Morphological and functional evidence of increased excitatory signaling in the prelimbic cortex during ethanol withdrawal

Abstract

Excessive alcohol consumption in humans induces deficits in decision making and emotional processing, which indicates a dysfunction of the prefrontal cortex (PFC). The present study aimed to determine the impact of chronic intermittent ethanol (CIE) inhalation on mouse medial PFC pyramidal neurons. Data were collected 6-8 days into withdrawal from 7 weeks of CIE exposure, a time point when mice exhibit behavioral symptoms of withdrawal. We found that spine maturity in prelimbic (PL) layer 2/3 neurons was increased, while dendritic spines in PL layer 5 neurons or infralimbic (IL) neurons were not affected. Corroborating these morphological observations, CIE enhanced glutamatergic transmission in PL layer 2/3 pyramidal neurons, but not IL layer 2/3 neurons. Contrary to our predictions, these cellular alterations were associated with improved, rather than impaired, performance in reversal learning and strategy switching tasks in the Barnes maze at an earlier stage of chronic ethanol exposure (5-7 days withdrawal from 3-4 weeks of CIE), which could result from the anxiety-like behavior associated with ethanol withdrawal. Altogether, this study adds to a growing body of literature indicating that glutamatergic activity in the PFC is upregulated following chronic ethanol exposure, and identifies PL layer 2/3 pyramidal neurons as a sensitive target of synaptic remodeling. It also indicates that the Barnes maze is not suitable to detect deficits in cognitive flexibility in CIE-withdrawn mice.

Introduction

Alcohol use disorders (AUDs) represent a spectrum of pathological patterns of alcoholic beverage consumption, which are notably characterized by loss of control over drinking and the emergence of negative affect during abstinence (American Psychiatric Association, 2013). The transition from moderate alcohol consumption to AUD is characterized by a gradual switch from positive to negative reinforcement, whereby the motivation to drink alcohol is no longer driven by the rewarding effects of intoxication, but by relief from the aversive effects of withdrawal (Cooney et al., 1997; de Castro et al., 2007; Johnson and Fromme, 1994; Lee et al., 2013; Pombo et al., 2016; Sinha, 2013). During abstinence, the heightened incentive salience of alcohol-associated cues combined with deficits in inhibitory control contribute to the reinstatement of drinking (Koob and Volkow, 2016).

Abstinence and cue-induced craving in human alcoholics are associated with abnormal activity in a number of prefrontal cortex (PFC) subregions implicated in reward anticipation, decision making and impulse control, including the dorsolateral PFC, ventromedial PFC, orbitofrontal cortex (OFC) and anterior cingulate cortex (Cyders et al., 2014; Jasinska et al., 2014; Seo et al., 2016). AUDs also cause volume loss and impaired functional connectivity of prefrontal regions (see Dupuy and Chanraud, 2016; Oscar-Berman and Marinkovic, 2007 for reviews). In accordance with these imaging data, AUD subjects exhibit deficits in a number of cognitive processes relevant to decision making and emotional processing (see Le Berre et al., 2017 for review).

Rodents exposed to chronic intermittent ethanol (CIE) also exhibit impaired performance in various cognitive tasks when tested during withdrawal. For instance, they show deficits in behavioral flexibility (attentional set-shifting and reversal learning in discrimination tasks), excessive impulsivity (premature responding in the 5-choice serial reaction time task), and impaired fear extinction retrieval (Badanich et al., 2011; Holmes et al., 2012; Hu et al., 2015; Irimia et al., 2014; Irimia et al., 2015; Kroener et al., 2012; Trantham-Davidson et al., 2014). Some of these cognitive abilities are known to depend upon medial PFC (mPFC) functional integrity (Chudasama and Robbins, 2006; Dalley et al., 2004; Myers-Schulz and Koenigs, 2012; Seamans et al., 2008; Uylings et al., 2003). Accordingly, studies investigating the molecular basis of the behavioral phenotypes induced by chronic ethanol exposure have identified local changes in excitatory and inhibitory neurotransmission, as well as monoaminergic modulation, within the mPFC (Holmes et al., 2012; Irimia et al., 2017; Kroener et al., 2012; Pleil et al., 2015; Trantham-Davidson et al., 2014). These functional changes are paralleled by structural alterations in dendritic arborization and spines of mPFC pyramidal neurons (Holmes et al., 2012; Kim et al., 2015; Kroener et al., 2012; Navarro and Mandyam, 2015).

The rodent mPFC is subdivided into prelimbic (PL, dorsally located) and infralimbic (IL, ventrally located) areas that exhibit differential connectivity, neurochemistry and behavioral functionality (Dolan and Dayan, 2013; George and Koob, 2010; Gourley and Taylor, 2016; Heidbreder and Groenewegen, 2003; Moorman et al., 2015; Vertes, 2004). In instrumental learning the PL and IL cortices promote mutually exclusive behaviors such that goal-directed actions require the PL area, while habitual responding requires the IL area (Balleine and Dickinson, 1998; Coutureau and Killcross, 2003; Killcross and Coutureau, 2003). The antagonistic functions of the PL and IL cortices are best illustrated by fear conditioning experiments, whereby the expression of learned fear is reduced both by PL inactivation or IL stimulation, while the extinction of conditioned freezing response is prevented both by IL inactivation or PL stimulation (Milad and Quirk, 2002; Morgan and LeDoux, 1995; Quirk et al., 2000; Vidal-Gonzalez et al., 2006). A similar distinction between behavioral execution and inhibition being driven by the PL and IL cortices, respectively, has also been proposed for reward/drug seeking, although the dichotomy appears less rigid than for aversive conditioning (Moorman et al., 2015; Peters et al., 2009; Pfarr et al., 2015). Finally, the PL and IL cortices play complementary roles in behavioral flexibility and conflict resolution; the IL facilitates the testing of a previously irrelevant strategy upon rule-shifting and reversal learning, while the PL enables the selection and maintenance of the new strategy (Oualian and Gisquet-Verrier, 2010). In accordance with this functional dichotomy, the few studies that have examined ethanol-induced neuroadaptations in both the PL and IL cortices have observed differential alterations of dendritic spines and synaptic transmission in the two subregions (Gass et al., 2014; Jury et al., 2017b; Pleil et al., 2015).

