Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2021 Dec 20.
Published in final edited form as: Alcohol Clin Exp Res. 2020 Aug 20;44(9):1896–1905. doi: 10.1111/acer.14418

Selective Deficits in Contingency-Driven Ethanol Seeking Following Chronic Ethanol Exposure in Male Mice

Jacqueline M Barker 1, Kathleen G Bryant 2, Alan Montiel-Ramos 3, Benjamin Goldwasser 4, Lawrence Judson Chandler 5
PMCID: PMC8686207  NIHMSID: NIHMS1762109  PMID: 32735737

Abstract

Background:

People with alcohol use disorders exhibit an overreliance on habitual response strategies which may result from a history of chronic alcohol exposure. Although habits are defined by behavior that persists despite changes in outcome value and in action–outcome relationships, most research investigating the effects of ethanol exposure on habits has focused only on outcome devaluation. A clear understanding of the effects of chronic alcohol exposure on the ability to flexibly update behavior may provide insight into the behavioral deficits that characterize alcohol use disorders.

Methods:

To dissociate the effects of chronic intermittent ethanol (CIE) exposure on contingency-mediated sucrose versus ethanol seeking, adult male C57Bl/6J mice were assigned to 2 separate experiments. In the first experiment, mice were trained to self-administer ethanol prior to 2 cycles of interleaved CIE exposure by vapor inhalation. In a second experiment, mice were trained to self-administer sucrose and ethanol in separate training sessions prior to 4 cycles of interleaved CIE. The use of contingencies to mediate reward seeking was assessed using a contingency degradation paradigm.

Results:

In mice trained to self-administer only ethanol, 2 weeks of CIE resulted in escalated self-administration. At this time point, CIE-exposed mice, but not air-exposed controls, exhibited ethanol seeking that was insensitive to changes in action–outcome contingency, consistent with habitual ethanol seeking. In mice trained to self-administer ethanol and sucrose rewards in sequential sessions, no escalation in self-administration across 4 weeks of CIE was observed. Under these conditions, neither Airnor CIE-exposed mice reduced ethanol seeking in response to contingency degradation. In contrast, sucrose seeking remained goal-directed.

Conclusions:

Our results suggest that chronic ethanol exposure impairs contingency-driven ethanol seeking more readily than sucrose-seeking behavior. Further, these findings indicate that the transition from contingency-mediated ethanol seeking occurs more rapidly than for sucrose seeking under similar ethanol exposure conditions.

Keywords: Habit, Chronic intermittent ethanol, Self-administration, Contingency, Degradation, Goal-directed behavior

Although alcohol abuse and dependence are tremendous personal, social, and economic burdens, pharmacotherapeutic strategies aimed at facilitating termination of and preventing relapse to drinking have proven to be largely ineffective. While not all alcohol-seeking behaviors in individuals with alcohol use disorders are habitual (Grodin et al., 2019), the development of alcohol use disorders may involve a transition from casual drug seeking to inflexible, uncontrolled ethanol-seeking habits (Barker et al., 2015a; Barker and Taylor, 2014; Koob and Volkow, 2016; O’Tousa and Grahame, 2014). While considerable research has investigated reduction of ethanol consumption using animal models (Becker, 2012), many of the therapeutic strategies identified have proven ineffective at reducing ethanol seeking in human alcoholics. Habitual ethanol seeking may be less sensitive to disruption than goal-directed behaviors (Hay et al., 2013). As such, it is critical to investigate the conditions in which overreliance on habitual behaviors occurs.

Individuals with alcohol use disorders show an overreliance on habitual response strategies (Sjoerds et al., 2013), though it is not clear whether this aberrant habit formation is a result of chronic alcohol use or may have been a predisposing factor to the development of alcoholism. While individual differences in corticolimbic striatal circuitry that regulates flexible reward-seeking behavior have been identified to predict the development of addiction-related behaviors, including habitual ethanol seeking (Barker and Taylor, 2014; Barker et al., 2014), others observe that chronic intermittent ethanol exposure by vapor inhalation results in the development of habitual ethanol-seeking behaviors that are resistant to devaluation (Barker et al., 2017; Renteria et al., 2020). These findings suggest that while propensity toward habit formation may contribute to the development of inflexible ethanol use, ethanol exposure itself can act directly to promote the loss of cognitive control over actions that characterizes AUDs. Beyond maintenance of ethanol use, this may contribute to more general deficits in behavioral flexibility in individuals with a history of chronic ethanol exposure.

Habitual behaviors are characterized by both the loss of sensitivity to change in outcome value and the loss of sensitivity to change in action–outcome relationships. Early in acquisition, reward seeking is mediated by its relationship to reward; that is, an action is performed in explicit relationship to its outcome (Adams and Dickinson, 1981; Balleine and Dickinson, 1998). Habits are considered to be maintained by stimulus–response relationships, rather than the action–outcome contingency (Dickinson, 1985). Thus, by definition, habits are insensitive to either the loss of action–outcome contingency (i.e., the action no longer produces the expected outcome) or to fluctuation in outcome value.

To date, the preponderance of the research investigating ethanol-seeking habits has focused on ethanol-induced deficits in sensitivity to change in value through the use of outcome devaluation paradigms (Barker et al., 2010; Corbit et al., 2012; Dickinson et al., 2002; Renteria et al., 2020; Serlin and Torregrossa, 2014). In a majority of studies where contingency degradation and outcome value are assessed, loss of sensitivity to change in contingency and value appears to occur in parallel for both ethanol and nondrug rewards (e.g., Barker et al., 2014, 2017b, 2017c; Gourley et al., 2016). However, there is growing appreciation for the need to parse the discrete aspects of behavioral inflexibility that drive maintenance of reward seeking, including dissociation of sensitivity to change in contingency and to change in outcome value (Schreiner et al., 2019). Ethanol habits are known to develop more rapidly than sucrose-seeking habits (Dickinson et al., 2002), but data suggest that ethanol exposure can promote the development of sucrose-seeking habits as well (Corbit et al., 2012; Renteria et al., 2018a), suggesting that ethanol acts to dysregulate shared circuitry to promote habitual behavior. Based on findings that chronic intermittent ethanol (CIE) exposure by vapor inhalation resulted in insensitivity to outcome devaluation in ethanol seeking (Renteria et al., 2020) and that either voluntary ethanol self-administration (Corbit et al., 2012) or CIE exposure (Renteria et al., 2018a) facilitated loss of sensitivity to outcome value in sucrose-seeking behavior, we investigated the hypothesis that CIE would impair use of contingencies to guide sucrose and ethanol-seeking behavior.

