Chronic Intermittent EtOH Exposure.

Male C57BL/6J mice were exposed to a chronic intermittent ethanol (CIE) procedure designed to simulate the repeated episodes of heavy drinking and withdrawal that characterize alcohol abuse (25, 26). CIE involved 16-h continuous EtOH exposure followed by an 8-h withdrawal. A cycle of CIE consisted of four exposures (Monday–Friday), followed by a longer 80-h withdrawal (Saturday–Sunday). There were two cycles in total. Representative blood EtOH concentrations (BECs) from sentinel mice (n = 11) were in the range of 175 ± 25 mg/dL and did not decrease (i.e., exhibit metabolic tolerance) from the first to the second cycle (Fig. 1A).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Chronic intermittent EtOH exposure. (A) Representative blood EtOH concentrations obtained in sentinel mice on the first to the second CIE cycle. (B) CIE increased EtOH drinking in a two-bottle choice test. Air controls showed a transient increase in drinking (EtOH deprivation effect). Data are means ± SEM, *P < 0.05 CIE vs. air, ^#P < 0.05 vs. pre-CIE/same group.

To confirm that this CIE procedure increased EtOH consumption, as reported in prior studies (8), we tested mice for EtOH drinking in a two-bottle choice test over 12 d, beginning 3 d after CIE (i.e., postacute withdrawal). CIE mice exhibited significantly increased EtOH drinking relative to baseline levels recorded over 2 wk before exposure, and compared with air-exposed controls (ANOVA group × time interaction: F2,36 = 6.19, P < 0.01, followed by Fisher’s least significant differences (LSD) post hoc tests, n = 10). Increased drinking in CIE mice was evident during the first 1–6 d of post-CIE measurement (see Fig. 1B and Fig. S1 for detailed timeline). Air controls also showed an increase in drinking relative to preexposure baseline (t test: t (9) = 4.83, P < 0.01), likely due to an EtOH deprivation effect. By days 7–12 postexposure, consumption had returned to baseline in the air group but remained significantly elevated in CIE mice (t test: t (9) = 2.72, P < 0.05) (Fig. 1B). These data show that CIE produced lasting increases in EtOH drinking.

CIE Causes DLS Dendritic Hypertrophy.

Chronic psychostimulant exposure increases dendritic arborization in the PFC and ventral striatum (27). We recently found that four cycles of CIE produced dendritic hypertrophy in mouse PFC neurons (10), but it is not known whether neurons in the DLS are also affected. We therefore exposed mice to two cycles of CIE and collected brains 3 d later for impregnation with a Golgi–Cox stain to visualize the dendritic arbor of DLS neurons, as previously described (10, 28).

Sholl analysis revealed CIE-induced increases in dendritic length (ANOVA group × distance-from-soma interaction: F11,143 = 3.41, P < 0.01, group effect: F1,13 = 5.60, P < 0.05, distance-from-soma effect: F11,143 = 286.88, P < 0.01, n = 7–8 mice per group, n = 8 neurons per mouse) (Fig. 2A) and number of dendritic branches (ANOVA group × distance-from-soma interaction: F11,143 = 1.95, P < 0.05) (Fig. S2). There was a significant increase in the overall amount of dendritic material in DLS neurons after CIE, relative to air exposure. Moreover, CIE produced a significant increase in the length (t test: t (13) = 6.60, P < 0.05) and number (t test: t (13) = 5.91, P < 0.05) of terminal dendritic branches (Fig. 2 B–D).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

CIE causes DLS dendritic hypertrophy and altered DLS synaptic plasticity. (A) CIE increased dendritic length. (B) DLS neuron reconstructions in air and CIE mice. (Scale bar: 50 µm.) (C and D) CIE increased terminal dendritic length and number. (E) CIE prevented induction of LTD induced by local afferent high-frequency stimulation (HFS, denoted by arrows) of cortical inputs to DLS. (F) Traces from E. (G) CIE decreased CB1R agonist-stimulated functional binding in DLS. (H and I) CIE increased levels of 2-AG, not AEA, in DLS. Data are means ± SEM, *P < 0.05 CIE vs. air, ^#P < 0.05 vs. pre-HFS1 baseline.

These data show that DLS neurons undergo significant and selective dendritic remodeling following chronic exposure to EtOH. In addition, the CIE-induced dendritic hypertrophy was restricted to DLS neurons and did not extend to cortical or limbic regions. The dendritic arbors of pyramidal neurons in the orbitofrontal cortex and the basolateral amygdala (Fig. S3) or (as previously shown in ref. 10) the infralimbic and prelimbic subregions of medial PFC all showed no differences between mice exposed to two cycles of CIE and air controls.