The mPFC is also laminar, with superficial (2/3) and deeper (5) layers exhibiting differential connectivity (Bitzenhofer et al., 2017; DeNardo et al., 2015; Gabbott et al., 2005). While there is in vitro evidence of layer specificity in the effects of chronic ethanol on mPFC pyramidal neurons (Pava and Woodward, 2014), there is no study to date that has directly compared neuroadaptations occurring in layer 2/3 vs layer 5 neurons in vivo.

Therefore, the present study aimed to: 1) investigate spine alterations in layer 2/3 and layer 5 pyramidal neurons of the PL and IL cortices of CIE-exposed mice, 2) determine whether morphological changes were associated with alterations in intrinsic excitability or excitatory synaptic transmission, and 3) assess whether cellular changes were associated with deficits in cognitive flexibility as measured by reversal learning and strategy switching in the Barnes maze. Data were collected 3-8 days into withdrawal from vapor, a time frame when CIE-exposed mice display escalated ethanol drinking and other behavioral alterations (Becker and Lopez, 2004; Jury et al., 2017a; Kroener et al., 2012).

Materials and Methods

Animals

Male C57BL/6J mice were obtained from The Scripps Research Institute Rodent Breeding Colony at 8 weeks of age and were acclimated to reverse light cycle (12 h light/dark cycle, lights on at 10:15 PM) for at least 10 days before any testing or treatment was performed. Mice were housed in a temperature-controlled vivarium (22°C). Water (acidified) and food (autoclaved Teklad LM-485/Envigo 7012 diet) were available ad libitum. All procedures were carried out in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and were approved by The Scripps Research Institute Institutional Animal Care and Use Committee.

Chronic intermittent exposure to ethanol vapor (CIE)

Inhalation chambers (La Jolla Alcohol Research Inc., La Jolla, CA) consisted of standard plastic mouse cages closed with airtight lids and fitted with two inlets for ethanol vapor inflow in the front and passive exhaust in the back. Ethanol vapor was created by dripping 95% ethanol into a 2-L Erlenmeyer vacuum flask kept at 50°C on a warming tray. Air was pumped into the flask (HK40L, Hakko, Laguna Hills, CA) and ethanol vapor flowed into each sealed chamber at a rate of 6 L/min. During each CIE week, mice were placed in inhalation chambers and subjected to four cycles of intoxication (ethanol vapor inhalation for 16 h, from 5:00 PM to 9:00 AM, starting Monday afternoon and ending Friday morning) separated by 8 h periods of withdrawal (air inhalation), and then returned to their home cage. Before each onset of ethanol vapor exposure, mice were injected intraperitoneally with a solution of ethanol (1.5 g/kg) and pyrazole (1 mmol/kg, Sigma, St Louis, MO), with control mice receiving injections of pyrazole alone. Ethanol vapor concentration was adjusted by varying the ethanol dripping rate, so as to yield serum ethanol concentrations approximating 200 mg/dL. Blood samples were obtained once per week at the end of a 16 h intoxication period. Blood was collected from the tail vein with heparinized capillary tubes and centrifuged for 5 min at 13000 rpm. The supernatant was processed in a GM7 analyser (Analox Instruments, London, UK). Average serum ethanol concentration was 217.5 ± 22.6 mg/dL in the cohort used for dendritic spine analysis, 211.8 ± 19.0 mg/dL in the cohort used for electrophysiological recordings, and 208.8 ± 14.3 mg/dL in the cohorts used for Barnes maze testing.

Dendritic spine analysis

Mice (Air, n=4; CIE, n=5) were exposed to a total of 7 weeks of air/CIE and euthanized 7 days into withdrawal. Mice were anesthetized with isoflurane and perfused with cold PBS followed by 4% paraformaldehyde (PFA). Brains were extracted and postfixed in 4% PFA at 4°C for 2 hours before being sectioned coronally into 100 μm slices on a vibratome (Leica VT1000S; Leica Biosystems). Sections were then labeled with 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate (DiI) using a biolistic delivery system (Helios Gene Gun, Biorad) and incubated in PBS for 3 days at 4°C before being mounted onto slides in PBS. Neurons in layers 2/3 and 5 of the prelimbic (PL) and infralimbic (IL) divisions of the mPFC were imaged by confocal microscopy (model LSM 710; Carl Zeiss MicroImaging) using a series of high-resolution optical sections (1,024 × 1,024-pixel format) that were captured for each neuron with a 63× water-immersion objective (1.2 numerical aperture), with 1× zoom at 1-μm step intervals (z-stack). All images were acquired under identical conditions. Each z-stack was collapsed into a single image by projection in Zeiss ZEN software, converted to a tiff file, encoded for blind analysis and spine counting was performed in Image J (NIH). Sections from at least 20 cells and between 40-58 dendritic segments were imaged per treatment condition.