MATERIALS AND METHODS

Subjects

All subjects were adult male (between 70 days and 100 days at initiation) C57BL/6J mice at the start of the experiment and were group-housed in the Medical University of South Carolina (MUSC) vivarium under reverse 12-hour:12-hour light conditions, and all behavioral testing took place during the dark cycle. Mice underwent brief water restriction and temporal food restriction as described below. Weights were maintained at >95% free-feeding weight throughout the experiment. All experiments were approved by the Institutional Animal Care and Use Committee (IACUC) at MUSC.

Operant boxes

Mice were trained in standard self-administration boxes from Med Associates (St. Albans, VT) with 2 plexiglas and 2 metal walls. Floors were metal grid floors, and each box had 2 retractable levers that were extended during instrumental training sessions. One was assigned as the “active” lever that would result in reinforcer delivery. On the same wall, reinforcers were delivered by syringe pump to a well in a magazine equipped with a lickometer to measure consummatory behavior.

Experiment 1. Effects of 2 weeks of CIE on contingency-mediated ethanol seeking.

Operant ethanol self-administration.

A postprandial consumption procedure was used to facilitate acquisition of ethanol self-administration. For this procedure, mice received limited access to food and water. Specifically, 1 hour prior to beginning the self-administration sessions, mice received 15-min access to standard home-cage chow (Envigo Teklad 2918) and the water bottles were removed from the cages during the self-administration sessions. Weights were monitored, and additional food was provided after the behavior session to maintain free-feeding weights. In these sessions, mice were trained to respond (lever press) on a fixed ratio 1 (FR1) schedule for access to a 10% unsweetened ethanol solution in a 30-minute session. After establishing baseline responding, mice were returned to ad libitum food and water access and responding was maintained on a FR1 schedule. After confirming stable responding for 5 days on FR1, mice for CIE experiments went through the procedure as described below and then were returned to an FR1 self-administration schedule for 2 days. Mice were then graduated to a random interval 30 seconds (RI30) schedule for 2 days, in which a 10% ethanol reinforcer was delivered at the first lever response after a randomly generated interval, averaging 30 seconds.

CIE procedure.

Mice were assigned in a counterbalanced fashion to either a CIE or Air-exposed control (Air) group based on response rates during acquisition. CIE mice underwent 2 cycles of CIE for 16 h/d for 4 consecutive days (Barker et al., 2015b; Lopez et al., 2014), with 8 hours of withdrawal separating each exposure. Each of the 2 cycles was separated by 5 days of operant self-administration. Immediately prior to placement in the vapor chambers, all mice received equivalent injections of the alcohol dehydrogenase inhibitor pyrazole (1 mmol/kg) to stabilize the levels of ethanol in the blood at elevated levels in the vapor-exposed group. Blood ethanol concentrations (BEC) were assessed weekly from retro-orbital blood using an Analox instrument. These assessments revealed that the BEC levels during the CIE procedure ranged from 150 to 350 mg/dl (BEC: 300.1 ± 27.41 mg/dl).

Contingency degradation test.

The use of action–outcome contingencies to guide behavior was assessed using a contingency degradation paradigm in which the relationship of the action and reinforcer delivery was disrupted through the provision of noncontingent reinforcement at a rate determined by each individual animal’s reinforcement rate on the prior day. To accomplish this, mice (n = 11 Air controls, 9 CIE-exposed; 3 mice were excluded for failing to maintain an average of >5 reinforcers earned per session) underwent testing in which conditions were identical to training except that reinforcer delivery was independent of responding. Responding on both the active and inactive levers was recorded, but did not result in, or prevent, reinforcer delivery. Rather, reward was delivered on a fixed time schedule, as determined by the rate of reward delivery earned on the previous session in which the contingency was intact. Thus, the number and rate of rewards matched the nondegraded test session and varied by animal.

Experiment 2. Comparison of 4 weeks of CIE on contingency-mediated sucrose versus ethanol seeking.

Operant self-administration and CIE procedure.

For these experiments, mice were trained during separate sequential sessions to respond for sucrose and ethanol. To accomplish this, mice were food-restricted as described above and then underwent 15-minute sucrose self-administration sessions in the first part of the dark cycle (AM) in which the mice were first trained to respond (lever press) on an FR1 schedule for access to 2% sucrose. This order was not counterbalanced to avoid sucrose self-administration during acute intoxication or withdrawal. This low concentration of sucrose was selected via pilot studies to match response rates for sucrose and ethanol. Only the active lever, on which a press resulted in sucrose delivery, was made available for responding, and no inactive lever was presented. At least 3 hours later, but still during the dark cycle, the same mice underwent the postprandial consumption procedure described above. After food exposure and water restriction, the mice were returned to the same operant chambers, but now the opposite lever was extended. Responding on the ethanol lever was also initially reinforced on an FR1 schedule, resulting in access to a 10% unsweetened ethanol solution. After establishing baseline responding for both sucrose and ethanol, mice were returned to ad libitum food and water access and responding was maintained on the FR1 schedule. After confirming stable responding, mice were matched by response rates for sucrose and ethanol to undergo either CIE or Air exposure as described above under Experiment 1, targeted a range between 150 and 350 mg/dl (mean BEC: 199.1 ± 7.97 mg/dl). Self-administration occurred between each cycle of CIE as for Experiment 1. Following 4 cycles of CIE, mice were then graduated to a RI30 schedule for 2 days prior to contingency degradation testing.

Contingency degradation test.

