Cessation of fluoxetine treatment increases alcohol seeking during relapse and dysregulates endocannabinoid and glutamatergic signaling in the central amygdala

Abstract

Administration of selective serotonin reuptake inhibitors (SSRIs), typically used as antidepressants, induces long-lasting behavioural changes associated with alcohol use disorder (AUD). However, the contribution of SSRI (fluoxetine)-induced alterations in neurobiological processes underlying alcohol relapse such as endocannabinoid and glutamate signaling in the central amygdala (CeA) remains largely unknown. We utilized an integrative approach to study the effects of repeated fluoxetine administration during abstinence on ethanol drinking. Gene expression, biochemical and electrophysiological studies explored the hypothesis that dysregulation in glutamatergic and endocannabinoid mechanisms in the CeA underlie the susceptibility to alcohol relapse. Cessation of daily treatment with fluoxetine (10 mg/kg) during abstinence resulted in a marked increase in ethanol seeking during re-exposure periods. The increase in ethanol self-administration was associated with: a) reductions in levels of the endocannabinoids N-arachidonoylethanolomine and 2-arachidonoylglycerol in the CeA, b) increased amygdalar gene expression of cannabinoid type-1 receptor (CB1), N-acyl phosphatidylethanolamine phospholipase D (Nape-pld), fatty acid amid hydrolase (Faah), c) decreased amygdalar gene expression of ionotropic AMPA (GluA2 and GluA4) and metabotropic (mGlu3) glutamate receptors and d) increased glutamatergic receptor function. Overall our data suggest that the administration of the antidepressant fluoxetine during abstinence dysregulates endocannabinoid signaling and glutamatergic receptor function in the amygdala, facts that likely facilitate alcohol drinking behavior during relapse.

INTRODUCTION

Alcohol use disorder (AUD) remains among the most common substance abuse problems worldwide. Chronic AUD is associated with the dysregulation of physical and motivational processes, facilitating an excitatory withdrawal condition in dependent individuals that may induce severe pathophysiological consequences¹. In addition, AUD is often accompanied by anxiety and/or depression^1,2, and the co-diagnoses of mood disorders is common among abstinent drinkers undergoing treatment³. The co-morbidity of AUD and depression is associated with dysregulated brain monoaminergic systems including the serotoninergic (5-HT) system, and these findings have informed pharmacological strategies targeting the restoration of deficient 5-HT signaling^4-6. Thus, treatments using selective serotonin reuptake inhibitors (SSRI) such as fluoxetine have proven to be effective in reducing clinical symptoms of negative affect in dependent individuals^7,8. For example, SSRIs ameliorate alcohol withdrawal symptoms^4,9,10, and fluoxetine reduces withdrawal-induced locomotor hyperactivity, anxiety and depression⁴. Although it has long been proposed the clinical use of antidepressants in the treatment of people with co-occurring depression and AUD, there are multiple contradictory reports suggesting that SSRIs are not effective in AUD patients¹¹. In fact, SSRI treatment in dependent individuals co-diagnosed with depression displays similar efficacy as with other mixed noradrenaline/serotonin uptake inhibitors (SNRI) such as nortriptyline⁸. Moreover, prolonged SSRI treatment (sertraline) resulted in poorer alcohol-related outcomes (e.g. less heavy drinking days) than placebo in early-onset/high-vulnerability alcoholics¹². In this regard, there are two important problems associated with fluoxetine treatment: first, in patients with comorbid depression and AUD, SSRIs are not more effective than SNRIs; and second, cessation of fluoxetine after withdrawal facilitates escalation of ethanol self-administration, indicating a profound change in motivational responses for alcohol stemming from altered 5-HT transmission during abstinence¹³. This finding was recapitulated in rats using a model of stable drinking to study relapse-like drinking¹³, suggesting that both chronic SSRI treatment and/or its discontinuation may facilitate alcohol relapse in individuals with AUD and co-occurring mood disorders¹⁴. These findings reflect a timely and critical concern because the few compounds currently approved for AUD treatment do not address conflating issues that may arise from co-morbid pathologies¹⁰.

