Dorsolateral striatum dopamine-dependent cocaine seeking is resistant to Pavlovian cue extinction in male and female rats

Abstract

Cue exposure therapy (CET) reduces craving induced by drug-associated cues in individuals with substance use disorders. A preclinical model of CET, cue extinction, similarly reduces cue-induced cocaine seeking in rodent self-administration models; however, those models may not capture the habitual or compulsive aspects of drug use. Thus, the effectiveness of cue extinction was tested in male and female rats trained to self-administer cocaine using second-order (SO) or fixed-ratio (FR) schedules of reinforcement to facilitate dorsolateral striatum (DLS) dopamine-dependent or - independent response strategies, respectively. Cue extinction significantly reduced drug seeking in FR-trained rats, replicating prior results, but was ineffective in SO-trained rats. SO-trained rats also showed increased indices of glutamate signaling in the DLS relative to FR-trained rats, despite comparable levels of cocaine intake. Furthermore, in SO-trained rats, antagonism of AMPA receptors in the DLS rescued the efficacy of cue extinction to reduce drug seeking. Finally, the effectiveness of cue extinction was also revealed in SO-trained rats when they were taught to perform a new, non-habitual response for cocaine cue presentation. Overall, our findings indicate that habit-like drug seeking leads to plasticity in the DLS and behavior that is resistant to cue extinction, but that the effects of cue extinction are restored when DLS glutamatergic signaling is blocked. These results have implications for the effectiveness of clinical cue exposure therapy.

1. Introduction

Drugs of abuse influence learning and memory systems, and drug-related memories are thought to contribute to continued drug use and relapse (Milton and Everitt, 2012; Torregrossa et al., 2011). Drug use begins as a goal-directed action, driven by knowledge that a behavior will procure the drug (Belin-Rauscent et al., 2012). Under certain conditions, drug seeking can become a stimulus-response habit, when the behavior is driven by environmental stimuli (Leong et al., 2016; Murray et al., 2012; Zapata et al., 2010). The involvement of stimulus-response behaviors in SUDs has been widely debated, but complex human drug seeking likely involves both action-outcome and stimulus-response behaviors (Everitt and Robbins, 2005; Hogarth et al., 2019; Watson and de Wit, 2018; Woodhead and Robbins, 2017). Drug seeking initiated by action-outcome or stimulus-response associations can be distinguished behaviorally or through pharmacological manipulation of distinct anatomical structures (Corbit et al., 2014; DePoy et al., 2016; Murray et al., 2015; Murray et al., 2012; O’Hare et al., 2018). Action-outcome, also known as goal-directed behaviors rely on dopamine in the dorsomedial striatum (DMS), whereas habitual behaviors initiated by stimulus-response associations rely on dopamine in the dorsolateral striatum (DLS) (Corbit et al., 2014; Faure et al., 2005; Hodebourg et al., 2019; Murray et al., 2015; Murray et al., 2012; Yin and Knowlton, 2004; Zapata et al., 2010). These behaviors also rely on distinct amygdala nuclei, where inhibition of the basolateral amygdala (BLA) reduces DLS-independent cocaine seeking, while inhibition of the central amygdala (CeA) reduces DLS-dependent cocaine seeking (Murray et al., 2015).

Pavlovian associations between environmental cues and drug reinforcement promote drug-seeking behavior and involve synaptic plasticity in the BLA (Bender and Torregrossa, 2020; Feltenstein and See, 2007; Rich et al., 2019). Drug-associated cues can themselves reinforce drug-seeking behavior or non-drug-related behaviors through conditioned reinforcement (Di Ciano and Everitt, 2004). Likewise, exposure to drug-associated cues can induce craving and relapse in patients with substance use disorders (SUDs) (Carter and Tiffany, 1999; Grant et al., 1996; Wang et al., 1999). Cue exposure therapy (CET), which involves repeated unreinforced exposure to drug-associated cues, is a proposed behavioral treatment for SUDs (Conklin and Tiffany, 2002). Several preclinical studies support the efficacy of cue extinction, an animal model of CET (Kearns et al., 2012; Madsen et al., 2017; Perry et al., 2016; Rich et al., 2019, 2016; Rich and Torregrossa, 2018), but clinical applications have yielded mixed results (Conklin and Tiffany, 2002; Mellentin et al., 2017; Taylor et al., 2009). One reason for the lack of translation is likely due to the context dependency of extinction learning (Kantak and Nic Dhonnchadha, 2011; Rich et al., 2019; Torregrossa et al., 2010). Yet, another factor could be that human SUDs often involve inflexible, habitual behaviors that may be initiated even if the association between the cue and the drug have been extinguished (Belin-Rauscent et al., 2016; Everitt, 2014; Everitt et al., 2018; Sjoerds et al., 2013; Volkow et al., 2006).

Our lab has shown that the effects of cue extinction are mediated by depotentiation of BLA synapses (Rich et al., 2019). This evidence, combined with the knowledge that DLS-dependent, habitual behavior no longer relies on the BLA (Murray et al., 2015), led us to hypothesize that DLS-dependent drug seeking would be resistant to cue extinction that otherwise effectively reduces drug seeking. Here, we use fixed-ratio (FR) and second-order (SO) schedules of reinforcement to facilitate DLS dopamine-independent, or -dependent, putatively habitual cocaine self-administration, respectively, in male and female rats. We evaluated the effects of cue extinction on cue-induced drug seeking and compared the expression of proteins involved in glutamate-dependent plasticity in the dorsal striatum. Finally, we determined how the restoration of action-outcome control via DLS glutamate antagonism impacted the efficacy of cue extinction. Our results replicate findings that cue extinction reduces cue-induced drug seeking in FR-trained rats, and provide the first evidence that SO-trained rats, using DLS dopamine-dependent response strategies to self-administer cocaine, are resistant to cue extinction unless action-outcome control is restored. These findings have important implications for the use of CET in SUD treatment.

2. Methods

2.1. Animals

Adult Sprague-Dawley rats (Envigo) weighed ~275 (male) or ~200 g (female) upon arrival (n=168; male n=120; female n=48). Animals were pair-housed in auto-ventilated racks with automated watering in a temperature- and humidity-controlled room maintained on a 12-hour light-dark cycle and had ad libitum access to food and water. Rats were given ≥5 days to acclimate to the facility before surgical procedures, after which they were housed individually. Rats were food-restricted 24 hours before the start of training, and were maintained at ~90% of their free-feeding body weight. Behavioral experiments were run in the light cycle and began within ~3 hours of the same time of day. Procedures were conducted in accordance with the National Institute of Health’s Guide for the Care and Use of Laboratory Animals and were approved by the University of Pittsburgh’s Institutional Animal Care and Use Committee.

2.2. Drugs

Cocaine hydrochloride (graciously provided by NIDA) was dissolved at 2 mg/ml in 0.9% sterile saline (Thermo Fisher) and filter-sterilized. Cis-flupenthixol hydrochloride (Cayman Chemical Company) was dissolved at 20 μg/μl in ddH₂O. NBQX (2,3-dihydroxy-6-nitro-7-sulfamoyl-benzo[f]quinoxaline, Cayman Chemical Company) was dissolved at 0.3 and 1 μg/μl in 0.9% sterile saline.

