Introduction

An increasing body of evidence indicates that agonists at gamma aminobutyric acidB (GABAB) receptors can modulate the reinforcing effects of cocaine. For example, the selective GABAB agonist baclofen decreases cocaine self-administration in rats responding under fixed ratio (FR), progressive ratio (PR) and discrete trials (DT) schedules of reinforcement (Roberts et al. 1996; Roberts and Andrews 1997; Shoaib et al. 1998; Campbell et al. 1999; Brebner et al. 2000a). Evidence for the role of GABAB receptors in these effects is derived from studies showing that the selective GABAB antagonist CGP56433A dose-dependently attenuates the effects of baclofen in these procedures (Brebner et al. 2002). In clinical populations, baclofen has been shown to reduce cocaine craving and decrease cocaine intake when combined with weekly group therapy in individuals with a history of cocaine dependence (Gudeman et al. 1996; Ling et al. 1998). Although these findings suggest that baclofen and related compounds may hold promise as eventual pharmacotherapies for cocaine dependence, diminished enthusiasm for their potential use is derived from their untoward side effects, which include sedation and motor impairment.

A potentially intriguing development in the use of GABAergic drugs in treating cocaine dependence was the recent identification of positive allosteric modulators of the GABAB receptor. Positive allosteric modulators of the GABAB receptor have no intrinsic activity of their own, but interact synergistically with GABA to enhance its effects. Because these drugs only produce pharmacological activity in systems where GABA is already present, they are expected to possess a better side-effect profile than conventional GABAB agonists. Two such compounds, CGP7930 and GS39783, were recently shown to potentiate GABA-stimulated GTPγ[35S] binding in Chinese hamster ovary cells stably expressing GABAB receptors (Urwyler et al. 2001, 2003). Importantly, CGP7930 and GS39783 were ineffective in the absence of GABA, or when the effects of GABA were blocked by a competitive antagonist. CGP7930 and GS39783 also potentiated GABA-induced activation of inwardly rectifying potassium channels in Xenopus laevis oocytes and increased Ca2+ signaling in human embryonic kidney cells, despite producing no effect in these assays on their own (Urwyler et al. 2001, 2003). Radioligand binding studies showed that CGP7930 and GS 39783 increased the affinity and efficacy of conventional GABAB agonists, but did so by acting at a site distinct from that of the agonists. At present, there are no published data documenting the effects of these and other positive allosteric modulators in vivo; consequently, the ability of these drugs to modulate the reinforcing effects of cocaine is not known.

In the present study, we examined the effects of two positive modulators of the GABAB receptor, CGP7930 and GS39783, on cocaine self-administration in rats responding under several schedules of reinforcement. First, the effects of various doses of CGP7930 and GS39783 were examined on responding maintained under a progressive ratio schedule of reinforcement, a schedule that is uniquely suited to evaluate changes in the reinforcing efficacy of a drug (Richardson and Roberts 1996; Arnold and Roberts 1997; Stafford et al. 1998). Next, we examined the ability of fixed doses of CGP7930 and GS39783 to alter responding for cocaine under an FR schedule of reinforcement, evaluating the ability of these drugs to alter rates of responding for threshold versus supra-threshold doses of cocaine. Finally, we examined the ability of CGP7930 and GS39783 to alter responding for cocaine under a discrete trials procedure, a procedure in which access to cocaine is limited to a discrete number of trials each hour throughout the light/dark cycle. This procedure engenders a predictable pattern of cocaine self-administration and permits an examination of the initiation and circadian pattern of self-administration behavior (Fitch and Roberts 1993; Roberts and Andrews 1997; Brebner et al. 1999; Roberts et al. 2002).

Materials and methods

Animals and apparatus

All subjects were male, experimentally naive, Sprague-Dawley rats (Harlan, Indianapolis, Ind., USA) weighing approximately 350 g at the start of the experiments. Rats were housed in a large colony room maintained on a 12-h light/dark cycle with food and drinking water available ad libitum in the home cages. All subjects were maintained in accordance with the guidelines of the Institutional Animal Care and Use Committee of Wake Forest University and the “Guide for the Care and Use of Laboratory Animals” (Institute of Laboratory Animals Resources 1996). After a minimum 3-day acclimation period, each rat was anesthetized with a combination of ketamine HCl (100 mg/kg, IP) and xylazine HCl (8.0 mg/kg IP), and surgically implanted with a chronic indwelling jugular cannula that exited the skin on the dorsal surface of the scapulae (Roberts and Goeders 1989). Butorphanol HCl (1.0 mg/kg IP) was given immediately after surgery as a post-operative analgesic. Following cannula implantation, rats were housed individually in identical 25×25×25 cm operant conditioning chambers. The cannula was connected through a stainless-steel protective spring to a counterbalanced swivel apparatus that allowed free movement within the chamber. Rats that lost cannula patency were replaced by new, experimentally naive rats.

