Oxytocin neurons in the paraventricular nucleus of the hypothalamus circuit-dependently regulates social behavior, which malfunctions in BTBR mouse model of autism

Oxytocin (OXT) a neuropeptide synthesized in the hypothalamic nuclei has a variety of function including socio-emotional processes in mammals. While the neural circuits and signaling pathways in central OXT converge in the paraventricular nucleus of the hypothalamus (PVN), we illuminate specific function of discrete PVN OXT circuits, which connect to the medial amygdala (MeA) and the bed nucleus of the stria terminalis (BnST) in mouse models. The OXTPVN→BnST projections are innervated from entire portions of the PVN, while those OXTPVN→MeA projections are asymmetrically innervated from the posterior portion of the PVN. Compared with OXT neurons in B6 wild type mice, BTBR mice that are recognized as a behavior-based autism model exhibited defect in the OXTPVN→BnST projection. We demonstrate that chemogenetic activation of OXTPVN→MeA circuit enhances anxiety-like behavior and facilitates social approach behavior, while activation of OXTPVN→BnST circuit suppresses anxiety-like behavior along with inhibiting social approach. This chemogenetic manipulation on the OXTPVN→BnST circuit proves ineffective in BTBR mice. Accordingly, chemogenetic activation of OXTPVN neurons that stimulate both OXT circuits induces OXT receptor expressions in both MeA and BnST as with those by social encounter in B6 mice. The induction of OXT receptor genes in the BnST was not observed in BTBR mice. These data support the hypothesis that OXT circuits serve as a regulator for OXT signaling in PVN to control socio-emotional approach/avoidance behavior, and a defect of OXTPVN→BnST circuit contributes to autism-like social phenotypes in BTBR mice.


Introduction
Social interaction is a fundamental behavior in all animal species for sustaining adaptive life and wellbeing. Engaging in positive social (i.e., prosocial) interactions robustly bene ts health across the lifespan [1][2][3]. The disruption of prosocial relationships has emerged as a major symptom in many psychiatric diseases and neurodegenerative diseases [4][5][6][7]. Oxytocin (OXT) is a neuropeptide involved in a key mechanism that regulates prosocial behavior, as evidenced in several different species, including humans [8][9][10]. Central OXT synthesis occurs in the paraventricular nucleus (PVN) and the supraoptic nucleus (SON) of the hypothalamus and in the accessory magnocellular nuclei of the hypothalamus, such as strio-and septo-hypothalamic nuclei [10][11]. The central OXT is mainly transported via axonal projections from the PVN, distributed to various brain regions, including the basal ganglia area, such as nucleus accumbens, septal nucleus, bed nucleus of the stria terminalis (BnST), medial amygdala (MeA), and paraventricular thalamic nucleus, and mid brain areas including the raphe nucleus and periaqueductal gray [12][13]. In particular, the OXT PVN → MeA projections are required for the processing prosocial approach behavior [14][15][16], while those OXT PVN → BnST contribute to social avoidance by circuitspeci c OXT action [17].
The PVN is a complex, multifunctional and multitransmitter nucleus, which modulate many CNS processes, including stress-related hormonal regulation [18]; thus, either social interaction or exploration of unfamiliar objects induces neural activation in the PVN [19]. Activation of the PVN neurons by these stimulus exposures also follows OXT release and OXT-ergic circuit-speci c processing, which would be determined by the type and intensity of stimuli exposed. Abnormalities in either the OXT peptide or its receptor have been associated with social de cits that are prevalent in many psychiatric and neurodevelopmental disorders, including autism spectrum disorders (ASD) [20][21]. While whole-brain (e.g., intranasal) treatment with OXT failed to affect prosocial behavior in mice [22] and alleviate ASD symptoms in clinical trials [23], circuit-speci c processing of OXT via the complex PVN network underlies for appropriate expressions of prosocial behaviors.
