2.1 Ethics

All animals were kept in Institutional Biosafety Committee (IBC) approved housing at the New York University Abu Dhabi (NYUAD) animal facility. All experimenters completed the Collaborative Training Initiative (CITI) Animal Care and Use Course, which meets United States Department of Agriculture (USDA) and Office of Laboratory Animal Welfare (OLAW) criteria for training in the humane care and use of animals in research. All animal protocols were in accordance with the National Institute of Health Guide for Care and Use of Laboratory Animals (IACUC Protocol: 150005A2) and have been approved by the NYUAD Animal Care and Use Committee.

2.2 Animal rearing and behavioral testing

All experiments were performed on male C57BL/6J mice (Jackson Laboratory). Upon arrival, mice were ear-marked and allowed to acclimatize for 1 week before the onset of experiments. Animals were weighed weekly to ensure healthy weight and weight-matching within cage groups. Behavioral experiments commenced when animals were 7 weeks old and were completed by 15 weeks of age. Retired male CD1 breeders (Charles River Laboratory) were used as resident aggressors during the CSDS paradigm. All mice were maintained under standard housing conditions at a humidity of 50 ± 10%, temperature of 23 ± 2°C and a 12 hr light/dark cycle (7 a.m.–7 p.m.), with ad libitum access to food and water. Wood shavings were used as enrichment in the home cage (green line IVCs, size: 391 × 199 × 160 mm) with social housing, unless isolation was required by the experimental protocol. Only the mice tested for the effect of stress on subsequent hierarchy formation were single housed upon arrival to avoid confounding effects of prior hierarchical rank. Within the experiment, variance was reduced by using all male mice that were age- and weight-matched per cage. The same female experimenter handled mice prior to and during testing whenever possible to reduce stress and anxiety responses.

All behavioral tests were conducted during the light period of higher activity (2–7 p.m.), and the mice were habituated to the recording room and lighting conditions for at least 30 min prior to testing. Between animals, the behavioral apparatus was cleaned with MB-10 solution (active ingredients: 20.8% sodium chlorite and 7.0% sodium dichloroisocyanurate dehydrate) for disinfection and elimination of olfactory cues.

All mice used to investigate the effects of rank on stress susceptibility underwent virus injection surgeries for neural projection tracing (data not shown). These surgeries were performed at 6-week-old mice who were allowed 1 week of recovery before re-housing into novel weight-matched groups of four. All animals recovered well from anesthesia and the surgery were healthy and displayed normal behavior. We therefore believe the surgery did not confound the experimental results reported here.

2.3 Social dominance tests

A minimum of 2 weeks of cohabitation were allowed for the formation of stable hierarchies, as defined by earlier studies (Varholick et al., 2018). The validity of the TT has been previously critiqued based on whether it is a true measure of dominance considering the possible confounding effects of sensorimotor capacity, learning ability, and spatial context (Bernstein, 1981; Miczek & Barry, 1975). Therefore, we have used a battery of additional dominance tests that displayed high consistency of ranking results with those obtained by the TT, thus validating that dominance was indeed the underlying variable measured in the context of our study. The rank was established first in the TT and then followed by three supplementary dominance tests as described previously (Wang et al., 2011). The supplementary dominance tests were scored blinded to the result of the TT ranking. We have only used male animals in this study as female mice are not commonly used in assessments of social dominance based on territoriality and vocalizations since females rely more on intrinsic attributes and social feedback to establish a hierarchy rather than prior social experience (Van Den Berg et al., 2015). However, some studies were able to show stable linear hierarchies in female mice, although the effect of estrus stages has to be taken into account (Rienecker et al., 2020).

2.3.1 Social confrontation TT

Our customized automated TT system (Clever Sys Inc.) consisted of a clear plexiglass tube 55 cm in length and 2.5 cm in diameter, sufficiently wide for one mouse to walk through but not for two mice to pass each other. The tube was connected to a 10 × 10 cm box on each side. The box in which a mouse was initially placed was called the “starting box,” for which the box at the other end would be the “goal box.” Automated doors were placed at the box exits and in the middle of the tube. All animals underwent a two-day training phase to learn how to enter the tube, be within 3 cm of the middle door to open it and initiate a trial, and then to pass to the goal box on the other side. On each of the 2 days, animals had to complete a total of 10 trials, five starting on each side of the tube, making a total of 20 trials for the whole training phase. If the mouse remained stationary for longer than 5 s, or began retreating, a gentle push from behind was used to direct movement toward the middle door. The use of food deprivation and food reward previously showed no effect on animals’ motivation to complete the task (Fan et al., 2019) and were therefore not used in this study.

