Introduction

A central feature of drug addiction is the high rate of relapse during abstinence (Hunt et al, 1971; O'Brien, 2005). Over the years, results from many correlational human studies have suggested that this relapse is often provoked by stress (Box 1) (Khantzian, 1985; Shiffman and Wills, 1985; Kosten et al, 1986). These clinical studies have led to the development of a rat model, which is based on the operant conditioning reinstatement model (Stewart and de Wit, 1987), to study mechanisms of stress-induced relapse to drug seeking (Shaham et al, 2000a). In the reinstatement model, laboratory animals are tested for reinstatement of drug seeking induced by drug priming, discrete cues, discriminative cues, contextual cues, or stressors, following drug self-administration training (typically lever pressing or nose poking for drug infusions) and subsequent extinction of the drug-reinforced responding (Shaham et al, 2003; Stewart and de Wit, 1987). The first study using the reinstatement model was published in 1995 using male and female rats that were trained to self-administer heroin for 12 days for 12 h/day. In this study, acute exposure to 10 min of intermittent footshock stress immediately before the test sessions reinstated heroin seeking after 1–2 weeks of extinction training and after an additional 4- to 6-week drug-free period (Shaham and Stewart, 1995) (Figure 1).

Figure 1
figure 1

Effect of intermittent footshock stress on reinstatement across drug classes. Data are mean±SEM non-reinforced active lever presses after exposure to intermittent footshock or no shock immediately before the reinstatement tests in rats with a history of heroin, cocaine, alcohol, nicotine, or methamphetamine self-administration. The session duration was 3 h for heroin and cocaine, 1 h for alcohol and nicotine, and 2 h for methamphetamine. The intermittent footshock (0.5 s ON; mean inter-shock-interval 40 s) duration was 10 min for heroin and cocaine, 5 min for alcohol and methamphetamine, and 15 min for nicotine. Data were redrawn from Buczek et al (1999), Erb et al (1996), Le et al (1998), Shaham and Stewart (1995), and Shepard et al (2004). *Different from the no stress condition, p<0.05.

PowerPoint slide

Full size image

In this review, we summarize results from studies published during the last 20 years in which investigators have used different variations of the stress-induced reinstatement model to study behavioral, neuropharmacological, and neuroanatomical mechanisms of stress-induced reinstatement of drug seeking. We also summarize results from human studies that have determined the degree to which the results from the preclinical studies ‘translate’ to the human condition. The material in our review overlaps to some degree with previous reviews (Bossert et al, 2005, 2013; Lu et al, 2003; Mantsch et al, 2014; Shaham et al, 2003; Shalev et al, 2010, 2002; Sinha et al, 2011b), but provides the only comprehensive review uniquely dedicated to the phenomenon of stress-induced reinstatement across drug classes since an early review in 2000 (Shaham et al, 2000a). Since 1995, over 200 papers have been published on this phenomenon (PubMed, Figure 2). Due to space limitations, we do not cover all published papers on this topic in our review.

Figure 2
figure 2

Publications on stress-induced reinstatement. Number of preclinical publications per 5-year period on stress-induced reinstatement of drug self-administration and CPP since 1995.

PowerPoint slide

Full size image

Generality of Stress-Induced Reinstatement

Generality to Other Drug and Non-Drug Rewards

Early studies showed that the effect of intermittent footshock on reinstatement of heroin seeking generalized to male rats with a history of cocaine (Erb et al, 1996), alcohol (Le et al, 1998), nicotine (Buczek et al, 1999), and methamphetamine (Shepard et al, 2004) self-administration (Figure 1). In contrast, several early studies showed that exposure to an intermittent footshock stressor that reinstates drug seeking (Shaham et al, 2000a) does not reinstate food reward seeking in male rats trained to self-administer food pellets (Ahmed and Koob, 1997), sucrose pellets (Le et al, 1998; Mantsch and Goeders, 1999), or sucrose solution (Buczek et al, 1999). However, more recently, Chen et al (2014b) showed that intermittent footshock modestly reinstates palatable food (high carbohydrate pellets) seeking in female rats with a history of ‘cafeteria diet’ consumption during adolescence. There is also evidence that intermittent footshock stress reinstates operant responding previously reinforced by brain stimulation reward (Shalev et al, 2000); this finding is in agreement with results from an early study (Deutch and Howarth, 1962). In the studies described above, and in others reviewed below, the intermittent footshock stressor and other stressors (with the exception of 1 day food deprivation) were administered acutely just before the reinstatement test sessions.

The reasons for the ‘selective’ effect of intermittent footshock stress on reinstatement of drug vs food reward are unknown. As previously discussed, one possibility is that while certain physiological effects of intermittent footshock promote reward seeking, other effects of footshock induce physiological states that inhibit food intake and seeking (Shaham et al, 2000a). In this regard, one potential mechanism is hypothalamic corticotropin releasing factor (CRF), which is released after exposure to footshock and other stressors (Bale and Vale, 2004; Sawchenko et al, 2000), and causes inhibition of food intake (Glowa et al, 1992; Morley and Levine, 1982; Sekino et al, 2004). As discussed below (section ‘Brain areas and circuits of stress-induced reinstatement’), this classical effect of CRF on the hypothalamic-pituitary adrenal (HPA) axis (Dallman et al, 1995) contrasts with CRF actions on extrahypothalamic systems, which promote stress-induced reinstatement of drug seeking (Erb et al, 2001b). It is also possible that the ‘putative’ selective effect of intermittent footshock on reinstatement of drug vs food seeking is due to the fact that the above negative studies with food were performed under experimental conditions in which footshock stress-induced reinstatement is often not observed with drugs like cocaine (eg, limited access daily sessions) (see section ‘Methodological considerations’).

Generality to Other Behavioral Procedures

In the studies described in section ‘Generality to other drug and non-drug rewards’, footshock stress-induced reinstatement was determined using the operant version of the extinction-reinstatement model (Shaham et al, 2003). The effect of this stressor was also demonstrated in the conditioned place preference (CPP) variation of the model using cocaine (Wang et al, 2000). In the CPP reinstatement model, laboratory animals are trained to associate a distinct environment with drug injections, undergo extinction training during which they are exposed to the drug environment in the absence of drug, and then tested for reinstatement of drug CPP after stress exposure (Sanchez and Sorg, 2001; Wang et al, 2000). There is also evidence that intermittent footshock can re-establish morphine or cocaine CPP under experimental conditions in which the drug CPP is no longer observed after the passage of time without extinction training (Lu et al, 2000, 2001). Reinstatement of drug seeking after extinction has also been assessed in the runway operant model (Ettenberg, 2009; Ettenberg et al, 1996). To date, however, we are not aware of published studies on the effect of intermittent footshock or other stressors on reinstatement of drug seeking in this model.

Generality to Other Stressors

Over the years, investigators have determined whether stressors other than intermittent footshock reinstate operant drug seeking or drug CPP. We discuss these results separately for the self-administration and CPP variations of the reinstatement model. Table 1 provides a summary of the text in section ‘Generality to other stressors’.

Table 1 Effects of Different Stressors on Reinstatement of Operant Drug Seeking and Conditioned Place Preference (CPP)
Full size table

Stress-induced reinstatement of drug self-administration

Acute 1-day food deprivation reinstates heroin and cocaine seeking (Shalev et al, 2000, 2003); these results are in agreement with an earlier finding that food restriction (about 30–40% of daily food ration) reinstates cocaine seeking in rats with a history of exposure to this stressor during the training phase (Carroll, 1985). Acute 1-day food deprivation also reliably reinstates cocaine seeking in mice (Highfield et al, 2002). Exposure to cues previously paired with social defeat stress, but not acute exposure to social defeat, modestly reinstates alcohol seeking (Funk et al, 2005). Conrad et al (2010) showed that acute exposure (4–5 min) to cold swim stress 24 h before the reinstatement tests reinstates cocaine seeking. Most recently, Twining et al (2015) showed that acute intraoral delivery of an aversive bitter quinine solution reinstates cocaine seeking.