2.3. Behavioral Apparatus

Experiments were conducted in 24 standard operant conditioning chambers (MedAssociates) using MedPC software (MedAssociates). Each animal underwent all training and testing in the same chamber. Each chamber was equipped with bar floors, a house light, two cue lights above two levers, a tone generator, a head-entry magazine, and a syringe pump connected to a swiveled leash and housed in a sound-attenuating box with a fan for background noise. All boxes had 2 plexiglass walls, one wall containing the levers, magazine, and cue lights, and another wall containing nose-poke apertures. Half of the boxes were equipped with 2 nose-poke apertures, while the other half were equipped with 5 nose-poke apertures with a removable opaque plexiglass cover.

2.4. Surgery

2.4.1. Anesthesia

Rats were fully anesthetized via intramuscular injections of ketamine (87.5–100 mg/kg, Henry Schein) and xylazine (5 mg/kg, Butler Schein) and were then given subcutaneous injections of the analgesic Rimadyl (5 mg/kg; Henry Schein) and 5 ml of Lactated Ringer’s solution. Surgical sites were shaved and betadine (povidone iodine, 5%; Henry Schein) and 70% ethanol were applied to all incision sites as previously described (Rich et al., 2019).

2.4.2. Intravenous catheterization

All rats were implanted with a chronic indwelling intravenous catheter into the right jugular vein and fed subcutaneously to exit the midscapular region where a bent cannula (PlasticsOne) exited through a round incision, as previously described (Torregrossa and Kalivas, 2008). Catheters were capped to prevent blockage.

2.4.3. Intracranial cannulation

Immediately following jugular vein catheterization, rats used in experiments involving intra-DLS infusions were placed in a stereotaxic frame. A small injection of lidocaine (0.3–0.4 ml; Henry Schein) was used as a local anesthetic to the scalp. Two 22 gauge guide cannula (cut 6 mm below an 8 mm pedestal, PlasticsOne) were implanted bilaterally, aimed 1 mm dorsal to the anterior DLS (in mm from Bregma, anterior and posterior (AP): +0.8; medial and lateral (ML): +/− 3.0; dorsal and ventral (DV): −4.0)) and secured to the skull with 3 screws and OrthoJet dental cement (Lang Dental). Once dry, dummy cannula (C313DC, PlasticsOne) the length of the guide cannula were inserted to prevent obstruction.

2.4.4. Post-operative care

On the two days following surgery, rats were administered Rimadyl (5 mg/kg) subcutaneously. Catheter patency was maintained by daily infusion of 0.2–0.4 ml of a 0.99% sterile saline solution containing Gentamicin (3 mg/ml; Henry Schein), heparin (30 USP/ml; Henry Schein), and, for the week following surgery only, streptokinase (9.33 USP/ml; MP Biomedicals).

2.5. Behavioral Procedures

2.5.1. Cocaine self-administration

Rats were trained to self-administer cocaine (1mg/kg/infusion) in 1-hour daily sessions. Each session began with the illumination of the house light, insertion of the active and inactive lever (counterbalanced between animals), and start of the fan. All cocaine infusions were paired with a 20-second audiovisual cue of a tone and the illumination of the cue light above the active lever, and initiated a 20-second time-out period when the house light was extinguished and levers retracted. Inactive lever presses were recorded, but had no programmed consequence. Each session was terminated after 1 hour or the delivery of 30 infusions.

Rats were trained on schedules designed to facilitate different behavioral strategies during self-administration using a protocol modified from experiments previously described (Murray et al., 2015; Murray et al., 2012). All rats were initially trained to self-administer cocaine on a fixed-ratio 1 (FR1) schedule for 7 days, and then on an FR3 schedule for 3 days. Figure 2A illustrates how rats were then divided into groups that were either trained to maintain action-outcome drug seeking (FR-trained) or to develop DLS dopamine-dependent behavior through second-order (SO) schedule training (SO-trained). Following the three FR3 training sessions, the FR-trained rats were maintained on an FR3 schedule for an additional 5 days and were then switched to an FR5 schedule for 5 more days to increase responses required for each infusion. The SO-trained rats were switched to an SO schedule for a total of 10 days, with 5 days on an FR5(FR2S) followed by 5 days on an FR7(FR2S) schedule. Rats were matched for responding before splitting into FR-trained and SO-trained groups. Based on our own pilot experiments and evidence that just two weeks of training on an FR10(FR4S) SO schedule facilitates DLS dopamine-dependent cocaine seeking, we chose this level and time-course of training to maximize acquisition and catheter patency while minimizing experiment length and differences in responding between FR- and SO-trained rats (Murray et al., 2015).

Figure 2: DLS dopamine-dependent cocaine-seeking behavior is resistant to cue extinction.

Experimental timeline (A). Rats (n=37; 19 male, 18 female) were trained to self-administer cocaine on either FR or SO schedules of reinforcement (B-D). They underwent cue extinction (120 CS) or a control (0 CS) procedure followed by a cue-induced drug-seeking test (E-F). Females self-administered more cocaine than males, and FR-trained rats self-administered more cocaine than SO-trained rats (B), but there were no differences between rats to be assigned to different cue extinction groups (C). As rats learned to self-administer cocaine, their active lever presses increased, while inactive lever presses remained low, and SO-trained rats pressed the active lever more than FR-trained rats (D). In a cue-induced drug-seeking test, there was a main effect of cue extinction and training schedule, but no interaction, on ratio of responding (E), and a main effect of training schedule, but no main effect of cue extinction or interaction, on active lever presses (F). Planned comparisons revealed that previous cue extinction (120 CS) resulted in reduced ratio of responding (E) and active lever presses (F) in FR-trained, but not SO-trained, rats relative to a 0-cue control group. Graphs show group means ± SEM and individual data points. Open symbols indicate females, and closed symbols indicate males. Vertical gray bars indicate changes in reinforcement schedule (B-D). *p<0.05. **p<0.01. ****p<0.0001.

Under an FR1 schedule, each active lever press resulted in a cocaine infusion paired with the audiovisual cue and the initiation of the time-out period. Under an FR3 and FR5 schedule, every third or fifth lever press, respectively, resulted in an infusion, the cue, and the time-out period. Under an FR5(FR2S) SO schedule, every two lever presses (indicated by FR2S) resulted in a short 1-second presentation of the audiovisual cue, and the 5th completion of this schedule (indicated by FR5), or a total of 10 lever presses, resulted in cocaine delivery, the audiovisual cue, and time-out period. Lever presses during the 1-second cue presentations were recorded, but did not contribute toward the completion of the schedule. Under an FR7(FR2S) schedule, every two lever presses resulted in a short 1-second cue presentation, and the cue, cocaine, and time-out occurred after the seventh completion of that schedule (14 total lever presses).