Mice are a highly social species that provide an exceptional tool for elucidating the neural mechanisms underlying prosocial behavior and its de cits [24]. Multiple lines of evidence have indicated that BTBR T + Itpr3tf/J (BTBR) mouse serves as a behavior-based ASD model resembling the characteristics of human ASD, including impaired prosocial interaction and persistent, repetitive behaviors [25][26]. A decreased neural activation of the PVN in BTBR mice in response to the presence of social stimuli compared to that in control mouse strain (e.g., C57BL/6J; B6 mice) has been documented [27][28][29].
Altered OXT neuronal morphology in the PVN was also evident in both sexes of BTBR mice compared to B6 mice [27,29]. We hypothesized that (1) OXT signaling in the PVN is relayed via circuit-speci c processes to control appropriate expression of prosocial behavior, and that (2) mal-formation or disconnection of these OXT circuits underlies social de cits exhibited by BTBR mice. Using immunohistochemical mapping of neural activity patterns and cell-type speci c tracing of OXT projections, we identify modi ed neural activity and altered OXT projecting patterns in the PVN of BTBR mice. Our data obtained by chemogenetic manipulation of PVN neurons demonstrate that two massive OXT projections, OXT PVN → MeA and OXT PVN → BnST , discretely regulate social and non-social behaviors, and a modi cation of these projections and binding receptor expressions are determinants for social de cits in BTBR mice. In summary, the data demonstrate a circuitry role for PVN OXT signaling in the expression of prosocial behavior, and reveal a possible OXT mechanism underlying de ciency in BTBR mice as an ASD mouse model.

Materials And Methods
Animals BTBR T + Itpr3tf/J (BTBR) mice (stock # 002282) and C57BL/6J (B6) mice (stock # 000664) of both sexes were originally purchased from The Jackson Laboratory (Bar Harbor, ME, USA), were used as subject and stimulus mice. They were bred and maintained in the colony room of a facility at the University of the Ryukyus Faculty of Medicine, Japan. The mice were housed in groups of three or four in standard shoebox cages (28 × 22.5 × 14 cm height) with water and food provided ad libitum. The vivarium was maintained at 50% humidity and 23 ± 1˚C, with a 12-h light/dark cycle (lights on at 08:00).
All procedures including animal care and use were performed in line with the National Institutes of Health Guide for the Care and Use of Laboratory Animals that comply with the ARRIVE guidelines and approved by the Institutional Laboratory Animal Care and Use Committee at the University of the Ryukyus Graduate School of Medicine.

Experimental design
For full description of methods, see Supplementary information. Brie y, we rstly measured neural activity in the PVN sub-portions (rostral to caudal) and related brain regions of B6 (n = 22) and BTBR mice (n = 21) in response to a novel social or object encounter. Then we conducted a series of behavioral tests to assess social and non-social pro les relevant to autism-like behaviors in B6 and BTBR mice. Our test battery consisted of a set of anxiety/locomotor task (the elevated plus maze), social behavior task (the modi ed three-chamber test), and defensive behavior task (odor defense test). These subject mice including male and female received a bilateral AAV injection for DREADD manipulation and assigned into Testing was performed in tandem with small groups of mice consisting of three-to-four cohorts of animals at different times. Following a completion of behavioral testing, AAV transduction in the targeted brain sites was histologically veri ed by microscopy with transduced uorescence and immunohistochemical staining of OXT neurons. Finally, mRNA expressions of targeted genes in social processing brain regions following a social encounter or chemogenetic activation of OXT PVN neurons were evaluated in B6 (n = 32) and BTBR mice (n = 30).

