Chronic Physical and Vicarious Psychosocial Stress Alter Fentanyl Consumption and Nucleus Accumbens Rho GTPases in Male and Female C57BL/6 Mice

Chronic stress can increase the risk of developing a substance use disorder in vulnerable individuals. Numerous models have been developed to probe the underlying neurobiological mechanisms, however, most prior work has been restricted to male rodents, conducted only in rats, or introduces physical injury that can complicate opioid studies. Here we sought to establish how chronic psychosocial stress influences fentanyl consumption in male and female C57BL/6 mice. We used chronic social defeat stress (CSDS), or the modified vicarious chronic witness defeat stress (CWDS), and used social interaction to stratify mice as stress-susceptible or resilient. We then subjected mice to a 15 days fentanyl drinking paradigm in the home cage that consisted of alternating forced and choice periods with increasing fentanyl concentrations. Male mice susceptible to either CWDS or CSDS consumed more fentanyl relative to unstressed mice. CWDS-susceptible female mice did not differ from unstressed mice during the forced periods, but showed increased preference for fentanyl over time. We also found decreased expression of nucleus accumbens Rho GTPases in male, but not female mice following stress and fentanyl drinking. We also compare fentanyl drinking behavior in mice that had free access to plain water throughout. Our results indicate that stress-sensitized fentanyl consumption is dependent on both sex and behavioral outcomes to stress.


INTRODUCTION
Repeated and severe stress has long been associated with the emergence of psychiatric disorders including substance use disorders (Sinha, 2008;Hollon et al., 2015;Sapolsky, 2015;Newman et al., 2018;Wemm and Sinha, 2019). Opioid use disorder (OUD) is of particular concern as rates of opioid misuse have skyrocketed, especially in North America (GBD 2016 Alcohol andDrug Use Collaborators, 2018). Accompanying the increase in OUD is an alarming number of overdose deaths and decreased average lifespan (Dowell et al., 2017). Synthetic opioids are a major contributor to increased death rates, and in 2020, caused ∼75% of drug overdose deaths in the United States (Ahmad et al., 2021;Skolnick, 2021).
A wealth of literature has sought to uncover how stress drives increased susceptibility to substance use, and the neurobiological underpinnings of comorbid stress and substance use disorders [For recent reviews see: Newman et al. (2018) and Calarco and Lobo (2020)]. Psychosocial stress is of particular interest as it is thought to more closely mimic human stressors. However, the effects of psychosocial stress on substance use are often divergent and depend on the drug, stress duration, and species [reviewed in Neisewander et al. (2012)]. For example in rats, brief social stress increases cocaine self-administration and motivation for cocaine (Haney et al., 1995;Tidey and Miczek, 1997;Covington and Miczek, 2001;Burke and Miczek, 2015), while prolonged social stress reduces cocaine self-administration . In mice, prolonged social stress either promotes or suppresses cocaine self-administration (Yap and Miczek, 2007;Han et al., 2015Han et al., , 2017Arena et al., 2019) but we recently showed this depends on individual stress-response and social housing conditions (Engeln et al., 2021). There are comparatively fewer studies on psychosocial stress and opioid self-administration. In rats, social stress does not influence heroin self-administration (Cruz et al., 2011), while in mice it is associated with increased morphine preference (Cooper et al., 2017).
To investigate how stress influences vulnerability to synthetic opioid use, we adapted an oral fentanyl paradigm in male rats (Shaham et al., 1992) to male and female mice. In the Shaham model, rats were subjected to daily immobilization stress and presented with fentanyl in the homecage drinking water for 4 days ("forced consumption"). Then, rats experienced alternating periods of a choice between fentanyl and water, and additional forced consumption periods. Over time and as concentrations increased, stressed rats increased their fentanyl preference relative to unstressed rats. We sought to replicate this finding in male and female C57BL/6 mice using a similar alternating forced/choice protocol with increasing concentrations. We chose the C57BL/6 mouse since many transgenic tools are developed in this strain (Gong et al., 2007). Like Cooper et al. (2017), we used the standardized chronic social defeat stress (CSDS) procedure (Berton et al., 2006;Krishnan et al., 2007;Golden et al., 2011;Fox et al., 2020a,b). CSDS produces anhedonia in the majority of mice (∼60%, termed "stress-susceptible"), along with a decrease in motivated behaviors such as social interaction (Krishnan et al., 2007;Der-Avakian et al., 2014;Heshmati and Russo, 2015;Fox and Lobo, 2019). The other ∼40% are termed "stress-resilient" and do not display deficits in motivated behavior.
When studying stress and OUD vulnerability, it is also paramount to account for pain-related confounds (Dib and Duclaux, 1982) due the analgesic effects of opioids. Thus, to eliminate pain associated with physical injury (Cooper et al., 2017), we also used chronic witness defeat stress (CWDS), in which mice witness the social defeat of a C57BL/6 mouse and experience "vicarious" social stress (Warren et al., 2013(Warren et al., , 2014Iñiguez et al., 2018). Like CSDS, CWDS produces susceptible and resilient cohorts marked by decreased or unaltered social interaction and motivation, respectively. Since anhedonia and reduced motivation are associated with substance use and dependence (Hatzigiakoumis et al., 2011), we included stress-susceptibility as a factor in our analysis. Numerous brain regions are sensitive to both drugs and stress, including the nucleus accumbens (NAc) (Hollon et al., 2015;Newman et al., 2018;Calarco and Lobo, 2020). Indeed, both chronic stress and opioid exposure cause structural changes in the NAc that are thought to drive stress-susceptibility or increased drug intake (Matsubara et al., 1999;Robinson et al., 2002;Spiga et al., 2005;Diana et al., 2006;Christoffel et al., 2011a,b;Golden et al., 2013;Pal and Das, 2013;Graziane et al., 2016;Guegan et al., 2016;Kobrin et al., 2016;Francis et al., 2017;Cahill et al., 2018;Geoffroy et al., 2019;Fox et al., 2020a,b). These structural changes are primarily driven by Rho GTPases, and we and others have shown opioid withdrawal (Cahill et al., 2018) and CSDS (Francis et al., 2017;Fox et al., 2020a) engage and alter NAc RhoA signaling. We thus examined the consequences of our combined stress and fentanyl paradigm on expression of GTPases associated with dendritic remodeling (Nakayama et al., 2000;Negishi and Katoh, 2005;Newey et al., 2005;Chen and Firestein, 2007).
Here, we test three primary hypotheses. First, that psychosocial stressors will increase fentanyl consumption in mice, Second, that stress-sensitized consumption is dependent on stress response, and Third, that fentanyl and stress exposure will alter Rho GTPase expression in the NAc.

