Low-dose polypharmacology targeting dopamine D1 and D3 receptors reduces cue-induced relapse to heroin seeking in rats

Abstract

Chemical compounds that target dopamine (DA) D1 or D3 receptors have shown promise as potential interventions in animal models of cue-induced relapse. However, undesirable side effects or pharmacodynamic profiles have limited the advancement of new compounds in preclinical studies when administered as independent treatments. In this series of experiments, we explored the effects of co-administration of a D1-recepter partial agonist (SKF 77434) and a D3-receptor antagonist (NGB 2904) in heroin-seeking rats within a ‘conflict’ model of abstinence and cue-induced relapse. Rats were first trained to press a lever to self-administer heroin and drug delivery was paired contingently with cues (e.g., light, pump noise). Self-initiated abstinence was facilitated by applying electrical current to the flooring in front of the levers. Lastly, a relapse response was provoked by noncontingent presentation of conditioned cues. Prior to provocation, rats received a systemic injection of SKF 77434, NGB 2904, or a combination of both compounds to assess treatment effects on lever pressing. Results indicated that the co-administration of low (i.e., independently ineffective) doses of both compounds was more effective in reducing cue-induced relapse to heroin seeking than either compound alone, with some evidence of drug synergism. Follow-up studies indicated that this reduction was not due to motoric impairment nor enhanced sensitivity to the electrified flooring and that this treatment did not significantly affect motivation for food. Implications for the treatment of opiate use disorder and recommendations for further research are discussed.

1. Introduction

The opioid epidemic remains a global crisis. Current pharmacological treatments for opiate use disorder¹, such as methadone and buprenorphine/naloxone, are only partially effective and exhibit extremely high rates of relapse.^2,3 Stimuli that have become associated with opiate use (cues) are believed to contribute to relapse cycles; cue exposure enhances self-reports of craving and dysphoria in humans with a history of chronic opiate use^4,5 and reliably increases opiate-seeking behaviors in animal models of relapse.⁶

Opiate-associated cues enhance transmission of mesolimbic dopamine (DA), indicating DA receptors as potential targets for pharmacological interventions. DA D1 receptors (D1R) are implicated in the experience of opiate- and cue-induced reward⁷ and D1R partial agonists and antagonists have been shown to be effective in reducing opiate-seeking behaviors.^8–11 However, efficacious doses of these agents can elicit motoric/aversive side effects,^12–14 possibly due to incidental activity at DA D2 receptors (D2R)^15,16 or to the wide expression of D1R in the central and peripheral nervous system.^17,18 By contrast, DA D3 receptors (D3R) are largely isolated to the nucleus accumbens (NAcc) shell, olfactory tubercle, and the islands of Calleja.^17,18 There is evidence that D3R play a central role in conditioned responses to drug cues^19,20 and antagonists at D3R have been shown to attenuate opiate self-administration and cue-driven behaviors.^21–24

These findings lend support for these compounds as potential interventions for opiate relapse; indeed, the National Institute on Drug Abuse (NIDA) has indicated selective D3R ligands as one of the ‘ten most wanted’ pharmacological therapies for battling the opioid crisis.²⁵ However, factors such as poor selectivity, specificity, and bioavailability have led to inconsistent results from new candidates and limited their advancement in clinical trials.^26–29 Of particular concern, even selective D1R and D3R agents can exhibit affinity at D2R when administered at higher doses;^15,16,30 it is well-documented that humans treated with D2R antagonists (e.g., antipsychotic medications) often report pronounced, persistent dysphoria, a side effect that is highly correlated with the initiation of substance use and relapse.^31,32

At this time, the vast majority of preclinical studies have emphasized DA receptor-modulating agents as an intervention for cocaine seeking; while promising, this research has suffered similar limitations.^33–36 Previous work from our lab attempted to respond to this problem by exploring a novel, poly-pharmacological treatment in rats with a history of cocaine self-administration.³⁵ After testing the efficacy of independent treatment with a D1R partial agonist (SKF 77434) or a D3R antagonist (NGB 2904), Galaj et al.³⁵ then explored systemic combination treatments using low (ineffective) doses of both compounds. These combinations were more effective in reducing cue-induced reinstatement of cocaine seeking than all independent doses tested, an effect that could not be attributed to non-specific, motoric side effects, such as impairment in lever-pressing capabilities. The authors proposed that these benefits were due to synergistic activity at D1R-D3R heterodimers, which form only within the few structures where these receptors co-localize, such as the NAcc shell.^35,36

It is currently unknown if this poly-drug approach could be effective in treating opiate-seeking behaviors. It has been proposed that D1R-D3R interactions play a crucial role in opiate reward and tolerance,²⁴ but it is reasonable to suspect that the response to pharmacological interventions may differ in opiate-seeking animals from those with a history of cocaine use.^24,37,38 Nevertheless, given the largely overlapping neural circuitry underlying both psychostimulant and opioid reward and cue-driven behavior, it is also reasonable to expect similar effects of this treatment.⁷ The following series of experiments attempted to demonstrate the efficacy of SKF 77434 and NGB 2904, administered independently and in combination, in preventing cue-induced relapse after a period of abstinence from opiate-seeking. We hypothesized that the poly-drug combinations of a D1R partial agonist and a D3R antagonist would be more effective at reducing cue-induced relapse than the individual compounds themselves.

