Self-Administration of Benzodiazepine and Cocaine Combinations by Male and Female Rhesus Monkeys in a Choice Procedure: Role of α1 Subunit-Containing GABA_A Receptors

Abstract

Rationale.

Compounds lacking efficacy at the α1 subunit-containing GABA_A (α1GABA_A) receptor appear to have reduced abuse potential compared with those having measurable efficacy at this receptor, though their self-administration in nonhuman primates is dependent upon past drug experience.

Objectives.

We used a drug vs. drug choice procedure to evaluate the hypothesis that L-838,417, a compound lacking efficacy at α1GABA_A receptors, would not enhance cocaine choice in monkeys trained to self-administer cocaine. We also hypothesized that zolpidem, a compound with preferential modulation of α1GABA_A receptors and midazolam, a nonselective benzodiazepine, would enhance cocaine choice in this procedure.

Methods.

One female and three male rhesus monkeys chose between cocaine alone (0.1 mg/kg/injection) vs. the same dose of cocaine combined with midazolam (0.003-0.1 mg/kg/injection), zolpidem (0.003-0.3 mg/kg/injection), or L-838-417 (0.01-0.1 mg/kg/injection). In addition, we evaluated choice between saline and L-838,417 at select doses to determine whether L-838,417 would function as a reinforcer on its own.

Results.

Consistent with our hypotheses, midazolam- and zolpidem-cocaine mixtures were chosen over cocaine alone at sufficiently high doses. However, L-838,417-cocaine mixtures also were chosen over cocaine alone in 3 of 4 subjects with at least one dose. When available alone vs. saline, L-838,417 did not function as a reinforcer in any subject.

Conclusion.

Compounds that lack efficacy at α1GABA_A receptors may have low abuse potential compared to classic benzodiazepines, but self-administration of these compounds is context dependent.

Introduction

Benzodiazepines exert their therapeutic and abuse-related effects by binding the γ-aminobutyric acid type A (GABA_A) receptor which consists of α, β, and γ subunit families arranged in the sequence αβαβγ (Rudolph and Knoflach 2011; Sieghart and Savic 2018; Tan et al. 2011 for review). It has been proposed that the therapeutic and abuse-related effects can be dissociated by targeting selective GABA_A receptor subtypes (see Engin et al. 2018; Rudolph and Knoflach 2011). Benzodiazepines bind to an extracellular site on the native GABA_A receptor at the interface of the γ2 subunit and either α1, α2, α3, or α5 subunits but not α4 or α6 (e.g., Olsen 2018; Sieghart and Savic 2018). A growing body of preclinical evidence indicates that some behavioral effects of benzodiazepines like anxiolysis, sedation, and abuse potential can be attributed to specific α subunits or to groups of α subunits (see Engin et al. 2018; Rudolph and Knoflach 2011; Sieghart and Savic 2018 for review), raising the possibility of developing compounds to target individual α subunit-containing GABA_A receptors to increase therapeutic selectivity.

Tan and colleagues (2011) hypothesized that the neural basis for the abuse potential of benzodiazepines results from disinhibition of dopamine (DA) neurons mediated by α1 subunit-containing GABA_A (α1GABA_A) receptors. GABA interneurons in the ventral tegmental area (VTA) primarily express α1GABA_A receptors and send inhibitory signals to DA neurons that project to the nucleus accumbens (NAc). Nonselective benzodiazepines like diazepam bind to GABA_A receptors on GABA interneurons, decreasing the activity of GABA interneurons via activation of α1GABA_A receptors, in turn leading to disinhibition of DA neurons and an increase in the rate of firing of DA neurons, a physiological response generally associated with reinforcing properties of drugs (Tan et al. 2011). Based on these observations, Tan et al. (2010, 2011) proposed that compounds lacking activity at the α1GABA_A receptor, hereafter referred to as “α1-sparing”, should lack abuse potential, and some evidence is available to support this (Ator et al. 2010; Schwienteck et al. 2017; Tan et al. 2010).