2.5.2. Drug-seeking tests

For the experiment involving intra-DLS infusion of cis-flupenthixol, separate groups of rats underwent a 15-minute drug-seeking test immediately before self-administration on day 9 (during FR training) or day 13 (during SO training) (Figure 1A). Rats were given bilateral intra-DLS infusions of vehicle (ddH₂O) or cis-flupenthixol (10 μg) 5 minutes prior to testing, a timepoint for which this dose of cis-flupenthixol has previously been shown to reduce habitual drug seeking (Murray et al., 2015; Murray et al., 2012). Rats then began the 15-minute drug-seeking test, during which cues and timeouts were presented contingently as previously described on the current training schedule (FR3 for day 9; FR5(FR2S) for day 13), but cocaine was withheld. Lever presses were recorded, and a 1-hour self-administration session immediately followed to prevent extinction learning.

Figure 1: Second-order schedule training facilitates the development of DLS dopamine-dependent, putatively habitual cocaine-seeking behavior.

Experimental timeline for self-administration (SA) and drug-seeking tests (A). Rats (n=38, male) were trained to self-administer cocaine (1 mg/kg/infusion) for 15 days (B-C; G-H), and the reliance of drug-seeking on DLS dopamine was analyzed by intra-DLS infusion of the nonspecific dopamine antagonist cis-flupenthixol (10 μg) in a 15-minute drug-seeking test, with no cocaine available, on either SA day 9, after FR training (B-F), or SA day 13, after SO training (G-K). There was no effect of treatment group on infusions (B, G) or active lever presses (C, H) during self-administration. On SA day 9, after FR training, DLS dopamine antagonism did not affect the ratio of active lever presses during test to the previous day of SA (D), active lever presses (E), or inactive lever presses (F) during a 15-minute drug-seeking test. On SA day 13, after SO training, DLS dopamine antagonism reduced the ratio of active lever presses during test to the previous day of SA (G) and active lever presses (H) compared to vehicle controls, but did not affect inactive lever presses (I). Graphs show group means ± SEM and individual data points. Vertical gray bars indicate changes in reinforcement schedule (B-C). *p<0.05. **p<0.01. ****p<0.0001.

2.5.3. Pavlovian cue extinction

On the day immediately following the final day of self-administration, rats underwent Pavlovian cue extinction or a control procedure, during a 1-hour session when 0 or 120 20-second audiovisual conditioned stimuli (CSs), separated by 10 seconds, were presented non-contingently in the same context as self-administration, with levers retracted.

2.5.4. Cue-induced drug-seeking test

One day following cue extinction, rats underwent a 1-hour cue-induced drug-seeking test, during which cues were presented contingently on the rat’s previous training schedule, but no cocaine was delivered. In one experiment (Figure 5A), rats received intra-DLS infusions of either vehicle (0.9% bacteriostatic saline) or NBQX (0.3 or 1 μg/μl) 5 minutes prior to the start of the session. Although this cue-induced drug-seeking test is typically referred to as cue-induced reinstatement, these rats did not undergo typical instrumental extinction to avoid its potential degradation of stimulus-response associations important for maintenance of DLS-dependent behaviors. Previous findings suggest that cue extinction’s behavioral and biological effects are not dependent on instrumental extinction (Rich et al., 2019), but in order to uphold precise terminology, we will refer to this test as a cue-induced drug-seeking test because of the omission of instrumental extinction. Rats used for the drug-seeking tests after cis-flupenthixol or vehicle were also used to examine the effects of a lower dose of NBQX (0.3 μg/μl). Because there were no differences between vehicle rats between experiments, data were collapsed across cohorts for analysis.

Figure 5: A cocaine-associated discrete cue is not required for the maintenance of DLS dopamine-dependent cocaine-seeking behavior despite the cue having reinforcing properties.

Experimental timeline (A). A subset of rats from the previous experiment (Figure 4) returned to SA for 2 days, with cues removed on the second day. They then underwent conditioned reinforcement training, when they learned to nose-poke for cues on an FR7(FR2S) schedule, and then underwent cue extinction and conditioned reinforcement testing (A). Removal of the audiovisual cue during cocaine SA had no effect on active (B-C) or inactive lever presses (D). During conditioned reinforcement training, rats made more active than inactive nose pokes (E). After cue extinction, during conditioned reinforcement testing, extinction rats made fewer active nose pokes than control rats (F) but their inactive nose pokes were not affected (G). When comparing active nose-pokes during training to testing, there was a significant day × extinction interaction (H). Graphs show group means ± SEM and individual data points. Open symbols indicate females, and closed symbols indicate males. *p<0.05.

2.5.5. Re-training and self-administration without cue

On the day following the cue-induced drug-seeking test, a subset of rats used in the NBQX or vehicle experiment returned to self-administration for one day of re-training on an FR7(FR2S) schedule. In order to examine if the discrete audiovisual cue was important for the maintenance of established, DLS dopamine-dependent cocaine seeking, these rats then underwent a self-administration session where the discrete audiovisual cue was removed, but the same number of lever presses resulted in a cocaine infusion. In this session, 14 lever presses (FR14, no cues) resulted in a cocaine infusion and a 20-second time-out period, during which the house light was extinguished and the levers were retracted, but no discrete audiovisual cue was presented.

2.5.6. Conditioned reinforcement

The same rats that were re-trained and used for self-administration without cues were then used to examine the reinforcing properties of the discrete audiovisual cue. Rats were returned to the self-administration context and underwent conditioned reinforcement training. During these sessions, the house light was illuminated, but levers were retracted. Head entries into the active nose-poke aperture (counterbalanced) resulted in presentation of discrete cues on an FR7(FR2S) schedule, with no cocaine infusions. Cues were presented for 1 second after every two presses and for 20 seconds upon the completion of the second order schedule on the same cue light above the previously active lever. In operant conditioning chambers containing 5 nose-poke apertures, the left-most and right-most apertures were used as the active and inactive nose pokes. Active nose pokes were recorded, and inactive nose pokes were recorded after the first active nose poke because prior to the first active nose poke the rats have no knowledge of the contingencies in place on either aperture. This avoids the possibility that a rat could spend the majority of the session poking the inactive aperture without discovering that the active aperture would produce the cue, which would then skew the results. Rats were matched for responding and underwent a second cue extinction procedure on the following day as described above (counterbalancing for previous cue extinction group), and on the following day underwent conditioned reinforcement testing, which did not differ from the training session.

2.6. Intra-DLS infusions

For experiments involving intra-DLS infusions, drug or vehicle was infused through a 28-guage internal cannula extending 1 mm below the guide cannula into the DLS using a 10 μl Hamilton syringe connected to a syringe pump (Harvard Apparatus). Cis-flupenthixol was infused at 0.33 μl/min for 90 seconds. NBQX was infused at 0.3 μl/min for 60 seconds. Internal cannula were left in place for 1 minute following infusion.

