Comparison of chronic physical and emotional social defeat stress effects on mesocorticolimbic circuit activation and voluntary consumption of morphine

Chronic social defeat stress (CSDS) is a well-established rodent model of depression that induces persistent social avoidance. CSDS triggers molecular adaptations throughout the mesocorticolimbic reward circuit, including changes in the activity of dopamine neurons in the ventral tegmental area (VTA), that may also influence drug reward. One limitation of traditional, physical CSDS (PS) is that injury complicates the study of opiate drugs like morphine. Thus, we sought to characterize a variation of CSDS, termed emotional CSDS (ES), that eliminates this confound. We assessed the effect of PS and ES on mesocorticolimbic circuit activation, VTA gene expression, and morphine intake. We found that PS and ES similarly induced ΔFosB in the hippocampus, but only PS significantly increased ΔFosB expression in the prefrontal cortex and striatum. In contrast, cFos expression was similarly reduced by both PS and ES. Interestingly, we found that PS and ES similarly increased voluntary morphine consumption immediately following stress, despite differences in the magnitude of the depressive phenotype and striatal ΔFosB expression at this time point. Combined, these data suggest that both stress paradigms may be useful for investigation of stress-induced changes in drug behavior.

shown to produce similar behavioral changes to physical CSDS (hereafter, PS), such as decreased social interaction and sucrose preference, and also appears to induce similar molecular changes in the VTA, as evidenced by RNA sequence analysis (RNAseq) 13 . Though ES offers many advantages for the study of the interaction between mood-disorders and opiate addiction, complete characterization of the effects of ES on both behavior and reward circuitry signaling is needed to fully interpret such studies.
We performed a side-by-side comparison of PS-and ES-induced effects on immediate-early gene activation, VTA gene expression, and opiate reward and consumption to determine whether ES and PS elicit similar changes. First, we compared ΔFosB and cFos induction following PS and ES. Additionally, we used RT-PCR to validate ES upregulation of VTA genes previously identified with RNAseq 13 . Lastly, we determined the presence of an immediate increase in opiate reward and consumption following both ES and PS, despite differences in social avoidance phenotypes at this time point. These data suggest that the molecular changes occurring immediately after either stressor directly affect opiate sensitization and that mechanisms responsible for long-term neuroadaptations that maintain depressive-like behavior may play less of a role, as morphine preference was normalized two weeks following CSDS. Together, this study demonstrates the similar molecular and behavioral responses to both CSDS paradigms and highlights ES as a useful model in the study of stress-induced changes in opiate reward and intake.

Results
Validation of Physical and Emotional CSDS paradigm. Male mice were subjected to 10-day physical (PS) or emotional (ES) CSDS and social interaction (SI) ratios were assessed 1, 14, or 28 days post-stress (Fig. 1). Consistent with previous results, there were significant main effects of both stress (F (2,92) = 30.82, p < 0.0001) and time (F (2,92) = 3.276, p = 0.0422), while the stress × time interaction was not significant (two-way ANOVA). Specifically, PS mice had significantly reduced SI ratios compared to controls on D1, D14 and D28, and were also significantly different from ES mice on D1 and D14 (Tukey's test, *p < 0.05). ES mice have reduced SI ratios compared to controls on D1 (Tukey's test, *p < 0.05), although they are also significantly different from PS on D1 and D14, illustrating a more modest phenotype. This time-course is similar to published results, where PS mice have a maximal reduction in SI ratio on D1 but ES do not display a similarly robust social avoidance phenotype (~0.6 SI ratio) until D28 13 . Due to the increased variability in both stress groups and controls on D28, and to avoid possible confounds of prolonged social isolation, we used D14 as an intermediate time point for long-term effects in behavioral studies.

PS-and ES-induced Changes in ΔFosB and cFos in Mesocorticolimbic Brain Regions. It has
been shown that PS robustly increases FosB and ∆FosB (a splice variant of FosB) expression throughout the mesocorticolimbic circuit 24 hr following social defeat [15][16][17] . Additionally, ΔFosB protein levels (western blot) in PS mice correlated with extent of social avoidance in susceptible animals in NAc and PFC 15 . Moreover, direct alteration of ΔFosB levels in NAc and PFC via viral-mediated overexpression was sufficient to alter depressive-like behaviors, supporting at link between ∆FosB expression and depressive-like phenotypes 17,18 .