However, the study conducted by Rowlett et al. (2005) with rhesus monkeys also provides evidence contradictory to Tan and colleagues’ (2010, 2011) hypothesis. While the α1-sparing compound L-838,417 was not self-administered as robustly as the other compounds tested, it still was self-administered above saline levels (Rowlett et al. 2005), suggesting some, albeit potentially reduced, abuse potential of an α1-sparing compound, a finding replicated with other α1-sparing compounds with selectivity for α2, α3, and/or α5 subunit-containing GABA_A receptors (α2GABA_A, α3GABA_A, and α5GABA_A receptors, respectively; Shinday et al. 2013). In addition, using rodent ICSS and/or two-bottle choice procedures, others have proposed a role of α2GABA_A and α3GABA_A (Reynolds et al. 2012) subunit-containing receptors or primarily α2GABA_A (Engin et al. 2014) in the reward-related effects of benzodiazepines.

There are some important contextual factors to note in the nonhuman primate self-administration studies cited above. In Ator and colleagues’ (2010) experiments, baboons were trained to self-administer cocaine prior to substitution testing with different compounds, and in Rowlett and colleagues’ (2005) experiments, rhesus monkeys were trained to self-administer methohexital, a short-acting barbiturate, prior to substitution testing with different compounds. Importantly, Shinday and colleagues’ (2013) directly evaluated drug history in groups of rhesus monkeys trained to either self-administer cocaine or midazolam under a PR schedule of reinforcement prior to testing different compounds for their reinforcing effects. Only rhesus monkeys trained and maintained under a midazolam baseline self-administered α1-sparing compounds, while those trained and maintained under a cocaine baseline did not self-administer any α1-sparing compound.

Overall, it appears that benzodiazepine reinforcement may be diminished by a history of cocaine experience, and with the exception of Bergman and Johanson (1985), this diminished effect is restricted to α1-sparing compounds. We do not know whether α1-sparing compounds lack reinforcing effects in cocaine-experienced subjects because those compounds are neutral or have punishing effects. Since α3GABA_A receptor subtypes exist on VTA DA neurons, it is conceivable that compounds lacking activity at α1GABA_A receptors could result in a net decrease in DA in the NAc, possibly resulting in decreased reinforcing effects or resulting in punishing effects (cf. Tan et al. 2011). Previous self-administration procedures used to evaluate reinforcing properties of α1-sparing compounds in cocaine-experienced nonhuman primates cannot distinguish between neutral and punishing stimuli, but choice procedures can be arranged to detect whether a drug is a reinforcer, a neutral stimulus, or a punisher (Freeman et al. 2014; Weed et al. 2017; Woolverton 2003). To investigate these issues, the current study used a drug-drug choice procedure to evaluate the hypothesis that an α1-sparing GABA_A compound, L-838,417, a partial positive allosteric modulator at α2, α3, and α5GABA_A receptors and an antagonist at α1GABA_A receptors, would function as a punisher of cocaine choice in rhesus monkeys trained to self-administer cocaine. We also hypothesized that zolpidem, a compound with preferential modulation of α1GABA_A receptors, and midazolam mixed with cocaine would be chosen over cocaine alone.

Materials and Methods

All procedures were approved by the University of Mississippi Medical Center’s (UMMC) Institutional Animal Care and Use Committee and were conducted in accordance with the National Research Council’s Guide for Care and Use of Laboratory Animals (8^th edition, 2011).

