Molecular fingerprints in the hippocampus of alcohol seeking during withdrawal

Alcohol use disorder (AUD) is characterized by pathological motivation to consume alcohol and cognitive inflexibility, leading to excessive alcohol seeking and use. Due to limited understanding of the molecular basis of the disease, there are few pharmacological interventions available to combat AUD. In this study, we aimed to investigate the molecular correlates of impaired extinction of alcohol seeking during alcohol withdrawal using a mouse model of AUD implemented in the automated IntelliCage social system. This model enabled us to distinguish between animals exhibiting AUD-prone and AUD-resistant phenotypes, based on the presence of ≥ 2 or < 2 criteria of AUD, respectively. We utilized new generation RNA sequencing to identify genes that were differentially expressed in the hippocampus and amygdala of mice meeting ≥ 2 or < 2 criteria, as these brain regions are implicated in alcohol motivation, seeking, consumption and the cognitive inflexibility characteristic of AUD. To complement the sequencing studies, we conducted ex vivo electrophysiology experiments. Our findings revealed significant dysregulation of the hippocampal genes associated with the actin cytoskeleton and synaptic function, including actin binding molecule cofilin, during alcohol withdrawal in mice meeting ≥ 2 criteria compared to those meeting < 2 criteria. Moreover, this dysregulation was accompanied by impaired synaptic transmission in the molecular layer of the hippocampal dentate gyrus (ML-DG). Additionally, we demonstrated that overexpression of cofilin in the polymorphic layer of the hippocampal dentate gyrus (PoDG) inhibited ML-DG synapses, increased motivation to seek alcohol, impaired extinction of alcohol seeking and increased correlation between AUD behaviors, resembling the phenotype observed in mice meeting ≥ 2 criteria. Overall, our study uncovers a novel mechanism linking increased hippocampal cofilin expression with the AUD phenotype.


INTRODUCTION
Alcohol use disorder (AUD) is a progressive and debilitating psychiatric disease characterized by pathological alcohol craving and motivation to consume alcohol, as well as cognitive rigidity.This condition results in an excessive focus on alcohol procurement and alcohol use in daily routines [1].Although AUD is one of the leading causes of premature deaths globally, pharmacological interventions aiming to control alcohol misuse are limited, with significant negative side effects, and therfore infrequently prescribed and used [2][3][4].To identify a new therapeutic approach to battle AUD, a neurobiology of the disease must be elucidated.So far most of the molecular studies focused on the quantitative aspects of alcohol misuse [5]-several candidate genes and molecular pathways that affect the amounts of consumed alcohol both in humans and animals have been identified [6][7][8][9].Moreover, in recent years an accumulating number of studies focus on the biology of complex alcohol-related behaviors, such as compulsivity [10][11][12][13][14], cognitive inflexibility [15], or choice between alcohol and natural rewards [16][17][18].Still, the molecular processes that affect behavioral hallmarks of AUD beyond alcohol consumption remain poorly understood.To develop a successful prevention and therapeutic control of AUD progression, the neuronal basis of all AUD-related behaviors must be recognized.
Here we focused on the molecular correlates of excessive alcohol seeking induced in the alcohol-predicting context during withdrawal.Such behaviour reflects individual focus on alcohol procurment and use, as well as cognitive inflexibility characteristic for AUD patients [1,19].Toward this end we developed a mouse model of AUD that has been pharmacologicallyvalidated [20][21][22] and proven to have significant translational value [23][24][25].The model is based on four DSM-5 criteria of the disease [1,20].(i) Craving, or a strong desire or urge to use alcohol.We measured motivation to obtain alcohol in a progressive-ratio schedule.(ii) The subjects spents a great deal of time in activities necessary to obtain alcohol.We measured the extinction of alcohol seeking during periods of forced abstinence.(iii) The subject takes alcohol in larger amounts than intended.We measured alcohol intake during alcohol relapse after abstinence.(iv) Unsuccessful efforts to control alcohol use.We measured alcohol seeking induced by alcohol-predicting cues and during signalled periods of alcohol nonavaliability.The model allowed us to distinguish the animals that exhibit consistent AUD-prone phenotype, as they were positive (uppermost 35% of the population) in at least two AUDrelated tests (≥ 2 crit mice), and the AUD-resistant mice that were positive in none or one test (< 2 crit animals).Next, we used the new generation RNA sequencing (RNA-seq) to characterize differentially expressed genes in the hippocampus and amygdala of ≥ 2 crit and < 2 crit animals seeking for alcohol during alcohol withdrawal.We focused on the hippocampus as aberrant hippocampal synaptic plasticity is causally linked with cognitive and motivational aberrations characteristic for AUD, including drug seeking induced by drug-predicting contexts and cues [21,22,[26][27][28][29][30].In particular, the manipulations that ablate adult neurogenesis in DG increase drug consumption and motivation to seek for drugs, as well as invigorate drug seeking induced by associated cues and contexts [29,30].Moreover, the CA1 area and subiculum have been implicated in drug-induced place preference and context-induced alcohol and drug seeking [26][27][28][31][32][33][34].On the other hand, the amygdala has been identified as a key region of the neural circuits implicated in the regulation of incentive salience of alcohol-and drug-associated cues, as well as cue-induced reinstatement of drug seeking [23,24,[35][36][37][38][39][40][41].Furthermore, the amygdala was implicated in regulation of alcohol consumption despite negative consequences and alcohol choice over natural rewards [11,16].The molecular processes that underlie the functions of the hippocampus and amygdala in AUD are still largely unknown.
Our data showed that the variance in the transcriptome between the < 2 crit and ≥ 2 crit drinkers after alcohol withdrawal primarily involves hippocampal genes related to the cytoskeleton and synaptic function, including actin binding molecule, cofilin (Cfl) [42].
Accordingly, ex vivo electrophysiology was used to characterize pre-and post-synaptic changes in the hippocampus of the < 2 crit and ≥ 2 crit mice.Finally, we investigated the role of the hippocampal Cfl in the AUD pathology using the local expression of Cfl delivered by adeno-associated viral vectors.Overall, our study identifies transcriptomic differences between the AUD-prone vs -resistant drinkers during alcohol withdrawal.We also describe a novel mechanism that links Cfl-regulated synaptic plasticity in the hippocampus with AUD phenotype characterized by high motivation to seek for alcohol and impaired extinction of alcohol seeking during withdrawal.

