Intranasal oxytocin drives coordinated social approach

Coordinated responses to challenge are essential to survival for bonded monogamous animals and may depend on behavioral compatibility. Oxytocin (OT) context-dependently regulates social affiliation and vocal communication, but its role in pair members’ decision to jointly respond to challenge is unclear. To test for OT effects, California mouse females received an intranasal dose of OT (IN-OT) or saline after bonding with males either matched or mismatched in their approach response to an aggressive vocal challenge. Pair mates were re-tested jointly for approach response, time spent together, and vocalizations. Females and males converged in their approach after pairing, but mismatched pairs with females given a single dose of IN-OT displayed a greater convergence that resulted from behavioral changes by both pair members. Unpaired females given IN-OT did not change their approach, indicating a social partner was necessary for effects to emerge. Moreover, IN-OT increased time spent approaching together, suggesting behavioral coordination beyond a further increase in bonding. This OT-induced increase in joint approach was associated with a decrease in the proportion of sustained vocalizations, a type of vocalization that can be associated with intra-pair conflict. Our results expand OT’s effects on behavioral coordination and underscore the importance of emergent social context.

Playback tracks. We investigated California mouse approach behavior towards playbacks of loud, aversive bark calls, adapted from 38 . In a separate cohort used exclusively to produce the playback stimuli, individual male and female mice were placed in a single-chambered plexiglass apparatus (50 × 30 × 30) under normal food and water conditions for 24 h, a length of time demonstrated to be sufficient for the formation of residency behavior, and in which the arena becomes the individual's territory 36 . After 24 h, residents were introduced to a same-sex intruder for an 8-min aggressive encounter period similar to previous studies 32,33,36,45,46 . Each intruder was only used for a single encounter, and otherwise had no previous aggression testing experience. During the encounter we recorded defensive-aggressive barks using an ultrasonic microphone (Emkay/Knowles FG series, detection range: 10-120 kHz) with a 250 kHz sampling rate and 16-bit resolution, placed 30 cm above the bottom of the apparatus. Spectrograms were produced using a 512 fast Fourier transform in Avisoft SASlab pro (Avisoft Bioacoustics, Berlin, Germany) in order to identify barks. Barks appear as short, high-amplitude calls with an upside down U shape that begins and ends in the audible range for humans 16,47 (Fig. 1A). We created playback tracks using these spectrograms by selecting only bark calls. Calls could not be distinguished between the resident and the intruder during the encounter, therefore both resident and intruder barks were used to construct playback tracks. Playback tracks were 2 min in duration and contained 120 ± 5 bark calls. Output gain/volume was maintained across playback tracks. The ambient noise track control was a 2 min recording of the quiet testing room with all lights off and no mice present. We used 8 unique tracks from 8 different sets of individuals and assigned tracks to individuals randomly with each track used between 20 and 23 times 38 , ensuring that no individual heard the same track more than once over the course of the two tests (to avoid habituation and maintain consistency). We did not observe differences in bark features (frequency modulation, bark length) between males and females used in the production of bark tracks and therefore did not use the sex of the bark producer as a factor in analyses in this study.
Pre-pairing playback approach test. Mice were first tested for response to bark playbacks as nonbonded, sexually naïve individuals ( Fig. 2A, left). Mice were placed in the testing apparatus for 5-10 min to habituate to all three chambers. One male and two females did not enter all three chambers by 10-min and so were removed from the experiment (resulting in 47 males and 46 females). Two-min playback tracks were played from speakers at opposite ends of the apparatus behind wire mesh, with one speaker playing a bark track and the other an ambient noise track concurrently. The bark track and ambient noise track speaker locations were randomized across trials. Video and audio recordings were made of their behavior. Throughout the 2 min approach test we www.nature.com/scientificreports/ recorded time spent in the chamber closest to the bark speaker ("approach chamber") as an approach score, and time spent in the chamber closest to the ambient noise speaker ("avoid chamber") as an avoidance score. See below for evidence of repeatability in unpaired mice.
