Male C57BL/6J mice and β2^−/− knockout mice purchased from Charles Rivers Laboratories (L'Arbresle Cedex, France) were used in the present study. β2^−/− mice were generated from a 129/Sv Embryonic Stem cell line as previously described (Picciotto et al., 1995) and backcrossed onto the C57BL/6J strain for 20 generations. As they were shown to be at more than 99.99% C57BL/6J by a genomic analysis using 400 markers, C57BL/6J mice were used as control of β2^−/− animals.

All experimental procedures were carried out in accordance with the European Communities Council Directive of 24 November 1986 (86/609/EEC), EU Directive 2010/63/EU, Decree N 2013-118 of February 1st, 2013, and the French National Committee (87/848). Experiments were conducted in order to reduce the number of mice used and their level of discomfort. C57BL/6J and β2^−/− mice, 10–11 weeks old, were group-housed (four mice/cage; size of the cages L: 26.5 cm, l: 16.5 cm, H: 13.5 cm) at their arrival to acclimatize to the animal facilities (food and water ad libitum, room temperature 20–22°C) under a 12/12 h light/dark cycle (light on at 7.30 am).

The social interaction task was performed in a transparent Plexiglass box (L: 50 cm × l: 25 cm × H: 31 cm) located in an experimental room with light level set at 100–110 Lux by indirect white bulbs. For each experiment, the floor of the cage was covered with clean sawdust. The box was placed under a camera connected to a computer located outside of the experimental room.

Social behaviors were studied in dyads of male mice composed by an Isolated Host mouse (IH) and a Social Visitor mouse (SV) brought together for the first time. IH mice were individually housed 4 weeks before the social interaction task, while SV mice remained group-housed. IH mice (either C57BL/6J or β2^−/− mice) were submitted or not to an acute stress. SV mice were always not stressed C57BL/6J male mice. The choice of 4 weeks social isolation was made after pilot experiments. All behavioral experiments were performed from 9.00 a.m. to 2.00 p.m. (i.e., in the lit phase of the dark/light cycle). Each isolated animal randomly assigned to the stressed groups was placed for 45 min in a Falcon® tube. After 5 min in its home cage, IH mouse was placed in the experimental box for 30 min exploration. Then, a SV mouse was gently introduced in the box, in the corner opposite to the IH mouse. The delay from the beginning of the stress to the end of the social task (i.e., 84 min) was compatible with expression of the c-fos protein in the PFC (Weinberg et al., 2007; Avale et al., 2011), an area previously shown to be critically involved in social interaction (Avale et al., 2011). Control mice were treated similarly, except that they were not submitted to stress (Supplementary Figure 1). In this study, 19 isolated C57BL/6J mice were used (11 were not stressed, i.e., C57BL/6J mice, 8 were subjected to the acute stress, i.e., C57BL/6J Stress mice) and 16 isolated β2^−/− mice (9 were not stressed, i.e., β2^−/− mice and 7 were submitted to an acute stress, i.e., β2^−/− Stress mice). In addition, 35 group-housed C57BL/6J animals served as SV mice.

Social interactions between SV and IH mice were videotaped and analyzed using Mice Profiler software (de Chaumont et al., 2012). Affiliative behaviors studied are explained in Supplementary Table 1. We distinguished contact events, relative positions between mice, dynamic complex events initiated by SV or IH mice, and stop events. Dynamic events were dissociated in first order events (escapes, approaches or follows), second order events (first-order event preceded or followed by a contact) and third order events (2 successions of 2 second-order events). The number and duration of behavioral events of each mouse of the dyad were quantified during the 4 min of the social interaction task.

Chronograms and density graphs were built to illustrate the evolution of the quantitative data in the course of mice dyad interactions. They show behavioral variation of a specific event within an animal and between animals over time. As such, they allow to perceive the stability or instability of given behaviors. For each mouse or each dyad, chronograms depict the frame-by-frame occurrence and duration of a particular event, and thus provide an immediate visualization of its frequency. Density graphs (built from chronograms) represent the temporal evolution of the density of a given event, i.e., the sum of the time in which an event is happening vs. the total time of the bin (grouped here in bins of 30 s). As such, they represent the same information as chronograms but averaged for all mice of a specific group. We also built transitional graphs to illustrate the probability of incidence of one event or posture on the next or the previous one in a sequence.

Finally, we scored dominance and aggressiveness by off line manual analysis. Dominance index was defined as the number of “paw control” (i.e., the number of times IH mouse placed its forepaw on the head or on the back of the SV mouse). Aggressiveness was indexed by the cumulated number of tail rattling and attacks over the 4 min of the experiment. Latencies to first attack and first tail rattling were also scored.

