]. Unexpectedly, no significant correlation was found between OxtRs and oxytocin neuronal axons in mouse brains [47]. A similar result was also reported in a transgenic mouse line with labeled OxtR-positive neurons [2, 49]. In addition, Hu and colleagues found OxtR-positive neurons in the mouse retina, but few oxytocinergic axons were detected [50]. The mechanisms and possible physiological roles of the different distributions of oxytocinergic axons and OxtRs remain unclear. OxtR expression is changed during development and the changing pattern is not the same in distinct brain regions of mice [47, 49, 51] and humans [52], which supports the diverse roles of oxytocin signals at different stages [53].

Neurons can be classified based on neurotransmitter/neuromodulator synthesis and release. OxtR expression in distinct neuronal types has been investigated in some brain regions (Fig. 2), which is important for predicting the possible functions of oxytocin signals in information processing. In the mouse medial prefrontal cortex (mPFC), OxtR is expressed in both glutamatergic pyramidal neurons and GABAergic interneurons, especially somatostatin (SST) interneurons [54–56]. One recent preprint study in rats showed that > 90% of OxtR-positive neurons in the mPFC are GABAergic, including parvalbumin (PV), calbindin, calretinin, and SST interneurons [57]. In the mouse auditory cortex, ~40% of SST and PV interneurons are OxtR-positive [58]. In young mice, Oxtr mRNA is exclusively co-localized with GABA neurons, including SST interneurons in the visual cortex (V1), but it is expressed in both GABA neurons and glutamate neurons in the somatosensory cortex (S1) [59]. OxtRs are also expressed in neurons of other mouse cortical regions, such as the piriform cortex and motor cortex [47, 60], but the subtypes of OxtR-positive neurons have not been investigated. In addition to the cerebral cortex, OxtR-positive neurons in some subcortical neurons have also been identified. In the hippocampus, OxtR is not only expressed in pyramidal neurons but also in different types of interneurons [61], and OxtRs are highly expressed in the CA2 and distal CA3 regions of the dorsal hippocampus in mice [47, 62]. In the LS, BNST, and PAG, OxtR neurons are colocalized with GABAergic neurons [22, 46, 63]. OxtR-positive neurons in the mouse paraventricular thalamus send glutamatergic axons to the NAc and basolateral amygdala [64], and OxtR-positive neurons are exclusively glutamatergic in rat PVN [65] and mouse arcuate nucleus (ARC) [66]. Some monoaminergic neurons are reported to express OxtRs. Oxtr mRNA is expressed in both dopamine and non-dopamine neurons in the mouse VTA [67], but mainly in GABA neurons in the SNc [23, 68]. Many serotonergic neurons in the mouse raphe nucleus are reported to be OxtR-positive [46]. In addition to the neurotransmitter-labeled neurons, OxtR is also reported to be expressed in neuromodulator-synthesizing (including neuropeptide) neurons. In the ARC, > 50% of OxtR neurons are pro-opiomelanocortin (POMC)-positive [66, 69], and ~40% of OxtR-positive neurons are corticotropin-releasing factor (CRF)-positive in the PVN [65]. Therefore, oxytocin can broadcast its influence by regulating the synthesis and release of various neurotransmitters/neuromodulators via OxtR expression in different neurons.

In addition to neurons, OxtR is expressed in non-neurons, including microglia and astrocytes [25, 73–75]. Microglia act as immune cells and play an important role in regulating synaptic connectivity. OxtR mRNA and protein were reported to be expressed in CD11b-positive cells assessed using flow cytometry in mice [76]. BV-2 microglia express OxtRs and OxtR agonists inhibit microglial activation in some disease models in vitro and in vivo [77, 78]. However, these studies did not directly show OxtR expression in brain microglia. Utilizing fluorescent in situ hybridization and transgenic mice, Jiang and colleagues provided direct evidence showing the expression of OxtR in mouse hippocampal microglia cells [73]. Astrocytes serve as a pivotal regulator of neuronal synapse functions and bi-directionally crosstalk with neurons. OxtRs have been detected not only in cultured astrocytes [73], but also in some brain regions, such as mouse auditory cortex [47], mouse NAc [70], mouse LS [70], mouse and rat central amygdala [70, 72], rat dorsal hippocampus [79], and rat striatum [71]. Glial cells are much more than supporting cells [80], but the functional expression of OxtRs in non-neurons in brain functions and diseases remains to be deeply investigated.