As described above, the use of action–outcome contingencies to guide behavior was assessed using a contingency degradation paradigm. Contingency-driven sucrose and ethanol-seeking behaviors were assessed on separate days. While all mice underwent identical training procedures, with one session of sucrose responding and one session of ethanol responding each day, a subset of mice acquired responding for only one reward. Thus, for these mice, the use of contingencies to guide behavior was only assessed for one reward. This was evenly distributed for sucrose and ethanol rewards. Groups consisted of 8 Air control and 7 CIE-exposed mice for assessment of ethanol seeking and 6 air-exposed controls and 10 CIE-exposed mice for sucrose-seeking behavior.

Statistical Analyses

Data were analyzed in GraphPad Prism v8.2 (San Diego, CA, USA) using 1-sample t-tests (comparison to baseline), unpaired t-tests (Air vs CIE groups), or ANOVA. Day of training and testing condition were handled as repeated-measures variables. Ethanol exposure was a between-subjects variable. Mixed-models ANOVA was used to handle missing data. Significant findings were followed with Sidak’s (for contingency degradation data) or Dunnett’s (for acquisition data to compare to day 1 of training) post hoc tests to deconstruct effects.

RESULTS

Experiment 1

Escalation of self-administration of ethanol across chronic intermittent ethanol exposure.

To assess responding for ethanol across training in CIE-exposed and Air control mice, lever presses were analyzed using a mixed-model ANOVA to accommodate missing data (CIE × session; Fig. 1A). Main effects of CIE, F(1, 20) = 8.561, p = 0.0084, and session, Greenhouse–Geisser corrected; F(3.825, 70.76) = 7.685, p < 0.001, were observed, as well as a significant CIE × session interaction, F(12, 222) = 3.054, p < 0.001. Corrected post hoc comparisons indicate that there were no individual sessions during which responding was significantly different in CIE vs Air control mice (all p values >0.1 except Session 9 where p = 0.0829), suggesting responding for ethanol was elevated throughout training in the CIE-exposed animals. For Air control mice, responding was never significantly enhanced compared to the first session, indicating that responding did not escalate in these animals (all p values >0.1 with the exception of session 7, 8, and 9 in which the p values were 0.0692, 0.0563, and 0.0994, respectively). For CIE-exposed mice, response rates were significantly higher on session 5 (p = 0.0424), 6 (p = 0.0342), 7 (p = 0.0296), 8 (p = 0.474), 9 (p = 0.0182), 12 (p = 0.014), and 13 (p = 0.0014), consistent with escalations in responding. Also consistent with this, the number of total lever presses across all conditioning sessions was significantly higher in CIE-exposed mice than in Air controls, 2-tailed unpaired t-test; t (20) = 3.136, p = 0.0052. This change was observed during the FR1 sessions that proceeded the first cycles, but was moderated following the second cycle of CIE when reinforcement schedules were transitioned from FR1 to RI30. This may suggest response schedule-dependent sensitivity to CIE-induced escalation of self-administration.

Fig. 1.

Fig. 1.

Open in a new tab

Two weeks of chronic ethanol exposure promotes loss of sensitivity to changes in ethanol-seeking contingency. (A) CIE-exposed mice exhibit higher rates of ethanol self-administration than Air controls. (B) At this testing time point, the control mice that did not undergo CIE by vapor inhalation were sensitive to changes in action–outcome contingency and reduced their response rates when the contingency was degraded. In contrast, CIE-exposed mice were insensitive to loss of action–outcome contingency and did not reduce responding. Because baseline response rates were significantly different between Air controls and CIE-exposed mice, we applied 2 normalization strategies. Air control mice reduced their responding compared to baseline (C), while CIE-exposed mice did not. Similarly, Air controls responded at a lower percentage of their baseline than CIE-exposed mice. When assessing distribution of responding across a test session where the contingency was intact versus degraded (D), the Air control mice responded more in the nondegraded session than the degraded session, while CIE-exposed mice distributed responding evenly. n = 11 Air-exposed, 9 CIE-exposed. *p < 0.05, **p < 0.01 vs control, †p < 0.05 vs baseline, Δp < 0.05 vs day 1.

In contrast to recent findings reporting significant escalation in ethanol consummatory behavior (i.e., licking) following CIE (Renteria et al., 2020), we did not observe an escalation in licking following CIE (data not shown). A mixed-model ANOVA indicated no significant CIE × session interaction, F(12, 176) = 0.9448, p = 0.5037. A trend toward a main effect of CIE was observed, F(1, 17) = 3.960, p = 0.0629, such that there was a nonsignificant increase in licking in CIE-exposed mice across training. A main effect of session was observed, F(2.949, 43.26) = 4.227, p = 0.0108; Greenhouse–Geisser corrected. Notably, although differences in licking were not present, a significant main effect of CIE, mixed-model ANOVA; F(1, 20) = 6.160, p = 0.0221, and a CIE × session interaction on the overall number of reinforcers earned were observed, F(12, 222) = 7.051; p = 0.003. A main effect of session was also observed, Greenhouse–Geisser corrected; F(3.783, 69.99) = 7.051, p < 0.001. For Air controls, reinforcer number was not significantly greater than the first session for any subsequent sessions (all p values >0.05). For CIE-exposed mice, significant increases in reinforcer delivery were observed on session 8 (p = 0.0484), 12 (p = 0.0250), and 13 (p = 0.0365). Although initial training sessions were on a FR1 schedule, the differential number of lever presses and reward deliveries may reflect nonreinforced presses that occurred during the 1.67 seconds following a lever press when the reward was being delivered. This may suggest differential burst patterns of responding in CIE versus Air control mice that should be elucidated in future research. As no measurable ethanol volumes were left in wells, grams of ethanol consumed per body weight was estimated based on reinforcer deliveries (mean for Air controls: 0.036 ± 0.005 g ethanol/kg body weight; CIE-exposed mice: 0.066 ± 0.011 g ethanol/kg body weight). We observed a CIE × session interaction on g/kg ethanol consumed, mixed model; F(12, 228) = 2.590, p = 0.003, as well as main effects of session, Greenhouse–Geisser corrected; F(3.950, 75.05) = 7.335, p < 0.001, and CIE exposure, F(1, 20) = 6.114, p = 0.0225. Post hoc analyses indicate that g/kg ethanol consumption did not escalate in Air controls (all p values> 0.1). Mice that underwent CIE exhibited significantly greater g/kg ethanol consumption versus the first session in sessions 8 (p = 0.0481), 12 (p = 0.0244), and 13 (p = 0.0381).