2.7. Western blot analysis

2.7.1. Sample preparation

Rats used in the 0 vs. 120 cue extinction experiment were killed by decapitation immediately after the cue-induced drug seeking test. Brains were flash-frozen in isopentane on dry ice and kept at −80 °C. The anterior DLS (AP +2 mm to +0.8 mm) and posterior DMS (AP +1 mm to −0.4 mm) were dissected over dry ice from coronal sections made ~1 mm thick on a stainless steel brain matrix (Braintree Scientific). We chose to focus our analysis on the anterior portion of the DLS and the posterior portion of the DMS based on previous publications suggesting this distinction is important (Furlong et al., 2018; Yin et al., 2005). Tissue was fractionated into membrane- and non-membrane-bound components as previously described (Bañuelos et al., 2014). Protein concentrations for each sample were determined using a Thermo Scientific Pierce Micro BCA Protein Assay (Thermo Fischer Scientific) as previously described (Kirschmann et al., 2017). For each sample, 20 μg of protein was diluted in 30 μl containing 8 μl of sample buffer (a 9:1 mixture of 4x Laemmli protein sample buffer [Bio-Rad] to 2-Mercaptoethanol [Millipore Sigma]).

2.7.2. Immunoblotting

Membrane fractions were resolved on ice by SDS-PAGE on 4–12% Tris-glycine gels (Invitrogen) and electrophoretically transferred to a PVDF membrane (Immuno-Blot PVDF Membrane, Bio-Rad). Membranes were blocked for 1 hour at room temperature by incubation in 5% bovine serum albumin (Thermo Scientific Pierce) and then incubated overnight at 4 °C with specific primary antibodies (Millipore) against the following proteins: NMDA receptor subunits GluN2A (1:1000, 04–901) and GluN2B (1:1000, 05–920), AMPA receptor subunits GluA1 (1:1000, MAB2263) and GluA2/3 (1:500, 07–598), vesicular glutamate transporter VGluT1 (1:1000, ABN1647), and β-actin (1:1000, Cell Signaling Technology, 3700S). Antigen binding was visualized by incubating membranes for 1 hour at room temperature in secondary fluorescent antibodies (IRDye 800 CW anti-rabbit, 1:5000, LI-COR Odyssey, 926–32211; IRDye 680 CW anti-mouse, 1:5000, LI-COR Odyssey, 926–68070). All antibodies were diluted in blocking solution (1:1 LI-COR Odyssey blocking buffer to 1x PBST). Protein expression was quantified using LI-COR Odyssey imaging and ImageStudio software. Each sample was normalized to its own β-actin expression, and expression for the SO-trained group was normalized to average levels of the FR-trained group within each gel.

2.8. Histology

Rats with cannula were killed by CO₂ followed by decapitation, and brains were removed and immersed in 10% buffered formalin phosphate (Fisher Chemical) for ~24 hours. Brains were then moved to a cryoprotectant solution (30% sucrose) for ≥2 days before they were frozen and sectioned on a cryostat (Leica CM1950) at 50 μm. Every third section was immediately mounted on a slide, and cannula tracts were visualized and their location noted. Histological misses were characterized by placement more than 2.2 mm anterior to Bregma, less than 0.6 mm anterior to Bregma, dorsal to the striatum, or medial or lateral to the DLS.

2.9. Exclusion Criteria

Rats were excluded from analysis due to death or illness after surgery (n=7), histological misses in drug-treated rats (n=8), failure to acquire cocaine self-administration (n=10) (>4 infusions on the final day of self-administration, note hat the low infusion number is due to the high response requirement), or loss of catheter patency (n=33) (determined by a 0.1 ml intravenous infusion of 10 mg/ml sodium brevital at the end of self-administration training). Note, that the high rate of loss due to catheter patency was due to the development of a new catheter system in the lab at the time. Attrition rates are not commonly that high, and were improved by the addition of streptokinase (9.33 U/ml) to the post-operative flushing solution.

2.10. Quantification and Statistical Analysis

Behavioral data were collected using MedPC software. For 2 rats, self-administration data for day 17 were unavailable due to a power outage, so data for the missing day were entered by averaging behavioral responses on day 16 and 18. When possible, experimenters were blinded to rats’ treatment conditions. All statistical analyses were performed using GraphPadPrism and SPSS Statistics software. When split into groups, animals were matched for responding. For 15-minute drug-seeking tests, the ratio of responding during test to previous responding was calculated by dividing the active lever presses during test by the average number of active lever presses per 15 minutes in the previous self-administration session: active lever presses during test / (active lever presses on previous day / 4). For 1-hour cue-induced drug-seeking tests, the ratio of responding was calculated by dividing the number of active lever presses during test by the number of active lever presses in the previous self-administration session: active lever presses during test / active lever presses on previous day. We use the ratio of responding, where a ratio of 1 indicates no change in responding during the cue-induced drug-seeking test compared to the previous day of self-administration to accurately visualize, with y axes of the same scale, the effects of cue extinction on animals responding on different schedules of reinforcement that produce different rates of responding. We additionally show active lever presses, but note that plotting raw active lever presses requires the use of y axes of different scales and occludes valuable information about how rats using different response strategies differentially respond under extinction conditions compared to during self-administration. Despite this, we show that ratio of responding and active lever presses reveal similar results, but ultimately draw our major conclusions from data showing the ratio of responding.

For all statistical analyses, significance was set at p<0.05. All data were determined to be normally distributed using the Shapiro-Wilk test, and Bartlett’s test was used to determine that there were no significant differences in the estimated variance between groups. Infusions were analyzed by using either a two-way or three-way rmANOVA, using time as one factor and either lever, sex, training schedule, drug treatment, treatment day, or future extinction groups as the other factor(s) as indicated. Protein expression, ratio of responding, active lever presses, inactive lever presses, active nose pokes, or inactive nose pokes were analyzed using either a student’s t-test (with paired measures when indicated) or a two-way ANOVA as indicated. When an interaction was detected by two-way ANOVA analysis, significant effects were further analyzed by Sidak’s post-hoc multiple comparisons analysis. Student’s t-tests were used to make planned comparisons between FR-trained and SO-trained rats.

3. Results

3.1. Cocaine seeking relies on DLS dopamine after training on SO, but not FR, schedules of reinforcement

To evaluate the effects of cue extinction on DLS dopamine-dependent cocaine seeking, we sought to adapt previously established methods utilizing SO schedules to facilitate the formation of DLS dopamine-dependent, putatively habitual behavior (Murray et al., 2015; Murray et al., 2012). Under SO schedules, short presentations of the drug-paired cue sustain responding over longer periods when the drug is not available. It has been proposed that SO schedules may better model the human experience of encountering drug-associated cues both in the presence and absence of drug reinforcement (Belin-Rauscent et al., 2016). Our first goal was to verify that we could replicate prior studies indicating that cocaine self-administration reinforced on an FR schedule would not require DA signaling in the DLS, while responding under an SO schedule would be dependent on DLS DA. Because these previous experiments used male rats, we used only male rats for this initial investigation.