Experimental Subjects
All experiments were approved by the Institutional Animal Care and Use Committee at the University of Maryland School of Medicine (UMSOM) and performed in accordance with NIH guidelines for the use of laboratory animals. Mice were given food and water ad libitum and housed in the UMSOM vivarium on a 12:12 h light: dark cycle. Experimental mice were 8-9 weeks old male and female C57BL/6 mice bred at UMSOM. Male CD-1 retired breeders (Charles River, >4 months) were used as the aggressors for CSDS/CWDS. Mice were randomly assigned to control or stressed groups. The forced/choice cohort (see below) included n = 12 unstressed males, 12 unstressed females, 23 CSDS males, 19 CWDS males, and 20 CWDS females. The choice cohort included n = 11 unstressed males, 13 unstressed females, 31 CSDS males, 17 CWDS males, and 15 CWDS females. We excluded 1 unstressed male mouse from this experiment due to abnormally low social interaction behavior.

Chronic Social Stress
Social stress was performed as in our previous work (Fox et al., 2020a,b;Engeln et al., 2021;Morais-Silva et al., 2021), by using the modifications described in Warren et al. (2013) and Iñiguez et al. (2018) for vicarious resident-intruder stress. In chronic social defeat stress (CSDS), a male mouse (intruder) is physically defeated by an aggressive CD-1 (resident) for 10 min in a hamster cage containing woodchip bedding and a perforated divider. Only male mice are used in CSDS because CD-1s will not defeat female mice without modification to either the female mouse or the CD-1 (Takahashi et al., 2017;Harris et al., 2018). In chronic witness defeat stress (CWDS), a male or female mouse is housed on the opposite side of the perforated divider and allowed to witness the agonistic resident-intruder interactions for 10 min. The CSDS mouse is then housed opposite the resident, and the CWDS mouse is removed and housed opposite to a new CD-1 resident. Following 24 h of sensory interaction, the CSDS mouse is defeated by a new CD-1 resident while a different CWDS mouse witnesses the agonistic interaction. This process repeats for 10 days for 10 novel CD1-CSDS-CWDS pairings. Unstressed control mice are pair-housed across perforated dividers in cages containing woodchip bedding with a sex-matched conspecific for 10 days. Immediately following the last stressor, both stressed and unstressed mice are housed individually in woodchip bedding cages.