Biochemical measures of plasma corticosterone were carried out in a pseudo random order by an experimenter blind to mice genotype and stress conditions. Experiments were performed in the same time scale than behavioral ones (10 a.m.–1.30 p.m., i.e., in the lit phase of the dark/light cycle), animals being treated similarly than for behavioral experiments (see Supplementary Figure 1). Trunk blood was collected after mice decapitation. Sacrifices were done 60 min after transfer of mice from animal facility, i.e., after habituation to the experimental room (condition A, n = 4 for both genotypes), just after the restraint stress (condition B, n = 6 for each genotype) and at the end of the exploration period, i.e., at the time at which social interaction should take place (condition C, n = 6 not stressed and 7 stressed C57BL/6J mice, n = 7 not stressed and 9 stressed β2^−/− mice). Knowing that baseline levels were already measured 60 min after the exit of the animal facility (condition A), waiting 45 min more was extremely unlikely to increase plasma corticosterone levels by itself, so we didn't measure it in additional groups of not stressed animals (condition B). After centrifugation (5000 g 10 min at 4°C), the supernatant was stored at −20°C. Corticosterone levels were quantified using an enzyme immunoassay commercial kit (DetectX®, Arbor Assays). Samples were included into the linear part of the standard curve. Limit of detection was determined as 0.17 μg/dl.

Normal distribution (assessed by Shapiro-Wilkonson test) and equality of variance (evaluated by equal variance test) were first tested using Sigmaplot 12.0. When both were statistically significant, main factor effects and interactions were tested with Two-Way ANOVAs followed by Fisher LSD Method tests when appropriate. When normality and/or equal variance were not statistically significant, Kruskal-Wallis One-Way ANOVA then Mann-Whitney U-tests were performed when appropriate. Results were reported as means ± SEM. P values ≤ 0.05 were considered statistically significant. Correlations between behavioral events were performed using Spearman test and a Bonferroni correction was applied for multiple comparisons. Statistically threshold was thus set at p ≤ 0.0022.

We found that, in all 4 types of dyads, SV stops always occurred at the starting point of contact events, postures, or dynamic event, regardless of the emotional status or genotype of its partner. In addition, for all dyads, IH stops and follow events were key points in the social repertoire.

We first confirmed here previous data as β2^−/− mice made longer contacts and more numerous follows, spent less time at rest and were more dominant (Granon et al., 2003; Maubourguet et al., 2008; de Chaumont et al., 2012; Coura et al., 2013). We also confirmed that key events, crucial for initiating various behaviors, differed between the two mouse genotypes. We thus explored thoroughly their social consequences in both genotypes. We found that β2^−/− mice exhibited a strong and constant interest for contacting their social partner as the experiment progressed while C57BL/6J animals, decreasing their interests for social contact, became accustomed to the social situation over time (Supplementary Figure 2). Furthermore, follows in β2^−/− mice were only correlated with oral-oral contact and posture behind the SV mouse while they were correlated with various actions (contacts, escape, dominance) in C57BL/6J mice (Figure 6). This indicated that follows are associated to a restricted social repertoire in β2^−/− mice but to a large panel of behavior with no fixed consequences in C57BL/6J. Also, correlation analyses showed no links between different contact subtypes in β2^−/− mice while many of them were related in C57BL/6J animals (Figure 6). Stop events allowed mice to scan their environment and to acquire and evaluate information about potential risks (Quartermain et al., 1996; Maubourguet et al., 2008). We showed here that their relative absence in β2^−/− mice, consistent with their hyperactivity (Avale et al., 2008), had strong consequences on the social repertoire: in β2^−/− mice, stop events only induced oral-genital contacts while in C57BL/6J mice, stop events triggered multiple events, including contacts, relative position events, and follows (Figure 7). All these data reinforced the fact that β2^−/− mice exhibited less flexible behavior than C57BL/6J and showed more stereotyped and restricted actions (Granon et al., 2003; Maubourguet et al., 2008; de Chaumont et al., 2012). It also showed that the behavioral repertoire may be possessed by both genotypes but may not have the same “value.”