Rats that undergo long-term alcohol drinking with repeated deprivation periods develop a compulsive drinking phenotype that facilitates relapse as a means to avoid negative mood states during withdrawal^15,16. In this regard, an association between acute withdrawal of chronic ethanol consumption, decreases in striatal 5-HT levels and depression-like status have been widely described in a variety of experimental models⁴. In these models, fluoxetine also acts as an antidepressant by increasing synaptic 5-HT levels and reduces other physical signs of ethanol withdrawal such as locomotor hyperactivity, stereotyped behaviors, tremor, wet dog shakes, agitation and audiogenic seizure in alcohol-dependent rats⁴.

Chronic ethanol exposure and withdrawal induces neurochemical-specific neuroadaptations associated with glutamatergic and endocannabinoid signaling in numerous brain regions^17-24 including the amygdalar complex^25-28. Specifically, the central nucleus of the amygdala (CeA) contributes to both anxiety and ethanol-drinking phenotypes^21,22,27,28. AUD and withdrawal result in enhanced glutamate transmission^18,19,24 and affect the expression and function of α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptors (AMPAR) and metabotropic glutamate receptors (mGluR)^17,20,24-26, as well as extracellular levels of glutamate in cortical and limbic areas of the brain^22,23,28. Relapse to alcohol self-administration is also associated with altered glutamatergic receptor function, further supporting the role of dysregulated glutamatergic signaling in AUD and alcoholism^29-31.

Similarly, the endogenous cannabinoid (eCB) system plays a critical role in the behavioral neuroadaptations associated with AUD and depression^32-38. Changes in eCB signaling-related gene expression and interstitial eCB concentrations during AUD have been observed in the amygdala^28,39,40. Ethanol withdrawal is associated with a profound decrease in extracellular concentrations of 2-AG associated with enhanced glutamate release. These alterations disappear after ethanol self-administration is resumed^28,39,40. However, it is currently unknown how alcohol elicits this eCB response and whether serotonergic modulation of amygdala neurotransmission does play a role .

In the present study, we used an animal model of stable drinking based on the alcohol deprivation effect with minor modifications^13,15. This model is considered to have optimal predictive validity in relation to alcohol consumption^13,15. Here, we explored the hypothesis that the cessation of antidepressant treatment enhances susceptibility to alcohol relapse during abstinence. We also examined whether these behaviors were associated with underlying impairments in endocannabinoid signaling that serve as an important modulator of glutamatergic activity in the amygdala. Towards this goal, we employed a multi-disciplinary approach combining behavioral, biochemical, genetic and electrophysiological methods in rats given repeated administration of fluoxetine during abstinence. Specifically, we assessed the effects of fluoxetine cessation during alcohol re-exposure on 1) ethanol self-administration and anxiety-like behavior, 2) mRNA levels of components of the glutamatergic and eCB systems, and 3) glutamatergic neurotransmission in the CeA.

MATERIALS AND METHODS

The animal procedures were approved by The Scripps Research Institute Institutional Animal Care and Use Committee and were consistent with the National Institutes of Health Guide for the Care and Use of Laboratory Animals, as well as the National Research Council, Neuroscience CoGftUoAi, Research B, 2003. The research was also approved by the ethics committee of Complutense University (Spain), following the European Community Council Directive (86/609/EEC).

Animals

We used 48 adult male albino Wistar rats (Charles River); 16 rats (Cohort A) were used at Complutense University (Spain) and 32 rats (Cohort B) for experiments at the Scripps Research Institute (USA). Rats were individually housed and allowed to acclimatize (12-h light/dark cycle, 22 ± 1°C) to the housing and behavioral facilities for 1 week prior to the experiments (Supplementary Information).

Drugs

Alcohol solution (10% ethanol w/v solution) was prepared daily from 99% ethanol. Fluoxetine HCl (Toronto Research Chemicals, Canada) solution was injected (10 mg/kg, i.p.) in a volume of 2 mL/kg (0.9% saline) following previous studies¹³. CGP 55845A and bicuculline were purchased from Sigma (St. Louis, MO, USA).

Operant ethanol self-administration

A total of 32 animals, divided into two batches (Cohort A and Cohort B), were trained on two paradigms of ethanol self-administration, as previously described¹³ for Cohort A and with minor modifications for Cohort B performing a shorter protocol (Supplementary Information). After training, ethanol operant sessions lasted 30 min/day over 5 days/week (Monday to Friday) until steady levels of self-administration were achieved (baseline: 30 presses, ~% SEM).