Oxytocin may target non-OxtRs, such as vasopressin receptors and the TRPV1 channel, to execute actions. Since OxtR shares high levels of sequence alignment with vasopressin receptors (V1a, V1b, and V2) [81, 82], the activity of oxytocin at the V1a and V1b receptors has been reported in rodents [83], such as the midbrain [23] and dorsal root ganglion [84]. In the mouse spinal cord, oxytocin interacts with TRPV1 to attenuate capsaicin-induced nociception [85]. In addition, other neuropeptides, such as vasopressin, can activate OxtRs. All of these factors further complicate the modulatory modes of oxytocin signals. As the structural information of OxtR is dissected with cryo-electron microscopy [47, 86–88], it is possible that potent and specific OxtR-targeted agonists will be discovered or re-constructed.

Oxytocin, Social Behaviors, and Autism

Social behaviors are essential for animal survival and propagation. Oxytocin is known as the love hormone because of its role in promoting social interaction between lovers, parents, and offspring. In addition to the impairment in milk injection, one of the consistent functional disorders in oxytocin- and OxtR-deficient mice is social amnesia [9, 10]. A recent systematic review provided a detailed summary of the role of the oxytocin system in sociality-related behaviors [103]. Here, we mainly focused on recent neural circuitry studies about oxytocinergic modulation of social behaviors under physiological and pathological conditions.

Fiber photometry recording has shown that the activity of PVN oxytocin neurons is enhanced during juvenile social interaction, and activating oxytocin neurons promotes mouse social-reward learning [104] and facilitates parental behavior of virgin male mice [35]. Directly silencing PVN oxytocin neurons is reported to impair social-related behaviors, including impairments of the short- and long-term social recognition memory in male rats [105] and reduced maternal preference in mice, rats, and rhesus monkeys [106]. PVN oxytocin neurons are strongly activated by stressful stimuli, including social aggression, and inhibiting these neurons during restraint stress in mice prevents social cognitive impairment [107]. Although most studies have suggested the prosocial role of oxytocin signals, one study with wireless optogenetic devices found that repeated activation of PVN oxytocin neurons in male mice dynamically elevates both prosocial and agonistic behaviors, depending on the social context [108]. Thus, PVN oxytocin neurons are involved in modulating social behaviors, but their role in promoting sociality depends on context and environment.

Oxytocin signals in various social regions of the brain, such as the VTA, NAc, PFC, amygdala, and LS, modulate social behaviors [103]. Oxytocin release in the VTA modulates dopamine neuronal activity [23, 104], which promotes mouse social reward behavior [104]. Oxytocin signals in the NAc together with dopamine signals mediate pair bonding in the prairie vole [109], and oxytocin-mediated serotonin release in the NAc promotes social reward in the mouse [96]. In the mPFC, OxtR-positive neurons modulate sociosexual behavior in female mice and anxiety-related behaviors in male mice [56]. In female rats, oxytocin facilitates social behavior via modulating mPFC interneurons [57]. Oxytocin signals in the rat hippocampus, medial amygdala, and supramammillary nucleus are linked with social memory [105, 110, 111]. In the medial amygdala of male mandarin voles, distinct OxtR-positive neurons project to the anterior insula and ventromedial hypothalamus to control consolation and aggression, respectively [112]. A series of studies conducted by Neumann and colleagues reported the role of the LS oxytocin signal in modulating rat social memory and aggression, and they found that oxytocin in the LS is involved in modulating social memory retrieval, social fear extinction, and female aggression [113–116]. Oxytocinergic modulation in the anterior subdivision of the ventromedial hypothalamus was found to control aversive social learning in mice and manipulation of OxtR-positive neurons in this region bidirectionally regulates social avoidance [117]. A study with a genetically encoded oxytocin sensor (GRAB_OT) found that endogenous oxytocin release in the VTA, mPFC, and PVN differently varies during the stages of male mouse mating behavior [24]. Aggressive behavior is elevated in OxtR-deficient mice [9], but knocking down OxtRs and V1a receptors in estrogen receptor alpha-positive neurons of the ventromedial hypothalamus was found to reduce aggressive behavior [118]. Social behaviors are complex, comprising enriched stages and subtypes, and the specific roles of oxytocin signals, such as different oxytocin neurons and neural circuits, in various target regions in regulating social behaviors remain to be investigated.