In order to determine whether PS and ES induce similar changes in brain ΔFosB immunoreactivity, the average number of ΔFosB-positive cells (per mm 2 ) was assessed in select brain regions on D1 following standard PS and ES (Table 1). Consistent with previous reports, we found robust increases in ΔFosB expression in the PFC and dorsal and ventral striatum of PS mice, but interestingly, this was not observed in ES mice, where expression was similar to controls (Table 1, Fig. 2A,B). Specifically, there was significantly increased ΔFosB expression in PS mice compared to both controls and ES mice within the Cg1 ( Fig. 2A Given that we also observe a significant difference in SI ratio between PS and ES mice at this time point, we next completed a correlational analysis to determine the relationship between ΔFosB expression and SI score. We found that ΔFosB counts were significantly negatively correlated with SI ratio in the PFC (Cg1: r = −0.45, Figure 1. Establishment of chronic social defeat stress (CSDS). Social interaction (SI) was assessed at three time points (Day (D) 1, D14, D28) in separate cohorts following physical (PS) and emotional (ES) stress. D1: control n = 20, PS n = 27, ES n = 24; D14 and 28: control, PS, and ES n = 5 mice/group; *p < 0.05 compared to control, $ p < 0.05 compared to PS, Tukey's post-hoc test. Fig. 2G; PrL: r = −0.43, p = 0.014, data not shown), such that mice with lower SI ratios (i.e. more "depressed") had higher ΔFosB expression. This was also the case in the dorsal striatum (CPu: r = −0.39, p = 0.0220, Fig. 2H) and the NAc core displayed a similar trend (r = −0.19, p = 0.28). Together, these data support a link between FosB expression in the PFC and striatum and depressive-like behavior, and suggest that the moderate phenotype in ES mice may be related to minimal activation of these brain regions.
While ES did not increase ΔFosB expression in the PFC and striatum, it did increase ΔFosB expression in the hippocampus (Table 1, Fig. 2C). Specifically, we found that there was significantly increased ΔFosB expression in the dorsal hippocampus DG region of both PS and ES mice compared to controls ( Fig. 2F: F (2,30) = 4.98 p = 0.0136, Tukey's test *p < 0.05). In contrast to the PFC and striatum, ΔFosB counts were not significantly correlated to SI ratio in the hippocampus (Fig. 2I). This may suggest that ΔFosB expression in dorsal hippocampus is not linked to depressive-like behavior, but that perhaps other CSDS-related behavioral outputs, such as anxiety-like behaviors, may be driven by hippocampal ΔFosB. We also assessed ΔFosB expression in the VTA and DR as these are major dopaminergic and serotonergic nuclei within the brain and alteration of dopaminergic and serotonergic signaling has been implicated in the etiology of depression. However, we did not observe any statistical differences between groups ( Table 1), suggesting that chronic ES and PS do not robustly induce ΔFosB in these brain regions.
We next examined the number of cFos-positive cells in separate tissue sections from the same mice (Table 2). Given the differences in the time-courses of cFos and ΔFosB expression 19 , it is likely that cFos induction is driven by the experience of the social interaction test and not the effect of CSDS per se. The pattern of cFos regulation differed from that of ΔFosB in three main ways. Firstly, we observed a general decrease in cFos expression following CSDS, in contrast to ΔFosB expression which was generally increased throughout the mesocorticolimbic circuit. Secondly, we found that exposure to PS and ES always produced similar changes in cFos expression. And thirdly, we did not observe a correlation between SI ratio and cFos expression in any of the regions assessed, suggesting that induction of a depressive-like phenotype was not required for the decreased cFos expression, but that the experience of previous CSDS itself was sufficient.
Overall, we found significant ΔFosB and cFos expression changes within the mesocorticolimbic circuit of both PS and ES mice. Changes in ΔFosB expression were more robust and complex compared to cFos expression, with brain region-specific differences (PFC and striatum) and similarities (hippocampus) observed between PS and ES mice.

