Prior alcohol use enhances vulnerability to compulsive cocaine self-administration by promoting degradation of HDAC4 and HDAC5

To distinguish the effects of alcohol preexposure from alcohol coexposure, we compared the outcomes of cocaine self-administration in the alcohol-primed group ( “AP”; Fig. 1A) with two control groups: (i) an alcohol-naïve (AN) group that drank water instead of alcohol and (ii) an alcohol-concurrent (AC) group that started using alcohol and cocaine on the same day. Animals in the AP and AC groups did not differ in their daily alcohol intake patterns (Fig. 1B). Prior exposure to alcohol did not increase the daily intake of cocaine during the acquisition or maintenance phase of the sequential paradigm (fig. S1). Prior exposure to alcohol did not affect the acquisition rate [duration of time required to learn to lever press for cocaine injection on a fixed ratio of 5 (FR5)]. We found that the animals learned to lever press on an FR5 schedule after 5.9, 6.2, and 6.6 days in the alcohol-naïve, alcohol-primed, and alcohol-concurrent groups, respectively (Fig. 1C). Similarly, the groups did not differ in the average number of lever presses for cocaine during a 7-day maintenance phase (Fig. 1D). Overall, our behavioral findings on the effect of prior alcohol use on the early stages of cocaine self-administration (days 11 to 21; Fig. 1A) are consistent with the observations of Fredriksson et al. (12). Using a 12-day cocaine self-administration protocol, they reported no change in acquisition of cocaine self-administration in animals that had an extensive 7-week history of prior alcohol use.

Cocaine addiction is characterized not only by the choice to self-administer cocaine but also by the complex behaviors that define addiction, such as increased motivation for the drug, persistence of drug seeking in the absence of reward, and continued drug use despite negative consequences. To determine whether animals with a history of prior alcohol use have enhanced cocaine addiction–like behaviors, we assayed these key behavioral indices using an approach modeled after Deroche-Gamonet et al. (13) and Belin et al. (14). To examine the persistence of cocaine seeking in the absence of reward, we measured the number of lever presses during “no drug periods” of daily cocaine self-administration (two 15-min intervals during which the levers are available but are not rewarded). Animals in the alcohol-naïve group averaged 18 lever presses during the time-out periods. Animals in the alcohol-primed group had significantly enhanced persistence in comparison to alcohol-naïve animals, averaging 58 lever presses. Animals in alcohol-cocaine concurrent control averaged 38 lever presses during the time-out period but were not significantly different from the alcohol-naïve group (Fig. 1E).

We next measured motivation for cocaine self-administration on a progressive ratio schedule of reinforcement, where the animal was required to make increased number of lever presses to earn each subsequent intravenous cocaine reward (0.8 mg/kg per injection), until it reached the breakpoint where it ceased lever pressing (15). We found that alcohol-naïve animals worked for cocaine until an average breakpoint of 310 lever presses, whereas alcohol-primed animals worked for cocaine until they reached an average breakpoint of 563 lever presses. Animals in the alcohol-concurrent group had an average breakpoint of 317 lever presses, similar to the alcohol-naïve group (Fig. 1F).

One of the core characteristics of drug addiction in humans is compulsivity, as reflected in continued drug use despite negative consequences (13, 16). Prior studies have found that most rats stop self-administering cocaine when challenged with an aversive footshock 20 days after initiating cocaine self-administration (16). But when rats have had long-term exposure to cocaine (~60 days of daily cocaine self-administration), a subset of animals continue to seek and take cocaine despite an aversive footshock (13, 17). We reasoned that exposure to alcohol before cocaine would reveal a compulsive phenotype that would not otherwise occur in the absence of priming by alcohol.

To test for compulsive drug use, we introduced an aversive 2-s foot shock (0.1 to 0.3 mA) to the cocaine self-administration session, with an increase in shock intensity occurring every 40 min. We found that animals in the alcohol-naïve and alcohol-concurrent groups declined similarly, earning 73 and 76% of baseline reward at 0.1 mA, 39 and 31% of baseline reward at 0.2 mA, and 16 and 14% of baseline reward at 0.3 mA. By contrast, animals preexposed to alcohol were significantly more resistant to punishment, earning 90, 68, and 29% of baseline reward at 0.1, 0.2, and 0.3 mA of footshock, respectively (Fig. 1G).