Behavioral type and pairing. 3-7 days following the pre-pairing test, 46 females were selectively paired with 46 males in a stratified randomized design based on approach scores. Mice were paired such that they were "matched" in their approach scores (i.e., their approach scores were ≤ 10 s different from each other, N = 17) or they were "mismatched" (i.e., their approach scores were ≥ 30 s different from each other, N = 29). There were no pairs formed that had an approach score difference between 10 and 30 s. In a follow-up experiment, 20 female mice were left unpaired following the first playback approach test.
Post-pairing playback approach test. All mice underwent a second playback test to determine if pairing alters responses to bark calls. Pairs were retested 10-11 days after pairing (13-17 days after the pre-pairing test). At 7 days post-pairing, pairs exhibit hallmarks of pair-bonding 48 , including increased affiliation and decreased aggression, indicating that 10-11 days is sufficient for pair-bond formation. The playback procedure was the same as in the pre-pairing test except that paired mice were tested together as a pair ( Fig. 2A, right). Both mice were placed into the central chamber and required to enter all three chambers prior to testing. Behavior scoring was the same as in the first test and, additionally, pairs were scored for time spent together in the same chamber. Females generally give birth within 35 days of pairing when housed in cages of similar size to those used in this experiment 37 , so were expected to be pregnant at the time of the post-pairing test (gestation 31-32 days 49 ). Thus, the time point for the post-pairing test of 10-11 days following pairing was deemed appropriate, as it is approximately 3 times the length of time that would be expected for latency to first mate. The procedure for unpaired mice was the same as the pre-pairing test. Two unpaired females were removed as they did not enter all three chambers by 10 min (resulting in 18 females).  50 and approximates a weight-adjusted   51 . Intranasal administration is an established, non-invasive route of delivery for OT 52 , and intranasally-administered OT enters the brain in house mice (Mus musculus) at behaviorally-relevant concentrations 53 . In California mice, IN-OT produces behavioral effects similar to centrally-administered OT, suggesting that it reaches the brain in this species 41,54 . Following pairing, mice were randomly assigned to either saline (mismatched pairs: N = 15; matched pairs: N = 8) or IN-OT treatment (mismatched pairs: N = 14; matched pairs: N = 9) groups. Unpaired females in the follow-up experiment were also randomly assigned to either saline (N = 10) or IN-OT treatment (N = 8) prior to the retest. Immediately prior to the 5-10 min habituation preceding the retest, females were given either IN-OT or control saline, while all males were administered saline, as this time course has previously been shown to result in behavioral effects in California mice 40 . To administer, a mouse was scruffed and 25 μl of solution was administered to the nostrils using a blunt needle attached to cannula tubing at the end of a Hamilton syringe. Individual droplets were beaded at the end of the syringe, applied to the surface of the nose, and allowed to absorb into the nasal mucosa. Administration lasted less than 30 s per animal. Only females were given IN-OT because of unpublished data suggesting that IN-OT shifts the convergence/division of labor ratio for pairs defending their territories when given to females, but not males 55 .

