Alcohol inhibits sociability via serotonin inputs to the nucleus accumbens

Social interaction is a core component of motivational behavior that is perturbed across multiple neuropsychiatric disorders, including alcohol use disorder (AUD). Positive social bonds are neuroprotective and enhance recovery from stress, so reduced social interaction in AUD may delay recovery and lead to alcohol relapse. We report that chronic intermittent ethanol (CIE) induces social avoidance in a sex-dependent manner and is associated with hyperactivity of serotonin (5-HT) neurons in the dorsal raphe nucleus (DRN). While 5-HTDRN neurons are generally thought to enhance social behavior, recent evidence suggests that specific 5-HT pathways can be aversive. Using chemogenetic iDISCO, the nucleus accumbens (NAcc) was identified as one of 5 regions that were activated by 5-HT DRN stimulation. We then employed an array of molecular genetic tools in transgenic mice to show that 5-HT DRN inputs to NAcc dynorphin neurons drive social avoidance in male mice after CIE by activating 5-HT2C receptors. NAcc dynorphin neurons also inhibit dopamine release during social interaction, reducing the motivational drive to engage with social partners. This study reveals that excessive serotonergic drive after chronic alcohol can promote social aversion by inhibiting accumbal dopamine release. Drugs that boost brain serotonin levels may be contraindicated for individuals with AUD.

weeks, which resulted in an average ethanol intake and preference of 13.09 ± 2.36 g/kg/24-h and 46.8 ± 8.9%, respectively, for males and 26.12 ± 4 g/kg/24-h and 57± 6% for females at the 8-week mark (Extended Data Figure 1A -B). Despite lower ethanol consumption and preference overall, male mice exhibited social de cits in the social interaction test (Bonferroni t 36 =7.33, p<0.0001) after CIE whereas females did not. Females also spent less time in social interaction relative to males (Bonferroni t 36 =3.41, p<0.01) (Extended Data Figure 1C-E). No anxiety-like behaviors or locomotor de cits were observed in the elevated plus maze (Extended Data Figure 1F-I). In the sucrose preference test, there was no 3-way CIE x sex x time interaction, but there was a two-way CIE x sex interaction (F 1,36 =4.44, p<0.05) as well as a main effect of sex (F 1,36 =19.82, p<0.0.0001) and time (F 3.81, 136.4 =2.72, p<0.05), but not CIE (Extended Data Figure 1J). There was also no change in time spent in the center of an open eld in either sex, but males exhibited hypoactivity in the open eld which may be indicative of anxiety (Bonferroni t 36 =3.20, p<0.05) (Extended Data Figure 1K-O). Since the ethanol-induced social de cits were limited to males, we used male mice throughout the remainder of the study.
Mapping 5-HT DRN -stimulated brain-wide activation patterns with chemogenetic iDISCO Since 5-HT DRN neurons have been linked to social behavior 13 , we rst asked whether CIE could alter the excitable properties of these neurons. Ex vivo electrophysiological recordings revealed that 5-HT neurons in CIE mice had lower mean rheobase (i.e. action potential threshold) (t 15 =2.73, p<0.05) and increased current-induced spiking which is indicative of enhanced excitability (Interaction F 20,280 =3.57, p<0.0001; Area under the curve: t 14 =4.02, p<0.01) (Extended Data Figure 2). We then paired chemogenetic stimulation of 5-HT DRN neurons with whole brain c-fos immunolabeling and volumetric imaging (iDISCO) 14 to map neuronal activity in downstream targets. Sert-cre mice were stereotaxically injected with a Cre-inducible Gq-coupled DREADD (hM3Dq) in the DRN, and their brains were harvested for iDISCO 3 hours after CNO injection. Volumetric images were aligned to the Allen Brain Atlas and c-fosimmunoreactive (IR) cells were counted using the ClearMap spot detection algorithm. C-fos-IR neurons were elevated in the central lateral nucleus of the thalamus (CL thalamus), nucleus accumbens (NAcc), ventral anterior cingulate cortex (ACC, layer 1), primary auditory cortex (layer 6a), and bed nucleus of the stria terminalis (BNST) (Figure 1). The NAcc has been previously implicated in social behavior in the context of autism spectrum disorder and morphine withdrawal 15,16 , so we focused on the 5-HT projections to the NAcc throughout the remainder of the study.
Neuronal activity in NAcc-projecting 5-HT neurons drives social de cits after CIE We next targeted Gq-coupled DREADDs to NAcc-projecting 5-HT neurons by injecting ethanol-naïve Sertcre mice with the retrograde virus AAVrg-hsyn-DIO-hM3Dq-mCherry or a control virus (AAVrg-hsyn-DIO-mCherry) in the NAcc. Sert-cre::hM3Dq 5-HTàNAcc mice spent signi cantly less time in social interaction relative to controls after CNO (F 1,17 =21.73, p<0.001; Bonferroni t 34 =5.26, p<0.0001) (Figure 2A-C). There were no group differences in anxiety-like behavior in the open eld, although there was a trend toward a reduction in locomotor activity, which may be indicative of increased anxiety (Extended Data Figure 3).