Alcohol is a nonspecific and potent neurotoxin, and could decrease sensitivity to footshock by decreasing peripheral pain sensitivity or fear memory. We found, however, that alcohol use did not enhance compulsive lever pressing for sugar pellets in food-restricted animals (Fig. 1H). Thus, the priming effects of alcohol on cocaine compulsivity appear to be mediated by processes that do not generalize to natural rewards. Together, our results indicate that voluntary alcohol use increases vulnerability to cocaine by enhancing persistence of drug seeking, motivation, and continued use despite negative consequences.

To determine whether prior exposure to cocaine enhances alcohol preference, we reversed the order of drug exposure so that cocaine self-administration preceded the start of voluntary alcohol use by 10 days (fig. S2A). Animals in the control group with no prior history of cocaine showed increasing preference for alcohol in a two-bottle free choice paradigm, whereas animals with a prior history of cocaine use actually had decreased alcohol intake and decreased preference for alcohol (fig. S2, B and C).

When we analyzed the daily alcohol intake of animals in our sequential alcohol-to-cocaine paradigm (Fig. 1B), we found that animals in the alcohol-primed group had a transient but significant decline in daily alcohol intake during the first 2 days of cocaine self-administration, and after reaching a nadir of alcohol intake, the animals required 4 to 5 days before reaching the prior baseline of daily alcohol intake of 1 to 1.2 g/kg per day (fig. S3). Animals that started alcohol self-administration concurrently with cocaine use also required 4 to 5 days before reaching a daily alcohol intake of 1 to 1.2 g/kg. This is much slower than cocaine-naïve animals that began drinking 1 g/kg by the second day of alcohol use (alcohol-primed group, days 1 to 5). These results suggest that cocaine use may decrease the preference for alcohol even without a priming period.

Our findings are consistent with previous reports that prior cocaine use does not enhance alcohol use in rats (18) and that, in mice, prior cocaine use results in decreased alcohol preference (19). Together, our findings are consistent with a unidirectional cross-sensitization as predicted by the gateway hypothesis: from alcohol to cocaine, but not from cocaine to alcohol.

To gain insight into the molecular mechanisms by which prior alcohol use potentiates the behavioral effects of cocaine self-administration, we examined the epigenetic effects of alcohol use in the nucleus accumbens, a brain region critical for addiction-related learning and memory. In our previous studies with nicotine, we found that nicotine inhibits histone deacetylase (HDAC) activity in the striatum. Inhibition of HDAC activity resulted in global histone acetylation and a permissive epigenetic environment for cocaine-induced gene expression (4). To determine whether alcohol exposure similarly primes the transcriptional response to cocaine, we allowed animals to drink alcohol for 10 daily 2-hour sessions and followed this with a single injection of cocaine (20 mg/kg intraperitoneally) 18 hours after the last alcohol ingestion (Fig. 2A). Accumulation of the ΔFOSB protein in the nucleus accumbens has been found to have a key role in enhancing cocaine self-administration (20), and is induced by cocaine under voluntary (self-administered) conditions as well as experimenter-administered conditions (21). We now asked: Does prior alcohol use enhance the cocaine-induced transcription of this key marker? We carried out quantitative polymerase chain reaction (PCR) using primers specific to the ΔFosB splice variant and found that alcohol preexposure significantly enhanced cocaine-induced expression of ΔFosB in the nucleus accumbens compared to alcohol-naïve animals that also received an injection of cocaine (Fig. 2B).

To determine whether alcohol facilitates acetylation locally at the FosB gene locus, we next performed chromatin immunoprecipitation (ChIP) in lysates of the nucleus accumbens of alcohol-treated animals. A search of the encyclopedia of DNA elements (ENCODE) revealed that acetylation of H3 lysine 27 (H3K27) in the first intronic region of FosB (the gene giving rise to the splice variant ΔFosB) may be a key epigenetic regulatory mark for this gene in several tissues, including brain (fig. S4) (22). We found that alcohol facilitated robust acetylation of H3K27 residues locally at the FosB gene—fourfold above the (water) control group (Fig. 2C). Because the level of global H3 acetylation in the nucleus accumbens has been found to be positively correlated with motivation for cocaine self-administration in rats [although no such effect has been observed for H4 acetylation (23)], we next immunoblotted total histone lysates and found that alcohol also increased global acetylation of H3K27 in the nucleus accumbens (Fig. 2D).