Autism or autism spectrum disorder (ASD) is a neurodevelopmental disorder with social dysfunction as one of the core symptoms [119]. Oxytocin is thought to be a potential drug to treat ASD because of its actions in promoting social reward and social learning. Given that ASD is a complex human genetic disorder and some genetic risk factors have been discovered [120, 121], the effects of oxytocin on alleviating social behaviors have been reported in rodent autism models based on risk genes. Mutation of the synaptic adhesion molecule neuroligin3 (Nlgn3) is associated with ASD, and in Nlgn3-knockout mice, restoring the impaired oxytocin signaling in midbrain dopaminergic neurons alleviates social novelty [122]. Selective activation of OxtR neurons in the LS also restores the impaired social memory in Nlgn3^R451C knock-in mice [123]. Magel2 is one of the genes strongly associated with ASD risk, and subcutaneous oxytocin administration in Magel2 mutant neonates has been reported to restore neuronal alterations in the hippocampus and social memory [124]. In Magel2-knockout mice, oxytocin and vasopressin signals from SON to LS are deficient, which disrupts social fear extinction by disinhibiting SST neurons [125]. In a fragile X syndrome mouse model (Fmr1-knockout mice), neonatal administration of oxytocin relieves the over-activation of dorsal hippocampal CA3 neurons and prevents social dysfunction [126]. In a mouse model that carries the ASD patient-derived de novo mutation in the pogo transposable element derived with zinc finger domain (POGZ^WT/Q1038R mice), the oxytocin system is impaired, and intranasal administration of oxytocin alleviates social behavior deficits [127]. Mutations in the Contactin-associated protein-like 2 (CNTNAP2) gene can cause a syndromic form of ASD. In Cntnap2-knockout mice, several studies found that PVN oxytocin neurons are reduced and exogenously administered oxytocin or elevating endogenous oxytocin release reverses the aberrant connectivity patterns and rescues social interaction by targeting OxtRs in the NAc [128, 129]. In the SH3 and multiple ankyrin repeat domains 3 (Shank3) gene deficiency mouse and rat models [130, 131], oxytocin administration has been reported to alleviate social interaction and social memory, and oxytocin treatment restores LTP induction at hippocampal and hippocampal-prefrontal synapses in Shank3-deficient rats [130]. A recent study identified a new risk gene for ASD, secretagogin (SCGN), a synaptic transmission regulator, and showed that oxytocinergic signaling is dysregulated in ASD children with SCGN mutation and in Scgn deficient zebrafish and mice [132]. Moreover, oxytocin administration attenuates the ASD-associated defects caused by SCGN deficiency [132]. These animal studies indicate that many brain regions are involved in the social deficits of ASD, and oxytocin may target distinct regions to alleviate the associated social behaviors in different genetic mutations of ASD.

Clinical trials of intranasal oxytocin therapy in ASD have been conducted. However, the results from different trials remain controversial. An early study reported that plasma oxytocin levels were reduced in male autistic children [133]. For adult ASD or Asperger’s disorder, oxytocin administration was reported to increase the retention of social cognition [134]. Guastella and colleagues provided the first evidence of the improvement of emotion recognition by oxytocin nasal spray in young individuals with ASD [135]. Some subsequent clinical trials have also reported the ameliorating effect of oxytocin on social impairments in ASD patients [136–138]. However, no significant effect in ameliorating the social symptoms of ASD has been reported in some studies [139–141], especially one recent clinical trial with 290 ASD children and adolescents [141]. Several possible aspects may contribute to these inconsistent conclusions [137, 142, 143]: (1) different genetic subtypes of ASD; (2) paired oxytocin with behavioral therapy; (3) the age of oxytocin interference; and (4) the dosage of oxytocin. Dissecting the influences of these factors on the efficiency of oxytocin therapy will provide valuable information for the precision-targeted therapy of ASD.