Rats learned to self-administer cocaine (1 mg/kg/infusion) for 20 days, and their full self-administration data is included in Figure 4, as these rats were also used for those experiments. Here, we show their self-administration data for the first 15 days, which is relevant for these experiments. Vertical gray lines on graphs indicate changes in self-administration reinforcement schedule outlined in Figure 1A. On days 9 or 13 respectively after either 8 (FR-group) or 12 (SO-group) days of self-administration, rats were given intra-DLS infusions of vehicle or cis-flupenthixol (10 μg) before a 15-minute drug-seeking test, when cocaine was unavailable (Figure 1A). The drug seeking test was immediately followed by the appropriate self-administration session when cocaine was available. In the first group of rats, which underwent only FR training prior to DLS infusion of vehicle or cis-flupenthixol before the 9^th day of self-administration, analysis of daily self-administration behavior revealed no main effect of treatment group on infusions (F_(1,16)=0.001817, p=0.9665) (Figure 1B) (two-way rmANOVA). Active lever presses per session increased as the schedule of reinforcement required more presses per infusion, and inactive lever presses remained low. There was a main effect of training day (F_{(14, 224)}=25.97, ****p<0.0001), lever (F(1,16)=33.81, ****p<0.0001), and a training day × lever interaction (F_{(14, 224)}=24.75, ****p<0.0001) (Figure 1C) (three-way rmANOVA). After only FR training, during the 15-minute drug-seeking test, cis-flupenthixol had no effect on ratio of responding (p=0.9186, η²=0.0006736) (Figure 1D), active lever presses (p=0.5210, η²=0.02621) (Figure 1E), or inactive lever presses (p=0.2630, η²=0.07761) (Figure 1F) (independent samples t-tests), indicating that lever pressing did not rely on DLS dopamine and suggesting the maintenance of action-outcome behavior. Note that a subset of animals in each group expressed a low number of active lever presses during this 15-minute test, which likely reflects individual differences in acquisition of the FR3 contingency, but these individual differences do not impact the results.

Figure 4: AMPA receptor antagonism during cue-induced drug seeking in previously SO-trained rats reveals an effect of cue extinction.

Experimental timeline for self-administration, cue extinction, and cue-induced drug-seeking tests (A). Rats underwent 20 days of training to facilitate DLS dopamine-dependent cocaine self-administration, and there was no main effect of sex (B) or future drug treatment group (C) on infusions throughout training. After cue extinction, rats were given intra-DLS infusions of vehicle, a low dose (0.3 μg/μl), or a high dose (1 μg/μl) of the AMPA antagonist NBQX prior to a 1-hour cue-induced drug-seeking test. During the drug-seeking test, there was a main effect of cue extinction, no main effect of drug treatment, and an extinction × drug treatment interaction on ratio of responding (D). Post-hoc analyses revealed no differences between the 0-cue control and extinction rats in the vehicle groups or the low-dose NBQX groups, but rats in the high-dose NBQX group that underwent cue extinction had a reduced ratio of responding compared to 0-cue controls also given a high dose (D). For active lever presses, there was a main effect of cue extinction, but no main effect of drug treatment or interaction (E). Placement of DLS cannula was histologically evaluated for all rats used in cis-flupenthixol and NBQX experiments (total n=73), and each on-target placement is indicated by a black dot, while off-target misses are indicated by black Xs (F). Graphs show group means ± SEM and individual data points. Open symbols indicate females, and closed symbols indicate males. Vertical gray bars indicate changes in reinforcement schedule (B-C). **p<0.01.

A separate group of rats was infused with cis-flupenthixol or vehicle on the 13^th day of training, when responding had previously been reinforced on an FR5(FR2S) SO schedule. Analysis of daily self-administration behavior revealed no main effect of treatment group on infusions (F_(1,18)=0.05192, p=0.9665) (Figure 1G) (two-way rmANOVA). Active lever presses per session increased as the schedule of reinforcement required more presses per infusion, and inactive lever presses remained low. There was a main effect of training day (F_{(14, 252)}=28.57, ****p<0.0001), lever (F_(1,18)=74.54, ****p<0.0001), and a training day × lever interaction (F_{(14, 252)}=29.09, ****p<0.0001) (Figure 1H) (three-way rmANOVA). After SO training, during the 15-minute drug-seeking test on day 13, DLS dopamine antagonism led to a reduced ratio of responding (**p=0.0071, η²=0.3390) (Figure 1I) and active lever presses (*p=0.0144, η²=0.2894) (Figure 1J) compared to vehicle controls, but had no effect on inactive lever presses (p=0.2593, η²=0.07010) (Figure 1K) (independent samples t-tests), thus suggesting that SO training facilitated the formation of DLS dopamine-dependent behavior, similar to prior results observed in other labs. Because rats were used for later experiments, DLS cannula placements for rats in this experiment are included in Figure 5F.

3.2. Cocaine seeking is resistant to cue extinction in SO-trained rats

Having established that self-administration on an SO schedule facilitates DLS dopamine-dependent cocaine seeking, we sought to compare the effect of cue extinction on cue-induced drug seeking in rats trained on either FR or SO schedules of reinforcement. In this experiment we used both males and females and all rats were trained for 20 days, with the last five days being an FR5 schedule for the FR group or an FR7(FR2S) schedule for the SO group. Rats were then divided into those that underwent cue extinction (120 CS presentations), or a control no cue extinction group (0 CS presentations), and then were given a 1-hour cue-induced drug seeking test (Figure 2A). During self-administration training, main effects of sex and training schedule were found, where females self-administered more cocaine than males (F_(1,33)=5.928, *p=0.0205), and FR-trained rats self-administered more cocaine than SO-trained rats (F_(1,33)=4.308, *p=0.0458) (three-way rmANOVA) (Figure 2B). When evaluated based on future extinction group and training schedule collapsed across sex, FR-trained rats again self-administered more cocaine than SO-trained rats (F_(1,33)=4.650, *p=0.0384), but there were no differences between rats in to-be extinction vs. to-be control groups (F_(1,33)=0.3307, p=0.5691) (three-way rmANOVA) (Figure 2C). Active lever presses per session increased as the schedule of reinforcement required more presses per infusion, and inactive lever presses remained low, demonstrated by a main effect of training day (F_(19,1330)=64.7, ****p<0.0001), lever (F_(1,70)=170.5, ****p<0.0001), and a training day × lever interaction (F_(19,1330)=58.39, ****p<0.0001) (three-way rmANOVA) (Figure 2D). There was a main effect of schedule (F_(1,70)=31.49, ****p<0.0001) and a schedule × lever interaction (F_(1,70)=29.57, ****p<0.0001) (three-way rmANOVA) (Figure 2D). In a cue-induced drug-seeking test after cue extinction, there was a main effect of training schedule (F_(1,33)=6.0196, *p=0.0196, partial η²=0.154), a main effect of cue extinction (F_(1,33)=4.267, *p=0.0468, partial η²=0.115), and no significant interaction (F_(1,33)=1.751, p=0.1948, partial η²=0.050) on the ratio of active lever presses during test to active lever presses on the final day of self-administration (two-way ANOVA) (Figure 2E). For raw active lever presses, there was a main effect of training schedule (F_(8.582,)=8.582, **p=0.0061, partial η²=0.206), but no main effect of cue extinction (F_(1,33)=2.424, p=0.1290, partial η²=0.068) or interaction (F_(1,33)=0.01412, p=0.9061, partial η²=0.000) (two-way ANOVA) (Figure 2F). Planned comparisons between 0 CS controls and 120 CS extinction groups indicated an effect of cue extinction on ratio of responding in FR-trained (*p=0.0426, η²=0.2203), but not SO-trained rats (p=0.5601, η²=0.02166) (Figure 2E) (independent samples t-tests). Planned comparisons showed that cue extinction also reduced raw active lever presses in FR-trained (**p=0.0098, η²=0.2203) but not SO-trained (p=0.4120, η²=0.2203) (Figure FH) rats (independent samples t-tests). These results suggest that cue extinction does not impact responding in SO-trained rats, but reduces responding in FR-trained rats. Although these planned comparisons do not correct for multiple comparisons, these findings have been replicated in multiple experiments. These experiments were not powered to detect sex differences in the effects of cue extinction, but no obvious differences between males and females were observed, as indicated by the individual, differentially shaded, data points in the bars.