To determine whether the local acetylation at the FosB locus and the global acetylation in the nucleus accumbens are due to changes in HDAC activity, we next tested the enzymatic HDAC activity in nuclear lysates of the nucleus accumbens of animals exposed to 10 days of daily alcohol use. Prior studies have found that the effects of alcohol on HDAC activity are dynamic and highly sensitive to both the acute exposure and abstinence stages of alcohol use (24). Therefore, we tested HDAC activity in nucleus accumbens lysates of animals sacrificed at sequential time points after the 10th daily 2-hour alcohol session (on day 10) (Fig. 3A). To determine whether the effects are specific to long-term alcohol exposure, we performed a similar analysis of nuclear HDAC activity in the nucleus accumbens of animals after only 2 days of alcohol exposure (Fig. 3B).

Ten days, but not 2 days, of alcohol use resulted in a significant decrease in HDAC activity in comparison to water control (Fig. 3, C and D). Ten days of alcohol use caused a progressive decrease in HDAC activity after the end of the alcohol drinking session (Fig. 3E). We observed no significant change in HDAC activity at the time of acute alcohol exposure (“time 0,” during which blood alcohol levels were 193 ± 43 μg/ml; see table S1) and a significant decrease in HDAC activity at time points during which no serum alcohol was detectable (18 and 22 hours; Fig. 3E and table S1).

In line with this decrease in HDAC activity, we also observed an increase in global H3K27 acetylation in the nucleus accumbens, starting at 12 hours after the last alcohol intake (Fig. 3F). Two days of alcohol use did not have a significant effect on H3K27 acetylation in the nucleus accumbens (Fig. 3G). The finding that short-term alcohol use does not affect HDAC activity or histone acetylation suggests that HDACs behave quite differently under short-term versus long-term exposure conditions, fostering a permissive epigenetic environment only after long-term exposure. Together with the findings observed in our earlier studies with nicotine, the present findings suggest that HDAC inhibition, global acetylation in the nucleus accumbens, and specific hyperacetylation of key genes in the nucleus accumbens may be general mechanisms of action for gateway drugs (5).

We next asked: Which specific HDACs are involved in alcohol-induced acetylation, and how are these inhibited by alcohol? Eleven HDACs have been identified in mammals. An extensive body of literature has demonstrated key roles of these enzymes in integrating a diverse array of molecular responses to stimulants such as cocaine (25, 26). We focused on the class IIa HDACs HDAC4 and HDAC5 because they have been previously implicated in regulating the behavioral effects of cocaine, including cocaine self-administration (23, 27, 28). The class II HDACs shuttle between the cytoplasm and nucleus in an activity-dependent manner (29, 30), and are heavily regulated by posttranslational modifications such as phosphorylation, carbonylation, SUMOylation, and ubiquitination (31). Although the class II HDACs carry a mutation in their catalytic domains that render them inactive as deacetylases (32, 33), genetic and pharmacological knockdown of class II HDACs causes decreased HDAC activity (34) and global histone hyperacetylation and enhanced neuronal plasticity (35–37). Class II HDACs mediate histone acetylation by forming a multiprotein co-repressor complex that includes the class I HDAC, HDAC3 (38). HDAC3 is a potent deacetylase. Focal deletion of HDAC3 has been found to enhance long-term memory (39) and to facilitate conditioned place preference to cocaine (40). The catalytically inactive C terminus of HDAC4 is crucial for the recruitment and deacetylase activity of the multiprotein complex, which forms only in the nucleus of the cell. Thus, class II HDACs may function as activity-dependent scaffolds that target the co-repressor complex to specific promoters (38). HDAC4 has been found to be critical for activity-dependent regulation of H3K27 deacetylation in specific gene promoter regions such as BDNF (brain-derived neurotrophic factor) (41).