During the testing phase, each group of animals underwent the TT once and repeated it daily thereafter. Our pilot experiments used a seven-day testing phase, but the animals started displaying signs of stress and reduced task motivation in the later days, so we opted for the four-day testing instead (i.e., each group did four TTs in total in four consecutive days). On each day of the testing phase, each mouse explored the tube once from each side prior to starting the confrontation trials. Using a round-robin design, all pairs of mice from the same cage were tested (six pairs per social group of four mice; each mouse undergoes three TTs in total on 1 day). During the trial, both mice were guided into the tube simultaneously from their respective starting boxes. The starting side for each mouse alternated between trials. When both mice were within 3 cm of the middle door, the door opened, and the social confrontation trial began. The trial ended when one of the mice retreated with all four paws to its starting box, therefore becoming the “loser” or the subordinate (Figure 1a). The mouse that forced its cage mate to retreat was termed the “winner” or the dominant. In between trials, mice were kept in separate clean holding cages. The confrontation trials were repeated for four consecutive days with a randomized order of the pairs and the cages. The experimenter remained stationary during each trial in a designated position in the room to maintain cue consistency.

The winning ratio was calculated as the number of all trials won by that mouse divided by the total number of trials. This determined the index of overall dominance where rank 1 and rank 2 mice (winning ratio >0.5) were the dominant, while rank 3 and rank 4 mice (winning ratio <0.5) were the subordinate mice.

2.4 Inducing and characterizing chronic social stress

2.5 Anxiety tests

All animals were handled for a minimum of 2 days prior to onset of baseline anxiety measurements, allowing the mice to habituate to the experimenter interaction. TopScan (Clever Sys Inc.) video tracking system recorded the time spent in each zone, as well as bouts of entering each zone and total distance travelled as measures of exploration.

2.6 Statistical analyses

Animals were randomly assigned to treatment groups, but cage groups were matched by weight. Where possible, the experimenter was blinded to treatments. Statistical analyses were performed using GraphPad Prism software v7 (https://www.graphpad.com, RRID:SCR_002798). All values are given as a mean ± SEM. All statistical tests were two-tailed and the significance was assigned at p < .05. The D’Agostino-Pearson omnibus normality test and Brown–Forsythe test were used to test normality and equal variances between group samples, respectively. When normality and equal variance between sample groups was achieved, ordinary one-way ANOVA (followed by Tukey's multiple comparisons test), repeated measures two-way ANOVA, unpaired or one sample t-tests were used. Where normality or equal variance of samples failed, Mann–Whitney U test, Kruskal–Wallis test with Dunn's post hoc multiple comparison, or Wilcoxon signed rank test was performed. Linear regression and Fisher's exact tests were used for correlation and contingency analyses.

2.7 Experiment 1: Effect of CSD on novel hierarchy formation

After a baseline SI test, one cohort of animals (n = 80) underwent 15 days of CSDS, followed by another SI test. A subset of animals also underwent SPT and a one-week wheel-running assay. Based on the SI scores (see Section 2.4.2), CSDS-exposed animals were split into two groups, the stress-resilient or stress-susceptible. One mouse died during CSDS making n = 76 for subsequent analyses (as three stress-naïve males due to be the cage mates were also excluded). The mice were then housed in weight-matched groups such that there was one CSDS-stressed mouse together with three stress-naïve controls in one cage (total of N = 19 cages). After 2 weeks of hierarchy formation, TT and supporting dominance tests were performed as described in Section 2.3. Anxiety was measured prior to starting and after completing all behavioral tests.

2.8 Experiment 2: Effect of social status on stress susceptibility

To delineate whether newly formed hierarchies would predispose a certain rank to greater stress susceptibility, a second cohort of mice (n = 68) was first matched by weight and group-housed immediately after trait anxiety tests (total of N = 176 cages). After ranks were determined in dominance tests and a baseline SI was recorded, all ranks underwent CSDS with a subset of mice (n = 12) serving as CSDS-controls (and therefore being excluded from further comparison based only on CSDS-exposed mice). We report death of one control mouse during CSDS due to unrelated sickness (its dominance data were still used for TT validation). This did not impact subsequent analysis of rank 1 versus rank 4 differences. The same tests of stress were performed as in Experiment 1, followed by state anxiety recordings. To minimize the variability due to different hierarchical structures, main comparison was limited to the clear alphas or omegas of a group, the rank 1 (n = 17) and rank 4 mice (n = 16) respectively.