Social Interaction Testing
Twenty-four hours after the last stressor, mice were tested for stress-susceptibility in a 3-chamber social preference test (Morais-Silva et al., 2021). Mice were placed in an arena (60 × 40 cm, white walls and floor) divided into three chambers (20 × 40 cm) by perforated clear acrylic dividers. The two outer chambers contain wire mesh cups, while the central chamber is empty. The experimental mouse is placed in the central chamber of the arena with two empty wire mesh cups and allowed to explore for 5 min. Then the experimental mouse is allowed to explore the arena for an additional 5 min, this time with unfamiliar sex-matched adult conspecific in one of the wire mesh cups. The amount of time spent in the chamber containing the cups (empty or novel mouse) is measured with video tracking software (TopScan Lite, CleverSys, Reston, VA, United States), and used to determine stress-phenotype. Mice spending <170 s in the mouse-paired chamber were deemed "susceptible, " and >170 s were "resilient." This 170 s cutoff was chosen based on the average time unstressed mice spend in the mouse-paired chamber.

Homecage Fentanyl Administration
Following social interaction testing, mice were weighed then pair-housed across a perforated divider with a sex and stressphenotype matched conspecific in a woodchip bedding cage. Each mouse was provided two 50 mL conical tubes with rubber stoppers and ballpoint sipper tubes (Ancare, Bellmore, NY, United States). All tubes were weighed daily to determine liquid consumption, and the volume consumed was normalized to individual mouse weights.
On day 16, fentanyl solutions were replaced with water and mice were housed individually. Mice were then reassessed for social interaction behavior on day 18. Fentanyl preference was calculated as a percent of total liquid intake.

RNA Isolation
Four 14-gauge NAc tissue punches per mouse were collected 24 h after the last social interaction and stored at −80 • C until processing. RNA was extracted as described previously (Calarco et al., 2021) with TRIzol (Invitrogen; #15596018) and the EZNA MicroElute Total RNA kit (Omega Bio-Tek, Norcross, GA, United States; #R6831-01) with a DNase step (Qiagen, Germantown, MD, United States; #79254). RNA concentration and quality were determined with a NanoDrop 1000 spectrophotometer (Thermo Fisher Scientific). 400 ng of complementary DNA (cDNA) was synthesized using the reverse transcriptase iScript complementary DNA synthesis kit (Bio-Rad, Hercules, CA, United States; # 1708891), then diluted to a concentration of 2 ng/µL. Relative mRNA expression changes were measured by quantitative PCR using Perfecta SYBR Green FastMix (Quantabio, Beverly, MA, United States; #95072) with a Bio-Rad CFX384 qPCR system. Primer sets are available in Table 1. Fold change expression was calculated using the 2 −ddCt method with Gapdh as the reference gene. Data were normalized to the respective unstressed male or female controls.

Statistics
All statistical analysis was conducted in GraphPad Prism (version 9, San Diego, CA, United States) and JASP (Version 0.16) 1 . For CWDS mice, we used repeated-measures ANOVA with sex and stress-phenotype as between-subject factors, and epoch as within-subject factors. For CSDS mice, we used RMANOVA with stress-phenotype as between-subject and epoch as within-subject factors. When sphericity assumptions were violated we employed the Greenhouse-Geisser correction. Post hoc testing employed Holm's correction unless noted otherwise. For gene expression analysis, we used two-way ANOVA in CWDS mice, and unpaired t-test for CSDS mice. To compare social interaction between the forced/choice cohort and the choice cohort, we used threeway ANOVA using sex, stress-phenotype, and future-cohort as factors. We used Pearson's correlation to examine relationships between social interaction and fentanyl preference/consumption.