Subjects
Ten-week old female C57BL/6J mice were purchased from the Medical University of Bialystok, Poland.We used only females as they not only show lower levels of aggression when group-housed in the IntelliCages but more importantly drink significantly more alcohol as compared to males 40 .Animals were housed under a 12/12 hr light/dark cycle in standard mouse home cages with ad libitum access to water and food.Experiments were approved by the Animal Protection Act of Poland guidelines and the 1 st Local Ethical Committee in Warsaw, Poland (no.117/2016, 421/2017, 884/2019).All experiments were planned to reduce the number of animals used and to minimize their suffering.

Animal model of AUD in the IntelliCages
After 1 week of acclimatization, the mice were injected subcutaneously (s.c.) with unique microtransponders (11.5 mm length, 2.2 mm diameter; Trovan, ID-100) under brief isoflurane anesthesia.The mice were then allowed to recover for 3 days, and the animals with properly located microtransponders were introduced to the IntelliCage system (NewBehavior AG, Zürich, Switzerland) (https://www.tse-systems.com/service/intellicage/),15 animals per system.The IntelliCage consists of a large standard rat cage (20.5 cm high, 40 cm × 58 cm at the top, 55 cm × 37.5 cm at the base).In each corner, a triangular learning chamber is located with two bottles.To drink, only one mouse can go inside a plastic ring (outer ring: 50 mm diameter; inner ring: 30 mm diameter; 20 mm depth into outer ring) that ends with two 13 mm holes (one on the left, one on the right) that provide access to bottle nozzles.Each visit to the corner, nose-poke at the doors governing access to the bottles, and licks were recorded by the system and ascribed to a particular animal.
The training in the IntelliCage was based on published protocol [20] and composed of the following phases: initiation of alcohol consumption in increasing concentrations, free access to 10% alcohol, motivation test, persistence test, withdrawal, cue relapse and alcohol relapse.The timelines of the experiments are on the figures.Adaptation phase.All mice had free access to all bottles with water in both active corners.All doors were open.After 24 hours, when all mice visited and licked from both corners, the doors were closed.Under a fixed ratio of reinforcement (FR 1), each nose-poke was rewarded by a 5 second access to the bottles with water.
Initiation of alcohol consumption.During the test, 2 corners were active, each with two bottles available.In one corner, the animals had access to water ("water corner"), and in the other ("reward corner") animals had access to ethanol solution (Alcohol group) at increasing concentrations (4, 8, and 12% ethanol changed every 3 days, prepared from 96% ethanol and tap water) or water (Alcohol-naive mice).When alcohol (or water) was available, it was signaled by a green light turned on in the "reward corner" each time a mouse entered the corner.All liquids were available under an FR1 schedule.During free alcohol access phases mice had unlimited access to water in one corner and 4, 8 and 12% ethanol (or water) in the "reward corner" (each concentration tested for 3-5 days).Access to water and alcohol was under FR1.10% alcohol was chosen based on maximal alcohol consumption in g/kg/day during initiation of alcohol consumption.Daily alcohol consumption (g/kg/day) was calculated with the following formula: (number of licks of 12% alcohol per day × lick volume × 0.12 × 1 g/ml) / animal weight).To calculate the average lick volume, water consumption (in μl) was measured for 3 consecutive days.The average volume of one lick was measured as the total volume consumed / number of licks.According to these calculations, an average lick volume was established as 1.94 ± 0.2 μl.In the control groups, tap water was presented in the reward corner through the whole experiment.
Motivation for alcohol tests.During the test, two corners were active and available to animals.
The animals had to perform an increasing number of nosepokes (2, 4, 8, 12, 16, 20, 24, 28, 32, and 36) spaced by less than 1 s during one visit to open the door and be allowed for a 5 s access to the reward bottles.The number of required instrumental responses (nosepokes) increased when an animal performed 10 sets of responses of a given ratio.The tests were terminated when 90% of animals did not change the FR level during the last 24 hours.The FR level reached during the test was used as an index of motivation.Extinction of alcohol seeking followed by cue and alcohol relapse.The extinction periods were signaled as the "no-reward" periods and lasted 7 days.The door to reward was closed and nosepokes to the reward corner were without scheduled consequences.Average daily number of nosepokes performed in the "reward corner" during extinction, and a difference in the average number of nosepokes during extinction vs. the last day before the test, were used as indices of alcohol seeking during withdrawal.Each extinction was followed by a 24-hour cuerelapse.A green cue light (reward-predicting cue) in the reward corner was presented each time a mouse entered the reward corner.However, nose-pokes to the reward doors had no scheduled consequences.Average daily number of nosepokes performed in the "reward corner" during cue relapse, and a difference in the average number of nosepokes during cue relapse vs. the last day of extinction, were used as indices of alcohol seeking during cue relapse.This test was followed by reward relapse when bottles with reward (alcohol or water) were added into the active "reward corner".During the test each nose-poke into the reward door opened the door for 5 s.Amount of reward drank during the first day of relapse (number of licks) was used as an index of relapse.
Persistence in reward seeking tests.Each persistence test lasted 3 days, starting at the beginning of the dark phase and was composed of six, 6-hour long "active periods" (A) altered with 6-hour long "non-active periods" (nA)."Active" periods (A) were signaled by the green cue light in the reward corner.During the "active" periods, nosepokes at all doors opened the door for 5 s (FR 1).The "non-active" periods were signaled by elimination of the green cue light.
During the "non-active" periods, nose-pokes on the reward side were not followed by any scheduled consequences.Number of "reward" nose-pokes performed during the test, as well as the difference of nose-pokes performed during nA and A reward periods, were used as indices of persistence.
Establishment of mouse subpopulations.The addiction index was calculated as previously described [20] and was based on five behaviors: (i) the breakpoint reached during the motivation test, (ii) persistence in alcohol seeking during the persistence test, (iii) alcohol seeking during the cue-induced relapse and (iv) extinction and (v) alcohol consumption during the alcohol relapse.An individual was arbitrarily considered positive for an AUD-like criterion when its score in the test was in the uppermost 35% of the population.The scoring allowed us to divide the mice into groups according to the number of fulfilled AUD-like criteria: "AUDprone" who fulfilled 2 or more criteria (≥ 2 crit); "AUD-resistant" who were positive for one or none of the criteria (< 2 crit).Moreover, since the addiction index may neglect mice performance in some tests, we developed addiction score.To calculate addiction score each of the addiction-like behaviors was normalized and summed up to calculate individual addiction score according to formula: AS = (Vi (individual score)mean(population))/SD(population).This allowed us to distinguish mice which show consistent behavioral patterns towards alcohol.