Re-exposure to ethanol self-administration following abstinence

After reaching baseline, the rats were withdrawn from alcohol self-administration sessions (abstinence) and treated daily with fluoxetine for either 14 days (Cohort A) or 7 days (Cohort B). Twenty-four hours after the last fluoxetine dose, ethanol self-administration sessions were resumed (re-exposure period) and consumption patterns were monitored for 3 weeks (Cohort A) and 2 weeks (Cohort B), respectively. Food intake and body weight were also monitored during the re-exposure period of the Cohort 2 (Supplementary Information). The experimental groups of Cohorts A and B were ethanol-vehicle and ethanol-fluoxetine groups (n=8/group). A control-vehicle (0.9% saline) group and a control-fluoxetine group were also considered in the Cohort B for comparison.

Sample collection

Thirty minutes after the final session of self-administration, rats were euthanized and whole brains were extracted and immediately frozen. The amygdala (Bregma −1.60 mm to −3.14 mm, aprox.) was dissected bilaterally on dry ice using a brain matrix and 2-mm tissue biopsy punches (Zivic). Frozen tissue collections were stored in a −80 °C freezer until RT-qPCR (brains from Cohort A) and mass spectrometry (brains from Cohort B) analyses.

RNA isolation and RT-qPCR analysis

We performed real-time PCR (TaqMan, ThermoFisher Scientific, Waltham, MA, USA) as described³⁹ using specific sets of primer probes from TaqMan® Gene Expression Assays (Table S1). After the reverse transcript reaction from 1 μg of mRNA, a quantitative real-time reverse transcription polymerase chain reaction (qPCR) was performed in a CFX96TM Real-Time PCR Detection System (Bio-Rad, Hercules, CA, USA) and the FAM dye labeled format for the TaqMan® Gene Expression Assays (ThermoFisher Scientific). Values obtained from the amygdala samples were normalized to Actb levels (n=8/group from Cohort A).

Lipid extraction and quantification

Levels of the eCBs N-arachidonoylethanolamine (AEA), 2-arachidonoylglycerol (2-AG), as well as the N-acylethanolamides (NAE) palmitoylethanolamide (PEA) and oleoylethanolamide (OEA) were compared in frozen tissue dissections of the CeA (n=5/group from Cohort B). For these experiments, we used lipid extraction and liquid chromatography tandem mass spectrometry (LC-MS/MS) procedures to determine amygdalar tissue content of the eCBs/NAEs on a triple quadruple mass spectrometer^41,42 (Supplementary Information).

Electrophysiology

Electrophysiological recordings in the medial subdivision of the CeA were performed on ex vivo brain slices collected from 3 animals from each experimental group of Cohort B thirty minutes after the last ethanol self-administration session during the second week of re-exposure (see Supplementary Information for detailed description of slice preparation, whole-cell patch-clamp recordings and data analysis). Briefly, we assessed neuronal membrane properties, excitability and pharmacologically-isolated spontaneous excitatory postsynaptic currents (sEPSCs) (see Supplementary Information).

Data analysis

Data are presented as means ± S.E.M. and “n” indicates the number of animals or recorded cells. For statistical analysis, we used GraphPad Prism 6.0 (GraphPad Software, San Diego, CA, USA) and applied one-sample Student’s t-test, repeated measures analyses of variance (ANOVA; time as the within-subject factor and experimental group as between-subject factor) or two-way ANOVA (ethanol and treatment as factors) followed by Bonferroni-corrected tests (simple effect) where appropriate. p<0.05 indicates statistical significance.

RESULTS

Fluoxetine administration for 14 days increased ethanol self-administration for 3 weeks following abstinence

The Cohort A was initially performed in order to tune up the paradigm in rats treated with fluoxetine. After assessing the influence of fluoxetine on alcohol drinking and relevant signaling systems in the amygdala, we developed the Cohort B. Rats from Cohort A were daily treated with fluoxetine (10 mg/kg) during an ethanol deprivation period of 14 days (Fig. 1A). Then, ethanol self-administration sessions were resumed (re-exposure period) for 3 weeks. Repeated measures ANOVA indicated a main effect of treatment on the daily ethanol self-administration (number of operant lever presses and ethanol intake for 30 min) during re-exposure (Week 1: F_1,70>17.30, p<0.0001; Week 2: F_1,70>21.69, p<0.0001; Week 3:F_1,70>37.22, p<0.0001) (Fig. 1B). No interaction between treatment and daily self-administration was found, suggesting that fluoxetine did not significantly modify the increased ethanol intake across 3 weeks of re-exposure. Simple effect analysis showed that the ethanol-fluoxetine group displayed an increase in ethanol self-administration beginning on the third day of testing and onward. However, we observed an interaction between treatment and weekly self-administration (F_3,56>3.11, p<0.05), indicating that fluoxetine differentially affected the increased ethanol self-administration across 3 weeks of re-exposure (Fig. 1C). Post-hoc analyses confirmed an increase in the weekly average consumption of ethanol in fluoxetine-treated rats during the first (*p<0.05), second (**p<0.01) and third weeks (**p<0.01) of re-exposure as compared to vehicle controls (Fig. 1C), and relative to their respective baseline levels (^#/##p<0.05/0.01) (Fig. 1C).

