Neuroplasticity in the mesolimbic dopamine system and cocaine addiction

ERK: psychomotor sensitization

Extracellular signal-regulated kinase activity and subsequent ERK-mediated downstream gene transcription is critical for the DEVELOPMENT OF PSYCH induced by cocaine exposure (Girault et al., 2007). In contrast, acute cocaine-induced ERK activity does not mediate the EXPRESSION OF PSYCHO after withdrawal from the drug. Systemic injections of SL327 or VTA injections of PD98059 before daily cocaine injections (during the development of psychomotor sensitization) attenuate sensitized cocaine-induced locomotion during tests for expression of psychomotor sensitization, performed after several weeks of withdrawal (Pierce et al., 1999; Ferguson et al., 2006; Valjent et al., 2006b). The VTA is a critical site for the development of psychomotor sensitization (Vanderschuren and Kalivas, 2000; Vezina, 2004). However, in previously cocaine-sensitized mice, acute systemic pretreatment with SL327 prior to a cocaine injection did not alter the expression of drug psychomotor sensitization (Valjent et al., 2006b). Recently, Boudreau et al. (2007) reported that in rats demonstrating cocaine psychomotor sensitization after several weeks of withdrawal from the drug, increased ERK2 activity is associated with increased accumbens AMPA (α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid) receptor (AMPAR) surface expression. No increases in ERK2 activity and AMPAR surface expression were observed in rats that did not demonstrate cocaine psychomotor sensitization after repeated non-contingent cocaine exposure and a period of withdrawal. The relevance of these interesting correlational findings to enduring cocaine psychomotor sensitization is a subject for future research.

ERK: conditioned place preference

Accumbens ERK activity likely serves two distinct roles in mediating the rewarding effects of psychostimulant in the CPP procedure: during CPP training, accumbens activity mediates consolidation of learned associations between the drug's unconditioned rewarding effects and the drug-paired context; during CPP testing, ERK activity mediates acute expression of cocaine-cue-conditioned responses. Systemic SL327 injections before cocaine CPP training prevent cocaine-induced accumbens (and dorsal striatum) ERK phosphorylation and subsequent expression of cocaine CPP (Valjent et al., 2000). PD98059 accumbens injections given either before or after CPP training sessions block subsequent amphetamine CPP expression (Gerdjikov et al., 2004). Exposure to the cocaine-paired context during CPP testing increases ERK phosphorylation in accumbens core, and U0126 accumbens core injections block both ERK phosphorylation and expression of cocaine CPP in rats (Miller and Marshall, 2005). In addition, methamphetamine CPP expression is associated with increased accumbens (and dorsal striatum) ERK phosphorylation, and PD98059 accumbens injections prevent this CPP expression. The CPP procedure was also used to demonstrate a role of accumbens ERK in RECONSOLIDATION of memories for cocaine cues. Here, if rats are injected with U0126 or PD98059 into accumbens core immediately after the test for the expression of cocaine CPP, subsequent preference for the drug is impaired (Miller and Marshall, 2005). Valjent et al. (2006a) also provided evidence for a role of accumbens ERK in reconsolidation of memory for cocaine cues.

ERK: cocaine self-administration and cocaine seeking

On the basis of limited clinical evidence, Gawin and Kleber (1986) hypothesized that cue-induced cocaine craving progressively increases over the first several weeks of abstinence and remains high over extended periods. An analogous phenomenon, which was termed incubation of cocaine craving, has been demonstrated in rats and mice: time-dependent increases in cue-induced cocaine seeking over the first months of withdrawal from cocaine (Neisewander et al., 2000; Grimm et al., 2001; Lu et al., 2004b; Lee et al., 2006; Mead et al., 2007). Glutamate-mediated ERK activity in the central amygdala plays an important role in the expression and persistence of incubation of cocaine craving (Lu et al., 2005). Exposure to cocaine cues in extinction tests increases ERK phosphorylation in the central but not basolateral amygdala after 30 but not 1 withdrawal days. After 30 withdrawal days, inhibition of central amygdala ERK phosphorylation by U0126 or the NMDAR antagonist AP-5 decreased cue-induced cocaine seeking; the behavioural effect of AP-5 is mimicked by LY379268 (Lu et al., 2007), an mGluR2/3 agonist that decreases glutamate transmission. After 1 withdrawal day, stimulation of central amygdala ERK phosphorylation by NMDA increased cocaine seeking, an effect reversed by U0126. Exposure to cocaine cues in extinction tests also increases ERK activity in the accumbens (L Lu and Y Shaham, unpublished data) and ventral and dorsal medial prefrontal cortex (mPFC; Koya et al., 2008) after 30 days (but not 1 day) of withdrawal. However, the relevance increased ERK activity in accumbens, and mPFC to incubation of cocaine craving is currently unknown.