Oxytocin, Energy Homeostasis, and Obesity

The pivotal role of hypothalamic neurons in regulating and maintaining energy homeostasis has been widely investigated. As one of the important neuroendocrine hormones synthesized in the hypothalamus, oxytocin is involved in regulating food intake and energy expenditure [144–146]. The anorexigenic role of oxytocin by intracerebroventricular injection in rats was reported more than 30 years ago [147]. Subsequent studies in different species reported that oxytocin, administered ventricular [148–150], intraperitoneally [151, 152], or subcutaneously [153], inhibits food intake as well as increases energy expenditure in a dose-dependent manner. The internal state is reported to influence the effect of oxytocin on regulating food intake. For example, oxytocin administration increases food intake in pregnant rats [154], and clinical results have shown that intranasal oxytocin administration only reduces food intake in the fasted state but not in normal-weight men [155]. Eating behavior is strongly influenced by social factors, called the “social facilitation of eating”. Oxytocin administration in the rat dorsal hippocampus does not affect food intake under isolated conditions, yet significantly increases consumption in the presence of a familiar conspecific [111].

Numerous studies directly manipulated oxytocinergic neurons to investigate the role of endogenous oxytocin in regulating energy homeostasis, but the results are inconsistent. Oxytocin-deficient mice were reported to develop late-onset obesity without any alteration in food intake [156]. Ablating oxytocin neurons in adult mice also did not have significant effects on body weight, food intake, or energy expenditure on a regular diet, but male adult mice after ablating oxytocin neurons were more sensitive to high-fat diet-induced obesity [33]. In addition, directly modulating PVN oxytocin neurons by optogenetic or chemogenetic methods did not change food intake in food-deprived mice [157, 158]. Specifically knocking down ErbB4, the receptor of the adipose tissue-enriched secreted factor neuregulin 4 (Nrg4), in PVN oxytocin neurons had no significant effect on body weight on a regular diet, but aggravated obesity when shifted to a high-fat diet [159]. These studies indicate that hypothalamic oxytocin neurons may not be involved in regulating regular food intake. However, Atasoy and colleagues found that activating PVN oxytocin neurons suppressed the feeding behavior induced by activating ARC AgRP (agouti-related peptide) neurons [157]. Two recent studies found that virus-mediated ablation of PVN oxytocin neurons or conditional knockout of the oxytocin gene in the PVN, but not in the SON, increases mouse food intake and induces hyperphagic obesity [152, 160], and exogenous oxytocin treatment rescues the related phenotypes. In addition, deletion of transient receptor potential channel 5 (Trpc5) from PVN oxytocin neurons causes obesity in male and female mice, and intraperitoneal injection of oxytocin reverses the food intake and body weight changes in OXT-Trpc5 knockout mice [161]. Since these studies used different experimental strategies to manipulate oxytocin neurons, the compensation effect and functional diversity of hypothalamic oxytocin neurons may contribute to the inconsistent effects on food intake and body weight. A recent study found that, in one subtype of PVN oxytocin neurons, their firing rate changes with feeding states [1]. We infer that different subgroups of oxytocin neurons may play mutually antagonistic roles in regulating feeding behaviors, and there may be a specific group of feeding-sensitive oxytocin neurons, similar to the fear-sensitive oxytocin neurons [162].