Figure 1.

Timeline of an extended paradigm of animal model of stable drinking performed in the rats from Cohort A (A). Daily operant lever presses and ethanol self-administration (B) were monitored to assess baseline and ethanol re-exposure for 3 weeks following abstinence and fluoxetine treatment (10 mg/kg/day) for 14 days. Timeline of a shorter paradigm of animal model of stable drinking performed in the rats from Cohort B (D). Daily ethanol self-administration (E) were monitored to assess baseline and ethanol re-exposure for 2 weeks following abstinence and fluoxetine treatment (10 mg/kg/day) for 7 days. Weekly average of active lever presses for alcohol and alcohol consumption are shown in C (Cohort A) and F (Cohort B), respectively (Supplementary Information). Data are represented as mean ± S.E.M. (n = 8). Statistical analysis (simple effects): */**/***p<0.05/0.01/0.001 vs vehicle; ^#/##p<0.05/0.01 vs same group at baseline.

Fluoxetine administration for 7 days increased ethanol self-administration for 2 weeks following abstinence

Rats from Cohort B were daily treated with fluoxetine (10 mg/kg) during an ethanol deprivation period of 7 days (Fig. 1D). Then, ethanol self-administration sessions were resumed (re-exposure period) for 2 weeks. Fluoxetine treatment decreased food intake during ethanol deprivation (Fig. S1) and also ameliorated the anxiogenic effects observed during re-exposure (Fig. S2). Despite the shorter duration of the paradigm, a similar drinking behavioral phenotype was obtained in the rats from Cohort B as compared to those of Cohort A. In Cohort B, we observed an effect of treatment on daily ethanol self-administration during re-exposure (Week 1: F_1,70=15.71, p=0.0002; Week 2: F_1,70=31.15, p<0.0001). Similar to Cohort A, we did not find an interaction between treatment and daily self-administration. Simple effect analysis showed that the ethanol-fluoxetine group displayed an increase in ethanol self-administration beginning on the second day of testing and onward (*/**p<0.05/0.01, Fig. 1E). Repeated measures ANOVA also indicated a treatment effect on weekly ethanol intake and active lever responses following abstinence (F_1,42=6.36, p=0.015; F_1,42=8.84, p=0.004, respectively). No interaction between treatment and weekly self-administration was found. Simple effect analysis showed an increase in the weekly average consumption of ethanol and number of active lever responses in ethanol-fluoxetine rats during the weeks of re-exposure as compared to vehicle controls (*/**p<0.05/0.01), and relative to their respective baseline levels (^#/##p<0.05/0.01) (Fig. 1F).

Fluoxetine decreased gene expression of specific glutamatergic receptor subunits in the amygdala of rats re-exposed to ethanol self-administration for 3 weeks

mRNA levels for GluA2, GluA4 and mGlu3 were lower (*p<0.05; Fig. 2A) in the amygdala of fluoxetine-treated rats re-exposed to ethanol self-administration for 3 weeks after abstinence (Cohort A). No further differences were found.

Figure 2.

Amygdalar mRNA levels of the endocannabinoid (eCB) (A) and glutamatergic (B) signaling systems 30 minutes after the last session of ethanol self-administration of the re-exposure period. Data are represented as mean ± S.E.M. (n = 8). Student’s t test: */***p<0.05/0.001 vs vehicle-ethanol.

Fluoxetine increased gene expression of eCB system-related components in the amygdala of rats re-exposed to ethanol self-administration for 3 weeks

mRNA levels for CB1, Pparα, Nape-pld, Faah and Daglβ were elevated (*/***p<0.05/0.001; Fig. 2B) in the amygdala of fluoxetine-treated rats re-exposed to ethanol self-administration for 3 weeks after abstinence (Cohort A). No further differences were found.