Pharmacological induction of a stress-like state by ventricular CRF injections reinstates heroin (Shaham et al, 1997), alcohol (Le et al, 2002), and cocaine (Erb et al, 2006) seeking. Similarly, systemic injections of the alpha-2 adrenoceptor antagonist yohimbine, which induces stress- and anxiety-like responses in humans and laboratory animals (Bremner et al, 1996a and b), reinstate methamphetamine (Shepard et al, 2004), alcohol (Le et al, 2005), cocaine (Feltenstein and See, 2006), and nicotine (Feltenstein et al, 2012) seeking in rats, and cocaine seeking in monkeys (Lee et al, 2004). However, the relevance of the potent effect of yohimbine on reinstatement of drug seeking to the general phenomenon of stress-induced reinstatement has been questioned in a recent study (Chen et al, 2014c) (see Box 2).

Another class of putative pharmacological stressors is kappa opioid receptor (KOR) agonists, which induce stress-like states in humans and laboratory animals (Bruchas et al, 2010; Van't Veer and Carlezon, 2013). In rats, systemic injections of the KOR agonist U50,488 reinstate nicotine and alcohol seeking (Funk et al, 2014; Grella et al, 2014). In monkeys, systemic injections of KOR agonists (spiradoline and enadoline) reinstate cocaine seeking (Valdez et al, 2007). Another potential demonstration of the effect of pharmacological stressors on reinstatement is the finding that metyrapone potently reinstates heroin seeking (Shaham et al, 1997). Metyrapone is a corticosterone synthesis inhibitor (Jenkins et al, 1958) that has been used in numerous endocrinology studies (Haynes, 1990). However, the drug also activates extrahypothalamic brain areas, including the central amygdala (CeA) nucleus (Rotllant et al, 2002), which has an important role in stress-induced reinstatement (section ‘Brain areas and circuits of stress-induced reinstatement’). Finally, another class of pharmacological stressors, GABAa antagonists or inverse receptor agonists like pentylenetetrazole or FG7142, had no effect on reinstatement of cocaine (Waters and See, 2011), heroin, or alcohol (Y Shaham, AD Le, unpublished observations) seeking.

One study showed that a cue previously paired with footshock (a conditioned fear manipulation) reinstates alcohol seeking (Liu and Weiss, 2003). However, earlier studies reported that conditioned fear exposure has no effect on reinstatement of heroin or cocaine seeking (Shaham et al, 2000a). A possible reason for the negative results is that the predominant effect of cues paired with shock is freezing (LeDoux, 2000), a behavior that is incompatible with lever-pressing behavior. Exposure to restraint stress in a novel context had no effect on reinstatement of heroin seeking (Shalev et al, 2000). As intermittent footshock stress is effective when given in the drug self-administration context, but has no effect on reinstatement of heroin seeking when given in a novel context, a potential account for the negative results with restraint stress is that the phenomenon of stress-induced reinstatement is context-specific (Shalev et al, 2000). Finally, the effect of footshock on reinstatement of heroin seeking is mimicked by reversible inactivation of the medial septum (Highfield et al, 2000). Previous studies have shown that septal lesions cause stress-like physiological and psychological responses (Gray, 1987; Holdstock, 1967).

Stress-induced reinstatement of CPP

Exposure to several stressors induces reinstatement of drug CPP after extinction in rats and mice. Swim stress reinstates cocaine CPP in mice (Carey et al, 2007; Kreibich and Blendy, 2004; Redila and Chavkin, 2008) and morphine CPP in rats (Li et al, 2013; Ma et al, 2007). Restraint stress reinstates cocaine CPP in rats (Sanchez et al, 2003) and morphine CPP in mice (Ribeiro Do Couto et al, 2006). Tail pinch and social defeat reinstate morphine CPP in mice (Ribeiro Do Couto et al, 2006). A conditioned fear manipulation reinstates cocaine CPP in rats (Sanchez and Sorg, 2001). Additionally, as in the operant reinstatement model, systemic injections of the putative pharmacological stressors U50,488 (a KOR agonist) (Redila and Chavkin, 2008) and yohimbine (Mantsch et al, 2010) reinstate cocaine CPP in mice. However, 1-day food deprivation, which reinstates operant heroin seeking in rats (Shalev et al, 2000), has no effect on reinstatement of morphine CPP (Ma et al, 2007).

Methodological Considerations

In the sections above, we summarize results from studies in which stressors reinstate drug seeking and CPP. However, particularly in the case of the intermittent footshock stressor, whether or not the stressor reinstates operant drug seeking is dependent on several critical methodological factors. One factor is the context of footshock exposure: under identical experimental conditions, intermittent footshock reinstates heroin seeking when it is administered in the drug self-administration context but not in a novel context (Shalev et al, 2000). These findings are in agreement with results from conditioned fear studies where a footshock previously paired with a conditioned stimulus (CS) reinstates the response to the CS after extinction only when rats are exposed to the stressor in the training context, but not in a novel context (Bouton and King, 1983; Bouton and Swartzentruber, 1991).

A second factor is the duration of the withdrawal period. Shalev et al (2001) used a variation of the reinstatement model in which extinction of operant responding and subsequent footshock-induced reinstatement is determined in the same daily session (Shalev et al, 2002) to demonstrate the time-dependent changes in this reinstatement after withdrawal from heroin. The stressor reinstated heroin seeking after 6, 12, 25, and 66 withdrawal days but not after 1 day. In this study, the magnitude of stress-induced reinstatement was maximal after 6 or 12 days.

A third factor is the amount of drug intake during drug self-administration training. For both heroin (Ahmed et al, 2000) and cocaine (Mantsch et al, 2008), intermittent footshock-induced reinstatement of drug seeking is higher in magnitude and more robust/reliable in rats with a history of extended daily drug access (6–11 h/day) than in rats with a history of limited daily drug access (1–2 h/day).

A fourth factor is the availability of discrete infusion cues during all phases of the experiment (training, extinction, and reinstatement). Shelton and Beardsley (2005) showed that under identical experimental conditions, intermittent footshock reinstates cocaine seeking in the presence but not in the absence of a discrete infusion cue previously paired with cocaine infusions during training. This finding is in agreement with the notion that a potential mechanism for footshock-induced reinstatement is that the stressor reinstates the incentive motivational effects of drug cues (Shaham and Stewart, 1995). Additional support for this notion is that intermittent footshock potentiates cue-induced reinstatement of cocaine and alcohol seeking (Buffalari and See, 2009; Liu and Weiss, 2002a). In the cue-induced reinstatement procedure, an operant response reinforced by a discrete drug cue is determined after drug self-administration training in the presence of the cue and extinction training in its absence (Meil and See, 1996).

A fifth factor is baseline stress levels before footshock exposure during the reinstatement tests. In general, while this variable has not been experimentally manipulated like the other factors described above, anecdotal evidence suggests that footshock-induced reinstatement is not readily observed when the baseline stress level is high, because under these conditions the predominant effect of the stressor during the reinstatement test is prolonged freezing. This idea has led to the development of a modified stress-induced reinstatement procedure in which rats are first given several days of extinction training, and during the subsequent reinstatement test they are given several hours of extinction training before footshock exposure (Erb et al, 1996; Kupferschmidt et al, 2011). Presumably, intermittent footshock reliably reinstates drug seeking in this procedure because the rat’s basal stress level is relatively low several hours after being exposed to non-specific stressors (eg, handling and transferred from the animal facility to the test chambers).