Importantly, this treatment was evaluated within the context of an animal ‘conflict’ model of abstinence and relapse.^39,40 There has been some debate as to the external validity of extinction/reinstatement procedures when investigating drug abstinence and cue-induced relapse, as extinguished drug seeking after drug removal fails to emulate the human experiences of conflict and choice.^41–43 In the ‘conflict’ model, drug seeking continues to be reinforced by drug delivery, but self-initiated abstinence can be facilitated using electrified flooring adjacent to the drug-taking manipulandum (e.g., levers), forcing an animal to subject itself to foot shocks in order to obtain the drug. We believe that this model, which presents the animals with the choice to avoid the negative repercussions of drug seeking, more closely emulates the human experience of abstinence and relapse than extinction/reinstatement procedures.^40,42 Thus, if this poly-drug treatment is indeed effective at reducing cue-induced relapse in rats, this model may enhance support for the intervention as a relapse prevention treatment for humans with opioid use disorder.

2. Materials and Methods

2.1. Subjects and Facilities

Subjects were adult male Long-Evans rats (Charles River, Kingston, NY) that were individually housed in standard laboratory ‘shoebox’ cages with ad libitum access to water and food (LabChow diet). Cages were kept in windowless rooms with an automated 12-h light/12-h dark cycle and rats were only engaged in experimental procedures during the dark (active) phase. These experiments were conducted in accordance with the National Research Council’s Guide for the Care and Use of Laboratory Animals and approved by the Queens College Institutional Animal Care and Use Committee.

2.2. Drugs

The heroin hydrochloride used in this study was a gift from the National Institute on Drug Abuse (NIDA). In intravenous self-administration (IVSA) studies, heroin was delivered via an implanted catheter in a saline solution (0.9% NaCl) in infusions of 0.125 mL and at an estimated dose of 0.05 mg/kg/infusion. In the nociception studies, a heroin-saline solution was delivered via intraperitoneal (IP) injection at a dose of 0.5 mg/kg. Methohexital sodium (Brevital; Southern Anesthesia & Surgical, Inc., South Carolina, USA) was administered via IVSA catheter assembly following relapse test sessions (1 mg in 0.1 mL of saline) to verify catheter patency. NGB 2904 and SKF 77434 were purchased in powder form from Bio-Techne (Minnesota, USA); NGB 2904 was dissolved in 10% 2-Hydroxyprolyl β-cyclodextrin and SKF 77434 in distilled water; both were delivered via IP injection (1 mL/kg at various concentrations).

2.3. Equipment

Operant conditioning chambers were operated by Med Associates interfacing and software (Georgia, VT). IVSA chambers (Experiment 1) featured a fluid line connected to a Razel syringe pump (3.33 r.p.m.) and a constant-current aversive stimulator (Model ENV-414; Med Associates) connected to the stainless-steel flooring rods. Food chambers (Experiment 2) were equipped with an electronic food dispenser and trough. All were equipped with two levers (counterbalanced active/inactive) and a white cue light above each lever. Tail-flick latency assays (Experiment 3) were conducted using a tail-flick analgesia meter (IITC Inc. Life Science, California, USA). For further specifications, see our previous work.^40,44

2.4. Surgery

The animals in Experiment 1 required surgical preparation as previously described.⁴⁰ In short, dental acrylic was used to mount a steel cannula to the skull; the subcutaneous end of the cannula was connected to a silastic catheter that was implanted into the right jugular vein to enable fluid release into the superior vena cava. Rats received 3 recovery days post-surgery, during which they received a daily infusion of gentamicin sulfate solution (0.4 mg in 0.1 mL saline) through the catheter assembly to prevent infection, followed by a heparin-saline solution (0.05 mL, 50 U.S.P.) to prevent blood clotting within the catheter. After 3 days and for the remainder of the study, rats received only heparin post-session to maintain catheter patency.

2.5. Experiment 1: Cue-Induced Relapse to Heroin-Seeking

After recovery from surgery, each rat (N = 129) was randomly assigned to a single operant conditioning chamber in which to complete all study sessions. Prior to addressing our primary aims regarding DA receptor modulating treatments, a pilot study (details below) was conducted using a sub-group of these animals to validate the use of a proportionally defined electrical current to be used during the cue-induced relapse ‘test’ session.

2.5.1. Training Phase

During the training phase, rats underwent daily 3-h sessions with unimpeded access to both levers (i.e., the electric barrier was inactive). Presses on the inactive lever were counted but produced no consequences. Presses on the active lever were programmed on a fixed ratio one (FR1) schedule of reinforcement; each press activated the syringe pump to deliver an infusion of heroin and illuminated the cue light above the active lever for 20 s. During this 20-s period, any additional presses on the active lever were counted but produced no consequences. Presses on both levers were recorded in 30-min increments. Daily training sessions were conducted for a minimum of 15 days (one session per day) and until a stable pattern of heroin self-administration was observed as previously described.⁴⁰

2.5.2. Electric Barrier/Abstinence (EBA) Phase

During the EBA Phase, animals underwent daily 35-min sessions as described previously.⁴⁰ At session onset, the two-thirds of flooring adjacent to the levers was electrified; the third of flooring furthest from the levers provided a non-electrified ‘safe zone’. In order to reach the active lever to self-administer a heroin infusion, the rat needed to leave this ‘safe zone’ and subject itself to foot shocks.

