Preclinical and clinical evidence for suppression of alcohol intake by apremilast

Treatment options for alcohol use disorders (AUDs) have minimally advanced since 2004, while the annual deaths and economic toll have increased alarmingly. Phosphodiesterase type 4 (PDE4) is associated with alcohol and nicotine dependence. PDE4 inhibitors were identified as a potential AUD treatment using a bioinformatics approach. We prioritized a newer PDE4 inhibitor, apremilast, as ideal for repurposing (i.e., FDA approved for psoriasis, low incidence of adverse events, excellent safety profile) and tested it using multiple animal strains and models, as well as in a human phase IIa study. We found that apremilast reduced binge-like alcohol intake and behavioral measures of alcohol motivation in mouse models of genetic risk for drinking to intoxication. Apremilast also reduced excessive alcohol drinking in models of stress-facilitated drinking and alcohol dependence. Using site-directed drug infusions and electrophysiology, we uncovered that apremilast may act to lessen drinking in mice by increasing neural activity in the nucleus accumbens, a key brain region in the regulation of alcohol intake. Importantly, apremilast (90 mg/d) reduced excessive drinking in non–treatment-seeking individuals with AUD in a double-blind, placebo-controlled study. These results demonstrate that apremilast suppresses excessive alcohol drinking across the spectrum of AUD severity.


Mouse strains and animal care
To test whether apremilast reduces binge-like drinking and the motivation for alcohol, we used selectively bred High Drinking in the Dark (HDID-1, HDID-2) and inbred -HDID-1 (iHDID) mice of both sexes. Female and male HDID-1 were also used to determine whether PDE4 inhibition in the NAc alone is sufficient to reduce binge-like ethanol drinking. The above lines were bred and Tween-80 (1.875% v/v) and saline and administered intraperitoneally (i.p.) in a volume of 10 mL/kg mouse body weight. For the chronic intermittent ethanol (CIE) study, apremilast was prepared in an identical manner and administered 10 mL/kg by oral gavage, 2 hours prior to measured drinking. For the intra-NAc apremilast study, 1 µL of 2 µg/µL apremilast (dissolved in 35% DMSO and Dulbecco's phosphate buffered saline, DPBS) was administered bilaterally into the nucleus accumbens (where cannula placement was verified using methylene blue (7)). For electrophysiology, frozen aliquots of 50 mM apremilast in 100% dimethyl sulfoxide (DMSO; Fisher) were defrosted and added to artificial cerebral spinal fluid (ACSF), yielding final concentrations of 1 µM apremilast and 0.002% DMSO. Vehicle solutions for each study were prepared identically without the addition of rolipram or apremilast.

Effects of chronic binge-like drinking on Pde4 gene expression
HDID-1 female mice (n = 46-48/fluid group) were subjected to chronic binge-like drinking (8-weeks of DID), where mice were offered measured 20% ethanol or water 4-days/week and had access to water (unmeasured) at all other times. 21-hrs after the last DID exposure (when there is no alcohol in the system, reducing possible confounding effects of dose-related issues) in week 8, mice were euthanized by cervical dislocation and rapid decapitation. Whole brains were frozen on powdered dry ice, cryostat sectioned (200 µm), and frozen NAc tissue punches were collected and processed for RNA isolation (using PureZOL and Aurum total RNA fatty and fibrous tissue kit; Bio-Rad). This tissue was generated to determine whether target genes of interest were altered by chronic alcohol drinking. mRNA was reverse transcribed into cDNA using iScript cDNA synthesis kit (Bio-Rad). Quantitative real-time PCR (qPCR) reactions were performed using a CFX384 Touch qPCR system. Reactions were performed in duplicate for each sample to determine relative levels of 18s, Pde4a, and Pde4b [Bio-Rad PrimePCR probe assay, targeting mouse genes: Pde4a (5'HEX,3' Iowa Black® FQ, Unique Assay ID:qMmuCIP0030727), Pde4b, 5'TEX 615,3' Iowa Black® FQ, Unique Assay ID:qMmuCIP0036026), and 18s (RPS18, 5' 6-FAM,3' Iowa Black® FQ, Unique Assay ID: qMmuCEP0053856). Data was analyzed using the ddCT method for determining relative gene expression (8,9).