Summary

The phenomenon of intermittent footshock stress-induced reinstatement of drug seeking, originally reported with heroin-trained rats, generalizes to other drugs of abuse and is also observed in the CPP variation of the reinstatement model. This phenomenon also generalizes to certain environmental and pharmacological stressors but not others. Additionally, it appears that the number of stressors that are effective in the CPP variation of the reinstatement model is more diverse than in the operant variation of the model (Table 1). Furthermore, several methodological factors are critical for a reliable effect of intermittent footshock stress on reinstatement of drug seeking. These factors include the context of footshock exposure, high drug intake during training, the withdrawal period duration, the presence of the extinguished discrete drug cues during testing, and basal (baseline) stress levels before footshock exposure in the reinstatement tests.

Finally, in all the published studies on stress-induced reinstatement, the reinstating stressors have been both uncontrollable and unpredictable. Thus, a question for future research is whether these features of stressors are critical for their effect on reinstatement of drug seeking. Based on previous studies investigating the inhibitory effect of both stress controllability and predictability on physiological and psychological stress responses in laboratory animals (Maier, 1984; Seligman and Maier, 1967) and humans (Cohen et al, 1986; Glass and Singer, 1972), we predict that controllable or predictable stressors will have a weak effect on reinstatement of drug seeking. Additionally, in the studies reviewed above, the stressors were administered acutely before the reinstatement test session. Thus, another question for future research is whether prior exposure to stress, which often leads to subsequent sensitization of stress responses (Antelman et al, 2000; Kvetnansky et al, 2009) and cross sensitization of the behavioral and neurochemical responses to drugs (Kalivas and Stewart, 1991), would potentiate the stressors’ effect on reinstatement of drug seeking.

Brain Areas and Circuits of Stress-Induced Reinstatement

Since 1995 many studies have determined the effects of systemic and intracranial injections of pharmacological agents targeting different receptor classes on stress-induced reinstatement of drug seeking or CPP. Due to space limitations, we summarize most of the systemic pharmacology data in Tables 2 and 3 but do not discuss them in detail in our review. The brain circuitry data that we discuss below are primarily based on the use of intermittent footshock in the self-administration- and CPP-based reinstatement models with different drug classes in rats, and the use of swim stress in the CPP-based procedure with cocaine in mice. During the last several years, many investigators have also determined the neuropharmacological mechanisms underlying yohimbine-induced reinstatement of drug seeking and CPP (Bossert et al, 2013; Mantsch et al, 2014; See and Waters, 2010). However, as discussed in Box 2, yohimbine’s effects on operant responding during reinstatement testing are likely independent of the history of contingent drug self-administration and may not be related to the commonly assumed stress-like effects of yohimbine. Therefore, we only mention results from yohimbine-related studies when a given neuropharmacological manipulation affected both intermittent footshock- and yohimbine-induced reinstatement.

Table 2 Effect of Systemic Injections of Different Pharmacological Agents on Stress-Induced Reinstatement of Operant Drug and Food Seeking
Full size table
Table 3 Effect of Systemic Injections of Different Pharmacological Agents on Stress-Induced Reinstatement of Conditioned Place Preference
Full size table

Bed Nucleus of Stria Terminalis and CeA

The bed nucleus of stria terminalis (BNST) and the CeA are known to have important roles in stress responses (Davis et al, 2010; LeDoux, 2000) that involve activation of noradrenaline and CRF systems (Davis et al, 1997; Flavin and Winder, 2013; McGaugh, 2004; Silberman and Winder, 2013). Studies using systemic or ventricular drug injections have shown that CRF receptor antagonists and alpha-2 adrenoceptor agonists like clonidine (which inhibit noradrenaline cell firing and release) decrease stress-induced reinstatement of drug seeking and drug CPP (Mantsch et al, 2014; Shaham et al, 2000a; Shalev et al, 2010) (Tables 2 and 3). Conversely, ventricular injections of CRF or noradrenaline mimic the effect of intermittent footshock stress on reinstatement (Brown et al, 2009; Erb et al, 2006; Le et al, 2000; Shaham et al, 1997). Studies using endocrinology methods (adrenalectomy, corticosterone replacement) have also shown that CRF’s role in footshock-induced reinstatement does not involve stress-induced activation of the HPA axis, indicating a role of extrahypothalamic CRF systems (Erb et al, 2001a; Lu et al, 2003). Based on these results, investigators have studied the involvement of CRF and noradrenaline in BNST and CeA in stress-induced reinstatement and CPP. We describe these studies below. (Note: in Box 3 we discuss in more detail the role of corticosterone and the HPA axis in stress-induced reinstatement.)

The CRF projections from CeA to BNST

Erb and Stewart (1999) showed that injections of the CRF receptor antagonist D-Phe CRF12–41 into the BNST (ventral region) but not CeA decrease footshock-induced reinstatement of cocaine seeking. They also showed that CRF injections into the BNST but not CeA reinstate cocaine seeking; this finding was confirmed by Vranjkovic et al (2014). In agreement with these findings, Wang et al (2006) showed that BNST but not CeA injections of the selective CRF1 receptor antagonist CP-154,526 (Schulz et al, 1996) decrease footshock-induced reinstatement of morphine CPP. However, while blockade or activation of CeA CRF receptors had no effect on footshock-induced reinstatement, Erb et al (2001a) reported data suggesting that a CRF projection from CeA to BNST (Sakanaka et al, 1986) has a role in this reinstatement. They used an asymmetric anatomical disconnection method (Gold, 1966) and showed that injections of tetrototoxin (TTX, a sodium channel blocker that inhibits neuronal activity) into the CeA in one hemisphere and D-Phe CRF12–41 into the BNST in the contralateral hemisphere decrease footshock-induced reinstatement of cocaine seeking.

Taken together, these results suggest that stress-induced activation of CRF projections from CeA to BNST and subsequent activation of CRF receptors (presumably the CRF1 subtype) in the BNST are critical for stress-induced reinstatement of drug seeking and CPP. However, the circuitry (projection) data are based on the use of TTX, which also inhibits fiber of passage and non-CRF neurons that project to BNST (Cassell et al, 1999). Therefore, it is likely that within the BNST, CRF-related mechanisms that are independent of the CeA to BNST CRF projection also have a role in stress-induced reinstatement.

Noradrenaline in CeA and BNST and interaction between noradrenaline and CRF in BNST

Leri et al (2002) showed that injections of a mixture of beta-1 and beta-2 adrenoceptor antagonists (betaxolol+ICI-118,551) into both the BNST (the injections were in both dorsal and ventral BNST) and the CeA decrease footshock-induced reinstatement of cocaine seeking. Subsequently, Vranjkovic et al (2014) showed that within the ventral BNST, this effect is mediated by the beta-2 adrenoceptor: local ICI-118,551 but not betaxolol injections decrease stress-induced reinstatement. Additionally, they showed that ventral BNST injections of the beta-2 adrenoceptor agonist clenbuterol, but not the beta-1 adrenoceptor agonist dobutamine, mimic the effect of footshock on reinstatement. Additional evidence for a role of BNST and CeA noradrenaline in stress-induced reinstatement is that BNST and CeA injections of clonidine decrease footshock-induced reinstatement of morphine CPP (Wang et al, 2001) and footshock-induced reinstatement of nicotine seeking (Yamada and Bruijnzeel, 2011). These data, however, are difficult to interpret in the absence of additional anatomical and pharmacological controls, because clonidine is highly lipophilic and it can readily diffuse to the nearby areas, or in the case of the BNST, to the ventricles. Finally, based on the earlier results of Leri et al (2002), an unexpected finding of Yamada and Bruijnzeel (2011) is that CeA injections of the non-selective beta adrenoceptor antagonist propranolol have no effect on footshock-induced reinstatement of nicotine seeking.