We revealed a crucial importance of dominance behaviors: in C57BL/6J mice dominance was associated with a large panel of social contact but this was not the case in β2^−/− mice, for which dominance seemed an isolated compound (Figure 6), although quantitatively over represented (Table 2). This data matched with our previous findings, which showed that dominance tendencies were strongly modulated by the nicotinic-cholinergic system (Coura et al., 2013). For the first time, we showed that behavioral differences between C57BL/6J and β2^−/− mice were not due to different plasma corticosterone levels, either measured in a control condition (60 min after exit from animal facility), or after exploration of the novel environment. Baseline measures were very low, in agreement with others results (Thoeringer et al., 2007). This showed that animal isolation per se did not generate a chronic stress. However, as compared to baseline values, novelty exploration increased corticosterone in all not stressed animals (Table 3), as had been reported by others in C57BL/6J animals (Roozendaal, 2002; Chauveau et al., 2010; Beerling et al., 2011; Chaouloff and Groc, 2011), showing that exploration of a novel environment induced activation of physiological parameters normally associated to stress responses (De Boer and Van der Gugten, 1987; Beerling et al., 2011). Besides, we detected that both mouse genotypes similarly and largely preferred to be behind SV mice all along the experiment. In contrast, the opposite posture gradually appeared more often over time, particularly in C57BL/6J mice (Figure 2). These findings can be explained by the fact that SV mice were always C57BL/6J, group-housed, not stressed and never dominant in our experimental condition. Indeed, it was shown that taking place behind a partner is a posture likely to be exhibited by dominant animals (Arakawa et al., 2007). Thus, our results suggested that relative positions events in this task could be used as an index of a mouse's tolerance to another mouse.

In summary, the present study showed that each mouse genotype exhibited a specific complex social behavioral pattern and developed distinct social strategies. It highlighted major behavioral difference between C57BL/6J and β2^−/− mice, hence the important role of the cholinergic nicotinic system in the social strategy used by animals to organize their behavior. The differences between C57BL/6J and β2^−/− mice in this task may rely on a different trade-off between exploitation and exploration: C57BL/6J mice would favor exploration of various options, including social contact when confronted to a novel environment, and β2^−/− mice would favor exploitation of the social reinforcement. Favoring exploitation over exploration has been shown to be accompanied by rigid behaviors (Cohen et al., 2007). This view is consistent with the involvement of the monoaminergic ascendant system on one hand and of the cholinergic system on the other hand in the governance of the balance between exploration and exploitation (Cohen and Aston-Jones, 2005; Yu and Dayan, 2005). As the regulation of this balance has been shown fundamental for adaptive and flexible behaviors (Kehagia et al., 2010), our data supported the need for functional nAchRs in social context, complex, changing and uncertain by nature (for review, Cohen et al., 2007).

Although SV mice were not subjected to stress procedure, it indirectly altered their behavior: it shorted their stops and their tolerance when facing a stressed C57BL/6J mouse, indicating that SV was influenced by the emotional status of its partner. This can be due to the increase of follows, dominance and aggressiveness exhibited by stressed mice. This adaptation to their social partner, in turn, deeply modified their entire social repertoire as seen in transition graphs. In another social paradigm (3 chambers), Yang et al. (2012) showed that C57BL/6J mice adjusted their behavior, depending on the genotype of their partner. Also, Varlinskaya et al. (1999) showed that adolescent rats changed their social behaviors depending on the social activity of their play partner, as observed in other species (for review, Blanchard et al., 2003) or in other social tasks (Nyberg et al., 2004; Moy et al., 2008; Fairless et al., 2013). However, this is the first study to report that such a profound, dynamic and rapid adaptation occurred in a dyad of adult mice that had never interacted before. Future studies should investigate how this adaptation might be achieved; it may involve communication, via acoustic (Panksepp and Lahvis, 2011; Chabout et al., 2013), olfactory (Arakawa et al., 2008a,b; Deschenes et al., 2012), visual (Clark et al., 2006), and/or tactile (Clark et al., 2006; Wolfe et al., 2011; Deschenes et al., 2012) cues (Adolphs, 2010; Silverman et al., 2010).

Stress had three main consequences in IH C57BL/6J mice: it impoverished the social repertoire, increased dominance, and fostered aggressive behaviors. The social impoverishment was illustrated by the reduction of complex sequences with concomitant increase in single events such as approaches and follows, the shortening of some contact events and of mutual tolerance (Table 1). Stress also reduced the duration of stop events that were shown to be keys in the decision process for making various social actions. In consequence, their impact on subsequent social behaviors was abolished (Figure 7). Therefore, by shortening the time spent scanning their environment during stops, stressed mice reduced the panel of their possible behavioral choices. Interestingly, such stress effect has also been reported in humans to promote habits (Schwabe and Wolf, 2009) and to reduce the panel of strategies during decision-making tasks (for review, Starcke and Brand, 2012). Also, we showed that follow behaviors increased after stress and appeared crucial for initiating multiple short sequences. Their consequences changed as compared to not stress condition: the relationships between contacts and follows were completely abolished by stress (Figure 6). These features were those observed in β2^−/− mice without stress. As consequence, the complexity and the length of behavioral sequences were reduced in stressed C57BL/6J mice, and social behaviors appeared scattered and less fluently organized. Our data thus showed that the meaning of behaviors that are part of the normal social repertoire (e.g., follows) may depend on the emotional status of the animals. It implies that the gross measurement of their existence is not sufficient to characterize a social deficit in mice models.