5-HT DRNàNAcc neurons have collateral projections to other cortical and subcortical regions
Our data suggest that inhibiting 5-HT DRNàNAcc neurons rescued social de cits after CIE, but these neurons may have collateral projections to other regions that drive social behaviors. In order to address this question, we injected Tph2-p2A-pO mice with AAVrg-fDIO-Cre in the NAcc and AAV8.2-hef1α-DIOsynaptophysin-eYFP in the DRN to label axonal processes in downstream regions that may represent collateral projections of 5-HT DRNàNAcc neurons. We were particularly interested in regions where we had observed elevated c-fos expression after chemogenetic stimulation of 5-HT DRN neurons, including the central lateral nucleus of the thalamus (CL thalamus), NAcc, ACC (layer 1), primary auditory cortex (layer 6a) and BNST. Of those regions, we detected eYFP-labeled processes in the NAcc, BNST, and CL thalamus. The BNST is of particular interest because there is elevated 5-HT 2C receptor expression after CIE 11 , although a recent study found that BNST 5-HT 2C receptors did not mediate behavioral dysregulation following binge alcohol consumption 18 . We also examined regions outside of this network and found eYFP-labeled axons in the ventral pallidum, ventral tegmental area (VTA), hypothalamus, amygdala, insular cortex, entorhinal cortex, and medial prefrontal cortex. (Extended Data Figure 5).
However, since these regions did not exhibit altered c-fos immunoreactivity following chemogenetic stimulation of 5-HT DRN neurons, it is unclear whether they are involved in behavioral de cits following CIE.

Serotonin release is enhanced in the NAcc during social interaction
To further examine the role of the 5-HT DRNàNAcc pathway in social de cits after CIE, we next asked whether 5-HT is released in the NAcc during social interaction in ethanol-naïve mice. C57BL/6J mice were stereotaxically injected with a genetically encoded 5-HT biosensor (AAV9-hsyn-5-HT3.5 or GRAB  and implanted with optical bers in the NAcc for ber photometry recordings. We observed a signi cant increase in area under the curve (AUC) in the peri-event plot during the rst social approach relative to empty cage investigation (t 7.82 =3.45, p<0.01 with Welch's correction), suggesting that enhanced 5-HT release in the NAcc coincides with the onset of social interaction. Enhanced activity in 5-HT DRNàNAcc neurons after CIE may increase basal 5-HT activity in the NAcc and ultimately dysregulate phasic 5-HT release during social interaction, resulting in social de cits (Extended Data Figure 6).

5-HT neurons in the DRN form synaptic connections with NAcc dynorphin neurons and increase neural activity via 5-HT 2C receptors
We next injected Sert-cre mice with AAV8.2-ef1α-DIO-synaptophysin-eYFP and found that serotonergic bers originating in the DRN were robustly expressed in the ventral NAcc in close proximity to dynorphin neurons ( Figure 3A -B). We also observed synaptic junctions between eYFP-labeled presynaptic terminals and dynorphin A-labeled post-synaptic terminals in the ventral NAcc using transmission electron microscopy (TEM) ( Figure 3C). To con rm that ventral NAcc dynorphin neurons receive direct synaptic input from 5-HT DRN neurons, we injected Sert-cre mice with the anterograde transsynaptic tracer AAV9-CAG-WGA-ZsGreen in the DRN 19 . This resulted in robust labeling of dynorphin neurons in the ventral NAcc ( Figure 3D-F).
A previous study reported that 5-HT 2C receptors were enriched in dynorphin-positive cell clusters in the striatum 20 , so we asked whether CIE upregulated 5-HT 2C receptors within NAcc dynorphin neurons.
Fluorescence in situ hybridization (FISH) analysis revealed that CIE increased the percentage of neurons that were double-positive for prodynorphin and 5-HT 2C ( Figure 4A-C). In contrast, other 5-HT receptors including the 5-HT 1B or 5-HT 2A were not signi cantly modi ed after CIE (Extended Data Figure 7). Striatal 5-HT 2C receptors have been previously implicated in anxiety and dysphoria and may promote escalated drinking in ethanol-dependent mice 21 , but a direct link to human AUD has not yet been established. We then asked whether 5-HT 2C receptors colocalize with dynorphin neurons in human NAcc sections obtained from the University of Iowa Neurobank (demographic details in Table 1). Overall, 23.25 ± 2.2% of NAcc neurons were 5-HT 2C -positive, 13.92 ± 2.2% were prodynorphin-positive, and 10.61 ± 1.71% were double positive for 5-HT 2C and prodynorphin ( Figure 4D-F). These data suggest that the distribution of 5-HT 2C receptors in the human NAcc is similar to the mouse, and these receptors may play similar roles in behavioral regulation.
Given that the 5-HT 2C receptor is a G-protein-coupled receptor (GPCR) that activates G q/11 signaling and promotes membrane depolarization, we expect this upregulation of 5-HT 2C receptors to enhance excitability of NAcc dynorphin neurons. As such, we crossed Pdyn-cre mice to L10AeGFP ox/ ox mice to use as reporters for ex vivo slice electrophysiology experiments ( Figure 4G-H). Dynorphin neurons from CIE mice had lower action potential thresholds and enhanced current-induced spiking relative to controls, and this effect was abolished by pre-application of the 5-HT 2C receptor antagonist RS102221 (Rheobase: F 1,9 =10.78, p<0.01, Bonferroni t 21 =2.79, p<0.05; Current-induced spiking: F 20,420 = 2.80, p<0.0001). We also calculated the area under the curve (AUC) of the action potential frequency vs. applied current plot and found AUC was increased in CIE mice but normalized by RS102221 pretreatment (F 1,21 =10.44, p<0.01, Bonferroni t 21 =4.86, p<0.001) ( Figure 4I-K). There were no group differences in spontaneous synaptic transmission (inhibitory or excitatory) or synaptic drive ( Figure 4L-N), suggesting that ethanol exposure primarily affects the intrinsic excitability of NAcc dynorphin neurons in a 5-HT 2C receptordependent manner.