Progressive ratio

During initial training, each lever response was reinforced on a fixed ratio 1 (FR1) schedule of reinforcement. On this schedule, each lever press activated an injection pump that delivered 1.5 mg/kg per injection cocaine over a 4- to 5-s infusion duration (based on body weight). Concurrent with the start of each injection, a stimulus light located above the lever signaled a 20-s time-out during which the lever was retracted. During initial training, rats could obtain a maximum of 40 reinforcers during each 24-h session. Once 40 responses were completed, the lever was retracted and no further cocaine was available until the beginning of the next session. When rats exhibited a stable pattern of responding on the FR1 schedule over 5 consecutive days (defined as 40 infusions within 6 h and with regular post-infusion pauses), a progressive ratio (PR) schedule of reinforcement was introduced. On this schedule, the number of responses required for cocaine reinforcement (1.5 mg/kg per injection) incremented through the following progression: 1, 2, 4, 9, 12, 15, 20, 25, 32, 40, 50, 62, 77, 95, 118, 145, 178, 219, 268, 323, 402, 492 and 603 (Richardson and Roberts 1996). Break point was defined as the number of increments completed before a 1-h time period elapsed with no infusions. Responding on the PR schedule was considered stable when 3 consecutive days elapsed in which break points varied by no more than three increments. Once this criterion was met, doses of CGP7930 and GS39783 were administered in a random order, 15 min prior to the session. At least 2 days of baseline cocaine self-administration separated each test.

Fixed ratio

Rats were trained to respond for cocaine as described above. Once responding stabilized, cocaine was made available on an FR1 schedule of reinforcement during daily, 4-h test sessions. The effects of CGP7930 (10 mg/kg) and GS39783 (10 mg/kg) on responding for various doses of cocaine were then evaluated. Four unit doses of cocaine (0.19, 0.38, 0.75 and 1.5 mg/kg per injection) were tested in random order. At least two baseline sessions preceded each session in which CGP7930 or GS39783 was administered. The effects of vehicle administration were always examined on the day immediately prior to CGP7930 or GS39783 testing. CGP7930 and GS39783 were always administered 15 min prior to the session.

Discrete trials

Experimentally naive rats were trained to respond for cocaine on an FR1 schedule of reinforcement. When responding stabilized (see above), cocaine was made available on a discrete trials (DT) schedule, using parameters that engender a circadian pattern of responding (Fitch and Roberts 1993; Roberts et al. 2002). On this schedule, discrete trials were 10 min in duration and occurred at 20-min intervals (i.e. 3 times per hour) for the duration of the 24-h session. Each trial began with the introduction of a retractable lever into the cage. Completion of one lever press (FR1) resulted in an injection of cocaine (1.5 mg/kg per injection), illumination of a stimulus light for 20 s, retraction of the lever, and termination of the trial. Failure to emit a response within 10 min led to the termination of the trial. When a circadian pattern of responding stabilized (approximately 7–10 days), the effects of various doses of CGP7930 and GS39783 on cocaine self-administration were evaluated. Doses of CGP7930 and GS39783 were administered in a random order, at the beginning of the dark phase and 10 min prior to the first discrete trial of the session. At least 3 days of baseline cocaine self-administration separated each test.

Drugs

Cocaine HCl was obtained from the National Institute on Drug Abuse (Research Triangle Institute, Research Triangle Park, N.C., USA). CGP7930 and GS39783 were generously supplied by Novartis Pharma (Basel, Switzerland). Cocaine was dissolved in sterile saline; CGP7930 and GS39783 were suspended in a solution of methyl cellulose (0.05%) and distilled water. Doses of CGP7930 and GS39783 were administered IP in a volume of 1.0–3.0 ml/kg.

