Neurobiology of Relapse to Heroin and Cocaine Seeking: A Review

1. Cocaine Priming.

The effect of cocaine priming on reinstatement is mimicked by systemic injections of amphetamine (a DA reuptake blocker and a DA releaser), DA reuptake blockers (GBR-12909, methylphenidate) and D2-like receptor agonists (7-hydroxy-2-dipropylaminotetralin, quinpirole, bromocriptine) (de Wit and Stewart, 1981; Wise et al., 1990; Self et al., 1996; De Vries et al., 1999; Schenk and Partridge, 1999). On the other hand, mixed DA agonists (apomorphine) or direct D1-like agonists (SKF 82958, SKF 81297, ABT 431) do not mimic the effect of cocaine priming on reinstatement (Self et al., 1996, 2000; De Vries et al., 1999) (Figs.3 and 4). Surprisingly, these direct D1-like agonists, some of which are self-administered by rats and monkeys (Self and Stein, 1992; Weed et al., 1993; but see Caine et al., 1999b), also block cocaine-induced reinstatement in rats (Self et al., 1996, 2000) and monkeys (Khroyan et al., 2000). Norman et al. (1999) reported that the D1-like antagonist, SCH 23390, attenuates cocaine-induced reinstatement. In addition, using the runway model in amphetamine-trained rats, Ettenberg (1990) reported that the D2-like receptor antagonist, haloperidol, attenuates drug-induced reinstatement. These pharmacological data indicate that DA plays a crucial role in cocaine-induced reinstatement. In addition, although activation of D2-like receptors provokes cocaine seeking, activation of D1-like receptors inhibits it. The reasons for this pharmacological dissociation are not clear in light of the literature on the similar behavioral effects of the D1- and D2-like agonists on locomotor activity (Self et al., 1996) and cocaine reinforcement (Self and Nestler, 1995).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Reinstatement following long-term extinction (3 weeks) of heroin and cocaine self-administration: mean (± S.E.M.) number of nose poke responses in the previously active (drug-paired) and inactive hole during the 2-h test for reinstatement. A, priming effects in rats previously trained to self-administer heroin; B, priming effects in rats previously trained to self-administer cocaine. Saline, cocaine (10 mg/kg, i.p.), amphetamine (1.0 mg/kg, i.p.), or heroin (0.25 mg/kg, s.c.) were injected 10 min before the start of the reinstatement session. ∗, significantly different from preceding saline injection (p < 0.05). Data are from De Vries et al., 1998, reprinted ©1998 with permission from Blackwell Science Ltd.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

Effects of intraperitoneal (i.p.) injections with vehicle (saline), the D2-like dopamine agonist 7-hydroxy-2-dipropylaminotetralin, or the D1-like agonist SKF 82958 on reinstatement of nonreinforced lever-press responding. Priming injections were given after extinction from 2 h of intravenous cocaine self-administration when only intravenous saline injections were available. A, lever-press responding in a representative animal. Hatch marks denotes the times of each self-infusion of cocaine in the cocaine phase and of saline in the saline phase. B and C, mean number (± S.E.M.) of nonreinforced lever-press responses during the final hour of the saline phase in the drug-reinstatement paradigm. Asterisks indicate significant differences from saline treatment (∗,p < 0.05; ∗∗, p < 0.01). Data are from Self et al., 1996. Copyright 1996 with permission from Science (Wash DC).

Studies using intracranial drug injections also provide evidence for the role of the mesocorticolimbic DA system in cocaine priming. Stewart (1984) found that intra-VTA infusions of morphine reinstate cocaine seeking (Fig. 5). A likely mechanism for this effect is that morphine inhibits GABAergic neurons in the VTA, which provide tonic inhibition of DA neurons (Di Chiara and North, 1992). More recently, Self et al. (1998) studied the effect of manipulations of the intracellular signaling of the D1- and D2-like receptors by using an activator (S_P-cAMPS) and an inhibitor (R_P-cAMPS) of protein kinase A (PKA). These agents mimic activation of the D1- and D2-like receptors by receptor agonists, respectively (Kebabian and Calne, 1979). They found that intra-NAc infusions of R_P-cAMPS reinstate cocaine seeking, whereas infusions of S_P-cAMPS nonselectively increase responding on both levers. The data with the PKA inhibitor are in agreement with those obtained with systemic injections of direct D2-like agonists. However, it cannot be ruled out that the effect of R_P-cAMPS on reinstatement is mediated via its action on non-DA receptors in the NAc that also inhibit cAMP formation (e.g., μ-opioid receptors) (Taylor and Fleming, 2001). The mechanisms underlying the effect of the PKA activator on the lever-pressing behavior and their relationship to DAergic activity are even less clear, as this effect is different from that observed with D1-like agonists. Cornish and Kalivas (1999) found that activation of AMPA receptors in the NAc reinstate cocaine seeking, and PKA modulates AMPA receptor activity (Banke et al., 2000). Thus, the effect of the PKA activator on cocaine seeking may be due to its action on the AMPA receptor. More recently Cornish and Kalivas (2000)reported that direct infusions of DA into the NAc reinstate cocaine seeking. Surprisingly, these authors found that intra-NAc infusions of the nonselective DA antagonist, fluphenazine, had no effect on cocaine-induced reinstatement. These data, together with the data described above suggest that DA neurotransmission in components of the mesocorticolimbic DA system other than the NAc is involved in cocaine-induced reinstatement. In agreement with this idea, McFarland and Kalivas (2001) recently found that infusions of fluphenazine into the medial prefrontal cortex (mPFC) attenuate cocaine-induced reinstatement.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5

Responses (mean ± S.E.M.) on the previously active (cocaine- or heroin-paired) lever following priming injections of morphine (Mor) into the ventral tegmental area (cell body region of the mesolimbic dopaminergic neurons), in rats previously trained to self-administer cocaine (left panel) or heroin (right panel). Rats were tested with (Nal-Mor) or without (Mor) pretreatment with the opioid receptor antagonist naltrexone (Nal), administered i.p. Data are fromStewart, 1984, reprinted ©1984 with permission from Elsevier Science.

Other evidence for the role of DA in cocaine-induced reinstatement comes from studies using in vivo microdialysis and electrophysiology, and immediate early gene expression (IEG) methods. Using in vivo microdialysis, it was reported that priming injections of cocaine increase extracellular dopamine levels in the NAc (Neisewander et al., 1996) and the amygdala (Tran-Nguyen et al., 1998), a terminal projection of the mesocorticolimbic DA system (Fallon and Ciofi, 1992).Di Ciano et al. (2001) reported that in rats trained to self-administerd-amphetamine, priming injections of the drug increased DA signal in the NAc, as measured by chronoamperometry. Finally,Neisewander et al. (2000) reported that priming cocaine injections increase Fos protein expression, a cellular marker of neuronal activity (Morgan and Curran, 1991), in the VTA and several terminal DA projections [the caudate putamen, central nucleus of the amygdala (CeA), lateral amygdala and the anterior cingulate cortex].

Other indirect evidence for the role of DA in reinstatement comes from studies demonstrating that caffeine (an antagonist of adenosine receptors) reliably reinstates cocaine seeking (Worley et al., 1994). This effect may be due to an interaction between adenosine A2a and D2-like receptors, which are negatively coupled (Fuxe et al., 1998). Thus, blockade of adenosine A2a receptors by caffeine can lead to activation of D2-like receptors and consequently to reinstatement of cocaine seeking. It has been shown that D2-like receptor antagonists can attenuate the effects of adenosine receptor antagonists (Garrett and Holtzman, 1995).

Converging evidence implicates the mesocorticolimbic DA system in cocaine-induced reinstatement. Surprisingly, D1- and D2-like receptors play fundamentally different roles in this effect. Finally, the recent data of Cornish and Kalivas (2000), together with the previous pharmacological data, suggest that the action of DA in regions of the mesocorticolimbic DA system (e.g., the prefrontal cortex), other than the NAc, mediates cocaine-induced reinstatement. Nevertheless, DA in the NAc still plays a role as intra-NAc infusions of DA reinstate cocaine seeking (Cornish and Kalivas, 2000).

2. Heroin Priming.

Indirect DA agonists (cocaine, amphetamine) were found to reinstate heroin seeking after prolonged withdrawal periods (3 weeks) (De Vries et al., 1998a, 1999) (Fig. 3). Intra-NAc infusions of amphetamine also reinstate heroin seeking in the within-session reinstatement procedure (Stewart and Vezina, 1988) (Fig.5). In contrast, using the within-session procedure, it was reported that cocaine does not reliably reinstate heroin seeking (de Wit and Stewart, 1983). The discrepant results between the effect of cocaine in the between- and within-reinstatement procedures may be related to the development of sensitization to the DA-dependent behavioral activating effects of cocaine following prolonged, but not short, withdrawal periods (De Vries et al., 1999; Vanderschuren and Kalivas, 2000).