On the first session of the EBA Phase, an electric current strength of 0.25 mA was used for all rats. Before each session, rats were placed centrally within their assigned chambers and for the first 5 min of the session the levers were retracted. After 5 min the levers were inserted, permitting heroin IVSA for the remaining 30 min. If a rat pressed on the active lever during the session, the subsequent session’s current strength was increased by 0.07 mA. If a rat did not press on the active lever, the same current strength was used during the subsequent session. This procedure continued until the rat did not press on the active lever for 3 consecutive sessions (i.e., achieved ‘abstinence’) or the rat continued drug-seeking at a maximum current strength of 1.5 mA (these rats were discontinued; n = 2). The number of days spent in the EBA phase and electric current strength at completion was recorded.

2.5.3. Cue-Induced Relapse Test

During the test session (35 min), rats were again placed centrally within their respective chambers prior to session onset. For the first 5 min the levers were withdrawn and the electric barrier (where applicable) was active. After 5 min the levers were inserted, and animals were presented with non-contingent, heroin-associated cues: the cue light above the assigned active lever illuminated for 20 s and the syringe pump activated to deliver an infusion of saline. These cues (light, sound of the pump motor, infusion) were presented non-contingently at 3-min intervals, as well as contingently with presses on the active lever (FR2). Presses on both levers were recorded for analysis.

Pilot Study to Confirm Test Parameters.

For this study, rats (n = 7) were randomly assigned to one of two groups. All rats received an IP injection of saline 30 min prior to the cue-induced relapse test session. For the No Barrier group (n = 3), the electric barrier was inactive for the duration of the session. For the Barrier group (n = 4), the strength of the electric barrier was set to 25% of the amperage at which the rat achieved abstinence criteria.

The Effects of NGB 2904 and SKF 77434 on Cue-Induced Relapse.

The primary study was organized into three arms. All rats underwent surgical procedures, heroin IVSA training, and the EBA phase as described above until achieving ‘abstinence’ criteria. In the NGB 2904-Only arm, n = 45 rats were randomly assigned to receive an IP injection of vehicle or 0.25, 1, 1.5 or 2 mg/kg of NGB 2904 30 min prior to the cue-induced relapse test session. In the SKF 77434-Only arm, n = 40 rats were randomly assigned to receive an injection of vehicle or 0.25, 0.5, 1 or 2 mg/kg of SKF 77434. In the ‘Combination’ arm, n = 42 rats were randomly assigned to receive an injection of both vehicles (0.5 mL/kg of each) or one of four unique combinations of both compounds; these combinations consisted of doses that were deemed ineffective when administered independently in the first two arms (expressed as NGB 2904+SKF 77434: 0.25 mg+0.25 mg, 0.25 mg+0.5 mg, 1 mg+0.25 mg, 1 mg+0.5 mg/kg). In all three arms, the strength of the electric barrier was set to 25% of the amperage at which each rat achieved abstinence.

2.6. Experiment 2: Food Self-Administration Under a Progressive Ratio (PR) Schedule

To test the possibility that any observed differences in ‘relapse’ activity may be due to motoric side effects (i.e., impairment in lever pressing capability) of NGB 2904 or SKF 77434, we evaluated the effects of these compounds on lever pressing reinforced by food, a procedure that produces many times more lever presses than typically observed in the cue-induced relapse paradigm. Naive rats (N = 24) were placed on a restricted diet 3 days prior to beginning daily food self-administration training sessions to maintain weights at 85% of their free-feeding values. After each rat completed 3 consecutive sessions (20 min; FR1) with >100 active lever presses, the rats were transitioned to 60-min daily sessions under a progressive ratio (PR) schedule of reinforcement: after each food reward, the number of active lever presses required to receive the next reward was increased to a value of [(5 * e^0.2X) − 5], where X represents the number of prior rewards (e.g., 1, 2, 4, 6, 9, …). A ‘break-point’ (BP) was operationally defined as the number of rewards received prior to a 20-min period in which no further rewards were earned. PR sessions continued until each animal was recorded having 3 consecutive sessions where BPs did not differ by more than 2.

To test the effects of NGB 2904 and SKF 77434 on lever pressing for food, rats were randomly assigned to receive an IP injection of one of two combination doses utilized in Experiment 1 (per Fig. 2D, treatment ‘A+C’, n = 8; treatment ‘B+D’, n = 8) or a combination of the respective vehicles (n = 8) 30 min prior to the test session. Test sessions were 30 min (modeling the window of lever accessibility during the cue-induced relapse tests in Experiment 1) but were otherwise identical to prior PR sessions. Activity on both levers was recorded.

Figure 2.