PS-and ES-induced Changes in Voluntary Morphine Consumption.
ΔFosB induction in the mesocorticolimbic circuit has been widely studied as a molecular mechanism underlying of aspects of addictive behavior 20 and given that PS alters drug reward 4-6 , we sought to determine whether ES also affects drug consumption. Given the increase in the prevalence and use of pain-relieving opioid drugs, we chose to assess the voluntary intake of morphine using a two-bottle choice assay. C57Bl6/J mice will voluntarily consume  Importantly, there was no difference in the total intake volume between groups given a choice (morphine/quinine in 0.2% sucrose) and vehicle solution (0.2% sucrose), indicating that the presence of morphine/quinine did not impact overall consumption of fluids (d1-d2 prior to choice: control group = 5.62 ± 0.20 ml, choice group = 5.74 ± 0.23 ml; d3-5 during choice: control group = 5.46 ± 0.14 ml, choice group 5.63 ± 0.22 ml). Next we monitored voluntary morphine consumption following exposure to PS and ES. Within this cohort, at D1, PS mice had significantly reduced SI ratios compared to controls and ES mice (F (2,28) = 40.22, p < 0.0001, Tukey's test *p < 0.05), while ES mice did not significantly differ from controls (Fig. 5A). While PS and ES groups had significantly different SI ratios, both groups exhibited a similar significant increase in morphine preference compared to controls ( Fig. 5B: F (2,28) = 7.265, p = 0.0029, Tukey's test *p < 0.05). There was also a significant negative correlation between SI ratio and morphine preference (Pearson r = −0.3596, p = 0.047), such that mice with greater social avoidance also displayed higher morphine preference (Fig. 5C). We also examined the total volume of morphine consumed and found that stress increased morphine intake, with ES exhibiting significantly increased intake compared to controls (Control = 3.91 ± 0.32 ml, PS = 4.49 ± 0.22 ml, ES = 4.97 ± 0.19 ml; F (2,27) = 4.1, p = 0.0279, Tukey's test *p < 0.05). These changes were not dependent on body weight as there were no differences in body weight between groups during morphine choice assays (Control = 22.3 ± 0.5 g, PS = 22.8 ± 0.3 g, ES = 23.0 ± 0.5 g; F (2,28) = 0.76, p = 0.48). These data are consistent with previous work that found increased cocaine CPP in PS-susceptible mice 4 . Moreover, while morphine preference was significantly correlated with SI score, a susceptible phenotype is not necessary to increase morphine reward, as ES mice exhibited increased morphine preference without a significant decrease in SI score. Further, the data suggest that the increase in morphine intake of PS mice is not necessarily dependent on physical trauma or analgesic effects on the resulting pain, as ES, a non-physical stress exposure, was sufficient to increase morphine preference as well.
Given that CSDS induces a long-lasting change in SI ratio, we examined whether changes in morphine reward also persist using a second cohort of mice. SI ratio was assessed on D1 and again PS mice had a significant reduction in SI ratio compared to controls, while ES mice did not differ significantly from controls (Control: 1.2 ± 0.05, PS: 0.51 ± 0.13, ES: 1.1 ± 0.05; F (2,37) = 14.23, p < 0.0001, Tukey's test *p < 0.05). We reassessed SI on D14 before drinking studies began and confirmed similar SI scores to those from D1 ( Fig. 6A: F (2,37) = 12.39, p < 0.0001 Tukey's test **p < 0.01). At this time point, however, the morphine preference did not significantly differ between groups ( Fig. 6B: F (2,36) = 0.77, p = 0.47), but we observed a significant increase in morphine intake in PS mice compared to controls (Control = 4.09 ± 0.23 ml, PS = 4.95 ± 0.22 ml, ES = 4.51 ± 0.18 ml; F (2,36) = 3.907, p = 0.0291, Tukey's test *p < 0.05). The discrepancy between these two measures was likely due to differences in body weight between groups at D14, as PS and ES mice weighed significantly more than control mice (Control = 24.74 ± 0.5 g, PS = 26.6 ± 0.3 g, ES = 26.4 ± 0.3 g; F (2,36) = 6.92, p = 0.0029, Tukey's post-hoc test p < 0.01), and when intake was normalized to body weight, there was no longer a significant difference between groups (F (2,36) = 2.07, p = 0.14). This difference in body weight was surprising, as previous reports indicate no difference in body weight of PS or ES mice at D28 4,13 . Together these data suggest that while PS and ES similarly increase morphine reward shortly following defeat, this phenotype is not maintained at later time points, despite the persistence of the depressive-like phenotype.