PVN neuronal activity in response to social and non-social encounter
We rstly con rmed neural responses of the PVN sub-regions to social and non-social stimuli in B6 and BTBR mice. To map neural activity patterns in the PVN following exposure to novel social and non-social (object) stimuli, we performed immunohistochemistry for c-Fos across the rostral-caudal divides of the PVN (Fig. 1A,C) and its projecting regions; MeA (Fig. 1F), BnST (Fig. 1D,E), and LS ( Fig. 1G) in B6 and BTBR mice that were exposed to a social (a novel same-sex mouse) or non-animated object, or remained in a home cage as a no exposure control (Fig. 1H)(see Fig. S1). For B6 mice, across the rostral to caudal portions of the PVN, signi cantly more cells expressing c-Fos in response to either social or non-social stimuli compared to the control condition were documented (Fig. 1A,B,C). For BTBR mice, however, the induction of c-Fos was found through the middle to caudal portions of the PVN only by an object, but not social, exposure. This effect mapped c-Fos induction disappeared in the rostral portion of the PVN. In the projecting regions from the PVN, the numbers of c-Fos + cells were increased both by social and nonsocial (object) exposures in B6 mice, and these were blunted in BTBR mice. Accordingly, sni ng/contact investigation towards exposed stimuli was demonstrated in both B6 and BTBR mice; however, these were attenuated in BTBR mice when confronted with a social stimulus (Fig. 1I). Therefore, a blunted c-Fos induction in targeted regions in response to social encounter exhibited by BTBR mice is covariant with decreased social investigation.
Oxytocinergic neural projection patterns reveal strain difference OXT neurons of the PVN are primarily responsible for the innervation of forebrain regions [30]. While dendritic OXT release are also evident [31], OXT neurons of the PVN send axonal projections to various distances of brain regions, including the basal forebrain area, such as septal nucleus, BnST, and MeA, and several thalamic and hind brain regions [12][13]. We chose two massive OXT projecting regions, MeA and BnST, based on functional contribution to the regulation of social behavior [14,[16][17]. Immunohistchemical analysis con rmed that OXT neurons are distributed across the rostral to caudal portions of the PVN in both B6 and BTBR mice, while densities of OXT neurons are lower in BTBR mice than those in B6 mice constantly across the PVN sub-portions ( The OXT PVN → MeA circuits facilitate anxiety-like behavior differs from the OXT PVN → BnST circuits The EPM is a popular test method to measure anxiety along with locomotor activity in rodent models [32]. BTBR mice display a variable pro le of anxiety-like behaviors in the EPM (e.g., increase or decrease) when compared with B6 strains [25,[33][34]62]. In present study, both male and female BTBR mice exhibited a low anxious pro le in the EPM, with an increased locomotor activity and enhanced open arm time compared with control B6 mice (Fig. S3). Otherwise, sex differences of the behavioral performance in BTBR strain as well as B6 strain were not consistent in this test.
The PVN is a center brain region for stress response [35] and anxiety behavior [36]. While OXT is known to play a modulatory role in anxiety behavior, the effects of exogeneous OXT are still variable [37] and the brain regions that transmit OXT information from the PVN for proper execution of anxiety behavior have not been identi ed. We applied chemogenetic manipulation on the PVN neurons to determine the cell-type and projection-speci c function of the PVN circuits. In particular, we bilaterally injected AAV vector with low transduce tropism via ubiquitin promoter (AAV8-EF1a:hM3Dq; Ns PVN, labeling global PVN neurons) or OXT speci c promoter (AAV8-pOXT:hM3Dq; OXT-PVN) into the PVN (Fig. 3A and Fig. S2). We also injected AAV vectors with retrograde tropism via OXT promoter (retro-AAV-pOXT:hM3Dq) at the MeA (OXT-MeA) and the BnST (OXT-BnST) solely, to manipulate OXT PVN → MeA and OXT PVN → BnST projections, respectively. In this scenario, cell expressing hM3(Dw) are categorized as various cell types in the PVN  (Fig. 3E). There was no impact observed on locomotor activity in both mouse strains under the current DREADD manipulation conditions (Fig. 3B,C). These data suggest that activation of OXT PVN → MeA neurons is consistently anxiogenic in both B6 and BTBR mice, while OXT PVN → BnST neurons is anxiolytic in B6 mice but no effect in BTBR mice.

OXT circuits determine controlling prosocial behavior
The OXT neurons play a critical role in processing social approach and avoidance in a variety of contexts, which implies circuit-speci c OXT action. The OXT PVN → MeA neurons are responsive to facilitate social approach to conspeci cs [15][16], while the OXT PVN → BnST neurons promote social avoidance in female California mice [38] and reduce social approach in stressed mice [17]. We provide evidence that circuitspeci c control of OXT in prosocial behavior is also strain-dependent.