Mean lever presses (±SEM) during the cue-induced relapse test session. A) results for arm 1; labels ‘V1’, ‘A’, and ‘B’ were assigned to the vehicle, 0.25 mg/kg, and 1 mg/kg of NGB 2904, respectively; B) results for arm 2; labels ‘V2’, ‘C’, and ‘D’ were assigned to the vehicle, 0.25 mg/kg, and 0.5 mg/kg of SKF 77434, respectively; C) results for arm 3; labels indicate which doses from arms 1 and 2 were combined to treat each dose group (e.g., ‘V1+V2’ received the vehicles for both compounds); D) comparison of all relevant dose groups to a control group comprised of ‘V1’, ‘V2’, and ‘V1+V2’ group data combined. * p < .05; ** p < .01; markers indicate significant differences when compared to controls.

2.7. Experiment 3: Pro/Anti-Nociceptive Effects - Tail-Flick Latency Assays

Potential pain modulation by NGB 2904 and SKF 77434 was tested in naïve rats (N = 38) by comparing latency of heat-induced tail-flick responses. For a detailed description of assay procedures, see our previous work.⁴⁰ As a pilot study, 8 rats were utilized to assess nociception following administration of the highest treatment doses that were used in Experiment 1. One assay was conducted before treatment to serve as a baseline, then rats were randomly assigned to Treatment or Control groups. The Treatment group (n = 4) received IP injections of NGB 2904 (2 mg/kg) and SKF 77434 (2 mg/kg) and Controls (n = 4) received the respective vehicles for each compound. Post-treatment assays occurred at 0 min (30 min after treatment injections), 15 min, and 30 min.

The remaining animals (n = 30) were assigned to test for potential interactions between these compounds and heroin in modulating nociceptive response. All rats underwent a baseline assay before random assignment to groups. The animals were then pre-treated with either a combination of both compounds (NGB+SKF; per Fig. 2D, treatment ‘B+D’) or a combination of the respective vehicles (Veh). After 30 min, all rats were treated with either heroin (0.5 mg/kg, IP) or saline. Assignment of pre-treatments and treatments resulted in 3 unique groups (n = 10 each): Veh-Saline, Veh-Heroin, and NGB+SKF-Heroin. Subsequent tail-flick assays were conducted 15 min, 30 min, 60 min and 90 min after treatment.

2.8. Data Analysis

Behavioral data (e.g., lever pressing, tail-flick latencies) and metrics used to determine pre-test group homogeneity (e.g., electrical current strength) were predominantly analyzed with one-way or mixed factorial ANOVAs or with Student’s t-tests; Groups/Doses were treated as between-subjects measures and Lever (active vs. inactive) as a within-subjects variable. Data from the cue-induced relapse tests were log-transformed prior to analysis to reduce skewness. Parametric tests of simple effects and/or pairwise comparisons were conducted when ANOVAs resulted in significant (p < .05) main effects or interactions. Bonferroni corrections were applied to pairwise comparisons. To detect evidence of drug synergism or antagonism, a series of ‘response additivity’ analyses were conducted on the effects of each dose combination vs. their respective independent compound doses on active lever pressing (for commentary on test selection, see Discussion). These analyses are conducted using a 2×2 factorial ANOVA (Drug A: Yes/No x Drug B: Yes/No); the significance of the Drug A x Drug B interaction is used to evaluate the null hypothesis that the combined effect of two drugs is merely additive.⁴⁵ All analyses were performed using IBM SPSS Statistics 26.

3. Results

3.1. Experiment 1: Cue-Induced Relapse to Heroin-Seeking

3.1.1. Pilot Study to Confirm Test Parameters

Prior to the experimental manipulation (no electric barrier vs. a proportionally defined barrier on test day) the two groups did not differ in the number of days spent in abstinence nor in average electric barrier strength at completion of the EBA phase (data not shown). During the cue-induced relapse test session, both groups pressed more on the active lever than the inactive lever, as shown in Fig. 1. The rats that had no electric barrier pressed more on the active lever (M = 98.67; 95% CI = 79.01, 118.32) than those with a barrier set to 25% of the strength at which the EBA phase was completed (M = 34.33; CI = 14.68, 53.99) but the groups did not differ in inactive lever presses. A mixed factorial ANOVA indicated a significant main effect of Lever, F(1,4) = 805.476, p < .001, and a significant Lever x Group interaction, F(1,4) = 2670.08, p < .001. Analysis of simple effects revealed a significant difference in active (p = .003) but not inactive lever pressing (p = .645).

Figure 1.

Results for the pilot study in Experiment 1: mean lever pressing (±SEM) during the cue-induced relapse test session. ** p < .01.

3.1.2. The Effects of NGB 2904 and SKF 77434 on Cue-Induced Relapse

Analyses of pre-test data found the groups in each arm to be equivalent in lever presses and infusions received during the training phase, in the number of days spent in the EBA phase, and in average electric barrier strength when ‘abstinence’ criteria were met (data not shown). Results from cue-induced relapse test sessions are presented in Fig. 2. In all arms, rats pressed more on the active lever than the inactive lever. When NGB 2904 or SKF 77434 were administered independently, there was a dose-dependent reduction in presses on both levers, with large reductions observed only in the higher dose groups. In both of these arms, the two lowest dose groups did not appear to differ in lever pressing when compared to their controls (see Fig. 2A and B). ANOVA results (Table 1) reflected significant main effects of Lever and Group (all ps < .05), but no significant Lever x Group interactions (both ps > .088).

