Oxytocin and vasopressin increase male-directed threats and vocalizations in female macaques

In a previous study, we reported that intranasal delivery of both oxytocin (OT) and arginine vasopressin (AVP) to male macaques relaxes spontaneous social interactions, flattens the existing dominance hierarchy, and increases behavioral synchrony with other monkeys. Here we report that intranasal OT and AVP administration modulates the behaviors of female macaque monkeys, but in robustly different ways from males. Most notably, both neuropeptides increase threatening and vocalization behaviors of females when they encounter males, and these behaviors effectively increase the social status of females over males. While OT and AVP heighten the confrontational nature of intersexual encounters, both peptides relax interactions between females. Finally, as previously reported for males, treating an individual female monkey with OT or AVP significantly modulates the behavior of her non-treated partner. Together, these findings show that OT and AVP can either inhibit or promote aggression, depending on sex and behavioral context, and call for a more careful, systematic examination of the functions of these neuropeptides in both sexes, especially in the context of therapeutics for human social disorders.

To bridge the gap between rodent and human studies on the interaction of sex and neuropeptide function, we examined the effects of intranasal application of aerosolized OT and AVP on spontaneous social behavior in rhesus macaques. Like humans, macaques live in large, hierarchical, mixed-sex groups 70 , engage in complex social interactions 71,72 , and largely use visual and vocal signals to communicate 73 . These behaviors are mediated by a network of cortical and subcortical brain areas that appear to be homologous with the human social brain network [74][75][76] . Together these factors make macaque monkeys the ideal animal model for studying the neurobiology of social cognition and social deficits associated with psychiatric disorders [77][78][79] . In a prior study 80 , we found intranasal treatment of male macaques with both OT and AVP relaxed social encounters, flattened the existing social hierarchy, and enhanced the temporal synchrony of reciprocal behaviors between individuals. Here we used the exact same experimental design to examine the effects of intranasal OT and AVP on the behavior of female macaques. Instead of replicating the prosocial effects observed in males, we found to our surprise that both neuropeptides produced the opposite effects in females. Most importantly, both OT and AVP increased females' threatening and vocalization behaviors towards males but not towards other females.

Results
Baseline female behavior towards males and females. To maximize the ecological validity and translational potential of our study, we probed the effects of OT and AVP on spontaneous, naturally-occurring social behaviors. We recorded a series of 5-minute long videos of pairs of adult macaques facing each other in close proximity. They were free to interact without danger of physical contact (see Fig. 1A, Methods as well as Supplementary Video). Prior to each session, one monkey (M1) inhaled saline (serving as a baseline for comparison), OT, or AVP via a pediatric nebulizer 81 , whereas the other monkey (M2) did not receive any treatment. M1 was always a female monkey (n = 4), whereas M2 could be either a male (n = 3), a female (n = 3), or an empty primate chair as a nonsocial control. Both monkeys' behaviors were rated offline by 1-3 independent observers and subsequently converted to a set of activity budget plots (overall concordance across observers = 0.85, see Methods). At the population level (n = 60 female-male and 60 female-female interactions under saline), there was no significant difference in female M1s' staring behavior when facing male or female M2s (F-M = 39.23 ± 3.95 s; F-F = 46.48 ± 5.54 s; P = 0.193, Wilcoxon rank sum test) ( Fig. 2A, left), but female-male pairs looked at the same objects less often than female-female pairs (shared attention, F-M = 50.21 ± 4.14 s; F-F = 78.99 ± 4.87 s; P = 0.000, Wilcoxon rank sum test) ( Fig. 2A, right). There was also no significant difference between male and female M2s' staring (male = 38.37 ± 3.95 s; female = 31.03 ± 3.23 s; P = 0.137, Wilcoxon rank sum test) ( Fig. 2A, insert). By contrast, other behaviors of male and female M2s differed more dramatically from each other. Specifically, compared with female M2s, male M2s turned away more (sign of subordination, male = 12.38 ± 0.40 s; female = 3.36 ± 0.14 s; P = 0.000, Wilcoxon rank sum test), lip smacked more (sign of affiliation, male = 2.44 ± 0.77 s; female = 0.47 ± 0.22 s; P = 0.001, Wilcoxon rank sum test), and yawned more (associated with dominance, male = 2.43 ± 0.52; female = 0; P = 0.000, Wilcoxon rank sum test). By contrast, female M2s vocalized more when compared with males (male = 0; female = 4.07 ± 0.87; P = 0.000, Wilcoxon rank sum test) (Fig. 2B).