3.3. FR and SO cocaine self-administration training results in differential expression of plasticity-related proteins in the dorsal striatum

Next, we wanted to determine if there were molecular indices of altered plasticity in the DMS or DLS of SO-trained rats GluA2/3 (p=0.5045, η²=0.01723), or VGluT1 (p=0.0561, η²=0.1242) between FR-trained and SO-trained rats, but GluA1 (*p=0.0496, η²=0.3249) was significantly increased of moderate effect size (independent sample t-tests, not corrected for multiple comparisons) (Figure 3B). These results relative to FR-trained rats. We analyzed tissue (Figure 3) from the male and female rats used in the previous cue extinction experiment (n=37, Figure 2) by western blot analysis to compare the membrane-bound expression of NMDA receptor subunits GluN2A and GluN2B, AMPA receptor subunits GluA1 and GluA2/3, and vesicular glutamate transporter VGluT1. We chose these proteins because they have previously been reported to change in response to a methamphetamine administration procedure that facilitates habit formation (Furlong et al., 2018). In the DMS, there were no differences in the expression of GluN2A (p=0.5006, η²=0.01526), GluN2B (p=0.6412, η²=0.006870), GluA1 (p=0.4212, η²=0.03260), or VGluT1 (p=0.2879, η²=0.0673), but SO-trained rats had increased expression of GluA2/3 of small effect size (*p=0.0343 η²=0.1961) compared to the FR-trained group (independent sample t-tests, not corrected for multiple comparisons) (Figure 3A). In the DLS, there were no significant differences in the expression of GluN2A (p=0.0653, η²=0.1401), GluN2B (p=0.1343, η²=0.07563),indicate that training on different reinforcement schedules may be sufficient to lead to differential expression of glutamate receptor subunits in the membrane in the dorsal striatum. There was no correlation between total cocaine exposure and protein expression for any protein in either region, and there were no differences between males and females in either region. Each protein was detected at its expected molecular weight, as indicated by representative images of gels from samples from the DLS alongside a ladder (Figure 3C). Proteins with overlapping molecular weights were detected using different secondary antibodies and were imaged on the same membrane at different wavelengths.

Figure 3: FR and SO training to self-administer cocaine results in differential expression of plasticity-related proteins in the dorsal striatum.

Western blot analysis was used to quantify the membrane-bound expression of glutamate plasticity-related proteins in the DMS and DLS in rats trained to self-administer cocaine on FR vs. SO schedules or reinforcement. In the DMS, SO-trained rats had increased expression of GluA2/3 in the membrane compared to FR-trained rats, but there were no differences in GluN2A, GluN2B, GluA1, or VGluT1 (A). In the DLS, SO-trained rats had significantly increased GluA1 in the membrane, but no differences in GluN2A, GluN2B, GluA2/3, or VGluT1 (B). Note that p values were not corrected for multiple comparisons. Representative bands indicate the protein of interest above the β-actin loading control for one animal in each group (A, B). Each protein was detected at its expected molecular weight, indicated on examples of DLS membranes with samples alongside a ladder (C). Dotted boxes indicate pairs of samples used as representative bands in panel B. Graphs show group means ± SEM and individual data points. Open symbols indicate females, and closed symbols indicate males. *p<0.05.

3.4. Restoration of action-outcome control reveals an effect of cue extinction

We next wanted to determine if inhibiting glutamatergic signaling in the DLS would be sufficient to reveal an effect of cue extinction learning in rats trained on the habit-promoting SO schedule. We hypothesized that DLS-dependent stimulus response associations formed during self-administration on an SO schedule resulted in the initiation and persistence of lever pressing during cue-induced drug seeking tests despite cue extinction learning; but, in the absence of DLS signaling, action-outcome responding would be restored and lever pressing would decrease after cue extinction learning. Rats completed the full 20 days of self-administration training, underwent cue extinction (0 or 120 cues), and then their cue-induced cocaine seeking was tested after intra-DLS infusion of vehicle or the AMPA antagonist NBQX (Figure 4A). NBQX in the DLS has been shown previously to restore action-outcome behavior in rats that orally self-administered alcohol (Corbit et al., 2014). Rats used in the cis-flupenthixol experiment (n=23, male) were initially used to evaluate the effects of vehicle or a lower dose of NBQX (0.3 μg/μl) because we were concerned that NBQX might suppress overall responding. When no overall suppression was observed, additional rats (n=35; 21 male, 14 female) were used to evaluate the effects of vehicle or a higher dose of NBQX (1 μg/μl). Because there were no differences between rats in the vehicle groups, results were collapsed for analyses. There were no differences in infusions throughout training between sexes (F_(1,56)=0.5050, p=0.4803) or between future drug treatment groups (F_(2,56)=0.01573, p=0.9844) (two-way rmANOVA) (Figure 4B–C). During cue-induced drug-seeking after cue extinction, when vehicle or NBQX was infused in the DLS, there was a main effect of cue extinction (F_(1,50)=7.255, **p=0.0096, partial η²=0.127), no main effect of drug treatment (F_(2,50)=0.8011, p=0.4545, partial η²=0.031), and an extinction × drug treatment interaction (F_(2,50)=3.456, *p=0.0393, partial η²=0.121) on ratio of responding (two-way ANOVA) (Figure 4D). There was a main effect of cue extinction on active lever presses (F_(1,50)=7.178, **p=0.0100, partial η²=0.126), but no main effect of drug treatment (F_(2,50)=0.7278, p=0.4880, partial η²=0.028) or interaction (F_(2,50)=2.057, p=0.1386, partial η²=0.076) (two-way ANOVA) (Figure 4E). Further post-hoc analyses (Sidak’s multiple comparisons) on ratio of responding revealed no differences between the control and extinction rats in rats treated with vehicle (p=0.9539, d=0.503) or the low-dose of NBQX (p>0.999, d=0.013), but in rats treated with the high-dose of NBQX, there was a significant difference in ratio of responding between the cue extinction and 0-CS control groups (**p=0.0078, d=1.839) (Figure 4D). Note that there was no difference between the 120-vehicle group and the 120-high dose NBQX group (p=0.9948, d=0.668) (Figure 4D). These results suggest that when action-outcome control is restored by a sufficient level of AMPA antagonism in the DLS, the effect of cue extinction is uncovered.