CIE promotes social de cits via 5-HT 2C receptors and prodynorphin in the NAcc
We next asked whether genetic ablation of 5-HT 2C receptors in NAcc dynorphin neurons could mitigate social de cits in CIE mice. Pdyn-cre mice were stereotaxically injected with AAV8-Sico-ef1α-eGFP-Htr2c[shRNA#1] in the NAcc, which expressed eGFP in the absence of Cre and a short-hairpin RNA (shRNA) directed against 5-HT 2C receptors in the presence of Cre recombinase. A control scrambled shRNA was also injected into the NAcc of water and CIE mice (shRNA sequence details in Table 2). This Chemogenetic inhibition of these NAcc dynorphin neurons produced similar effects (Interaction:  Figure 5D-F), suggesting that 5-HT may articulate its aversive effects on social behavior by activating NAcc dynorphin neurons. We then wanted to con rm that dynorphin peptide was mediating these effects since dynorphin-expressing neurons in other brain regions coexpress other neuropeptides that are abundant in the striatum including corticotropin-releasing factor (CRF) 22-25 . To do this, we knocked down dynorphin expression in the NAcc with shRNA directed against prodynorphin. AAV8-CMV-shPDYN or scrambled shRNA was stereotaxically injected into the NAcc of C57BL/6J mice exposed to CIE (shRNA sequence details in Table 2). We found that pdyn knockdown in the NAcc normalized time spent in social interaction in CIE mice (Interaction F 1,16 =5.51, p<0.05; Bonferroni post-tests: Water-shControl vs. CIE-shControl, t 16 =3.88, p<0.01; Water-shPDYN vs. CIE-shPDYN, t 16 =0.56, ns) ( Figure 5G-J), con rming that dynorphin peptide mediates social de cits after CIE.
Chemogenetic inhibition of Pdyn NAcc neurons abolishes social de cits induced by 5-HT DRN neuronal stimulation.
Pdyn NAcc neurons form local circuits and inhibit dopamine release during social interaction.
Dynorphin has been previously shown to inhibit dopamine release in the NAcc via presynaptic kappa opioid receptors 26 , but the precise origin of these dynorphinergic inputs is unknown. In another study, optogenetic stimulation of dynorphin in the ventral NAcc induced aversive behavior, but it was suggested that this may be mediated by long-range projections to VTA dopamine neurons 27 . Dynorphin neurons in the NAcc may also inhibit local dopamine release, which would provide a mechanistic framework for understanding their role in social avoidance during ethanol withdrawal. First, we asked whether NAcc dynorphin neurons form local synaptic connections in the NAcc itself by injecting Pdyn-cre mice with the anterograde transsynaptic tracer AAV9-CAG-DIO-WGA-ZsGreen to map their post-synaptic targets. A sizable population of non-dynorphin neurons labeled with ZsGreen was detected in the NAcc, suggesting the existence of a local NAcc dynorphinergic circuit ( Figure 6A-D). Surprisingly, we did not detect any ZsGreen-labeled cells in the vicinity of dopamine neurons in the VTA, so Pdyn NAcc inputs to VTA dopamine neurons may not be as robust as previously thought ( Figure 6E).
We then asked whether chemogenetic stimulation of pdyn NAcc neurons could inhibit local dopamine release in the NAcc in mice actively engaged in social interaction. Pdyn-cre mice were co-injected with AAV8-hsyn-DIO-hM3Dq-mcherry and AAV1-CAG-dlight1.1 in the NAcc so that we could simultaneously activate local dynorphin neurons with CNO and detect dopamine release with ber photometry ( Figure 7A-C). The frequency of dopamine transients increased when the animal was actively engaged in social interaction (SI) relative to exploration of empty cage (EC) or when exploring other aspects of the environment (F 3,63 =7.56, p<0.001; Bonferroni post-tests: SI vs EC: t 63 =3.88, p<0.01; SI vs non-EC: t 63 =4.04, p<0.001; SI vs non-SI: t 63 =3.68, p<0.01), but the peak amplitude was unaffected ( Figure 7D-F). In a separate session, the mice received intraperitoneal injections of CNO to activate NAcc dynorphin neurons while the mice were engaged in social interaction. Here we found that chemogenetic activation of NAcc dynorphin neurons reduced the peak amplitude of dopamine transients during social interaction (Mann-Whitney U=4, p<0.01) ( Figure 7G) but not during exposure to the empty cage (Mann-Whitney U=16, ns) ( Figure 7I). On the other hand, there was no effect of chemogenetic manipulation on the frequency of dopamine transients during social interaction (t 13 =0.87, ns) or exploration of the empty cage (t 12 =0.05, ns). The increase in frequency of dopamine transients during social interaction may re ect an increase in burst ring of dopamine neurons in the VTA that project to the NAcc 28 , which is not affected by local dynorphin activity. However, the ability of NAcc dynorphin neurons to reduce the amplitude of these dopamine transients suggests that dynorphin may be acting locally to inhibit release from dopamine terminals. This is consistent with previous microdialysis studies indicating that kappa opioid receptor agonists into the NAcc can reduce extracellular dopamine levels 26 .