Stress-Susceptibility Is Predictive of Choice Fentanyl Consumption in Chronic Witness Defeat Stress Mice
Given that stress-susceptible mice consumed more fentanyl, we next sought to strengthen the relationship between stressphenotype and fentanyl preference. We performed Pearson's correlations using social interaction data from individual mice. In CWDS mice, we found a negative correlation between social interaction and mg/kg fentanyl consumed during Choice 1, indicating that the less time spent interacting with a novel conspecific (i.e., more stress-susceptible), the more fentanyl the CWDS mouse consumed during Choice 1 (Pearson's r = −0.382, p = 0.026; Figure 4A). Similarly, we found a trending negative correlation between stress-susceptibility and mg/kg fentanyl consumed during Choice 2 (Pearson's r = −0.270, p = 0.097; Figure 4A). By Choice 3, stress-susceptibility was no longer correlated with mg/kg fentanyl consumption in CWDS mice (p = 0.372, Figure 4A), however, it remained a predictor of overall choice intake (Pearson's r = −0.371, p = 0.019). Of note, there was no significant relationship between social interaction and mg/kg fentanyl consumption or % preference in unstressed mice (not shown, p = 0.18). By contrast, stresssusceptibility was not significantly correlated with mg/kg fentanyl consumed during any choice period in CSDS mice (Choice 1 p = 0.94, Choice 2, p = 0.287, Choice 3, p = 0.484 Figure 4B), however, there was a trend toward stress-susceptibility increasing fentanyl consumption during forced epoch 1 (r = −0.383, p = 0.07, Figure 4B). The strongest relationship between choice consumption in CSDS mice was forced consumption, indicating physical dependence may play more of a role in CSDS mice (e.g., total choice and forced epoch 2, r = 0.60, p ≤ 0.001, Figure 4B).

Abstinence From Homecage Fentanyl Increases Stress-Susceptibility Dependent on Stress-Phenotype
Opioid abstinence and withdrawal can promote anhedonia and social interaction deficits (Bai et al., 2014;Becker et al., 2017;Bravo et al., 2020). To test how our stress and fentanyl paradigm influenced social interaction, we next replaced the fentanyl solutions with plain water for 2 days and single housed mice prior to reassessing social interaction behavior. We chose this ∼56 hr time point to avoid acute withdrawal symptoms during testing. We found significant effects of stress-phenotype (RMANOVA, F 2 , 57 = 8.4, p < 0.001), time (F 1 , 57 = 11.4, p = 0.001), and a time × stress-phenotype interaction (F 2 , 57 = 14.7, p < 0.0001; Figure 5A), but not sex (p = 0.1). All unstressed and CWDSresilient groups showed reduced time interacting with a novel sex-matched conspecific relative to the first social-interaction (before vs. after fentanyl: unstressed p ≤ 0.001, CWDS-resilient p = 0.036, Figure 5A). CWDS-susceptible mice showed a nominal increase in social interaction but remained susceptible (p = 0.14). CSDS mice also reduced time interacting with a novel conspecific relative to the first social interaction, but this was only statistically significant in CSDS-resilient mice, possibly indicating a "floor effect" in CSDS-susceptible mice (RMANOVA, time: F 1 , 32 = 10.4, p = 0.003; stress-phenotype: F 2 , 32 = 12.90, p < 0.001; time × stress-phenotype: F 2 , 32 = 8.9, p ≤ 0.001; before vs. after fentanyl: CSDS-susceptible p = 0.87; CSDS-resilient p = 0.002, Figure 5B). Together, this indicates fentanyl exposure and abstinence alone is sufficient to generate a stress-like phenotype.