Ascending noradrenaline systems originate from two clusters of neurons, the locus coeruleus (LC, A6) and the lateral tegmental nuclei (A1, A2 (located in nucleus tractus solitarii, NTS), A4); these neurons project to many brain areas via the dorsal and ventral noradrenergic bundles, respectively (Kvetnansky et al, 2009; Moore and Bloom, 1979). The lateral tegmental nuclei, which provide the main source of noradrenaline to BNST and CeA (Aston-Jones et al, 1999; Fritschy and Grzanna, 1991), but not the LC, are critical for stress-induced reinstatement of drug seeking and CPP. 6-OHDA lesions of the ventral noradrenergic bundle decrease footshock-induced reinstatement of heroin seeking (Shaham et al, 2000b) and morphine CPP (Wang et al, 2001). In contrast, injections of clonidine or its charged analog ST-91 into the LC have no effect on this reinstatement (Shaham et al, 2000b; Wang et al, 2001).

Finally, there is evidence that an interaction between CRF and noradrenaline in the BNST is critical for footshock-induced reinstatement of cocaine seeking. In an initial study, Brown et al (2009) showed that reinstatement of cocaine seeking induced by ventricular injections of noradrenaline is decreased by pretreatment with the CRF receptor antagonist D-Phe CRF12–41. In contrast, reinstatement induced by ventricular injections of CRF is not blocked by the alpha-2 adrenoceptor agonist clonidine. These data suggest that the role of CRF in stress-induced reinstatement is downstream from stress-induced activation of noradrenaline systems. The likely brain site for this noradrenaline-CRF interaction is the BNST. Vranjkovic et al (2014) showed that reinstatement of cocaine seeking induced by ventral BNST injections of the beta-2 adrenoceptor agonist clenbuterol is blocked by local injections of the CRF1 receptor antagonist antalarmin (Bornstein et al, 1998).

Taken together, stress-induced activation of lateral tegmental noradrenergic nuclei and subsequent actions of noradrenaline in both the BNST and CeA are critical for stress-induced reinstatement of drug seeking and CPP. Additionally, within the ventral BNST, stress-induced activation of CRF and subsequent stress-induced reinstatement are downstream from stress-induced activation of lateral tegmental noradrenergic nuclei that project to this brain area.

Other evidence for a role of BNST and CeA in stress-induced reinstatement

Several other lines of evidence also indicate critical roles for the BNST and CeA in stress-induced reinstatement of drug seeking and CPP. Reversible inactivation of ventral BNST or CeA with GABAa and GABAb receptor agonists (muscimol+baclofen) decreases footshock-induced reinstatement of cocaine seeking (McFarland et al, 2004). Additionally, TTX injections into either the BNST or CeA decrease footshock-induced reinstatement of heroin seeking (Shaham et al, 2000a) and this reinstatement is associated with increased CRF mRNA in both brain areas (Shalev et al, 2001). However, both sets of data should be interpreted with caution. As mentioned above, a methodological issue with TTX inactivation data is that the drug inactivates both cell bodies and fibers of passage. Additionally, the relevance of the CRF mRNA expression data to CRF’s role in stress-induced reinstatement is unclear, because within the BNST footshock selectively increases expression in dorsal but not in ventral BNST, while the pharmacological data described above are primarily based on ventral BNST injections.

Other relevant evidence is that systemic injections of the beta-2 adrenoceptor antagonist ICI-118,551 decrease swim stress-induced reinstatement of cocaine CPP and prevent increases in CRF mRNA in BNST (but not CeA) in mice (McReynolds et al, 2014). However, under somewhat different experimental conditions, exposure to swim stress increases CRF mRNA in CeA in cocaine-experienced mice (Cleck et al, 2008). In agreement with the latter findings, Nawata et al (2012) showed that a CRF receptor antagonist (alpha-helical-CRF9-41) decreases footshock-induced reinstatement of methamphetamine seeking in rats; additionally, using enzyme-linked immunosorbent assay (ELISA) they showed that CRF protein levels in CeA are increased after 10 but not 5 days of withdrawal from methamphetamine. The relevance of the ELISA data to the pharmacological data with the CRF receptor antagonist is unknown.

Finally, studies using markers of neuronal activity such as Fos and pCREB (a downstream target of Fos activation) showed that swim stress-induced reinstatement of cocaine CPP in mice is associated with increased CREB activity (phosphorylation) in amygdala but not in BNST (Kreibich and Blendy, 2004). In a subsequent study, however, Briand et al (2010) showed that swim stress increases Fos expression in both ventral and dorsal BNST and that reversible inactivation of the BNST with lidocaine decreases swim stress-induced reinstatement of cocaine CPP. Additionally, Buffalari and See (2011) showed that reversible inactivation with muscimol+baclofen of the BNST decreases yohimbine-induced reinstatement of cocaine seeking and yohimbine-induced potentiation of cue-induced cocaine seeking.

Summary

The studies reviewed above indicate that the BNST (primarily the ventral BNST) and CeA have important roles in stress-induced reinstatement of drug seeking across drug classes and are also involved in stress-induced reinstatement of morphine and cocaine CPP. These roles include stress-induced activation of the lateral tegmental nuclei that provide the main noradrenergic innervation to BNST and CeA, and potentially the CRF projection from CeA to BNST. Within the BNST, stress-induced CRF activation and subsequent stress-induced reinstatement are likely downstream from stress-induced activation of lateral tegmental noradrenaline neurons.

Ventral Tegmental Area, Media Prefrontal Cortex, Orbital Frontal Cortex, and Nucleus Accumbens

Neuronal activity in the mesocorticolimbic dopamine system, which comprises cell bodies in ventral tegmental area (VTA) that project to medial prefrontal cortex (mPFC), orbital frontal cortex (OFC), nucleus accumbens (NAc), amygdala, BNST, and other brain areas, has an important role in reinstatement of drug seeking induced by discrete cues (See, 2005), discriminative cues (Weiss, 2005), contextual cues (Marchant et al, 2014), and drug priming (Kalivas and McFarland, 2003; Pierce and Kumaresan, 2006; Self, 2004). Early studies have shown that different stressors strongly and preferentially activate the dopaminergic projection from VTA to mPFC (the mesocortical dopamine system) (Bannon and Roth, 1983; Deutch and Roth, 1990; Thierry et al, 1976). Subsequent studies, however, showed that stressors can also activate the dopaminergic projection from VTA to NAc (a component of the mesolimbic dopamine system) (Cabib and Imperato, 1996; Kalivas and Stewart, 1991).

Based on these findings, the first two studies on intermittent footshock-induced reinstatement of drug seeking explored the role of mesolimbic dopamine. In the first study, Shaham and Stewart (1995) showed that footshock-induced reinstatement of heroin seeking is associated with increased dopamine release in NAc. In the second study, they showed that systemic injections of the mixed dopamine receptor antagonist flupenthixol decrease this reinstatement (Shaham and Stewart, 1996). In a subsequent anatomical mapping study, McFarland et al (2004) showed that reversible inactivation with muscimol+baclofen of VTA, dorsal mPFC, NAc core and shell, and ventral pallidum (a downstream output of NAc via the ‘indirect’ pathway) decreases footshock-induced reinstatement of cocaine seeking. Below, we describe subsequent mechanistic studies on the role of different brain areas of the mesocorticolimbic system in stress-induced reinstatement of drug seeking and CPP.

Glutamate, CRF, and Kappa/dynorphin in VTA

Glutamate input to VTA arrives from several brain regions (Geisler et al, 2007; Omelchenko and Sesack, 2007) and controls the activity of VTA dopamine neurons under basal conditions (Westerink et al, 1996; White et al, 1995) and after stress exposure (Enrico et al, 1998; Takahata and Moghaddam, 1998). Based on these studies, Wise and colleagues studied the role of VTA glutamate in stress-induced reinstatement of cocaine and heroin seeking. They showed that footshock-induced reinstatement of both heroin and cocaine seeking is associated with increased VTA glutamate release and that this reinstatement is decreased by local blockade of ionotropic glutamate receptors with kynurenic acid (Wang et al, 2005, 2012).

