Oxytocin neurons mediates the effect of social isolation via the VTA circuits

Social interaction during adolescence strongly influences brain function and behaviour, and the recent pandemic has emphasized the devastating effect of social distancing on mental health. While accumulating evidences have shown the importance of the reward system in encoding specific aspects of social interaction, the consequences of social isolation on the reward system and the development of social skills later in adulthood are still largely unknown. Here, we found that one week of social isolation during adolescence in mice increased social interaction at the expense of social habituation and social novelty preference. Behavioural changes were accompanied by the acute hyperexcitability of dopamine (DA) neurons in the ventral segmental area (VTA) and long-lasting expression of GluA2-lacking AMPARs at excitatory inputs onto DA neurons that project to the prefrontal cortex (PFC). Social isolation-dependent behavioural deficits and changes in neural activity and synaptic plasticity were reversed by chemogenetic inhibition of oxytocin neurons in the paraventricular nucleus (PVN) of the hypothalamus. These results demonstrate that social isolation has acute and long-lasting effects on social interaction and suggest that these effects are mediated by homeostatic adaptations within the reward circuit.


Introduction 1 2
The experience of social interaction during postnatal development and adolescence is 3 fundamental for setting the basis for social life, and the deprivation of social experience (hereby 4 defined social isolation) impacts the survival of all species, as suggested by the adverse effects 5 of the massive social isolation imposed by the COVID-19 crisis on mental health 1 . Identifying 6 the possible neural mechanisms that underly the negative consequences of social isolation may 7 help to prevent and treat mental disorders. 8 In rodents, depending on the duration of juvenile social isolation, increased or decreased 9 sociability in adulthood has been reported, suggesting that adolescence is a sensitive period for 10 the establishment of social behaviour later in life [2][3][4] . Few studies have examined the acute 11 effects of social isolation and the cellular and circuit mechanisms that regulate the long-lasting 12 effects of social isolation. For example, short-term isolation has been shown to increase social 13 interaction in rats 5,6 . 14 Social interaction is a rewarding experience with reinforcing properties, and recent studies have 15 highlighted the necessity and sufficiency of dopamine (DA) neurons of the ventral segmental 16 area (VTA) to promote social interaction 7,8 . Brief periods of acute social isolation have been 17 reported to activate midbrain regions in humans 9 and to increase the activity of DA neurons 18 within the dorsal Raphe nucleus (DRN) 10 . Intriguingly, while in rodents, 24 hr of social 19 isolation does not change synaptic strength onto DA neurons of the VTA 10 , in humans, the 20 response of the VTA to social cues after brief isolation is increased 9 . These data suggest that 21 changes within DA neurons in the VTA may be the substrate for social craving caused by acute 22 social isolation. The neuronal mechanisms and long-term consequences remain uninvestigated. 23 DA neurons of the VTA contribute to reward-seeking behaviour, motivation and reinforcement 24 learning, and their activity is controlled upstream by several brain structures 9 , each of which 25 may contribute to distinct behavioural aspects. Interestingly, DA neuron activity is not only 26 under the control of glutamatergic and GABAergic inputs but also tightly regulated by 27 neuromodulators that act on G protein-coupled receptors (GPCRs). Oxytocin is a neuropeptide 28 released by neurons within the paraventricular nucleus (PVN) of the hypothalamus that directly 29 projects to the VTA. Oxytocin in the VTA activates oxytocin receptors on DA neurons, 30 regulates their activity 11-13 and favours social reward. Indeed, the presence of a conspecific 31 activates the oxytocin system, which increases oxytocin release, activates DA neurons of the 32 VTA and therefore promotes the initiation and maintenance of social interaction 14,15 . 33 Interestingly, although it has been hypothesized that oxytocin senses changes in the environment and facilitates behavioural stability to better adapt to changes 16 , the role of 1 oxytocin neurons in the behavioural consequences of social isolation remains largely unknown. 