Using the within-subjects method, the direct D2-like agonist, bromocriptine, was found to reinstate heroin seeking (Wise et al., 1990). Interestingly, De Vries et al. (1999, 2002) found that the D2-like agonist, quinpirole, reinstates heroin seeking following short withdrawal periods (within 1 week) but not following 3 weeks of withdrawal. The nonselective DA agonist, apomorphine, and the D1-like receptor agonist, SKF 82958, do not reinstate heroin seeking (de Wit and Stewart, 1983; De Vries et al., 1999). In other studies it was found that D2-like receptor antagonists block heroin-induced reinstatement (Ettenberg et al., 1996; Shaham and Stewart, 1996; Lu et al., 2001b). Blockade of the D1-like receptors by SCH 23390 also attenuated heroin-induced reinstatement (Shaham and Stewart, 1996). However, nonspecific sedative effects of the relatively high dose of the D1-like antagonist (0.05–0.1 mg/kg, i.p.) may have contributed to this effect.

Together, the available data suggest that DA is involved in heroin-induced reinstatement. The recent data of De Vries and colleagues suggest that activation of D2-like receptors plays an important role in heroin reinstatement during early but not late withdrawal. A role of D1-like receptors in heroin-induced reinstatement, however, has not been established.

1. Cocaine Priming.

The role of opioid receptors in cocaine priming has not been clearly established and the data reviewed below mirror the conflicting literature on the role of opioid receptors in cocaine reinforcement. The preferentially μ-opioid antagonist, naltrexone, has no effect on cocaine-induced reinstatement (Comer et al., 1993). In addition, systemic injections of μ-opioid receptor agonists (heroin, morphine, etonitazene), a mixed μ-agonist/κ-antagonist (buprenorphine) or a mixed μ/κ-agonist (butorphanol) do not reliably reinstate cocaine seeking (de Wit and Stewart, 1981; Comer et al., 1993; De Vries et al., 1998a; Lynch et al., 1998). Furthermore, butorphanol, etonitazene, and buprenorphine (Comer et al., 1993; Lynch et al., 1998), but not morphine (Lynch et al., 1998), attenuate cocaine-induced reinstatement. However, the interpretation of these data is complicated because in opioid-naive rats, systemic injections of opioid agonists have a biphasic effect on behavior: an initial sedative effect that is followed by behavioral activation (Babbini et al., 1975). These initial sedative effects may have masked the expression of the motivational effects of the opioid receptor agonists on reinstatement. Thus, when morphine is infused acutely into the VTA, where it induces behavioral activation (Joyce and Iversen, 1979), it reinstates cocaine seeking (Stewart, 1984). More recently, Schenk et al. (1999, 2000) reported that the κ-opioid receptor agonist, U69593, decreases cocaine-induced reinstatement. This effect may be related to the inhibitory effect of κ-opioid receptor activation on DA release (Spanagel et al., 1990; Shippenberg and Elmer, 1998).

Despite the fact that opioid agonists and mixed agonists have been shown to attenuate cocaine priming, the potential sedative effects of these compounds in opioid-naive rats, together with the data on the lack of effect of naltrexone on cocaine-induced reinstatement, suggest that activation of μ-opioid receptors is not involved in cocaine reinstatement. However, it appears that alterations of DA neurotransmission by opioid agents can, under certain conditions (e.g., direct intra-VTA infusions of morphine), lead to cocaine seeking.

2. Heroin Priming.

The effect of heroin priming on reinstatement is dependent on activation of μ-opioid receptors. Priming injections of μ-opioid receptor agonists such as morphine mimic the effect of heroin on reinstatement when given systemically (de Wit and Stewart, 1983; Stewart and Wise, 1992) or intra-VTA but not intra-NAc (Stewart, 1984) (Fig. 5). Naltrexone, a preferentially μ-opioid receptor agonist, blocks reinstatement induced by either systemic injections of heroin (Shaham and Stewart, 1996) or intra-VTA morphine (Stewart, 1984). Finally, chronic occupation of the opioid receptors with heroin given via Alzet osmotic minipumps attenuates heroin-induced reinstatement (Shaham et al., 1996). These data suggest that activation of DA neurons in the VTA mediates heroin-induced reinstatement. However, there are recent reports that infusions of opioid and GABAergic agents into the VTA also have DA-independent rewarding effects (Nader and Van der Kooy, 1997; McBride et al., 1999). Thus, it cannot be ruled out that DA-independent mechanisms within the VTA are involved in heroin-induced reinstatement.

1. 5-Hydroxytryptamine.

Manipulations of brain 5-HT systems can alter the behavioral effects of cocaine, including drug self-administration and discrimination (Walsh and Cunningham, 1997). Several studies examined the effect of 5-HT agents on cocaine priming-induced reinstatement. The selective serotonin reuptake inhibitor (SSRI), fluoxetine, which increases 5-HT levels in terminal regions (Perry and Fuller, 1992), has no effect on cocaine-induced reinstatement (Baker et al., 2001). The 5-HT_1aantagonist, WAY 100635, which increases 5-HT cell firing and release (Mongeau et al., 1997), attenuates cocaine-induced reinstatement. In addition, the 5-HT_2C agonist Ro 60-0175 (Grottick et al., 2000), but not the 5-HT₂ antagonist ritanserin (Schenk, 2000), attenuates cocaine-induced reinstatement. Ro 60-0175 may attenuate cocaine-induced reinstatement by decreasing DA levels in the NAc and frontal cortex (Millan et al., 1998; Di Matteo et al., 1999). Finally, Tran-Nguyen et al. (2001) reported that 5-HT lesions by 5,7-dihydroxytryptamine shift the dose-response curve of cocaine priming to the left. These data suggest that 5-HT acts on 5-HT_2C receptors to attenuate cocaine-induced reinstatement. However, the observation that fluoxetine has no effect on cocaine priming is not in agreement with this idea. Thus, the role of 5-HT in cocaine-induced reinstatement remains to be determined.

2. Corticosterone.

The stress hormone, corticosterone, which is released following activation of the hypothalamic-pituitary adrenal (HPA) axis (Selye, 1956), plays an important role in cocaine reinforcement (Piazza and Le Moal, 1998). Cocaine activates the HPA axis (Sarnyai et al., 2001), and inhibition of circulating corticosterone decreases intravenous cocaine self-administration in rats (Goeders, 1997; Piazza and Le Moal, 1997). The data reviewed below, however, suggest that corticosterone secretion does not play a major role in cocaine- (or heroin-) induced reinstatement. The removal of corticosterone by adrenalectomy (ADX) or the administration of CRF receptor antagonists had no effect on cocaine- or heroin-induced reinstatement (Shaham et al., 1997b; Erb et al., 1998). In addition, synthesis inhibitors of corticosterone (ketoconazole or metyrapone) had no effect on cocaine- or heroin-induced reinstatement (Shaham et al., 1997b; Mantsch and Goeders, 1999b). Recently, however, the nonselective CRF receptor antagonist, α-helical CRF, and the selective CRF₁ receptor antagonist, CP-154,526 (Schulz et al., 1996) but not the CRF₂ receptor antagonist antisauvagine-30, were reported to attenuate reactivation of morphine CPP by drug priming after 28 days of withdrawal (Lu et al., 2000). Lu et al. (2001a) also reported that α-helical CRF, but not CP-154,526 or antisauvagine-30, attenuates reactivation of cocaine CPP by cocaine priming. The reasons for the different effects of the CRF receptor antagonists in the CPP model versus the self-administration model are not clear.

3. γ-Aminobutyric Acid.

Roberts and Brebner (2000) found that the GABA_B receptor agonist, baclofen, attenuates cocaine reward. Campbell et al. (1999) reported that baclofen also attenuates cocaine-induced reinstatement (Campbell et al., 1999). This effect may be due to the inhibitory action of baclofen on DA neurons in the VTA (Westerink et al., 1996).

4. Noradrenaline.

Several studies found that manipulations of central NA have an effect on opioid and psychostimulant self-administration behavior (Davis et al., 1975; Harris et al., 1996). However, based on other studies, it is generally believed that the involvement of central NA neurons in drug reinforcement is minimal (Wise, 1978, 1996b). Erb et al. (2000) found that the α-2 adrenoceptor agonists, clonidine and lofexidine, have no effect on cocaine-induced reinstatement at doses that decrease NA release in the amygdala and prefrontal cortex. These data suggest that the action of cocaine on the NA transporter (Blakely et al., 1994) is not involved in its effect on reinstatement. The role of NA in heroin-induced reinstatement has not been determined.

5. Acetylcholine.

Cholinergic neurons modulate mesocorticolimbic DA neurotransmission in the NAc and the VTA (Di Chiara et al., 1994; Sarter et al., 1999). Nicotine increases DA release in the NAc (Damsma et al., 1989), an effect mediated via the activation of nicotine receptors in the VTA (Nisell et al., 1994). The available data, however, do not implicate nicotinic acetylcholine receptors in cocaine-induced reinstatement. Schenk et al. (1999) found some effect of nicotine on reinstatement of cocaine seeking, but Wise et al. (1990) did not.

6. Endocannabinoids.

The endocannabinoid system has been implicated in several neuropsychiatric conditions, including drug addiction (Gardner and Vorel, 1998; Piomelli et al., 2000). The active ingredient of marijuana, Δ9-THC, activates the mesocorticolimbic DA system (Chen et al., 1991; Tanda et al., 1997). In addition, DA through activation of D2-like receptors, releases endogenous cannabinoids in the striatal complex (Giuffrida et al., 1999). Based on these findings, two studies determined the effect of activation or blockade of cannabinoid receptors on cocaine seeking. Using the within-session method, Schenk and Partridge (1999) found that Δ9-THC has no effect on cocaine seeking. In contrast, using the between-session method, De Vries et al. (2001) found that the nonselective cannabinoid agonist, HU210, potently reinstates cocaine seeking following 2 to 3 weeks of withdrawal, whereas the CB1 receptor antagonist SR141716A attenuates cocaine-induced reinstatement. These discrepant results may be due to the time of testing (several hours versus several weeks of withdrawal) and the longer duration of action of HU210 compared with Δ9-THC.