Fluoxetine treatment differentially altered amygdalar eCB/NAE contents in rats with a history of ethanol self-administration

The levels of 2-AG, AEA, OEA and PEA were measured in the amygdala of rats from Cohort B at the end of the second week of ethanol re-exposure. We observed fluoxetine treatment-associated effects on 2-AG and AEA levels (F_1,16=11.88, p=0.0033; F_1,16=7.09, p=0.0017, respectively), and ethanol effects on OEA and PEA levels (F_1,16=10.41, p=0.0053; F_1,16=10.93, p=0.0045, respectively). No interactions between ethanol and fluoxetine treatment were found. Simple effect analyses indicated no change in 2-AG and AEA contents (Figs. 3A,B), but decreased concentrations of OEA and PEA (***p<0.001; Figs. 3C,D) in the amygdala of ethanol self-administering rats. Fluoxetine also decreased 2-AG levels in the amygdala of both control and ethanol-exposed rats relative to vehicle controls (*^/#p<0.05; Fig. 3A). Fluoxetine decreased AEA levels and increased PEA levels in the amygdala of rats exposed to ethanol compared to those of the ethanol-vehicle group (^#p<0.05, Figs. 3B,D). In addition, fluoxetine treatment partially counteracted the decreased OEA levels induced by alcohol self-administration, as no differences between control-vehicle and ethanol-fluoxetine groups were found (Figs. 3C).

Figure 3.

Amygdalar concentrations of 2-AG (A), AEA (B), OEA (C) and PEA (D) 30 minutes after the last session of ethanol self-administration of the re-exposure period in rats from Cohort B. Data are represented as mean ± S.E.M. (n = 5). Statistical analysis (simple effects): */***p<0.05/0.001 vs vehicle controls; ^#p<0.05 vs vehicle-ethanol group.

Fluoxetine and ethanol self-administration prolong sEPSC rise and decay times in the CeA of rats

Based on the changes in glutamatergic subunit expression we sought to determine potential changes in glutamatergic transmission in the CeA. Compared to vehicle controls, ethanol and fluoxetine did not alter excitability of CeA neurons (Fig. S3) but induced significant changes in sEPSC kinetics (Figs. 4A-F) without affecting frequencies (Fig. 4C) and amplitudes (Fig. 4D; Table S2). We observed chronic ethanol effects on sEPSC rise and decay times (F_1,238=10.85, p=0.0011; F_1,238=6.78, p=0.0098, respectively) as well as a fluoxetine treatment effect on sEPSC rise time (F_1,238=4.18, p=0.041). Specifically, both sEPSC rise and decay times were significantly longer in ethanol-fluoxetine rats compared to vehicle control rats (**/***p<0.01/0.001) and fluoxetine control rats (*p<0.05) (Figs. 4E,F). In addition, sEPSC rise times were longer in ethanol rats (*p<0.05) than in vehicle control rats (Fig. 4E). We did not observe significant changes in CeA membrane properties between the groups except a main effect of chronic ethanol treatment on the membrane time constant (Fig. S3). Overall, these findings suggest a significant change in glutamatergic receptor function (longer time of opening and closing of the receptor) with fluoxetine treatment without changing excitability.

Figure 4.

Representative voltage clamp recordings obtained from CeA neurons (A), superimposed averages of sEPSCs (B), and sEPSC characteristics for the indicated animal groups from Cohort B including sEPSC frequencies (C), amplitudes (D), rise (E) and decay (F) times obtained by whole-cell electrophysiology in ex vivo brain slices prepared 30 minutes after the last session of ethanol self-administration during the second re-exposure period (Fig. S3, Table S2 and Supplementary Information) are depicted. Data are represented as mean ± S.E.M. for vehicle controls (n = 64), fluoxetine controls (n = 58), vehicle-ethanol group (n = 58) and fluoxetine-ethanol group (n = 62 cells). Statistical analysis (simple effects): */**/***p<0.05/0.01/0.001 between groups.