Electrophysiological recordings

Mice (Air, n=5; CIE, n=5) were exposed to a total of 7 weeks of air/CIE and euthanized 6-8 days into withdrawal. After euthanasia with 3-5% isoflurane, the brains were placed in cold, oxygenated (95% O₂/5% CO₂) high sucrose solution (pH 7.3-7.4): 206.0 mM sucrose; 2.5 mM KCl; 0.5 mM CaCl₂; 7.0 mM MgCl₂; 1.2 mM NaH₂PO₄; 26.0 mM NaHCO₃; 5.0 glucose; 5.0 HEPES, as previously described (Bajo et al., 2015; Salling and Harrison, 2014; Schweitzer et al., 2016). Coronal slices (300 μm) containing the PL and IL cortices were sectioned (Leica VT1000S) and incubated in oxygenated artificial cerebrospinal fluid: 130 mM NaCl, 3.5 mM KCl, 2 mM CaCl₂, 1.25 mM NaH₂PO₄, 1.5 mM MgSO₄, 24 mM NaHCO₃, and 10 mM glucose for 30 min at 32 °C and then for 30 min at room temperature. PL and IL neurons were visualized with infrared-differential interference contrast (IR-DIC) optics, a w60 water immersion objective (Olympus BX51WI) and a CCD camera (EXi Aqua, QImaging), and layer 2/3 pyramidal neurons located 100-300 μm from the pial surface were identified by their characteristic size and shape (Salling and Harrison, 2014). Whole-cell recordings from a total of 175 neurons were performed in gap-free acquisition mode with a sampling rate per signal of 20 kHz and low-pass filtered at 10 kHz, using a Multiclamp 700B amplifier, Digidata 1440A and pClamp 10.2 software (all Molecular Devices). Recording pipettes (4-6 MΩ resistance) were filled with internal solution: 145 mM K-gluconate, 0.5 mM EGTA, 2 mM MgCl₂, 10 mM HEPES, 2 mM Na-ATP, and 0.2 mM Na-GTP. Neuronal membrane properties (resting membrane potential and input resistance) and excitability (action potential threshold and rheobase, recorded in current-clamp mode using a series of current injections of 5 pA/600 ms in cells held at −70 mV) were assessed, and the recordings analyzed using NeuroExpress software developed by Dr. A. Szucs. For the voltage-clamp experiments, cells held at −70 mV were bathed in GABA receptor antagonists (1 μM CGP 55845A and 30 μM bicuculline) to isolate spontaneous glutamatergic excitatory postsynaptic currents (sEPSCs). Due to the magnesium concentration in the ACSF and our cell holding potential, we expect these sEPSCs to be primarily AMPA receptor-mediated currents. To confirm this, the NMDA receptor antagonist DL-2-Amino-5-phosphonopentanoic acid (30 μM DL-AP5) was subsequently applied to the bath in a subset of recordings from PL and IL neurons of Air and CIE mice. sEPSC recordings with a series resistance >15 MΩ or a >20% change in series resistance, as monitored with a 10 mV pulse, were excluded. sEPSC frequencies, amplitudes and kinetics were analyzed over a 3 min period using Mini Analysis software (Synaptosoft Inc.) and visually confirmed, with events <5 pA excluded and cells with less than 60 events/3 min interval excluded.

Barnes maze

Two independent cohorts of mice were tested in the Barnes maze to examine cognitive flexibility 5-7 days into withdrawal from 3-4 weeks of CIE exposure (first cohort, n=24 mice; second cohort, n=12 mice). The Barnes maze test was performed as described previously (Bach et al., 1995; Barnes, 1979; Holmes et al., 2002; Sunyer et al., 2007). Mice were given 5 min to escape into a small dark recessed chamber placed under 1 of 20 small holes (2” diameter) evenly distributed along the perimeter of an elevated (36” high), brightly lit (400 lux), circular open platform (36” diameter). At the beginning of each trial, the mouse was placed in the center of the maze under an opaque cylinder (4” high). After 10 s, the cylinder was lifted and the trial began. Once the mouse entered the escape box, the hole was covered to prevent the mouse from resurfacing onto the maze and the mouse was returned to its home cage 1 min later. If the mouse failed to enter the escape box within 5 min, it was gently guided toward the target hole and allowed to enter the escape box. Mice were given two trials per day, the first at the beginning of the dark phase and the second approximately 3 h later. A habituation session was first performed, which entailed placing the mouse in the escape box for 1 min. The mice then underwent four testing phases: spatial learning, memory retention/re-acquisition, reversal learning and strategy switching (Fig. 5A). During the acquisition phase, the location of the escape box was constant for a given mouse (but randomized across mice), and distal visual cues were available for the mouse to navigate. The path of the mouse was recorded by a ceiling-mounted video camera linked to Any-maze video tracking software (Stoelting). The time elapsed, distance traveled, and number of holes visited prior to entering the escape box were used as performance indexes. Path length was the primary endpoint as this measure is not confounded by velocity (Contet et al., 2001). Mice were tested as many times as needed for their performance to stabilize and for their search strategy to be serial or direct rather than random (20 trials in first cohort, 24 trials in second cohort). One mouse was excluded because it systematically jumped from the platform and five other mice were excluded because they were still using a random strategy by the end of the acquisition phase. Mice were then exposed to three weeks of air (n=14) or CIE (n=16) inhalation. Five mice died 24-72 h into withdrawal after their first week of CIE exposure. After three air/CIE weeks, a Barnes maze trial was conducted 3 days into withdrawal from vapor to assess spatial memory retention. Mice were then retrained under the same conditions until their performance stabilized (5 trials, withdrawal days 3-5). The reversal learning phase was then started, whereby the location of the escape box was moved to the hole diametrically opposed to the hole used during the acquisition phase. Mice were then tested until their performance stabilized (5 trials, withdrawal days 5-7). The mice were then exposed to one more week of air or CIE inhalation, and a Barnes maze trial was conducted 3 days into withdrawal from vapor using the same conditions as during the reversal learning phase. Mice were then retrained under the same conditions until their performance stabilized (6 trials, withdrawal days 3-5). The strategy switching phase was then started, whereby a beacon (orange plastic racket-shaped toy, 3.5” high, 2” wide) was affixed on the rim of the platform, next to the escape hole, and the location of the escape hole changed at every trial (i.e., mice now had to rely on a proximal cue rather than distal cues for navigation). Mice were then tested until their performance stabilized (5 trials, withdrawal days 5-7). Mice from the second cohort were then exposed to an additional 3 weeks of CIE following completion of the Barnes maze experiment and were used for dendritic spine analysis.