1. Discriminative Drug Cues.

Using a runway model with heroin-trained rats, McFarland and Ettenberg (1995, 1997, 1998) found that the opioid antagonist naloxone or the preferentially D2-like receptor antagonist haloperidol have no effect on heroin seeking provoked by discriminative heroin cues. Naloxone and haloperidol, however, blocked drug seeking on a test day conducted 24 h after last exposure to heroin (Fig. 8). These data suggest that the motivational processes underlying relapse induced by the discriminative cues and the drug itself are dissociable (McFarland and Ettenberg, 1997). Ciccocioppo et al. (2001) and Weiss et al. (2001) reported that SCH 23390 and SCH 39166 (a newer selective D1-like receptor antagonist that binds with low affinity to 5-HT receptors) block reinstatement of cocaine seeking induced by discriminative cues (Fig. 9). Weiss et al. (2001) also reported a similar effect with the D2-like antagonist nafadotride and the D2-like agonist PD 128,907 but not the D1-like agonist SKF 81297. These data should be interpreted with caution because the investigators used a single drug dose of each compound, the effect of the compounds on ongoing extinction behavior (i.e., baseline responding) was not determined, and the same group of rats was tested with the four compounds. The data with PD 128,907 also are different from previous reports on reinstatement of cocaine seeking by D2-like agonists (Self and Nestler, 1998). Finally, in two important studies,Weiss et al. (2000) and Ciccocioppo et al. (2001) reported that cocaine cues increase DA release in both the amygdala and the NAc and that D1-like receptor antagonists decrease cues-induced Fos expression in the BLA and the mPFC (areas Cg1/Cg3).

• Download figure
• Open in new tab
• Download powerpoint

Figure 8

Mean run times (± S.E.M.) in an operant runway for each treatment group during 3 consecutive days: the final day of extinction (Baseline), a single reinstatement trial (Test), and two post-treatment (Post+ and Post−) conditions, both of which occurred on the same day 24 h after the test trial. Panels A, C, and E show mean run times for animals that experienced the heroin-associated olfactory stimulus (S⁺) during the reinstatement test, whereas Panels B, D, and F show the data from animals that experienced the saline-associated olfactory stimulus (S⁻) on this trial. Subjects were pretreated with lactic acid vehicle (A and B), 0.15 mg/kg haloperidol (C and D), or 0.3 mg/kg haloperidol (E and F) prior to behavioral testing on reinstatement (Test) day. ★, significantly different from baseline (p < 0.05). Data are from McFarland and Ettenberg, 1997, reprinted ©1997 with permission from Springer-Verlag.

• Download figure
• Open in new tab
• Download powerpoint

Figure 9

Effects of the D1 antagonist SCH 23390 on the number of responses (mean ± S.E.M.) induced by the discriminative stimulus for cocaine (S⁺). SCH 23390 dose dependently reversed the effects of the S⁺. For comparison, the figure also shows the average number of responses during the last 3 days of the self-administration (SA) and extinction (EXT) phases, as well as responses in the presence of the stimulus associated with nonreward (S⁻). ★, significant linear trend over dose levels (p < 0.05); +, different from the EXT and S⁻ conditions (p < 0.05); &, different from cocaine-reinforced responses (SA; p< 0.01). Data are from Ciccocioppo et al., 2001, reprinted ©2001 with permission from the National Academy of Sciences, U.S.A.

These findings identify an important role of D1-like receptor activation in relapse induced by discriminative drug cues, whereas a role for the D2-like receptors in relapse induced by these cues is less clear. Specifically, D2-like receptor antagonists have different effects on cue-induced reinstatement of heroin seeking in the runway model versus cocaine seeking in the traditional reinstatement model. The microdialysis and the Fos protein data further implicate the BLA, and possibly the mPFC in discriminative cues-induced reinstatement of cocaine seeking.

2. Contextual Drug Cues.

As mentioned above, we found that, in rats trained to self-administer speedball, exposure to the drug self-administration environment, after extinction of the drug-reinforced behavior in a different context, leads to renewal of drug seeking (Crombag and Shaham, 2002). We recently tested the effect of selective D1-like (SCH 23390) and D2-like (raclopride) antagonists on context-induced reinstatement of cocaine seeking (H. Crombag, J. W. Grimm, and Y. Shaham, manuscript submitted). Pretreatment with the D1-like or the D2-like receptor antagonists attenuated context-induced renewal of cocaine seeking at doses that had minimal impact on high rate of operant responding for a sucrose reinforcer. However, raclopride, but not SCH 23390, also decreased lever pressing on the previously active lever in a control group that remained in the extinction context. These data indicate that activation of D1-like receptors is involved in context-induced reinstatement of cocaine seeking. These data also suggest that activation of D2-like receptors is involved in this effect, but we cannot rule out that some nonspecific effects of raclopride were involved to some degree.

1. Corticosterone.

We found that manipulations of corticosterone levels by ADX or by pretreatment with a synthesis inhibitor of corticosterone, metyrapone, have no effect on footshock-induced reinstatement of heroin seeking (Shaham et al., 1997b). In cocaine-trained rats, ADX attenuates footshock-induced reinstatement. However, in rats with ADX + corticosterone replacement (to maintain basal levels of the hormone) footshock reinstates cocaine seeking (Erb et al., 1998). Thus, it appears that although footshock-induced corticosterone release is not involved in reinstatement of either heroin or cocaine seeking by the stressor, basal levels of the hormone are required for footshock-induced reinstatement of cocaine seeking. Two other studies, however, appear to challenge the above conclusion. Deroche et al. (1997) found that corticosterone infusions reinstate cocaine seeking. Mantsch and Goeders (1999a) reported that inhibition of corticosterone synthesis with the antimycotic agent, ketoconazole, which one of its actions is to inhibit corticosterone synthesis, attenuates footshock-induced reinstatement of cocaine seeking. The results of this study, however, are difficult to interpret due to lack of specificity of ketoconazole, which also acts on several other neurotransmitter and hormonal systems.

There is no evidence that corticosterone is involved in footshock-induced relapse to heroin seeking. In contrast, corticosterone appears to play some modulatory role in footshock-induced relapse to cocaine seeking, but this hormone is probably not the main mediator of this effect.

2. Corticotropin-Releasing Factor.

Acute injections of CRF induce reinstatement of heroin seeking (Shaham et al., 1997b). Moreover, the CRF receptor antagonists, d-Phe CRF and α-helical CRF, attenuate footshock-induced reinstatement of heroin and cocaine seeking (Shaham et al., 1997b; Erb et al., 1998; Lu et al., 2000, 2001a). The effect of CRF on reinstatement appears to be mediated by the CRF₁ receptor subtype. Administration of CP-154,526, a selective CRF₁ receptor antagonist, attenuates footshock-induced reinstatement of heroin and cocaine seeking (Shaham et al., 1998; Lu et al., 2000, 2001a). In contrast, administration of the CRF₂ receptor antagonist, antisauvagine-30, has no effect on footshock-induced reactivation of CPP for morphine and cocaine (Lu et al., 2000, 2001a). These data, together with those from the studies of the lack of effect of ADX (heroin-trained rats) or ADX + corticosterone replacement (cocaine-trained rats) on footshock-induced reinstatement, suggest that the effect of the CRF receptor antagonists on reinstatement is mediated via their action on extra-hypothalamic sites, independent of their effect on the HPA axis.

Erb and Stewart (1999) studied the role of CRF in two brain sites involved in stress response and the behavioral effects of CRF, the bed nucleus of the stria terminalis (BNST) and the amygdala (Davis et al., 1997; Koob and Heinrichs, 1999). They found that infusions of CRF into the ventral BNST reinstate cocaine seeking, whereas intra-BNST infusions of d-Phe CRF attenuate footshock-induced reinstatement (Fig. 10). In contrast, infusions of CRF into the amygdala (infusions were aimed at the CeA) have no effect on reinstatement and intra-amygdala infusions ofd-Phe CRF do not attenuate footshock-induced reinstatement (Erb and Stewart, 1999). In a subsequent study with heroin-trained rats, however, we found that reversible inactivation of both the CeA and the BNST with TTX blocks footshock-induced reinstatement of heroin seeking (Shaham et al., 2000a). The CeA contains high densities of cell bodies and projection neurons for CRF, but unlike the BNST, the density of the CRF receptors in this area is low (Gray and Bingaman, 1996; Van Pett et al., 2000). Thus, it may not be surprising that CRF or the CRF receptor antagonists had little effect on reinstatement when infused into the CeA (see also Lee and Davis, 1997). Sakanaka et al. (1986)described a CRF-containing projection from the CeA to the BNST, and it has been speculated that this projection may account, in part, for the increase in CRF release in the BNST following exposure to stress (Lee and Davis, 1997).