VTA dopamine and glutamate transmission is also controlled by local CRF signaling (Kalivas et al, 1987; Ungless et al, 2003) from unspecified sources and, as discussed above, CRF has a critical role in stress-induced reinstatement of drug seeking and CPP. Based on these findings, Wise and colleagues studied the role of VTA CRF in stress-induced reinstatement of cocaine seeking. In the first study, they showed that VTA perfusion (via a microdialysis probe) of alpha-helical CRF9–41 decreases footshock-induced reinstatement and that local CRF perfusion reinstates cocaine seeking (Wang et al, 2005); see also Wang et al (2009) for a replication of the VTA CRF effect on reinstatement. Next, they showed that footshock-induced reinstatement of cocaine seeking is decreased by VTA perfusion of a selective CRF2 receptor antagonist or an inhibitor of the CRF binding protein but not by a selective CRF1 receptor antagonist. Moreover, they showed that VTA perfusion of CRF or CRF2 receptor agonists with strong affinity for CRF-BP (binding protein) causes reinstatement of cocaine seeking, whereas CRF receptor agonists that do not bind CRF-BP are ineffective (Wang et al, 2007).

These are surprising findings from two perspectives. The first is that previous studies discussed above (see also Tables 2 and 3) implicate CRF1 but not CRF2 receptors in stress-induced reinstatement. The second is that anatomical studies indicate that CRF1 receptors are the predominant postsynaptic CRF receptors in VTA and that CRF2 receptor expression is not detected by traditional in situ hybridization methods (Van Pett et al, 2000). Another issue to consider in studies using putative selective CRF2 antagonists is that their CRF2/CRF1 ratio selectivity is moderate, and some commonly used CRF2 antagonists like antisauvagine-30 have behavioral effects in CRF2 knockout mice, indicating actions on other receptors (Zorrilla et al, 2013b).

More recent studies demonstrated a role of CRF1 receptors in stress-induced reinstatement of cocaine seeking. Blacktop et al (2011) showed that VTA injections of the CRF1 receptor antagonists antalarmin or CP-376395 but not the CRF2 receptor antagonists astressin-2B or antisauvagine-30 decrease footshock-induced reinstatement of cocaine seeking in rats. Additionally, VTA injections of the CRF1 receptor agonist cortagine but not the CRF2 receptor agonist rat urocortin II mimic the effect of footshock stress on reinstatement. Chen et al (2014a) showed that local viral-mediated knockdown of VTA CRF1 receptors decreases food deprivation stress-induced reinstatement of cocaine seeking in mice. Vranjkovic et al (2014) also showed that injections of the beta-2 adrenoceptor antagonist ICI-188,551 into the BNST in one hemisphere and the CRF1 receptor antagonist antalarmin into the VTA of the contralateral hemisphere decrease footshock-induced reinstatement of cocaine seeking. These data suggest a potential role of the CRF projection from BNST to VTA (Rodaros et al, 2007), which is activated by noradrenergic projections to BNST (Aston-Jones et al, 1999), in stress-induced reinstatement of cocaine seeking. In this regard, the BNST projection to VTA is also activated during swim stress-induced reinstatement of cocaine CPP in mice, as assessed by double labeling of the activity marker Fos with the retrograde tracer fluorogold gold (injected in VTA) in ventral BNST (Briand et al, 2010).

Two recent studies by Kauer, Pierce and colleagues demonstrate a role of kappa/dynorphin signaling in VTA in cold swim stress-induced reinstatement of cocaine seeking in rats (Graziane et al, 2013; Polter et al, 2014). In this stress reinstatement procedure, short (4–5 min) exposure to cold swim reinstates cocaine seeking 24 h later (Conrad et al, 2010). These studies were inspired by earlier studies on the regulation of VTA dopamine transmission by local kappa/dynorphin signaling (Ford et al, 2007; Margolis et al, 2003) and the findings that systemic injections of KOR antagonists decrease stress-induced drug seeking in rats and stress-induced reinstatement of drug CPP in mice (Tables 2 and 3) (Beardsley et al, 2005; Redila and Chavkin, 2008). In the first study, Graziane et al (2013) showed that systemic injections of the KOR antagonist Nor-BNI rescue swim stress-induced abolishment of long-term potentiation (LTP) at GABAergic synapses on VTA dopaminergic neurons. Next, they showed that VTA injections of Nor-BNI decrease stress-induced reinstatement of cocaine seeking. In the second study, Polter et al (2014) showed that acute exposure to swim stress blocks normal LTP at GABAergic synapses on VTA dopaminergic neurons for up to 5 days, an effect dependent on KOR, and that surprisingly, systemic injections of Nor-BNI several hours after swim stress exposure also decrease the stressor’s effect on reinstatement 24 h later.

Summary

The studies reviewed above indicate that footshock stress-induced activation of CRF and glutamate transmission in VTA is critical for the stressor’s effect on reinstatement of drug seeking. A recent study by Vranjkovic et al (2014) further suggests that glutamatergic (Georges and Aston-Jones, 2002) and/or CRF (Rodaros et al, 2007) projections from BNST to VTA are the source of this activation. There is also evidence that KOR-mediated stress-induced alterations in GABAergic transmission in VTA have a critical role in the effect of cold swim stress on reinstatement of cocaine seeking.

Dopamine and glutamate in NAc, mPFC, and OFC

Studies using intracranial injections of different dopamine receptor antagonists have demonstrated an important role of dopamine transmission in dorsal mPFC (prelimbic area) and OFC in stress-induced reinstatement of cocaine seeking. Capriles et al (2003) showed that dorsal (but not ventral or infralimbic area) mPFC and OFC injections of the D1-family receptor antagonist SCH23390 but not the D2-family receptor antagonist raclopride decrease footshock-induced reinstatement of cocaine seeking. McFarland et al (2004) showed that dorsal mPFC injections of the mixed dopamine receptor antagonist flupenthixol decrease this reinstatement. Sanchez et al (2003) showed that SCH23390 injections into dorsal mPFC decrease restraint stress-induced reinstatement of cocaine CPP. Dorsal mPFC dopamine and neuronal activity in this brain area also have a role in yohimbine-induced reinstatement of palatable food seeking in male and female rats (Calu et al, 2013; Cifani et al, 2012; Nair et al, 2011) and acute food deprivation-induced reinstatement of heroin seeking (Tobin et al, 2013). As VTA dopamine neurons project to mPFC and OFC (Moore and Bloom, 1979), the results from the studies reviewed above indicate that stress-induced activation of the dopaminergic projections from VTA to mPFC (Thierry et al, 1976) and OFC is critical for stress-induced reinstatement. This activation involves D1-family receptors which, in the mPFC, are expressed on pyramidal cells where they promote excitability primarily by increasing NMDA receptor-mediated effects (Lewis and O'Donnell, 2000; Seamans et al, 2001).

McFarland et al (2004) provided additional evidence that the circuit controlling footshock-induced reinstatement of cocaine seeking includes the dopaminergic projection from VTA to dorsal mPFC (prelimbic area), and also provided evidence that the glutamatergic projection from dorsal mPFC to NAc core (Sesack et al, 1989) and potentially the GABAergic projection from NAc shell to VTA (Zahm et al, 2001) (the ‘direct’ striatal pathway) have a role in this reinstatement. They showed that (1) reversible inactivation of dorsal mPFC decreases both footshock-induced reinstatement and glutamate release in NAc core, (2) reversible inactivation of NAc shell and core, and as mentioned above the VTA, decreases this reinstatement, and (3) that reversible inactivation of NAc shell also decreases stress-induced dopamine release in dorsal mPFC. These authors also showed that reversible inactivation of the ventral pallidum, a main output region of the striatal indirect pathway (Gerfen, 1992), decreases footshock-induced reinstatement. This circuit scheme, however, is speculative in the absence of more direct evidence from studies using classical anatomical asymmetrical disconnection procedures (Setlow et al, 2002) or more recent projection-specific inhibition using optogenetic (Stefanik et al, 2013) or chemogenetic (Mahler et al, 2014) methods.