Data analysis

Progressive ratio data were analyzed via one-way ANOVA. Data from the fixed ratio experiment were analyzed via repeated-measures ANOVA, with dose of cocaine and treatment (drug versus vehicle) serving as factors. Data from the discrete trials procedure were also analyzed via repeated-measures ANOVA, with time and treatment (drug versus baseline) serving as factors.

Results

Progressive ratio

Under baseline conditions, cocaine engendered break points (±SEM) of 16.2 (±1.8) and 18.5 (±1.4) in rats treated with CGP7930 and GS39783, respectively. Administration of methyl cellulose (i.e. vehicle) did not significantly alter breakpoints in either group of rats. As shown in Fig. 1, CGP7930 produced robust decreases in cocaine-reinforced break points across all doses tested. A one-way ANOVA revealed a significant effect of dose [F(4,74)=6.474, P<0.001], and post-hoc tests revealed that doses of 10 and 30 mg/kg CGP7930 significantly decreased break points relative to baseline conditions (P<0.05). Figure 1 also shows that GS39783 decreased cocaine-reinforced break points, albeit to a lesser extent than that observed with CGP7930. A one-way ANOVA revealed a trend for dose [F(4,39)=79.167, P=0.074], with 30 mg/kg GS39783 significantly decreasing break point relative to baseline (P<0.05).

Fig. 1
figure 1

The effects of various doses of CGP7930 (left panel) and GS39783 (right panel) on cocaine self-administration in rats responding under a progressive ratio schedule of reinforcement. Ordinates reflect cocaine-reinforced break points on left and final ratio values on right; abscissae reflect doses of CGP7930 and GS39783 in mg/kg body weight. Points above “BL” represent the baseline values; points above “V” represent the effects of vehicle (0.05% methyl cellulose). Asterisks (*) indicate data points significantly different from baseline values. Vertical lines on data points represent the SEM

Full size image

Analysis of data from individual rats showed that CGP7930 and GS39783 reduced cocaine-reinforced break points in one of two ways. One subset of rats began responding immediately once the session began, but completed fewer ratios before reaching their final break point; a second subset of rats reached their typical break point, but only after waiting several hours before beginning to respond. Rats in this latter category were assigned a break point of 0 (see Materials and methods). Figure 2 depicts cumulative records of representative rats from each subgroup treated with 30 mg/kg CGP7930. Similar but less pronounced effects were also observed in rats treated with 30 mg/kg GS39783 (data not shown).

Fig. 2
figure 2

Cumulative records of cocaine-maintained responding in two representative rats treated with 30 mg/kg CGP7930 and responding under a progressive ratio schedule of reinforcement. Ordinates reflect cumulative number of responses; abscissae reflect time (h) since start of session. Short downward deflections represent cocaine infusions. Note: In one rat, CGP7930 decreased break point and number of infusions obtained, but did not alter latency to begin responding (left panel). In a second rat, CGP7930 increased the latency to begin responding, but did not alter break point or number of infusions obtained (right panel)

Full size image

Fixed ratio

Under the FR schedule of reinforcement, doses of 0.19, 0.38, 0.75 and 1.50 mg/kg per injection cocaine maintained markedly different response rates in rats treated with CGP7930 and GS39783 (Table 1). These values were not significantly altered by methyl cellulose administration in either group of rats. Figure 3 shows that a dose of 10 mg/kg CGP7930 decreased cocaine-maintained responding at lower but not higher unit doses of cocaine. A repeated-measures ANOVA revealed main effects of cocaine [F(3,61)=26.507, P<0.001], CGP7930 [F(1,61)=17.810, P<0.001], and a significant cocaine×CGP76930 interaction [F(3,61)=10.251, P<0.001]. Post-hoc tests revealed that CGP7930 significantly decreased responding for 0.19 mg/kg per injection cocaine (P<0.05). Figure 3 also shows that 10 mg/kg GS39783 decreased responding for a low dose of cocaine, but did not alter responding for higher doses. A repeated-measures ANOVA revealed a significant main effect of cocaine [F(3,63)=14.974, P<0.001], a significant cocaine×GS39783 interaction [F(3,63)=2.276, P=0.090], and a trend for GS39783 [F(1,63)=4.025, P=0.050]. Similar to that seen with CGP7930, post-hoc tests revealed that GS39783 significantly reduced responding for 0.19 mg/kg per injection cocaine (P<0.05). Doses of cocaine lower than 0.19 mg/kg per injection failed to maintain responding in the majority of rats tested (data not shown).