RNA Sequencing
The hippocampal tissue was quickly dissected on ice from the fresh brain, homogenized and stored in RNAlater solution (Invitrogen, AM7020) at 4°C, for 24 hrs and kept then at -20°C till further use.Total RNA was extracted using the RNeasy Mini kit (Qiagen, 74104) according to the manufacturer's recommendations.RNA concentration, quality and integrity were determined using a Nanodrop 1000 (Thermo Scientific) and a Bioanalyzer (Agilent).RNA libraries for sequencing were prepared using a KAPA Stranded mRNA Sample Preparation Kit (KK8401-07962169001, Kapa Biosystems, Wilmington, MA, USA).The libraries were sequenced after onboard cluster generation using HiSeq Rapid SBS Kit v2 (200 cycles) and HiSeq PE Rapid Cluster Kit v2 (Illumina) on a HiSeq 1500 (Illumina).The row data are available at GEO NCBI (GSE221166).Power (defined as the expected proportion of identified differentially expressed genes among all the truly differentially expressed genes, given at least one gene is truly differentially expressed in the data) was modeled using the ssizeRNA R package [73] assuming 80% of non-differential genes, significance threshold FDR = 5%, dispersion = 5% and fold change following normal distribution.Raw sequencing data were processed by trimmomatic [74], adapter contamination and bad quality reads or read fragments were removed.Processed reads were mapped to mm10 reference genome by STAR algorithm [75].Subsequently Picard MarkDuplicates algorithm was used to mark optical duplicates.Featurecounts from the subread R package [76] was used to count the number of fragments assigned to genes.Data normalization and statistical analysis was done by the NOIseq R package [77].RNA-seq quality control analysis was performed using the RSeQC package and STAR (log files).Sequencing fragments distribution, sequencing quality and percent of unique mapping fragments was evaluated.Significantly deregulated genes (FDRadjusted p-value < 0.05) obtained from the RNA sequencing by NOISeq [77] analysis were used as input for the pathway analysis.By ShinyGO 0.76 [46] functional enrichment analysis was performed to find affected pathways in the gene ontology cellular component (GO:CC), molecular function (GO:MF) platforms and in the Kyoto Encyclopedia of Genes and Genomes database (KEGG) [47].
Extracellular field potential recordings were conducted in a submerged chamber perfused with recording ACSF in RT.The synaptic potentials were evoked with a custom built stimulus isolator using a concentric bipolar electrode (FHC, 30200) placed in the perforant path.The stimulating pulses were delivered at 0.033 Hz and the pulse duration was 0.3 ms.
The amplitude and slope of fEPSPs were measured using AxoGraph 1.7.4 software (Axon Instruments, U.S.A).

Stereotaxic surgery
Mice were anaesthetized with isoflurane (5% for induction, 1.5-2.0%for maintenance of general anesthesia), fixed in the stereotactic frame (51503, Stoelting, Wood Dale, IL, USA), and their body temperature was maintained using a heating pad.Stereotactic injections were performed bilaterally into the dDG region of the hippocampus using coordinates from the Bregma: ML, ±1.0 mm; AP, -2.0 mm; DV, -2.0 mm according to (Paxinos and Franklin, 2001).0.5 µl of virus solution was microinjected through a beveled 26 gauge metal needle, attached to a 10 µl microsyringe (SGE010RNS, WPI, USA) connected to a microsyringe pump (UMP3, WPI, Sarasota, USA) and its controller (Micro4, WPI, Sarasota, USA), at a rate 0.1 µl/min.The needle was left in place for an additional 10 minutes following injection to prevent leakage of the vector.Mice were injected with adeno-associated viral vectors (AAV2.1)encoding wild-type cofilin protein under CaMKII promoter fused with HA (AAV2.1:CaMKII_cofilin_HA)(Cfl) (viral titer: 7.29 × 10 8 gc/µl), or a control eGFP-coding vector (AAV2.1_CaMKII_eGFP)(eGFP) (viral titer: 6.9 × 10 8 gc/µl).The viruses were prepared by the Laboratory of Animal Models at Nencki Institute of Experimental Biology, Polish Academy of Sciences.After the surgery, animals were given 14 days to recover before training in the IntelliCages.After training, the animals were perfused with 4% PFA in PBS and brain sections from the dorsal hippocampus were immunostained to detect Cfl and HA, and imaged with Zeiss Spinning Disc confocal microscope (magnification: 10x) to assess the extent of the viral expression and proteins level (ImageJ software).