3.5. The conditioned reinforcing properties of the drug-paired cue are sensitive to cue extinction in rats trained on SO schedules to self-administered cocaine

Cue extinction’s lack of effect on cocaine seeking in SO-trained rats brings into question both the necessity of the cue to maintain DLS dopamine-dependent responding and the reinforcing properties of the cue. Therefore, a subset of rats from the previous experiment (n=15; 4 male, 11 female) were used to address these questions. Rats returned to self-administration for two days, but on the second day, audiovisual cues were removed from the session (Figure 5A). There were no differences in active (p=0.7469, η²=0.007678) or inactive (p=0.3268, η²=0.06883) lever presses between self-administration days with and without cues (paired t-tests) (Figure 5B–D), suggesting the drug-paired audiovisual cue is not required to maintain DLS dopamine-dependent self-administration. To assess whether or not the cue maintained conditioned reinforcing properties after SO self-administration, rats were trained for 1 day on an acquisition of a new response task, when rats had the opportunity to nose-poke for the cue alone, in the absence of drug reinforcement, on the same SO schedule (Figure 5A). During conditioned reinforcement training, rats made more active than inactive nose-pokes (*p=0.0121, η²=0.3723) (paired t-test), suggesting the cue retained its conditioned reinforcing properties, which was revealed because the rats were asked to perform a new action for cue presentation (Figure 5E). Rats then underwent cue extinction (120 CS) or control (0 CS) procedures and were re-tested for conditioned reinforcement. Rats that underwent cue extinction made fewer active nose pokes during test than control rats (*p=0.0183, η²=0.3586), but there were no differences between groups in inactive nose pokes (p=0.1428, η²=0.1576) (unpaired t-test) (Figure 5F–G). There was a day × extinction interaction, where rats in the 0-cue control group tended to increase responding on the second day of conditioned reinforcement testing, while rats in the cue extinction group primarily decreased their active nose pokes (F_(1,13)=8.912, *p=0.0105) (2-way ANOVA) (Figure 5H). These results suggest that the reinforcing properties of the drug-paired cue can be extinguished in a new action-outcome task in rats that previously trained on SO schedules to self-administer cocaine.

4. Discussion

Overall, we find that cue extinction reduces cue-induced drug seeking when rats are trained on an FR schedule that promotes action-outcome behavior, as previously published (Rich et al., 2019, 2016; Torregrossa et al., 2013). However, we find that cue extinction is ineffective in rats trained on SO schedules that promote DLS-dependent behavior, despite comparable length of training and cocaine intake. Although a lack of significant interaction between training schedule and cue extinction prevents the conclusion that cue extinction is more effective in FR-trained than SO-trained rats, cue extinction’s efficacy has been well demonstrated in FR-trained rats in several previous publications (Madsen et al., 2017; Rich et al., 2019, 2016; Torregrossa et al., 2013). Thus, we provide new evidence that putatively habitual drug seeking is resistant to cue extinction. We also show that training rats on these different schedules of reinforcement leads to differential expression of membrane-bound glutamate receptors in the dorsal striatum, though this effect may have been more pronounced were we able to examine expression in a cell-type specific manner. Additionally, we show that blocking AMPA receptors in the DLS or requiring rats to produce a novel nose-poke response for cocaine cues leads to an apparent restoration of action-outcome behavior and reveals the suppressive effect of cue extinction.

Pavlovian cocaine-cue associations are dependent on the BLA, and we have shown that the effects of cue extinction are mediated by depotentiation of BLA synapses (Rich et al., 2019). Prior research indicates that the CeA, not the BLA, is necessary for DLS dopamine-dependent, habitual cocaine seeking (Murray et al., 2015). Our findings are consistent with these prior studies, and suggest that cue extinction’s lack of effect on DLS-dependent cocaine seeking is due to reduced reliance on the BLA for cocaine seeking behavior. An alternative interpretation for the lack of effect of cue extinction after SO training includes the “partial reinforcement effect,” which is known to occlude extinction learning (Chan and Harris, 2019). While this is a possibility, this does not completely explain our findings because DLS inhibition revealed an effect of cue extinction, suggesting that extinction is not prevented, but its effects are masked by DLS-dependent neural signaling. Furthermore, the efficacy of cue extinction on the conditioned-reinforcement task refutes the theory that responding on an SO schedule alone leads to resistance to cue extinction, suggesting that the neural circuit underlying behavioral control, and not the particular schedule of reinforcement, is what masks the effects of cue extinction. Previous work from our lab has shown a positive correlation between the number of cocaine-cue pairings and the strength of thalamo-BLA synapses, suggesting that fewer cocaine-cue pairings could result in increased susceptibility to cue extinction (Rich et al., 2019). Interestingly, we found that SO-trained rats were resistant to cue extinction despite taking significantly less cocaine, and therefore receiving fewer cocaine-cue pairings, than FR-trained rats. This finding provides additional evidence that it is the reduced reliance on this circuitry, and not a failure of cue extinction to impact BLA synaptic strength, which results in unaltered responding in SO-trained rats. Although unlikely, it is possible that the results of the present study reveal differences in the underlying Pavlovian processes of cue learning that are unrelated to habit circuitry, but these findings remain relevant to understanding how drug-cue associations guide behavior, especially because SO schedules may better model how drug-associated cues can be encountered in the absence of subsequent drug procurement.

Here, we used concentrations of the AMPAR antagonist NBQX in the DLS that restored action-outcome control in rats that orally self-administered alcohol (Corbit et al., 2014). It has been established that dopamine antagonism in the DLS can reduce overall rates of responding, whereas AMPA receptor antagonism restores action-outcome control without reducing general responding (Corbit et al., 2014; Murray et al., 2015; Murray et al., 2012). The mechanisms by which dopamine or AMPA antagonism in the DLS differentially affect behavioral output are unclear, but are likely due to differential effects of manipulating midbrain dopaminergic inputs versus cortical and subcortical glutamatergic inputs.

Although other schedules of reinforcement, such as random ratio and random interval schedules, can also produce different patterns of behavioral control over drug seeking, we used SO and FR schedules for a few reasons. SO schedules may better model how humans with SUDs encounter drug-associated cues (Belin-Rauscent et al., 2016). Additionally, we sought to remain consistent with the methods used in the experiments that motivated our hypothesis, and previous experiments determined that while SO training facilitates DLS dopamine-dependent cocaine self-administration, an FR schedule maintains DLS dopamine-independence of behavior (Belin-Rauscent et al., 2016; Murray et al., 2015; Murray et al., 2012; Rich et al., 2019). In agreement with previous studies, we found that SO schedule training facilitates DLS dopamine antagonist-sensitive cocaine seeking, suggesting our training methods also facilitated habit formation. Interestingly, we found that just a few days of SO training rendered responding sensitive to DLS dopamine antagonism, suggesting the neural circuitry involved in drug seeking may shift more rapidly than expected. How this shift occurs is still unclear, but likely involves the dramatic increase in lever presses that are sustained by the intermediate cues presented on an SO schedule. Future experiments utilizing in vivo imaging in the striatum and its input regions throughout this shift would provide further insight into the time-course and mechanism. One way these experiments deviated from previous research was by increasing the response requirement for FR-trained rats to FR5 to increase their overall responding (Murray et al., 2012). Although FR-trained rats still maintained fewer active lever presses than SO-trained rats, this adjustment rules out the possibility that cue extinction only affects low levels of responding. One limitation to these experiments is the use of a pharmacological manipulation to define habitual behavior, but this was necessitated because there is no well-established behavioral method, such as devaluation or contingency degradation, for defining habitual IV drug self-administration. Given the results of the present experiments, which suggest a role of Pavlovian associations in guiding behavior relying on goal-directed but not habitual response circuitry, we propose that future research should examine the application of sensitivity to the effects of cue extinction as a viable behavioral assay to distinguish between goal-directed and habitual response strategies, which would be easier to implement in IV drug self-administration models.