3.1 Hierarchy formation and TT validation

Following 2 weeks of agonistic activity, establishment of social hierarchies among cage mates was determined in a battery of social dominance tests. In the TT, the measured hierarchies were consistent over the four-day testing period, both within and across groups respectively (Figure 1b,c), and exhibited a linear trend from the most dominant to the most subordinate animal (Figure 1d: One-way ANOVA F(3,140) = 361.9, p < .0001: post-test for linear trend slope = −0.2134 ± −0.0065, R² = .8858, F(1,140) = 1,086, p < .0001). In a linear hierarchy, the top individual (“alpha”) dominates over all others. Each subsequent rank is singly occupied, down to the most subordinate mouse (“omega”) that is dominated by all other members. Nearly 78% of hierarchies observed in this experiment were linear (Figure 1e), but other hierarchical structures were observed, such as nontransitive and despotic (one alpha with other members not having clear ranks). The TT rank is not induced by the testing procedure, as the time spent in the tube during the training phase was not an indicator of success during testing (Figure 1f: Kruskal–Wallis H(3) = 3.83, p = .2804) and neither was the weight profile before (Figure 1g: One-way ANOVA F(3,88) = 0.3467, p = .9576) or after testing (Figure 1h: One-way ANOVA F(3,88) = 0.1442, p = .9331). The validity of TT-obtained ranks is supported by correlation with other dominance measures that highlight different manifestations of dominance behavior. Dominance ranks from the TT were consistent with ranks obtained via three other methods: territoriality in the UMA, spontaneous fighting in the agonistic behavior test (ABT), and resource competition in the WST (Figure 1i: Fisher's exact t-tests (2-sided) UMA p = .0509, ABT p = .0229, WST p = .0090). The degree of correlation ensures that dominance is the common underlying factor being measured. Ranked mice belonged to two experimental groups based on exposure to CSDS before or after hierarchy formation, as described in Sections 2.7 and 2.8.

3.2 Experiment 1: Effect of CSD on novel hierarchy formation

The full experimental timeline is shown in Figure 2a.

3.3 Experiment 2: Effect of social status on stress susceptibility

The full experimental timeline is shown in Figure 3a.

4.1 Chronic exposure to social defeat did not render mice subordinate

We anticipated that CSDS exposure may have differential effects on hierarchy formation such that susceptible mice would be more prone to subordination as they exhibit compromised ability to handle stressful conflict situations. In contrast, we predicted that resilient mice would exhibit higher dominance as they may have acquired a more adaptive strategy for adjusting to new social cohorts, which enabled them to overcome any putative adverse effects of CSDS. However, we found that both susceptible and resilient mice exhibited comparable dominance levels, as evidenced by the equal distribution of the winning ratios.

While depression is mainly associated with despondency and social withdrawal, aggression is also a common symptom in human depressive states (Van Praag, 1998). Moreover, the chronic unpredictable stress model was reported to increase aggression, hostility, and social dominance in rodents (Yang et al., 2015). Our study did not quantify push, retreat, and resistance behaviors of animals within the tube. Therefore, we cannot exclude the possibility that CSDS-exposed mice win via a more or less effortful strategy than the stress-naïve controls, for example by “freezing” in the tube instead of pushing until the opponent retreats. CSDS could also lead to an increase in the levels of glucocorticoids, which might play a key role in shaping the behavioral trajectories leading to social rank attainment via glucocorticoid receptor activation, specifically in the nucleus accumbens (Papilloud et al., 2020). Additional measures of animals’ agonistic propensity could delineate whether increased aggression or other behavioral strategies would account for these observations.

Glucocorticoids are reported to modulate neurobehavioral plasticity engaged in shaping social subordination (Weger et al., 2018). Thus, we predicted that chronic stress, which would lead to long-term elevation of glucocorticoids, would increase the probability of mice exhibiting subordinate behavior when exposed to new colonies because of reduced levels of aggression. However, this was not the case, either because CSDS was not as stressful for these animals, especially considering they did not show any signs of anhedonia or increased state anxiety, or because the animals used other strategies to develop a good social ranking. As we did not directly measure aggression during hierarchy formation, we cannot confirm how the aggression levels of CSDS-exposed animals were compared to the stress-naïve individuals. However, the relationship between glucocorticoid levels and social rank is complex and has been described as highly dependent on social context (Weger et al., 2018). We do not expect social isolation prior to CSDS to have had significant effects on glucocorticoids, as endocrine changes in isolated compared to group-housed animals are very limited (Benton & Brain, 1981; Holson et al., 1991; Misslin et al., 1982).

To further determine the effects of stress on susceptible/resilient mice, we investigated whether diurnal activity was disrupted following CSDS exposure. Since a number of studies report that stress affects circadian rhythms and sleep-wake cycle (Bunney et al., 2015) we predicted that stress exposure following CSDS would lead to disruptions in daily rhythms. Daily wheel-running activity is a standard measure of internal rhythms where mice will typically exhibit low activity in the day and high activity at night. We did not observe any obvious effect of CSDS on total activity counts in daily wheel-running activity since both susceptible and resilient mice exhibited similar rhythms to stress-naïve mice.