• Download figure
• Open in new tab
• Download powerpoint

Figure 10

Mean (± S.E.M.) number of nonreinforced responses on the previously active (cocaine-paired) lever after exposure to 15 min of intermittent footshock in animals pretreated by intra-BNST injections of the CRF receptor antagonist, d-Phe CRF_12–41 (left panel), and after intra-BNST injections of CRF itself; no footshock was given (right panel). ∗, significantly different from the other conditions (p < 0.05). Similar manipulations in the amygdala had no effects. Data are from Erb and Stewart, 1999, reprinted with permission ©1999 from the Society for Neuroscience.

Erb et al. (2001) studied whether the CRF-containing projection from the CeA to the BNST is involved in footshock-induced reinstatement of cocaine seeking. They used an asymmetric lesion method (Parkinson et al., 2000; Easton and Gaffan, 2001) to functionally disconnect the CRF-containing pathway from the CeA to the BNST. This was achieved by simultaneously injecting TTX into the CeA in one hemisphere andd-Phe CRF into the BNST in the opposite hemisphere. Before testing for footshock-induced reinstatement, rats were given asymmetric injections of TTX (2.5 ng) and d-Phe CRF (50 ng) into the CeA and BNST, respectively, or they were given unilateral injections of either compound alone into the appropriate structure. Functional inactivation of the amygdala-BNST pathway significantly reduced footshock-induced reinstatement compared with that seen when only one hemisphere was manipulated. However, the fact that this attenuation of footshock-induced reinstatement was not complete implies the existence of an additional source of CRF in the BNST other than the CeA-BNST pathway, possibly from cells intrinsic to the BNST (Veinante et al., 1997).

These data suggest that activation of extra-hypothalamic CRF receptors is involved in footshock-induced reinstatement of heroin and cocaine seeking (see also Lê et al., 2000 for similar findings in alcohol-trained rats). In addition, this effect is probably mediated by activation of CRF₁ but not CRF₂ receptors. Finally, it appears that CRF receptors within the BNST and possibly the CRF projection from the CeA to the BNST are involved in footshock-induced reinstatement.

1. Drug Priming-Induced Reinstatement.

Activation of μ-opioid receptors is critically involved in heroin-induced reinstatement, and DA mechanisms (likely to be initiated within the VTA) also are involved in this effect. In the case of cocaine priming, the data indicate that D1- and D2-like DA receptors play fundamentally different roles: activation of D2-like receptors promotes cocaine seeking, whereas activation of D1-like receptors inhibits it. In addition, activation of AMPA receptors within the NAc is involved in cocaine-induced reinstatement. An important question for future research, therefore, is what are the brain sites through which DA mediates cocaine-induced reinstatement. Finally, although pharmacological agents acting at receptors of several other neurotransmitter systems (e.g., 5-HT_2C, GABA_B) can attenuate cocaine-induced reinstatement, this effect is probably mediated via the effect of these receptor manipulations on DA release.

2. Cue-Induced Reinstatement.

The D1-like receptor appears to play a critical role in cocaine reinstatement induced by discrete CSs, discriminative and contextual cues, a finding in agreement with the general role of this receptor in conditioned reward (Sutton and Beninger, 1999). In addition, it appears that the D2-like receptor also is involved in contextual or discriminative cues-induced reinstatement of cocaine, but not heroin seeking. A critical brain site that mediates discrete CSs-induced reinstatement is the BLA, in which D1-like, but not glutamate, receptors are involved in this effect. Indirect evidence from in vivo microdialysis and Fos expression studies also implicates the BLA in cue-induced reinstatement. In contrast, a role for the NAc in cocaine cues-induced reinstatement has not been established. Manipulations of 5-HT neurotransmission (lesions and 5-HT reuptake blockers) can alter lever pressing controlled by the cocaine cues during extinction, but the exact role of 5-HT in extinction behavior is not clear.

3. Stress-Induced Reinstatement.

The effect of footshock stress on reinstatement is opioid-independent and DA appears to play some modulatory role in this effect. In addition, footshock-induced corticosterone secretion is not involved in the effect of the stressor on reinstatement. Two main neurotransmitter systems, CRF and NA, and two main brain structures, the CeA and the BNST, are involved in footshock stress-induced reinstatement. The data also suggest that two neuronal pathways are involved in this effect: the VNAB, which originates from the lateral tegmental NA neurons, and possibly a CRF pathway from the CeA to the BNST. Finally, the hormone leptin is involved in reinstatement of heroin seeking induced by food deprivation stress.

An important conclusion from this review is that the neural systems involved in drug priming-, cue-, and stress-induced reinstatement are to a large degree dissociable. In the case of footshock stress and drug priming, a pharmacological double dissociation is evident. Selective D1- and D2-like receptors and μ-opioid antagonists that block heroin-induced drug seeking (Shaham and Stewart, 1996; McFarland and Ettenberg, 1997, 1998) have no effect on footshock-induced heroin seeking (Shaham and Stewart, 1996). Conversely, CRF receptor antagonists and α-2 adrenoceptor agonists that block footshock-induced reinstatement have no effect on drug-induced reinstatement (Shaham et al., 2000a). In the case of cue- and drug-induced reinstatement, Stewart et al. (1984) suggested that drug cues reinstate drug seeking by inducing a “drug-like” state similar to that induced by the drug itself, which is mediated in part by enhanced DA neurotransmission. However, whereas DA appears to be involved in both drug- and cue-induced reinstatement (see above), pharmacological and neuroanatomical dissociations were reported. For example, the NAc appears to be a critical substrate for cocaine-induced reinstatement (Cornish and Kalivas, 2000), but lesions of this structure have no effect on discrete CSs-induced reinstatement (Grimm and See, 2000). On the other hand, reversible lesions of the NAc, which block cocaine self-administration, have no effect on cue-induced reinstatement (Grimm and See, 2000). In addition, in the case of heroin seeking, DA and opioid receptor antagonists block heroin-induced drug seeking, but they have no effect on drug seeking provoked by discriminative heroin cues (McFarland and Ettenberg, 1997, 1998).

Another issue is the degree of overlap between the neuronal substrates underlying cocaine versus heroin seeking. In the case of drug priming, it appears that activation of the mesocorticolimbic DA system is involved in both heroin and cocaine seeking (Stewart, 2000). However, cross-reinstatement between opioid and stimulant drugs is to a large degree asymmetrical. Indirect DA agonists or D2-like agonists reinstate heroin seeking (Wise et al., 1990; De Vries et al., 1999), whereas μ-opioid receptor agonists (given systemically) do not reinstate cocaine seeking (de Wit and Stewart, 1981; Comer et al., 1993). However, the initial sedative effects of opioid agonists, given acutely to opioid-naive rats, may mask their motivational effect as infusions of morphine into the VTA reinstate cocaine seeking (Stewart, 1984). In the case of footshock-induced reinstatement, the pharmacological data suggest that similar neuronal mechanisms are involved in heroin or cocaine seeking (Shaham et al., 2000a). One exception is that basal levels of corticosterone are required for footshock-induced reinstatement of cocaine but not heroin seeking (Shaham et al., 2000a). There is very little information on similarities/differences in cue-induced reinstatement of heroin and cocaine seeking, but it is possible that these are not identical. For example, although the preferential D2-like antagonist haloperidol has no effect on discriminative cue-induced reinstatement of heroin seeking in the runway model (McFarland and Ettenberg, 1997), selective D2-like antagonists appear to attenuate discriminative (Weiss et al., 2001) and contextual (H. Crombag, J. W. Grimm, and Y. Shaham, manuscript submitted) cues-induced reinstatement of cocaine seeking. In addition, studies using the second-order schedule procedure have shown that manipulations that block drug seeking controlled by CSs in cocaine-trained rats (e.g., a D3 receptor partial agonist, BLA lesions) have no effect on cue-induced drug seeking in heroin-trained rats (Everitt and Robbins, 2000).

Finally, the data reviewed suggest that the neuronal substrates that mediate drug-induced reinstatement are to some degree different from those that mediate drug reinforcement. First, although D1-like receptor agonists substitute for cocaine in the drug self-administration method and some of these agents produce CPP (Self and Nestler, 1995; Abrahams et al., 1998), they do not reinstate cocaine seeking (Self and Nestler, 1998). Second, although NMDA receptor antagonists are self-administered directly into the NAc (Carlezon and Wise, 1996), an NMDA antagonist does not reinstate cocaine seeking (Cornish and Kalivas, 2000). Third,Cornish et al. (1999) found that whereas NMDA and AMPA agonists have a minimal effect on cocaine self-administration behavior, these agonists reinstate cocaine seeking. These investigators also showed that blockade of DA receptors in the NAc, known to be involved in cocaine reinforcement (Wise, 1996b), have no effect on cocaine-induced reinstatement. Fourth, although manipulations of corticosterone secretion have a profound effect on cocaine self-administration behavior (Piazza and Le Moal, 1996; Goeders, 1997), similar manipulations have a minimal effect on cocaine-induced reinstatement (Erb et al., 1998; Mantsch and Goeders, 1999b). Fifth, Vorel et al. (2001) showed that electrical stimulation of the medial forebrain bundle—known to be involved in reward processes (Wise, 1996a)—has no effect on reinstatement of cocaine seeking. In contrast, stimulation of the ventral hippocampus, a brain area that has not been reported to be involved in drug reward, potently reinstates cocaine seeking. Sixth, although blockade of brain CB1 receptors attenuates cocaine-induced reinstatement, this manipulation has no effect on cocaine self-administration behavior (De Vries et al., 2001). Finally, although DA receptor antagonists do not have a consistent effect on heroin self-administration behavior (Ettenberg et al., 1982; Mello and Negus, 1996), they reliably attenuate drug seeking induced by heroin priming (Shaham and Stewart, 1996; McFarland and Ettenberg, 1997).