We reported that single restraint stress increased dominant behaviors by more than three-fold in C57BL/6J mice, and dramatically altered their consequences. Social dominance is a normal behavior (Blanchard et al., 1984, 2003; Chichinadze et al., 2011). Without stress, dominance promotes social contacts whereas after stress, it favored only follows (Figure 6).

We showed that aggressiveness and dominance were dissociated social features that were never correlated, as previously reported (Coura et al., 2013). IH mice were isolated for 4 weeks prior to the experiments. It is questionable whether social isolation can itself promote aggression. To investigate this question, we compared our social task with other tasks studying aggressiveness. For example, in the resident-intruder task, classically used to trigger aggressiveness, social isolation is not the only component of the procedure: aggression is typically realized in the home cage of the isolated mouse and in some cases, males are previously exposed to females to increase territorial behavior, or are confronted with partners of a different size (Miczek and O'Donnell, 1978; Crawley, 2000; Miczek et al., 2001; Arregi et al., 2006; Neumann et al., 2010; Yohe et al., 2012; Marquez et al., 2013). In the present study, the task was never conducted in the home cage of a mouse. Also, the experimental cage was much larger than home cages, and it contained clean bedding; this environment favored novelty exploration and prevented strong territorial behavior as shown previously (Supplementary Figure 1 in Avale et al., 2011). Finally, our animals were never exposed to females. Therefore, our task differed dramatically from other tasks widely used to trigger aggressiveness. Moreover, we used C57BL/6J mice, which exhibit high sociability (Sankoorikal et al., 2006; Yang et al., 2012) and are generally not considered an aggressive strain (Puglisi-Allegra and Cabib, 1985; Roubertoux et al., 1999; Le Roy et al., 2000; Sankoorikal et al., 2006; de Chaumont et al., 2012), even after 8 weeks of social isolation (Puglisi-Allegra and Cabib, 1985). This non-aggressive disposition was confirmed in our study, because no aggression was observed in not stressed dyads. Thus, stress-induced aggressive behavior could not be attributed to the isolation of one of the adult mice in our experimental conditions. Also, in social contexts, aggression can be promoted by various experiences, including shocks (Puglisi-Allegra and Cabib, 1985), post-weaning isolation (Toth et al., 2011, 2012; Tulogdi et al., 2014), post-traumatic stress disorders (Lasko et al., 1994; Pavic et al., 2003; Haller et al., 2005), and early life stressful experiences (Veenema, 2009; Marquez et al., 2013). Prolonged or recurrent stress also contributed to aggressive behaviors (Wood et al., 2003; Yohe et al., 2012; Umukoro et al., 2013). To our knowledge, however, the present study was the first to assess aggressive behavior after delivering a single short stress to adult mice. As expected, immobilization stress increased plasma corticosterone levels. It is noticeable that plasma corticosterone levels remained steadily high when stress was followed by novelty exploration in C57BL/6J mice. Therefore, our stress procedure did not superimpose to a putative isolation-induced stress.

Social experiments reported by others were mostly performed during the light phase of the cycle, although some experiments were conducted during the dark phase. However, it has been reviewed recently that sociability in rodents can be efficiently evaluated in the light phase (Yang et al., 2007, 2008). It is well known that plasma corticosterone has a circadian rhythm, with a peak around 3 p.m., a largest second one around 7 p.m., and a relatively high level during the first 6 h of night (De Boer and Van der Gugten, 1987; Kalsbeek et al., 1996). However, we cannot exclude that our experiments, conducted when plasma corticosterone levels were the lowest, could have led to different behavioral results if conducted in the dark phase of the cycle. In our experimental conditions, aggressiveness cannot be due to changes in plasma corticosterone levels induced by novelty exploration nor to the stress itself. Since social behaviors are supported by the HPA axis (for review, Owings and Coss, 2007), it may be involved, at least partially, in the stress responses observed here. But how the adaptive adjustment of behavioral system develops and evolves remains to be determined.