Table 1 Mean (±SEM) number of responses over 4 h in rats responding on an FR1 schedule of cocaine reinforcement
Full size table
Fig. 3
figure 3

The effects of selected doses of CGP7930 (left panel) and GS39783 (right panel) on cocaine self-administration in rats responding under a fixed ratio 1 schedule of reinforcement. Ordinates reflect number of responses during the 4-h session; abscissae reflect unit doses of cocaine in mg/kg per injection. Closed circles represent the effects of vehicle (0.05% methyl cellulose); open circles represent the effects of CGP7930 (10 mg/kg) and GS39783 (10 mg/kg). Asterisks (*) indicate data points significantly different from vehicle control values. Vertical lines on data points represent the SEM; where not indicated, the SEM fell within the data point

Full size image

Discrete trials

As shown in Fig. 4, the discrete trials procedure engendered a circadian pattern of cocaine self-administration that was characterized by high rates of responding during the dark phase and low rates of responding during the light phase of the daily light/dark cycle. Administration of saline or methyl cellulose did not alter the circadian pattern of cocaine self administration or the total number of infusions obtained (data not shown). In contrast, selected doses of baclofen, CGP7930 and GS39783 significantly decreased cocaine self-administration during the dark phase of the cycle when response rates were high. In rats treated with 2.5 mg/kg baclofen, a repeated-measures ANOVA revealed a main effect of time [F(11,119)=2.870, P=0.006], and a significant baclofen×time interaction [F(11,119)=2.899, P=0.006]. Post-hoc tests showed that this dose of baclofen significantly decreased responding at 2, 4, 6 and 8 h after administration (P<0.05). In rats treated with 30 mg/kg CGP7930, main effects of CGP7930 [F(1,143)=27.839, P=0.003] and time [F(11,143)=5.902, P<0.001] were observed, as was a significant CGP7930×time interaction [F(11,143)=4.513, P<0.001]. Post-hoc tests revealed that CGP7930 significantly decreased cocaine self-administration 4, 6, 8, 10, 12 and 14 h after administration (P<0.05). Similarly, in rats treated with GS39783, a repeated-measures ANOVA revealed main effects of GS39783 [F(1,143)=19.179, P=0.007] and time [F(11,143)=4.103, P<0.001], and a significant GS39783×time interaction [F(11,143)=2.011, P<0.045], with post-hoc tests showing that cocaine self-administration was significantly decreased 6, 8 and 10 h after administration (P<0.05). Higher doses of GS39783 (10 and 30 mg/kg) as well as lower doses of CGP7930 (3.0 and 10 mg/kg) failed to significantly decrease cocaine self-administration, although trends were observed with the two higher doses of GS39783 (0.05<P<0.10; data not shown).

Fig. 4
figure 4

The effects of selected doses of baclofen (top panel), CGP7930 (middle panel) and GS39783 (bottom panel) on cocaine self-administration in rats responding under a discrete trials 3 schedule of reinforcement. Ordinates reflect number of infusions per 2-h period; abscissae reflect time (h) since start of session. Dark horizontal bar over hours 1–12 indicates data obtained during the dark phase; light horizontal bar over hours 13–24 indicates data obtained during the light phase. Closed circles represent control (i.e. baseline) responding; open circles represent the effects of baclofen (2.5 mg/kg), CGP7930 (30 mg/kg) and GS39783 (3.0 mg/kg). Asterisks (*) indicate data points significantly different from baseline control values. Vertical lines on data points represent the SEM

Full size image

Discussion

The purpose of the present study was to examine the efficacy of two novel compounds, CGP7930 and GS39783, to reduce cocaine self-administration in rats responding under different schedules of reinforcement. Both CGP7930 and GS39783 have previously been shown to function as positive allosteric modulators at GABAB receptors in Chinese hamster ovary cells, Xenopus laevis oocytes and human embryonic kidney cells (Urwyler et al. 2001, 2003). Given the ability of these drugs to facilitate the activity of GABA in vitro, it was hypothesized that they would decrease cocaine self-administration in a manner similar to that previously reported with baclofen and other GABAB agonists (Roberts et al. 1996; Roberts and Andrews 1997; Shoaib et al. 1998; Brebner et al. 1999, 2000a, 2002; Campbell et al. 1999).