While we report measures of sociability, anxiety, and locomotive behavior, it is nonetheless difficult to account for the complete array of side-effects that CSDS may have on cognition and physiology, and how these in turn may affect social dominance. Though our initial study measured the effect of stress on winning ratios in completely new social groups, another valuable question would be to examine how stress would affect integration of an individual in an already established group of stress-naïve conspecifics.

4.2 Dominant and subordinate mice are equally susceptible to the adverse effects of chronic social stress

The differences in social competitiveness were not related to the overall differences in stress susceptibility. However, subordinate, but not dominant, mice exhibited decreased average baseline social preference prior to CSDS. This may be indicative of social stressors experienced by these groups during hierarchy formation which coincides with recent suggestions of high levels of intrinsic stress in selectively bred socially submissive mice (Bairachnaya et al., 2019). Repeatedly defeated males showed classical stress responses such as decreased body weight gain, increased adrenals, and increased basal corticosterone levels (Haller et al., 1999). Like repeated defeat from aggressive residents in CSDS, low social status in social groups can also induce chronic stress in male rodents (Blanchard et al., 1995). Therefore, it is not surprising that subordinate males are already showing signs of chronic stress after 2 weeks of group-housing. CSDS following 2 weeks of group housing likely added further stress, driving the subordinate population to even lower sociability which indicates increased social stress. However, both hierarchical groups exhibited susceptible phenotypes as evidenced by their comparably low social interaction scores.

Defeat involves the loss of social status which would be more pertinent for those males that enjoyed a higher social status in original colonies, than for males who already lost and attained only low dominance over other group members. A recent study reported that dominant males were indeed more susceptible to CSDS (Larrieu et al., 2017) while we found that dominant mice exhibited the greater change in sociability when we compared SI scores before and after CSDS relative to the subordinates. However, we did not find any evidence that the dominants were more stressed than the subordinates. In comparison, Larrieu et al. used 5 weeks of cohabitation prior to dominance testing and CSDS, while we opted for 2 weeks of cohabitation as it has been shown to be sufficient for hierarchy formation. Therefore, the inconsistency in stress susceptibility may arise because the effects of group-housing are relatively mild at 2 weeks and/or because more profound effects are only observable after prolonged occupation of a certain rank. One hypothesis may be that the dominants suffer more severe stress when having to maintain their position throughout a longer period, while during the initial establishment of hierarchy the subordinates experience more stress. Our findings support this hypothesis since subordinate mice already exhibited a susceptible-like sociability phenotype prior to introduction of any additional stressors as evidenced by the significantly lower pre-CSDS SI ratio. Another recent study also reported subordinate mice having higher depressive-like behavior, as well as hormonal and expression levels of genes associated with stress, after only 2 weeks of group-housing (Horii et al., 2017). As baseline SI measurements were not reported by Larrieu et al. we cannot make a direct comparison with hierarchies maintained for a longer period.

It is possible that increased vulnerability to chronic stressors may be an effect of long-term dominance, arising from the struggles to maintain the rank position, similar to how long-term subordination in the visible-burrow system induces a stress-phenotype (Blanchard et al., 1995). Nonetheless, studies on hierarchy maintenance in mice showed that there was a large degree of variability between social groups in overall stability, time taken in establishing the hierarchy and in the degree of despotism of the alpha male (Williamson et al., 2016). As a result, we would expect rank-related differences to arise over a variable timescale, making duration of group-housing a significant contributor to the effects of social hierarchies on behavior and physiology. Moreover, another explanation for the differences between the studies may be due to the type and the strength of stressors used. Acute and chronic stressors can have very different effects on neurophysiological function where, for example, strong stressors lead to increased firing, but longer term weaker stressors induced decreased firing in the brain reward circuits (Chaudhury et al., 2013; Tye et al., 2013; Venzala et al., 2013). Thus, differences in the stress paradigms used in these studies likely induced different changes in neural circuits that will have different effects on behavioral processes such as motivation and aggression during hierarchy formation resulting in variations in observed responses (Zhou et al., 2017).

It was recently reported that dominant mice exhibited increased sleep fragmentation which were hypothesized to be due to the stressful effects of having to constantly maintain a dominant status (Karamihalev et al., 2019). Since sleep and circadian rhythms are intimately linked, we wondered whether dominant and subordinate mice would exhibit different daily rhythms following stress exposure. However, analysis of daily wheel-running activity did not show any difference in diurnal total activity counts per hour in dominant and subordinate mice.