Figure 5. Effect of CIE on spatial reference memory retention and cognitive flexibility in the Barnes maze.

A: Experimental design. Mice were first trained to acquire the spatial navigation task, then exposed to three weeks of air or chronic intermittent ethanol (CIE, 16 h/day for 4 consecutive days in each week) inhalation, and memory of the escape hole was evaluated. Following re-acquisition of the task, reversal learning was evaluated by moving the escape hole to the diametrically opposite location on the maze. Mice were exposed to an additional week of air or CIE inhalation and retrained on the task, and then their ability to switch from a spatial to a visual-cued strategy was tested by changing the location of the escape hole at each trial and signaling its location with a proximal cue. B: Cognitive performance measured by path length, i.e., distance traveled from starting point to escape hole. Data were analyzed by repeated measures ANOVA within each experimental phase and at each phase transition and results are reported in the main text (Air, n=14 mice; CIE, n=11 mice). Data are presented as mean ± SEM. * represents the effect of trials, ^# represents the effect of treatment. One symbol, p<0.05; three symbols, p<0.001.

Data analysis

Spine densities and electrophysiological measures were analyzed by unpaired two-tailed t-tests. Spine proportions were analyzed using two-way repeated-measures analysis of variance (ANOVA), with spine shape as the within-subject variable and treatment as the between-subject variable. Barnes maze data were analyzed by two-way repeated-measure ANOVA, with trial as the within-subject variable and treatment as the between-subject variable. Simple linear regression was performed to correlate dendritic spine parameters with cognitive flexibility performance in the Barnes maze, using the average distance traveled during all reversal learning and strategy switching trials as the independent variable. Data are presented as mean ± standard error of the mean (SEM), with the number of cells and mice used for each experiment reported in the figure legend.

Results

CIE increases the maturity of dendritic spines in PL layer 2/3 pyramidal neurons

We examined whether CIE exposure followed by 7 days of withdrawal produced dendritic spine remodeling in layer 2/3 and layer 5 pyramidal neurons of the PL and IL cortices. CIE did not alter spine density in either layer of both subregions (Fig. 1). We further analyzed the morphology of dendritic protrusions, which were classified as mushroom (most mature), stubby, thin and filopodia (least mature) (Fig. 2A). There was a significant interaction between spine type and treatment in PL layer 2/3 neurons (F_3,21 = 8.0, p<0.001), whereby CIE increased the proportion of mushroom spines (p<0.01) and decreased the proportion of thin spines (p<0.01) (Fig. 2B). A similar interaction was observed in IL layer 2/3 neurons (F_3,21 = 3.2, p<0.05), but post hoc analysis did not detect a significant change in the proportion of any spine type (Fig. 2D). There were no significant changes in the proportions of spine types in PL and IL layer 5 neurons (Fig. 2C and 2E).

Figure 1. Spine density in layer 2/3 and 5 of the prelimbic (PL) and infralimbic (IL) divisions of the medial prefrontal cortex.

A: Representative image of the dendrites imaged and analyzed. B–E: Spines were analyzed in pyramidal neurons of PL layer 2/3 (B) and layer 5 (C) and IL layer 2/3 (D) and layer 5 (E). Spine densities were analyzed by unpaired t-test and no effect of CIE was detected. Data are presented as mean ± SEM (Air, n=40 dendritic segments from 15 neurons from 4 mice; CIE, n=58 dendritic segments from 16 neurons from 5 mice).

Figure 2. Proportion of spine types in layer 2/3 and 5 of the prelimbic (PL) and infralimbic (IL) divisions of the medial prefrontal cortex.

A: Schematic showing the classification of the different spine types. B-E: Spine type proportions were analyzed using repeated measures ANOVA and there was a significant interaction between spine type and treatment in PL layer 2/3 (F_3,21=8.0, p<0.001) (B) and IL layer 2/3 neurons (F_3,21=3.2, p<0.05) (D), but not in PL layer 5 (C) or IL layer 5 neurons (E). **, p<0.05. Data are presented as mean ± SEM (Air, n=40 dendritic segments from 15 neurons from 4 mice; CIE, n= 58 dendritic segments from 16 neurons from 5 mice).

CIE does not affect the basal membrane properties or excitability of PL layer 2/3 pyramidal neurons

To determine whether the region- and layer-specific effect of CIE on spine maturity was associated with altered neuronal function, we performed in vitro whole-cell electrophysiology in PL and IL layer 2/3 pyramidal neurons 6-8 days into withdrawal from CIE (Fig. 3A). In both mPFC subregions, there were no differences in resting membrane potential, input resistance, action potential threshold and rheobase between control and CIE-exposed mice (Fig. 3B–J), indicating that CIE did not affect neuronal membrane properties or excitability.

Figure 3. Membrane properties and excitability measures in layer 2/3 neurons of the prelimbic (PL) and infralimbic (IL) cortices.