IN-OT
USV analysis. We recorded USVs with two ultrasonic microphones (Emkay/Knowles FG, detection range 10-120 kHz) placed 85 cm apart at opposite corners of the apparatus and 30 cm from the apparatus floor, www.nature.com/scientificreports/ with one microphone placed in the approach chamber and the other placed in the avoid chamber. Microphone channels were calibrated to equal gain (− 60 dB noise floor). WAV files were produced using RECORDER software (Avisoft Bioacoustics, Berlin, Germany). Recordings were made using a 250 kHz sampling rate with 16-bit resolution and spectrograms were produced with a 512 fast Fourier transform made using Avisoft SASLab Pro (Avisoft bioacoustics). USVs were differentiated by visual and auditory inspection of WAV files with sampling rates reduced to 4% of normal speed (11,025 kHz). Beyond barks, California mice produce a variety of ultrasonic vocalizations (USVs) 39 important for communication and behavioral coordination 56 . Because barks are only typically produced when aggressive physical contact is made 16 and were not produced by mice in this study, two call types, sustained vocalizations (SVs) and sweeps, were analyzed, because they may play a role in pair bond formation and maintenance 48,57,58 . SVs are low-bandwidth calls with low modulation, a peak frequency of 20 kHz and a duration of 100 to 500 ms for each individual syllable (Fig. 1B). Shorter SVs are used during aggressive encounters 16 . Sweeps are of relatively short durations of 30 to 100 ms with upwards and downwards modulation ranging from 25 to 100 kHz and potentially including multiple inflection points 16 (Fig. 1C). Because audio was recorded for both pair members simultaneously, it was impossible to differentiate calls by each individual so vocalizations were analyzed at the pair level. Both the total number of USV calls produced and the proportion of each USV individual call type produced relative to all call type production were analyzed within this dataset. Four audio files were unusable due to recording setup error (mismatched saline pairs: 11; matched saline pairs: 8; mismatched OT pairs: 14; matched OT pairs: 9). The length of SVs was also analyzed, and eight pairs that did not produce SVs were removed from this analysis only (mismatched saline pairs: 11; matched saline pairs: 5; mismatched OT pairs: 12; matched OT pairs: 6).
Statistics. All statistics were analyzed using R (version 3.6.2). We analyzed the female-male approach difference before pairing by taking the absolute value of the difference between female (F) and male (M) approach, |F 1 -M 1 | and regressing it on pair type (mismatched or matched) in a general linear model (GLM). We analyzed the degree to which pairs converged by taking the initial difference between pair members, |F 1 -M 1 |, and subtracting it from the difference at the second test, |F 2 -M 2 |, and regressed it on pair type, treatment type (saline or IN-OT), and their interaction in a GLM. We analyzed individual changes by subtracting initial approach from approach at the second test (F 2 -F 1 and M 2 -M 1 ) and regressed it on pair type, treatment type, and their interaction in a GLM. We likewise analyzed the amount of time pair members spent within the same chamber using pair type, treatment type, and their interaction as factors in a GLM. For changes in vocalizations, we summed vocalizations produced by pairs before pairing and subtracted them from the vocalizations produced in the second test, again using pair type, treatment type, and their interaction as factors in a GLM. For the proportion of sustained vocalizations (SVs), we looked only at vocalizations produced at the post-pairing test, because there were so few vocalizations produced prior to pairing. We analyzed the proportion of SVs using a GLM with pair type, treatment, and their interaction as factors. We used linear regressions to test if behavior predicted call type proportion. For the separate experiment using unpaired females, we regressed second approach on initial approach, and also analyzed the effects of treatment on approach during the second test, using a GLM. p values were Holm-Bonferroni corrected for multiple comparisons where appropriate.

Results
Mismatched pairs had a greater difference in approach scores before pairing than did matched pairs. Prior to pairing, a wide range in approach to bark playbacks was found in females (  From pre-pairing to post-pairing, mismatched and matched IN-OT and saline groups all contained pairs that had pair-averaged approach scores that decreased and pair-averaged approach scores that increased, suggesting that convergence may be driven differently by individuals within pairs (Fig. 3B). To determine whether males and females contribute differently to the overall convergence of pair approach scores, we regressed change in approach from the before-pairing test to the after-pairing test for each individual on the interaction of pair type, treatment, and sex. This GLM revealed no 3-way interaction (F(1,84)  There was also no effect of location (F(1,88) = 0.389, p = 0.535) or the interaction of treatment and location (F(1,88) = 0.996, p = 0.321). Additionally, we did not observe any classic rodent courtship behaviors, including anogenital sniffing or following behavior (i.e. pursuing), that also occur in California mice 59 , nor did we observe any aggressive behavior between pair mates.    (Fig. 4E, Table 1E). These results indicate that SV proportion may be involved in pair convergence. Unpaired females respond to bark playbacks in a consistent manner. In order to demonstrate that approach behavior is repeatable and to investigate the effects of the IN-OT treatment in unpaired individuals, a separate cohort of unpaired females was tested for initial response and response 13-17 days later, the same length of time as that between the pre-pairing and post-pairing test for paired mice. A GLM revealed that initial approach reliably predicts later approach (F(1,14) = 14.59, p = 0.002; Fig. 5A), with no significant main effect of treatment (F(1,14) = 0.313, p = 0.585) or interaction (F(1,14) = 0.227, p = 0.641). A separate GLM indicated no effect of treatment on approach during the second test (F(1,16) = 0.415, p = 0.529; Fig. 5B).