The nucleus of the solitary tract (NTS), parabrachial nucleus (PBN), LH, paraventricular nucleus of the thalamus (PVT), ARC, VTA, and NAc are well-known brain regions involved in feeding regulation [163, 164]. The possibility of oxytocin targeting these regions to regulate food intake has been investigated. The NTS is an important relay center for incoming central and peripheral visceral sensory inputs, and oxytocin infusion into the rat's fourth ventricle and NTS reduces food intake and body-weight gain [150, 165]. Chemogenetic activation of mouse PBN OxtR-positive neurons does not affect 24-h fasting-induced food intake but suppresses food intake during mild dehydration, indicating the role of the PBN oxytocin signal in mediating dehydration-induced anorexia [166]. Neuronal circuitry-tracing studies described the dense distribution of oxytocinergic axons in the mouse LH [2, 20], and oxytocin was suggested to excite LH melanin-concentrating hormone (MCH)-positive GABA neurons [167]. Surprisingly, few studies reported the possible roles of oxytocinergic modulation in LH on feeding behavior. The PVT plays an important role in regulating feeding motivation. Oxytocin can activate PVT OxtR-positive neurons projecting to the NAc to promote mouse feeding motivation and meal frequency, but not total food intake [64, 168]. Direct oxytocin microinjection into the NAc decreases deprivation-induced chow intake and reduces palatable food consumption in nondeprived rats [169]. The ARC is a key node of hypothalamic control of feeding behavior, and the PVN is one of the important downstream regions of the ARC. Activating mouse PVN oxytocin neurons partially blocks the orexigenic effect of activating ARC AgRP neurons [157]. Some ARC glutamatergic neurons express OxtRs and ~50% of ARC OxtR neurons express POMC, and activating OxtR-positive neurons rapidly reduces mouse dark-cycle feeding [66] and knocks out mouse food intake and the body weight [152] increase by ARC OxtRs. VTA dopamine neurons regulate feeding behavior via their projections to the PBN and NAc in mice [170, 171]. The regulation of VTA dopamine neurons by PVN oxytocin neurons has been well investigated [23, 104], but the effect of oxytocinergic regulation in the VTA on food intake remains unknown [172]. Feeding behavior comprises several subtypes (homeostatic and hedonic feeding) and stages (procurement, consumption, and termination) [163]. Oxytocinergic modulation in different brain regions seems to have distinct functions in regulating feeding behavior, therefore it is necessary to explore and dissect the roles and neural mechanisms of oxytocin neurons in regulating feeding behavior. In addition to the central regulation, a recent study identified a subpopulation of tyrosine hydroxylase-positive sympathetic neurons innervating adipose tissue (WAT) in mice that release oxytocin and regulate adipose and systemic metabolism by activating OxtRs expressed on adipocytes [7]. How are the oxytocin signals in the peripheral system involved in regulating energy homeostasis? What is the functional interaction between central and peripheral oxytocin signals? These questions require further investigation.

Oxytocin signaling impairments are correlated with obesity in non-human primates and humans. Serum oxytocin levels are lower in obese adolescents and polymorphisms in the OXTR gene are associated with overeating and obesity [173, 174]. In macaques and humans with diabetes, oxytocin neurons in the PVN, but not in the SON, are significantly decreased [175, 176]. The functional mechanisms underlying the endogenous oxytocin modulation of glucose homeostasis are still under investigation [177]. In Prader-Willi syndrome (PWS), a complex multisystem genetic disorder with hyperphagia and obesity, oxytocin neurons are deficient in the PVN [178] and OxtR gene function is impaired [179]. Clinical trials have shown that short-term intranasal oxytocin administration improves the syndromes of feeding, anxiety, and socialization in children with PWS [180, 181]. However, similar to ASD, no significant effect of oxytocin on PWS was reported in some clinical trials [182]. Although the intervention period of oxytocin is critical [183], dissecting the mechanisms underlying the reduction of PVN oxytocin neurons and how endogenous oxytocin signals play different roles is also important for optimizing the strategy of treatment with oxytocin.