Altered DLS Synaptic Plasticity and CB1R Down-Regulation After CIE.

To test whether the CIE-induced structural abnormalities in DLS neurons were accompanied by functional changes, we examined long-term depression (LTD) at cortical glutamatergic inputs to DLS synapses, a key form of synaptic plasticity in the DLS (29). Three days after CIE, mice were assessed for LTD via local afferent high-frequency stimulation (HFS) of cortical inputs to DLS, comprising 2 × 1-s 100-pulse trains applied three times, once every 20 min. There was a borderline significant effect of CIE on LTD of evoked field potentials (ANOVA effect of group: F1,18 = 4.39, P = 0.05, followed by Fisher’s LSD post hoc tests, n = 10). In air controls, HFS significantly reduced population spike amplitude after each of the three sets of trains whereas HFS completely failed to induce LTD in CIE mice (Fig. 2 E and F). There was also partial attenuation of the efficacy (population spike magnitude) of synaptically induced activation in DLS following CIE (ANOVA group × stimulation amplitude interaction: F14,182 = 1.70, P = 0.058) (Fig. S4). These data extend prior reports of loss of corticostriatal LTD in rats measured 1 d after chronic EtOH exposure (30), by showing that this effect persists into the postacute-withdrawal period.

Current findings are reminiscent of the loss of (glutamate-dependent) LTD seen following chronic exposure to psychostimulant drugs of abuse that occurs in the DS as well as the nucleus accumbens (31⇓⇓–34). By limiting synaptic rescaling, drug-induced loss of LTD has been posited to impair the responsivity of accumbal circuits to new (e.g., cortically derived) environmental information, and thereby contribute to behavioral inflexibility in addiction (33). In light of the current findings in the DS after chronic alcohol, LTD loss may be a mechanism that is common to multiple drugs of abuse and would appear to be pervasive across striatal regions. At least in the case of chronic alcohol and the DS, deficient LTD may also be coupled with changes in other forms of striatal plasticity, including N-methyl-D-aspartate (NMDAR)–mediated LTP (35). Previous studies have found that chronic EtOH up-regulates NMDAR function and plasticity in the DS (albeit medially), in a manner that underlies dependence drinking (20, 21, 36). Moreover, various forms of learning, including some of the tasks used in the current study, are critically dependent upon DS NMDARs (37, 38). Thus, chronic alcohol may disrupt LTD-mediated plasticity in response to cortical input while at the same time unmasking NMDAR-dependent LTP and strengthening the associability of a subset of DS synapses.

The potential effect of CIE on NMDAR-mediated LTP in the DS remains to be tested. The mechanisms underlying the absence of LTD after CIE are also unclear but could reflect either a basal enhancement of LTD that occluded induction by HFS or an impairment of mechanisms subserving this form of plasticity in DLS. Endocannabinoid (ECB) signaling through CB1 receptors (CB1R) (19, 29) is a critical, obligatory mechanism for LTD in DLS, and blockade of CB1R in the DLS impairs certain forms of learning mediated by this region (39). CB1R signaling is also linked to our finding of DLS dysmorphology by prior work showing that loss of CB1R causes dendritic hypertrophy (40). We therefore asked whether CIE-induced alterations in plasticity were associated with changes in ECB signaling by performing CB1R agonist (R-WIN552320-2)-stimulated [³⁵S]GTPγS binding in DLS membranes (41) 3 d after CIE. There was a significant dose-dependent reduction in agonist-stimulated binding in DLS samples from CIE mice, relative to air controls (ANOVA group × dose interaction: F6,111 = 4.00, P < 0.01, n = 9) (Fig. 2G). The maximum effect of the agonist (E_max) (air = 314 ±14, CIE = 208 ± 4, t test: t (16) = 3.79, P < 0.01), but not the negative logarithm of the half-maximal effective concentration (pEC₅₀) (air = 6.8 ±0.2, CIE = 7.1 ±0.1), values were significantly decreased after CIE. These data are in line with the decreased CB1R binding and signaling previously reported in the DLS of rats chronically exposed to EtOH (42⇓–44), and in the DS of alcohol-dependent patients (45, 46).