Both CGP7930 and GS39783 decreased cocaine self-administration on all three schedules of reinforcement examined. On the PR schedule, CGP7930 dose-dependently decreased cocaine-reinforced break points, with a dose of 30 mg/kg decreasing break points by over 50% of control (i.e. baseline) values. Similar, albeit smaller, effects were observed with GS39783. Testing was suspended at 30 mg/kg with both drugs because of limited solubility, and it is not known whether higher doses would have produced larger effects on cocaine-maintained responding under these conditions. We have previously reported that the GABAB agonists baclofen (Brebner et al. 2000a) and CGP44532 (Brebner et al. 1999) decrease cocaine-reinforced break points at doses that do not alter food-maintained responding, suggesting that these drugs attenuate the reinforcing effects of cocaine in the absence of gross motor impairment. Although food-maintained responding was not measured in the present study, no evidence of sedation or motor ataxia was observed in the hours immediately following CGP7930 and GS39783 administration.

Analysis of data taken from individual rats revealed that CGP7930 and GS39783 disrupted responding on the PR schedule in one of two ways. In a subset of rats, responding began with the onset of the session, but was terminated earlier, resulting in lower break points and fewer infusions. This pattern of responding is similar to that previously reported for baclofen and CGP44532, and appears to be the typical pattern of responding engendered by GABAB agonists in rats responding under PR schedules of cocaine reinforcement (Roberts et al. 1996; Brebner et al. 1999, 2000a). A second subset of rats completed their normal break point and received their normal number of infusions, but only after a 2- to 10-h period of non-responding at the beginning of the session. This pattern of responding was never observed after vehicle administration, indicating it was a pharmacological consequence of CGP7930 and GS39783 administration. The reasons these drugs produced a unique pattern of effects are not known, but it is possible that their novel mechanism of action played a contributing role.

Moderate doses of CGP7930 (10 mg/kg) and GS39783 (10 mg/kg) decreased responding for a low unit dose of cocaine (0.19 mg/kg per injection) on an FR1 schedule of reinforcement. Similar to that seen under the PR schedule, this effect was more robust in rats treated with CGP7930 than GS39783. Unit doses of cocaine lower than 0.19 mg/kg failed to maintain responding in the majority of rats tested, suggesting that the ability of CGP7930 and GS39783 to reduce cocaine-maintained responding on an FR1 schedule is limited to threshold doses. These effects are similar to those produced by baclofen, in that baclofen suppresses fixed-ratio responding for low (Shoaib et al. 1998; Campbell et al. 1999) but not high (Roberts et al. 1996; Brebner et al. 2000a) unit doses of cocaine. It is important to note that the effects of CGP7930 and GS398973 on this schedule differ markedly from those produced by dopamine antagonists. The D2 antagonist spiperone, for instance, shifts the dose-effect curve for cocaine to the right on FR schedules, increasing responding for all supra-threshold doses of cocaine (Hubner and Moreton 1991; Caine and Koob 1994). CGP7930 and GS39783, in contrast, shifted the cocaine dose-effect curve downward, suggesting that their effects in this procedure are not easily surmounted by compensatory increases in cocaine intake.

Consistent with previous studies (e.g. Fitch and Roberts 1993; Roberts and Andrews 1997; Brebner et al. 1999; Roberts et al. 2002), the DT procedure engendered a circadian pattern of responding that consisted of high numbers of infusions during the dark phase and low numbers of infusions during the light phase. We have previously shown that the GABAB agonists baclofen and CGP44532 decrease cocaine self-administration in a DT2 procedure, a procedure in which cocaine is made available during two 10-min discrete trials each hour (Roberts and Andrews 1997; Brebner et al. 1999). These effects are dose-dependent and readily apparent whether baclofen and CGP44532 are administered at the beginning of the dark cycle or 6 h after the dark cycle begins. Furthermore, these effects are observed under conditions in which food-maintained responding is not altered, ruling out the possibility of sedation or motor ataxia as contributing factors (Roberts and Andrews 1997; Brebner et al. 1999).