DISCUSSION

In the present study, we examined the hypothesis that the cessation of fluoxetine treatment after abstinence facilitates ethanol drinking behavior during re-exposure, and dysregulates canonical anti-stress mechanisms that affect amygdalar excitability. Towards this goal, we used an experimental approach of moderate drinking that models a chronic relapsing condition commonly observed in AUD patients, but which is notably enhanced in subjects who cease SSRI treatment^15,43. Since it is advised to avoid combining alcohol and antidepressants, most patients quitting alcohol discontinue antidepressant treatment when they relapse⁴⁴. Drug treatment in the present study is, therefore, compatible with the actual short abstinence periods observed in humans (in proportion), while also compatible with the chronic nature of effective antidepressant treatment and abrupt retirement after relapse to alcohol. Our animal model of stable drinking did not produce a typical alcohol deprivation effect because the transient increase in ethanol consumption after a period of deprivation was very modest¹⁵. It is well-established that the amount and duration of ethanol exposure (e.g. by ethanol vapour chamber procedure) prior to abstinence is an important determinant of withdrawal severity and alcohol deprivation effect¹⁵. Thus, the modest allostatic change in reward observed in our model appears to be consistent with a history of moderate ethanol consumption. The main finding of the present study is that ethanol deprivation resulted in escalated alcohol consumption when alcohol self-administration is resumed after cessation of fluoxetine treatment. Importantly, this effect is associated with a) increased expression of eCB biosysnthesis/degradation enzymes (Nape-pld, Faah), b) alterations in eCB (2-AG and AEA) and NAE (OEA and PEA) levels, and b) changes in glutamatergic receptor subunit expression and glutamatergic transmission in the amygdala. These data provide compelling evidence for dysregulated mechanisms underlying lower eCB tone and glutamatergic hyperactivity in critical circuits of the amygdala. However, the interpretation of these results may be limited by the lack of a control group in which fluoxetine treatment would continue along relapse to alcohol. So far, the cessation of fluoxetine treatment appears to sensitize these neurobiological mechanisms that likely enhance susceptibility to alcohol relapse. These results challenge the therapeutic efficacy of fluoxetine to treat AUD patients who display comorbid pathologies with other mood disorders, such as depression.

To gain insight into potential mechanisms underlying this paradoxical increase of ethanol consumption, we evaluated changes in gene expression in glutamate and eCB systems. Our results show that the cessation of fluoxetine treatment (10 mg/kg), administered for 7 and 14 days during abstinence, increased ethanol self-administration during re-exposure, an effect resembling that of alcohol deprivation that occurs from 2-3 days after resumption and is maintained for up to 3 weeks. The delayed effect of alcohol deprivation observed after the cessation of fluoxetine treatment in both cohorts is possibly due to: 1) the long half-life (4-6 days) of fluoxetine activity; and 2) the SSRI activity of the main metabollite of fluoxetine norfluoxetine (serproxetine) with a half-life of 4-16 days⁴⁵; which long acting effects may diminish the antidepressant withdrawal. This result is in accordance with previous reports on SSRIs¹³ and those of other antidepressants such as the SNRI venlafaxine and the noradrenaline-dopamine uptake inhibitor atomoxetine on alcohol consumption^13,46, suggesting that noradrenaline may have similar effects. However, anxiety-like symptoms that often co-occur following cessation of ethanol intake were ameliorated by fluoxetine, despite ethanol re-exposure⁴⁷. Importantly, here fluoxetine did not change the anxiety-phenotype of ethanol-naïve rats. This observation together with previous reports^4,10 suggests that fluoxetine differentially acts on neuroadaptive mechanisms involved in the development of physical dependence and the anxious-like phenotype, raising the possibility that the motivational responses to alcohol can be dissociated from the anxious state induced by chronic alcohol exposure. These paradoxical effects could stem from possible long-term, biphasic side-effects of the chronic use of SSRIs affecting different neurocircuits and related mechanisms that contribute to AUD pathology.