Two weeks of chronic ethanol exposure impairs use of contingencies to guide ethanol-seeking behavior.

To determine whether 2 weeks of chronic intermittent ethanol exposure by vapor inhalation impacted the ability to use contingencies to guide behavior, male mice underwent a within-subjects assessment of reward-seeking behavior under contingency-degraded versus control conditions (Fig 1 B). A rmANOVA (CIE × degradation) of total lever presses indicated a significant CIE condition by degradation session interaction, F(1, 18) = 4.541, p = 0.0471, but no main effects of CIE, F(1, 18) = 0.6311, p = 0.4373, or degradation, F(1, 18) = 3.303, p = 0.0858, were observed. Post hoc comparisons using Sidak’s test indicate that in Air controls, response rates were significantly lower during contingency-degraded test conditions than under a session when responses were intact (p = 0.0173), which is consistent with the use of contingencies to regulate behavior. In contrast, no difference was observed between the degraded and nondegraded tests in CIE-exposed mice, indicating a deficit in the use of contingences to regulate ethanol-seeking behavior (p = 0.9728). As differences in response rates during training sessions were observed in CIE- and Air control mice, we performed 2 normalization assessments. First, we normalized responding in the degraded session as a percent baseline responding in the nondegraded test session. Analysis of these data revealed that CIE-exposed mice responded at a higher percentage of their baseline than Air controls, 2-tailed unpaired t-test; t(18) = 2.652, p = 0.0162; Fig. 1C. Using this normalization, we found that responding for Air controls was significantly lower than 100% of their own baseline, 1-sample t-test; t(10) = 4.118, p = 0.0021, while responding for CIE-exposed mice was not different from 100%, t(8) = 0.6037, p = 0.5628. Because the order of testing was not counterbalanced, a second normalization was performed to account for any potential shift in responding between days as the order of testing was not counterbalanced (Fig. 1D). In this analysis, a “habit score” (Renteria et al., 2018a) was calculated by dividing the responding in the nondegraded test session by the sum of all responding in the degraded and nondegraded test sessions [habit score = degraded/(degraded + nondegraded)], where a score of 0.5 is consistent with equal distribution of responding in the degraded and nondegraded conditions. This revealed that CIE-exposed mice had higher habit scores compared to Air controls, t(18) = 2.504, p = 0.0221. Similar to findings comparing baseline response rates, we observed that Air controls had a greater proportion of their responding during the nondegraded session than during the degraded session, 1-sample t-test; t(10) = 3.524, p = 0.0055, while CIE-exposed mice distributed their responding evenly between these 2 test sessions, t(8) = 0.02015, p = 0.9844. No significant differences in consummatory behavior were observed during the contingency degradation session, rmANOVA on CIE × degradation; main effect of CIE: F(1, 15) = 0.3224, p = 0.5786; main effect of degradation: F(1, 15) = 0.6121, p = 0.4662; interaction: F(1, 15) = 0.4139, p = 0.5297.

Habitual behavior and loss of sensitivity to change in contingency may result from increased experience with action–outcome relationships. Because overall higher rates of responding in CIE-exposed mice were observed during the acquisition period, we next assessed whether habit scores or percent baseline responding was significantly correlated with the number of lever presses. In both cases, no significant correlations were observed (all p values >0.05), though a trend was observed in the relationship between percent baseline pressing and total lever pressing in Air control mice (r2 = 0.3070, p = 0.0966) (data not shown). Although this relationship was not statistically significant, this apparent trend may indicate that higher response rates in CIE-exposed mice were associated with increased habit-like behavior.

Experiment 2

Escalation of neither sucrose nor ethanol self-administration was observed in mice trained to self-administer both rewards in sequential sessions.

Findings from other groups suggest that ethanol exposure in CIE paradigms or self-administration can promote habitual sucrose seeking as assessed by outcome devaluation (Corbit et al., 2012; Renteria et al., 2018a). Habitual ethanol seeking also develops more rapidly than sucrose-seeking habits (Dickinson et al., 2002). We sought to assess whether sucrose-seeking and ethanol-seeking behavior became contingency-insensitive under similar training conditions when mice received matched ethanol exposure. Thus, mice were trained to self-administer sucrose and ethanol in separate sessions. To determine whether 4 weeks of CIE resulted in escalated responding for sucrose or ethanol in mice trained to self-administer both rewards in different 15-min-long sessions, data were analyzed by either mixed-model ANOVA (degradation × CIE; ethanol; Fig. 2A) or rmANOVA (degradation × CIE; sucrose; Fig. 2B). For ethanol rewards, a main effect of session, Greenhouse–Geisser corrected; F(4.484, 60.54) = 6.159, p < 0.001, was observed, but no main effect of CIE, F(1, 14) = 0.6372, p = 0.4381, or CIE × session interaction, F(22, 297) = 0.9382, p = 0.5471. Post hoc comparisons indicate that responding in sessions 17, 20, 24, and 25 was significantly higher than the first session (Sidak’s, p = 0.0041, p = 0.0073, p = 0.0106, p = 0.0178, respectively). Similarly, for sucrose rewards, a main effect of session, Greenhouse–Geisser corrected; F(3.967, 55.54) = 9.521, p < 0.001, was observed. No main effect of CIE, F(1, 14) = 1.279, p = 0.2771, or CIE × session interaction, F(22, 308) = 0.6041, p = 0.9205, was present. Only responding during sessions 13, 14,18, and 19 was significantly greater than responding on day 1 (Sidak’s post hoc; p = 0.0461, p = 0.0178, p = 0.0095, p = 0.0004, respectively). Differences in reinforcer delivery paralleled differences in responding. For ethanol rewards, no main effect of CIE, mixed-model analysis; F(1, 14) = 0.004014, p = 0.9504, or CIE × session interaction on reinforcer number was observed, F(22, 296) = 0.5585, p = 0.9476. There was a significant main effect of session on ethanol reward number, Greenhouse–Geisser corrected; F(2.628, 35.35) = 4.053, p = 0.0176. Post hoc analyses indicate that reinforcer numbers were greater than session 1 during session 17 (p = 0.0065), 20 (p = 0.0096), and 21 (p = 0.0150). Similarly, for sucrose rewards, no effect of CIE, mixed-model ANOVA; F(1, 15) = 1.009, p = 0.3311, or CIE × session interaction on sucrose reinforcer number was observed, F(22, 312) = 1.073, p = 0.3747. A main effect of session on number of sucrose rewards was observed, Greenhouse–Geisser corrected; F(3.439, 48.77) = 6.538, p < 0.001. Post hoc comparisons indicate that sucrose reinforcer delivery number was greater than session 1 on session 14 (p = 0.0358), 18 (p = 0.0167), and 19 (p = 0.0012). We further examined training- and CIE-induced differences g/kg of ethanol consumed (mean ethanol consumption for Air controls: 0.035 ± 0.004 g ethanol/kg body weight; CIE-exposed: 0.036 ± 0.004 g ethanol/kg body weight). As with reinforcer number and lever presses, results indicated no effect of CIE, mixed model; F(1, 14) = 0.02307, p = 0.8814 or session × CIE interaction, F(22, 295) = 0.6053, p = 0.9195. A main effect of session was observed, Greenhouse–Geisser corrected; F(2.765, 37.08 = 3.942, p = 0.0177. Post hoc analyses indicate that g/kg consumption was greater than session 1 on sessions 17 (p = 0.0066), 20 (p = 0.0102), and 21 (p = 0.0240).