As consonantly shown in various literatures, both male and female BTBR mice demonstrate behavioral traits relevant to autism, including reduced social approach towards familiar (c.f., same strain) or unfamiliar (c.f., different strain) social stimuli, compared to B6 controls (Fig. S3). Using chemogenetic manipulation, we uncovered cell-type and circuit-dependent regulation of these social behavioral traits in B6 and BTBR mice. When confronted with a same-sex familiar opponent (i.e., B6 stimulus mouse), activation of OXT PVN neurons (OXT-PVN), but not those of global PVN neurons (ns PVN), decreased social approach of B6 mice (Fig. 3H). An activation of speci c OXT PVN → BnST neurons (OXT-BnST) but not OXT PVN → MeA neurons (OXT-MeA) reduced social approach in B6 mice (Fig. 3H). However, these chemogenetic activations of PVN neurons had little impact on social approach to familiar (i.e., BTBR stimulus mouse) opponents in BTBR mice (Fig. 3I). These behavioral pro les in response to chemogenetic activation of PVN neurons differ dependent on familiarity of the opponents. When Distinct PVN circuits regulate non-social behavioral pro le relevant to autism-like traits BTBR mouse model has been associated with de cits in social communication and a pronounced engagement in repetitive behavior [25,39]. Excessive self-grooming and persistent digging and burying the oor bedding materials (e.g., sawdust or corncob) or novel obstacles (e.g., embedded marbles) are considered to represent these later pathological states [40,41]. The characteristics of BTBR mice suggest that variant circuitry controlling exploratory and defensive behaviors underlying their atypical behaviors needs to be elucidated. We conducted an olfactory defense test, in which mice display defensive behaviors towards exposures of neutral (ie. banana odor) and predatory (ie., fox urine compound) odor cues [42]. B6 mice displayed exible sniff-immobile defensive patterns in tune with the type of odor stimuli, by decreased sni ng and increased immobile toward predatory aversive odor compared to those toward neutral odor (Fig. S3E). While BTBR mice exhibited reduced susceptibility of sniff/immobile response, they showed attenuated immobile time compared to B6 mice (Fig. S3F). In addition, BTBR mice exhibited a heightened levels of digging the bedding material and self-grooming compared to B6 mice (Fig. S3G,H).
To rule out the PVN circuitry function in these defensive components, we assessed the impact of chemogenetic manipulation on these behaviors. B6 mice altered their sniff-immobile pro le following an activation of OXT PVN → BnST neurons (i.e., PVN-BnST) when confronted with a neutral natural odor (Fig. 4A,C), and following an activation of the global PVN neurons (i.e., ns PVN) when confronted with predatory odor (Fig. 4E,G). In contrast, BTBR mice did not show any change in sniff-immobile strategy towards odor stimuli via either chemogenetic PVN activations (Fig. 4F,H). Furthermore, an activation of the global PVN neurons (i.e., ns PVN) stimulated increased digging and self-grooming behavior in B6 mice (Fig. 4I,K) and self-grooming in BTBR mice (Fig. 4J,L). These data indicate that the activation of OXT PVN → BnST neurons induces enhanced exploratory, but not defensive properties in B6 mice and this circuitry may not be responsive in BTBR mice. Global, but non OXT, PVN activation produces modi cation of other defensive behaviors, including immobile defense and digging and grooming behaviors, suggesting that these behavioral components are regulated by non OXT PVN neurons. mRNA expression pro les following chemogenetic OXT PVN activation and social encounter To validate OXT circuitry reaction patterns, we conducted quantitative polymerase chain reaction (qPCR) analyses on selected brain regions of B6 and BTBR mice exposed to either a social stimulus (a novel same-sex mouse) or a chemogenetic activation of OXT PVN neurons (OXT-PVN) (Fig. 5). We analyzed genes that are relevant to OXT neuronal activation via the PVN circuit, including c-fos (as a neural activity marker), OXT, OXT receptors (OXTR), vasopressin (AVP), and vasopressin 1a receptors (AVPR1a) in the PVN and its projecting regions, MeA, BnST, and septum.