Testing the effects of PDE4 inhibition on binge-like ethanol drinking
Drinking in the Dark (DID) Test: As in Rhodes et al. (2005), mice were habituated to individual housing and a new sipper bottle for one week prior to testing (10). For the ethanol DID assay, mice were offered a single bottle of 20% ethanol (v/v, in tap water), 3 hours after lights off. The general approach used was to measure fluid intake daily for either 2 hrs (first three days) or 4 hrs (fourth day); see specific experimental details below. Water was then restored for the remainder of the day (unmeasured). Details for each study follow below. Mice were pseudorandomly assigned to treatment groups based on their alcohol intake during the first 3 days of DID to achieve comparable drinking levels prior to testing on the fourth day. Treatment group assignment was fixed and did not change.
Rolipram: HDID-1 male and female (n = 6-7/sex/rolipram treatment) mice were habituated to individual housing for one week as described above. Mice were then subjected to ethanol DID, where they received access to 20% ethanol for 2-hrs on days 1, 2, and 3. On day 4, mice were treated with 0, 5, 7.5, or 10 mg/kg rolipram 30-min prior to a 4-hr DID session. Immediately after the drinking session, a 20 µL peri-orbital sinus blood sample was collected for determination of blood alcohol level (BAL) by gas chromatography using methods described previously (11).
Apremilast: HDID-1 male and female mice (n = 10-12/sex/apremilast treatment) were habituated to individual housing for one week as described above. Mice were then subjected to 3 weeks of DID testing (4 days/week), where mice were offered 20% ethanol in week 1, water in week 2, and 9.2 mM saccharin in week 3. Ethanol consumption is determined by many factors, including palatability; therefore, we tested the effects of apremilast on intake of ethanol, as well as water and a sweet tastant solution (saccharin). In week 1, mice received access to 20% ethanol for 2hrs on days 1, 2, and 3. On day 4, mice were treated with 0, 20, or 40 mg/kg apremilast 30-min prior to a 4-hr ethanol DID session. Immediately after the ethanol drinking session, a 20 µL periorbital sinus blood sample was collected for determination of BAL by gas chromatography using methods described previously (11). Water intake in week 2 and 9.2 mM saccharin intake in week 3 were measured identically to ethanol, whereby volumes were measured for 2-hrs on days 1, 2, and 3. On day 4, mice were treated with 0, 20, or 40 mg/kg apremilast 30-min prior to a 4-hr water DID session.

Effects of apremilast on the motivation for ethanol
To determine whether PDE4 inhibition reduces the motivation for ethanol and/or aversion resistant responding for ethanol, we tested the effects of apremilast on oral ethanol selfadministration during a 1) progressive ratio (PR) schedule of reinforcement and for 2) quinineadulterated ethanol. Mice were trained to acquire lever pressing for a food pellet reinforcer (food training) prior to measuring for responding and ethanol intake during fixed ratio responding (FR1, FR3), and PR for access to a 10 mL volumetric sipper filled with 20% ethanol or quinineadulterated 20% ethanol. In total, 72 mice [n = 10-12/sex/apremilast treatment (0, 20, 40 mg/kg apremilast)] were used.
One week prior to testing, mice were habituated to individual housing and sipper bottles (matching those used for operant self-administration). 24 operant testing chambers (Med Associates, VT, USA) were used, which were housed in light and sound attenuating boxes.
Chambers were operated using MedPC IV software (Med Associates, VT, USA). Each chamber was outfitted for lever responding for food (trough) or liquid (sipper) reinforcement, and contained a trough for pellet delivery, an ethanol extender (retracted between reinforcers), a lickometer, two levers with cue lights above, and one house light.