Western blot
Mice were decapitated under isoflurane anesthesia, hippocampi were isolated and sliced into 1 mm-thick slices.dDG was cut from the slices with a razor blade.The tissue was homogenized in ice-cold lysis buffer (25mM HEPES, pH 7.4; 500mM NaCl; 2mM EDTA; 20mM NaF; 1× protease inhibitor cocktail tablet; 0.1% (v/v) Nonidet P-40).After sonication and spinning (20000xg 4°C) the supernatant was stored at -80 °C until further analysis.For Western-blot equal amounts of total protein from each sample were mixed with a Laemmli buffer containing DTT (50 mM) and left to denature at 70°C for 10 minutes.The mixture was loaded on TGX precast gel wells, that contain trihalo compounds allowing stain-free visualization of total proteins (Bio-Rad #4568083), and ran until the loading buffer reached the bottom of the gel.The analyzed protein levels were normalized to the total protein levels.Next, the proteins were transferred to membranes.The membranes were blocked by 5% or 10% (depending on the antibody) milk diluted in TBST (Tris-buffered saline with Tween 20), and incubated with the primary antibody (cofilin, Cell Signaling, cat.5175S, 1:3000; HA-Tag, Santa Cruz, sc-7392, 1:1000; GAPDH, Merck Millipor, No. AB2302, 1:2000) for 12 hours.After washing in TBST the membranes were incubated in a secondary antibody with HRP (antirabbit, Vector pI-1000, 1:5000; anti-mouse, Santa Cruz, sc-2005, 1:5000) and washed again.

Statistical analysis
The sample sizes of the experimental groups, and details of the statistics, are placed on the graphs or in the legends.The data with normal distribution and equal variance are presented as the mean with SEM and were analyzed with Student's t-test, one-way, two-way or repeated measure two-way ANOVA.Post hoc Šídák's and Tukey's tests for multiple comparisons were used.Alternatively, multiple repeated Mann-Whitney's tests corrected with Holm-Sidak method for multiple comparisons were used.Correlations were analyzed using Pearson correlation.The difference between the experimental groups was considered significant at p < 0.05.All statistical analyses were performed using GraphPad Prism 9.3.1 Software.

Characteristics of AUD-prone and -resistant mice
To identify AUD-prone and -resistant mice we used a mouse model of the disease in the social context of IntelliCages [20].C57BL/6J mice (n = 58) went through a long-term training consisting of the introduction of alcohol (4, 8, 12%, days 1-12) and alcohol free access period (FA, 10%, days 13-47).During the 4-12% and FA mice had unlimited access to alcohol in the reward corner.Alcohol availability was signaled by the cue light presented each time a mouse entered the corner and each nosepoke in the corner gave access to alcohol for 5 seconds (fixed ratio 1, FR1).Next, we assessed behaviors that resemble DSM-5 criteria for AUD [1,20]: high motivation to drink alcohol was measured as a number of nose-pokes in the reward corner performed in a progressive-ratio schedule of reinforcement test when mice had to make an increasing number of nosepokes (FR2, 4, 8, 12, 16, 20, 24, 28…) in order to get access to alcohol for 5 seconds (Motivation); excessive alcohol seeking was measured as number of nosepokes in the alcohol corner when the corner was inactive and nosepokes had no programmed consequences (Extinction); reactivity to alcohol-predicting cues was assessed as nosepokes in the alcohol corner during presentation of the cue light when alcohol was not available (Cue relapse) [43]; lack of control over alcohol consumption was assessed as alcohol consumption (g/kg/day) when the alcohol corner was activated after withdrawal (Alcohol relapse); while lack of control over alcohol seeking was measured as the change of nosepokes number to the alcohol corner during the non-active vs. active phases of the test (Persistence) (Figure 1A).AUD score was calculated as a sum of normalized scores from all AUD tests, and AUD index as a sum of positive results (top 35%) in all tests [20,44].Mice were distinguished based on the DSM-5 criteria [1,20]: AUD-prone drinkers were positive in at least two AUD tests (AUD Index ≥ 2 crit), AUD-resistant drinkers were positive for none or one criterion (AUD Index < 2 crit).Overall, 38% of the mice were indicated as AUD-prone drinkers (Figure 1B).
Retrospective analysis of the mice behavior showed that the ≥ 2 crit group had higher AUD score as well as scores in all AUD tests (Supplementary Figure 1) as compared to < 2 crit animals.AUD index correlated with AUD score (Figure 1C) and all AUD-related behaviors (Figure 1D).Moreover, when < 2 crit (n = 36) and ≥ 2 crit mice (n = 22) were analyzed separately, we found that for the ≥ 2 crit group all AUD behaviors predicted AUD score, they generally correlated with each other and heavily loaded on the main factor (PC1) in the principal components analysis (PCA) suggesting that these behaviors are measures of a single factor that may reflect compulsive alcohol use (Figure 1E, Supplementary Tables 1-2).In particular, extinction of alcohol seeking during withdrawal and persistence (Spearman r = 0.70 for both) were the best predictors of AUD-prone phenotype.On the other hand, in the < 2 crit group AUD behaviors did not correlate with AUD score, only cue relapse correlated with motivation (Figure 1E, left) and AUD behaviors loaded differently on principal components in PCA (Figure 1E, right; Supplementary Tables 3-4) indicating that they are driven by different factors.Hence, tight correlation between AUD behaviors characterises only AUD-prone mice, a fraction of mice drinking alcohol.Finally, despite the profound difference in addiction-like behavior scores between < 2 crit and ≥ 2 crit mice, AUD index did not predict alcohol consumption during FA or mice activity (total NPs) (Figure 1F-G).
Thus, the AUD model allowed for the identification of the mice that demonstrate a consistent AUD-like phenotype and AUD-resistant drinkers.As extinction of alcohol seeking during alcohol withdrawal was one of the best predictors of AUD phenotype in our model, in the following step we focused on transcriptomic differences between the < 2 crit and ≥ 2 crit mice following withdrawal.