A: Schematic illustrating the location of PL and IL areas (left panel). Pipettes were directed at layer 2/3 (×4, middle panel), and pyramidal neurons were identified by their triangular shape with IR-DIC optics (×60, right panel). B–C: Group comparison by unpaired t-test showed no differences in the resting membrane potential (B) or input resistance (C) of PL layer 2/3 pyramidal neurons between control and CIE mice (Air, n=28 cells from 5 mice; CIE, n=25 cells from 5 mice). D: Representative traces of current-injected firing elicited by a stimulus waveform (5 pA steps, 600 ms duration) from PL layer 2/3 pyramidal neurons in control and CIE mice. The depicted sweeps resulted from injection of 0, 20, 40, 60, 80, 100, 120, 140, 160, 180 and 200 pA into both cells. E–F: Group comparison by unpaired t-test showed no differences in the action potential threshold (E) or rheobase (F) of PL layer 2/3 pyramidal neurons in control and CIE mice (Air, n=28 cells from 5 mice; CIE, n=25 cells from 5 mice). G–J: Group comparison by unpaired t-test showed no differences in the resting membrane potential (G), input resistance (H), action potential threshold (I) or rheobase (J) of IL layer 2/3 pyramidal neurons between control and CIE mice (Air, n=24 cells from 5 mice; CIE, n=27 cells from 5 mice). Data are presented as mean ± SEM.

CIE increases postsynaptic glutamate receptor function at PL layer 2/3 synapses

We then investigated whether CIE altered glutamatergic signaling at PL and IL layer 2/3 synapses by recording pharmacologically-isolated spontaneous excitatory postsynaptic currents (sEPSCs). In PL layer 2/3 pyramidal neurons, CIE-exposed mice displayed a similar sEPSC frequency to that of control mice, but had a significantly higher sEPSC amplitude (t₄₅=3.31, p<0.01; Fig. 4A–C). There was also a significant increase in the sEPSC decay time of PL layer 2/3 pyramidal neurons after exposure to CIE (t₄₅=3.10, p<0.01; Fig. 4A and 4D). In contrast, CIE had no effect on the sEPSC frequency, amplitude or kinetics of IL layer 2/3 pyramidal neurons (Fig. 4E–H). Generally, increased sEPSC frequencies reflect higher glutamate release probabilities, whereas increased amplitudes and kinetics indicate enhanced glutamate receptor sensitivity (via changed receptor number and/or subunit composition) (De Koninck and Mody, 1994; Otis et al., 1994). Therefore, CIE increased postsynaptic glutamate receptor function at PL layer 2/3 synapses, but not IL layer 2/3 synapses, without affecting glutamate release in either subregion. These postsynaptic effects were primarily mediated by AMPA receptors, as the NMDA receptor antagonist DL-AP5 (30 μM) was applied to a subset of recorded cells (n=16 cells from both subregions of Air and CIE mice) and no differences in sEPSC characteristics were observed (Supplementary Figure S1).

Figure 4. Glutamatergic transmission in layer 2/3 neurons of the prelimbic (PL) and infralimbic (IL) cortices.

A: Representative sEPSC traces (left panel) and scaled average sEPSCs (right panel) from PL layer 2/3 pyramidal neurons of control and CIE mice. B–D: Group comparison by unpaired t-test showed no difference in baseline sEPSC frequency (B), but a significantly higher sEPSC amplitude (C) and decay time (D) in CIE vs. control mice (Air, n=24 cells from 5 mice; CIE, n=23 cells from 5 mice). E: Representative sEPSC traces (left panel) and scaled average sEPSCs (right panel) from IL layer 2/3 pyramidal neurons of control and CIE mice. F–H: Group comparison by unpaired t-test showed no differences in baseline sEPSC frequency (F), amplitude (G) and kinetics (H) in CIE vs. control mice (Air, n=24 cells from 5 mice; CIE, n=24 cells from 5 mice). Data are presented as mean ± SEM. **, p<0.01.

CIE does not impair memory retention, reversal learning or strategy switching in the Barnes maze

We sought to identify a behavioral correlate of the morphological and functional changes identified in PL layer 2/3 pyramidal neurons. To do so, we evaluated the effect of CIE on cognitive flexibility in the Barnes maze spatial reference memory assay (Fig. 5A–B). We first trained the mice to learn the location of the escape hole using distal cues. We then exposed the mice to 3 consecutive weeks of air or CIE inhalation. Both groups showed an equivalent reduction in performance (as measured by path length) at the first post-CIE trial (effect of trial, F_1,23=5.6, p<0.05; effect of treatment, F_1,23=0.2, n.s.; trial × treatment, F_1,23=0.004, n.s.) and re-acquired the task at the same rate (effect of trial, F_4,92=6.8, p<0.001; effect of treatment, F_1,23=1.6, n.s.; treatment × trial interaction, F_4,92=2.2, n.s.). We then assigned the target hole to the diametrically opposite location for each mouse (reversal learning phase, withdrawal days 5-7). Both groups showed an equivalent reduction in performance at the first reversal learning trial (effect of trial, F_1,23=23.4, p<0.001; effect of treatment, F_1,23=0.3, n.s.; treatment × trial interaction, F_1,23=1.9, n.s.). Both groups learned the new location at the same rate but CIE-exposed mice performed better throughout the reversal learning phase (effect of trial, F_4,92=5.9, p<0.001; effect of treatment, F_1,23=5.1, p<0.05; treatment × trial interaction, F_4,92=0.4, n.s.). The mice were then exposed to an additional week of air or CIE inhalation. There was no significant effect of the 10-day testing pause on performance (effect of trial, F_1,23=2.2, n.s.; effect of treatment, F_1,23=0.3, n.s.; treatment × trial interaction, F_1,23=0.01, n.s.). Both groups improved their performance at the same rate during re-acquisition trials, but the CIE-exposed mice again performed better than their air-exposed counterparts throughout re-acquisition (effect of trial, F_5,115=2.4, p<0.05; effect of treatment, F_1,23=4.2, p=0.05; treatment × trial interaction, F_5,115=0.1, n.s.). We then randomly changed the location of the target hole at each trial and signaled the target hole with a proximal cue (strategy switching phase, withdrawal days 5-7). Both groups showed a major reduction in performance at the first strategy-switching trial, but CIE-exposed mice again performed better (effect of trial, F_1,23=26.3, p<0.001; effect of treatment, F_1,23=6.6, p<0.05; treatment × trial interaction, F_1,23=2.5, n.s.). Both groups learned the new strategy at the same rate (effect of trial, F_4,92=2.9, p<0.05; effect of treatment, F_1,23=4.1, p=0.056; treatment × trial interaction, F_4,92=0.5, n.s.). Similar data were obtained using latency to escape or number of holes visited as a measure of performance (Supplementary Figure S2A–B and Tables S1–S2).