Table 1.

Descriptive statistics (untransformed)					ANOVA results (log-transformed)
Arm	Dose mg/kg	active lever M (95% CI)	inactive lever M (95% CI)		Arm	Source	SS	df	MS	F	p
NGB 2904 only	Veh (V1)	39.11 (29.96, 48.26)	16.11 (5.08, 27.14)		NGB 2904 only	LEVER	1.86	1	1.85	23.59	<.001
	0.25 (A)	36.33 (7.35, 65.32)	15.56 (2.89, 28.22)			LEVER*GROUP	0.69	4	0.17	2.18	0.88
	1.0 (B)	28.22 (4.43, 52.01)	14 (2.4, 25.6)			Error	3.15	40	.08
	1.5	17 (−3.49, 37.49)	13 (−3.62, 29.62)			GROUP	8.96	4	2.24	2.70	.044
	2.0	10.33 (−1.12, 21.79)	2.89 (−1.83, 7.6)			Error	33.16	40	0.83
SKF 77434 only	Veh (V2)	47.75 (18.28, 77.22)	24.38 (2.34, 46.41)		SKF 77434 only	LEVER	1.33	1	1.33	22.32	<.001
	0.25 (C)	47.25 (7.79, 86.71)	16.5 (0.69, 32.31)			LEVER*GROUP	0.49	4	0.12	2.07	.106
	0.5 (D)	29.38 (8.08, 50.67)	17.75 (0.48, 35.02)			Error	2.08	35	0.06
	1.0	20.63 (−3.2, 44.45)	8.50 (−1.81, 18.81)			GROUP	12.31	4	3.08	3.15	.026
	2.0	0.88 (−1.19, 2.94)	0.38 (−0.51, 1.26)			Error	34.20	35	0.98
Combination Treatment	V1+V2	30.13 (15.67, 44.58)	17.38 (4.34, 30.41)		Combination Treatment	LEVER	0.34	1	0.34	9.69	.004
	A+C	11.75 (−0.26, 23.76)	12.25 (−5.96, 30.26)			LEVER*GROUP	0.58	4	0.15	4.09	.008
	A+D	9.56 (−0.49, 19.6)	6.78 (0.48, 13.08)			Error	1.31	37	.04
	B+C	12.44 (0.96, 23.93)	7.11 (−0.25, 14.47)			GROUP	6.42	4	1.60	2.18	.090
	B+D	3.25 (−1.12, 7.62)	7.13 (−4.84, 19.09)			Error	27.19	37	0.74
All	Vehicles Groups Combined	39.0 (29.33, 48.67)	19.16 (11.37, 26.95)		All (Vehicles Only)	LEVER	3.02	1	3.02	24.12	<.001
						LEVER*GROUP	0.05	2	0.02	0.18	.834
						Error	2.75	22	.13
						GROUP	0.12	2	0.06	0.13	.883
						Error	10.14	22	0.46

The two lower doses from the NGB 2904-only and SKF 77434-only arms were selected to be administered in unique combinations in the third arm. As shown in Fig. 2C, when the low doses of these compounds were administered in combination, all four treatment groups pressed less on the active lever than the control group, but inactive lever presses were similar across all groups. A mixed factorial ANOVA (Table 1) indicated a significant Group x Lever interaction (p = 0.008) but no main effect of Group (p = 0.090). Tests for simple effects of group revealed a significant group effect on active lever pressing, F(4,37) = 3.60, p = .014, but not inactive lever pressing (p = .391). Results from pairwise group comparisons indicated significantly lower active lever presses in all groups receiving combination treatments when compared to the control group (all ps < .021).

To facilitate a direct comparison of the individual compounds’ effects to those of the combination treatments, data from the relevant groups in all three arms of Experiment 1 were isolated for analysis. No significant differences were observed in lever pressing across the groups that received only vehicles in each arm (labeled ‘V1’, ‘V2’, and ‘V1+V2’ in Fig. 2; see Table 1 for ANOVA results). Therefore, data from these three vehicle groups were combined (n = 25) to serve as the control in subsequent analyses. As shown in Fig. 2D, lever pressing in the 4 groups that received single-drug treatments resembled that of the control rats. However, all four combination treatment groups appeared to respond less on the active lever than control rats, with relatively little reduction in inactive lever pressing. A mixed factorial ANOVA analyzing lever pressing across all groups in Fig. 2D found a significant main effect of Lever (p < .001) and of Group (p = .019) and a significant Lever x Group interaction, F(8,84) = 3.96, p = .001. Tests for simple effects of group at each lever revealed a significant effect at the active lever, F(8,84) = 3.44, p = .002, but not at the inactive lever, F(8,84) = 0.94, p = .490. Pairwise comparisons between groups on active lever responding indicated that each combination treatment group pressed significantly less than controls (all ps < .014).