Subjects and Apparatus

Three male and one female adult rhesus monkeys (Macaca mulatta) served as subjects. Subjects 105-2008 (male) and 309-2009 (female) were experimentally naïve at the start of the experiment. These subjects were trained to lever press with food pellets prior to catheterization, and following catheterization, they were trained to track an injection of cocaine (0.1 mg/kg/injection) on one lever vs. saline or no programmed consequences on the opposite lever. The other subjects had experimental histories with food and cocaine. Most recently, subject 274-2009 (male) had a history choosing between cocaine injections available under fixed- (FR) and variable-ratio (VR) schedules (Huskinson et al. 2017), and subject 321-2009 (male) had a history choosing between food pellets available under FR and VR schedules (unpublished data). Subjects received unlimited access to water and were fed standard biscuits (Teklad 25% Monkey Diet, Harlan/Teklad, Madison, WI) to maintain healthy body weights. Fruit, foraging materials, multivitamins, light cycle, jacket/tether system, and homecage, were similar to that described previously (Huskinson et al. 2015) except that custom-designed double-lumen polyvinyl chloride catheters (70A PVC tubing; ID: 0.66 mm; OD: 2.4 mm; WALL: 0.38 mm; A.P. Extrusion Inc., Salem, NH) were used. Drug infusions were delivered over a period of 3 s at a rate of 0.18 ml/s via 30-ml plastic syringes seated in two infusion pumps (Razel Scientific, St. Albans, VT), one connected to each lumen of the catheter system to allow delivery to two separate drug solutions.

Surgery

Pre-surgery, subjects were injected with atropine sulfate (0.04 mg/kg, i.m.) and ketamine hydrochloride (10-20 mg/kg, i.m.) followed by inhaled isoflurane and preoperative antibiotics (cefazolin; 20-25 mg/kg, i.m.) and analgesics (carprofen, 2-4 mg/kg, s.c. and buprenorphine SR, 0.05 mg/kg, s.c.). Under aseptic conditions, the catheter was implanted into a major vein with the tip terminating near the right atrium. The distal end of the catheter was passed subcutaneously to the mid-scapular region, where it exited the subject’s back. The catheter was then threaded through the tether, out the back of the cubicle, and connected to a double-lumen swivel (Lomir Biomedical, Inc., Malone, NY). Postoperative analgesics (carprofen 4 mg/kg, p.o) were given daily for at least 3 days, and antibiotics (usually Keflex, 22.2 mg/kg, p.o. or i.m.; Eli Lilly & Company, Indianapolis, IN) were given when recommended by veterinary staff. Catheters were flushed daily with heparinized saline (40-100 U/ml). If a catheter became nonfunctional, it was removed, and a new catheter was implanted in a different vein once health was verified by veterinary staff.

General Procedure

Sessions began at the same time each day, and when responding was maintained by cocaine or saline, sessions were conducted 7 days/week. When any test compound (midazolam, zolpidem, or L-838,417) was available for presses on either lever, sessions were conducted 5 days/week to reduce the likelihood of dependence. Sessions began with 2 sample trials, 1 on each lever in a random order. Sample trials were signaled by illumination of the corresponding set of white lever lights. Following 5 consecutive lever presses, the consequence associated with the active lever was available. Sample trials ensured exposure to the contingencies programmed for each lever. Following sample trials, choice trials began. During choice trials, both sets of white lever lights were illuminated, both levers were active, and consequences associated with both levers were available.

For all trials, 5 consecutive lever presses (FR 5) had to occur on a single lever to result in delivery of the drug or vehicle solution associated with that lever. If switching between response levers occurred before the FR 5 was complete, the FR contingency was reset. Following completion of 5 consecutive responses on a single lever, all white lights were darkened, and the red lights above the corresponding lever were illuminated during the infusion(s) associated with that lever. After the injection, all lever lights were darkened for a 30-min timeout. During the injection and timeout, responses on either lever were recorded but had no programmed consequences.

Cocaine vs. Cocaine + Test Compound.