Dendritic spine morphology in PL layer 2/3 pyramidal neurons correlates with Barnes maze performance

Finally, we examined whether the performance of mice during cognitive flexibility testing predicted dendritic spine morphology after exposure to an additional 3 weeks of CIE and 7 days of withdrawal. For PL layer 2/3 neurons, the average distance traveled during reversal learning and strategy switching trials was negatively correlated with the percentage of mushroom spines (R²=0.594, p<0.05) and positively correlated with the percentage of thin spines (R²=0.587, p<0.05). There was also a trend for a negative correlation with spine density (R²=0.551, p=0.056). There was no significant correlation with any of the spine parameters measured in the other populations of mPFC pyramidal neurons.

Discussion

Our study shows that chronic exposure to ethanol followed by one week of withdrawal elicits morphological and functional changes indicative of enhanced AMPA receptor-mediated excitatory transmission in layer 2/3 pyramidal neurons of the PL subregion of the mouse mPFC. Specifically, we found that subjecting mice to 7 weeks of CIE followed by 6-8 days of withdrawal did not affect dendritic spines in PL layer 5 neurons or IL neurons, but selectively increased spine maturity in PL layer 2/3 neurons. While there was no effect of CIE on basal membrane properties or excitability of layer 2/3 pyramidal neurons in either mPFC subregion, CIE increased the postsynaptic glutamate receptor function of PL layer 2/3 neurons, but not IL layer 2/3 neurons, and this effect was primarily mediated by AMPA receptors. Contrary to our predictions, these cellular alterations were associated with improved, rather than impaired, performance in reversal learning and strategy switching tasks in the Barnes maze, and there was a positive correlation between PL layer 2/3 spine maturity and cognitive flexibility.

Spine morphology plays a critical role in regulating excitatory synaptic function, as the maturation of dendritic spines and enlargement of spine heads is associated with increased spine stability and increased synaptic strength (De Roo et al., 2008; Sala and Segal, 2014). Mushrooms spines have larger postsynaptic densities, which anchor more AMPA receptors and make these synapses functionally stronger (Bourne and Harris, 2007). Accordingly, chronic ethanol binge drinking increased the average spine area and the proportion of mushroom spines in medium spiny neurons of the nucleus accumbens shell, and these changes were associated with the insertion of GluA2-lacking AMPA receptors (Laguesse et al., 2017). Another critical component of ethanol-induced plasticity at glutamatergic synapses involves the upregulation of synaptic NR2B-containing NMDA receptors and subsequent expansion of the postsynaptic density via PSD-95 scaffolding capacity (Kroener et al., 2012; Mulholland and Chandler, 2007). In our study, withdrawal from CIE increased the proportion of mushroom spines at the expense of thin spines selectively in PL layer 2/3 neurons. In agreement with the known relationship between spine shape and function, these structural changes were accompanied by an increase in sEPSC amplitude and decay time, which reflects enhanced postsynaptic glutamate receptor function and was primarily mediated by AMPA receptors as our recording conditions did not allow for the measure of NMDA receptor-mediated currents. Importantly, CIE did not affect spine maturity nor excitatory transmission in IL layer 2/3 pyramidal neurons, which further supports the correlation between spine morphology and synaptic function.

Our morphological and functional results therefore add to a growing body of literature indicating that glutamatergic activity in the PFC is upregulated following chronic ethanol exposure (see Burnett et al., 2016 for review). Importantly, we observed a higher sensitivity of PL layer 2/3 neurons compared to layer 5 or IL neurons. This finding corroborates a recent study by Jury et al. that reported no changes in total spine density in mouse PL and IL layer 2/3 neurons, but a selective increase in the head width of wide spines on PL layer 2/3 neurons (i.e., mushroom spines, based on the classification we used in the present study) after 3 days of withdrawal from CIE (Jury et al., 2017b). In contrast to these and our data, CIE-exposed rats exhibited a higher overall spine density, but no change in mushroom spine density, in mPFC layer 2/3 neurons (PL and IL neurons were not distinguished) 3 h after the last vapor exposure (Kim et al., 2015).