2 3 Results 4 5 We first characterized the acute consequences of short-term social isolation on social 6 interaction during adolescence. Male mice were weaned at post-natal day (P) 21 and then 7 isolated between P28 and P35 ( Figure 1A). The last day, we exposed the experimental mice to 8 an unknown sex-matched juvenile conspecific or a novel object in a direct free-interaction task 9 ( Figure 1B). Isolated mice spent more time interacting with the conspecific than the grouped 10 control mice ( Figure 1C). Conversely, object exploration did not differ between the two groups, 11 indicating that social isolation preferentially affects social exploration ( Figure 1B To further investigate the consequences of social isolation on different aspects of social 16 behaviour, we used a three-chamber interaction task to characterize sociability and social 17 novelty preference ( Figure 1E). We found that both the socially isolated and grouped mice 18 spent significantly more time investigating a juvenile conspecific over a novel object ( Figure  19 1F, Figure 1 supplement 1 C-E). In the second part of the test, in contrast to the control mice, 20 the socially isolated mice spent the same amount of time interacting with familiar and unknown 21 conspecifics ( Figure 1G, Figure 1 supplement 1 C,F,G), indicating that social isolation 22 induces deficits in social novelty preference. Interestingly, after social isolation, mice presented 23 deficits in habituation ( Figure 1H) when repeatedly exposed to the same juvenile conspecific 24 for 4 consecutive days ( Figure 1I, J). Novel object recognition (Figure 1supplement 2 A-F) 25 and behaviour in the elevated plus maze (Figure 1 supplement 2 G-K) did not differ between 26 the socially isolated and control mice. To investigate whether social isolation-dependent 27 behavioural deficits are age dependent, we investigated isolated mice during adulthood (7 days 28 of isolation; P53-P60, Figure 1 supplement 3A). We found that isolated adult mice spent less 29 time interacting with a conspecific than the control mice, while object exploration ( were no different. 32 Altogether, these data indicate that adolescence is a critical period for the development of social 33 behavioural skills and that social isolation during this period increases social interaction at the expense of an impaired social novelty preference and altered habituation to interact with a 1 familiar conspecific. 2 In humans, acute social isolation has been reported to have a rebound effect on conspecific 3 interaction, accompanied by an increase in the response of reward circuits and, in particular, 4 the VTA in response to social cues 9 . In line with these discoveries and to further investigate 5 the related neural mechanisms, we measured the excitability of DA neurons in the VTA after 6 social isolation. 7 Mice were isolated or maintained in group housing from P28 to P35, and acute brain slices 8 were subsequently prepared. (Figure 2A). We first performed whole-cell patch clamp recording 9 to measure the excitability of putative DA neurons in the VTA. We observed an increase in 10 excitability after social isolation without a change in the resting membrane potential ( Figure  11 2B, D). Aiming to identify the key upstream brain regions responsible for regulating the 12 activity of DA neurons, we performed cFos analysis of brain slices 7 days after social isolation 13 and observed an increase in PVN neurons immunopositive for cFos, suggesting increased 14 activity ( Figure 3A, B). Since PVN neurons have been shown to regulate the activity of DA 15 neurons in the VTA, we tested the hypothesis that these neurons are the master regulator of DA 16 neuron activity during social isolation. To test our hypothesis, mice were first injected with 17 CTB-488 in the VTA on P21 and then isolated between P28 and P35, after which PVN brain 18 slices were prepared. PVN neurons projecting to the VTA (PVN-VTA) showed increased 19 excitability compared to that of the control group when recorded ex vivo via the whole-cell 20 patch clamp technique ( Figure 3C-E). 21 Because of their role in social behaviour and the regulation of DA neuron activity 11,18 , we then 22 focused our analysis on oxytocin neurons within the PVN. By confocal imaging quantification, 23 we observed an increase in oxytocin neuron density in the PVN after social isolation ( Figure  24 3F, G), suggesting that the increase in oxytocin signalling during isolation can promote DA 25 neuron hyperexcitability and lead to increased social interaction. To prove the causal link 26 between oxytocin neurons and DA neuron activity during social isolation, we crossed OXT-27 Cre mice with hM4Di-loxP mice (OXT-Cre:hM4Di-loxP) to express a designer inhibitor 28 receptor activated exclusively by designer drugs expressed under the Cre promoter in OXT-29 positive cells ( Figure 3H). Clozapine-N-oxide was dissolved in drinking water and 30 administered during social isolation (or in parallel to grouped control mice). We obtained 31 whole-cell patch clamp recordings from putative DA neurons in the VTA and observed the 32 rescue of excitability in cells recorded from isolated mice treated with CNO but not isolated 33 mice treated with vehicle ( Figure 3I-M, Figure 3 supplement 1). Finally, to prove causality between neuronal hyperexcitability and behaviour, we treated OXT-Cre:hM4Di-loxP mice 1 with CNO or vehicle and compared the time spent in interaction with a novel conspecific after 2 social isolation ( Figure 3N). We also used grouped mice treated with CNO or vehicle as 3 controls. As expected, we found an increase in interaction time in isolated versus grouped mice, 4 while no difference was observed between mice treated with CNO independent of housing 