It appears that multiple and dissociable brain systems are involved in relapse to heroin and cocaine seeking induced by drug priming, conditioned cues, and stress. Somewhat surprisingly, it also appears that the neuronal events that mediate heroin- or cocaine-induced reinstatement are to some degree different from those involved in their reinforcing effects.

1. Drug Priming.

Several explanations for the drug priming effect have been put forward over the years. It has been suggested that the discriminative stimuli properties of drugs mediate in part drug-induced reinstatement (Stolerman, 1992; Bergman and Katz, 1998). Rats can readily learn to discriminate drugs from saline in a drug discrimination model, wherein deprived rats learn to press one lever for food/water following drug injections and a different lever following saline injections (Stolerman, 1992). The drug discrimination procedure is regarded as an animal model for the subjective effects of drugs in humans (Preston, 1991). According to a discriminative stimulus account of reinstatement, exposure to the self-administration drug during testing elicits a selective increase in responding on the previously active lever because certain drug effects signal the rat that pressing this lever but not the inactive lever will lead to drug infusions. Proponents of this view typically point out that as in the case of drug discrimination response generalization is most likely to occur within the same drug class.

However, studies on the neuronal substrates that mediate the discriminative stimulus effects of heroin suggest that they are different from those that mediate drug-induced reinstatement. Intra-VTA injections of morphine—a manipulation that induces reinstatement of heroin seeking (Stewart, 1984)—do not induce heroin-appropriate responding in the drug discrimination method (Shaham and Stewart, 1995a; Jaeger and van der Kooy, 1996; but see Shoaib and Spanagel, 1994). Jaeger and van der Kooy (1993, 1996) further showed that the rewarding and discriminative effects of morphine are anatomically dissociable and that the discriminative drug effects are neither necessary nor sufficient for morphine's rewarding effects. Finally, the observations that indirect DA agonists and D2-like agonists reinstate heroin seeking (Wise et al., 1990; De Vries et al., 1999) is not in agreement with most studies from the drug discrimination literature, in which generalization to the training drug is typically observed within a given drug class (Stolerman et al., 1989, 1995).

Stewart and colleagues hypothesized that reinstatement by drug priming is due to the ability of the drug to induce an incentive motivational state, which leads to resumed drug seeking (Stewart et al., 1984). This incentive motivational state refers to the arousing/activating state that occurs following exposure to an appetitive unconditioned stimulus or to stimuli classically conditioned to the unconditioned stimulus (Stewart et al., 1984). They postulated that drug seeking during tests for reinstatement occurs because the presence of the drug in the body (and brain) enhances the incentive value of the extinguished drug cues, previously paired with the drug's rewarding effect. The findings that both heroin and cocaine priming reinstate drug seeking via activation of the mesocorticolimbic DA (Stewart, 2000), thought to be involved in incentive motivation and appetitive goal-direct behavior (Stewart et al., 1984; Robinson and Berridge, 1993; Di Chiara, 1995), provide support to this idea. Additional support for Stewart's hypothesis is the observation that the removal of the drug cues previously associated with amphetamine during testing attenuates amphetamine-induced reinstatement (Stretch et al., 1971). Finally, as argued by Leri and Stewart (2001), the findings that drug priming reinstates an approach behavior toward contextual cues previously associated with the drug's rewarding effects after extinction in the CPP model is consistent with an incentive motivation view but not with a discriminative stimulus view. Specifically, in the CPP procedure the drug is not associated with a specific instrumental responding. Thus, it is unlikely that drug priming acts to induce a habitual form of responding that is elicited by its discriminative stimulus properties (see Bickel and Kelly, 1988).

It appears unlikely that the discriminative stimulus properties of the priming drug injections can account for their effect on reinstatement. In contrast, the available data appear to fit an incentive motivation account of drug priming. It should be pointed out, however, that this state of affairs might be due to the fact that incentive motivation is a theoretical construct, and as such, it is difficult to design experiments that can refute or support its role in reinstatement. It is also likely that incentive motivation is only one of several processes involved in relapse behavior. The data reviewed indicate that multiple neuronal systems are involved in relapse induced by drug priming, cues, and stress, implying that multiple psychological processes are likely to be involved. Finally, as argued by Tiffany (1990), relapse often occurs as a result of automated processes, which are independent of motivational processes.

2. Drug Cues.

It has long been established that conditioned drug cues can provoke relapse in rats and monkeys (Goldberg, 1976). Furthermore, drug-associated stimuli can modulate the behavioral and physiological effects of drugs of abuse (Siegel, 1989;Stewart, 1992; Robinson et al., 1998). In the last 6 years, several laboratories have used established learning procedures, including conditioned reinforcement, discrimination learning, and renewal (Catania, 1992; Bouton, 1993) to study neuronal substrates underlying relapse induced by discrete CSs (previously paired with each drug injection) and discriminative and contextual cues, which predict drug availability but are not specifically paired with each drug injection. This distinction is important since previous studies indicate that different neuronal circuits underlie the behavioral effects of discrete CSs versus contextual stimuli (Phillips and LeDoux, 1992; Holland and Bouton, 1999).

In studies using lever-pressing behavior as the dependent measure, it is often difficult to differentiate between the relative contribution of discrete versus contextual cues in extinction behavior or cue-induced reinstatement. For example, a retractable lever that extends at the start of the training sessions can serve as a contextual cue that predicts drug availability. However, the depression of the lever and the associated auditory click can also serve as discrete CSs that are associated with drug injections. The situation is even more complicated in studies in which both discrete CSs associated with drug/saline injections and discriminative cues that predict drug/saline availability are simultaneously manipulated during discrimination training (Gracy et al., 2000). Interestingly, this experimental manipulation results in a persistent effect of the drug cues on reinstatement over many test sessions and up to 4 months of withdrawal periods (Ciccocioppo et al., 2001).

Even with the renewal procedure, in which one set of contextual cues (e.g., specific smell, floor texture) is associated with drug self-administration training and the other set of cues is paired with extinction training while keeping the discrete CS (e.g., cue light, sound of the pump) constant during training and extinction, the interpretation of the data is not straightforward. One possibility is that the effect of the drug context on renewal of drug seeking is due to a nonspecific effect of switching rats to a different environment. This possibility is not likely becasue we found that rats exposed to a novel environment during testing do not resume lever-pressing behavior (Crombag and Shaham, 2002), a finding in agreement with previous reports (Bouton and King, 1983; Goddard, 1999; Nakajima et al., 2000). Alternatively, contextual stimuli, because they reliably signal drug availability, acquire excitatory conditioned stimulus (CS+) qualities, including incentive motivational properties. In our studies, extinction training occurred in a different environment and therefore contextual cues would have retained their motivational value. Exposure of rats to these (nonextinguished) contextual cues could have elicited cocaine seeking.

Another possibility is that drug-associated contextual stimuli provoke relapse because of their occasion setter (Holland, 1992) or modulator (Rescorla et al., 1985) properties. Unlike traditional CSs, occasion setters do not elicit behavior by themselves, but rather modulate the behavioral effects of other conditioned stimuli (Holland, 1992). According to this view, contexts function as retrieval cues in cases where the meaning of the discrete CSs is ambiguous because they have been paired with both reinforcement and nonreinforcement (Bouton, 1993), as is the case in our studies. Thus, a question that emerges from the above behavioral analysis is whether the DA receptor antagonists block context-induced reinstatement of cocaine seeking by interfering with the putative occasion setting properties of the context or by acting on neuronal systems that underlie conditioned behaviors elicited by discrete CSs.

Thus, unlike studies using classical conditioning paradigms on mechanisms underlying contextual versus discrete cues on behavior (e.g., conditioned fear; conditioned activity), in studies on drug cues-induced reinstatement, it is difficult to differentiate between the relative contribution of discrete CSs associated with drug infusions versus that of discriminative and/or contextual cues associated with drug availability. It is also likely that on many occasions an interaction between discrete CSs and contextual cues underlies cue-induced relapse to drugs.

3. Stress.

Several mechanisms have been proposed to explain the effect of footshock stress on relapse to drug seeking. It has been suggested that activation of the mesocorticolimbic DA system by stressors underlies their effect on reinstatement, thus mimicking drug priming (Robinson and Berridge, 1993; Shaham and Stewart, 1995a). This hypothesis, however, is not likely as footshock stress- and drug-induced reinstatement can be dissociated pharmacologically and anatomically (Shaham et al., 2000a). It has also been argued that footshock stress might reinstate cocaine seeking by mimicking certain interoceptive cues of cocaine that were present during training (Ahmed and Koob, 1997). Footshock exposure results in cocaine-appropriate responding in drug discrimination tasks (Mantsch and Goeders, 1999a). However, footshock is often a more effective stimulus for reinstatement than drug priming (Shaham, 1996), an observation that would not be predicted if stressors act by mimicking the cue properties of the drug. In addition, Mantsch and Goeders (1999a) reported that a pharmacological manipulation (ketoconazole administration) that blocks footshock-induced reinstatement does not alter footshock-induced cocaine-appropriate responding in a drug discrimination task. Finally, the pharmacological dissociation between footshock- and drug-induced relapse does not support the idea that the stressor mimics interoceptive drug cues.