For these conditions, 0.1 mg/kg/injection of cocaine was associated with one lever and the same 0.1 mg/kg/injection of cocaine plus a test compound (midazolam, zolpidem, or L-838,417) was associated with the other lever. At the beginning of the experiment, the cocaine + test compound option was delivered as a mixture in one syringe and the cocaine alone option was delivered in a separate syringe but mixed with the same vehicle as the mixture. For practical reasons, this arrangement was eventually changed so that 0.1 mg/kg/injection of cocaine, dissolved in saline, was always delivered from one syringe, through one lumen of the catheter, regardless of which lever was pressed. When the combination lever was chosen, the test compound, dissolved in saline or varying concentrations of propylene glycol, was delivered from a second syringe, through the second lumen of the catheter immediately following the cocaine injection. This arrangement was deemed to be optimal for catheter longevity because it restricted propylene glycol delivery to one lumen of the catheter rather than both. Depending on the approach used, mixture vs. sequential delivery, appropriate control conditions were conducted. Thus is, cocaine vs. cocaine + saline or cocaine vs. cocaine + vehicle. The caption for Figure 1 indicates which approach was used for each subject and condition. Importantly, results were consistent regardless of which approach was used.

Fig. 1. Cocaine vs. Cocaine + Test Compound.

Mean (+/−SEM) percent choice of cocaine (0.1 mg/kg/injection) combined with saline (open circles), vehicle (open triangles; 40-100% propylene glycol), 0.003-0.1 mg/kg/injection of midazolam (closed circles; nonselective), 0.003-0.3 mg/kg/injection of zolpidem (gray triangles; α1GABA_A-preferring), and 0.01-0.1 mg/kg/injection of L-838,417 (open squares; α1GABA_A-sparing) when the same dose of cocaine alone (0.1 mg/kg/injection) was available on the opposite lever. Data are shown for individual monkeys in each panel; subject number is indicated in the bottom left of each panel. Each data point is the average of the final 3 sessions of the initial lever-injection pairing and the final 3 sessions of the reversal, and error bars are one standard error of the mean (SEM). The shaded area represents choice between 25-75% and was used to indicate doses that were considered punishing (≤25%), neutral (25-75%), or reinforcing (≥75%); the dashed line represents 50%. For 309-2009, the cocaine + test compound combination was delivered as a mixture (see text for details) for all data points except vehicle, which was delivered sequentially (see text for details). For 105-2008, the combinations were delivered as mixtures for all data points except L-838,417 at the 0.03 and 0.1 mg/kg/injection doses and vehicle, which were delivered sequentially. For 274-2009 and 321-2009, all combinations were delivered sequentially

The compounds tested were midazolam (0.003-0.1 mg/kg/injection), zolpidem (0.003-0.3 mg/kg/injection), and L-838,417 (0.01-0.1 mg/kg/injection). Subjects experienced each drug in a counterbalanced order and within a drug, experienced each dose in an irregular order. Between each drug, a cocaine (0.1 mg/kg/injection) vs. saline condition was conducted to: 1) ensure subject behavior tracked a known reinforcer, 2) determine how many sessions it took subjects to track the cocaine injection on each lever, and 3) serve as a “washout” period during which no test compounds were available.

All conditions were in effect for at least 3 sessions and until choice was stable, though more than 3 sessions were typically needed to meet criteria (group mean=6.6 sessions, range of subject means= 6.0-7.4 sessions). Stability required choice of one lever to be within 20% of the 3-session mean for 3 consecutive sessions, no upward or downward trends over 3 consecutive sessions, and completion of all trials in a session. There were two exceptions to these stability criteria: 1) a trend was acceptable if a ceiling or floor effect was obtained, for example, 8, 9, and 10 choices over 3 consecutive sessions or 2, 1, and 0 choices over 3 consecutive sessions, and 2) if stability criteria were reached at indifference (40-60%), at least one more session was conducted to ensure subjects were not in the process of switching from one lever to the other. After this additional session, if stability criteria were met, the condition was ended. If the additional session revealed a trend, the condition was continued until stability criteria were again satisfied, and the indifference criterion was not applied more than once per condition. Once choice was stable, the schedule and reinforcer associated with each lever was reversed, and choice was re-determined according to the same stability criteria. Each data point represents the average of 3 stable sessions of a lever-injection pairing and 3 stable sessions of its reversal.