PS-and ES-induced Changes in VTA Gene Expression.
It has been well established that VTA activity is critical for susceptibility to PS 4,8,9 , and changes in VTA gene expression have been implicated as a potential mechanism. Furthermore, we have identified changes in VTA gene expression via RNA-sequencing analysis following chronic morphine exposure, including induction of serum and glucocorticoid-regulated kinase 1 (SGK1), one of the few genes similarly increase by both morphine and cocaine exposure 24 . Therefore, we compared PS-and Mice underwent PS or ES and VTA tissue was collected for analysis 24 hours following SI testing on D1 as described in previous studies 4,9,13 . Using RT-PCR, we confirmed that SGK1 gene expression was significantly increased in both PS and ES mice (Fig. 7A: F (2,67) = 5.75, p = 0.0050, Tukey's test *p < 0.05). However, SGK1 expression did not correlate with SI ratio (data not shown), suggesting that while SGK1 is induced following chronic stress, it may not drive social avoidance behavior. We also examined SGK1.1, a brain-specific SGK1 isoform known to influence neuronal excitability 25,26 and found that SGK1.1 expression was not significantly altered by stress exposure (Fig. 7B: F (2,51) = 1.048, n.s.) nor did it correlate with SI ratio (data not shown). We also assessed expression of other candidate genes involved in neuronal excitability identified from previous RNA sequencing (Prkcd, Kcnj13, Ramp3, Gabrd), but did not observe any significant differences between groups (Table 3). These data suggest that alterations in VTA gene expression driven by social stress may be subtle, and further work will be required to uncover VTA genes that drive cellular and behavioral responses to stress.

Discussion
The goal of these studies was to determine whether a variant of CSDS that eliminates physical contact, termed emotional or vicarious CSDS 13, 27 , elicits similar changes in mesocorticolimbic circuitry, reward behavior, and VTA gene expression to those caused by traditional physical CSDS. Similar global changes in VTA gene expression have been previously observed 13 , even at a time point (2 days post-stress) when there are differences in the magnitude of depressive-like phenotypes between PS and ES, suggesting that there may be overlap in the neurobiological underpinnings. Our immunohistochemistry data largely confirm this, as we observed similarly decreased cFos expression in PS and ES mice as well as similar induction of ΔFosB in the hippocampus following both forms of stress. However, we did observe differences in ΔFosB induction between PS and ES in the cortex and striatum, where ΔFosB was significantly increased in PS mice compared to ES mice and correlated with SI score. This suggests that ΔFosB signaling in these regions may be more closely linked with avoidance behavior, with signaling in other areas more indicative of a generalized stress response, and perhaps the anxiety-like phenotype. For example, increasing ΔFosB levels in the PrL region of the PFC is sufficient to increase susceptibility to the social withdrawal phenotype of physical CSDS, without significantly affecting PS-induced anxiety-like behaviors 17 . Moreover, biochemical data show a negative correlation between NAc ΔFosB levels and SI score in susceptible PS mice, where more "depressed" mice have higher NAc ΔFosB protein expression 15 , which is consistent with the increase in NAc ΔFosB-positive cells and correlation with SI score in our current study. However, CSDS-induced changes in NAc ΔFosB expression are complex, as it has also been shown that increasing NAc ΔFosB expression induces resilience to the social withdrawal effects of PS without affecting baseline anxiety-like behaviors, although stress-induced anxiety was not measured 18 . The apparent contradiction between NAc ΔFosB playing a role in both susceptibility and resilience to PS is likely due to induction in separate subsets of medium spiny neurons (MSNs): susceptible mice display increased ΔFosB in D2-type MSNs, while resilient mice show an  Table 3. RT-PCR analysis of candidate gene expression in the VTA of control, physical stress, and emotional stress mice 24 hr after social interaction testing (~48 hours after last day 10 d of chronic social defeat stress). Data are expressed as fold change +/− SEM, n = 11-25 mice/group. increase in ΔFosB in D1-type MSNs 16 . Thus, our lack of ΔFosB induction in the PrL and NAc of ES mice is consistent with a modest depressive-like phenotype at this time point, and is also consistent with recent biochemical data that similarly observed no change in NAc ΔFosB protein in ES mice immediately following CSDS 27 . Future studies could potentially explore whether the lack of ΔFosB induction in PFC and striatum of ES mice results from (or perhaps drives) a more subtle depressive phenotype, both by examining a larger cohort of mice with a wider range of SI scores and by virally manipulating ΔFosB expression or DNA binding in ES mice.