Finally, the effects of combination treatments on active lever pressing were compared to that of independent doses to determine if the drug interaction was antagonistic, additive, or synergistic. Visual appraisal of group means suggested that the reduction in active lever pressing after combination treatments was greater than the sum of reductions brought on by individual compounds (i.e., a synergistic effect; for an example of one such comparison, see Fig. 3). To find statistical support for this observation, a series of five response additivity analyses were performed as described above: one for each unique combination treatment and one comparing all 4 combinations simultaneously (for a summary of factors and group membership, see Fig. 4). The interaction results for each unique dose combination failed to reach significance (all ps > .091), indicating additive effects of NGB 2904 and SKF 77434. However, when treatment data were collapsed into 4 aggregated treatment groups (factors: NGB [Yes vs. No] x SKF [Yes vs. No]), a significant NGB 2904 x SKF 77434 interaction was found, F(1,89) = 3.97, p = .049, indicating significantly lower active lever presses in combination-treated rats than would be expected if the combined effect of these compounds was merely additive.

Figure 3.

Sample depiction of the potential interactive effects of NGB 2904 and SKF 77434 using the highest dose combination treatment group and its associated independent dose groups in Experiment 1. Bar height represents the difference in mean active lever presses for each group when compared to that of all vehicle groups combined (M_Vehicles – M_{Dose Group}; ±SEM). The grey bar represents hypothetical data; its height shows the sum of the reductions induced by single-compound treatments, representing the null hypothesis that the combination has only an additive (not antagonistic or synergistic) effect on drug-seeking. The observed effect was n.s. (p = .309).

Figure 4.

Each graph depicts mean active lever presses after treatment with vehicle(s) only, NGB 2904 only, SKF 77434 only, or a combination of both compounds. The first four graphs compare lever presses from each combination treatment group (arm 3) to its associated independent dose groups (arms 1 and 2) and to the combined vehicle group. Dose labels are defined in the caption for Fig. 2. The final graph (right) shows aggregated data: all rats that received some dose of NGB 2904 only, all rats that received SKF 77434 only, and all rats that received a combination treatment (vs. controls). Although all graphs appear to suggest a synergistic effect of these compounds on relapse-related lever pressing, response additivity analyses only provided statistical evidence of an NBG x SKF interaction when group data were aggregated. This may indicate that there was insufficient statistical power to detect a synergistic effect when each combination treatment was analyzed independently. * p = .049.

3.2. Experiment 2: Food Self-Administration Under a PR Schedule

Analyses involving pre-test data indicated that the groups were equivalent in the number of days spent in each training phase and in average BPs reached prior to receiving treatment (data not shown). Lever pressing activity from the test session is presented in Fig. 5; to facilitate comparison of motoric behavior, PR/Food data are presented side-by-side with relapse test data from the rats that received vehicles in Experiment 1 (IVSA/Heroin). In the PR/Food experiment alone, animals responded much more on the active lever than the inactive lever, and all groups appeared to respond at similar levels on each lever (Fig. 5). A mixed 2×3 ANOVA supported this observation by revealing a main effect of Lever (p < .001) but no Group effect (p = .584) nor Lever x Group interaction (p = .725). When compared with the relapse-related lever pressing of the IVSA/Heroin control rats, all groups in the PR/Food study produced many more presses on the active lever, but inactive lever presses appeared to be equivalent across all groups. A 2×4 mixed factorial ANOVA indicated significant main effects of Lever and Group (both ps < .001) and a significant Lever x Group interaction, F(3,45) = 17.95, p < .001. Tests of simple effects revealed a significant difference in active lever presses, F(3,45) = 18.35, p < .001, but not in inactive lever presses (p = .336). Post-hoc analyses indicated that all groups in the PR/Food study pressed significantly more on the active lever than the IVSA/Heroin control group (all ps < .001) but did not differ from each other (all ps > .216).

Figure 5.

Mean lever presses (±SEM) from the PR/Food test session in Experiment 2 is compared to that of control rats in the cue-induced relapse test session in Experiment 1. Group labels for the PR/Food data are defined in the caption for Fig. 2 and indicate the combinations of vehicles or of NGB 2904 and SKF 77434 received prior to the test session. *** p < .001; markers indicate significant differences in active lever presses when compared to controls from Experiment 1.

3.3. Experiment 3: Pro/Anti-Nociceptive Effects – Tail-Flick Latency Assays

3.3.1. Pilot study: The Effects of NGB 2904 and SKF 77434 on Nociception

Mean tail-flick latencies at each time point are presented in Fig. 6. An independent-measures t-test was performed on baseline data (prior to treatment) to verify homogeneity in group assignments; no significant differences were found across groups, t(6) = 0.26, p = 0.807. As shown in Figure 6, although all groups were observed to react slightly more quickly on average over time, tail-flick latencies appeared similar between groups across time points. A 2×4 mixed factorial ANOVA failed to find significant effects of Group (p = .310), Time (p = .083), or a Group x Time interaction (p = .515).

Figure 6.

Mean tail-flick latencies (±SEM) across time during the pilot study in Experiment 3. Group data shown at 0-min represent recordings taken 30 minutes after receiving treatment with a combination of NGB 2904 (2mg/kg) and SKF 77434 (2mg/kg; n = 4) or of the respective vehicles for each compound (n = 4). Analyses of Group and Group x Time effects were n.s.