L-838,417 vs. Saline.

Cocaine combined with L-838,417 was unexpectedly chosen over cocaine alone in 3 of 4 subjects, so we evaluated whether this compound would function as a reinforcer when delivered alone. Following a cocaine vs. saline condition, saline was made available on the lever previously paired with cocaine, which was the lever most recently chosen on greater than 80% of trials. L-838,417 was made available on the opposite lever, which was the lever most recently chosen on fewer than 20% of trials. This approach was taken because it is a more stringent test of a functional reinforcer if a subject will switch its choice to the opposite lever from the lever most recently selected on the majority of trials. The L-838,417 dose used for this condition was individually determined and was a dose that, when combined with cocaine, was chosen on ≥75% for 3 of the subjects. This dose was 0.03 mg/kg/injection for the subject for whom L-838-417 was not chosen when mixed with cocaine. For subject 274-2009, we also evaluated a lower dose that was not chosen when combined with cocaine (0.056 mg/kg/injection). These sessions were conducted for the minimum number of sessions required to reach stability in the cocaine + L-838,417 condition, unless the subject stopped completing sessions (309-2009) or percent L-838,417 choice met stability criteria at ≥75%.

Data Analysis

In cocaine vs. cocaine + test compound conditions, mean percent choice from the stable sessions of each lever-injection pairing and its reversal were plotted for each condition and were used to evaluate data from individual subjects. A dose of a test compound combined with cocaine was considered to be a reinforcer if choice of the lever associated with the test compound was ≥75% of the total choice trials for that condition. Conversely, a dose of a test compound was considered to be a punisher if choice of the lever associated with the test compound was ≤25% of the total choice trials for that condition. If choice was between 25-75%, that dose was considered neutral. In the L-838,417 vs. saline condition, L-838,417 did not function as a reinforcer in any subjects in terms of choice behavior. Despite not functioning as a reinforcer, 3 of 4 subjects were still completing sessions. For this reason, session time was also evaluated to determine if responding had decreased over time. Demonstrably longer session times would be expected if both outcomes were non-reinforcers

Drugs

Cocaine hydrochloride was provided by the National Institute on Drug Abuse (Rockville, MD), midazolam (Hospira, Inc., 5 mg/mL) was purchased from the pharmacy at the University of Mississippi Medical Center (Jackson, MS), and zolpidem and L-838,417 were purchased from Tocris-Cookson (Ellisville, MO). Cocaine and midazolam were prepared using 0.9% sterile saline, except during some mixture conditions when cocaine was prepared using sterile saline mixed with 20-80% propylene glycol to match the cocaine + test compound vehicle delivered through the opposite lumen. Zolpidem and L-838,417 were dissolved in propylene glycol and diluted in sterile saline to a final concentration of 20-80% propylene glycol (depending on drug concentration). Injections were delivered over a 3-s infusion at a rate of 0.18 ml/s. All solutions were passed through a 0.22 μm Millipore filter prior to administration.

Results

Cocaine vs. Cocaine + Test Compound.

Figure 1 shows percent mixture-lever choice for individual subjects. As expected, when choice was between the same dose of cocaine (0.1 mg/kg/injection; data points above “C 0.1”) available for presses to either lever, choice was indifferent between the two options for all subjects tested. No data point is shown for subject 321-2009, because all conditions were conducted under the sequential-delivery procedure. Similarly, when choice was between the same dose of cocaine, but for one option the cocaine injection was immediately followed by a saline or vehicle injection (40-100% PG; data points above Veh), choice also was indifferent for all subjects. During midazolam and zolpidem conditions, all subjects chose midazolam or zolpidem combined with cocaine over cocaine alone (≥75%) for at least one dose. At a sufficiently small doses for 3 of 4 subjects, choice was considered neutral, falling between 25 and 75%. Exceptions include subject 274-2009 with midazolam and subject 321-2009 with zolpidem. During L-838,417 conditions, 3 of 4 subjects chose the combination over cocaine alone on ≥75% of trials for at least one dose. Subject 105-2008 did not choose the combination over cocaine alone at any doses tested, and larger doses were not tested because of solubility issues.