In contrast, we found that PS and ES similarly induced ΔFosB in the DG region of the hippocampus. In our immunohistochemistry studies, we did not separately analyze mice susceptible and resilient to PS, as we had a low proportion of resilient mice (2/12 mice) in this cohort. Nevertheless, we observed a significant increase in ΔFosB expression, putting our work slightly at odds with previous data that found ΔFosB expression in the DG was only induced in mice resilient to PS 17 . However, our observation of increased ΔFosB expression in both PS and ES mice is consistent with both models inducing a similar anxiety-like phenotype at this time point 13 . Moreover, the idea of separate mechanisms underlying the depressive-and anxiety-like phenotypes induced by CSDS is supported by work in PS mice, as both susceptible and resilient mice exhibit similar increases in anxiety-like behavior, despite their differences in depressive-like tests (SI and sucrose preference) 4 . Finally, the DG is of particular interest as a candidate for CSDS-induced changes in anxiety behavior given recent data that increasing ΔFosB in mouse DG is sufficient to induce an anxiety-like phenotype 28 . Unfortunately, we were not able to correlate ΔFosB expression with anxiety-like behavior in these studies since we perfused animals shortly after the SI test. However, given the similar magnitude of anxiety-like behavior reported for PS-and ES-mice at this time point, we might predict a positive correlation between these measures and this should be explicitly tested in future studies. Additionally, investigation of whether direct manipulation of ΔFosB expression or activity is sufficient to promote or prevent PS-and ES-induced changes in anxiety-like behavior would be of interest. Overall, our data support the idea that ES induces changes in ΔFosB expression consistent with the PS-induced ΔFosB patterns previously correlated with depressive-and anxiety-like behaviors.
Interestingly, cFos expression was decreased following SI testing in stressed mice. To our knowledge, this was the first study to examine both ΔFosB and cFos changes immediately after SI testing, instead of the typical 24 hour time point following SI testing. Our approach allows a direct assessment of the cFos-derived regional activity caused by social interaction in stressed mice. The cFos data suggest that the pattern of brain activation during SI testing is different between controls and animals that have experienced either PS or ES, but is not dependent on a susceptible, depressive-like phenotype. It would be interesting to investigate whether the decrease in cFos activation in stressed mice also occurred following other behavioral tests, for example assays of anxiety-like behaviors such as elevated plus maze. This might help delineate whether activation changes are specific to exposure to social targets, or are generally induced by exposure to an environment that may induce a physiological stress response. This correlation may also be investigated by measuring stress hormones such as corticosterone, given that mice exposed to CSDS (whether susceptible or resilient) display increased corticosterone and anxiety-like behavior. In contrast to our observation of a general decrease in cFos in the mesocorticolimbic circuit following CSDS in mice, increased cFos has been observed following chronic intermittent stress in rats 29 , an interesting difference given that the models produce other similar neuroadaptations in the mesocorticolimbic circuit 30 . This may be due to procedural differences between the two studies (our study examined cFos 1 day after the last stress episode, immediately following SI testing while the study from Nikulina et al. examined cFos 1 week following the last stress and immediately after a saline injection), but further exploration is warranted.