Differentially expressed genes in the hippocampus of ≥ 2 crit and < 2 crit mice during extinction of alcohol seeking.
We hypothesized that transcriptomic differences drive the variance between the < 2 and ≥ 2 crit mice in extinction of alcohol seeking during alcohol withdrawal.To test this hypothesis 16 mice were trained to drink alcohol in the IntelliCages.For the molecular analysis ten individuals with the highest (≥ 2 crit, n = 5) and lowest AUD index (< 2 crit mice, n = 5) were selected (Figure 2A).They differed in AUD score as well as all AUD behaviors including extinction of alcohol seeking during withdrawal (Figure 2B and Supplementary Figure 2).The hippocampus and amygdala tissue was collected immediately after the second alcohol extinction test (day 90, Figure 2A), total RNA was extracted and used for a new generation high-throughput RNA sequencing (RNA-seq).We focused on these brain regions as the hippocampus has been implicated in context-induced alcohol and drug seeking during withdrawal [26][27][28][31][32][33][34], while the amygdala, use here as a control region, was implicated in alcohol consumption despite negative consequences, alcohol choice over natural rewards, alcohol motivation as well as cue relapse, rather then alcohol seeing in alcohol-predicting contexts [11,16,23,24,35,45].
The hippocampal transcriptome analysis of ≥ 2 crit and < 2 crit mice yielded 1107 differentially expressed genes (DEGs); 412 genes were upregulated and 695 downregulated in ≥ 2 crit as compared to < 2 crit animals (Figure 2C).On the other hand, we found only 4 DEGs in the amygdala; 1 gene (Snora2b) was upregulated and 3 transcripts (Prl, Gm25014 and Gm23711) downregulated in the ≥ 2 crit as compared to < 2 crit drinkers.Therefore, in the following steps of the analysis we focused on the hippocapal transcriptom.
These results indicate that the reorganization of the cytoskeleton and changes in synaptic function in the hippocampus may contribute to differences in extinction of alcohol seeking between the < 2 and ≥ 2 crit mice.
Cfl is upregulated in the hippocampus of the ≥ 2 crit mice during extinction of alcohol seeking.
Among the top DEGs associated with the cytoskeleton function, we found upregulation of cofilin (Cfl) transcripts in the ≥ 2 crit mice as compared to < 2 crit animals (Figure 2F-H).
Previous studies have shown that Cfl severs actin filaments, leading to increased actin cytoskeletal dynamics [48].This mechanism not only regulates postsynaptic function but also synaptic vesicle mobilization and exocytosis [49][50][51].Additionally, active Cfl can bind to Factin and form stable actin rods, which can impede axonal trafficking [52].Since RNA-seq analysis suggests that these processes may be dysregulated in the hippocampus of ≥ 2 crit mice during alcohol withdrawal (Figure 2D-E) we chose to focus on Cfl in the subsequent steps of our study.
To verify distinctive expression of Cfl in the ≥ 2 crit and < 2 crit groups during extinction test, mice were trained to drink alcohol in the IntelliCages.The ≥ 2 crit and < 2 crit animals were identified and they were sacrificed after 7-day alcohol withdrawal (extinction, day 90) (Supplementary Figure 3).The brains were sliced and immunostained with specific antibodies.We analyzed Cfl levels on the brain slices as integrated mean gray values of the microphotographs.Significant upregulation of Cfl in the ≥ 2 crit, as compared to < 2 crit mice, was observed in the dentate gyrus of the hippocampus (DG), but not CA1 area, basolateral amygdala (BLA), central nucleus of the amygdala (CeA), nucleus accumbens (NAc) and caudate putamen (CaPu) (Supplementary Figure 3D-E).
To confirm this observation the experiment was repeated with a new cohort of mice.The ≥ 2 crit and < 2 crit mice were sacrificed during free alcohol access period (alcohol, day 83) or after 7-day extinction test (extinction, day 90) (Supplementary Figure 4, Figure 3A and B).We also used alcohol-naive mice as a control.We focused on the analysis of Cfl in the DG layers: the granule cell layer (GCL), polymorphic layer of dDG (PoDG), as well as the molecular layer of GC dendrites (ML) (Figure 3C).Overall, Cfl levels were increased in all mice drinking alcohol as compared to alcohol-naive animals.Furthermore, the levels of Cfl were increased in the ML and PoDG after extinction test in the ≥ 2 crit mice, as compared to the < 2 crit extinction animals and the ≥ 2 crit alcohol group (Figure 3D).
As RNA-seq analysis suggested deregulation of the synaptic proteins and proteins related to synaptic vesicle cycle in the ≥ 2 mice (Figure 2E), we also analyzed colocalization of Cfl with synaptotagmin 1 (Syt1) (Ca 2+ sensor in the membrane of the pre-synaptic axon terminal involved in both synaptic vesicle docking and fusion with the presynaptic membrane; upregulated in RNA-seq data, Supplementary Table 1) and PSD-95/Dlg4 (post-synaptic scaffold protein; upregulated in RNA-seq data, Supplementary Table 1) (Figure 3E-F).
Overall, synaptic Cfl levels (Cfl colocalizing with Syt1 and PSD-95) were increased in mice drinking alcohol as compared to alcohol-naive mice.We also observed increased levels of Cfl colocalized with Syt1 in the ML in the ≥ 2 crit mice after extinction, as compared to the ≥ 2 crit alcohol group and the < 2 crit extinction animals (Figure 3E).There was no significant effect of the training and AUD on the levels of Cfl co-localised with PSD-95 (Figure 3F).Altogether, our analysis shows that alcohol training upregulates Cfl in DG.Furthermore, extinction of alcohol seeking upregulates Cfl in PoDG and ML in the ≥ 2 crit mice as compared to the < 2 group; and the upregulated Cfl in ML colocalized with the pre-rather than post-synaptic compartments.