5 ( Figure 3O). Altogether, the data presented here indicate that increased social interaction after 6 social isolation is the consequence of the increased excitability of oxytocin neurons of the PVN 7 and suggest that this effect is mediated by the increased activity of DA neurons within the 8 VTA. 9 We next investigated whether social isolation during adolescence has long-lasting 10 consequences and the consequent neural mechanisms. After 7 days of social isolation from P28 11 to P35, the mice were regrouped, and social behaviour was then tested during adulthood ( Figure  12 4A). We still observed an increase in social interaction in mice isolated during adolescence 13 compared to the control group ( Figure  habituation were similar between grouped and regrouped mice ( Figure 4E, G), although during 16 the first interaction, regrouped mice interacted more ( Figure 4F). These behavioural data 17 indicate that acute social isolation during adolescence leads to a long-lasting increase in social 18 interaction during adulthood. Remarkably, inhibition of oxytocin neuron activity during social 19 isolation was sufficient to block the long-lasting consequences of social isolation and restore 20 social behaviour ( Figure 4H-I). These data not only indicate that an isolation-dependent 21 increase in social interaction occur during adulthood but also support the role of oxytocin 22 neurons in regulating social craving after isolation. 23 To investigate the related neural mechanisms underlying the long-lasting consequence of social 24 isolation, we performed whole-cell patch clamp recording of putative DA neurons in the VTA. 25 Intriguingly, neuronal excitability in adulthood did not differ between isolated and control mice 26 , suggesting that the long-lasting behavioural consequences of 27 social isolation, although induced by neuronal excitability, are exerted by a different neuronal 28 mechanism. Neurons undergo different mechanisms of homeostatic adaptation to overall 29 changes in neuronal activity, and many studies have reported in vivo scaling triggered by 30 sensory manipulation and exerted by the regulation of calcium-permeable (CP)-AMPARs 19 . 31 We therefore characterized whether a long-lasting increase in social interaction after social 32 isolation during adolescence is accompanied by changes at the level of synaptic transmission 33 at excitatory inputs onto DA neurons in the VTA. We obtained whole-cell patch clamp recordings from putative DA neurons while pharmacologically isolating excitatory 1 transmission. Considering the output-dependent heterogeneity of DA neurons and the 2 previously identified neuronal type specificity in the form of experience-dependent synaptic 3 plasticity 20,21 , we decided to characterize long-lasting, isolation-dependent effects on the 4 synaptic plasticity of DA neurons in the VTA in an output-specific manner. To that end, we 5 injected choleratoxin in the prefrontal cortex (PFC) or nucleus accumbens (NAc) and then 6 obtained whole-cell path clamp recordings from the identified neuronal population ( Figure 5A, 7 F). While we observed no difference in the ratio of AMPAR-and NMDAR-mediated currents 8 between control and isolated mice ( Figure 5B, C), we observed an increase in rectification 9 index (RI) at excitatory inputs onto DA neurons projecting to the PFC in isolated mice ( Figure  10 5D, E). No change in the AMPA/NMDA ratio or the RI was observed at excitatory inputs onto 11 DA neurons projecting to the NAc ( Figure 5G-J). We then tested whether the expression of 12 CP-AMPARs on DA neurons projecting to the PFC is the consequence of the increased 13 excitability of oxytocin neurons in the PVN during social isolation. We isolated OXT-14 Cre:hM4Di-loxP mice during adolescence, treated them with CNO or vehicle during isolation, 15 regrouped them after 7 days until adulthood and recorded excitatory transmission from DA 16 neurons projecting to the PFC from acute VTA slices ( Figure 5K). Notably, we observed that 17 RI was normalized when the activity of the oxytocin neurons was chemogenetically reduced 18 during social isolation ( Figure 5L, M). These data indicate that social isolation during 19 adolescence leads to long-lasting effects on social interaction that are accompanied by oxytocin 20 neuron-dependent changes in synaptic transmission at excitatory inputs onto DA neurons 21 projecting to the PFC. 22 Finally, to assess the causal link between the presence of CP-AMPARs and isolation-dependent 23 changes in social interaction, the VTA in each mouse was cannulated, and we injected a CP-24 AMPAR antagonist (NASPM) into the region 10 min prior to the direct interaction task ( Figure  25 6A, B). While NASPM did not affect the interaction time in control mice, the inhibition of CP-26 AMPARs in the VTA was sufficient to normalize social interaction in the regrouped mice 27 ( Figure 6C). The data presented here show that the increased activity of oxytocin neurons 28 during social interaction induces synaptic scaling exerted by the presence of CP-AMPARs and 29 that these receptors are responsible for increased social interaction during adulthood.