In both B6 and BTBR mice, an increase in c-fos mRNA expression was observed in all selected regions following a social encounter, and in the PVN by chemogenetic OXT PVN activation. Social encounter induced OXT gene expression in the PVN of both strains and OXTR mRNA in the MeA and BnST of B6 mice, but not of BTBR mice. In addition, the social encounter increased AVPR1a gene expression in the septum of B6 and BTBR mice. Chemogenetic activation of OXT PVN neurons downregulated OXT gene expressions in the PVN and AVPR1a expressions in the BnST of both strains. The chemogenetic OXT PVN activation also upregulated OXTR expressions in the MeA of both strains, while in the BnST, changes in the OXTR mRNA expression (upregulation) were only observed in B6 mice but not in BTBR mice, which represents only strain difference in mRNA expression pattern following a chemogenetic activation examined in the present study.

OXT circuit-dependent regulation of socio-emotional behavior
Hypothalamic neuronal network plays a predominant role in valence processing of socio-emotional behaviors [43][44]. OXT neurons in the PVN are responsive to dynamic control of approach/avoidance behaviors in social and non-social contexts with a circuit-dependent fashion. We provide strong evidence that the OXT PVN → MeA circuit activates social approach behavior, along with anxiogenic pro le, while the OXT PVN → BnST circuit inhibits social approach, accompanied by anxiolytic response. Social encounter with an unfamiliar counterpart induces c-Fos across the rostral-to-caudal portions of the PVN and its projecting regions, including the BnST (only anterior portion) and MeA, in which the induction of OXT receptor genes was also evident. Chemogenetic stimulation of OXT PVN neurons that concurrently activates both OXT PVN → BnST neurons and OXT PVN → MeA neurons upregulates the OXT receptor gene expression in both the BnST and MeA; however, suppresses social approach behavior. Therefore, OXT PVN → BnST signal that suppresses social approach is signi cantly effective than OXT PVN → MeA signal that generally facilitates social approach when both circuits are simultaneously activated. Social encounter per se induces approach behavior accompanied by the upregulation of OXT receptor gene expression and c-Fos in the MeA and anterior BnST; thus, the OXT PVN → MeA signal takes priority to control their behaviors over those OXT PVN → BnST signal in a prosocial context. A possible key determinant for OXT circuit balance would act by an inhibitory mediator in the downstream of OXT pathway in the BnST. The posterior BnST and MeA are anatomically characterized as a set of interneuronal nuclei, called as the extended amygdala [45]. It is likely that certain interneuronal connection relayed between the BnST and MeA acts to adjust output balance of neural signals from investigatory approach to non-social stimuli (i.e., OXT PVN → BnST signal) to craving approach to prosocial stimuli (i.e., OXT PVN → MeA signal).
Accordingly, the MeA-posterior BNST circuits projecting to the hypothalamus are responsible for innate social and predator-defense behaviors [46][47]. Further research is needed to elucidate whether molecular pathway of the OXT circuits maintains balancing of their output within the PVN interneuronal connections or across the inter-regional connections, such as the MeA-posterior BnST circuits.
Defect of OXT neural circuits accompanied by social de cits in an ASD mouse model BTBR mice exhibit typical behavioral pro le resembling the characteristics of human ASD, including impaired prosocial interaction and persistent, repetitive behaviors [25][26]. Although a target neuronal circuit responsible for controlling these social and non-social de cits exhibited by ASD animal models have been investigated for more than a decade, little is known about the circuit mechanism underlying these behavioral impairments. BTBR mice exhibit decreased densities of OXT neurons across the PVN sub-portions. Furthermore, the OXT PVN → BnST circuit, which is massively expressed in B6 mice, appears to be defected in BTBR mice. These malformations of OXT circuit would be linked to blunted responses of neuronal activity patterns across the PVN and its projecting regions following a social encounter. While there is certain dissociation between protein level (e.g., c-Fos IHC) and mRNA level (e.g., c-fos qPCR) in response to a social stimulus exposure, the OXT PVN → BnST circuit is documented to be defective but the OXT PVN → MeA circuit is still intact in BTBR mice.