Possible explanations for this divergent finding include the different species and different withdrawal time-point used in that study. Likewise, Pleil and colleagues reported a trend for increased sEPSC amplitude in mouse PL layer 2/3 neurons recorded 48 h into withdrawal from three weeks of CIE (Pleil et al., 2015), and it is possible that this trend would have reached significance had the mice experienced longer CIE exposure and/or withdrawal as in the present study. Finally, in an earlier report, CIE increased spine maturity in mouse mPFC layer 5 neurons and this increase persisted after one week of withdrawal, which we did not replicate in our study (Kroener et al., 2012). Although the experimental conditions were for the most part identical to those used here, we speculate that the discrepancy arises from a differential extent of CIE exposure (three weeks in Kroener et al., 2012 vs seven weeks in the present study), which may correspond to different stages of ethanol dependence. Further studies will be needed to address the molecular mechanisms underlying increased spine maturity and glutamatergic transmission in PL layer 2/3 neurons following CIE exposure. A promising approach would entail the proteomic analysis of synaptic fractions, as was recently done in the nucleus accumbens and hippocampus (Uys et al., 2016; Wills et al., 2017). These pioneering studies identified a number of scaffolding proteins, cytoskeletal proteins, and other proteins known to regulate cellular morphology, which may contribute to dendritic spine plasticity during ethanol intoxication and withdrawal.

Based on the morphological and functional alterations we detected in PL layer 2/3 pyramidal neurons, we predicted that CIE-exposed mice would show impaired cognitive flexibility in the Barnes maze. Although spatial reference learning and memory is primarily dependent on the hippocampus, behavioral flexibility in spatial navigation tasks requires the mPFC (de Bruin et al., 1994; Deacon et al., 2002; Deacon and Rawlins, 2002; Delatour and Gisquet-Verrier, 2000; Lacroix et al., 2002; Latif-Hernandez et al., 2016; Livingston-Thomas et al., 2015; McDonald et al., 2008; Morris et al., 1982; Ragozzino et al., 1999). In particular, mPFC lesions in the rat impair reversal spatial learning and shifting from a spatial to a visual strategy in the Morris water maze (de Bruin et al., 1994; Ragozzino et al., 1999). Our experimental design aimed to detect CIE-induced mPFC deficits by testing these two aspects of cognitive flexibility in a spatial navigation task. We selected the Barnes maze because it is less stressful for mice than the Morris water maze (as measured by plasma corticosterone levels) and the spatial learning ability of mice is higher in dry-land tasks than in swimming-pool tasks (Harrison et al., 2009; Whishaw and Tomie, 1996). A limitation of our study is that cognitive flexibility was examined after 3-4 weeks of CIE, while cellular analyses were conducted after 7 weeks of CIE. Despite this discrepancy, Barnes maze performance during reversal learning and strategy switching predicted dendritic spine maturity in PL layer 2/3 neurons, but not other populations of mPFC pyramidal neurons.

Overall, we found that CIE exposure reduced Barnes maze performance to the same extent as an equivalent break in training in air-exposed mice, indicating that CIE does not affect the retention of spatial reference memory. This result is consistent with the lack of effect of CIE on the retention of previously learned discrimination tasks (Badanich et al., 2011; Kroener et al., 2012). In contrast, our observation that CIE-exposed mice performed as well, or even better, than their air-exposed counterparts in the reversal learning and strategy switching phases of the Barnes maze experiment is at odds with studies that detected impaired cognitive flexibility in discrimination tasks 3-7 days into withdrawal from 3 weeks of CIE (Badanich et al., 2011; Kroener et al., 2012). Several possibilities can be considered to explain the incongruity between our results and those previous reports.

First, distinct experimental procedures of cognitive flexibility assessment are known to engage different prefrontal subregions, which may be differentially impacted by CIE (Dalley et al., 2004). For instance, in tasks using distinct sensory modalities, the PL is critical for attentional processes related to visual inputs, while the OFC processes olfactory and gustatory information (see Uylings et al., 2003 for review). Furthermore, reversal learning in simple discrimination tasks recruits the OFC, while the mPFC is necessary for reversal learning in spatial navigation tasks, as well as for task shifts that involve more effortful processing (“higher order rules”) (Brown and Bowman, 2002; Chudasama and Robbins, 2006; de Bruin et al., 1994; Floresco et al., 2008). In support of this view, reversal learning and attentional set-shifting deficits previously reported in CIE-exposed mice were identified using distinct procedures. Badanich et al. used an odor + digging medium discrimination task and detected deficits in reversal learning, but not set-shifting, 3 days into withdrawal from CIE (Badanich et al., 2011). In contrast, Kroener and colleagues used a T-maze discrimination task based on either egocentric response or visual cue and detected deficits in set-shifting, but not reversal learning, 3-7 days into withdrawal (Hu et al., 2015; Kroener et al., 2012). Deficits in attentional set-shifting were also observed in CIE-exposed rats tested in an operant discrimination task (Trantham-Davidson et al., 2014). Our data indicate that the Barnes maze is not suitable to detect either type of CIE-induced deficit in cognitive flexibility, even though it revealed reversal learning deficits in adult mice exposed to ethanol binges during adolescence (Coleman et al., 2014). Although the mPFC is critical to cognitive flexibility in spatial navigation tasks, it is possible that the particular mPFC neurons recruited during task shifts are not those affected by CIE, namely PL layer 2/3 pyramidal neurons.