Discussion
Selective serotonin reuptake inhibitors (SSRIs) are commonly prescribed drugs for treating depression and anxiety which frequently present in individuals with AUD 29 . While these drugs are generally thought to mitigate depressive symptoms in these individuals, they can also increase the risk of alcohol relapse during periods of abstinence [30][31][32] . Although the neural mechanism behind this observation is not abundantly clear, there is some consensus around the idea that alcohol exposure increases serotonergic activity in the DRN and alters neural responses to serotonin release in downstream circuits 6,8,11,18,21,[33][34][35][36] . Here we demonstrate that ethanol-induced social de cits in males are orchestrated by DRN serotonergic neurons that project to the NAcc and activate 5-HT 2C receptors, which are upregulated in dynorphin neurons following CIE. These Gq-coupled 5-HT 2C receptors drive excitatory, postsynaptic signaling and potentiates dynorphin neuronal activity, which in turn inhibits dopamine release during social interaction. Thus serotonin, acting via the 5-HT 2C -dynorphin circuit, has an inhibitory effect on dopamine release in the NAcc, which may reduce the motivation to engage in social interaction. Interestingly, both serotonin and dopamine release increased during social interaction in ethanol-naïve animals. This suggests that serotonin release during social interaction may act as a molecular brake on dopamine release that eventually turns off this signal via the 5-HT 2C receptor as the social encounter becomes less novel and more familiar. After chronic ethanol, however, this 5-HT DRN àpdyn NAcc circuit goes into overdrive, which impairs dopamine release in the NAcc and degrades the reward value of social interaction.
5-HT DRNàNAcc neurons also have collateral projections to other regions that could potentially drive social behaviors, including the BNST which is enriched in 5-HT 2C receptors that are upregulated in mice following CIE 11 . However, our RNAi data suggests that CIE-induced social de cits are primarily driven by 5-HT 2C receptors in NAcc dynorphin neurons. Interestingly, it has also been reported that 5-HT inputs to the BNST promote fear and anxiety-like behaviors in ethanol-naïve mice 37 , so elevated 5-HT 2C signaling in the BNST could mediate some other behavioral endpoints associated with excessive alcohol use.
The idea that excessive serotonergic drive in the NAcc inhibits social behavior seems to contradict some of the previous literature in which 5-HT inputs to the NAcc alleviate social de cits in mouse models of autism 15 . However, it is important to note that 5-HT can have divergent effects depending on the distribution and density of speci c 5-HT receptor subtypes, which can be pre-or post-synaptic and activate different G protein-coupled signaling pathways. In 16p11.2 mice, the socially rewarding effects of 5-HT DRN stimulation were mediated by 5-HT 1B receptors, which are presynaptic Gi-coupled receptors that inhibit release of other neurotransmitters. In the context of CIE, there is upregulation of 5-HT 2C receptors in NAcc dynorphin neurons which tends to increase dynorphinergic output and cause a net inhibitory effect on phasic dopamine release during social behavior. While 5-HT signaling in the NAcc may promote social reward when 5-HT 1B receptors are abundant, upregulation of 5-HT 2C receptors after chronic alcohol may convert 5-HT to an aversive signal.
In total, these studies demonstrate that 5-HT neurons that project to the NAcc can modulate social behavior by activating 5-HT 2C receptors in dynorphin neurons. Chronic alcohol engages this 5-HT DRN àDyn NAcc circuit by enhancing activity in 5-HT DRN neurons and increasing 5-HT 2C receptor signaling in NAcc dynorphin neurons, which ultimately suppresses dopamine release during social interaction and promotes social de cits ( Figure 8). Overall, the results of this study strongly suggest that SSRIs should be used with caution in individuals with AUD, especially in those that are recently abstinent from alcohol as these individuals may experience more negative side effects from taking antidepressants that could ultimately lead to relapse. Future studies would be warranted to determine whether SSRIs could be more effective in individuals with AUD when combined with short-term use of a kappa opioid or 5-HT 2C receptor antagonist.