In conclusion, stressed C57BL/6J mice managed social contact with a novel social partner by reorganizing their behavior; moreover, they reduced or eliminated complex sequential behaviors. Interestingly, the dominance level in stressed C57BL/6J mice increased to the levels observed in β2^−/− mice. This finding suggested that stress may alleviate behavioral inhibition mediated by the cholinergic system (Picciotto et al., 2012) or by the prefrontal monoaminergic systems (Arnsten, 2009; Ginsberg et al., 2011), because β2^−/− mice exhibited constitutive major increases in the monoaminergic and cholinergic tones in the prefrontal cortex (PFC) (Coura et al., 2013). Previous studies showed that aggressiveness also resulted from lesions in the NA system (Cambon et al., 2010) or NA PFC depletion (Coura et al., 2013), and was associated with brain cholinergic and NA systems (Maxson, 2009; Ginsberg et al., 2011). Thus, the effects of stress in C57BL/6J mice may have been due to an acute increase in acetylcholine and/or monoamine PFC tone (Das et al., 2000; Del Arco and Mora, 2009; for review, Picciotto et al., 2012).

Stressed β2^−/− mice acquired C57BL/6J-like social behavior. That is, in β2^−/− mice, stress drastically reduced social contacts to the levels observed in C57BL/6J mice, and it normalized approaches, escapes and the number of stops. However, stress had no effect on tolerance, follows, or complex behaviors. After stress, follows were totally disconnected from all behavioral events (Figure 6). This disconnection may contribute to the reduced contact durations observed after stress. Sequences initiated by stop events were restricted, thus behavioral complexity was reduced (Figure 7). Finally, stress induced no additional dominance, but boosted aggressiveness. Individual data on chronograms and temporal evolution curves showed that the occurrence of aggression did not trigger systematic changes in social events; thus, these behavioral features were independent. In addition, emergence of aggressive behaviors seemed incompatible with elaboration of complex social sequences (Figure 7). Exploration was not associated to additional increase in plasma corticosterone in stressed β2^−/− mice. On the contrary, we observed a significant decrease in this stress marker at the end of the exploration period, as compared to the one measured immediately after immobilization. Thus, as for 57BL/6J mice, aggressiveness cannot be due to increases in plasma corticosterone amount. After stress, plasma corticosterone in β2^−/− mice tended to be higher (although this was not significant) than in C57BL/6J mice. This suggested that β2^−/− mice may be more sensitive to restraint stress (physical discomfort) than to stress generated by novelty exploration. Thus, stress would induce a focus on the internal state rather than on external stimuli. This focus on their internal state could produce attentional defects (Guillem et al., 2011) and trigger behavioral rigidity (Granon et al., 2003; de Chaumont et al., 2012). It was previously shown that elevation of catecholamines in the PFC inhibited prefrontal functional activity during stress (Arnsten, 2009) and that β2^−/− mice exhibited constitutive PFC noradrenaline elevation (Coura et al., 2013). A schematic of the catecholamine effect on PFC physiology exhibited a classical inverted U-shape (Arnsten, 2009); this relationship may underlie the inflexible social behavior observed in β2^−/− mice, as well as their even more inflexible behavior after stress.

In conclusion, stress in β2^−/− mice produced a dramatic reduction in the social repertoire, a clustering of behavioral sequences, emergence of aggressiveness, and apparent social normalization.

C57BL/6J and β2^−/− mice exhibited similarly low baseline levels of corticosterone. Also, their levels of plasma corticosterone were comparable after immobilization stress and after stress followed by novelty exploration. This reinforced the idea that the behavioral differences reported here between stressed and not stressed C57BL/6J and β2^−/− mice could not be accounted for by plasma corticosterone absolute levels. However, variation in these levels differed in both genotypes: it remained stable whether animals explored novelty or were only stressed in C57BL/6J mice, whereas it significantly dropped after novelty exploration despite previous stress in β2^−/− mice. β2^−/− mice seemed to be more reactive to restraint stress than C57BL/6J mice. As mentioned above, this provided evidence that β2^−/− mice were more sensitive to their internal status than to external cues, leading, in consequences, to a larger variation in their corticosterone levels in between different behavioral conditions. The comparison between stressed C57BL/6J and stressed β2^−/− mice showed that acute stress tended to stereotype social actions of both mouse genotypes, and thus, it made both of them less flexible. This finding goes along with recent data obtained in humans, which showed that cold stress imposed before an instrumental task promoted habits (Schwabe and Wolf, 2009). As discussed above, this stress response also reminded behavior observed after NA PFC depletion, suggesting that acute stress may cause an imbalance in the NA levels in the PFC or, more generally, it may reflect impairment in PFC functions (Arnsten, 2009; Del Arco and Mora, 2009). This interpretation would be consistent with our previous findings that flexible social behaviors required functional nAChRs within the PFC (Avale et al., 2011), as they control the release of multiple ascending neurotransmitters in this brain region (dos Santos Coura and Granon, 2012).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.