Oxytocin, Emotion, and Anxiety/Depression

Anxiety and depression are the most common affective disorders that seriously affect the daily life and health of patients [184]. Both peripheral and central oxytocin levels in humans are negatively correlated with anxiety symptoms [185, 186]. Although the reduction of oxytocin levels may be a susceptibility factor for the increased likelihood of developing affective disorders including depression, some studies report no alterations or even elevated plasma and cerebrospinal fluid oxytocin levels in subjects with major depression [187–189]. For the beneficial effects of oxytocin on anxiety and depression, clinical results are inconsistent [189, 190], which may be due to many factors, including small sample sizes, oxytocin dosage, and the heterogeneity of affective disorders.

The influences of oxytocin on affective disorders have been widely investigated in animal models. Chronic exposure to stressors induces anxiety and depression. Fiber photometry in vivo recording results have shown that PVN oxytocin neurons are activated by acute stressors [107]. PVN oxytocin neurons may resist the stressor-induced negative effect when facing acute stressors. Chronic inflammatory pain decreases the excitability of PVN oxytocin neurons and induces anxiety in mice [191]. One recent study showed that the Trpc5 (transient receptor potential channel 5) gene is expressed in mouse PVN, but not SON oxytocin neurons, and deleting Trpc5 in oxytocin neurons results in anxiety-like behaviors and causes postpartum depressive behavior, which is restored by overexpressing this gene in oxytocin neurons [161]. PVN neurons, including oxytocin neurons, express OxtRs, and locally delivering oxytocin into the PVN has an anxiolytic effect in a dose-dependent manner [192]. In addition to the PVN, the anxiolytic and antidepressant effects of oxytocin infusion have been reported in other brain regions. Oxytocin infusion into the mPFC relieves depression-like behaviors in both naïve and social defeat stress-induced depressed male mice via the potentiation of dopaminergic synaptic transmission [193]. Activating OxtR-positive neurons in the mPFC decreases anxiety-like behavior in male, but not female mice [56]. In the mouse anterior cingulate cortex (ACC), oxytocin administration reduces chronic pain-induced anxiety by reducing presynaptic long-term potentiation [194]. Oxytocin administration in the mouse median raphe nucleus reduces anxiety-related behavior by elevating OxtR-mediated serotonin release [46]. The amygdala is one of the most important brain regions for emotion regulation. Microinjection of oxytocin into the central nucleus of the amygdala (CeA) or activating oxytocinergic axons in the CeA prevents inflammatory pain-induced anxiety-like behaviors in mice [191]. Some CeA astrocytes also express OxtRs, and the OxtR-mediated modulation of the CeA astrocytes-neuronal network promotes a positive emotional state and reduces anxiety [72].

Several studies in both humans and rodents have reported the effect of oxytocin in promoting anxiety. Intranasal oxytocin delivered at 24 IU elevated the sensation of social stress in human subjects despite no change in salivary cortisol level [195]. In another randomized controlled trial, oxytocin application increased the defensive responses to unpredictable shocks in humans, suggesting that oxytocin enhances anxiety in the context of uncertainty [196]. Oxytocin infusion into the BNST in California mice increases social vigilance and reduces social approach, and knockdown of oxytocin prevents social stress-induced increases in social vigilance and decreases in social approach [197]. One preprint study in BTBR mice found that activating the PVN oxytocinergic projection to the medial amygdala increases anxiety-like behavior while activating the oxytocinergic projection to the BNST suppresses anxiety-like behavior and social approach [198]. Hence, the functional roles of oxytocin and oxytocinergic projection are inconsistent, even opposite, depending on many factors, such as context, gender, species, and neural circuits.