AUD contributes to neuroadaptations in amygdalar glutamatergic signaling including compensatory upregulation in expression and function of synaptic proteins, channels and receptors^22,24,48,49. Specifically, studies report enhanced AMPAR-mediated neurotransmission in the amygdala and related brain nuclei^24,50-53, increased protein expression of mGluR1 and mGluR5 in the CeA^23,54, and increased [³⁵S]GTPγS binding elicited by the mGluR2/3 agonist LY379368 in the CeA of alcohol-dependent rats³¹. Moreover, amygdalar mGluR2/3 activation has a functional role in modulating the discriminative stimulus effects of alcohol⁵⁵. In our previous work, AUD and withdrawal increased expression of NMDA subunits in the CeA⁴⁹. Chronic ethanol (0.5 g/kg, i.p. for 13 days) also induced a significant increase in non-NMDA postsynaptic activity in the CeA, suggesting AMPA receptors as a major transmission mechanism required for ethanol-induced reward behaviors^56,57. In this study, we found that fluoxetine in ethanol-exposed rats decreased amygdalar mRNA levels of the ionotropic AMPA receptor subunits GluA2 and GluA4 as well as the metabotropic glutamatergic receptor mGlu3. Although, there is evidence that fluoxetine induces synaptic plasticity^58,59, very little is known about its effects on CeA glutamatergic synaptic transmission. We observed prolonged sEPSC rise and decay times in the CeA of fluoxetine treated ethanol-exposed rats, suggesting increased postsynaptic glutamate receptor-mediated synaptic responses^48,56,57. Notably, under our recording conditions (i.e. physiological magnesium in the bath solution and holding potential of −60 mV) NMDA receptors are mostly inactivated^48,49, and the recorded sEPSCs are predominantly mediated by AMPA receptors^53,60,61. Thus, these differences in baseline sEPSC properties suggest an increased postsynaptic glutamatergic receptor activity that may contribute to the increased ethanol consumption after fluoxetine cessation and that the altered current kinetics may reflect changes in AMPA receptor subunit expression/composition^62,63. Further research is needed to elucidate whether fluoxetine withdrawal induces these potential compensatory mechanisms that contribute to the increased alcohol consumption. The activation of amygdalar serotonin receptor (e.g., 5HT_2C) subtypes by fluoxetine induces negative states⁶⁴ that might be expected to potentiate alcohol drinking. Similarly, fluoxetine administration and abstinence may influence anxiolytic responses in animals displaying elevated levels of alcohol intake. Future research should more conclusively examine the contribution of the 5-HTergic system in potentiating mechanism involved in mood disorders induced by AUD.

To further understand the alterations associated to fluoxetine-induced alcohol deprivation effect, we also studied changes in gene expression of the endocannabinoid system, a lipid transmitter-based system that modulates glutamatergic and GABAergic neurotransmissions⁶⁵. In the present study, fluoxetine cessation increased mRNA levels of CB1, Pparα, the N-acylethanolamine (NAE)-related enzymatic machinery Nape-pld and Faah, and the 2-AG-related biosynthesis enzyme Daglβ in the CeA of ethanol-exposed rats. Furthermore, decreased levels of 2-AG and AEA and increased levels of PEA were found in the amygdala of ethanol-exposed rats treated with fluoxetine. The signaling systems related to both, eCBs (2-AG and AEA) and NAEs (OEA and PEA), are also involved in ethanol-related neuroadaptations and behaviors^28,33,66. In previous reports, CB1 antagonism reduced acquisition and maintenance of alcohol intake, relapse responding, and alcohol-seeking behavior^10,35, whereas enhanced NAE tone and CB1 activation increased ethanol intake in rats with forced consumption of an alcoholic diet⁶⁷. A previus report demonstrated that cannabinoid-induced increases in relapse-like drinking could be mediated by NMDA receptors⁶⁶. Indeed, the cannabinoid receptor agonist WIN 55.212-2 administered during alcohol deprivation decreases CB1 and GluN1 mRNA expression in the amygdala of rats during alcohol re-exposure⁶⁸. These findings contrast other studies describing anxiety-like effects of CB1 receptor blockade⁶⁹ as well as the reversion of stress-induced anxiety and activation of corticolimbic and hypothalamic circuits after an enhancement of the eCB tone^25,70. Indeed, fluoxetine treatment produces an increase in AEA levels in the basolateral amygdala, likely through a decreased activity of FAAH, resulting in an amplification of endocannbinoid-mediated inhibitory transmission and fear extintion⁷¹. We recently reported that deficient eCB signaling in the CeA contributes to AUD-related anxiety-like behavior and excessive alcohol intake and CeA glutamate content²⁸, and to the propensity for co-morbid anxiety and alcohol preference in genetically selected rats⁴². Greater withdrawal-associated deficits in mRNA expression of CB1 and CB2 in rat amygdala were evident following 6, 24 and/or 72 hours of alcohol deprivation^39,40. Moreover, AUD induced by 21 days of exposure to intermittent alcohol vapor inhalation and 12 hours of abstinence reduced mRNA expression of genes that are associated with the eCB enzymatic machinery (Faah, Daglα, Daglβ, Mgll) in the rat CeA²⁸. Because bidirectional effects of stress on brain levels of AEA and 2-AG have been reported⁷², further studies are needed to elucidate whether the reduced eCB tone in the CeA fully explains the decrease in withdrawal-induced anxiogenic-like responses in rats during re-exposure.