In the present study, a dose of 2.5 mg/kg baclofen decreased cocaine self-administration on a DT schedule in which cocaine was made available during three 10-min discrete trials each hour (DT3). Consistent with its effects on a DT2 schedule, baclofen produced its most robust effects in the 8 h immediately following its administration. Similarly, 30 mg/kg CGP7930 and 3.0 mg/kg GS39783 decreased cocaine self-administration during the dark phase of the DT3 schedule. The effects observed with CGP7930 were robust and evident up to 14 h after administration, considerably longer than that observed with baclofen. In contrast, the effects observed with GS39783 were generally modest and only apparent after a significant delay. Indeed, the number of cocaine infusions was not significantly decreased until 6 h after GS39783 administration. Also of interest was the finding that higher doses of GS39783 (10 and 30 mg/kg) failed to significantly alter cocaine self-administration on this schedule. GABAergic drugs generally produce dose-dependent effects in this procedure (Roberts and Andrews 1997; Brebner et al. 1999), and thus the failure to observe dose dependency with GS39783 was unexpected. The reasons GS39783 produced these atypical effects are not known, but pharmacokinetic factors related its central distribution might have played a role (see below).

Although both CGP7930 and GS39783 decreased cocaine self-administration under all three schedules examined, CGP7930 was generally more efficacious across the three procedures. Whether these differences in efficacy reflect differences in pharmacodynamics or differences in pharmacokinetics after systemic administration remain to be determined. Limited drug solubility limited the dose range of both drugs that could be examined, and identical doses of CGP7930 and GS39783 were tested in all three procedures. We recently collected data indicating that the two drugs are similarly effective at increasing baclofen-induced [35S]GTPγS binding across comparable dose ranges (unpublished observations), suggesting that the efficacy of these drugs in vitro are comparable. It is possible that the bioavailability of CGP7930 is greater than that of GS39783 after systemic administration, which would account for its greater efficacy to reduce cocaine self-administration in vivo. Consistent with this possibility, concentrations of CGP7930 in rat brain are greater than those of GS39783 after oral dosing (R. Aicholz and J. Kühnöl, Novartis Pharma AG, unpublished observations). In four rats tested with 30 μM/kg, brain concentrations of CGP7930 reached 309 (±148) and 337 (±121) nM/kg after 1 and 4 h, whereas concentrations of GS39783 reached only 106 (±20) and 133 (±37) nM/kg. It is likely that factors related to the central distribution of these drugs will ultimately determine their utility in clinical populations.

The mechanisms by which drugs acting at the GABAB receptor reduce cocaine self-administration are not known. It is unlikely that these effects can be attributed to gross motor impairment or a general decrease in consummatory behavior, as we have previously shown that GABAB agonists decrease cocaine self-administration at doses that do not suppress food-maintained responding (Roberts et al. 1996; Roberts and Andrews 1997; Brebner et al. 1999, 2000a). GABAB receptors are one of the most abundant classes of receptors in the CNS (Bischoff et al. 1999; Margeta-Mitrovic et al. 1999), and thus it is difficult to identify a particular locus of action for their role in cocaine reinforcement. It has been postulated that GABAB agonists produce their effects on cocaine self-administration by attenuating cocaine-induced increases in nucleus accumbens dopamine (Ashby et al. 1999). Dopamine release in the accumbens is under tonic inhibitory control by GABAergic interneurons that make synaptic contact with dopaminergic cell bodies in the ventral tegmental area, and we have previously reported that intra-VTA baclofen attenuates cocaine self-administration (Brebner et al. 2000b). Although CGP7930 and GS39783 act at a binding site distinct from that of baclofen, it is likely that they interact with a similar population of GABAB receptors to attenuate cocaine-maintained responding.

In conclusion, two novel positive allosteric modulators of the GABAB receptor, CGP7930 and GS39783, were effective at decreasing cocaine self-administration under three different schedules of reinforcement. These effects are similar to those reported previously for the GABAB agonist baclofen, and support the potential utility of positive allosteric modulators as pharmacotherapies for cocaine dependence. Differences in efficacy were observed between the two drugs, with CGP7930 consistently producing more robust effects than GS39783. These differences in efficacy may reflect differences in drug bioavailability after systemic administration, although further studies will be needed to confirm this possibility. It is recommended that future studies also examine the ability of these drugs to enhance the effects of baclofen and other GABAB agonists in vivo, particularly in regard to their effects on cocaine craving and intake.