Fig. 2.

Fig. 2.

Open in a new tab

Four weeks of CIE does not alter self-administration of either sucrose or ethanol. In mice trained to self-administer ethanol in 15-minute sessions across 4 weeks of CIE, escalation was not observed for self-administration of either ethanol (n = 8 Air-exposed, 7 CIE-exposed; A) or sucrose (n = 6 Air-exposed, n = 10 CIE-exposed; B).

Similar to findings following 2 weeks of CIE, no escalation in licking behavior was observed for ethanol or sucrose rewards following 4 weeks of CIE (data not shown). For ethanol reward, no CIE by session interaction, F(22, 254) = 1.133, p = 0.3118 or main effect of CIE, F(1, 12) = 0.09205, p = 0.7668, was observed. However, there was a main effect of session, F(2.997, 34.60) = 7.335, p < 0.001; Geisser–Greenhouse corrected. For sucrose reward, no CIE × session interaction, F(22, 268) = 0.74, p = 0.7887, or main effect of CIE, F(1, 13) = 0.0685, p = 7976, was observed. Again, there was a main effect of session, F(3.354, 40.86) = 0.1525, p < 0.001; Geisser–Greenhouse corrected.

Following alternated self-administration and 4 cycles of CIE, neither Air- nor CIE-exposed mice reduced responding in contingency degradation tests.

To determine whether differences in the use of contingencies to drive reward-seeking behavior existed following protracted self-administration and 4 cycles of CIE, sensitivity to contingency degradation for an ethanol reward and a sucrose reward was assessed. Following 4 weeks of self-administration with intermittent CIE, both Air controls and CIE-exposed mice exhibited ethanol-seeking behavior that was not reduced following contingency degradation. A rmANOVA (degradation × CIE) indicated no main effect of CIE, Fig. 3A; F(1, 13) = 0.6531, p = 0.4335, or test, F(1, 13) = 2.121, p = 0.1690, nor a CIE by degradation interaction, F(1, 13) = 1.006, p = 0.3342. Similarly, when assessing normalized data as a % baseline, Fig. 3 B; t(14) = 2.027, p = 0.0622, or as a habit score, Fig. 3 C; t(14) = 0.9973, p = 0.3355, no differences were observed between the CIE-exposed vs Air controls. While the difference in percent baseline analyses was not statistically significant, these findings suggest a trend toward higher responding in the degradation session by Air controls than CIE-exposed mice. Furthermore, both means were greater than 100%, consistent with a failure to reduce responding when the contingency was degraded in all mice. To expand upon analyses comparing responding in CIE-exposed versus Air control mice, 1-sample t-tests were performed to determine whether responding differed from their own baseline (Fig. 3B) or was differentially distributed between degraded and nondegraded test session (Fig. 3C). When determining if responding was significantly different from 100% of baseline, a 1-sample t-test indicated that the Air controls elevated responding for ethanol during the degradation session as compared to baseline, Fig. 3B; 1-sample t-test; comparison to 100% t(8) = 2.684, p = 0.0277, while response rates in CIE-exposed mice did not differ from 100%, 1-sample t-test; comparison to 100%; t(6) = 0.1438, p = 0.8904. This indicates that nondependent mice escalated response rates when the contingency was degraded, which is potentially consistent with frustrative, escalated responding resulting from detection of the loss of contingency rather than an insensitivity to change in contingency. Similarly, Air controls did not distribute their responding evenly across the test sessions, responding at higher rates for ethanol during the degradation test when the action–outcome relationship was degraded, 1-sample t-test, significantly different than habit score of 0.5; Fig. 3 C; t(8) = 2.761, p = 0.0256, while response rates in CIE-exposed mice were evenly distributed, 1-sample t-test; comparison to habit score of 0.5; t(6) = 0.3122, p = 0.7655. No significant differences in consummatory behavior were observed during the contingency degradation session [rmANOVA on CIE × degradation; main effect of CIE: F(1, 11) = 0.0676, p = 0.7996; main effect of degradation: F(1, 11) = 0.1302, p = 0.7251; interaction: F(1, 11) = 0.0874, p = 0.7729.

Fig. 3.

Fig. 3.

Open in a new tab

Ethanol-seeking behavior after 4 weeks of chronic ethanol exposure. (A) Air control and CIE-exposed mice made similar numbers of responses during degraded and nondegraded test sessions. (B) Air control mice elevated responding over baseline when the action–outcome contingency was degraded, while CIE-exposed mice continued to respond at baseline levels. (C) Air controls made a greater proportion of their responses during a degraded test session than a nondegraded test session, whereas CIE-exposed mice made similar numbers of responses in both tests. n = 8 Air-exposed, 7 CIE-exposed. †p < 0.05 vs baseline.