Female Mice Exhibit Greater Fentanyl Preference in an All-Choice Paradigm
An important caveat in this work is that by forcing mice to consume fentanyl, some mice may have increased fentanyl preference due to tolerance or may consume fentanyl to mitigate withdrawal symptoms. We thus repeated the experiment in a separate cohort of mice that were given a choice between fentanyl and water for the entire experiment ("choice cohort") As with the mice that received forced exposure ("forced/choice cohort"), we divided choice cohort mice into susceptible and resilient as described above (For CWDS, two-way ANOVA, stress-phenotype, F 2 , 50 = 9.36, p = 0.0004; For CSDS, Welch's ANOVA, W 2 , 17 . 43 = 30.67, p < 0.0001; Table 2). Prior to the fentanyl exposure, we had a different number of mice classified as susceptible in the choice cohort as compared with the forced/choice cohort, however, there was no effect of futurecohort (three-way ANOVA, F 1 , 107 = 0.88 p = 0.3516) or futurecohort × stress-phenotype (F 2 , 107 = 2.10, p = 0.13) in CWDS mice, indicating the cohorts were equivalent. However, for CSDS mice, there was a future-cohort × stress-phenotype interaction (two-way ANOVA, F 2 , 71 = 6.02, p = 0.0038) such that CSDSsusceptible mice in the choice cohort had greater mean social interaction time compared with mice in the forced/choice cohort (Holm-Sidak post hoc, p = 0.004). We include the data from this group in the present manuscript, but it is important to note this difference when comparing the two cohorts.

DISCUSSION
Here we show both physical (CSDS) and vicarious (CWDS) psychosocial stressors impact homecage opioid consumption and preference in a stress and sex-dependent manner. First, susceptibility to either CWDS or CSDS increases forced fentanyl consumption in male mice regardless of fentanyl concentration. Second, female CWDS-susceptible mice show increased fentanyl consumption during the choice periods, and greater stress-susceptibility is associated with greater fentanyl preference. However, only female CWDS-susceptible mice show increased fentanyl preference during the final choice period. Physical CSDS-susceptibility is associated with increased choice fentanyl consumption, but this effect is decoupled from social interaction behavior and more closely associated with prior forced consumption. Finally, we show   CWDS-resilient 6.1 ± 0.8 6 Females consume more relative to males in non-forced epoch 2 (p = 0.038) and non-forced epoch 3 (p < 0.001).