We first asked: Does either HDAC4 or HDAC5 interact with the FosB gene locus? To address this question, we performed ChIP experiments in the rat striatum with antibodies against HDAC4 and HDAC5 and found that HDAC4, but not HDAC5, interacted with the FosB locus (fig. S5A). This finding is consistent with that of Wang et al. (23), who reported that overexpression of HDAC4 in the nucleus accumbens resulted in a decreased expression of FosB. Having found that HDAC4 is present at the FosB gene locus, we next examined the specific regulation of HDAC4 by asking whether alcohol regulates acetylation at the FosB promoter by altering the subcellular localization of HDAC4 in the nucleus accumbens. Another class IIa HDAC, HDAC5, limits cocaine reward by shuttling from the cytoplasm to the nucleus following acute exposure to cocaine (28). Although purified class II HDACs have weak deacetylase activity (32), Fischle et al. (38) found that, in vivo, nuclear HDAC4 (but not cytoplasmic HDAC4) exists in an enzymatically active multiprotein complex. To explore the effects of alcohol use on subcellular localization of HDAC4, we performed immunohistochemistry on striatal slices of animals exposed to alcohol for 10 days (fig. S5, B and C). We found that an acute exposure to alcohol (time 0) did not result in a significant change in HDAC4 subcellular localization compared to control animals. The HDAC4 stain completely and precisely overlaid the nuclear stain in 81% of nucleus accumbens cells in both the water control group and the acutely exposed alcohol group (time 0). By contrast, 18 hours after the last alcohol exposure on day 10 (when blood alcohol was not detectable; table S1), a significantly lower number of cells had HDAC4-positive nuclei, suggesting that HDAC4 accumulated less in the nucleus during the alcohol abstinence interval (fig. S5, B and C).

We next immunoblotted for HDAC4 in nuclear and cytoplasmic fractions of nucleus accumbens cells from animals sacrificed at four successive time points after the last alcohol ingestion (0, 12, 18, and 22 hours). Similar to immunohistochemical stains, we found that acute alcohol exposure (time 0) did not lead to a change in the nuclear accumulation of HDAC4, whereas the lack of alcohol resulted in a significant decrease in the nuclear accumulation of HDAC4 at 18 and 22 hours after the last alcohol ingestion (Fig. 4A). In parallel, we also performed immunoblotting studies of HDAC5 following a 10-day alcohol exposure, which revealed a similar decrease in nuclear accumulation 18 hours after the last alcohol ingestion (fig. S6B). Although HDAC4 may be a specific regulator of FosB, it is possible that other HDACs may be involved and may account for the changes that we observed in nuclear HDAC activity and global histone acetylation.

Surprisingly, we found that alcohol-induced decrease of nuclear HDAC4 accumulation was not paralleled by a comparable increased accumulation in the cytoplasmic compartment, as might be expected for an enzyme that shuttles between the nucleus and the cytoplasm (Fig. 4, A and B). In addition, quantitative analysis of HDAC4 mRNA revealed no change at the level of transcription in response to alcohol exposure (Fig. 4C). These findings suggested that the observed decrease in nuclear HDAC accumulation might be due to the degradation in the nucleus, not to shuttling to the cytoplasm. Studies by Potthoff et al. (42) have demonstrated that the activity-dependent differentiation of mature skeletal muscle is mediated by selective proteasomal degradation of class IIa HDACs in the nucleus of muscle fibers. To test for this possibility, we placed in-dwelling cannulae into the nucleus accumbens of rats bilaterally and allowed the animals to drink either alcohol or water for 10 days. Twelve hours after the last exposure to alcohol, we injected a covalent inhibitor of the proteasome in one hemisphere [one-time injection of 200 μM lactacystin, 0.4 μl into the nucleus accumbens, as described by Massaly et al. (43)] and injected dimethyl sulfoxide (DMSO) vehicle in the contralateral, control hemisphere. Lactacystin rescued the decrease in nuclear HDAC activity (Fig. 4D) and kept nuclear HDAC4 and HDAC5 protein levels at control levels, with no change in the vehicle-treated contralateral hemisphere (Fig. 4E and fig. S6B). Lactacystin infusion did not affect HDAC4 or HDAC5 accumulation in the cytoplasm, suggesting that the degradation of these HDACs is occurring in the nuclear compartment (Fig. 4F and fig. S6C). Proteasome-mediated degradation of HDAC4 in the nuclear compartment has been found to mediate growth factor–induced cell motility (44) and activity-dependent striatal muscle differentiation (42). Our findings therefore suggest that, in addition to the well-described phosphorylation-dependent shuttling of the class II HDAC (28), other posttranslational modifications, such as SUMOylation (45) and ubiquitinization (44), may have key mechanistic roles in the way in which class II HDACs regulate vulnerability to drug addiction. Moreover, these data suggest that the class II HDACs, HDAC4 and HDAC5 may act as nodal regulators, integrating environmental stimuli (alcohol use) with behavioral response (potentiated reward-based learning, resulting in addiction-like behavior).