Four weeks of CIE is not sufficient to drive loss of contingency-mediated sucrose seeking.

In the same mice trained to self-administer both sucrose and ethanol, the use of contingencies to mediate sucrose seeking was assessed following the same 4-week exposure paradigm described above. In contrast to responding for ethanol, responding for sucrose seeking remained sensitive to change in contingency for Air control and CIE-exposed mice. A rmANOVA (CIE × degradation) indicated a main effect of degradation, Fig. 4A; F(1, 14) = 18.51, p < 0.001, consistent with reduced responding when the action–outcome relationship was degraded. Although a trend toward a main effect of CIE was observed, F(1, 14) = 3.336, p = 0.0892, that may suggest increases in responding in both test sessions in Air controls compared to CIE-exposed mice, this did not reach significance. No CIE by degradation interaction was observed, F(1, 14) = 0.5036, p = 0.4896. No differences in percent baseline responding, Fig. 4B; t(14) = 0.4140, p = 0.8419, or habit score were observed, Fig. 4 C; t(14) = 0.1565, p = 0.8779, suggesting similar reductions in responding in the face of changed contingencies in Air control and CIE-exposed mice. When comparing responding to their own baselines, Air control mice significantly reduced responding for sucrose in the degraded test session, Fig. 4B; t(9) = 3.193, p = 0.011, while response rates in the CIE-exposed mice were not significantly different than their own baselines, t(9) = 1.502, p = 0.1672. This observation is consistent with reduced sensitivity to change in contingency in CIE-exposed mice when responding for sucrose. When using the alternative normalization method for comparing distribution of responses across the degraded and nondegraded test sessions, both Air controls and CIE-exposed mice responded less in the degraded test session than the session where the contingency was intact, Fig. 4 C; Air: t (5) = 2.591, p = 0.0488, CIE: t(9) = 2.481, p = 0.0349. No significant differences in consummatory behavior were observed during the contingency degradation session [rmANOVA on CIE × degradation; main effect of CIE: F(1, 12) = 2.356, p = 0.1507; main effect of degradation: F(1, 12) = 0.1547, p = 0.7010; interaction: F(1, 12) = 1.855, p = 0.1982.

Fig. 4.

Fig. 4.

Open in a new tab

Sucrose-seeking behavior following 4 weeks of chronic ethanol exposure. (A) Air controls and CIE-exposed mice reduced responding for sucrose when the action–outcome relationship was degraded. (B) Air control mice reduced responding compared to baseline when the action–outcome contingency was degraded. This did not reach significance in CIE-exposed mice. (C) Both Air controls and CIE-exposed mice made a greater proportion of their responses in a nondegraded test session. n = 6 Air-exposed, n = 10 CIE-exposed. ***p < 0.001, †p < 0.05 vs baseline.

DISCUSSION

The results of the present study demonstrate that ethanol dependence impairs the ability to use action–outcome contingencies to guide ethanol-seeking behavior. At a time point where control, nondependent mice were able to flexibly regulate their behavior in response to changing contingencies, mice exposed to 2 weeks of CIE continued to respond at baseline levels. This suggests that CIE induces habitual ethanol seeking that is insensitive to changes in contingency. Following 4 weeks of CIE that were interspersed with ethanol self-administration, both controls and dependent mice developed ethanol-seeking behaviors that were consistent with habitual behavior. Importantly, this indicates that the Air control mice may have developed contingency-insensitive habits following repeated training on these schedules, but that CIE facilitated the transition from actions to habits. This interpretation is complementary to findings from others that CIE induces outcome devaluation-insensitive ethanol-seeking behaviors (Renteria et al., 2020), indicating that a history of chronic ethanol exposure may facilitate habitual behaviors that are both outcome and contingency-insensitive. This is further consistent with reports that ethanol-seeking habits develop more rapidly than sucrose-seeking habits (Dickinson et al., 2002). Alternatively, while Air controls do not reduce responding when the contingency is degraded—the common definition of habitual behaviors—it does appear that they are sensitive to this change in contingency and elevate their responding when the action–outcome relationship is degraded. This suggests that Air controls are indeed tracking the contingency even after 4 weeks of responding, but their adaptive behavior has transitioned; rather than reducing response rates in the face of the loss of contingency, they escalate their ethanol-seeking behavior. This is potentially consistent with an extinction burst, known to occur in a subset of conditions when the action–outcome relationship is removed (Harris et al., 2007; Lerman and Iwata, 1995). Notably, we have not previously observed extinction bursts in ethanol-seeking behavior (Barker et al., 2012), but this task is distinct from traditional extinction paradigms as reward is delivered in a noncontingent manner and may yield distinct patterns of behavior following extensive training. This pattern was not present in CIE-exposed mice that did not modify responding when the contingency was degraded.

Under these training parameters, 4 weeks of CIE did not drive the development of contingency-insensitive sucrose-seeking behavior. This is particularly interesting as it suggests that total ethanol exposure alone does not necessarily result in habitual behavior, as mice in these studies both underwent CIE and ethanol self-administration. Notably, while BECs were higher in Experiment 1 than Experiment 2, in both instances BECs were consistent with those that induce outcome devaluation-insensitive sucrose seeking (Renteria et al., 2018). In these studies, sucrose self-administration always occurred earlier in the day to avoid effects of acute ethanol intoxication or withdrawal; however, this procedural difference may contribute to the differences in the transition to habitual behavior. This may suggest that the specific timing of ethanol self-administration relative to task performance or that the experience of the action–outcome contingency is a key contributor to impairments in subsequent use of contingencies. Alternatively, ethanol reinforcers may facilitate habit formation in a way that is independent of the effects of ethanol on learning and memory.