While the role of VTA DA neuronal activity in susceptibility and resilience to CSDS has been well established 11, 12, 31 , we did not observe changes in ΔFosB expression in the VTA of CSDS mice, consistent with previous reports 17 . Although there is evidence that regulation of VTA potassium channels can contribute to the depressive-like phenotype 4, 12, 32 , the molecular mechanisms underlying the increase in VTA DA activity remain largely undefined. One promising candidate for stress-induced adaptations is serum-and glucocorticoid-regulated kinase 1 (SGK1), which is increased in select brain regions following chronic unpredictable mild stress 33 . Indeed, we observe that both PS and ES robustly increase SGK1 gene expression in the VTA, consistent with published RNAseq results 13 . While SGK1 targets in the brain have not been elucidated, SGK1 regulates the expression and activity of a variety of ion channels and transporters in the renal system 34 , suggesting activation of SGK1 in the VTA may play a role in changes in stress-induced changes in cell excitability. Future studies that examine the protein levels of SGK1, as well as direct manipulation of SGK1 expression and activity, are critical to determine what, if any, role VTA SGK1 plays in stress-induced behaviors.
Interestingly, SGK1 gene expression is also robustly increased in the VTA by administration of drugs of abuse, such as cocaine and morphine 24 . CSDS is known to impact drug-related behaviors, with responses generally consistent with increased drug intake, reward, and sensitization. Here, we show that chronic ES also increases drug reward, with ES mice displaying an increase in morphine preference in a two-bottle choice oral morphine assay. Given the established role for NAc ΔFosB in mediating cocaine and morphine CPP 35,36 , it is interesting that we observe an increase in ES morphine preference without a significant increase in ΔFosB expression. This may suggest that distinct but overlapping mechanisms can drive CSDS-drug cross-sensitization. This is further supported by data that ES and PS mice displayed a similar increase in morphine preference immediately following stress, despite differences in the severity of depressive-like behavior at this time point. However, given previous data the mice resilient to PS do not display changes in cocaine CPP, it suggests that exposure to CSDS alone is not sufficient to alter drug reward, and that induction of at least a modest depressive-phenotype may be necessary. Together, the ability of both ES and PS to increase morphine preference and intake is consistent with the broader literature that exposure to a variety of acute and chronic stressors is sufficient to increase opiate reward, intake and reinstatement in rodents [37][38][39][40][41] , consistent with reports of stress-associated increases in craving in humans 42 . However, the data available on stress-induced changes in opiate reward and self-administration is much more limited than that of cocaine, and some differences have been noted. For example, social defeat stress has been shown to promote escalation, or "binge" self-administration of cocaine, but not heroin, in rats 43 , consistent with data that neurobiological mechanisms underlying opiate and psychostimulant addiction can be distinct 44 , and that findings generated using stimulant models may not translate to opiate drugs.
When we examined morphine intake 14 days following PS and ES, morphine preference of stressed mice no longer differed from controls, suggesting that long-term changes in neural activity and/or gene expression that maintain the depressive-like phenotype are not sufficient for a persistent change in morphine reward in adult mice. While a significant amount of work has focused on the role of stress-induced changes in mesocorticolimbic circuit function in drug responses, much remains unknown about the molecular mechanisms responsible and their persistence, partially due to the variety of stress models and species studied 7 . Thus, identification of candidate mechanisms induced in multiple stress models, such as VTA SGK1, are critical for future studies in order to better characterize the effects of stress on both the depressive-like phenotype and on sensitized drug responses and eventually translate findings to humans.

Methods
Animals. Adult (8-9 wk) male C57BL/6J mice (Jackson Laboratory) and CD1 retired breeders (Charles River) were habituated to the animal facility > 1 week prior to experiments. All animals had ad libitum access to standard chow and water and were kept on a 12 hr light-dark cycle. All experiments were approved by the Michigan State University Institutional Animal Care and Use Committee (IACUC) and were carried out in accordance with the guidelines set in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health.