Evidence for a role of NAc dopamine in stress-induced reinstatement is somewhat mixed. Shaham and Stewart (1995) showed that footshock-induced reinstatement of heroin seeking is associated with increased dopamine levels in NAc. In contrast, McFarland et al (2004) showed that footshock-induced reinstatement of cocaine seeking is not associated with increased dopamine release in NAc and that flupenthixol injections into NAc core have no effect on this reinstatement. However, Xi et al (2004) showed that injections of the preferential D3 receptor antagonist SB-277011A into NAc (injections targeted primarily NAc shell) decrease footshock-induced reinstatement of cocaine seeking. Additionally, Tobin et al (2013) showed that SCH23390 injections into NAc shell but not core decrease food deprivation-induced reinstatement of heroin seeking. Thus, one possible account of these potentially discrepant results is that NAc shell but not core dopamine is critical for stress-induced reinstatement of drug seeking.

Finally, a very recent study by Wheeler and colleagues suggests that at least for one stressor, a bitter quinine solution, reinstatement of cocaine seeking is associated with decreased tonic NAc dopamine levels, as assessed by in vivo voltammetry (Twining et al, 2015). They showed that in cocaine-experienced rats, exposure to quinine reinstates cocaine seeking, and that in drug-naïve rats, the same manipulation decreases tonic dopamine levels in NAc. Both of these effects were reversed by blockade of VTA CRF1 receptors. Together, with the previous results reviewed above, the results from this study suggest that both increases and decreases in NAc dopamine can have a role in stress-induced drug seeking.

Summary

D1 receptor-mediated dopamine transmission in dorsal mPFC and the glutamatergic projection from this brain area to NAc core are critical for stress-induced reinstatement of cocaine seeking. This reinstatement also likely involves activation of the NAc GABAergic projections to VTA (the direct pathway) and ventral pallidum (the indirect pathway), as well as increased dopamine transmission in NAc shell. A recent study using intraoral delivery of quinine solution as the stressor suggests that stress-induced reinstatement can also be triggered by decreases in NAc dopamine tone.

Median and Dorsal Raphe Nuclei

Many studies over the last several decades have implicated central serotonin in stress responses (Chaouloff, 1993; Maier, 1984; Maier and Watkins, 2005; Stone, 1983). This literature has led to studies on the role of serotonin in the two main cell body regions of brain serotonergic neurons—median raphe nucleus (MRN) and dorsal raphe nucleus (DRN) (Hale and Lowry, 2011; Vertes, 1991; Vertes et al, 1999)—in stress-induced reinstatement of alcohol seeking in rats and stress-induced reinstatement of cocaine and morphine CPP.

Role of MRN in stress-induced reinstatement of alcohol seeking

Le et al (1999) showed that systemic injections of the serotonin reuptake blocker fluoxetine, which increases brain serotonin levels, decrease footshock-induced reinstatement of alcohol seeking, indicating that serotonin negatively regulates this reinstatement. This negative regulation involves the MRN. Local injections of the selective 5-HT1a autoreceptor agonist 8-OH-DPAT, which decrease serotonin cell firing and release (Mongeau et al, 1997), mimic the effect of intermittent footshock on reinstatement (Le et al, 2002). Additionally, these authors showed that MRN injections of the non-selective CRF receptor antagonist D-Phe CRF12–41 decrease footshock-induced reinstatement of alcohol seeking while MRN injections of very low doses of CRF (10 ng) partially mimic the stressor’s effect on reinstatement. A physiological mechanism that connects the similar effects of 8-OH-DPAT and low-dose CRF injections in MRN on reinstatement is that low CRF doses, like 8-OH-DPAT, decrease serotonin cell firing and release (Kirby et al, 2000; Price et al, 1998). However, to date, this effect has only been reported for DRN but not for MRN neurons. Finally, Le et al (2008) showed that reversible inactivation of the MRN with the GABAa agonist muscimol reinstates alcohol seeking. However, the relevance of this observation to serotonin’s role in stress-induced reinstatement is questionable because this effect is not reversed by lesions of serotonin-containing neurons in the brain.

Role of DRN in stress-induced reinstatement of morphine and cocaine CPP

Results from CPP reinstatement studies from the Kirby and Chavkin laboratories implicate the DRN serotonergic cell body region in stress-induced reinstatement. Staub et al (2012) showed that swim stress-induced reinstatement of morphine CPP is associated with increased GABAergic postsynaptic neuronal activity that leads to inhibition of DRN serotonin neurons, and that this inhibitory effect involves local CRF signaling. In a follow-up study, Li et al (2013) showed that increasing GABAergic inhibitory transmission by DRN muscimol injections reinstates morphine CPP while inhibition of GABAergic neurons by bicuculline (a GABAa receptor antagonist) decreases swim stress-induced reinstatement.

There is also evidence for a role of kappa/dynorphin signaling in DRN in stress-induced reinstatement of cocaine CPP in mice. Land et al (2009) showed that DRN injections of the KOR antagonist Nor-BNI decrease social stress-induced reinstatement of cocaine CPP in mice. Their results from experiments examining the role of kappa/dynorphin in DRN in aversive behavior also suggest that the serotonergic projection from DRN to NAc likely has a role in the local inhibitory effect of Nor-BNI in DRN on stress-induced reinstatement of cocaine CPP; however, this possibility was not empirically tested. In a subsequent study, the Chavkin laboratory showed that KOR-mediated social stress-induced activation of p38 mitogen-activated protein kinase (MAPK) in DRN serotonergic neurons has a critical role in stress-induced reinstatement of cocaine CPP (Bruchas et al, 2011). In this study, the authors first showed that exposure to social defeat stress, which induces reinstatement of cocaine seeking, activates p38MAPK in DRN, and that this effect is reversed by Nor-BNI and is not observed in KOR knockout mice. They then showed that viral-mediated selective deletion of p38MAPK in DRN serotonin neurons decreases social stress-induced reinstatement of cocaine CPP.

Summary

The results of the studies reviewed above indicate an important role for an interaction between CRF and serotonin in MRN in footshock stress-induced reinstatement of alcohol seeking. There is also evidence for a role of DRN serotonergic neurons, which involves both CRF and kappa/dynorphin. A question for future research is which serotonergic projection areas contribute to the distinct roles of MRN and DRN in stress-induced reinstatement in the operant model and the CPP model (Figure 3).

Figure 3
figure 3

The circuitry of stress-induced reinstatement of drug seeking. The figure depicts brain regions, neuronal projections, and neurotransmitters implicated in stress-induced reinstatement of drug seeking and stress-induced reinstatement of drug CPP (see text for details). Filled lines represent established projections and dashed lines represent putative projections. Abbreviations: CeA, central nucleus of the amygdala; DRN, dorsal raphe nucleus, MRN, median raphe nucleus; mPFC, medial prefrontal cortex; NAc: nucleus accumbens; OFC, orbitofrontal cortex; vBNST, ventral BNST; VP, ventral pallidum; VTA, ventral tegmental area.

PowerPoint slide

Full size image

Translational Human Research

In this section, we describe results from several human laboratory studies and two double-blind clinical trials in which pharmacological agents that decrease stress-induced reinstatement of drug seeking were tested for their effects on drug craving and relapse in humans. These include the alpha-2 adrenoceptor agonists, clonidine, guanfacine, and lofexidine, the alpha-1 adrenoceptor antagonist, prazosin, and the CRF1 receptor antagonist, pexacerfont. Figures 4 and 5 show comparative data on the effect of alpha-2 adrenoceptor agonists and CRF1 receptor antagonists on stress-induced reinstatement of drug seeking in rats and stress-induced craving and time to first lapse in humans.