These findings stand in contrast to data suggesting that under similar ethanol vapor exposure conditions, 4 weeks of CIE impaired the use of outcome value-mediated sucrose seeking (Renteria et al., 2018a). While one interpretation of these findings is that CIE impairs value-driven sucrose seeking while sparing contingency-driven sucrose seeking, there are several key caveats to consider. First, in our studies, the initial action–outcome relationship was acquired ahead of chronic ethanol exposure. In studies from Renteria and colleagues (Renteria et al., 2018a), mice underwent CIE exposure prior to acquisition of the action–outcome relationship. The timing of CIE may play a substantial role in subsequent task performance. For example, we previously reported that when stimulus–outcome learning occurred following CIE, deficits in cue-guided behavior were observed (Barker et al., 2015b). Similarly, others have found that CIE impaired Pavlovian-to-instrumental transfer when acquisition followed ethanol exposure, but that behavior was intact if acquisition occurred prior to CIE (Depoy et al., 2014). This suggests that the initial acquisition of sucrose self-administration before CIE in our study versus after CIE in the Renteria study may contribute to this dissociation. However, there are instances in which loss of sensitivity to outcome value and to action–outcome contingency dissociates (Corbit et al., 2002; Lex and Hauber, 2010), suggesting that these aspects of habitual behavior may be differentially impacted by ethanol. A key difference between contingency degradation and outcome devaluation paradigms includes the presentation of the reinforcer. While the reinforcer is absent during outcome devaluation paradigms, it is presented during the contingency degradation paradigm. This may serve as a cue that promotes goal-directed reward seeking. Thus, this may drive an escalation in responding that appears consistent with habitual behavior but may indeed represent a competing response. It is possible that a history of CIE renders mice more susceptible to initiation of reward seeking by presentation of the reward itself, thus driving maintained responding. These task parameters are important to consider and a complete dissociation of the mechanisms underlying forms of inflexible behavior requires their consideration.

In addition to dependence-inducing ethanol exposure paradigms, others have found that self-administration of lower doses of ethanol can also promote habitual sucrose seeking as assessed by outcome devaluation (Corbit et al., 2012). In these studies, the timing of ethanol exposure was in close apposition to sucrose self-administration—each day, following sucrose self-administration rats received access to ethanol 4 hours after administration. Thus, ethanol consumption occurred during the late consolidation of the memory, potentially impacting ongoing learning, rather than before task acquisition (Renteria et al., 2018a, 2020), or after acquisition as in the current work.

Together, these results indicate that chronic ethanol exposure by vapor inhalation selectively impairs contingency-mediated ethanol seeking in adult male mice when initial task acquisition occurs prior to chronic ethanol exposure by vapor inhalation. The results presented here were limited to male mice, and further characterization of similarities and differences in the consequences of ethanol exposure on behavioral flexibility in female mice is necessary for a complete understanding. Female mice have been shown to be relatively resistant to ethanol, but not sucrose, habit formation in the absence of CIE (Barker et al., 2010; Quinn et al., 2007), while others found no obvious sex differences in the effects of CIE on behavioral flexibility (Renteria et al., 2018b), suggesting that sex differences may be reinforcer-, exposure-, and task-dependent. Beyond identification of the aspects of inflexible reward seeking that may be particularly susceptible to perturbation of reward-seeking behavior, these findings highlight the need for precision in interpretation of behavioral paradigms and the integration of behavioral analyses with drug exposure paradigms. While extant literature suggests that the neural substrates underlying habitual ethanol and nondrug rewards overlap (Corbit and Janak, 2016; Corbit et al., 2012; Fanelli et al., 2013; Gianessi et al., 2019, 2020; Morisot et al., 2019), it will be critical to precisely dissociate the impacts of ethanol on these substrates and their discrete contributions to contingency- versus value-insensitive behaviors.

ACKNOWLEDGMENTS

This research was supported by NIH grant AA024499 (JMB) and P50-AA10761 (LJC) and Pilot Funds from the Charleston Alcohol Research Center (LJC and JMB).

Footnotes

CONFLICT OF INTEREST

The authors declare no conflict of interest.

Contributor Information

Jacqueline M. Barker, Department of Pharmacology and Physiology, Drexel University College of Medicine, Philadelphia, Pennsylvania.

Kathleen G. Bryant, Department of Pharmacology and Physiology, Drexel University College of Medicine, Philadelphia, Pennsylvania.

Alan Montiel-Ramos, Department of Neurosciences, Medical University of South Carolina, Charleston, South Carolina..

Benjamin Goldwasser, Department of Neurosciences, Medical University of South Carolina, Charleston, South Carolina..

Lawrence Judson Chandler, Department of Neurosciences, Medical University of South Carolina, Charleston, South Carolina..