Whitehead (1974) suggested that stressors induce a withdrawal-like state, which leads to relapse to heroin use. There are similarities between the stress response and symptoms of opioid withdrawal (Redmond and Huang, 1979). However, this idea seems unlikely as withdrawal precipitated by an opioid receptor antagonist does not reinstate heroin seeking and footshock reinstates drug seeking in the presence of a maintenance dose of heroin (Shaham and Stewart, 1995b; Shaham et al., 1996). Finally, Highfield et al. (2000) provided some evidence that footshock stress may provoke relapse by interfering with a neuronal processes underlying response inhibition, whose function is to stop ongoing activity when reinforcers are not available, as is the case during extinction (Pavlov, 1927; Gray, 1987). Four outcomes suggestive of the interruption of inhibitory processes, and which would be predicted by a “disinhibitory” account of footshock-induced reinstatement, were obtained (Highfield et al., 2000). These include: 1) reinstatement of heroin seeking by inactivation of the medial septum, a brain area involved in response inhibition (Grossman, 1977;Gray and McNaughton, 1983); 2) attenuation of footshock-induced reinstatement by stimulation of the medial septum; 3) summation between reduced activation of the medial septum and mild shock; and 4) increased resistance to extinction by footshock stress.

Another unresolved question concerns the mechanisms underlying food deprivation stress-induced reinstatement. We found that leptin attenuates food deprivation-induced, but not footshock- or heroin-induced reinstatement (Shalev et al., 2001b). Thus, it is possible that the neuronal events that mediate acute food deprivation-induced reinstatement are dissociable from those involved in reinstatement induced by footshock or heroin priming. This hypothesis can be confirmed (or refuted) by determining the effect on food deprivation-induced reinstatement of pharmacological agents that block reinstatement by drug priming or footshock.

It appears that footshock stress reinstates drug seeking by mechanisms that are probably unrelated to the ability of the stressor to induce drug-like or withdrawal-like states. An alternative explanation is that acute exposure to footshock disrupts neural processes that normally exert inhibitory control over prepotent behaviors. Finally, the mechanisms underlying food deprivation-induced reinstatement are yet to be elucidated.

4. Summary.

Although a great deal of knowledge has been accumulated on the neuronal mechanisms underlying relapse behavior, the psychological processes involved in relapse induced by drug priming, drug cues, and stressors are not clear. In the case of drug-induced reinstatement, the available data appear to support the idea that incentive motivational processes underlie this effect. In contrast, the data reviewed do not support the idea that the discriminative stimulus effects of drugs mediate drug-induced reinstatement. In the case of cue-induced reinstatement, an unresolved issue is the degree of overlap between the neuronal mechanisms underlying relapse induced by discrete CSs paired with drug injection versus discriminative or contextual cues that predict drug availability. Finally, little is known about the mechanisms underlying stress-induced reinstatement. The available data do not support the idea that footshock reinstates by inducing drug-like or withdrawal-like states. The stressor, however, may reinstate drug seeking by its actions on neuronal processes involved in response inhibition. Finally, it is not known whether food deprivation reinstates drug seeking by acting on neuronal systems involved in drug priming or footshock, or by its action on a different neuronal circuit that is yet to be elucidated.

1. Prior Training for Food Reinforcement.

In many studies, rats are trained to lever press for food prior to drug self-administration training. Although these conditions facilitate drug self-administration training (Carroll, 1999), they may introduce confounds in reinstatement studies. Specifically, when the same experimental setup is used for food training and drug self-administration training, the latter condition comprises a component of extinction training for food. Thus, resumption of lever-pressing responding during testing may be due to reinstatement of food seeking rather than drug seeking. This alternative explanation can be ruled out by using a control group of rats that lever press for food (de Wit and Stewart, 1981; Shaham et al., 1997a), but this condition has rarely been employed in reinstatement studies.

2. Noncontingent Priming Injections during Training.

A common practice in drug self-administration studies is to start the training sessions with one or two priming noncontingent drug injections (Caine and Koob, 1994). As in the case of food training, this manipulation facilitates drug self-administration training but it introduces confounds in the interpretation of the data. When noncontingent drug injections are given at the start of each training session they may become discriminative cues (Catania, 1992) that predict drug availability. As these priming injections are not given during the extinction phase, when drug priming is reintroduced during testing it may reinstate drug seeking because of its discriminative stimulus properties (i.e., it informs the rat that the drug is now available or that subsequent lever presses will lead to contingent drug injections). Thus, when priming injections are given during training, their effect on reinstatement after extinction may be due to their discriminative stimulus effects, incentive motivational effects (see Section V.B.1.), or both.

3. Response Rates during Training.

The effect of pharmacological manipulations on operant behavior maintained by drugs or nondrug reinforcers is dependent on baseline rates of responding, the rate dependence effect (Sanger and Blackman, 1976; Witkin, 1994). For example, DA receptor antagonists increase lever pressing for a high unit dose of cocaine when each lever press is reinforced, a fixed ratio-1 (FR-1) schedule (Wise, 1978; Ettenberg et al., 1982). On the other hand, under an intermittent schedule of reinforcement (FR-15), which leads to a high rate of responding, DA receptor antagonists decrease lever pressing for cocaine (Caine and Koob, 1994). An important question, therefore, is whether the rate of responding during training leads to qualitative differences in the behavioral effects of drug priming, drug cues, and stressors during testing. Response rates during training can be manipulated by changing either the schedule of reinforcement, the unit dose (i.e., the dose per infusion) of the training drug, or both.

Despite the fact that only several studies have addressed this issue empirically (see below), we argue that it is not likely that the response rate during training is a major methodological concern in reinstatement studies. The main reason for this argument is that regardless of the rate of responding during training (and extinction), tests for reinstatement are conducted under low rates of responding after extinction. The available data on the effect of rate of responding during training on reinstatement appear to support this argument. Comer et al. (1995) found that under an FR-1 schedule, the training dose of cocaine (0.2, 0.4, or 1.0 mg/kg/infusion, which leads to high, moderate, or low response rates, respectively) does not alter reinstatement induced by cocaine priming in rats. In addition, regardless of the training dose, food restriction enhanced the effect of cocaine priming on reinstatement. Leri and Stewart (2001) trained rats to lever press for heroin (0.025, 0.05, 0.1, or 0.2 mg/kg/infusion; FR-1 schedule) or cocaine (0.25, 0.5, 1.0, or 2.0 mg/kg/infusion; FR-1 schedule) on alternate days. They found that response rates during training (which are high with low doses and low with high doses) do not predict the response to cocaine or heroin priming after extinction. In contrast, Kruzich et al. (1999) reported that the response rate during training was negatively correlated with the magnitude of cue-induced reinstatement. These data should be interpreted with caution as they are based on correlational analyses rather than on experimental manipulations.

Most important, it was found that regardless of the response rates during training, pharmacological agents have similar effects on reinstatement. For example, the effect of D1- and D2-like agents on reinstatement of cocaine seeking in rats trained on a high unit dose and an FR-1 schedule (Wise et al., 1990; Self et al., 1996), conditions that lead to low response rates, generalizes to monkeys trained on a second-order schedule, which leads to high rates of responding (Khroyan et al., 2000). Furthermore, it has been shown that pharmacological manipulations (e.g., CRF receptor antagonists) that attenuate footshock-induced reinstatement in an operant procedure (Shaham et al., 2000a), also attenuate reactivation of CPP by the stressor in a classical conditioning, rate-independent paradigm (Lu et al., 2000,2001a).

In conclusion, it does not appear that differences in response rates within and between studies lead to qualitatively different effects on reinstatement of drug seeking. The fact that findings from reinstatement studies based on operant self-administration behavior generalize to those from rate-independent CPP studies further supports this conclusion. However, in agreement with learning studies on the partial reinforcement effect, the increased resistance to extinction after training with intermittent versus continuous reinforcement schedules (Gray, 1987), it appears that reinstatement by drug or nondrug stimuli can be enhanced when rats are trained under intermittent schedules of reinforcement. Thus, training rats on intermittent schedules of reinforcement leads to higher rates of responding after exposure to cocaine priming (Tran-Nguyen et al., 1998) or footshock stress (Mantsch and Goeders, 1999a) than those observed in studies in which an FR-1 schedule of reinforcement was used.

4. Amount of Drug Exposure during Training.

Several recent studies indicate that the amount of drug intake during training can influence the effect of drug priming and footshock stress on reinstatement. The effect of prior drug exposure during training on reinstatement, however, is quantitative rather than qualitative. Compared with rats trained for 6 days, Deroche et al. (1999) reported that rats trained to lever press for cocaine for 29 days demonstrated a shift to the left in the dose-response curve in response to cocaine priming (0.2–1.6 mg/kg, i.v.). Sutton et al. (2000) and Baker et al. (2001) reported that total cocaine intake during training is correlated with the magnitude of reinstatement induced by amphetamine or cocaine priming, respectively. In the study of Sutton et al. (2000), however, the amount of cocaine intake during training was not correlated with reinstatement induced by drug cues or a mild footshock. In contrast,Ahmed et al. (2000) reported that, compared with rats trained for 1 h/day, rats trained for 11 h/day to self-administer heroin demonstrated an enhanced response to footshock stress during testing. Thus, it appears that the magnitude of reinstatement induced by drug priming is associated with the amount of cocaine intake during training. At present, however, a clear picture is yet to emerge on the relationship between drug intake and reinstatement induced by stressors and drug cues.