In summary, female-male interactions naturally differed from female-female interactions. In general, males were more alert and active during their interactions with females, which could be observed through both dominating and affiliative behaviors. Furthermore, females never threatened (and rarely yawned) whereas males never vocalized under these conditions. Intranasal OT and AVP increase female aggression towards males but not females. For the population (n = 120 saline, 120 OT, 120 AVP sessions), delivering OT or AVP intranasally (25 IU   To briefly sum up, OT and AVP both altered female-male interactions more substantially than female-female interactions. In terms of both staring (dominant) and turning (subordinate) behaviors, the most robust modulation was observed in the untreated M2s instead of the treated M1s. The overall effect of OT and AVP was a relative gain in dominance status for females when they encountered males, but not when they faced other females.
We next investigated how OT and AVP inhalation by female M1s induced changes in the behavior of their male partners, since there was no change in staring or turning away by female M1s themselves. We found both OT and AVP modulated lip smacking by female M1s differently depending on the sex of her partner. Specifically,  Remarkably, female M1s never threatened either male or female partners under saline, but threatened males following neuropeptide treatment, particularly OT. In contrast, there was no change in aggression towards other females after OT or AVP treatment (F-M saline = 0, OT = 0.33 ± 0.14, AVP = 0.05 ± 0.04;  In summary, we found that OT and AVP treatments in female monkeys increased the frequency of threats and vocalizations when they confronted male conspecifics, which were returned by males with an increase in dominant behavior (threat) and a decrease in affiliative behavior (lip smack). By contrast, interactions between females became more amicable following neuropeptide treatment, reflected by an increase in affiliative behavior (lip smack) (Fig. 4D).
OT and AVP also caused female M1s to threaten males more frequently (Fig. 4B). By contrast, OT and AVP caused female M1s to express more affiliative gestures (i.e. lip smacking) towards other females (Fig. 4A). Threats from female M1s were correlated with more staring by M1 and less staring by M2 (Fig. 5C, w/Threat).  Straight lines indicate linear regression fits for saline (grey), OT (pink), and AVP (red) data. (C) Compared to baseline, M1 making threats is linked to more prolonged staring from M1 and reduced staring from M2. In contrast, M1 lip smacking is linked to reduced staring from M1 and increased staring from M2. This is true for both F-M (left) and F-F (right) pairs. X axis: w/Threat: sessions during which M1 made threats; None: sessions during which M1 made neither threats nor lip smacks; Lip: sessions during which M1 made lip smacks. Error bars: mean ± SEM. (D) In F-M pairs (left), M1 making cooing calls is linked to more prolonged staring from M1 and reduced staring from M2. In F-F (right) pairs, increase in M1 calling is linked to increase in M1 staring as well as a smaller increase in M2 staring. X axis: w/No Call: sessions during which M1 did not make any calls; Call: sessions during which M1 made calls. Error bars: mean ± SEM.
To determine how neuropeptides influence the sequence of behaviors evoked during an interaction, we entered all significant behavioral variables into general linear models (GLMs) and found that different models accounted for female-male interactions and female-female interactions. During interactions between females and males, both turning away and threatening (in the absence of turning) predicted how staring by female M1s was modulated by neuropeptides (estimated coefficient for turn = −0.70 ± 0.18, P = 0.000; threat = 1.62 ± 2.99, P = 0.589; turn*threat = 3.27 ± 1.03, P = 0.002). By contrast, during interactions between females, turning away and calling predicted changes in staring by female M1s (estimated coefficient for turn = −0.74 ± 0.37, P = 0.050; call = −2.65 ± 0.99, P = 0.008) (Fig. 6A). We also found that both lip smacking and staring (in the absence of lip smacking) by M1 predicted changes in staring by male M2s (estimated coefficient for stare = −0.02 ± 0.05, P = 0.711; lip smack = −2.31 ± 1.97, P = 0.143; stare*lip smack = 0.23 ± 0.11, P = 0.034). By contrast, staring, lip smacking and calling by M1 predicted changes in staring by female M2s (estimated coefficient for stare = 0.15 ± 0.06, P = 0.016; lip smack = 0.67 ± 0.38, P = 0.081, call = 2.10 ± 0.66, P = 0.002) (Fig. 6B). These and other GLM results, viewed together with the reaction time distributions of different behaviors (Fig. 6C, note how lip smacking and threatening behaviors preceded staring whereas turning and calling often co-occurred with or followed staring) lead us to propose two distinct models for how the neuropeptides OT and AVP modulate intersexual and intrasexual interactions (Fig. 6D). We hypothesize that female-male interactions are much more complex and involve a greater number of behaviors. In addition, two critical behaviors displayed by female M1s, namely staring and lip smacking, have different effects on males and females. Importantly, staring by females towards males leads to less staring and lip smacking by males, whereas in female-female pairs the opposite is true. These findings demonstrate that the same female behaviors can be interpreted as aggressive or affiliative depending on the sex of the target.