Discussion 1 2 Social isolation is an adverse experience across social species that has long-lasting behavioural 3 and physiological consequences 22 . Although anxiety, depression and obsessive-compulsive 4 behaviour generally emerge after prolonged social isolation, it is evident that a short period of 5 deprivation from conspecific interaction may also trigger aversive consequences. The negative 6 effects of social isolation are particularly evident for adolescents. Indeed, adolescence is a 7 critical period for not only social interaction but also the emergence of psychiatric diseases, 8 and social isolation during adolescence can exacerbate mental health problems. Why is 9 adolescence especially affected, and what are the circuit mechanisms underlying the negative 10 consequences of social isolation? 11 Social interaction is a basic need across species, and rodents are excellent models in which to 12 investigate the neural bases of social interaction and social isolation. During adolescence, the 13 development of social and cognitive skills indeed depends on peer-to-peer interactions 23,24 , 14 and acute social isolation in rodents is sufficient to cause behavioural abnormalities, including 15 increased vulnerability to drug addiction and depressive-like behaviour, later in life. Here, we 16 show that one week of social isolation during adolescence is sufficient to generate a rebound 17 increase in social interaction in male mice independent of whether the conspecific is novel or 18 familiar. Indeed, the increase in interaction is accompanied by deficits in social novelty 19 preference and social habituation. These data support the assumption that social interaction is 20 a need and that the absence of social cues generates a craving response similar to that generated 21 by food cues after fasting 9 . Indeed, we could speculate that social isolation during adolescence 22 results in an intense urge to interact with whatever conspecific is present. 23 Previous studies have indicated the role of DA neurons in the VTA in social behaviour 7,25,26 24 and shown that chronic isolation alters DA levels in the NAc 27 . Remarkably, acute social 25 isolation during adulthood changes the synaptic strength at excitatory inputs onto DA neurons 26 of the DRN in rodents, leaving synaptic transmission onto DA neurons of the VTA unaltered 27 17 Interestingly, human exposure to a social cues after acute isolation evokes activity in the 28 VTA 9 , suggesting that while the loneliness-like brain state induced by social isolation is 29 mediated by DA neurons within the DRN, DA neurons in the VTA mediate social craving. Our 30 data indicate that one week of isolation during adolescence is sufficient to increase the overall 31 excitability of neurons within the mesocorticolimbic system. We next investigated the 32 underlying mechanisms and found that the activity of oxytocin neurons within the PVN is 33 causally linked to neuronal excitability within the reward system and to the behavioural consequences of social isolation. Previously, the release of oxytocin in the VTA was shown to 1 increase DA neuron firing within the VTA 18 . Furthermore, restoration of oxytocin signalling 2 in the VTA in an autism spectrum disorder (ASD)-related mouse model was sufficient to rescue 3 social novelty responses 28 . Taken together, these data suggest that oxytocin neurons represent 4 the neural signature of social craving. 5 We next chose to investigate whether acute social isolation during adolescence has long-lasting 6 consequences. Isolated mice spent more time interacting with a conspecific than control mice 7 did one month after being regrouped, indicating that the rebound increase in social interaction 8 is a long-lasting effect of isolation. Recently, it was reported that two weeks of social isolation 9 after weaning altered the pathway from the medial PFC to the posterior paraventricular 10 thalamus, leading to decreased sociability in adult mice 2 .Although the differences at the 11 behavioural level between the two studies can be explained by the duration of social isolation, 12 together, these data show that adolescence is a critical period for the establishment of social 13 behaviour and that isolation during this period alters sociability during adulthood. 