Figure 4
figure 4

Effects of alpha-2 adrenoceptor agonists on stress-induced reinstatement in rats and stress-induced craving in humans. (a) Rat studies: Data are mean±SEM non-reinforced active lever presses after exposure to intermittent footshock or no shock immediately prior to the reinstatement tests in rats with a history of heroin, cocaine, speedball (a heroin-cocaine combination), or alcohol self-administration. The session duration was 3 h for heroin, cocaine, and speedball, and 1 h for alcohol. The intermittent footshock duration was 15 min for heroin, cocaine, and speedball, and 10 min for alcohol. Clonidine and lofexidine were injected 25–45 min before intermittent footshock exposure. (b) Human studies: Left side: Data are mean±SEM subjective craving responses assessed on an ordinal Likert scale in opiate users (maintained on naltrexone) or cocaine users who were pretreated with lofexidine or clonidine, respectively, or placebo and exposed to a guided imagery stress or no stress (neutral guided imagery) manipulation. Right side: days to first lapse in human subjects that received chronic placebo or clonidine in a clinical trial. Data were redrawn from Erb et al (2000), Highfield et al (2001), Jobes et al (2011), Kowalczyk et al (2014), Le et al (2005), Shaham et al (2000b), and Sinha et al (2007). *Different from the vehicle (0 dose) in the stress condition or different from the placebo condition, p<0.05.

PowerPoint slide

Full size image
Figure 5
figure 5

Effects of CRF1 receptor antagonists on stress-induced reinstatement in rats and stress-induced craving in humans. (a) Rat studies: Data are mean±SEM non-reinforced active lever presses after exposure to intermittent footshock or no shock immediately prior to the reinstatement tests in rats with a history of heroin, cocaine, or alcohol self-administration. The session duration was 3 h for heroin and cocaine, and 1 h for alcohol. The intermittent footshock duration was 15 min for heroin and 10 min for cocaine and alcohol. The CRF1 receptor antagonist CP-154,526 was injected 30 min before intermittent footshock exposure. (b) Human study: Data are mean±SEM subjective craving responses assessed on an ordinal Likert scale in alcohol users who were pretreated with the CRF1 receptor antagonist pexacerfont and exposed to a guided imagery stress or no stress (neutral guided imagery) manipulation. Data were redrawn from Kwako et al (2015), Le et al (2000) and Shaham et al (1998). *Different from the vehicle (0 dose) in the stress condition, p<0.05.

PowerPoint slide

Full size image

Human Laboratory Studies

In 1999, Sinha and colleagues introduced a laboratory model to determine, in a cause-effect manner, the effect of stress on drug craving under controlled laboratory conditions (Sinha, 2001; Sinha et al, 1999). They and other investigators showed that certain stressors, including guided imagery stress, the Trier Social Stress Task, and systemic CRF injections induce physiological (eg, increased cortisol release) and psychological (eg, increased anxiety) stress responses and increase self-reports of drug craving (Back et al, 2010; Sinha, 2009; Sinha et al, 1999). From the perspective of drug relapse, an important finding from these studies is that physiological and psychological (including drug craving) stress responses in the laboratory predict, to some degree, subsequent drug relapse (Back et al, 2010; Higley et al, 2011; Sinha et al, 2006, 2011a).

Alpha-2 adrenoceptor agonists

In several studies, investigators determined the effects of alpha-2 adrenoceptor agonists on stress- and cue-induced craving in opiate, cocaine, heroin, and nicotine (smokers) users. In an initial preliminary study, Sinha et al (2007) showed that in opiate users maintained on naltrexone, lofexidine treatment (2.4 mg/day) decreases both stress- and cue-induced opiate craving. Similar findings were observed in a larger sample study with cocaine users in which Jobes et al (2011) showed that clonidine (0.1 or 0.2 mg) administration decreases stress- and cue-induced cocaine craving (Figure 3).

In another study, Fox et al (2012) determined the effects of guanfacine (up to 3 mg/day) on stress-, cue-, and stress+cue-induced drug craving in cocaine-dependent individuals who also use alcohol and nicotine (smokers). They showed that guanfacine decreases stress- and cue-induced nicotine craving. Guanfacine also modestly decreased cue-induced cocaine craving, anxiety, and arousal, but not stress-induced craving. However, the latter negative finding may reflect a floor effect, because the stress manipulation caused a very weak craving response in the placebo group.

More recently, McKee et al (2014) showed that in human smokers guanfacine treatment (3 mg/day) decreases both stress-induced drug craving and stress-induced smoking in the laboratory. In another study, Fox et al (2014) examined the effect of guanfacine (2 or 3 mg/day) on stress-, cue, and stress+cue-induced craving in male and female cocaine users who also use alcohol and nicotine (smokers). A surprising finding in this study was that guanfacine selectively decreased cocaine craving, alcohol craving, anxiety, and negative emotion following exposure to all three experimental conditions in females but not in males. In contrast, guanfacine decreased stress and cue-induced nicotine craving in both males and females. Finally, in a very recent study Moran-Santa Maria et al (2015a) showed that in cocaine users acute administration of guanfacine (2 mg) has no effect on cocaine craving in response to stress, cue, or stress+cue. In interpreting these data, two pharmacological factors should be considered. The first is that, unlike the previous guanfacine studies of Sinha and colleagues in which guanfacine was given chronically for several weeks, in the Moran-Santa Maria et al (2015a) study, the drug was given acutely. The second is that the data of Sinha and colleagues suggest that the minimal effective dose of guanfacine in the laboratory studies is ~3 mg, which is higher than the dose used in Moran-Santa Maria et al study.

Findings from three other human laboratory studies in which yohimbine was used to induce drug craving and seeking provide additional support to the translational value of stress-induced reinstatement studies in laboratory animals. In these studies, doses of yohimbine that induced physiological and psychological stress responses increased alcohol craving (Umhau et al, 2011), potentiated cue-induced cocaine craving (an effect that was more pronounced in females), and increased opiate seeking in a choice-based (hydromorphone vs money) progressive ratio task (Greenwald et al, 2013). Thus, despite the interpretation issues related to yohimbine’s effects in rat reinstatement studies (Box 2), in humans, the drug induces an aversive stress-like state of which one manifestation is increased drug craving and seeking.

Finally, as discussed elsewhere (Bossert et al, 2013), an unexpected finding in the human studies was that alpha-2 adrenoceptor agonists decreased both stress- and cue-induced drug craving. This was unexpected because in rat studies the alpha-2 adrenoceptor agonists selectively decreased footshock stress-induced reinstatement but not drug priming- or cue-induced reinstatement (Erb et al, 2000; Highfield et al, 2001), but see Smith and Aston-Jones (2011) for different results. A possible account for these different findings is that in rats cue exposure primarily induces an appetitive motivational state (See, 2005) while in humans, cue exposure also induces stress-like physiological and psychological states that are similar to those induced by stress exposure (Sinha et al, 1999, 2000).

Alpha-1 adrenoceptor and CRF1 receptor antagonists

Two studies determined the effects of the alpha-1 adrenoceptor antagonist prazosin or the CRF1 receptor antagonist pexacerfont on stress- and cue-induced alcohol craving. Fox et al (2012) showed that in alcohol users prazosin (16 mg/day) decreases stress-induced craving and negative emotions. Prazosin also decreased cue-induced negative emotion self-reports, but had no effect on cue-induced craving. The negative data might be due to a floor effect, because of a weak cue-induced craving in the placebo group.

Most recently, Kwako et al (2015) determined the effects of oral pexacerfont (300 mg/day for 7 days and 100 mg/day for 23 days) on stress- and cue-induced alcohol craving in ‘anxious’ alcohol-dependent subjects. This study was inspired by the large and remarkably consistent preclinical literature on the effect of CRF receptor antagonists on escalation of drug intake, stress-like physiological and psychological withdrawal symptoms, and stress-induced reinstatement of drug seeking across drug classes (Heilig and Koob, 2007; Koob, 2008; Shalev et al, 2010; Zorrilla et al, 2013a). In sharp contrast to the preclinical data, pexacerfont treatment had no effect on stress- or cue-induced alcohol craving, emotional responses, or anxiety (Figure 4). Pexacerfont also had no effect on neural responses (fMRI) to alcohol-related or affective-related cues.