Declarations Declaration of Competing Interest
The authors declare that they have no known competing nancial interests or personal relationships that could have in uenced or appeared to in uence this work.

Funding
This work was funded through NIH grants R00 AA024215 and R01 AA028931 as well as BBRF grant #27530 to C.A.M. K.M.K. was supported by T32 NS045549, T.J. was supported by T32 HL007638, and S.P. was supported by T32 GM067795.

Animals
In these experiments, we used adult male and female C57BL/6J mice from Jackson Labs (Bar Harbor, ME, USA), as well as transgenic Sert-cre (Jackson Labs, 014554), Pdyn-ires-cre (Jackson Labs, 027958), and EGFP-L10a mice (Jackson Labs, 024750) which were bred in-house. Tph2-p2A-pO mice were developed at the University of Iowa Genome Editing Core Facility. All mice were maintained at a temperature-and humidity-controlled AAALAC approved vivarium at the University of Iowa College of Medicine with ad libitum access to food and water and 24 hour on-call veterinary care.
Generation of Tph2-p2A-pO mice To be able to target subpopulations of 5-HT neurons, we generated a Tph2-p2A-pO mouse line that expresses pO recombinase from the endogenous Tph2 locus 38 . CRISPR-Cas9 endonuclease-mediated genome editing was used to insert a p2A-pO-recombinase sequence 3' in coding exon 11 of the Tph2 locus ( Figure 2I). The self-cleaving p2A sequence allows for expression of the endogenous Tph2 protein and pO recombinase. B6SJLF1/J mice (Jackson Labs, 100012) were used in the production of Tph2-p2A-pO mice. To validate this mouse line, we performed uorescence in situ hybridization (FISH) and immuno uorescence assays (Extended Data Figure 4). In FISH, RNAscope probes Mm-Tph2-C2 and Flp from ACDBio (Newark, CA, USA; Table 3) were used to assess colocalization of Tph2 with pO RNA in the DRN. In immunohistochemistry, we rst stereotaxically injected Tph2-p2A-pO mice with AAV5-ef1α-fDIO-mCherry in the DRN, and 4 weeks later, we examined colocalization of Tph2 with mCherry (see antibody details in Table 4).

Chemogenetic experiments
Clozapine-N-oxide dihydrochloride (CNO; Hello Bio, Princeton, NJ, USA) was dissolved in sterile, 0.9% saline to a concentration of 0.30 mg/ml and injected at 10 ml/kg, i.p., 30 min before the beginning of behavioral experiments for a dose of 3 mg/kg. For c-fos studies, CNO was injected 2 hours before perfusion and brain extraction. Deschloroclozapine dihydrochloride (DCZ; Hello Bio) was dissolved in sterile, 0.9% saline to a concentration of 10 µg/ml and injected at 10 ml/kg, i.p., 20 min before the beginning of behavioral experiments for a nal dose of 100 µg/kg, which was previously found to induce neural activity in hM3Dq-expressing neurons in mice and monkeys 17 .  Table 5 for details about viral constructs.

Stereotaxic surgeries
The mouse was placed in an Angle Two stereotaxic frame (Leica Biosystems, Wetzlar, Germany) and maintained on a heated platform under 5% (v/v) iso urane anesthesia as described previously 37

Chronic Intermittent access to ethanol (CIE)
Mice were singly housed starting at 7-8 weeks of age in specialty cages with 2 holes at the front of the cage designed to t two water bottles containing either tap water or 20% w/v ethanol. Water control mice were given tap water only. Water and ethanol bottles were given between 9:00-10:00 AM on MWF and replaced with two water bottles 24 hours later. The con guration of the bottles (left v right) was switched between drinking sessions to account for side preferences. Two drip bottles for water and ethanol attached to an empty cage were included to measure loss via drip. Water and ethanol consumption over each 24-hour drinking session was calculated as follows: where ΔW i is the weight change of the water or ethanol bottle in grams, ΔW drip is the weight change of the drip bottle, and BW is the body weight of the mouse in kg.