Our data may also result from counterintuitive consequences of mPFC dysfunction. In contrast to the effect of larger thermal lesions of the mPFC, electrolytic lesions confined to the PL cortex paradoxically improved reversal learning in the Morris water maze (de Bruin et al., 1994; Maaswinkel et al., 1996). This observation is particularly relevant to our observation of the selective impact of CIE on PL pyramidal neurons, rather than IL pyramidal neurons. The improved performance of CIE-exposed mice during the reversal learning and strategy switching phases may therefore be compatible with PL cortex dysfunction. Interestingly, CIE-exposed mice also performed slightly better in the intra-dimensional shift test conducted after reversal learning by Badanich et al., indicating a modest facilitation of cognitive flexibility in that particular task switch (Badanich et al., 2011). Further studies will be needed to elucidate how increased excitatory transmission in PL pyramidal neurons may produce behavioral phenotypes similar to those elicited by mPFC inactivation. A similar situation was previously observed in a mouse model of traumatic brain injury, whereby increased glutamatergic input onto PL layer 2/3 (but not layer 5) pyramidal neurons was associated with working memory deficits (Smith et al., 2015). Moreover, increased excitatory transmission may not necessarily increase neuronal activity in vivo and may instead be associated with hypoactivity, i.e., comparable to the effects of experimental inactivation. An example of such dissociation was reported in inflammatory and neuropathic pain models, whereby PL layer 2/3 pyramidal neuron dendritic branching, spine density and glutamatergic synaptic transmission were increased, yet the excitability of these neurons was reduced and the associated behavioral phenotypes were reversed by optogenetic activation of PL excitatory neurons (Metz et al., 2009; Wang et al., 2015).

A third possible explanation for the discrepancy between our data and previous findings relates to the potential interference of other behavioral alterations elicited by CIE withdrawal (e.g., negative emotional state) with Barnes maze performance. Importantly, in the studies by Badanich et al. and Kroener et al. mice were motivated to complete the cognitive task by earning a reward (palatable food), while in our study mice were motivated by escaping an aversive environment (brightly lit, open space). Anxiety-like behavior is therefore more likely to have influenced performance in our study by increasing motivation to locate the escape hole. In accordance with this hypothesis, CIE-exposed male mice exhibit anxiety-like behavior 3-6 days into withdrawal and may therefore be more sensitive to the aversive nature of the Barnes maze (Jury et al., 2017a). Interestingly, mPFC layer 2/3 neurons have been implicated in moderating the impact of stress on emotions (Shrestha et al., 2015). The cellular abnormalities we identified in this neuronal population could therefore promote anxiety in CIE-exposed mice. Layer 2/3 pyramidal neurons of the mouse PL cortex receive inputs from the contralateral mPFC, basolateral amygdala, midline thalamic nucleus and ventral hippocampus, indicating that CIE may increase excitatory drive from these areas to elicit anxiety (Cheriyan et al., 2016; Little and Carter, 2012).

Of note, the phenotype we observed in the Barnes maze may not necessarily generalize to other spatial navigation tasks. Anxiety-like behavior and plasma corticosterone levels negatively correlate with Morris water maze performance (Darcet et al., 2014; Harrison et al., 2009; Holscher, 1999; Miyakawa et al., 2001), suggesting that the anxiogenic-like effects of CIE withdrawal would likely worsen reversal learning abilities in this task. While this interaction has yet to be studied, the reversal learning deficits induced by adolescent ethanol binge drinking can be detected both in the Barnes and Morris water mazes, despite the presence of an anxious-like phenotype (Coleman et al., 2011; Coleman et al., 2014). Our results, in conjunction with these findings, indicate that adolescent binge drinking and adult CIE elicit distinct cognitive alterations.

Collectively, we have shown that glutamatergic activity in the PFC is upregulated following chronic ethanol exposure, and identified PL layer 2/3 pyramidal neurons as a sensitive target of synaptic remodeling. In the future, it will be interesting to determine whether inhibition (or activation) of PL pyramidal neurons can reverse some of the behavioral consequences of CIE. Further insights into the chain of events linking the PL cortex to cognitive and emotional disruption during ethanol withdrawal may also be provided by studying the impact of CIE on downstream circuitry, i.e., on those neurons whose activity is controlled by PL layer 2/3 pyramidal neurons, such as ventral striatum or basolateral amygdala neurons (Gabbott et al., 2005; Little and Carter, 2013).

Supplementary Material

1

Supplementary Figure S1. To pharmacologically isolate the AMPA receptor-mediated component of sEPSCs, the NMDA receptor antagonist DL-AP5 (30 μM) was co-applied to cells bathed in GABA receptor antagonists (1 μM CGP 55845A and 30 μM bicuculline). DL-AP5 had no effect on sEPSC frequency, amplitude and kinetics (n=16 cells from both mPFC subregions of Air and CIE mice).

Supplementary Figure S2. Effect of CIE on spatial reference memory retention and cognitive flexibility in the Barnes maze. A: Cognitive performance measured by latency to enter escape hole. B: Cognitive performance measured by number of holes visited prior to escape. Data were analyzed by repeated measures ANOVA within each experimental phase and at each phase transition and results are reported in Tables S1 and S2 (Air, n=14 mice; CIE, n=11 mice). Data are presented as mean ± SEM. * represents the effect of trials, ^# represents the effect of treatment. One symbol, p<0.05; two symbols, p<0.01; three symbols, p<0.001.

Table S1. Statistical analysis of latency to escape in the Barnes maze experiment. Data were analyzed by repeated measures ANOVA using trial as within-subject variable and treatment as between-subject variable. Data are shown in Supplementary Figure S2A.

Table S2. Statistical analysis of number of holes visited in the Barnes maze experiment. Data were analyzed by repeated measures ANOVA using trial as within-subject variable and treatment as between-subject variable. Data are shown in Supplementary Figure S2B.

Highlights.

• Withdrawal from chronic intermittent ethanol alters the medial prefrontal cortex.

• It increases spine maturity in prelimbic cortex layer 2/3 pyramidal neurons.

• It also increases excitatory transmission in these neurons.

• However, it does not impair cognitive flexibility in the Barnes maze.