Aggressive behaviors are often extremely sexually dimorphic. In many mammals, including humans, females are often less aggressive [99][100][101] and sometimes more cooperative 102-106 than males. Males are often aggressive towards both sexes to compete for a wide range of resources, whereas females are more aggressive towards males than females [99][100][101]107,108 , often within the context of reproduction [109][110][111] . This type of temporary behavioral change, termed 'maternal aggression, ' is considered critical for defending offspring against potential threats, especially infanticidal males [109][110][111][112] . OT has been implicated in all aspects of maternal behavior [113][114][115][116] . It specifically suppresses aggression towards one's own offspring and increases aggression towards intruders 115,117,118 . Although AVP is known to contribute to male aggression 39 , our understanding of its role in female aggression is very limited. Our findings regarding the sex-specific effects of exogenous OT and AVP in macaque monkeys nicely correspond to recent studies in humans 97,98,119,120 reporting neuropeptides promoting prosocial behaviors in males more than in females. It is worth noting that since exogenous OT and AVP very likely bind to each other's receptors to modify behavior 121 , it is difficult to directly compare the effects of these two neuropeptides to each other. That being said, echoing our previous observation in male macaque monkeys 80 , here we did not observe any systematic difference in the behavioral effects of OT versus AVP treatment. We conjecture that much like in male monkeys, the effects of OT on female monkeys are also likely to be partially mediated via binding to AVP receptors in the cortex.
Although we observed an increase in intersexual aggression following neuropeptide treatment of females, we also noted an increase in affiliative behaviors between females. OT has been linked to a preference for affiliation over aggression in females (i.e. the 'tend and befriend' strategy 106 ). This preference may help protect females against aggression by males. Increases in vocalizations following neuropeptide treatment of females may reflect attempts to recruit support of other females to protect against males [122][123][124] . Together, our results indicate that OT and AVP inhalation differentially affects female behavior towards males and females, possibly by recruiting different neural circuits related to reproductive behaviors. In baseline conditions, female-male and female-female interactions are readily distinguishable from each other, suggesting that rather than altering the nature of intersexual and intrasexual interactions, OT and AVP amplify pre-existing sex-specific patterns of interaction 125 .
In our previous study 80 , we reported that OT and AVP not only significantly reduce staring from both treated M1s and untreated M2s, but also shorten the latency with which dominant monkeys return stares and subordinate monkeys turn away to avert stares from other monkeys. This pattern suggested an increase in the efficiency and immediacy of communication between monkeys, consistent with an increase in social attention. In the current study, while it is somewhat surprising that neither OT nor AVP treatment significantly alters staring behavior in treated female M1s, we did find that neuropeptide treatment of female monkeys increases the frequency of threats (towards male partners) and lip smacks (towards female partners). GLMs together with reaction time analyses on different behaviors suggest that threats and lip smacks could potentially lead all other behaviors and set the tone for the subsequent interaction. Regardless of the mechanism, once OT or AVP alters the behavior of one monkey, however subtly, its effect is amplified by feedback from the untreated monkey, thus shaping the overall tenor of the social interaction. Through such chain reactions, OT and AVP regulate the nature of social interactions in profoundly sex-specific ways.

Methods
The majority of methods and analyses used in this study were described in detail in a previous paper 80  monkeys) for the duration of this experiment. Cages were arranged facing toward the center of the room, along two walls, permitting all animals to be in continuous visual and auditory contact. Two of the four females (B and C) were pair-housed. All animals were between the ages of 11 and 18 at the time of the experiments and had been in the colony for at least six months. Food grabbing tests as well as caretakers/handlers' ratings confirmed that the dominance order within this group was: O > C > B > F > S > Sch > D for the duration of the experiment. 2-way ANOVAs revealed no significant interactions between neuropeptide treatment and weight, age, or dominance order of the animals.