An experimental timeline is provided in supplemental figure 2A. Food training (FT) was used to ensure mice learned to lever-press for access to a reinforcer prior to ethanol self-administration, as shown in Jensen et al. 2021 (12). Mice started food restriction one day prior to training and were maintained at 85% of their free feeding weight throughout FT. In brief, mice were placed for 1-hr daily in an operant conditioning chamber equipped with a chow pellet dispenser and two levers, during which the house light (signaling availability of a reinforcer) and the cue light above the active lever were illuminated continuously. Training was under a fixed ratio 1 (FR1) schedule of reinforcement, whereby a single lever response (FR1) of the active lever press (left) delivered 1 chow pellet as a reinforcer. To avoid overtraining, mice were food trained until they reached 3 sessions earning ≥ 25 chow pellet reinforcers / session. All mice met criteria by 7 sessions of FT.
Operant ethanol self-administration behavior: iHDID mice were subjected to 10, 2-hr sessions of FR1 (5 days/week for 2 weeks) and 10, 2-hr sessions of FR3 (5 days/week for 2 weeks) of operant ethanol self-administration. Ethanol intake for each 2-hr session, number of active and inactive lever presses, and number of ethanol access periods (reinforcers) earned were recorded. Stability in responding was defined as < 20% variance within the mean reinforcers earned within the last 3 sessions of FR3. Mice not exhibiting greater than a 2:1 active: inactive lever pressing ratio (to ensure responses are specific to the reinforcer) or stable responding during FR3 did not continue to progressive ratio (PR) or quine-adulterated FR sessions.
The last responding ratio reached in the 4 hr session was defined as the breakpoint. In addition to the breakpoint, ethanol intake, number of active and inactive lever presses, and number of ethanol access periods (reinforcers) earned were recorded.
To determine whether apremilast reduces aversion-resistant responding for ethanol, the above female and male iHDID mice were subsequently tested for quinine-adulterated ethanol [similar to the design of Sneddon et al. (13)]. After completion of the PR test, mice subjected to three 2hr sessions of FR3 responding for 20% ethanol, followed by three 1-hr sessions of FR3 for 20% + quinine (0, 100, or 500 µM). Mice were pseudorandomized into treatment groups based on the average responses and intake during the last 3 sessions. Using a counterbalanced design, mice were tested for quinine-adulterated ethanol responding during two 2-hr sessions (on an FR3 schedule). Here, half of the group received apremilast (40 mg/kg, IP) 1-hr prior to testing on Day 1 [and vehicle (IP) on day 2], and the other half received vehicle on day 1 and apremilast on day 2. The 40 mg/kg was chosen because the results shown in figure 1 suggested it was more efficacious than 20 mg/kg at reducing intake and reinforcers earned. The number of ethanol access periods (reinforcers) earned, ethanol intake, and number of active and inactive lever presses were recorded.

Effect of intra-accumbens apremilast administration on binge-like drinking
Surgery: HDID-1 male mice (n = 19-20/infusion group; n = 10-11/infusion group for BAL) were administered anesthesia (100 mg/kg ketamine, 10 mg/kg xylazine) and underwent stereotaxic surgery for bilateral guide cannula implantation (coordinates relative to bregma: A/P +1.34, M/L -/+ 0.63, D/V -3.4). The implanted guide cannulae (33-gauge stainless steel, 10 mm long) were aimed at the NAc (core and shell) and were secured by Durelon carboxylate acrylic (3M) and anchored by one stainless steel screw (Small Parts) inserted into the skull. Stainless steel stylets (10 mm, 26 gauge) were inserted into the guide cannula to maintain patency. Mice were provided a nutrient supplement containing carprofen (Medi-Gel CPF, ClearH2O) for 3-days prior to surgery (for acclimation and prevention of neophobic response) and 3-5 days post-surgically to aid recovery. Mice received saline, enrofloxacin, and baytril immediately after surgery. Mice were individually housed after surgery to maintain cannula headmount.