It also is important to note the between-subject variability in the doses that functioned as reinforcers as subjects who were experimentally naïve at the start of the experiment tended to require larger doses of midazolam and zolpidem to function as reinforcers (Figure 1). In this regard, the minimally effective dose of midazolam for experimentally naïve subjects 309-2009 and 105-2008 were 0.1 and 0.03 mg/kg/injection, respectively. For subjects 274-2009 and 321-with experimental histories at start of experiment, minimally effective midazolam doses were ≤0.003 and 0.01 mg/kg/injection, respectively. For zolpidem, minimally effective doses for relatively naïve subjects 309-2009 and 105-2008 were 0.03 and 0.1 mg/kg/injection, respectively, and for relatively experienced subjects 274-2009 and 321-2009 were 0.01 and ≤0.003 mg/kg/injection, respectively. Collectively, this variability between relatively naïve vs. relatively experienced subjects constituted an approximate 10-fold difference in the minimally effective doses of midazolam and zolpidem. However, L-838,417 was less variable across subjects, and this is somewhat complicated because of the lack of reinforcing effect found with one subject.

L-838,417 vs. Saline.

Figure 2 shows choice of the drug-associated lever during the most recent cocaine vs. saline condition or between a dose of L-838,417 and saline. All subjects tracked 0.1 mg/kg/injection of cocaine, reaching stable cocaine choice above 75% within 2-5 sessions. Conversely, none of the subjects tracked L-838,417 despite conducting these sessions for much longer than the cocaine vs. saline condition. For subject 321-2009 (bottom right panel), L-838,417 choice reached 80% on session 16 and 90% on session 17 but did not meet stability criteria, and choice decreased to 50% on session 18. For one subject (105-2008), whose session time (see Figure 3) for L-838,417 vs. saline remained normal despite drug choice remaining consistently under 75%, we also completed a saline vs. saline condition to evaluate the pattern of choice that would result with a known non-reinforcer. For this subject, choice in the L-838,417 condition was similar to the saline vs. saline condition; this subject was the only subject for whom L-838,417 did not function as a reinforcer at any dose in the cocaine vs. cocaine + test compound conditions.

Fig. 2. L-838,417 vs. Saline.

Percent choice of the drug-associated lever (initial-lever pairing) when choice was between 0.1 mg/kg/injection of cocaine vs. saline (open circles) and when choice was between 0.03 (closed circles), 0.056 (gray triangles), or 0.1 (open squares) mg/kg/injection of L-838,417 vs. saline. One subject (105-2008, top right panel) also completed a saline vs. saline (gray squares) condition to determine what choice would look like with a known non-reinforcer. Data are shown for individual monkeys in each panel; subject number is indicated in the bottom right of each panel. Each data point represents a single session, and all sessions conducted for the initial lever-injection pairing are shown. The bottom of the shaded area begins at 25%, the dashed line represents 50%, and the shaded area ends at 75% (i.e., criterion to be considered a reinforcer).

Fig. 3. L-838,417 vs. Saline.

Session time (min) when choice was between 0.1 mg/kg/injection of cocaine vs. saline (open circles), when choice was between 0.03 (closed circles), 0.056 (gray triangles), or 0.1 (open squares) mg/kg/injection of L-838,417 vs. saline, and for one subject (105-2008, top right panel) when choice was between saline vs. saline (gray squares). Data are shown for individual monkeys in each panel; subject number is indicated in the bottom right of each panel. Each data point represents a single session, and all sessions conducted for the initial lever-injection pairing are shown. The dashed line represents 360 min or the minimum session time possible.