To determine directly whether nuclear degradation of HDAC4 and HDAC5 causes an increase in addiction-like behavior, similar to the behaviors induced by alcohol (Fig. 1), we treated animals to the class II–specific HDAC inhibitor MC1568 (46, 47). MC1568 selectively inhibits class II HDAC activity by promoting nuclear import and proteasome-mediated degradation of HDAC4 and HDAC5 in the nucleus of the cell (48). Several other HDAC inhibitors, including suberoylanilide hydroxamic acid, valproic acid, and quinidine, have been found to inhibit HDAC activity by promoting proteasome-mediated degradation of HDAC4, HDAC2, and HDAC1, respectively (48–50). Systemic administration of MC1568 is well tolerated by mice, and crosses the blood-brain barrier, at doses of up to 6.5 mg/kg (51). In addition, MC1568 promotes global H3 hyperacetylation and enhances neurite outgrowth in dopaminergic and sympathetic neurons in vitro (37).

Although the effects of MC1568 on class II HDACs have been well characterized in cancerous and noncancerous tissue (34, 46, 48), the effects of this drug on HDAC4 and HDAC5 have not been described in the nucleus accumbens in vivo. Thus, we first characterized the effects of MC1568 in the nucleus accumbens. We treated animals to 10 daily injections of MC1568 (0.5 mg/kg) and tested for nuclear HDAC activity in the nucleus accumbens 18 hours after the last treatment. We found that MC1568 causes a 35% decrease in nuclear HDAC activity (Fig. 5B). To determine whether MC1568 selectively decreases class II HDACs, we immunoblotted nuclear and cytoplasmic fractions of the nucleus accumbens cells for class II HDACs, HDAC4 and HDAC5, and for class I HDACs, HDAC1 and HDAC2. We found that 10 days of daily exposure to MC1568 caused a significant decrease in nuclear HDAC4 and HDAC5 in the nucleus accumbens, with no significant change in HDAC1 and HDAC2 (Fig. 5C). We also observed no changes in the cytoplasmic accumulation of HDAC4 and HDAC5 (Fig. 5D). These findings are consistent with previous studies showing that MC1568-mediated inhibition of class II HDACs is facilitated by the degradation of HDAC4 and HDAC5 in the nucleus (48, 52).

To explore the effect of selective degradation of HDAC4 and HDAC5 on cocaine addiction–like behaviors, we administered MC1568 (0.5 mg/kg), or vehicle, daily for 10 consecutive days before cocaine self-administration (Fig. 5A). Drug treatments were continued concurrently with cocaine self-administration but administered at a different time during the dark cycle (Fig. 5A). Treatment with MC1568 did not increase the daily intake of cocaine during the acquisition or maintenance phase of the sequential paradigm (fig. S7). We tested motivation for cocaine on a progressive ratio schedule of reinforcement, at three cocaine doses (0.8, 0.4, and 0.2 mg/kg), on days 15, 17, and 19 of cocaine self-administration, respectively. As expected, animals had fewer cumulative lever presses for cocaine with each successive decrease in cocaine concentration (23). Animals treated with the selective class II HDAC inhibitor, however, had significantly higher motivation for cocaine at a unit dose of 0.4 mg/kg and no significant difference in motivation at the lowest unit dose of 0.2 mg/kg (Fig. 5E). We measured persistence of cocaine seeking in the absence of the drug and found that animals exposed to MC1568 had significantly higher persistence than the control group (Fig. 5F).