Social isolation stress is one of the major risk factors for the incidence of anxiety and depression [199]. During the COVID-19 pandemic, people suffering from anxiety and depression increased by > 20% globally [200], among whom people living alone were more severely affected [201]. Oxytocin signals exhibit different changes after social isolation. Plasma oxytocin levels and the density of PVN oxytocin neurons are increased after 4 weeks of social isolation in rats and prairie voles [202, 203]. PVN oxytocin mRNA expression or PVN oxytocin neuronal activity has been reported to increase in female rats after 7 weeks post-weaning social isolation or in male mice after 1 week of acute social isolation [204, 205]. However, a decrease in the number of PVN oxytocin neurons was reported in single-housed male mice [206], and high-lonely individuals show reduced salivary oxytocin levels [207]. OxtR expression is reduced after social isolation in several brain regions, including the CeA and NAc [205, 208]. Oxytocin has been reported to alleviate social isolation-induced behavioral disorders, including anxiety and depression. Socially isolated mice exhibit heightened aggression that is powerfully and dose-dependently inhibited by oxytocin via targeting V1a receptors [83]. In socially isolated prairie voles, daily oxytocin injection mitigates glucocorticoid levels and depression-like behavior [209]. Local microinjection of oxytocin into the mouse CeA, but not intracerebroventricular injection, attenuates social isolation-induced depression, and anxiety-like behaviors [208]. Therefore, the oxytocin system is closely associated with social isolation-induced affective disorders, and uncovering the possible mechanisms underlying the improvement by oxytocin will promote the clinical translation of oxytocin treatment for social isolation-induced disorders.

Oxytocin and Sensory Perception

A fundamental link between the oxytocin system, sensory perception, and sensory information processing has been established. Grinevich and Stoop systematically reviewed the modulatory effects of oxytocin signals on diverse sensory systems [210]. In the past several years, the complex and close interactions between oxytocin signals and sensory systems, especially the somatosensory, visual, and auditory systems, in modulating and maintaining brain functions have been further uncovered, so we mainly summarize recent advances in the interaction between oxytocin signals and sensory perception.

The sense of smell is an important channel for social communication. OxtRs are widely expressed in the olfactory system, including the olfactory bulb, anterior olfactory nucleus, and piriform cortex [47, 60, 93, 211]. In rats, OxtR activation has been found to increase the excitability of neurons in the anterior olfactory nucleus, which exerts top-down control of the inhibitory tone to mitral/tufted cells in the olfactory bulb and improves the signal-to-noise of odor responses [211]. In a subsequent study of the mouse olfactory bulb, oxytocin was found to reduce the spontaneous activity but increase odor stimulation-induced neural activity in mitral/tufted cells, which elevated the signal-to-noise ratio of mitral/tufted cells and enhanced social novelty behavior [93]. A recent study combining optogenetic activation, functional magnetic resonance imaging, and in vivo electrophysiological recording showed that endogenous oxytocin release activates the mouse anterior olfactory nucleus to enable the formation of cortical memory traces of familiar odors during recognition [212]. Therefore, oxytocin signals in both the olfactory bulb and anterior olfactory nucleus are devoted to promoting odor discrimination. The piriform cortex is a phylogenetically old brain region that plays a key role in our sense of smell. Oxytocin signaling in the mouse piriform cortex is required for social learning [60]. Exposure to female odor has been reported to increase Fos expression in PVN oxytocin neurons depending on the sexual experience of male rats [213]. The possible neural circuits underlying the modulation of hypothalamic oxytocin neurons by odors remain largely unknown.

Consistent findings over the past decades have documented the analgesic effect of oxytocin. Activating PVN and SON oxytocin neurons in rats exerts anti-nociceptive and anti-inflammatory effects [214]. Intranasal, intrathecal, and intravenous oxytocin applications have been reported to relieve pain in humans and animal models [215–217]. Many studies have shown that oxytocin targets the spinal cord [16, 217] and peripheral tissues [218], such as dorsal root ganglia and skin, to attenuate pain perception. Recent studies uncovered distinct projections of PVN oxytocin neurons to central regions to modulate pain. An endogenous oxytocin signal in the rat prelimbic PFC enhances the nociceptive response and inhibits acute and inflammatory pain by reducing synaptic inhibition [219]. In the mouse ACC, oxytocin application and activating oxytocinergic terminals from the PVN have analgesic and anxiolytic effects by inhibiting presynaptic long-term potentiation in neuropathic pain [194]. Parvocellular oxytocin neurons in the rat PVN, different from the spinal cord-projecting oxytocin neurons, project to the periaqueductal gray and excite GABA neurons to promote analgesia in the context of both inflammatory and chronic neuropathic pain [22]. Thus, oxytocin has an analgesic effect by targeting diverse central and peripheral regions. Since the current methods for pain assessment and treatment fall short of clinical needs [220], efforts to decipher the translational value of oxytocin and its analogs in treating nociceptive, neuropathic, or neoplastic pain are needed.