Physical and Emotional Chronic Social Defeat Stress (CSDS). Physical and emotional CSDS were
performed as reported previously 14,45 . Briefly, retired male CD1 breeders were initially screened for consistent home-cage intruder aggression before use as aggressors in CSDS. Experimental mice (C57BL/6J) were either placed into the home cage of a CD1 aggressor (physical stress, PS), or placed on the opposite side of the perforated plexiglass divider where they could witness the aggressive encounter (emotional stress, ES) for 8-10 min. PS and ES mice were then separately co-housed with a CD1 aggressor, but kept physically separated from the aggressor via a perforated divider for the remaining 24 hr, allowing for continued sensory stress. PS and ES mice were exposed to a new CD1 aggressor daily, for 10 consecutive days, and all behavioral experiments were carried out between 8:00 and 11:00 AM.
Non-stress C57BL/6 J control mice (Con) were co-housed with each other in standard cages but physically separated by a perforated plexiglass divider. This allows for physical isolation with continuous sensory contact similar to that experienced by PS and ES mice. Control mice were exposed to a novel conspecific daily for the 10-day experiment period.
Social Interaction (SI) Testing. Social interaction (SI) testing was used to designate mice as susceptible or resilient to CSDS following standard procedures 14,45 . Following the 10 th CSDS exposure, mice were singly housed in standard cages for 24 hr prior to SI testing on post-defeat day 1 (D1). SI was also assessed at later time points (D14 and D28) in some experiments, with mice remaining singly housed for the duration of the experiment prior to SI testing. The social interaction arena (42 cm × 42 cm) contained a cylindrical Plexiglas/wire-mesh enclosure located in the center of the back wall. The interaction zone was defined as an 8 cm region surrounding the cylinder. Locomotor activity and time spent in the interaction zone were quantified by video tracking software (TopScan Suite, CleverSys) for 2.5 min of free exploration (target absent), followed by a 2.5 min session with the CD1 target present in the wire-mesh enclosure. SI ratio was calculated as: (time spent in the interaction zone, target present)/(time spent in the interaction zone, target absent). Scores < 1.0 were defined as "susceptible, " and scores ≥ 1.0 were defined as "resilient" 4 .
Note: Immunohistochemistry for ΔFosB was performed using an antibody that recognizes full-length FosB, ΔFosB, and Δ2ΔFosB, as well as other Fra antigens 15 . Therefore, we cannot attribute the changes we observe in staining to any specific FosB gene product. However, we chose to use the term ΔFosB throughout the text for simplicity, and because ΔFosB was the predominant isoform observed by Western blot in similar experiments performed by our group in the past 15 .
Real-time polymerase chain reaction (RT-PCR). PS and ES was performed as described above, and mice were sacrificed 24 hrs following SI testing and VTA tissue was microdissected (14-gauge midline punch) and stored at −80 °C until RNA processing following standard protocols 50 . RNA was extracted using RNeasy micro columns (Qiagen) and reverse-transcribed with a high capacity cDNA reverse transcription kit (Applied Biosystems). Expression changes of candidate genes were assessed via RT-PCR using SYBR green (CFX connect, BioRad). Samples were run in triplicate and normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH) using the ∆∆Ct method 3 . Forward and reverse primer pairs are listed in Table 4.
Morphine Drinking Studies. Voluntary drinking was assessed by a two-bottle choice test using 50 ml plastic conical tubes fitted with standard sipper tops following established procedures 51 . Briefly, bottle placements (left or right side of standard home-cage) were switched daily to account for individual side-preference. Total fluid consumption was measured daily by changes in total bottle weight and by marking of fluid levels on each bottle. Habituation to the bottles and basal fluid consumption was assessed using tap water 1-2 days prior to morphine choice experiments. Morphine sulfate was kindly provided by the NIDA Drug Supply Program and quinine hemisulfate salt was purchased from Sigma (22640).
To determine a morphine concentration which would elicit ~75% morphine preference in non-stress control C57Bl/6 J mice, we conducted a pilot study using naïve male mice. Based on published results 22,23 , bottles contained either 0.3 mg/ml morphine or 0.06 mg/ml quinine (as a bitter taste control) in 0.2% sucrose. Fluid consumption is reported as either the mean fluid intake (ml) or morphine preference calculated as (morphine solution consumed/total fluid consumed) × 100. To determine if morphine consumption changed over time, morphine preference and consumption were assessed on D3-6, or D16-19 in separate cohorts. The average intake or morphine preference over 4 consecutive days is reported.