CWDS and CSDS-susceptible males have downregulated Rho
GTPases in the NAc.
The main goal of this study was to determine if psychosocial stress in mice produced similar effects to restraint stress in rats on fentanyl preference using the alternating periods of forced and choice consumption described by Shaham et al. (1992). Consistent with this prior work, stressed mice consumed more fentanyl than unstressed mice. In addition to physical stress, vicarious stress was sufficient to increase drug-taking, similar Female mice consume more than male mice during choice 2 (p < 0.001) and choice 3 (p = 0.005).
to Ramsey and Van Ree (1993) and Cooper et al. (2017). We also found a negative correlation between social interaction and fentanyl preference, similar to the morphine preference findings of Cooper et al. (2017). Importantly, our mice were pair-housed throughout the fentanyl procedure. Previously, we showed that when mice are single-housed, CSDS-susceptibility is associated with increased early cocaine self-administration (Engeln et al., 2021). When pair-housed, CSDS instead decreases cocaine self-administration, analogous to increased anhedonia. Consistent with our cocaine work, Alexander et al. (1978) showed socially isolated rats consume more morphine relative to socially housed rats. Thus, the small effects we report here may reflect a kind of "social buffering" that decreases opioid consumption, similar to the protective effects of social interaction on heroin and methamphetamine craving (Venniro et al., 2018(Venniro et al., , 2019. Future work should investigate how stress interacts with social housing conditions to sensitize or blunt opioid consumption. Here we found only stress-susceptible female mice increase fentanyl preference over time. In humans, sex-differences in OUD are largely dependent on the opioid (Back et al., 2010;Lee and Ho, 2013). Recent evidence suggests that women have higher rates of prescription opioid use (Serdarevic et al., 2017) and are more likely to report using opioids to cope with negative affect relative to men (McHugh et al., 2013). Women also exhibit increased susceptibility to stress-related disorders (e.g., PTSD, depression) (Bangasser et al., 2019). While we do not see striking sex-differences in our behavioral readout of susceptibility in mice, the increased opioid consumption in our female mice mirrors that of humans and other rodents models (Alexander et al., 1978;Lynch and Carroll, 1999;Kokane and Perrotti, 2020;Hammerslag et al., 2021). In our study, female mice reliably consume more fentanyl than male mice, even when provided free choice throughout. This finding aligns well with recent work showing female mice are resistant to devaluation of oral fentanyl (Monroe and Radke, 2021), self-administer more oral oxycodone (Phillips et al., 2020), intravenous heroin (Towers et al., 2019), and remifentanil (Anderson et al., 2021). Further work is needed to establish if stress exacerbates the sex-differences in operant opioid selfadministration.
When we compared male mice from our modified forced/choice model (Shaham et al., 1992) to male mice that had free-choice throughout the entire procedure, stressed males do not appear different from unstressed males. It is tempting to speculate that in the forced/choice cohort, stressed male mice exhibit increased consumption due to either tolerance that develops from the forced epoch, and/or a desire to relieve the negative affect caused by abstinence/withdrawal (Koob, 2020). Our correlation data support this interpretation, as the strongest predictor of choice intake was forced intake in CSDS mice. It is impossible to rule out that the physically stressed CSDS mice consume fentanyl to relieve pain. However, we do not believe this to be the case given the modest differences between intake in choice cohort unstressed and CSDS mice.
Social withdrawal and other anhedonia-like behaviors are commonly exhibited following opioid abstinence and withdrawal (Bai et al., 2014;Becker et al., 2017;Bravo et al., 2020). In the forced/choice cohort, all unstressed and previously resilient mice exhibited decreased social interaction; in the choice cohort, only previously resilient mice decreased social interaction. Given that choice cohort unstressed mice consumed less fentanyl as compared with forced/choice unstressed mice, this suggests the social-withdrawal is dose-dependent. Opioid abstinence in previously resilient choice cohort mice-despite the overall dose being smallermay have been sufficient to tip them toward susceptibility. Regardless of forced/choice or choice, the CSDS-susceptible mice maintain their previous social interaction behavior, which we believe may reflect a "floor effect." Future work examining severity of withdrawal symptoms or other stresslike behaviors could address if susceptible mice exposed to fentanyl are more or less stressed than their resilient drug-exposed counterparts.
Targeting dendritic remodeling may prove useful for blunting stress-sensitized acquisition of drug taking (Rigoni et al., 2021). Given that NAc structural plasticity is an important determinant of synaptic strength, and heavily involved in both stress and drug-related behaviors, we examined Rho GTPase expression in the NAc. Interestingly, while female mice displayed the more persistent elevation in fentanyl preference after stress, we only found differential expression in stressed male mice. Furthermore, the changes to gene expression were consistent in susceptible mice regardless of physical or vicarious stress. We found decreased expression of Rac1, RhoA, and Cdc42 in NAc of stress and force/choice male mice. This is largely consistent with decreased Rac1 expression in NAc after CSDS (Golden et al., 2013), and decreased active Cdc42 in mPFC after chronic unpredictable stress (Luo et al., 2020). Both acute morphine withdrawal (Cahill et al., 2018) and CSDS (Francis et al., 2017;Fox et al., 2020a) engage NAc RhoA signaling. Thus, the downregulated RhoA expression seen here may reflect a compensatory mechanism in males that help protects against elevated late fentanyl preference. In agreement with compensatory downregulation, Rock1, a downstream effector of RhoA, is decreased in striatum after protracted morphine withdrawal (Drastichova et al., 2021). However, our data are difficult to interpret given the competing effects of fentanyl and stress. It is also possible that the sex and stressor-dependent changes in gene expression are related to sex differences in fentanyl consumption and stress responsivity. It is also worth noting that RhoA expression and dendritic remodeling after CSDS are cell subtype specific (Fox et al., 2020a,b), and there is no work on NAc GTPases after CWDS. Future work can dissect precise sex and cell-type specific adaptations that may explain the effects of CSDS/CWDS on subsequent fentanyl consumption.
In summary, we show stress-susceptibility is associated with increased fentanyl consumption that presents differently in male and female mice. We also found sex differences in the expression of dendritic-complexity molecules in the nucleus accumbens-a region important for both stress and drug-related behaviors. Our findings here, together with our updated mouse model, will provide the basis for more precise investigations on mechanisms of stress-sensitized opioid use in both sexes of transgenic mice.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author/s.

ETHICS STATEMENT
The animal study was reviewed and approved by the University of Maryland School of Medicine IACUC.

AUTHOR CONTRIBUTIONS
MF designed the study. DF, AW, and MF conducted the experiments and analyzed the data. ML provided resources for conducting the experiments. MF and DF wrote the manuscript with contributions from AW and ML. All authors contributed to the article and approved the submitted version.