Behavior
Social interaction test: The social interaction test was conducted 24 hours after the last ethanol drinking session and 30 min after injection with CNO (in chemogenetic experiments only), which has been described in detail before 11,12,34 . Brie y, the test was conducted in a three-chambered Plexiglas container (L × W × H: 64 × 42 × 21 cm) divided by 2 panels with a small opening at the base of each to allow movement between the chambers 39 . Behavior was recorded by an overhead camera with Media Recorder software (Noldus, Leesburg, VA, USA). At the start of each testing session, the test mouse was placed in the center chamber and allowed free movement/exploration throughout the 3 compartments for 10 min.
The mouse was then placed back in the center chamber with the doors closed and a holding cage (i.e., a wire mesh pen cup, 11 cm high, open end: 10 cm in diameter; closed end: 8 cm in diameter) was placed in the center of each of the two outermost chambers. A novel conspeci c of the same age and sex was placed inside the holding cage on one side while the other cage remained empty. The doors were then reopened and the test mouse was allowed to roam freely between chambers for 10 more min. The time spent interacting with the stranger mouse and investigating the empty cage was scored by a trained observer blinded to the experimental condition. The percentage of time spent in social interaction or investigating the empty cage during the 10-min session was then calculated for each mouse. The placement of the stranger was alternated between trials to control for side preferences.
Elevated plus maze (EPM) test: The Plexiglas elevated plus maze had two open arms (35 x 5 cm each) and two closed arms (35 x 5 cm each, with 20-cm-tall dark side walls), interconnected via a center zone (5 x 5 cm) and elevated by 60 cm from the oor. The test was conducted under dim light (~20 lux in the open arms and ~5 lux in the closed arms), starting when the mouse was being placed in the center zone.
The mouse was allowed to freely move and explore in the maze for 5 min, and their activity was recorded with a camera mounted above the maze. cm), situated in a custom-made sound-attenuating wood box. The arena was placed under dim light (~20 lux in the center of the arena), with a camera mounted above. The mouse was placed in a corner of the arena and allowed to freely move and explore for 30 min. All the activity was recorded and analyzed with Ethovision XT14 software, by comparing the following parameters: latency to the center, center entries, and total distance moved. The center was de ned as the central 15% of the arena.
Sucrose preference test: Mice were placed in a PhenoTyper 3000 observation cage (Noldus), consisting of a square chamber (30 x 30 cm) with clear Plexiglas walls, an overhead camera, a feeder with ad libitum access to chow, and 2 water bottles on the wall opposite to the feeder, which were both equipped with lickometers that automatically detect contacts made between the mouse and the water spouts. The data from the lickometers and the overhead camera were processed with Ethovision XT14 software. The mouse was allowed to freely explore the chamber for 1 hour on each of 4 training days, with ad libitum access to tap water and 5% sucrose solution from the 2 bottles. The placements of the water and sucrose bottles were switched every day to avoid place preferences. The nal test lasted for 20 min on the 5 th day.
Sucrose preference was calculated as follows: After this 5-day sucrose preference test, the mice were exposed to another cycle of CIE. At 24 hours after the last ethanol exposure (ethanol withdrawal; EW), they were given ad libitum access to 5% sucrose and water for 20 min.

Whole-cell patch clamp electrophysiology
Procedures for whole-cell patch clamp electrophysiology have been described in detail elsewhere 11,37,40,41 PO 4 ,10.0 glucose, and 26.0 NaHCO 3 , maintained at 30 ± 1 °C for at least 1 h before recording. The slices were then transferred to a recording chamber containing normal, oxygenated aCSF, maintained at 28-30 °C, delivered at a ow rate of 2 ml/min.
Neurons were visualized using infrared differential interference contrast (DIC), video-enhanced microscopy (Olympus, Tokyo, Japan) and their genetic identity as dynorphin or 5-HT neurons was con rmed using GFP and RFP lter cube sets (Olympus). Slices were perfused in normal aCSF in the recording chamber for at least 30 min before whole-cell patch clamp recordings were performed. Borosilicate electrodes were pulled with a Flaming-Brown micropipette puller (Sutter Instruments, Novato, CA, USA) and only those with a pipette resistance between 3 and 6 MΩ were used to patch cells. These Signals were acquired using a MultiClamp 700B ampli er and digitized with a Digidata 1550B digitizer (Molecular Devices, San Jose, CA, USA). Signals were then analyzed with pClamp 11 software (Molecular Devices). Excitability experiments were conducted in the current clamp mode with the holding current set to zero. The rheobase was determined using a current ramp protocol to measure the minimum current required to induce ring. The V-I plot was determined by the numbers of spikes generated within 250 ms during a series of discrete 10 pA current steps from 0 to 200 pA. In a subset of experiments, the 5-HT 2C receptor antagonist RS102221 (1 μM) was bath applied for 10 min before the rheobase and VI plot experiments were performed. Synaptic transmission was assessed in the voltage clamp mode.
Spontaneous excitatory postsynaptic currents (sEPSCs) were recorded continuously for 2 min while the membrane potential was maintained at -55 mV; in the same cell, spontaneous inhibitory postsynaptic currents (sIPSCs) were recorded while the cell was held at 10 mV. Synaptic drive was calculated as follows:

Fluorescence In situ hybridization (FISH)
Mouse tissue: Animals were deeply anesthetized with iso urane before decapitation. Brains were then rapidly isolated, frozen on dry ice, and embedded in tissue freezing media (Tissue-Tek O.C.T. compound, Sakura Finetek USA, Torrance, CA). 16 µm NAcc coronal sections were obtained with a Leica CM3050S cryostat and mounted onto SuperFrost Plus glass slides (VWR, Radnor, PA, USA). The sections were collected in a way that each slide contained rostral, middle, and caudal NAcc sections. Once mounted, slices were rapidly refrozen and stored at −80°C before the FISH procedure. Multiplexed FISH was performed using the ACDBio RNAscope V2 reagents and protocols (Cat No: 323100; Newark, CA, USA).
After the pretreatment steps, respective mRNA probes for mouse Pdyn, Htr1b, Htr2a and Htr2c were applied (see probe details in Table 3). A series of ampli cation and detection reagents were applied to the tissue sections according to the manufacturer's instructions. Sections were counterstained using DAPI provided in the detection reagent kits and mounted in ProLong Gold mounting media (P36930, Thermo Fisher Scienti c). The images were obtained using an Olympus FV3000 confocal laser scanning microscope with a 40X magni cation oil immersion objective. The confocal images were processed in ImageJ Fiji software. Cell detection and RNAscope puncta quanti cation were performed using a custom pipeline in CellPro ler (version 3.1.5) and QuPath (version 0.4.2).
Human tissue: Frozen NAcc brain blocks were obtained from the Iowa Neurobank for FISH analysis (Table 1). 20 µm sections were obtained and mounted on SuperFrost Plus glass slides in the cryostat chamber and the slides were stored at −80°C before the FISH procedure. A similar protocol was used as described above with human-speci c probes for Pdyn and Htr2c mRNA ( Table 3). The slides were scanned using an Olympus SLIDEVIEW VS200 digital slide scanner and the images were analyzed in QuPath (version 0.4.2), including cell detection and RNAscope puncta quanti cation.

Immuno uorescence and Confocal Imaging
Deeply anesthetized mice (with tribromoethanol, 250 mg/kg, i.p.) were transcardially perfused using PBS followed by 4% PFA. After post-xation for 24 h, the brains were cryopreserved and cryosectioned at 30 µm with a Leica SM2010R microtome. For immuno uorescence, the brain slices were washed 3X in PBS, permeabilized in 0.5% Triton X-100-PBS solution for 0.5 hour, and then blocked in blocking buffer (10% normal donkey serum/NDS, 0.1% Triton X-100-PBS solution) for 1 hour. Following the blocking procedure, slices were incubated with primary antibodies diluted in the blocking buffer (Table 4) overnight at 4°C with agitation. The next day, slices were washed 3X in PBS and then incubated with speci c uorescent dye conjugated secondary antibodies diluted in PBS (Table 4) for 2 hours. Finally, they were mounted with Vectashield antifade mounting medium (Vector Laboratories: H-1000-10, Newark, CA, USA) on glass slides. Fluorescent images were acquired via an Olympus FV3000 confocal microscope. Images were then processed with ImageJ Fiji software. Pseudo-coloring was applied in certain images to improve visibility.

iDISCO
Following perfusion and post-xation in 4% PFA, samples were rinsed with PBS 3X and stored in PBS with 0.02% NaN 3 at 4 C until the iDISCO protocol was performed. One hemisphere of each brain was used in iDISCO. Sample treatment, immunolabeling, and clearing were performed as described in the iDISCO protocol 14 (available at idisco.info), which are brie y outlined below. Brains were kept in individual 5mL Eppendorf tubes (Hamburg, Germany) during the process, and gentle agitation was applied at each step unless otherwise noted. to the Allen Brain Atlas, background was corrected, and cells were detected with a detection threshold of 20 voxels for cell size. Cell counts were compared using a 2-sample student t-test, assuming unequal variances.