Discussion
The social salience hypothesis argues that oxytocin (OT) attunes individuals to social context 24 , possibly explaining why OT is able to induce both social avoidance 19 and social approach 41 . However, the roles of affiliative bonds in this process have remained elusive. Here we used a pre-post pairing design and found that OT facilitates pair-bonding-induced changes in social behavior and coordination. Intranasal OT (IN-OT), when given to female California mice, produced pair-level shifts in social approach of an aggressive bark stimulus driving mismatched pair mates to become more similar. Analysis of time spent in proximity showed that IN-OT drove pairs to spend more time together approaching the stimulus, likely producing the shift towards pair similarity www.nature.com/scientificreports/ and indicating coordinating effects of OT beyond bonding in these already bonded pair members. Recordings of USVs revealed that IN-OT drove the negative correlation between the ratio of SVs and time spent together in approach. Finally, unpaired females demonstrated a high degree of repeatability in their approach, which was not affected by IN-OT, suggesting that it is an individual trait. This lack of effect with IN-OT in unpaired females also indicates that there was no increase in motivation for females to socially approach the stimulus produced by IN-OT, and if there is any motivation it is not revealed unless a male partner is present. Taken together, these data suggest that the social context of pair bonding alters social approach, that the degree to which an individual alters their social approach depends on their partner, and that OT facilitates pair-bonding-dependent changes in social approach. Our results are consistent with a social convergence theory, whereby OT unites pair-bonded individuals in the face of challenge, and that this coordination may be related to changes in the proportions of vocalization types produced.
Recently, two other Peromyscus species were shown to have repeatability in behavior and correlation between boldness behaviors 60 , and other consistent individual differences have been found in Peromyscus species 61 , including California mice 39 . We extend these findings to consider the change in behavior type resulting from the presence of a pair member, and vocalizations and OT signaling as behavioral and hormonal mechanisms, respectively, under affiliative social contexts. We observed that IN-OT given to females increased convergence in approach scores for initially mismatched pairs but had no effect on either the convergence of pairs that were initially matched or on total time unpaired females spent approaching the simulated unfamiliar conspecific, suggesting that OT changes in behavior may be related to a motivation to maintain partner proximity in response to a single stimulus. While convergence has been shown to be asymmetrical in convict cichlids, with the more reactive partners being more flexible than proactive partners 43 , we show that in California mice both individuals with low and high initial levels of social approach change their behavior in response to pairing. Our results can be interpreted as adjustments by both pair members to become more similar in behavioral type to their partners, and not solely within-individual variation in approach response from the pre-pairing to the post-pairing test, because unpaired individuals showed repeatable responses to the vocalization playback stimuli. Taken together, our study provides experimental evidence of pair type-dependent changes to social behavior that are facilitated by female OT signaling, and is consistent with prior evidence of convergence by mismatched pairs 38 . Moreover, it suggests that although individual unpaired females demonstrate reliable and consistent approach behavior even in response to IN-OT, the change in approach behavior induced by pairing mismatched individuals is malleable in response to IN-OT.