Extinction of alcohol seeking impairs ML synaptic function in ≥ 2 crit mice.
To test whether extinction of alcohol seeking induces synaptic changes in the ML of the ≥ 2 crit mice, we trained a new cohort of animals to drink alcohol in the IntelliCages.As previously, ≥ 2 and < 2 crit mice were identified and sacrificed during FA (alcohol) or after extinction test (Figure 4A and B, Supplementary Figure 5).Alcohol-naive mice were used as a control.Field excitatory postsynaptic potentials (fEPSPs) were recorded to evaluate synaptic function by measuring input-output and paired-pulse ratio (PPR) in the ML synapses of acute hippocampal slices when axons terminating in the ML were stimulated by monotonically increasing stimuli (Figure 4C).
The PPR slope was significantly decreased in the alcohol mice as compared to alcohol-naive animals, and increased in the ≥ 2 crit mice sacrificed after extinction test, as compared to the < 2 crit extinction animals and the ≥ 2 crit alcohol mice (Figure 4D-E).This data indicates higher presynaptic release probability in the alcohol mice, as compared to the alcohol-naives, and lower in the ≥ 2 crit mice after extinction as compared to other alcohol groups.Moreover, the input-output curves for the slope of fEPSP were increased in the alcohol mice as compared to alcohol-naives, and decreased in the ≥ 2 crit extinction mice compared to the ≥ 2 crit alcohol group and < 2 crit extinction mice (Figure 4G-I), suggesting more synaptic transmission in the alcohol groups as compared to alcohol-naive animals, and less synaptic transmission in the ≥ 2 crit extinction mice as compared to other alcohol groups.We also observed lower fiber volley (FV) responses in the ≥ 2 crit mice as compared to the alcoholnaive and < 2 crit animals indicating less activated axons in the ≥ 2 crit mice.
Overall, higher PPR and lower input-output in the alcohol-naive mice as compared to alcohol-trained animals indicate increased synaptic function after alcohol training.However, higher PPR and lower input-output in the ≥ 2 crit mice after extinction test, as compared to the < 2 crit extinction group and the ≥ 2 crit alcohol animals, indicate weakening of the ML synapses of ≥ 2 crit mice during alcohol withdrawal.This process is likely driven by presynaptic changes.

Overexpression of cofilin in PoDG weakens contralateral synapses in the ML of DG.
The main inputs to the ML originate from the entorhinal cortex and contralateral PoDG (Figure 4C).As we observed increased Cfl levels in PoDG of the ≥ 2 crit mice after extinction (Figure 3D) we hypothesized that weakened synaptic transmission in the ML of the ≥ 2 crit mice after extinction test, as compared to the ≥ 2 crit mice before extinction, is driven by the increase of pre-synaptic Cfl levels in PoDG.To address this hypothesis, alcohol-naive mice were unilaterally injected into DG with adeno-associated viral vectors (AAV2.1)expressing cofilin with hemagglutinin tag (HA) (Cfl) under CaMKII promoter [53].This resulted in Cfl overexpression in the PoDG cells (Cfl_PoDG) ipsilaterally to the injection, and in the PoDG axons in the ML (Cfl_ML) contralaterally to the injection (Figure 5A-D).The AAV2.1 encoding eGFP under CaMKII promoter was used as a control.The fEPSPs were recorded to measure input-output and PPR in the ML while axons terminating in the ML were stimulated (Figure 5D).
The PPR, analyzed as slope of fEPSP, was significantly increased in the Cfl_ML slices, as compared to the eGFP and the Cfl_PoDG sections (Figure 5E).The fEPSP slope of the input-output test was significantly decreased in the Cfl_ML slices compared to the eGFP and Cfl_PoDG slices.We did not observe the difference in the ML synaptic strength in the Cfl_PoDG slices compared to the eGFP sections (Figure 5F).We also did not observe any effect of the virus on FV amplitude (Figure 5F).Thus, overexpression of Cfl in the PoDG cells decreased the probability of synaptic release and synaptic transmission in the contralateral PoDG→ML synapses.This indicates that increased expression of Cfl in the PoDG→ML synapses is a plausible mechanism that decreases ML synaptic function in the ≥ 2 crit mice during alcohol withdrawal.

Overexpression of cofilin in PoDG impairs extinction of alcohol seeking and increases alcohol motivation.
To test whether PoDG Cfl affects AUD-related behaviors, mice were bilaterally injected with Cfl (n=13) or the control eGFP virus (n=12).Two weeks after the surgery the animals started long-term alcohol training in the IntelliCages (Figure 6A).Post-training analysis of the hippocampal sections showed that PoDG cells expressing Cfl had higher levels of Cfl and Factin as compared to the non-transduced cells analyzed in the same animals [Cfl(-)].Thus Cfl affected the actin cytoskeleton (Figure 6B-C).
Overexpression of Cfl in PoDG increased motivation for alcohol and impaired extinction of alcohol seeking during withdrawal.However, it had no effect on alcohol consumption during alcohol relapse, alcohol seeking during cue relapse and persistence in alcohol seeking (Figure 6D).Furthermore, PoDG Cfl had no effect on mice activity in the water corner, total activity and liquid consumption as well as alcohol seeking and consumption during 4-12% and alcohol periods (Supplementary Figure 6).Furthermore, we observed that the AUD behaviors of eGFP mice, as for < 2 crit animals, did not correlate with each other and loaded on different components in PCA (Figure 6E, Supplementary Table 8-9) (possibly due to the fact that in the general population there are significantly more < 2 crit than ≥ 2 crit mice).
On the other hand, in the Cfl group, we observed a significant correlation between motivation and cue relapse, as well as extinction and alcohol relapse (Figure 6F).Moreover, persistence, alcohol relapse, extinction and motivation similarly loaded on PC1 (r = 0.57 to 0.88) (Figure 6F, right; Supplementary Table [10][11] suggesting that they measure one factor that may reflect compulsive alcohol seeking.Thus our data indicate that increased Cfl expression in PoDG resulted in specific enhancement of alcohol seeking during extinction test that possibly resulted from the increased motivation for alcohol, and increased correlation between AUD behaviours that resembled the phenotype of the ≥ 2 crit mice.