In order to further characterize rats in different training groups, we utilized western blot analysis to compare the expression of glutamate receptor subunit and vesicular transporter proteins in the membrane in the DMS and DLS as markers of glutamatergic plasticity. We designed this experiment based on evidence that methamphetamine biases toward habitual response strategies with an associated downregulation of glutamate proteins in the DMS and an upregulation in the DLS (Furlong et al., 2018). Our experiments differ in that all rats self-administered cocaine, so there was no drug-naïve group, and thus training schedule is the distinction between groups. In SO-trained rats, we detected increased membrane-bound GluA2/3 in the DMS and GluA1 in the DLS compared to FR-trained rats. The GluA1 subunit of the AMPA receptor has been implicated in long-term potentiation (Boehm et al., 2006; Shi et al., 2001). Upregulated GluA1 in the DLS in SO-trained rats suggests increased glutamatergic plasticity and further supports the conclusion that SO training facilitates the formation of DLS dopamine-dependent, habitual behavior. Because there is no drug-naïve group to compare to, interpretation of upregulated GluA2/3 in the DMS of SO-trained rats is more difficult. It is possible that this is an effect of significantly reduced membrane insertion of GluA2/3 in FR-trained rats coinciding with a nonsignificant increase in GluA1, which overall suggests an increased reliance on the plasticity-related GluA1 subunit compared to GluA2/3 in FR-trained rats, which would support the hypothesis of increased reliance on the DMS.

In these studies, we expanded on previous findings with the use of both sexes. Although these experiments were not powered to detect sex differences, we show similar effects of cue extinction in male and female rats. In the experiment utilizing both FR and SO schedules, females self-administered significantly more cocaine than males, which agrees with the literature (Jackson et al., 2006; Swalve et al., 2016). This sex difference in cocaine infusions was not significant in the NBQX experiment. There were also no sex differences in the membrane-bound expression of plasticity-related proteins. Although these data suggest similar effects of cue extinction in males and females, future studies could further examine potential sex differences, as sex and sex hormones may impact cue-mediated drug seeking and habit formation (Barker et al., 2010; Fuchs et al., 2005; Schoenberg et al., 2019).

In addition to presenting raw lever press data, we chose to present data as a ratio of active lever presses during the cue-induced drug-seeking test to active lever presses on the final day of self-administration. Importantly, this involves comparing responding for the cocaine-paired cue when cocaine is not available to responding during normal self-administration, which are related, but not equivalent, measures. The ratio, or the relationship between responding under these different conditions, reveals interesting caveats to our findings that are otherwise occluded by the raw lever press data. In FR-trained rats, cue extinction appears to suppress increased responding that occurs in the 0-CS control group. This increased responding in the control group is likely due to intact potentiation of BLA synapses that drive an increase in cue-induced drug seeking in the absence of cocaine (Rich et al., 2019). Interestingly, SO-trained rats in the control treatment group do not increase their responding in the absence of cocaine, supporting the hypothesis that drug-cue associations, encoded by BLA synaptic strength, do not impact DLS dopamine-dependent responding. SO-trained rats continue to respond at the same rate independent of whether or not they underwent cue extinction or are receiving cocaine infusions, providing further evidence that their behavior is not driven by action-outcome associations. Therefore, restoration of action-outcome control with NBQX reveals an effect of cue extinction, but restoring action-outcome control after cue extinction does not significantly reduce responding below that of rats utilizing DLS-dependent response strategies that also underwent cue extinction.

Restoration of action-outcome control produces bidirectional effects on ratio of responding of SO-trained rats depending on their extinction history, indicated by a significant interaction between extinction experience and NBQX treatment, and the restored effect of cue extinction appears to primarily rely on an increase in responding in the 0-CS control group. Therefore, it is important to note that although restoring action-outcome control reveals an effect of cue extinction, and this gives insight into the synaptic and circuit mechanisms underlying DLS dopamine-dependent behavior’s resistance to cue extinction, it does not significantly reduce responding after cue extinction alone. Restoration of action-outcome control is the most likely explanation for NBQX’s bidirectional effects on cue-induced drug seeking because a previous study showed that the same dose of NBQX infused in the DLS restored action-outcome control of oral alcohol seeking, which was established using outcome devaluation to behaviorally distinguish between action-outcome and habit-like behavior (Corbit et al., 2014). Nevertheless, it is possible that NBQX could impact motivation for the drug or cue or impair retrieval of the cocaine-cue association. However, if NBQX could somehow impact retrieval of the cocaine-cue association, we would expect impaired retrieval to result in a decrease in cue-induced drug seeking, not the increase observed in the 0-cue control group. These findings have implications for the use of CET in SUD treatment. Although they show that DLS dopamine-dependent behavior is resistant to cue extinction, they also suggest that restoring action-outcome control alone could be harmful because, without other manipulations, it may increase cue-induced drug seeking and risk of relapse. The combination of restoring action-outcome control with other therapeutic methods including CET may be more effective and should be further studied.

A major obstacle in the use of CET to treat SUDs is context dependency (Kantak and Nic Dhonnchadha, 2011; Rich et al., 2019; Torregrossa et al., 2010). Even if this challenge is overcome, our findings suggesting that habitual behaviors are resistant to cue extinction, which may be important for the design and interpretation of clinical CET studies. There is ample evidence for a dissociation in the circuitry involved in DLS-independent and -dependent behaviors (Murray et al., 2015; Murray et al., 2012; Steinberg et al., 2020). Future experiments are required to determine how DLS dopamine-dependent response circuitry is strengthened, how it processes environmental stimuli to initiate behavior, and what role drug-associated cues play when they are reinforcing but not required for responding. Additionally, our findings reveal an interesting dissociation between the reinforcing properties of cues and how they guide behavior depending on response strategy, demonstrated by the susceptibility of conditioned reinforcement to cue extinction in SO-trained rats. These results further underscore the complex dynamics that may motivate drug seeking depending on environmental conditions.

Highlights.

• Habit-promoting second-order (SO) training results in resistance to cue extinction

• SO training results in cocaine seeking dependent on dopamine in the DLS

• Expression of AMPA receptor subunits is altered by SO training

• AMPA antagonism in the DLS reveals the effects of cue extinction