5. The Drug Withdrawal Period Prior to Tests for Reinstatement.

Recent studies indicate that the duration of the withdrawal period prior to tests for reinstatement is an important factor, which can lead to quantitative andqualitative differences in reinstatement by drug and nondrug stimuli. Tran-Nguyen et al. (1998) demonstrated that the response to cocaine priming is increased after 1 month of withdrawal compared with 1 day or 1 week (Fig.13A). In addition, the effect of D2-like receptor agonists on reinstatement of heroin seeking is critically dependent on the withdrawal period. Using a within-session model, Wise et al. (1990) found that the D2-like receptor agonist, bromocriptine, potently reinstates heroin seeking when given several hours after drug self-administration. De Vries et al. (2002) found that the D2-like agonist quinpirole reinstates heroin seeking after short periods (less than 1 week), but not after prolonged periods (greater than 3 weeks) of drug withdrawal.

There also are profound differences in response to drug cues and footshock stress at different time periods after drug withdrawal. In a study with cocaine-trained rats, we found that the response to the cocaine cues progressively increased following withdrawal from cocaine and that, surprisingly, reinstatement was not observed on day 1 of withdrawal (Grimm et al., 2001) (Fig. 13B). In another study with heroin-trained rats, we found that reinstatement of lever-pressing behavior by footshock followed an inverted U-shaped curve with maximal responding after 6 and 12 days of heroin withdrawal (Shalev et al., 2001a) (Fig. 13C). Surprisingly, footshock did not reinstate behavior on day 1 of heroin withdrawal.

6. Side Effects of the Pharmacological and Brain Manipulations.

An important issue in reinstatement studies is concerned with the interpretation of data from studies in which pharmacological agents and/or brain manipulations are given prior to tests for reinstatement. The question is what are the necessary control conditions to determine that decreases (or increases) in active lever responding during testing reflect the effects of the experimental manipulations on drug seeking rather than some other nonspecific effects. At present, there is no single behavioral measure that can adequately address this question. Thus, it is necessary to collect data from several dependent measures to rule out that nonspecific effects of the experimental manipulations led to changes in behavior.

A common practice is to determine the effect of the experimental manipulations on responses on the inactive lever as a measure of nonspecific activity. However, given that baseline rates of responding on this lever are low, nonspecific sedative effects cannot be adequately assessed. In addition, even if the experimental manipulations increase responding on the inactive lever, it does not necessarily indicate that it is due to nonspecific behavioral activation. When tests are conducted under extinction conditions, increased responding on the inactive lever may reflect response generalization, which commonly occurs during extinction (Catania, 1992).

Another way to assess nonspecific effects of the pharmacological/brain manipulations is to determine their impact on lever pressing after extinction but in the absence of the reinstating stimulus. However, as in the case of inactive lever responses, because response rates on the active lever are low after extinction, nonspecific sedative effects are difficult to assess. In addition, even if an experimental manipulation selectively increases responding on the active lever but not on the inactive lever, the increased responding may be in part due to nonspecific behavioral activation. Early studies on the rate dependence effects of methamphetamine have shown that the drug does not increase low rate baseline responding that was never reinforced (Verhave, 1958).

Potential sedative effects can be determined by examining the effect of the experimental manipulations on ongoing high rate of responding for a nondrug reinfrocer such as sucrose (Weissenborn et al., 1995; Erb et al., 2000). The specificity of the experimental manipulations on drug seeking, for example, can also be studied by determining their impact on reinstatement of food seeking (de Wit and Stewart, 1981; Cornish et al., 1999) or reinstatement of lever pressing previously maintained by another drug (De Vries et al., 1999). For example, Leri and Stewart (2001) have shown that in rats trained on alternate days to lever press for cocaine and heroin, each with its associated lever and distinct cues, cocaine or heroin priming selectively reinstates lever pressing on its drug-associated lever. However, although the interpretation of the data is clear if the experimental manipulations do not alter response rates under the above control conditions, a change in response rates for a nondrug reinforcer or for a different drug may not be necessarily due to nonspecific effects. To the extent that drugs of abuse act on brain systems underlying natural rewards (Wise and Rompre, 1989), experimental manipulations that decrease drug seeking may also decrease operant responding for food reward or reinstatement of responding previously maintained by other drug or nondrug reinforcers. Finally, the selectivity of a given manipulation can be studied by determining its effect on more than one reinstating stimulus. For example, CRF receptor antagonists selectively attenuate footshock-induced reinstatement but not reinstatement induced by drug priming (Shaham et al., 2000a).

The measurement of changes in responding on the inactive lever during tests for reinstatement is not an optimal measure of nonspecific effects. Thus, it is necessary to collect converging evidence from several measures. These may include the effect of the manipulations on high rate of operant responding for a nondrug reinforcer, reinstatement of food/water seeking, and reinstatement by other stimuli.

7. Summary.

A number of methodological issues should be considered in the design and interpretation of reinstatement studies. To avoid confounds in reinstatement studies, rats should not be food trained prior to drug self-administration training and should not be given noncontingent drug injections at the onset of the training sessions. Based on limited evidence, it appears that the response rate during training does not lead to qualitative changes in responding during tests for reinstatement. Training rats under intermittent schedules of reinforcement, however, can lead to increased responding during testing. Another factor associated with increased responding during tests for reinstatement is the amount of drug intake during training. Recent evidence indicates that the duration of the drug withdrawal period prior to testing is an important factor in reinstatement by drug or nondrug stimuli. Finally, to assess potential side effects of pharmacological and brain manipulations given during tests for reinstatement, data should be collected from several control conditions (e.g., operant responding for food, reinstatement of food seeking), in addition to inactive lever responses.

1. Does Drug Sensitization Contribute to Relapse to Heroin and Cocaine?

Repeated exposure to opioid and stimulant drugs results in enhanced behavioral and neurochemical responses to the drugs (sensitization) or stressors (cross-sensitization) after withdrawal periods (Kalivas and Stewart, 1991; Piazza and Le Moal, 1996). These sensitized behavioral and neurochemical responses to drugs and stressors peak at time points that are beyond the acute withdrawal phase and persist for many months, and the mesocorticolimbic DA system is involved in the manifestation of these sensitized responses (Pierce and Kalivas, 1997; White and Kalivas, 1998). These observations, and those demonstrating that pre-exposure to drugs and stressors facilitate the initiation of drug self-administration in rats (Goeders, 1997;Schenk and Partridge, 1997; Piazza and Le Moal, 1998), were the basis for the hypothesis that drug sensitization also contributes to drug relapse (Robinson and Berridge, 1993; Piazza and Le Moal, 1996; De Vries et al., 1998a; Kalivas et al., 1998).

2. Application of the Reinstatement Model to Mice.

The rat reinstatement model may not be the most suitable mean for studying genetic factors in drug addiction and relapse. Methods for over-expressing genes via viral vectors or inactivating genes by antisense oligonucleotides have contributed to the understanding of the mechanisms of drug reward in rats (Nestler, 2000, 2001). For several reasons, however, the mouse is a more appropriate species to study genetic factors in drug addiction. There are a number of inbred mouse strains and several molecular genetic techniques in mice that allow them to either inactivate (knockout methods) or over-express (transgenic methods) specific genes potentially involved in drug effects (Crabbe et al., 1999; Nestler, 2001). Mice self-administer opioid (Elmer et al., 1996; Roberts et al., 1997) and stimulant (Rocha et al., 1998; Caine et al., 1999a) drugs and also demonstrate CPP for these drugs (Cunningham et al., 1992; Laviola et al., 1992).

There are two published reports on reinstatement in mice, both of which used the CPP procedure. Manzanedo et al. (2001) trained mice (OF1 albino strain) to acquire morphine (40 mg, i.p.) place preference for 4 days. Subsequently, mice were repeatedly tested for morphine CPP in the absence of the drug 5, 15, 20, 35, and 45 days after last exposure to morphine (extinction). CPP for morphine was not observed on the last 3 test days, but a priming injection of morphine (40 mg/kg) led to reinstatement of CPP 46 days after training. Itzhak and Martin (2002)trained Swiss-Webster mice for cocaine CPP and then extinguished this behavior by administering saline injections (8 days) in the previously cocaine- and saline-paired compartments. Subsequently, they found reinstatement of CPP following injections of cocaine, methamphetamine, and methylphenidate but not phencyclidine.

We trained male mice of the 129X1/SvJ strain for 14 to 16 days to self-administer cocaine (0.75 mg/kg/infusion, 4 h per day; infusions were paired with a light-tone compound cue). Next, the lever-pressing behavior was extinguished in the presence or absence (for subsequent tests for cue-induced reinstatement) of the cocaine cue. Subsequently, tests for reinstatement were conducted in different groups of mice after exposure to priming injections of cocaine (0, 1.5, 3.0, and 6.0 mg/kg, i.v.), response-contingent presentations of the cocaine-associated cue or food deprivation stress (1 and 22 h). We found that the effect of cocaine priming on reinstatement was relatively modest and was only observed at the highest dose tested. On the other hand, reinstatement of cocaine seeking was observed following exposure to the cocaine-associated cue and food deprivation stress. These data tentatively suggest that factors contributing to relapse to drugs can be studied in the reinstatement model using the common 129X1/SvJ mouse inbred strain (D. Highfield, A. Mead, J. W. Grimm, B. A. Rocha, and Y. Shaham, manuscript submitted). The reasons for the weak effect of the priming cocaine injections in mice are not clear.