Decreases in CB1R functional binding after chronic EtOH exposure are associated with elevated levels of the ECB 2-arachidonoylglycerol (2-AG) and, in some studies, anandamide (AEA) (18, 19). To test whether similar changes were present after CIE, we quantified DLS levels of 2-AG and AEA (3 d after CIE) via liquid chromatography-tandem mass spectrometry. We detected significantly higher levels of 2-AG (t test: t (17) = 5.12, P < 0.05, n = 9–10) (Fig. 2H), not AEA (t test: P > 0.05) (Fig. 2I), in the DLS of CIE-exposed mice, relative to air controls. There were no comparable changes in ECBs in other brain regions (medial PFC, dorsal hippocampus), indicating changes specific to DLS (Fig. S5). Taken together, these data suggest that CIE increases 2-AG levels in the DLS, leading to a compensatory down-regulation of CB1R signaling and a resultant loss of CB1R-mediated plasticity (LTD). CB1R down-regulation may also account for CIE-induced DLS dendritic dysmorphology, given prior studies have demonstrated that CB1R gene deletion is sufficient to cause similar dendritic hypertrophy in other brain regions (e.g., amygdala) (40).

CIE Facilitates DLS-Dependent Learning.

We next examined whether the structural and functional effects of CIE in the DLS affected learning processes previously shown to be mediated by this brain region (37, 47). We first assessed the effects of CIE on pairwise visual discrimination learning (Fig. 3A). This task requires mice to make a discriminative response to stimuli on a touch-sensitive computer screen to obtain food reward and is severely impaired by discrete excitotoxic lesions of the DLS (47). To reach a criterion of 85% correct responding, CIE mice made significantly fewer repetitive (“correction’) errors (t test: t (22) = 2.38, P < 0.05, n = 12) and slightly fewer errors overall (t test: t (22) = 1.16, P > 0.05), compared with air controls (Fig. 3 B and C). This improvement in discrimination learning in CIE mice was independent of changes in reaction times (air = 4.17 ± 0.45, CIE = 4.00 ± 0.29 s) or reward retrieval latencies (air = 2.20 ± 0.20, CIE = 1.84 ± 0.10 s), discounting gross alterations in motivation or locomotor activity.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

CIE facilitates DLS-dependent learning. (A–C) CIE before pairwise visual discrimination training decreased the number of errors and correction errors to criterion performance. (D–F) CIE before reversal training decreased the number of errors and correction errors to criterion performance, specifically during the late testing stage when percent correct performance was greater than chance. (G–I) CIE before Pavlovian conditioned approach testing produced a more rapid decrease in approaches during omission training. (J and K) CIE before stimulus–reward extinction training produced faster extinction. Data are means ± SEM, *P < 0.05 CIE vs. air.

Another cohort was tested on a reversal learning variant of the task, whereby mice had to reduce responding to the previously rewarded stimulus and learn to respond to the second, newly rewarded stimulus (Fig. 3D). Performance on this task is also highly sensitive to lesions of the DLS (47). CIE mice showed improved learning, relative to air controls, making significantly fewer errors (t test: t (22) = 2.42, P < 0.05, n = 11–13) and correction errors (t test: t (22) = 2.83, P < 0.01) (Fig. 3 E and F). Notably, the CIE-induced improvements emerged during the late phase of reversal when performance was already above chance levels and new learning predominates over perseveration (48, 49). This effect on late reversal coincides with the phase of reversal learning when the DLS is most strongly activated, as evidenced by high immediate-early gene activity (37).

These data are consistent with a CIE-induced facilitation of learning processes mediated by the DLS. The precise nature of the learning processes that are altered remains to be determined. One possibility is that CIE promotes DLS-mediated forms of stimulus–response (S–R) learning, such as habit formation (23, 50). Prior work has shown that rats given prolonged access to EtOH develop habit-like imperviousness to drinking taste-adulterated EtOH (51). Moreover, instrumental responding for EtOH becomes insensitive to reinforcer devaluation following extended behavioral training (23, 52). We previously found that EtOH-naïve mice that well-trained on the reversal task show habit-like insensitivity to reinforcer devaluation (47). When probed for reinforcer devaluation when performing at chance levels, neither CIE mice nor air controls reduced performance or errors rates following malaise-induced devaluation, suggesting that both groups may have already developed habit-like behavior (Fig. S6). Although these data cannot address whether CIE enhanced habit-like behavior in this task, they do suggest that CIE did not facilitate reversal by promoting action-outcome learning, which would be predicted to increase sensitivity to devaluation.

To further elucidate the behavioral effects of CIE, we tested another cohort for Pavlovian conditioned approach (53) (Fig. 3G). Following repeated presentation of a touchscreen stimulus that noncontingently signaled reward, CIE mice and air controls similarly increased their approaches to the stimulus over a second stimulus never associated with reward (ANOVA session effect: F3,69 = 23.31, P < 0.01, n = 12–13) (Fig. 3H). After attaining the conditioned approach response, we tested mice under an omission schedule in which stimulus approaches had to be withheld to obtain a reward. Although both groups decreased approaches during omission training, CIE mice were significantly quicker to do so than air controls (ANOVA group × session interaction: F2,46 = 4.35, P < 0.05) (Fig. 3I). This finding demonstrates that CIE improved the ability not just to associate and respond to a stimulus to obtain reward (as in discrimination and reversal learning), but also to learn to inhibit responses to stimuli no longer associated with reward.