In this regard, we should consider the role of other non-cannabinoid NAEs such as OEA and PEA. Previous studies provided covergent evidence using mouse models and human data from GWAS for specific peroxisome proliferator-activated receptors (PPARs) in alcohol consumption and related behaviour^73,74. Our data show that fluoxetine attenuated the alcohol-induced reduction of the amygdalar concentration of PEA and partially reversed the decrease in OEA. These endogenous ligands of the peroxisome proliferator-activated receptor alpha (PPARα) were found to be antidepressants in alcohol-associated depression in rats⁷⁵. Thus, a potential regulation of these NAEs by fluoxetine might underlie the behavioral performance in the elevated plus-maze of ethanol-deprived fluoxetine-treated animals. Partially, this could be related to the antidepressant and anxiolytic-like effects of PEA (5-40 mg/kg) in animal models⁷⁶, an effect comparable to that of fluoxetine (20 mg/kg) in mice using the tail suspension test and the forced swimming test⁷⁷. This result is also interesting because it highlights the hypothesis that treating only anxiety in AUD will not reduce relapse nor will it completely abolish the risk of AUD in patients. A previous study demonstrated that the depressive-like state caused by corticosterone treatment was reversed by exogenous administration of the PPARα ligand PEA for 2 weeks⁷⁸. In the present study, increases in PEA and mRNA levels of Pparα in the CeA induced by fluoxetine cessation can be likely associated with the amelioration of the anxiogenic effect induced by ethanol withdrawal.

In summary, we report that the cessation of fluoxetine treatment, administered during alcohol abstinence, increases alcohol consumption when alcohol self-administration is resumed. This likely alcohol deprivation effect is associated with decreases in endocannabinoid (2-AG and AEA) levels in the amygdala and increased glutamatergic receptor function in the CeA. Collectively, our results argue against the efficacy of fluoxetine to reduce alcohol relapse and challenge the clinical relevance of this antidepressant to treat AUD patients.

Supplementary Material

Figure S1

Figure S1. Daily cumulative food intake during vehicle or fluoxetine treatment (10 mg/kg/day) and ethanol abstinence for 7 days (A) as well as during the first days of the first and second weeks of re-exposure (B) in rats from Cohort B. Body weight gain during the last four days of treatment and first 3 days of reinstatement are shown in C and D respectively (Supplementary Information). Data are represented as mean ± S.E.M. (n = 8). Statistical analysis (simple effects): */**/***p<0.05/0.01/0.001 vs vehicle controls; ^#/##/###p<0.05/0.01/0.001 vs vehicle-ethanol group.

Figure S2

Figure S2. Anxiogenic-like behavior was evaluated by the total entries (activity) into the open and closed arms (A), the ratio of open/close arm entries (B), the ratio of open/closed arm times (C), the entries into the open (D) and closed (E) arms, and time spent in the open (F) and closed (G) arms for 10 minutes in an elevated plus-maze. This test was performed 30 min after the self-administration session during the second week of re-exposure in the rats of Cohort B. Data are represented as mean ± S.E.M. (n = 8). Statistical analysis (simple effects): */**p<0.05/0.01 vs vehicle controls; ^#p<0.05 vs vehicle-ethanol group.

Figure S3

Figure S3. Basal membrane properties and excitability measures showing resistance (A), capacitance (B), membrane time constant (C), voltage sag slope (D), ADP slope (E), rheobase (F), number of spikes at maximal intensity (G), spike threshold (H) and spike amplitude (I) of neurons located in the medial subdivision of the central amygdala (CeA) in the rats from Cohort B by using whole-cell electrophysiology in ex vivo brain slices preparation 30 minutes after the last session of ethanol self-administration during the second re-exposure period (Supplementary Information). Data are represented as mean ± S.E.M. for vehicle controls (n = 14), fluoxetine controls (n = 11), vehicle-ethanol group (n = 30) and fluoxetine-ethanol group (n = 13 cells). Statistical analysis (simple effects): */**p<0.05/0.01 between groups.