Behavioral testing: Testing began 7-10 days after surgery. Mice were subjected to 3-weeks of DID testing (4 days/week), where mice were offered 20% ethanol in week 1, water in week 2, and 9.2 mM saccharin in week 3. This procedure was very similar to the above apremilast test, but with two modifications: 1) measured fluid access was limited to 2-hrs on all four days of DID testing and 2) on day 4 of each DID, apremilast was administered 15-20 minutes prior to measured fluid access by direct administration into the NAc. Cannula microinjectors (11 mm length, 26 gauge) were used to deliver 1 µl of vehicle or apremilast (2 µg) at a rate of 0.1 µL/minute for over 10-min. After infusion, microinjectors were left in place for 5-min to allow for diffusion before reinserting stylets. For a subgroup of mice, we collected a 20 µL peri-orbital sinus blood sample for determination of BAL by gas chromatography. After the completion of the study (3 weeks of DID testing), all mice were infused bilaterally with 1 µL of filter sterilized methylene blue dye (17 mg/mL) into the NAc. One hour after infusion, animals were euthanized to extract brains. Whole brains were post-fixed in fresh 4% paraformaldehyde in PBS for 2 days, cryoprotected in PBS + 30% glycerol for 24-hr and sectioned on a freezing microtome (100 µm).
Sections were mounted onto slides and inspected using a microscope (localization of methylene blue dye in the NAc was used to confirm injection placement).

Stress + Chronic intermittent stress (CIE):
An experimental timeline is provided in Supplemental data Figure 3A. Adult male C57BL/6J drank ethanol daily (Mon-Fri) for 1-hr/day starting three hours after lights off. Ethanol (15% v/v) and water (2-bottle choice; 2-BC), were available during this 1-hr access period. After 5-weeks of baseline drinking mice were separated into four groups [CTL (air control), FSS (forced swim stress), CIE, CIE+FSS) and entered the CIE ± FSS phase of the study. That is, weekly cycles of CIE/Air exposure were alternated with weekly test drinking cycles, with half the mice receiving 10-min FSS exposure 4-hrs prior to the drinking sessions (remaining mice were not disturbed) (3,14). Therefore, the four groups evaluated in this study were CTL, CIE, FSS (air-exposed mice that experienced FSS before drinking), and CIE+FSS (mice that experienced both CIE and FSS). During test cycle 3, these four groups were further separated into groups based on the apremilast dose condition (0, 20, 40 mg/kg; N= 9-10/group). During Baseline and the first two test cycles, all mice received vehicle (saline) injections (IP) 30-min before ethanol access. During test cycle 3, mice were injected with apremilast (0, 20, or 40 mg/kg) 30-min before the drinking sessions. Solutions were prepared fresh daily. Mice received IP injections in a 10 ml/kg volume.
CIE: An experimental timeline is provided in Supplemental Figure 3B. To test the effects of apremilast on a more chronic model of dependence induced escalations in ethanol drinking, female and male C57BL/6J mice (n = 10/sex/group) underwent a standard CIE protocol (15)(16)(17). In brief, baseline drinking consisted of 2-hr access to 15% ethanol and water, 5-days/week All recordings were filtered at 1 kHz, and digitized at 5 kHz via a Digidata 1440A interface board using Clampex 10.3 (Molecular Devices). Access resistance >30 MΩ or depolarized resting membrane potential (> -60 mV) were criteria for immediate termination of recording. Excitability was measured by applying depolarizing intracellular current steps (300 msec duration) of increasing amplitude, from 50 pA to 550 pA in 50 pA steps, once every 700 msec, in order to determine rheobase (minimum injected current that elicited an action potential) and the number of action potentials fired at each current amplitude. Input resistance was calculated from the average of the voltage responses to small hyperpolarizing current pulses (-50 pA; 100 ms duration) delivered 100 msec before each 300 msec current step. Action potential characteristics were determined from the first current step to elicit firing, with the exception of the maximum peak to peak frequency, for which the current step eliciting the maximum number of action potentials was used. The action potential threshold was determined as the membrane potential at which dV/dt ≥ 10 mV/msec. AHPs are reported as the difference between the threshold potential and the membrane potential at the respective time point post-action potential: fAHP was determined from the minimum membrane potential within 4 msec of AP threshold; mAHP and sAHP were measured 10 and 15 msec after AP threshold, respectively.