3.3.2. Combination Treatment + Heroin Interaction

Mean tail-flick latencies for the three groups across all time points are presented in Fig. 7. A one-way ANOVA was performed on baseline data (prior to receiving injections) to verify homogeneity in group assignments; no significant differences were found across groups, F(2,27) = 0.44, p = .646. As shown in Fig. 7, the Veh-Heroin and NGB+SKF-Heroin groups exhibited similar increases in average tail-flick latencies when compared to the Veh-Saline group 15 min after receiving treatment. Across time, both groups appeared to return to baseline latencies at a similar pace. A mixed factorial ANOVA revealed significant main effects of Time and Group (both ps < .001) and a significant Group x Time interaction, F(8,108) = 10.90, p < .001. Tests for simple effects of group at each time point revealed a significant group effect at 15 min (F[2,27] = 17.44, p < .001), 30 min (p = .001), and 60 min (p = .049). Post-hoc analyses indicated that the Heroin groups had significantly longer tail-flick latencies than the control group at 15 min (both ps < .001) and 30 min (both ps < .012). At 60 min, the NGB+SKF-Heroin group (but not the Veh-Heroin group) had longer tail-flick latencies than the Veh-Saline group (p = 0.016). The Veh-Heroin and NGB+SKF-Heroin groups did not significantly differ at any time point (all ps > .119).

Figure 7.

Experiment 3: nociception assessment to identify a potential interaction between the combination treatment (per Fig. 2, dose ‘B+D’) and heroin. Figure shows mean tail-flick latencies (±SEM) across time. Group data shown at 15-min represent recordings taken 15 minutes after treatment with heroin or saline (45 minutes after pre-treatment with NGB 2904+SKF 77434 or vehicles). * p < .05; *** p < .001; markers indicate significant differences when compared to controls.

4. Discussion

This series of experiments aimed to demonstrate that the co-administration of a D1R partial agonist (SKF 77434) and a D3R antagonist (NGB 2904) is more effective in reducing cue-induced relapse to heroin seeking in rats than independent treatment with either compound. When administered alone, both agents produced a dose-dependent reduction in active lever presses following cue presentation; however, similar reductions were observed on the inactive lever at higher doses, possibly indicating non-specific, motoric side effects. When low (ineffective) doses of each compound were administered simultaneously, all four unique combinations reduced active lever pressing with no significant effect on inactive lever presses.

Furthermore, exploratory analyses detected some evidence that the combined effect of NGB 2904 and SKF 77434 was greater than the sum of independent treatment effects (i.e., drug synergism). The results of response additivity tests⁴⁵ failed to reach statistical significance when evaluating each combination treatment separately, but this may have been due to insufficient statistical power (Fig. 3). The analysis of aggregated data (i.e., active lever presses for rats that receive some dose of NGB 2904, SKF 77434, or some combination of both) suggested a significant, synergistic interaction of these compounds in reducing drug-seeking behavior. However, it is important to note that these results are interpreted conservatively and are merely presented as evidence that further investigation is warranted. There are limitations to the interpretation of significant results from response additivity analyses⁴⁶, and this study lacks the necessary data (e.g., dose-response curves for each compound; a wider range of dose combinations) for more robust methods of exploring drug synergism, such as isobologram analyses and other dose-effect-based strategies.^46–48

Follow-up studies were performed to provide further evidence that the attenuation in cue-induced opiate seeking in combination-treated rats was not due to systemic effects that have been associated with DA receptor-modulating agents: specifically, motoric impairment and modulation of pain sensitivity.^15,49,50 In a food self-administration paradigm, animals that received poly-drug treatments were able to produce more than 7x (on average) the active lever presses as vehicle-treated rats in the heroin reinstatement experiment (Fig. 5); this indicates that combination-treated animals are as motorically capable of pressing a lever as non-treated rats, supporting the notion that the observed reduction in cue-induced drug seeking reflected reduced motivation for the drug. Interestingly, neither dose combination appeared to have a significant effect on food seeking, which may suggest that this treatment has minimal impact on motivation for natural rewards (discussed further below). Furthermore, we were able to demonstrate via tail-flick analgesia recordings that the compounds had no effect on nociception, even with high-dose combinations (Fig. 6); this indicates that these treatments did not prevent drug-seeking by enhancing sensitivity to the electrified flooring in front of the levers. While the mechanisms of heat-induced nociception may differ from that of electric foot-shock, several studies have found the two to be functionally similar measures of analgesia in rodents.⁵¹

This poly-pharmacological approach was first explored by our lab in the context of cocaine-seeking behaviors, finding that the treatment was effective in reducing cue-induced reinstatement after seeking behaviors had been extinguished.³⁵ There is substantial evidence that D1R and D3R play a crucial role in opiate-seeking behaviors (see Introduction), but to our knowledge, neither SKF 77434, NGB 2904, nor combinations of DA receptor agents have been investigated in the context of cue-induced opiate seeking prior to this study. Importantly, the present work was evaluated within a modified version of the animal ‘conflict’ model of abstinence/relapse,^39,40 which is capable of facilitating ‘true’ (self-initiated) abstinence despite continued availability of the drug. The ‘conflict’ model is believed to possess greater face validity in emulating human experience than extinction/reinstatement procedures,^39,40,42 thus results obtained with these methods may translate more readily to clinical trials.