Several hypotheses emerge for why IN-OT drives changes in behavior. First, it is possible that social approach reflects underlying differences in stress and stress susceptibility, as the stressed state of an individual influences both social approach and OT signaling. Social defeat stress decreases OT receptor (OTR) gene expression in female California mice 19 and reduced OTR binding in the NAc corresponds to reduced social approach following defeat 62 . Moreover, plasma OT levels have been positively linked to attachment anxiety in romantic human relationships 63 , and pair bonding influences affective state across taxa 64,65 . Because we did not directly measure stress or anxiety, we are unable to correlate individual differences in stress response to social approach or pair bonding-induced convergence. However, while IN-OT may have influenced anxiety-like behavior, the lack of IN-OT to change social approach in unpaired individuals suggests no baseline interaction between stress and OT.
Another hypothesis for why IN-OT drove changes in behavior is that it increased the motivation to bond. While it is possible that the behavioral convergence reflects a motivation to stay closer together to maintain the pair bond, the second test occurred at a time point sufficiently after bonding measures are typically displayed in this species 48 to suggest that pairs were already stably bonded. Moreover, we did not observe any classical rodent courtship behaviors (such as anogenital sniffing or following) as we have observed in previous studies 59 during the playback approach test. IN-OT likely induced coordination of behavior beyond bonding, as evidenced by the increase in time IN-OT pairs spent jointly approaching the stimulus, but not jointly avoiding. Further evidence is suggested by the vocalization data: In this study we tracked two call types, sustained vocalizations (SVs) and sweeps, because they may play a role in pair bond formation and maintenance 58 . While mismatched pairs produced a higher proportion of sustained vocalizations, indicating a potential behavioral mechanism underlying this coordination, IN-OT drove a negative correlation between SV proportion and time spent together in the approach chamber, suggesting that OT may be involved in the relationship between decreased SV production and behavioral convergence. We found no difference in SV length between groups, which has previously been shown to be a relevant characteristic for bonding pairs, with longer SVs produced over the first week of pairing 56 . However, different results might have occurred if we had recorded individual USVs as opposed to pair USVs. Moreover, the SVs in this study were relatively short compared to those found by other studies (an average of about 115 ms compared to, for example, an average of about 175 ms in 48 ), but similar to those observed during aggressive encounters (for example, an average of about 105 ms in 16 ), which may indicate negotiation or negative affect by pairs in this novel context.
Finally, another hypothesis for why IN-OT drove changes in behavior relates to pair coordination and synchrony. Synchrony, an emergent behavior that occurs between two or more individuals, such as eye contact, singing, or movement, is linked to greater affiliation and cooperation 66 . California mice are socially monogamous and form strong affiliative bonds that last for life 67 ; OT may increase the ability of pair-bonded individuals in monogamous, biparental species to perceive a partner's intent and thus provide complimentary behaviors to efficiently address environmental challenges such as foraging, defending against predators, and taking care of young.
It is interesting to note that IN-OT did not increase sweeps as it has previously been shown to do so when given to mothers interacting with their pups 40 . This effect was context-dependent such that OT increased sweeps in mothers only when mothers were in physical contact with their pups. In the current study we observed an increase in sweeps in response to pairing as expected by 48  www.nature.com/scientificreports/ reason for this is unknown, however, it is possibly because the playbacks resulted in a less affiliative interaction, or that IN-OT effects are restricted to mother-pup interactions.
Overall our results suggest that OT modulates social approach in a social context-dependent fashion by driving initially mismatched pairs to become more matched by increasing time spent in joint navigation, consistent with previous findings that OT signaling is critical to both bonding 68 and social coordination 23 . Taken together these results suggest that affiliative social context is critical for OT-regulated social approach and vigilance. An important direction for future studies will be to determine the neural circuitry overlap between social approach and affiliative bonding as well as using OT blockers to inhibit the natural increase in similarity that was observed in this study. Our results are in line with calls for researchers to consider social context when studying behavioral types 69,70 .

Data availability
Data available from the Open Science Framework: https:// osf. io/ 2a8gm/.