In addition to pain management, oxytocin is also involved in the gentle touch-associated positive valence. Skin-to-skin contact between mothers and infants stimulates oxytocin release in both mother and infant, which is associated with reduced maternal depressive symptoms and physiological stress [221]. In female rats, social physical contact, but not olfactory, visual, or auditory stimuli during social interaction, activates PVN parvocellular oxytocin neurons, further activating magnocellular oxytocin neurons to promote interfemale communication [21]. Although this study demonstrated the role of PVN oxytocin neurons in integrating somatosensory information and driving social interaction, the possible neural circuits underlying gentle touch information transmitted to PVN oxytocin neurons were unclear. Yu and colleagues found that social-touch-like tactile stimulation enhances mouse PVN oxytocin neuronal activity by elevating excitatory synaptic inputs from the periaqueductal gray to promote social interactions [222]. In addition to the periaqueductal gray, PVN oxytocin neurons also receive inputs from other somatosensory-related regions, such as the thalamus and ACC [2]; whether gentle touch may be transmitted to PVN oxytocin neurons by other neural pathways remains unknown.

OxtRs are expressed in the visual system, from the retina to the visual cortex in rodents [47, 50, 223]. A previous study reported that OxtRs are only expressed in retinal pigment epithelium in rodents and primates [223]. Hu and colleagues found the functional expression of OxtRs in mouse retinal GABAergic amacrine cells (ACs), especially dopaminergic ACs, and oxytocin reduces the activity of retinal bipolar cells via OxtRs and dopamine receptors [50]. The retinal oxytocin level exhibits a day/night rhythmic change [50, 224]. However, the source of retinal oxytocin and the physiological roles of oxytocin signals in the retina remain to be investigated. In addition to being regulated by oxytocin, one type of retinal ganglion cells (RGCs), intrinsically photosensitive RGCs have been found to modulate hypothalamic oxytocin neurons to promote brain development [225] and glucose metabolism [226]. The synaptic inputs from visual layers of the superior colliculus onto PVN oxytocin neurons enable the social transmission of maternal behavior [36]. In sensory cortices, including the visual cortex, oxytocin is crucial for wiring and refining neural circuits during development [59, 98]. Studies have shown considerable evidence of visual impairments in ASD, such as a hypersensitive pupillary light reflex [227], and oxytocin has been found to enhance pupil dilation and increase eye gaze toward novel stimuli [228, 229]. We infer that oxytocin signals may be involved in the visual impairments of ASD and oxytocin application may alleviate the visual functions of ASD patients.

A series of studies by Froemke and colleagues discovered the close intermodulation between oxytocin and the auditory system. They found that OxtRs are preferentially expressed in the left primary auditory cortex of female mice and oxytocin enhances the neuroplasticity of auditory cortical pup call responses for maternal recognition [58]. Infant calls with inter-syllable intervals from 75 ms to 375 ms elicit pup retrieval by maternal female mice, and co-housing naïve virgins with a dam and pups adjust virgins’ neurobehavioral responses to distressed pup calls via oxytocin-dependent synaptic plasticity in the auditory cortex [36, 230]. Infant cries increase the oxytocin level in human mothers [231]. Another study from Froemke’s lab dissected a noncanonical auditory circuit from the posterior intralaminar thalamus to PVN parvocellular oxytocin neurons, which relays infant distress sounds to induce the release of oxytocin and affects maternal behavior [232]. Hence, oxytocin and auditory systems closely interact to promote maternal behavior in mice. It is known that both oxytocin and auditory systems are also critical for many other brain functions, such as social-related behaviors, so they may be inter-modulated to maintain or promote other functions.