We recently reported a circuit dysfunction in BTBR mice that shows a compromised circuit in the posterior BnST to the lateral habenula, possibly via attenuating vasopressinergic signal/projection, which contributes to a lack of coordinative responses of social signaling by BTBR mice in a social context [29]. Our present data can extend it to an OXT circuit defect of BTBR mice for controlling prosocial behavior.
BTBR mice exhibits agenesis of several brain structures, including the absence of corpus callosum, atrophic of hippocampus, and glial cells overexpression in the cingulate cortex [29,48]. A MRI imaging analysis revealed that decreased cortical and thalamic grey matter volume along with a reduction of cortical thickness are documented in BTBR mice [49]. It is indicated that one of these morphological mutations occurs in developing brain network of BTBR mice, resulting in a subsided OXT signals and coordinative circuit function that are required to express appropriate approach/avoidance behavior in a social and non-social context. Non OXT neurons in the PVN are involved in controlling nonsocial, defensive components A pathological repetitive behavior is one of the core symptoms diagnosed as ASD [50]. An underlying mechanism for this exploited behavior would be distinguishable from those for social de cit, although it is still under debate [51][52]. In rodent models, excessive self-grooming and persistent digging/burying the cage bedding materials are considered to recapitulate pathological repetitive behavior states expressed in ASD patients [41]. BTBR mice have repeatedly con rmed these excessive behavioral pro les [25,33,53]. Interestingly, nursing with a different, non-ASD, strain can modify social de cits in BTBR mice, but cannot improve their excessive self-grooming [54], suggesting a possibility that different behavioral domains relevant to ASD symptom may be regulated by distinct brain mechanisms. Our data clearly demonstrated that OXT neurons in the PVN are not involved in the responsible mechanism for controlling excessive behavioral pro les, but rather non-OXT neurons in the PVN can stimulate these behaviors including self-grooming and digging/burying cage bedding.
A growing number of pharmacological manipulations on excessive self-grooming behaviors of BTBR mice have fault to archive consistent effects, including the use of cholinergic, glutaminergic, or serotonergic agents [53,[55][56][57]. Highly complicated, sequential organization of grooming (and related defensive) behaviors raises a hypothesis that various neural circuits regulating discrete components of complex behaviors (e.g., motor, strategy, and sequencing) are involved. Non-OXT PVN neurons offer to facilitate grooming and related obsessive components in both B6 and BTBR mice. The PVN is a multitransmitter nucleus controlling various biological functions [35]. A major stress-related hormone, corticotropin-releasing hormone (CRH) produced in the PVN, are known to induce self-grooming [58][59]. Furthermore, virally induced selective manipulation of vasopressin neurons in the PVN can modify selfgrooming in mouse models [60]. These suggest that stress-related, OXT-independent mechanism would be a key mediator for controlling compulsive and defensive components of behavior in the PVN.

Conclusions
The data demonstrate a circuit-dependent role for OXT signaling in the PVN neurons in controlling social and emotional behaviors and illustrate that the coordinative regulation of two discrete OXT circuits; the OXT PVN → BnST and OXT PVN → MeA circuits, underlies to control these behaviors. A functional defect of the OXT PVN → BnST signal coincides with prosocial behavioral de cits in BTBR mouse model of autism. The study provides novel information about the coordinative role of OXT circuits and establish a foundation for further investigations that delineate the circuit mechanisms of OXT signaling impacting prosocial processes and social de cits relevant to ASD symptom [61].   Non-OXT PVN neurons are su cient to mediate non-social, defensive behavior in both B6 and BTBR mice.
A,C When exposed to a novel neutral odor (isoamyl acetate), chemogenetic activation of OXT PVN