To determine whether selective degradation of class II HDACs enhances compulsivity of cocaine use, we used the variable shock paradigm described in Fig. 1, where we introduced an aversive 2-s footshock to the standard self-administration session, with an increase in shock intensity occurring every 40 min. Under mild aversive condition (0.1 mA of footshock), control and MC1568-treated animals did not differ, earning 83% of baseline reward. However, at 0.2 mA of punishment, the MC1568 group earned almost twice as much reward (65% versus 37% of baseline for drug-treated and vehicle, respectively); at 0.3 mA of footshock, the control animals earned no drug reward, whereas the MC1568 group earned 35% of baseline reward (Fig. 5G). We found that MC1568 did not enhance resistance to punishment in food-restricted animals pressing for sugar pellets (Fig. 5H). These data indicate that MC1568 does not disrupt fear circuits involved in decision-making, and suggest that, like alcohol, the priming effect induced by the selective HDAC inhibitor does not generalize to natural rewards. Together, these findings further substantiate our observation that the degradation of HDAC4 and HDAC5 increases vulnerability to cocaine addiction–like behaviors.

Sprague-Dawley rats (8 to 12 weeks, weighing 280 to 300 g at the beginning of experiments) (Taconic Laboratories) were used for all studies. Rats were single-housed on a reverse light/dark schedule (10 a.m. lights off, 10 p.m. lights on). All experiments were conducted during the dark cycle and in accordance with institutional guidelines.

Cocaine HCl (provided by the National Institute on Drug Abuse) was dissolved in 0.9% NaCl to a concentration of 5 mg/ml and filtered through 0.2-μm filters. Lactacystin (Sigma) was dissolved in 2% DMSO to a final concentration of 200 μM, as described by Massaly et al. (43). MC1568 (ApexBio) was dissolved in sterile saline/0.3% Tween/5% DMSO.

Animals were acclimated to ad libitum water access in 100-ml graduated feeding tubes (Dyets) for 4 days before initiation of alcohol treatment protocol. Alcohol treatments occurred each day, 3 p.m. to 5 p.m. (approximately 5 hours after the start of the dark cycle). We used a one-bottle, limited alcohol access paradigm, to obtain stable alcohol intake between animals, with fast acquisition of alcohol drinking. In the one-bottle paradigm, animals voluntarily drink from one bottle of 10% alcohol in their home cages, approximately 5 hours into the dark cycle. Animals were water-restricted for 18 to 24 hours before the first day of alcohol exposure to stimulate drinking. In all experiments, an overnight water bottle was left with the animal after the 2-hour alcohol treatment, to be removed the next morning at approximately 10:30 a.m. No liquids of any kind were given to animals between the hours of 10:30 a.m. and 3 p.m. to control for the animals that undergo behavioral studies and do not drink during this time period due to cocaine self-administration (11 a.m. to 2 p.m.).

Serum alcohol, cocaine, and cocaethylene levels in the sequential drug administration paradigm: When taken simultaneously, alcohol and cocaine can have synergistic effects: (i) Alcohol ameliorates the anxiogenic effects of cocaine, resulting in higher quantities of cocaine use during coadministration (7, 8); (ii) alcohol and cocaine form a metabolic by-product in the liver, cocaethylene, which by itself is euphorigenic (9, 10); and (iii) alcohol can inhibit cocaine metabolism, resulting in higher serum cocaine concentrations (9, 11). Access to the two drugs was therefore restricted to different times during the dark cycle, thereby limiting behavioral, metabolic, and pharmacokinetic interaction.

ChIP experiments were performed to study histone modifications (HM ChIP) and DNA interactions with HDAC ChIP at the FosB gene using a ChIP assay kit (EMD Millipore). Briefly, for HM ChIP, nucleus accumbens punches were homogenized and cross-linked for 10 min at room temperature in the presence of 1% formaldehyde. The cross-linking reaction was quenched with an excess presence of glycine. Following washes, the cell pellet was resuspended in SDS lysis buffer and sonicated to shear the DNA to approximately 1000 base pairs (bp). The sonicated cell supernatant was diluted, precleared with protein A agarose/salmon sperm DNA beads, and then incubated overnight at 4°C in the presence of 5 μg of an anti-H3K27 acetylation antibody (Active Motif, #39133). The antibody/histone complex was collected with protein A agarose/salmon sperm DNA beads and washed five times according to the assay’s protocol. The histone complex was then eluted from the antibody, histone-DNA cross-links were reversed overnight, and the eluate was treated with proteinase K. DNA was recovered using the Chromatin IP DNA Purification Kit (Active Motif), and the samples were analyzed in triplicates using qRT-PCR as previously described. Data were calculated as % input [100*2^(Adjusted input − C_t (IP)] and were then transformed to fold change differences relative to the water group.