All processing of raw data was performed blind to the cell subtype, treatment condition, and sex of the animal.

Statistics
All statistical analyses of mouse studies were conducted with GraphPad Prism Software Version 9.0. The sample sizes for each experiment are reported in the appropriate figure legend. If no significant sex X treatment interactions were observed, we performed statistical analyses on data collapsed across sexes. Data are presented as mean +/-SEM.
For behavioral experiments, post-hoc analyses used the Dunnett method to compare all doses to vehicle control group. To ensure that treatment groups did not differ meaningfully before drug treatment, we analyzed data for the first 3 days of each experiment. In experiments where animals were tested in subsequent weeks with a fluid other than ethanol, drug vs. vehicle groups were the same as during the ethanol DID test in week 1, but we analyzed each week's data separately. The principal dependent variables of interest were g/kg ethanol intake and BAL. Intakes for other fluids were analyzed as mL/kg body weight. For stress + CIE and CIE induced escalation in drinking, preliminary analyses of the data indicated that there were not significant variations in intake across the five days of drinking during baseline or each test cycle.
Therefore, data were averaged across the last five days of baseline and each test cycle before analysis. Data were analyzed with ANOVA (alpha set at 0.05) followed by Newman-Keuls posthoc tests whenever a main effect or interaction was significant. Student's t-test was used to analyze data for ethanol intake and BAL data from the intra-accumbens apremilast experiment.
NAc mRNA expression following chronic binge-like ethanol drinking in HDID-1 mice was analyzed as a 2-way ANOVA (fluid type x time of day). Because there were no time effects for either PDE4a and PDE4b NAc gene expression, data were collapsed across time and analyzed using a Student's t-test.
For electrophysiology experiments, data were analyzed using 2-or 3-way ANOVA in GraphPad Prism, with cell type (D1 or D2 MSN) and treatment condition (vehicle or apremilast) as between-groups factors. The effect of treatment within each MSN subtype was analyzed using Bonferroni's multiple comparison test. Institute Withdrawal Assessment for Alcohol-Revised (21) (CIWA-Ar) was used to assess severity of alcohol withdrawal symptoms; subjects were required to have a negative breathalyzer reading and a CIWA-Ar score < 9 at randomization, to eliminate acute alcohol or withdrawal effects.

Outcome Measures
Efficacy: The Timeline Follow Back Interview (22) (TLFB) was the primary measure used to assess daily intake of standard drinks consumed over the 11-day period of ad libitum drinking.
Results were also assessed as binary-coded heavy drinking days (4+ drinks for females, 5+ drinks for males) over the same period. A standard drink contains ~14 grams of alcohol, such that one 12 oz beer is equivalent to one 5 oz glass of wine and 1.5 oz of distilled spirits. A daily drinking diary was used to confirm TLFB data. Drinking urges were assessed by self-report using the Alcohol Craving Questionnaire-Short Form (23).
Safety: Subjective adverse drug experiences were recorded on standardized case report forms that depicted each side effect complaint in terms of its onset, duration, severity, relation to study medication and clinical action. Vital signs, routine urine and blood tests, EKG and physical exam were conducted pre-and post-treatment to verify that no clinically significant changes from baseline had occurred.
Medication Conditions: Apremilast was purchased from a retail specialty pharmacy, overencapsulated by Lake Hills Pharmacy, Austin, Texas, and matched with identical placebo capsules. Double-blind study drug was given in a standard 7-day titration to 90 mg/d apremilast or matched placebo, administered orally as two 30 mg capsules (60 mg) of apremilast or identical placebo in the AM, and one 30 mg capsule of apremilast or identical placebo in the PM, taken with or without food. Double-blind study drug was packaged in a blistercard with the subject's study ID number and the day and time of each dose indicated on the blistercard. Study drugs, packaging and dosing regimen were identical to preserve the double-blind. All participants, care providers, and those assessing outcomes were blinded to the identity of the assigned medication until after the trial was completed. Medication adherence was verified with returned blistercard and pill count. Apremilast plasma determinations were obtained on the last day of dosing, frozen (-80º C) and analyzed in batch after study completion to verify correct medication assignment per the randomization code and ingestion of active medication and were examined for an association with outcome on an exploratory basis, as apremilast has no established therapeutic plasma level. Determination of apremilast concentration in plasma was made by High Performance Liquid Chromatography (HPLC) in the laboratory of Dr. Esther Maier, Drug Dynamics Institute, TherapeUTex, University of Texas -Austin.