Fiber photometry
The social interaction assay was performed in a camera-monitored open eld arena (L x W x H: 50 x 25 x 25 cm) located in a dimly lit (~20 lux), sound-attenuating box. The day before behavioral testing, mice were gently handled and acclimated to the arena. In the 5-HT biosensor experiment, the Amuza wireless ber photometry system was used, and recordings were conducted with 470 nm LED excitation, 90 µW light power, and 100 Hz sampling rate. First, the mouse was mounted with a headstage, and placed in the arena equipped with a small, empty corner cage. Prior to recording, the mouse was allowed to acclimate to the arena for 20 min. The social interaction assay was split into two phases: a 10-min baseline phase during which the mouse was free to explore, followed by a 10-min social phase, which began with a novel C57BL/6J mouse of the same age and sex being carefully placed in the corner cage. Videos of the experiments were recorded using Ethovision XT 14 software and scored manually o ine. During the baseline phase, facing, sni ng, or climbing on the empty cage were scored as exploration. After introduction of the stranger mouse, all the behaviors speci ed above were scored as social interaction.
Fiber photometry data were acquired using TeleFipho software (Amuza) and analyzed o ine with MATLAB R2021b software (MathWorks, Natick, MA, USA). First, to correct for photobleaching, we created a linear regression model, using the MATLAB tlm function to t the raw data. Second, we normalized each uorescent intensity value to the corresponding value predicted by the linear regression model. Finally, we computed peri-event uorescent changes (ΔF/F) between the last 10 s pre-event (baseline) and the rst 10 s post-event, and compared area under the curve (AUC) in the peri-event plot during the rst 10 s of the rst social interaction versus empty cage investigation.
In the dopamine biosensor experiment, we utilized the same procedure except that a wired ber photometry system, Neurophotometrics FP3001, was employed. Therefore, instead of being mounted with a headstage, the mouse was tethered with a low-auto uorescence bundle branching ber-optic patch cord connected to FP3001, and ber photometry was recorded with interlaced 415 nm and 470 nm LED stimulation at 50 µW light power and 40 Hz sampling rate. All ber photometry data were acquired using Bonsai software (Open Ephys, Atlanta, GA, USA) and analyzed o ine with MATLAB. To normalize the raw data for photobleaching correction, we rst tted the isosbestic data (415 nm) to a two-term exponential model. Then, we applied the dopamine-dependent data (470 nm) to the isosbestic model. Finally, the dopamine-dependent uorescent intensity values were normalized to the corresponding values predicted by the model. Peaks were identi ed using MATLAB ndpeaks function. Frequency was de ned as the number of peaks during a bout divided by the length of the bout. Latency was de ned as the time from the start of the behavioral bout until the next uorescence peak. Amplitudes were normalized to the minimum and maximum for each recording.

Two-photon imaging
AAVs expressing the GRAB 5-HT3.5 biosensor (AAV9-hSyn-5-HT3.5) were stereotaxically injected into the DRN of C57BL/6J mice, as described above. Four weeks later, ex vivo two-photon uorescent imaging was conducted, using an Olympus FVMPE RS multiphoton microscope paired with Mai Tai Si:Sapphire laser (Spectra-Physics, Mountain View, CA, USA). First, acute brain slices containing the DRN were prepared. Then, they were transferred to a recording chamber lled with aCSF saturated with 95% O 2 /5% CO 2 , owing at the rate of 2 ml/min. The excitation wavelength was 920 nm, and uorescence was detected with a 495-540-nm lter. Data were analyzed using ImageJ and MATLAB software.

Transmission electron microscopy
Sert-cre mice were stereotaxically injected with AAV8.2-ef1a-DIO-synaptophysin-eYFP in the DRN, and transmission electron microscopy was utilized to observe synaptic junctions between eYFP-labeled presynaptic part and dynorphin A-labeled post-synaptic part in the ventral NAcc. Brie y, we rst performed immunohistochemical staining on free-oating 45 µm NAcc sections. They were incubated with chicken anti-GFP (1: ImmunoResearch) and 12 nm colloidal gold conjugated donkey anti-rabbit (1:100;; Jackson ImmunoResearch) secondary antibodies for 4 hours at room temperature. Following poststaining processing and dehydration 42 , the sections were embedded in resin. Then, serial ultrathin (50-100 nm) sections were made with a Leica EM UC7 ultramicrotome and mounted on mesh grids (4-6 sections/grid; 3.05 mm diameter). Finally, they were scanned under a 120 kV Hitachi HT7800 transmission electron microscope (Hitachi High-Tech, Tokyo, Japan).

Statistical analysis
Paired or unpaired student t tests were used for comparisons of two groups with one independent variable. One-way ANOVA was applied for comparisons of more than two groups with one independent variable and two-way or three-way ANOVAs were applied for data sets with more than one independent variable. If ANOVAs revealed signi cant main or interaction effects, post hoc Bonferroni tests were employed for pairwise comparisons. In the case of unequal variances, Welch's correction was applied.

Figure 8
Conceptual model depicting how chronic alcohol use can induce social de cits.
Chronic ethanol exposure drives hyperactivity in 5-HT DRN neurons, which activate 5-HT2C receptors in dynorphin neurons in the NAcc. This in turn inhibits dopamine release from dopamine terminals in the NAcc during social investigation, which ultimately inhibits that behavior. Parts of this gure were made using Biorender.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. ExtendedDataFiguresCombined3compressed3.pdf