Experimental Setup. In each experiment session, two monkeys faced each other face-on in an empty room, and were free to interact for 5 minutes. The monkeys sat in their respective primate chairs (Crist Instruments), and the two chairs were positioned close together without touching each other (~30 cm apart from edge to edge). This relatively unconstrained yet still well controlled setup afforded us much of the flexibility of natural social interactions without risking actual physical contact between monkeys. A video camera (Logitech, 60 fps) was positioned on the right side of M1/the left side of M2, and simultaneously recorded both monkeys' behaviors into an MP4 file. On each day, one M1 faced six different M2s sequentially. In addition, the same M1 also faced an empty chair for 5 minutes. The order in which M1 faced the other monkeys together with the empty chair was determined randomly each day.
Intranasal Nebulization. The procedure for intranasal OT delivery in macaque monkeys has been described in detail previously 80,81,87 . Briefly, monkeys were trained to accept a pediatric nebulizer mask (Pari Labs) over the nose and mouth. Through the nebulizer 1 ml of OT (25 IU/ml in saline; Agrilabs/Sigma Aldrich) or saline was delivered at a constant rate (0.2 ml/min) over a total of 5 minutes. Behavioral testing began 30 minutes after intranasal delivery and continued for 1-2 hours. This OT dosage and timing protocol were similar to that typically used in humans 20,89-92 and other non-human primate studies 81,87,126,127 . We followed the same procedure for intranasal AVP delivery (25 IU in 1 ml saline; Sigma Aldrich). The same amount of neuropeptide (25 IU) was delivered to all monkeys regardless of their weights (ranging from 8-12 kg at the time of the experiment), which resulted in a dosage of ~2.1-3.1 IU/kg. Neuropeptide and saline treatments were delivered on alternating days, with each monkey receiving no more than 5 treatments per week. The same monkey never received OT and AVP treatment within the same week. The order of treatments was counterbalanced across monkeys as well as within monkeys between weeks to mitigate any possible order effects. Furthermore, to rule out the possibility that the particular sequence with which different monkey pairs were tested (after saline or neuropeptide delivery) had any impact on behavior, 1-way ANOVAs were performed on different behaviors and did not reveal any significant order effect. In addition, general linear models were constructed for different behaviors as well, with the testing order being treated as a continuous variable and measured as the estimated time passed from drug administration to behavioral testing for each pair. This analysis revealed no significant order effect either.
Data Analysis. Each behavioral video was rated offline by 1 to 3 independent viewers, all of whom were blind to treatment conditions. Viewers used a Python GTK based, custom GUI to play and pause the video, adjust its speed, and code monkey behaviors by pressing a set of keys on the keyboard. Briefly, the ethogram used in this experiment included: (1) staring: one monkey facing forward and fixating directly on the other monkey (a sign of dominance); (2) turning away: one monkey turning at least 180° in the chair with its back facing the other monkey; (3) lip smacking: one monkey facing forward and quickly smacking its lips at the other monkey, sometimes accompanied by slight head tilt; (4) looking around: one monkey facing forward but not making direct eye contact with the other monkey; (5) yawning: one monkey opening its mouth widely and inhaling deeply; (6) threatening: one monkey staring at the other with its eyes wide open, often with open mouth and tense with the lips covering the teeth, sometimes accompanied by head jerk; (7) vocalizing: defined here as a soft, high pitch cooing, otherwise known as clear call; (8) shared attention: defined here as a portion of the looking around behavior where both monkeys' gaze was directed away from each other, but at the same point in space. The strings of identified behaviors and their corresponding time stamps were imported into MATLAB (Mathworks) and converted into a pair of activity budget plots via custom MATLAB scripts. When more than one viewer rated the same video (~40% of all the videos), their ratings were averaged to generate the activity budgets. For the same videos rating consistency was very high across different viewers (for example, fixation duration r = 0.66, P = 0.000; number of fixations r = 0.40, P = 0.001). The overall concordance across observers was 0.85. All subsequent data analyses were accomplished with custom MATLAB scripts.
All statistical tests were two-tailed. For hypothesis testing between two samples, a non-parametric Wilcoxon signed rank test (for paired samples) or Wilcoxon rank sum test (for un-paired samples) was used, as observed behavioral durations or frequencies often form skewed distributions with long tails 80 . For comparison among more than two samples, an ANOVA was used together with multiple comparisons (Tukey's HSD test) when appropriate. 1-way ANOVA was replaced with Kruskal-Wallies test if the assumption of normal distribution was significantly violated. 2-way ANOVA was used in combination with a bootstrapping test to determine the corresponding P values if the assumption of normal distribution was significantly violated. Correlation coefficients were estimated with Pearson's r. All reaction time distributions were fitted with Gamma distributions:

Data Availability
The datasets generated during and/or analyzed during the current study are available from the corresponding author upon request.