DISCUSSION
We employed an extensive mouse model of AUD in the social context of IntelliCages to identify AUD-prone and -resistant animals -mice positive for ≥ 2 crit and < 2 crit of AUD, respectively.Next, using RNA-seq, we characterized differences between the hippocampal and amygdala transcriptomes of the ≥ 2 crit and < 2 crit animals during alcohol withdrawal.
The differential expression of the hippocampal genes related to the reorganization of the actin cytoskeleton and synaptic vesicles cycle (e.g.Cfl1) significantly contributed to the distinction between the phenotypes.We also observed decreased function of the ML-DG synapses in the ≥ 2 crit drinkers during withdrawal, and such changes were not observed in the < 2 crit animals.Overexpression of Cfl in PoDG mimicked some aspects of the ≥ 2 crit phenotype including impaired function of the ML synapses, increased alcohol motivation, impaired extinction of alcohol seeking during withdrawal, and increased correlation between AUDrelated behaviors.

RNA-seq enables discovery of novel molecular mechanisms of psychiatric disorders.
In the context of AUD, massive transcriptomic analyses were conducted so far using either the brain tissue of AUD patients [54][55][56] or the animals with the alcohol consumption history [57][58][59].However, these approaches have important limitations.By analyzing human tissue, one cannot distinguish between the transcripts which contribute to the development of AUD and those that are altered only at the advanced stages of the disease.On the other hand, the transcriptomic analyses in animal models commonly used animals exposed to alcohol and alcohol-naive controls, without AUD diagnosis of the tested individuals.Hence, such comparisons did not allow for the distinction of the transcripts specific to alcohol exposure from those involved in AUD progression.Here, we analyzed for the first time the transcriptomic differences between the AUD-prone and AUD-resistant animals, all with free access to alcohol.These animals significantly differ in all tested AUD behaviours, but not alcohol consumption and overall activity.We found over 1000 transcripts differentially expressed in the hippocampus and only 4 in the amygdala after alcohol withdrawal.In particular, we observed differences in the hippocampal transcripts related to the cytoskeleton rearrangement, synapses, and synaptic vesicles cycle.Low number of DEGs in the amygdala during alcohol withdrawal is not surprising as this brain region was previously linked with regulation of alcohol consumption and response to alcohol-predicting cues rather than alcohol seeking in alcohol contexts [11,16,23,24,35,45].
Rearrangement of the actin cytoskeleton has been linked with AUD [60].However, the studies in animal models associated actin binding proteins and actin cytoskeleton mostly with sensitivity to sedative effects of ethanol [61][62][63] and ethanol consumption [64].To our knowledge, only one study so far linked actin-binding protein, Prosapip1, with alcohol seeking and reward [65].Thus, the role of actin cytoskeleton in the regulation of AUD-related behaviors beyond alcohol consumption remains poorly understood.Here we found that the differential expression of the hippocampal transcripts related to actin cytoskeleton distinguished the ≥ 2 crit and < 2 crit mice during alcohol withdrawal.In particular, we show that Cfl (a key molecule regulating actin cytoskeleton and synaptic physiology [66]) expression is increased during alcohol withdrawal in the DG of the ≥ 2 crit mice.Furthermore, we could replicate some features of the ≥ 2 crit phenotype by the local overexpression of Cfl in the PoDG.
Overexpression of Cfl increased alcohol motivation, impaired extinction of alcohol seeking during withdrawal and increased correlation between AUD-related behaviors.Previously, CFL1 has been implicated in the development of neurodegenerative diseases (Alzheimer's disease and Huntington's disease) [67], neuronal migration disorders (lissencephaly, epilepsy, and schizophrenia), neural tube closure defects [68] and memory consolidation during sleep [53].Mutations in CFL1 have been associated with impaired neural crest cell migration and neural tube closure defects [69].Here, we extend these findings by demonstrating the role of Cfl in the regulation of the core symptoms for the AUD diagnosis.
Actin-binding molecules were linked with alcohol consumption by their effects on actin cytoskeleton stabilization.For example, local deletion of Prosapip1 (actin regulatory protein) in the nucleus accumbens results in decreased F-actin levels in the animals treated with alcohol, compared to the alcohol-treated control mice, as well as decreased alcohol consumption [65].On the other hand, mice lacking Eps8 show increased ethanol consumption, while Eps8 null neurons are resistant to the actin-remodeling activity of NMDA receptors and ethanol [62].It remains unclear how actin-binding proteins regulate AUD-related behaviors beyond alcohol consumption.Here, we observed that overexpression of Cfl in PoDG increased PPR in the contralateral ML synapses and decreased fEPSP slope in input-output test, indicating impaired function of the ML synapses that is likely driven by lower presynaptic release probability.Similar physiological changes were observed in the ≥ 2 crit mice after extinction test, as compared to the < 2 mice sacrificed at the same time point.Thus increased levels of Cfl in the ML of ≥ 2 crit mice is likely the mechanism of synaptic weakening observed in these mice during withdrawal.In its unphosphorylated, active state, cofilin severs actin filaments and increases actin cytoskeletal dynamics [48].By this mechanism Cfl regulates both post-synaptic function and synaptic vesicle mobilization and exocytosis [49][50][51].
Moreover, active Cfl may also bind to F-actin and form stable actin rods that block axonal trafficking [52].Thus, the impaired function of the ML synapses after expression of Cfl in PoDG of the alcohol-naive mice, and in the ≥ 2 crit mice during alcohol withdrawal, may be driven by deregulation of synaptic vesicles exocytosis or/and generation of actin rods that impair axonal trafficking.The role of pre-synaptic compartments in the hippocampus in the regulation of extinction of alcohol seeking is also supported by our RNA-seq data showing differential expression of the genes related to axonal projections and synaptic vesicle cycle between the ≥ 2 crit and < 2 crit mice.Alternatively, Cfl may impair DG circuit by inhibiting PoDG postsynaptic compartments.However, this hypothesis is less likely as we did not observe significant changes in the levels of Cfl colocalization with PSD-95 in the ≥ 2 crit mice after extinction.
Our findings add up to previous studies showing that impaired synaptic transmission in DG drives AUD-related behaviors [21,22].In particular, our former studies found that the frequency of silent synapses (that lack functional AMPA receptors) generated during cue relapse in ML positively correlates with AUD index, while downregulation of PSD-95 in the granule cells in DG drives excessive alcohol seeking during cue relapse [21].Here we extend these findings demonstrating the role of pre-synaptic compartments in DG in the regulation of alcohol motivation and extinction of alcohol seeking during withdrawal.Our data are in agreement with former studies showing the role of DG in drug motivation [29,30] as well as extinction of memories and learning about contingencies in the context [70][71][72].Interestingly, our data also indicate that Cfl in PoDG increases overall correlation between AUD-related behaviors, the phenomenon also observed in the ≥ 2 crit vs < 2 crit mice, AUD patients vs healthy individuals, or patients with mild vs severe AUD diagnosis.In particular, by manipulating Cfl levels in PoDG we observed increased correlations between persistence, extinction and alcohol relapse.These behaviors loaded on one major component in PCA suggesting that they measure one factor that may reflect loss of control over alcohol use, and have a common neurobiological root related to the function of Cfl in DG.
Overall, by employing a multidimensional AUD model in conjunction with genomic, electrophysiological, and biochemical analyses, we have identified a novel molecular mechanism that drives increased alcohol motivation and impairs the extinction of alcoholseeking behavior during withdrawal.This mechanism is specific to individuals prone to AUD behaviours and involves the upregulation of Cfl at DG pre-synapses, leading to a weakening of synaptic transmission during alcohol withdrawal.Collectively, these findings may pave the way for novel therapeutic strategies aimed at enhancing synaptic transmission in patients diagnosed with AUD.