14 Increased social interaction during adulthood is accompanied and causally linked to the 15 increase in GluA2-lacking AMPARs at excitatory inputs onto DA neurons projecting to the 16 PFC. GluA2-lacking AMPARs are calcium permeable (CP) ionotropic receptors with larger 17 single-channel conductance and that can undergo voltage-dependent blockade by polyamines 18 29,30 . These receptors are inserted at excitatory synapses after exposure to addictive drugs 31 and 19 have been proposed to mediate the incubation of drug craving 32 . Interestingly, it has also been 20 shown that stress induces the insertion of these receptors, and local blockade of GluA2-lacking 21 AMPARs was shown to attenuate stress-induced behavioural changes in rodents 33,34 . 22 Furthermore, the presence of these receptors changes the rules for the induction of plasticity 35 23 and mediates in vivo synaptic scaling 36 . Interestingly, we show here that prolonged activation 24 of DA neurons induced by social isolation promoted the insertion of CP-AMPARs and that in 25 vivo blockade of these receptors was sufficient to rescue social isolation-induced behavioural 26 deficits. Together, all these findings indicate that GluA2-lacking AMPARs are important 27 therapeutic targets for the treatment of maladaptive motivated behaviour and suggest that 28 inhibitors of these receptors may counteract the negative consequences of social isolation. 29 By investigating the short-term and long-term consequences of social isolation during 30 adolescence, our work here contributes to our understanding of how social isolation impacts 31 neural circuits and behaviour. Since social isolation has a tremendous impact on mental health 32 and increases vulnerability to psychiatric diseases, our work is relevant to the identification of 33 new therapeutic approaches to fight loneliness and social craving.

Animals and experimental paradigms 3
The study was conducted with male wild-type (WT) and OXT-hM4Di mice, under 12h light-4 dark cycle (7 a.m. -7 p.m.) with free access to food and drink. For adolescenceadolesce social 5 isolation, mice were weaned at P21 and kept in groups until P28. Subsequently mice were 6 single used until P35. For 24h isolation mice were single used between P34 and P35. For 7 social isolation in adulthood, mice were kept in groups until P53 and then were single housed 8 until P60. For the long-lasting effect of SI, mice were single housed between P28 and P35 9 and subsequently regrouped until P60. When in group, mice were housed 2 by 2. The 10 experimental mice were randomly assigned to the different groups. All the procedures 11 performed at UNIGE complied with the Swiss National Institutional Guidelines on Animal 12 Experimentation and were approved by the Swiss Cantonal Veterinary Office Committees for 13 Animal Experimentation. 14 15

Direct free interaction task 16
Mice were let to freely explore for 10 minutes the arena (clean home-cagehomecage) and 17 subsequently an unfamiliar sex-matched conspecific social stimulus (1 week younger) or an 18 object was introduced and the interaction time was recorded for 5 minutes. All the trials were 19 recorded with a camera placed above the arena, and the interaction time was manually scored 20 when the experimental mouse initiated the action and when the nose of the animal was 21 oriented toward the social stimulus mouse only, or towards the object. 22 23

Three-chamber task 24
A three-chambered social interaction assay was used, comprising a rectangular Plexiglas 25 arena (60 × 40 × 22 cm) (Ugo Basile, Varese, Italy) divided into three chambers (each 20 × 26 40 × 22 (h) cm). Each mouse was placed in the arena for a habituation period of 10 min, when 27 it was allowed to freely explore the empty arena. At the end of the habituation, two enclosures 28 with metal vertical bars were placed in the center of the two outer chambers. One enclosure 29 was empty (serving as an object) whereas the other contained a social stimulus (stimulus1, 1 30 week younger unfamiliar sex-matched conspecific). The experimental mouse was allowed to 31 freely explore the apparatus and the enclosures for 10 min (social preference phase). 32 Subsequently, the empty enclosure was replaced with another contained an unfamiliar 33 conspecific social stimulus (stimulus2) and the experimental mouse was allowed to freely explore for 10 min the apparatus and the enclosures for other 10 min (social novelty phase). 1 The position of the empty vs. social stimulus1-containing or social stimulus1-containing vs 2 social stimulus2-containing enclosures alternated and was counterbalanced for each trial to 3 avoid any bias effects. Every session was video-tracked and recorded using ANY-maze 4 (Stoelting Europe, Dublin, Ireland) which provided an automated recording of the time in the 5 compartment and the distance moved. The time spent interacting with each enclosure was 6 manually scored and then used to determine the preference index for the object or social target 7 (stimulus1 and stimulus2). The stimulus interaction was scored when the nose of the 8 experimental was oriented toward the enclosures at a distance approximately less than 2 cm. 9 The arena was cleaned with 1% acetic acid solution and dried between trials. 10