We further tested learned inhibition of responding by measuring the extinction of the response to a visual stimulus that mice had previously (pre-CIE) learned to respond to, to obtain reward (Fig. 3J). CIE mice took significantly fewer sessions than air-exposed controls to reduce responding to a ≥77% extinction criterion (survival analysis: χ² = 3.94, df = 1, P < 0.05, n = 10–11) (Fig. 3K). Thus, CIE improved learned inhibition under extinction, suggesting that CIE might have enabled better predictions about the absence of reward. More generally, these behavioral results taken together demonstrate that the striatal neuroadaptations resulting from CIE are associated with improvements in learning across a range of diverse behavioral settings.

CIE Augments Learning-Related DS Single-Unit Activity.

We next sought to more directly connect the behavioral effects of CIE to changes in the function of neurons in the DS. To this end, we conducted in vivo single-unit neuronal recordings as mice performed the reversal learning task. After being trained to discrimination criterion, mice had 16-channel multielectrode arrays implanted into the DS, with one row of 8 channels targeting the most lateral part of DS and another row targeting the more medial area (Fig. 4A), and were then exposed to CIE. As no differences were observed between learning-related activity recorded from the two rows, data were collapsed for analysis.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

CIE augments learning-related DS single-unit activity. (A) Multielectrode arrays were implanted in DLS and a medial/DLS intermediate zone before CIE and reversal. (B) Recordings were made on early, mid, and late stages of reversal, matching CIE and air groups for percent correct performance. (C) Example raster plots during each state of reversal. (D) CIE produced a sustained increase in unit firing during late reversal. Data are means ± SEM, *P < 0.05 CIE vs. air.

Mice began reversal training 3 d after CIE. To assess how single-unit activity changes as reversal learning progressed, recordings were made on three separate sessions, corresponding to the start (“early”), midpoint (“mid”), and end (“late”) stages of training, when percent correct performance was poor, at chance or at criterion, respectively. In this way, CIE-exposed mice and air controls were matched for learning performance at each recording stage (Fig. 4B). Yoking the groups in this manner allowed for comparison of single-unit activity, while avoiding confounding effects of group differences in performance or elapsed time from CIE exposure (time since CIE: early training: air = 4 ± 1 d, CIE = 3 ± 1, midtraining: air = 12 ± 1, CIE = 11 ± 1, late-training: air = 26 ± 3, CIE = 23 ± 2).

At each training stage, recordings were made from seven CIE mice and six air controls (79–108 neurons per group). Baseline firing rate did not differ between the CIE (CIE = 4.4 ± 0.3 Hz) and air (4.3 ± 0.3 Hz) groups. To measure single-unit firing associated with learning, activity was aligned to touchscreen responses (1 s preresponse/6 s postresponse) and then segregated according to whether the response was correct or incorrect. Individual cells were categorized as event-related if there was a significant (t test) change in firing after a response was made, relative to preresponse baseline. The firing patterns in these units were broken out into the three learning stages and presented in perievent histograms (Fig. 4C).

Inspection of the histograms for all three stages revealed an increase in single-unit activity immediately before a correct response being made, followed by a rapid decrease and a second peak in activity ∼2 s postresponse. There was subsequently a steady decrease to below preresponse levels over the next 4 s. As learning progressed across stages, this decrease became more marked and the variability in activity lessened, suggesting greater coherence in activity within the network. The magnitude and dynamic profile of activity was essentially the same in CIE mice and air controls during the early stage (ANOVA effect of time: F139,12093 = 8.64, P < 0.01). On the midstage, there was a slight right-shifted delay in the postresponse peak in the CIE mice (ANOVA effect of time: F139,14317 = 26.16, P < 0.01). However, significant differences emerged on the late stage, with the CIE units showing sustained firing across the latter part of the epoch, relative to air controls, after mice made a correct response (ANOVA time × CIE interaction: F139,11537 = 1.26, P < 0.05). Post hoc analysis of the data in 1-s blocks revealed significantly higher firing in CIE than air controls during the final 3 s of the postresponse epoch (Fig. 4D). This increased firing in CIE mice during late reversal did not reflect a general hyperexcitability of recorded units because analysis of unit activity after incorrect responses was not different between groups during the late (or early or mid) stage (Fig. S7).

These data provide an in vivo correlate of the improved late-stage reversal learning after CIE and support the notion that CIE facilitates this and other forms of learning by strengthening DLS encoding of behavior.