Supplementary Files
This is a list of supplementary les associated with this preprint.Click to download. RNAseqandalcohol2023supplement.pdf

1994; 14 :
5325-5337.79.Rivero-Gutiérrez B, Anzola A, Martínez-Augustin O, de Medina FS.Stain-free detection as loading control alternative to Ponceau and housekeeping protein immunodetection in Western blotting.Anal Biochem.2014;467:1-3.(D) Spearman correlations between AUD Index (AI) and AUD-related behaviors.Each dot on the graphs represents one animal.Linear regression lines ± 95% confidence intervals are shown; Spearman correlation (r) and ANCOVA results are given for raw data.(E-F) Spearman correlations (r) between AUD score (AS) and AUD-related behaviors and principal component analysis (PCA) of AUD behaviors for (E) < 2 crit and (F) ≥ 2 crit mice.Each dot on the PCA graphs represents one behavioral measure [motivation to drink alcohol (M), extinction of alcohol seeking during withdrawal (E) and cue relapse (CR), alcohol drinking during alcohol relapse (AR), and alcohol seeking during a persistence test (P)].(G-H) Spearman correlations between AUD Index, (G) mice activity and (H) alcohol consumption during 4-12% and FA phases.Each dot on the graphs represents one animal.Linear regression lines ± 95% confidence intervals are shown; Spearman correlation (r) and ANCOVA results are given for raw data.

Figure 2 .
Figure 2. Transcripts related to cytoskeleton and synaptic function are differentially expressed in the hippocampus of the < 2 crit and ≥ 2 crit drinkers during extinction of alcohol seeking.(A-B) Experimental timeline.Mice were trained to drink alcohol in the IntelliCages (n = 16), classified as < 2 crit (n = 5) and ≥ 2 crit drinkers (n = 5) and sacrificed after a second alcohol extinction test (day 90) (AUD score: t(8) = 3.78, p = 0.003).Hippocampus and amygdala tissue was dissected from fresh brains for RNA-seq analysis.(C) A volcano plot illustrating DEGs in the hippocampus of < 2 crit and ≥ 2 crit drinkers.Only genes with FDR < 0.05 are shown.(D)Gene ontology analysis of molecular function (GO:MF) and cellular components (GO:CC) based on the genes deregulated (both up-and downregulated) in the hippocampus of < 2 crit vs. ≥ 2 crit drinkers (FDR cutoff 0.05).The network shows pathways in nodes.Two nodes are connected if they share 20% or more genes.Darker nodes are more significantly enriched gene sets while bigger nodes represent larger gene sets.Thicker connection between nodes represents more overlapped genes.(E) KEGG pathway analysis of the deregulated genes (both up-and downregulated) in the hippocampus of < 2 crit vs. ≥ 2 crit drinkers (FDR cutoff 0.05).(F)A volcano plot illustrating DEGs from Cytoskeletal protein binding,Tubulin binding and Actin binding nodes (GO:MF) in the hippocampus of < 2 crit vs. ≥ 2 crit drinkers.(G)Regulation of actin cytoskeleton pathway (KEGG) with indicated DEGs in the hippocampus of < 2 crit and ≥ 2 crit drinkers.(H)Cofilin (Cfl) mRNA levels in the hippocampus of < 2 crit and ≥ 2 crit drinkers (t(8)=2.49,p = 0.038).

Figure 4 .
Figure 4. Extinction of alcohol seeking impairs synaptic transmission in the ML of the ≥ 2 crit mice.

Figures Figure 1 AUD
Figures

Figure 3 C
Figure 3