REFERENCES

  1. Adams C, Dickinson A (1981) Actions and habits: Variations in associative representations during instrumental learning, in Information processing in animals: Memory mechanisms (Spear NE& Miller RReds) pp. 143–165. Erlbaum, Hillsdale, NJ. [Google Scholar]
  2. Balleine BW, Dickinson A (1998) Goal-directed instrumental action: contingency and incentive learning and their cortical substrates. Neuropharmacology 37:407–419. [DOI] [PubMed] [Google Scholar]
  3. Barker JM, Bryant KG, Osborne JI, Chandler LJ (2017) Age and sex interact to mediate the effects of intermittent, high-dose ethanol exposure on behavioral flexibility. Front Pharmacol 8:1–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Barker JM, Corbit LH, Robinson DL, Gremel CM, Gonzales RA, Chandler LJ (2015a) Corticostriatal circuitry and habitual ethanol seeking. Alcohol 49:817–824. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Barker JM, Glen WB, Linsenbardt DN, Lapish CC, Chandler LJ. (2017) Habitual behavior is mediated by a shift in response-outcome encoding by infralimbic cortex. eneuro 4(6):ENEURO.0337–17.2017. 10.1523/ENEURO.0337-17.2017 [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Barker JM, Lench DH, Chandler LJ (2015b) Reversal of alcohol dependence-induced deficits in cue-guided behavior via mGluR2/3 signaling in mice. Psychopharmacology 233:235–242. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Barker JM, Taylor JR (2014) Habitual alcohol seeking: modeling the transition from casual drinking to addiction. Neurosci Biobehav Rev 47:281–94. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Barker JM, Torregrossa MM, Arnold AP, Taylor JR (2010) Dissociation of genetic and hormonal influences on sex differences in alcoholism-related behaviors. J Neurosci 30:9140–9144. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Barker JM, Torregrossa MM, Taylor JR (2012) Low prefrontal PSA-NCAM confers risk for alcoholism-related behavior. Nat Neurosci 15:1356–1358. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Barker JM, Zhang H, Villafane JJ, Wang TL, Torregrossa MM, Taylor JR (2014) Epigenetic and pharmacological regulation of 5HT3 receptors controls compulsive ethanol seeking in mice. Eur J Neurosci 39:999–1008. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Becker HC (2012) Animal models of excessive alcohol consumption in rodents, in Behavioral Neurobiology of Alcohol Addiction. Current Topics in Behavioral Neurosciences (Sommer W, Spanagel Reds), Vol. 13. Springer, Berlin, Heidelberg. [Google Scholar]
  12. Corbit LH, Janak PH (2016) Habitual alcohol seeking: neural bases and possible relations to alcohol use disorders. Alcohol Clin Exp Res 40:1380–1389. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Corbit LH, Nie H, Janak PH (2012) Habitual alcohol seeking: time course and the contribution of subregions of the dorsal striatum. Biol Psychiatry 72:389–395. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Corbit LH, Ostlund SB, Balleine BW (2002) Sensitivity to instrumental contingency degradation is mediated by the entorhinal cortex and its efferents via the dorsal hippocampus. J Neurosci 22:10976–10984. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Depoy L, Daut R, Wright T, Camp M, Crowley N, Noronha B, Lovinger D, Holmes A (2014) Chronic alcohol alters rewarded behaviors and striatal plasticity. Addiction Biol 20:345–348. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Dickinson A (1985) Actions and Habits: The Development of Behavioural Autonomy. Philos Trans R Soc B Biol Sci 308:67–78. [Google Scholar]
  17. Dickinson A, Wood N, Smith JW (2002) Alcohol seeking by rats: Action or habit? Q J Exp Psychol Sect B 55:331–348. [DOI] [PubMed] [Google Scholar]
  18. Fanelli RR, Klein JT, Reese RM, Robinson DL (2013) Dorsomedial and dorsolateral striatum exhibit distinct phasic neuronal activity during alcohol self-administration in rats. Eur J Neurosci 38:2637–2648. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Gianessi CA, Groman SM, Taylor JR (2019) Bi-directional modulation of food habit expression by the endocannabinoid system. Eur J Neurosci 49:1610–1622. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Gianessi CA, Groman SM, Thompson SL, Jiang M, Stelt M, Taylor JR (2020) Endocannabinoid contributions to alcohol habits and motivation: Relevance to treatment. Addict Biol 25(3):e12768. 10.1111/adb.12768. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Gourley SL, Zimmermann KS, Allen AG, Taylor JR (2016) The Medial Orbitofrontal Cortex Regulates Sensitivity to Outcome Value. J Neurosci 36:4600–4613. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Grodin EN, Bujarski S, Venegas A, Baskerville W-A, Nieto SJ, Jentsch JD, Ray LA (2019) Reward, Relief and Habit Drinking: Initial Validation of a Brief Assessment Tool. Alcohol Alcohol 54:574–583. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Harris AC, Pentel PR, LeSage MG (2007) Prevalence, magnitude, and correlates of an extinction burst in drug-seeking behavior in rats trained to self-administer nicotine during unlimited access (23 h/day) sessions. Psychopharmacology 194:395–402. [DOI] [PubMed] [Google Scholar]
  24. Hay RA, Jennings JH, Zitzman DL, Hodge CW, Robinson DL (2013) Specific and Nonspecific Effects of Naltrexone on Goal-Directed and Habitual Models of Alcohol Seeking and Drinking. Alcohol Clin Exp Res 37:1100–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Koob GF, Volkow ND (2016) Neurobiology of addiction: a neurocircuitry analysis. The Lancet Psychiatry 3:760–773. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Lerman DC, Iwata BA (1995) Prevalence of the Extinction Burst and its Attenuation During Treatment. J. Appl. Behav. Anal. 28:93–94. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Lex B, Hauber W (2010) Disconnection of the entorhinal cortex and dorsomedial striatum impairs the sensitivity to instrumental contingency degradation. Neuropsychopharmacology 35:1788–96. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Lopez MF, Becker HC, Chandler LJ (2014) Repeated episodes of chronic intermittent ethanol promote insensitivity to devaluation of the reinforcing effect of ethanol. Alcohol. 10.1016/j.alcohol.2014.09.002. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Morisot N, Phamluong K, Ehinger Y, Berger AL, Moffat JJ, Ron D (2019) mTORC1 in the orbitofrontal cortex promotes habitual alcohol seeking. Elife 8: 10.7554/eLife.51333. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. O’Tousa D, Grahame N (2014) Habit formation: implications for alcoholism research. Alcohol 48:327–335. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Quinn JJ, Hitchcott PK, Umeda EA, Arnold AP, Taylor JR (2007) Sex chromosome complement regulates habit formation. Nat Neurosci 10:1398–1400. [DOI] [PubMed] [Google Scholar]
  32. Renteria R, Baltz ET, Gremel CM (2018a) Chronic alcohol exposure disrupts top-down control over basal ganglia action selection to produce habits. Nat Commun 9:211. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Renteria R, Cazares C, Gremel CM (2020) Habitual Ethanol Seeking and Licking Microstructure of Enhanced Ethanol Self-Administration in Ethanol-Dependent Mice. Alcohol Clin Exp Res 44:880–891. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Schreiner DC, Renteria R, Gremel CM (2019) Fractionating the all-or-nothing definition of goal-directed and habitual decision-making. J Neurosci Res 98:998–1006. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Serlin H, Torregrossa MM (2014) Adolescent rats are resistant to forming ethanol seeking habits. Develop Cogn Neurosci 16:183–190. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Sjoerds Z, de Wit S, van den Brink W, Robbins TW, Beekman ATF, Penninx BWJH, Veltman DJ(2013) Behavioral and neuroimaging evidence for overreliance on habit learning in alcohol-dependent patients. Transl Psychiatry 3:e337. [DOI] [PMC free article] [PubMed] [Google Scholar]

RESOURCES