Clinical Trials

The data from the laboratory studies described above provide evidence for the human ‘translational’ value of the stress-induced reinstatement model. However, from a perspective of human drug relapse, a critical test for the predictive validity of the model is whether medications identified in the rat model show efficacy in the drug user’s environment (Epstein et al, 2006). Kowalczyk et al (2014) recently determined in heroin users whether the clinical efficacy of buprenorphine would be enhanced by clonidine. In this relapse prevention study, the subjects first received contingency management to facilitate initial abstinence. Subjects who successfully maintained abstinence were randomized to receive clonidine (0.1 mg for 1 week, 0.2 mg for the next week, and 0.3 mg for the rest of the study, n=61) or placebo (n=57) for 14 weeks. The main dependent measures were time to lapse (defined as any opiate positive or missed urines) and time to relapse (defined as any consecutive lapses), as well as baseline craving, and stress- and cue-induced craving in the subjects’ home environment, as assessed by real-time ecological momentary assessment (EMA) (Epstein et al, 2009). The main findings of the study were that clonidine treatment increased the time to lapse (from 25 to 35 days) (Figure 4) and decreased baseline craving and stress-induced craving. In contrast, clonidine had no effect on relapse rates once a lapse had occurred.

In the context of our review, a study of Krupitsky et al (2013), which likely was not based on the stress-induced reinstatement studies in the animal model, is also relevant. In this 6-month clinical trial, these authors determined whether guanfacine (1 mg/day) would decrease relapse rates in opiate users maintained on naltrexone. The main finding was that guanfacine had no effect on relapse rates but decreased self-reports of stress and craving. While the study included a large number of subjects (n=75 per group) and was well designed, a potential methodological issue was the use of a low guanfacine dose (1 mg). The studies described in section ‘Human laboratory studies’ suggest that the effective dose of guanfacine in the laboratory studies is ~3 mg (Fox et al, 2014; McKee et al, 2014).

Summary

The results of the human studies with alpha-2 adrenoceptor agonists and prazosin appear to support the translational utility of the stress-induced reinstatement model. However, unlike the uniform effects of these agonists on stress-induced reinstatement in the rat model, the effects in the human studies are more variable, and some of the human studies reported negative results (Krupitsky et al, 2013; Moran-Santa Maria et al, 2015a), possibly due to the use of doses that were lower than those used in the studies that report positive effects. The human studies also suggest potential sex differences in the efficacy of alpha-2 adrenoceptor agonists (Fox et al, 2014), a factor that has not been considered in the preclinical studies. What might account for the more heterogeneous/variable results in the clinical preclinical studies?

We propose that four factors might be particularly important. The first is the fact that rat or mouse studies are invariably performed using highly homogenous subjects of the same or similar age, sex, drug exposure duration, and in many cases likely from a limited number of litters (and therefore a homogenous genetic background). In contrast, the human studies are performed with a highly heterogeneous subject population of different ages, duration of drug use, race, and also includes both sexes. The second factor is the more flexible nature of studies using animal models, which allows investigators to adjust both experimental parameters, including drug doses, to determine the effects of pharmacological agents on reinstatement under more than one experimental set-up. In contrast, human studies are typically performed double-blind from start to end, which increases statistical power but may result in negative data due to two main sources that have a significant role in the studies reviewed above: drug doses that were likely too low and experimental manipulations (stress- or cue-induced reinstatement) that in some studies had minimal effect on craving, making it difficult to detect a reliable effect of the medication. Third, a significant issue in assessing the efficacy of medications to treat drug addiction is compliance (Weiss, 2004), an issue that likely also played a role in the studies reviewed above. Fourth, the different types of stressors used in the preclinical studies (exposure to a physical aversive stimulus such as footshock, restraint, or cold water) vs the stressors used in the clinical studies (exposure to psychological stressors like stress imagery or the Trier Social Stress Task) is an issue to consider in ‘translational’ studies on stress-induced reinstatement of drug seeking in which different results are observed in human studies vs. animal models.

Finally, the negative results with pexacerfont (Kwako et al, 2015) indicate that it is unrealistic to expect that all potential medications identified in animal models of psychiatric disorders, including the stress-induced reinstatement model, will readily translate to the human condition. However, these negative data are particularly disappointing because, as mentioned above, they are based on unparalleled convergence from multiple animal models of drug addiction and results that were independently demonstrated in multiple laboratories using different drugs of abuse under different experimental conditions (Heilig and Koob, 2007; Koob, 2008; Sarnyai et al, 2001; Shalev et al, 2010).

Conclusions and Future Directions

The demonstration that a brief exposure to uncontrollable and unpredictable intermittent footshock stress reinstates heroin seeking in rats (Shaham and Stewart, 1995) has stimulated research by many preclinical investigators on behavioral and neuropharmacological mechanisms of stress-induced reinstatement of drug seeking and CPP. The preclinical model of stress-induced reinstatement and the demonstration that stress can provoke drug craving under controlled experimental conditions (Sinha et al, 1999) have also stimulated translational research to identify potential medications that specifically target stress-induced craving and relapse in humans. In this review, we summarized this field of research to consolidate the current state of knowledge within a single review. The results from the studies reviewed suggest at least four future lines of investigation.

The first is the investigation of the synaptic and cellular mechanisms of stress-induced reinstatement. Several studies have explored this question for swim stress-induced reinstatement of cocaine seeking in rats (Chirila et al, 2014; Kreibich and Blendy, 2004; Polter et al, 2014) and social defeat-induced reinstatement of cocaine CPP in mice (Bruchas et al, 2010, 2011; Land et al, 2009). At present, however, the synaptic and cellular mechanisms of stress-induced reinstatement are largely unknown.

The second is the investigation of circuitry mechanisms of stress-induced reinstatement, using neuroscience methods that go beyond classical neuropharmacological and neuroanatomical methods. In the ‘relapse/reinstatement’ field, investigators recently have begun to incorporate modern chemogenetic (Bossert et al, 2011; Cruz et al, 2014; Fanous et al, 2012; Mahler et al, 2014) and optogenetic (Lee et al, 2013; Ma et al, 2014; Stefanik et al, 2013) methods to study brain circuits of context-, cue-, and drug priming-induced reinstatement of drug seeking, but these methods have not yet been used for stress-induced reinstatement.

The third is the study of sex differences in stress-induced reinstatement of drug seeking. In this regard, footshock stress-induced reinstatement of drug seeking is observed in both male and female rats (Shaham and Stewart, 1995). Additionally, recent studies on sex differences and the role of ovarian hormones in stress-induced reinstatement have used yohimbine as a stressor and the results are mixed. Two studies showed that yohimbine effect on reinstatement of cocaine and methamphetamine seeking is somewhat stronger in female rats than in male rats (Anker and Carroll, 2010a; Cox et al, 2013). However, other studies have shown neither sex differences nor a role of ovarian hormones in yohimbine-induced reinstatement of nicotine or palatable food seeking (Cifani et al, 2012; Feltenstein et al, 2012; Pickens et al, 2012). A question for future research is whether sex differences would be observed in the mechanisms that mediate stress-induced reinstatement.

Finally, during human abstinence, stress exposure often occurs in the presence of contexts and cues previously associated with drug use and the drug itself in the case of an initial lapse (Epstein et al, 2009). Human laboratory studies also indicate that stress can potentiate cue-induced drug craving (Fox et al, 2014; George et al, 2008; Moran-Santa Maria et al, 2015b). Additionally, studies using the reinstatement model demonstrate that stress exposure potentiates the response to both cues (Buffalari and See, 2009; Liu and Weiss, 2002b) and drug priming injections (Graf et al, 2013). Thus, from a translational perspective, mechanistic studies on stress-induced potentiation of the response to drug cues or contexts after extinction or without prior extinction may lead to the identification of new targets for developing medications for relapse prevention.

Funding and Disclosure

The authors declare no conflict of interest.