Habituation task 11
A clean homecage was used as arena. The experimental mouse was placed in the arena with 12 a novel social stimulus (sex-matched conspecific mouse, 1-week younger compare to the 13 experimental mouse). The animals were let free to explore the cage and to interact with each 14 other for 15 minutes. At the end of the trial, the experimental and stimulus mice were returned 15 to their home-cagehomecage. For 4 consecutive days the experimental mouse was exposed to 16 the same social stimulus. All the trial were recorded with a camera placed above the arena. 17 Non-aggressive behavior was manually scored when the experimental mouse initiated the 18 action and when the nose of the animal was oriented toward the social stimulus mouse only. 19 20

Novel Object Recognition (NORT) 21
A squared arena was used for the task which consists of three phases: a first habituation phase 22 followed by a familiarization and the actual test phase. During the habituation phase the 23 experimental mouse is let to freely explore the arena (40x40x40 cm) for 10 min. During the 24 familiarization phase, the animal was exposed to two identical objects (object 1object 2) 25 and was let to freely interact for 10 minutes with both. After a retention delay of 20 min, the 26 mice were exposed for 10 minutes to one of the familiar objects (object 1) while the other was 27 replaced with a novel object (object 3). During the different phases of the test, the objects 28 were placed in the opposite sides of the cage, alternating the position of the respective objects. 29 Every session was video-tracked and recorded using ANY-maze (Stoelting Europe, Dublin, 30 Ireland). The time spent interacting with each object was manually scored and then used to 31 determine the preference index for the different objects. The stimulus interaction was scored when the nose of the experimental was oriented toward the objects at a distance approximately 1 less than 2 cm. The arena was cleaned with 1% acetic acid solution and dried between trials. 2

Elevated Plus Maze (EPM) 3
The elevated plus maze consisted of a platform of four opposite arms (40 cm week of treatment mice underwent to direct free interaction task or were used for whole-cell 9 patch clamp recordings. For long-lasting effect of SI, CNO treatment was stopped at P35, mice 10 were regrouped until P60 and normal water was given. Mice underwent to direct free 11 interaction task or were used for whole-cell patch clamp recordings. For the experiments with 12 1-Naphthylacetyl spermine trihydrochloride (NASPM), mice were infused using a Minipump 13 injector (pump Elite 11, Harvard apparatus, US). 10 minutes before each trial mice were either 14 infused with 4 µg of NASPM dissolved in 500 nL of aCSF (2 minutes of active injection at 15 250 nL/min rate, and 1 minute at rest) or aCSF only (vehicle). After infusion mice underwent 16 to to direct free interaction task. Statistical analysis was conducted with GraphPad Prism 9 (San Diego, CA, USA). Statistical 2 outliers were identified with the ROUT method (Q = 1) and excluded from the analysis. The 3 normality of sample distributions was assessed with the Shapiro-Wilk criterion and when 4 violated non-parametric tests were used. When normally distributed, the data were analyzed 5 with unpaired t-tests, paired t-tests, one-way ANOVA as appropriate. When normality was 6 violated, the data were analyzed with Mann-Whitney test. For the analysis of variance with 7 two factors (two-way ANOVA, RM two-way ANOVA), normality of sample distribution was 8 assumed, and followed by Bonferroni or Tukey post-hoc test as specified in each figure. Data 9 are represented as the mean ± SEM and the significance was set at 95% of confidence. All the 10 experimenters were blinded to perform behavioral manual score and analyses. 11 12