Figure 3 shows total session time (min) when choice was between cocaine and saline or between a dose of L-838,417 and saline. During the cocaine vs saline condition, session times for all subjects were close to 360 min, the shortest session time possible (group mean=364.7 min, range of group mean=361.2-372.8 min). For L-838,417 vs. saline, 3 of the 4 subjects had session times that tended to be much longer than during cocaine vs. saline sessions. For these subjects, session times were similar to cocaine vs. saline during the first 3-6 sessions of the L-838,417 vs. saline condition before becoming longer and variable from session to session. Subject 309-2009 stopped completing sessions altogether. For 321-2009, during sessions 16 and 17 when L-838,417 choice reached 80 and 90%, respectively (Figure 2), session times were 558 and 473 min, respectively. These times are 2-3 hours longer than cocaine vs. saline sessions. For 105-2008, session times were similar across cocaine vs. saline and L-838,417 vs. saline conditions. Session times were somewhat longer during saline vs. saline for this subject, but not as markedly as seen with the L-838,417 vs. saline conditions for the other 3 subjects.

Discussion

The most important and unexpected finding was that cocaine, when mixed with the α1-sparing compound, L-838,417, was chosen over cocaine alone, demonstrating that cocaine-trained animals will self-administer an α1-sparing compound under some circumstances. This finding is in direct contrast to previous studies with α1-sparing compounds under either a fixed-ratio or progressive-ratio schedule of reinforcement in which the compounds were substituted for the cocaine-training condition (Ator et al. 2010; Shinday et al. 2013). Moreover, this outcome cannot be accounted for within Tan and colleagues’ (2011) hypothesis that the mechanism of benzodiazepine reinforcement results from activation of the mesolimbic DA system by disinhibition of VTA interneurons via an α1GABA_A receptor mechanism. Without α1GABA_A receptor activation, a different mechanism likely contributed to L-838,417’s ability to maintain choice behavior when combined with cocaine. Similarly, the disinhibition of DA hypothesis via α1GABA_A receptors also cannot account for the ability of α1-sparing compounds to function as reinforcers with barbiturate-or benzodiazepine-experienced subjects (Rowlett et al. 2005; Shinday et al. 2013). We do not know what the mechanism is for the reinforcing properties of α1-sparing compounds under these specific contingencies, though it is feasible that it involves modulation by α2GABA_A and/or α3GABA_A receptors, either in the VTA or elsewhere in the mesolimbic reward pathway (cf. Engin et al. 2014; 2018).

In contrast, L-838,417 did not maintain choice behavior when it was available as a single drug. This finding replicates previous studies in which an α1-sparing compound was not self-administered by cocaine-experienced baboons or rhesus monkeys (Ator et al. 2010; Shinday et al. 2013). Importantly, this finding is what would be predicted by Tan and colleagues’ (2011) hypothesis described above. It is likely that multiple mechanisms, perhaps playing different roles under different contexts, contribute to the reinforcing properties of benzodiazepines. Whatever the mechanism, it appears that α1-sparing compounds like L-838,417 are relatively weak reinforcers with less abuse potential than traditional, nonselective benzodiazepines, an effect that is especially apparent in cocaine-experienced subjects.

Though not evaluated here, considerable research has pointed to α1-sparing compounds as candidates for improved anxiolytics, particularly with α2GABA_A, α3GABA_A, and/or α5GABA_A receptors implicated in mediating benzodiazepine-induced anxiolysis (Fischer et al. 2010; 2011; Licata et al. 2005; Low et al. 2000; McKernan et al. 2005; Rowlett et al. 2005; for review, see Engin et al. 2018). It may be productive for future research to develop α1-sparing compounds to be evaluated as anti-anxiety medications with reduced abuse potential with attention given to past drug experiences. There also is evidence from previous work that α1GABA_A receptors are involved in benzodiazepine tolerance and dependence (Ator et al. 2010; Fischer et al. 2013; Mirza and Nielson 2006; Vinkers et al. 2012). If compounds with functional selectivity for α2GABA_A and/or α3GABA_A receptors produce less tolerance and dependence (e.g., Ator et al. 2010), it would provide yet another reason for future work to develop and evaluate these compounds for use in the treatment of anxiety disorders.