Physiological Indicators: Apremilast is a selective PDE4 and TNF- inhibitor that acts on immune system targets. Hence, plasma for determination of cytokines (TNF-, CCL2, CXXL10) and cortisol concentration was obtained, frozen (-80º C) and assayed in batch after study completion for retrospective evaluation as potential physiological moderators of treatment response. Levels of endotoxin in serum were measured as a marker of leakage of the intestinal barrier. Results ruled out such leakage as an explanation for cytokine levels observed, thereby also ruling out endotoxins as potentially confounding treatment response to apremilast in AUD.
The following assays were run by the CSAR Cell and Immunology Core, Calibration curves were prepared by spiking apremilast into blank human plasma (0.025-0.5 µg/mL) with triplicate preparations across a minimum of six levels. Method selectivity was verified by extracting and analyzing unspiked blank samples. The chromatographic assay was performed with a Dionex Ultimate 3000 HPLC with UV detector using gradient elution on an Agilent Zorbax Eclipse Plus C18 3.5µ 3x150mm with an EC-18 guard column (3.0x5.0mm, 2.7μm), 30°C column temperature, 230 nm, and a run time of 15 minutes. A ternary gradient separation was performed using 0.1% phosphoric acid in water (v/v), acetonitrile, and methanol as the mobile phases, and a flow rate of 0.6 mL/min. Sample Size: There were no previous clinical studies of apremilast for AUD to use in power calculations. Accordingly, we used acamprosate as the reference compound for calculating sample size because it is the most recent drug approved for AUD and novel drugs would be expected to exceed its effect size to merit further development. Data from our acamprosate proof-of-concept study using cue reactivity with Visual Analogue Scale (VAS) craving as the outcome identified a moderate effect size for acamprosate (66.1%). This effect size is based on a total VAS craving score of 5.0 ± 1.40 in subjects tested on acamprosate (n=20) and 16.7 ± 3.83 in subjects tested on placebo (n=20), with a coefficient of variation of 100% for each group.
Based on these data, and assuming a similar or better effect size for drinking would be found, 20 completed subjects per treatment group provides adequate power (80%) to detect an effect size of 66.1% at a two-sided significance level of 0.05. Therefore, randomizing 50 subjects (25 per treatment group) allowed for a generous estimate (based on completed studies) of 2-3 noncompliant subjects and 2-3 dropouts for a total of 20 completed subjects per arm. Subjects were block-randomized to conditions using an urn randomization process to ensure balance, as implemented in the R package randomizr (24).
Statistical Analysis. Changes in drinks per day and probability of a heavy drinking day were specified over the 11-day period of ad libitum drinking using mixed-effect, latent growth models   Table 3. Pre-and post-treatment physiological indicators of treatment response.
Supplemental Table 4. Latent Growth Model of change in log-number of drinks per day over the 11 days of ad libitum drinking following start of medication/placebo, assuming a negative binomial error distribution, random intercepts, and random slopes.
*Results from hypothesis testing for change in number of drinks per day.
1. The time variable (day) was coded 0-10, and centered in the middle of the range, i.e., at 5.5, in order that the intercept zero point would provide a reasonable comparison of amount of drinking between drug and placebo groups.
2. Treatment was coded 0 for placebo group and 1 for apremilast group. 1. The time variable (day) was coded 0-10, and centered in the middle of the range, i.e., at 5.5, in order that the intercept zero point would provide a reasonable comparison of amount of drinking between drug and placebo groups.
2. Treatment was coded 0 for placebo group and 1 for apremilast group. Supplementary