We emphasize that cue-induced relapse was attenuated in this study using combined but relatively low doses of NGB 2904 and SKF 77434, with no significant impact on food seeking nor overt evidence of motoric or analgesic side effects. Previous studies have shown that food seeking behaviors can be dampened by D1R antagonists at sufficient doses,^52,53 and although food seeking appears to be unaffected by antagonism at D3R,^21,54 this raises a concern for the impact of a D1R/D3R ligand combination on motivation for natural rewards. However, food seeking was not significantly impacted in the present study, possibly due to the use of a D1R partial agonist (rather than an antagonist) or that it could be administered at low doses when combined with a D3R antagonist. With regards to potential motoric side effects, in previous studies, SKF 77434 and other D1R partial agonists have exhibited fairly narrow ‘dose windows’ for efficacy;^30,55 careful dosing can elicit desirable effects on drug seeking, but surpassing this window can lead to locomotor inhibition, thigmotaxis, or catatonia.¹⁵ This may be due to the relatively wide expression of D1R in non-reward-related structures¹⁷ or because some D1R agents, including SKF 77434, exhibit higher rates of binding at DA D2 receptors (D2R) when administered at high doses.^15,16 Incidental blocking at D2R would likely preclude a treatment from advancing in clinical trials, given the substantial evidence that D2R antagonists (e.g., antipsychotics) can cause pronounced dysphoria and motoric side effects in humans and contribute to drug relapse.^31,32,56 Similarly, NGB 2904 and other D3R antagonists have exhibited limited dose ranges for desirable outcomes, which has been attributed to dose-dependent differences in occupancy at pre- versus post-synaptic D3R.³⁰ It is worth noting that some single-compounds treatments that show affinity for receptors in both D1-like and D2-like receptor families, such as L-stepholidine and haloperidol, have shown some benefit in reducing drug-seeking behaviors;^9,44 however, these compounds offer little opportunity to ‘fine-tune’ activity at specific receptors by adjusting doses, thus increasing the risk of side effects. It is possible that the poly-drug approach used in this study would enhance the safety of DA-modulating treatments by providing a wider range of low but effective dose combinations.

Continuous efforts have been applied over the last 3–4 decades to developing compounds with greater selectivity and specificity for D1R or D3R to provide greater and more targeted control over their effects.^22,36 These efforts have been encouraged by NIDA, who listed D3R specifically as one of the “highest priority pharmacological targets for the development of novel therapeutics” for opiate addiction.²⁵ While considerable advances have been made, the results of this study provide evidence that a poly-pharmacological treatment using D1R and D3R ligands may be safer and more effective than single-drug approaches. It may be important to evaluate this approach using newly developed compounds and a wider range of dose combinations.

Further research will be needed to expand upon these results and to address potential limitations to their generalizability. For example, the present study did not investigate potential rewarding effects of the treatment itself; as seen in studies of DA receptor agonists, it is possible that these results may represent a substitution effect: inhibiting drug seeking by providing a ‘replacement’ for heroin reward.^57,58 Also, several of the studies cited herein suggest that pharmacological treatment efficacy may differ in female rodents or across species. Furthermore, at present it is unknown if abstinence that has been facilitated by avoidance of electrical shocks has unique neurophysiological implications. It may also be important to understand how these chemicals interact outside of dopaminergic synapses, as pointed out by Galaj et al.³⁵

A more promising opportunity to expand upon these results may be to investigate potential poly-drug treatment effects on the acquisition of opiate-seeking behaviors. It is well known that opiate dependence can be initiated by misuse of opiate medications prescribed for pain management,⁵⁹ but recently there has been evidence that D3R antagonists may dampen the reinforcing effects of prescription opiates while simultaneously enhancing⁶⁰ or having no impact²² on their analgesic properties. In Experiment 3 of the present study, we were able to expand on this research by demonstrating that co-administration of NGB 2904 and SKF 77434 had no significant effect on the analgesic effects of heroin (Fig. 7). Further research is needed to determine what factors contribute to enhancement of opiate analgesia vs. null effects (e.g., the specific ligands selected; some characteristic of the experimental model utilized), but this study lends further support for DA receptor agents as a potential supplement when treating pain with opiate medications.

Lastly, it will be important to explore potential interactions of this treatment when administered under the influence of other substances. For example, DA analogs have been shown to induce acute hypertension or other aversive side effects when administered with psychostimulants,^33,34,55 which may pose a risk to poly-substance users. Also, it is currently unknown if DA receptor modulation could impact the outcomes of established medication-assisted therapies (e.g., methadone and buprenorphine/naloxone), which are currently considered the gold standard in opiate use disorder treatment. Ideally, the co-administration of the present treatment could help to lower the extremely high relapse rates exhibited by these pharmacotherapies^2,3,61 provided adverse interactions can first be ruled out. It is possible that using relatively low doses of these pharmacological agents will reduce the risk of chemical interactions, providing relief from symptoms of opiate addiction with minimal risk of side effects.