As expected, when mixed with cocaine, both midazolam and zolpidem were chosen over cocaine alone in all 4 subjects with at least 1 dose tested. These findings replicate many prior studies, showing that cocaine-, barbiturate-, and benzodiazepine-experienced nonhuman primates will self-administer zolpidem and nonselective benzodiazepines (e.g., Ator et al. 2010; Bergman and Johanson 1985; Fischer et al. 2016; Rowlett et al. 2005; Shinday et al. 2013) and in a similar choice procedure, that rhesus monkeys will choose a remifentanil (ultra-short acting mu-opioid) + midazolam mixture over remifentanil alone (Weed et al. 2017). We did not evaluate midazolam and zolpidem alone as we did with L-838,417, because their effects as reinforcers is well established in nonhuman primates. As noted in the Results, we observed between-subject variability in the dose(s) that functioned as reinforcers. Subjects without an experimental history at the start of the experiment (309-2009 and 105-2008) typically required larger doses to obtain a reinforcing effect than the two subjects with experimental histories at the start of the experiment (274-2009 and 321-2009). It is possible that a longer experimental history with choice procedures enhanced more experienced subjects’ ability to discriminate subtle differences between the consequences associated with each lever, resulting in a leftward shift in the dose-response function. The experienced subjects also had a much longer history with cocaine self-administration than the naïve subjects. It is therefore also possible that the longer cocaine history altered subjects’ ability to discriminate subtle differences between the consequences associated with each lever.

Finally, our results with L-838,417 highlight one of the strengths of choice procedures (see Banks and Negus 2012) and in particular, of discrete-trial, drug-drug procedures. That is, these types of procedures are known for their sensitivity to small, yet meaningful effects; a manipulation need only to “tip the scale” to shift choice from indifference between alternatives to choice for one alternative over the other (e.g., Freeman and Woolverton 2011; Weed et al. 2017; Woolverton 2003). Using this choice procedure, we were able to show that co-administration of L-838,417 with cocaine was sufficient to shift choice from indifference to choice of the cocaine + L-838,417 alternative despite the fact that this compound does not, on its own, maintain choice behavior in cocaine-experienced subjects (Ator et al. 2010; Shinday et al. 2013; current study). It may have been possible to detect these effects in other (nonchoice) procedures. For example, mixing L-838,417 with cocaine in a PR procedure might shift the cocaine dose-effect function leftward, an effect that would be consistent with the result obtained in the current experiment. Indeed, relatively weak reinforcers like caffeine and nicotine, delivered as a pretreatment or under self-administration conditions, can enhance the reinforcing effects of cocaine in rats and monkeys while not reliably functioning as reinforcers when tested alone under similar conditions (e.g., Comer and Carroll 1996; Freeman and Woolverton 2009; Schenk et al. 1994). However, PR and similar procedures are susceptible to rate-dependent effects of drugs whereas choice procedures are relatively independent of such effects, therefore facilitating interpretation of changes in dose-response functions. Whether and how subtle reinforcing effects contribute to actual abuse, or polydrug abuse, warrants further investigation. Regardless, if we consider previous work on anxiolysis, abuse potential, tolerance, and dependence (Ator et al. 2010; Fischer et al. 2010; 2011; 2013; Licata et al. 2005; Low et al. 2000; McKernan et al. 2005; Mirza and Nielson 2006; Rowlett et al. 2005; Shinday et al. 2013; Vinkers et al. 2012) in light of our findings, the literature collectively supports future development of α1-sparing compounds as potential anti-anxiety medications with reduced abuse potential, and perhaps reduced dependence potential.