Although species differences in response to reinstating stimuli may exist, both operant and classical conditioning models of reinstatement can be used in mice. Studies using molecular genetic methods are likely to provide information on the role of specific genes in relapse to heroin and cocaine.

1. Implications for Addiction Theories.

Negative reinforcement theories postulate that compulsive drug use and drug relapse occur because the addict is seeking drugs to alleviate the aversive symptoms of drug withdrawal; symptoms that can also be elicited by cues previously paired with drug withdrawal (Himmelsbach, 1943; Lindsmith, 1947; Wikler, 1973). Over the years, several drug-opposite physiological and psychological adaptations have been hypothesized to underlie habitual drug use and relapse (Goldstein and Goldstein, 1968; Solomon and Corbit, 1974; Collier, 1980; Koob et al., 1989; Siegel, 1989). The self-medication hypothesis, which argues that compulsive drug use and relapse is due to the drug's effects on the individual's well being, is another form of a negative reinforcement model (Khantzian, 1985; Markou et al., 1998). It is beyond the scope of this paper to describe these negative reinforcement theories in detail and to discuss the degree to which they account for drug self-administration behavior (see Schuster and Thompson, 1969; Stewart et al., 1984; Wise and Bozarth, 1987). In the context of relapse to heroin and cocaine seeking as measured in the reinstatement model, however, we argue that there is little evidence that negative reinforcement models are compatible with the data on relapse induced by drug priming, drug cues, or stressors.

In many of the studies reviewed, rats were trained under conditions that are not sufficient to induce measurable withdrawal symptoms (i.e., 1–2 h/day of drug self-administration or only a single daily exposure to low drug doses in the runway and CPP models). As mentioned, attempts to induce relapse to heroin seeking by precipitating opioid withdrawal were not successful (Shaham and Stewart, 1995b; Shaham et al., 1996). Cues associated with drug administration can induce drug-opposite withdrawal-like effects (Siegel, 1989), but there is no evidence to date that such effect of cues can induce relapse in the reinstatement model. Finally, in our studies, in which rats were trained for 6 to 9 h/day leading to drug intakes that are sufficient to induce withdrawal symptoms during early withdrawal (1–2 days) from heroin (Shaham et al., 1996) and cocaine (Sarnyai et al., 2001), we found that extinction behavior, cue-induced reinstatement, and footshock-induced reinstatement were maximal at time points that are well beyond the acute withdrawal phase. These data are in agreement with those from a different model of drug seeking, the second-order schedule, in whichArroyo et al. (1998) found that during early withdrawal from 12-h cocaine self-administration, a decrease in lever-pressing behavior controlled by the cocaine cues is observed.

The limitations of negative reinforcement theories in explaining compulsive drug use led to the developments of several addiction theories, which were primarily based on the observations that drugs are powerful positive reinforcers and maintain behavior in the absence of physical dependence (Weeks and Collins, 1968; Schuster and Thompson, 1969). Wise and Bozarth (1987) argued that “a common brain mechanism of psychomotor activation is stimulated by all positive reinforcers and that the same mechanism mediates the psychomotor stimulant effects and the positive reinforcing effects of these agents” (Wise, 1988). These authors also argued that drugs elicit approach behavior or forward locomotion (Glickman and Schiff, 1967; Bindra, 1974), which leads to compulsive drug use. Stewart et al. (1984) argued that drugs or drug-associated cues elicit a positive incentive motivational state that leads to approach behavior toward these stimuli, and consequently to compulsive drug use and relapse upon exposure to these stimuli.Robinson and Berridge (1993) hypothesized that incentive sensitization processes or increased responsiveness to drug and drug-associated cues following repeated drug exposure may underlie compulsive drug use and relapse. More recently, Di Chiara (1999) argued that compulsive drug use occurs because repeated stimulation of DA in the NAc shell by drugs of abuse in the presence of drug-associated cues leads to the attribution of excessive motivational value and abnormal control over behavior by these cues. Again, it is beyond the scope of this paper to review these theories in detail, to discuss their similarities and differences, the degree to which they account for compulsive drug use, and the degree to which these theories fit with the data from studies on the role of the mesocorticolimbic DA system in goal-directed behavior (see Berridge and Robinson, 1998; Ikemoto and Panksepp, 1999;Schultz and Dickinson, 2000).

However, a common denominator of all four theories is that the mesocorticolimbic DA system is involved in compulsive drug use, albeit via the elicitation of somewhat different motivational processes, and that the presence of drugs in the body (and brain) rather than drug withdrawal is the critical factor in compulsive drug use and relapse.Stewart et al. (1984) and Robinson and Berridge (1993) also argued that drug cues elicit drug-like effects and consequently drug seeking via their action on the mesocorticolimbic DA system.

To a certain degree, the data reviewed are in agreement with these theories. As mentioned above, the available data tentatively support the view that the putative incentive motivational effects of heroin and cocaine priming may underlie their effect on reinstatement (Stewart et al., 1984). The findings of De Vries et al. (1998a, 1999), that the ability of drugs to induce locomotor activity is correlated with their effects on reinstatement, are in agreement with all four theories. In addition, as discussed in detail above (Section V.D.1.), recent data suggest that psychomotor sensitization (a phenomenon that served as the basis for the incentive-sensitization theory) is associated with drug- and possibly cue-induced reinstatement. Finally, the observations that the mesocorticolimbic DA system is involved in both drug priming- and cue-induced reinstatement is in agreement with the four theories.

Importantly, however, the anatomical (Grimm and See, 2000) and pharmacological (McFarland and Ettenberg, 1997) dissociations between reinstatement by drug priming and drug cues are not predicted by the theories of Stewart et al. (1984) and Robinson and Berridge (1993). In addition, all four theories would not predict the partial dissociation between the neuronal mechanisms underlying acute heroin or cocaine reinforcement and drug-induced reinstatement (Section V.A.). Furthermore, although Robinson and Berridge (1993) argued that stressors provoke relapse by mimicking the effect of drugs on the mesocorticolimbic DA system, this does not appear to be the case (Shaham et al., 2000a).

Finally, Koob and Le Moal (2001) proposed an “allostasis” model of drug addiction, in which they argued that repeated and compulsive use of high amounts of drugs induces changes in the drug's “hedonic” set point. Allostasis is a process wherein chronic disturbances of homeostasis of a given physiological (and psychological) system lead to a new stable set point (which is outside of the normal homeostatic range) to cope with the chronic internal and/or external demands (Schulkin et al., 1998). Ahmed and Koob (1998, 1999) provided evidence in support of this view by demonstrating that rats given access to cocaine for 6 h/day, but not 1 h/day, escalate their drug intake over time. However, it does not appear that the allostasis model can account for the results obtained in reinstatement studies. Specifically, following 35 days of withdrawal, cocaine intake returned to normal pre-escalating values in rats which previously self-administered cocaine for 6 h/day (Ahmed and Koob, 1998). On the other hand, in reinstatement studies, the time course of extinction behavior and cocaine- and cue-induced reinstatement has an opposite pattern, i.e., low responding on day 1 of withdrawal and high responding following 1 to 2 months of withdrawal (Tran-Nguyen et al., 1998; Grimm et al., 2001). In other words, drug seeking is inversely related to the changes in the putative hedonic set point.

It appears that negative reinforcement and opponent processes theories and the recent allostasis model cannot adequately explain the data obtained in reinstatement studies. In contrast, incentive motivation, incentive sensitization, and psychomotor theories of addiction can account for certain findings from reinstatement studies. However, even these theories cannot account for the pharmacological and anatomical dissociations between drug priming, drug cues, and stressors and for the partial neuronal dissociation between drug-induced reinstatement and drug reinforcement.

2. Implications for Treatment.

As mentioned in the Introduction, the reinstatement model appears to have good predictive validity because conditions that provoke drug relapse and craving in humans (drug re-exposure, drug cues, and stress) also reinstate heroin and cocaine seeking following prolonged withdrawal periods in laboratory animals. Over the past several decades, medication development studies have screened potential clinical compounds for their ability to attenuate drug withdrawal symptoms, to substitute for the abused drug in drug self-administration or discrimination models, or to block the reinforcing or discriminative stimulus effects of the abused drug in these models (Bhargava, 1994; Mello and Negus, 1996). The data reviewed here, however, suggest that effective compounds derived from the above screening methods may not prevent relapse to drug seeking. As mentioned, the data reviewed suggest that the neuronal processes that mediate drug-induced reinstatement are to some degree different from those involved in drug reinforcement or discrimination. Furthermore, even compounds that can effectively block relapse induced by drug priming or drug cues are not likely to block stress-induced relapse and vice versa. In addition, the recent data of McFarland and Ettenberg (1997, 1998) on the differential effect of naltrexone and haloperidol on heroin-induced drug seeking versus cue-induced heroin seeking suggest that potential medications screened for the effect on drug priming may not always alter relapse induced by drug cues. Finally, the data reviewed suggest that a pharmacological therapy that combines agents that are effective against relapse induced by drug or drug cues with agents that attenuate stress-induced relapse should be considered for relapse prevention in humans.