The primer pair used was targeting the first intron of the rat FosB gene to capture the region showing elevated H3K27 acetylation levels based on ENCODE’s mouse data (forward, 5′-ACGACGACCTTCAATTCCCC-3′; reverse, 5′-ATTCTTGTCAGGATCCGGCG-3′) (see fig. S4). For the HDAC ChIP analyses, the same procedure was used as before with the following exceptions: Dorsal striatal punches (n = 8) were homogenized and dual cross-linked with 5 mM disuccinimidyl glutarate for 1 hour, followed by 1% formaldehyde for 10 min at room temperature. Samples were sonicated to approximately 1000 bp, pooled, and divided equally into four samples that were incubated with 5 μg of either one of the following antibodies: HDAC4 (Santa Cruz Biotechnology, sc-11418 X), HDAC5 (Santa Cruz Biotechnology, sc-133106), rabbit immunoglobulin G (IgG) (Cell Signaling Technology, 27295), and mouse IgG (Santa Cruz Biotechnology, sc-2025). Both HDAC antibodies were first confirmed to work in IP experiments to not cross-react with each other. The eluted DNA was analyzed using qRT-PCR using the same primers that capture the first intron of FosB, which is the region showing elevated H3K27 acetylation levels. Because only the HDAC4 and input samples gave a detectable signal (C_t values for HDAC5 and both IgGs were undetectable), the data were visualized by running the samples on a 1% agarose gel.

Nucleus accumbens punches were processed using the NE-PER Nuclear and Cytoplasmic Extraction Kit (Thermo Fisher Scientific) to extract a cytoplasmic fraction and a nuclear fraction. The protocol was modified to also obtain a separate histone fraction: Following extraction of the cytoplasmic and nuclear fractions, the pellet was resuspended in nuclear extraction reagent and incubated with 300 U of micrococcal nuclease, in the presence of 5 mM CaCl₂, for 15 min at 37°C. To confirm efficient subcellular fractionation, Western/immunoblotting experiment was performed for proteins known to be present in all fractions (HDAC4; Santa Cruz Biotechnology, sc-11418), present in the cytoplasm (mitochondrial HSP60; Abcam, ab45134), present in the nucleus and cytoplasm (β-actin; Abcam, ab6276), and present in the histone fraction (histone H3; Abcam, ab10799). The results of the subcellular fractionation experiment are presented in fig. S8. For all Western blotting experiments, equal amounts of nuclear, cytoplasmic, or histone extracts were run on 4 to 20% gradient TGX precast gels (Bio-Rad), transferred to polyvinylidene difluoride membranes, and immunoblotted with the antibodies of interest: acetyl-H3K27 (Active Motif, #39133), HDAC1 (Abcam, ab7028), HDAC2 (Abcam, ab12169), HDAC4 (Santa Cruz Biotechnology, sc-11418), and HDAC5 (Santa Cruz Biotechnology, sc-133106). Total H3 (Abcam, ab10799) was used as loading control for the histone fractions, and β-actin (Abcam, ab6276) was used as loading control for the nuclear and cytoplasmic fractions.

HDAC activity was measured using the Epigenase HDAC Activity Assay, according to the manufacturer’s protocol (EpiGentek).

For behavioral analyses with more than one time point, group comparisons were performed using RM ANOVA, followed by Tukey or Sidak post hoc tests. For all remaining comparisons, Student’s t test was used to compare alcohol-treated groups to the untreated (water) control group. Student’s t test was also used to compare the alcohol + DMSO group with the water and the alcohol + lactacystin groups, respectively. Potential outliers were excluded from the statistical analyses. All statistical analyses were performed using GraphPad Prism, and a P value of less than 0.05 was considered significant.