ERK: conclusions

The role of ERK in cocaine's behavioural effects involves at least two distinct mechanisms with different time courses (Lu et al., 2006; Girault et al., 2007). One mechanism, which mediates cocaine psychomotor sensitization, and both consolidation and reconsolidation of memories for drug cues in the CPP model, may involves long-term stable alterations in molecules that can influence synaptic plasticity. These alterations may result from ERK-mediated activation of its downstream targets, including CREB, Elk-1, c-Fos and Zif268 (Valjent et al., 2000; Mattson et al., 2005; Miller and Marshall, 2005; Radwanska et al., 2006). The second mechanism involves rapid ERK-mediated effects that mediate acute expression of cocaine CPP and the expression of incubation of cocaine craving (Lu et al., 2006). Here, ERK activity mediates the acute behavioural effects of cocaine cues over short time periods (minutes) that likely do not involve gene transcription or stable changes in synaptic plasticity. For example, in hippocampal pyramidal neurons, ERK can acutely increase neuronal excitability via ERK-mediated inactivation of VOLTAGE-GATED POTASS (Yuan et al., 2002) or ERK-mediated potentiation of excitatory neurotransmission via AMPAR membrane insertion (Qin et al., 2005). Whether either of these mechanisms is engaged in neurons in mesocorticolimbic regions is unknown, although it is intriguing to speculate that this is the case based on evidence discussed above for parallel changes in AMPAR trafficking and ERK activation in accumbens in rats that demonstrated cocaine psychomotor sensitization after prolonged withdrawal from the drug (Boudreau et al., 2007).

Cocaine-induced ERK-mediated neuroplasticity is also critical for cocaine psychomotor sensitization because repeated blockade of cocaine-induced ERK activation in mesocorticolimbic areas prevents the development of cocaine psychomotor sensitization; thereby, satisfying the criterion outline above that reversal of cocaine-induced neuroplasticity of the physiological process to a drug-naive state leads to decreases in a specific behavioural effect of a drug. In the case of cocaine CPP and incubation of cocaine craving, whereas ERK activity plays a critical role in both phenomena, the available data do not satisfy this criterion. Both CPP (Figlewicz et al., 2007) and incubation of craving (Grimm et al., 2002, 2005) are reliably observed with non-drug reinforcers such as sucrose and other palatable foods. Thus, data from studies with cocaine-exposed rats may reflect a role of mesocorticolimbic ERK in the different learning, memory and motivational processes that underlie CPP and incubation of craving that are independent of cocaine-induced neuroplasticity of the ERK pathway. In the context of the hypothesis that ERK is a mediator of cocaine-induced neuroplasticity that contributes to cocaine-seeking behaviours, data from studies on the role of mesocorticolimbic ERK in CPP and incubation of craving with non-drug reinforcers is an important subject for future research.

BDNF: psychomotor sensitization

Daily BDNF injections into VTA or chronic BDNF administration via osmotic minipumps into VTA or accumbens potentiate the acute locomotor-activating effects of cocaine during the development of psychomotor sensitization phase (Horger et al., 1999; Pierce et al., 1999). Furthermore, BDNF heterozygote mice are less sensitive to the locomotor-activating effects of cocaine during this phase (Horger et al., 1999). However, the degree to which BDNF-induced potentiation of cocaine-induced locomotion is associated with cocaine-induced neuroplasticity that underlies enduring cocaine psychomotor sensitization is unclear. VTA injections of BDNF during the development of psychomotor sensitization phase had no effect on subsequent expression of this sensitization after withdrawal from cocaine (Pierce et al., 1999). Pu et al. (2006) recently reported that BDNF-mediated LTP in VTA dopamine neurons is associated with the expression of cocaine psychomotor sensitization in mice. However, the degree to which this interesting finding underlies enduring expression of psychomotor sensitization has not been established.

BDNF: conditioned place preference

Brain-derived neurotrophic factor heterozygote mice, which were less sensitive to the effects of acute cocaine-induced locomotion, were also less sensitive to the rewarding effects of cocaine in the CPP model (Hall et al., 2003). CPP is a contextual learning task, and there is evidence that hippocampal BDNF plays a role in contextual learning (Tyler et al., 2002). Thus, the decreased cocaine CPP in BDNF heterozygote mice may be due learning deficits in these mice. This possibility is unlikely because, although the heterozygote BDNF mice showed impaired CPP with 10 mg kg⁻¹ of cocaine, they demonstrated normal CPP with 20 mg kg⁻¹ cocaine (Hall et al., 2003). However, in the absence of data demonstrating lack of decreases in CPP induced by non-drug rewards, a potential interpretation of the data of Hall et al. is that decreased cocaine CPP in heterozygote BDNF mice reflects a more general reward deficit. Finally, additional evidence for a role of BDNF in psychostimulant CPP is that amphetamine CPP is associated with increased expression of TrkB (tyrosine receptor kinase-B; which mediates BDNF effects on intracellular signalling) in hippocampus (CA3/dentate gyrus) and accumbens shell, and that amphetamine CPP is blocked by CA3/DG injections of K-252a, a non-selective Trk inhibitor (Shen et al., 2006).

Taylor and colleagues (Horger et al., 1999) provided one of the most dramatic illustrations of the role of mesolimbic BDNF in cocaine behavioural effects. It is well established that systemic or accumbens injections of psychostimulant drugs potentiate the rat's response to cues previously paired with natural rewards (Robbins, 1975; Taylor and Robbins, 1984). Horger et al. showed that chronic injections of BDNF into accumbens (via minipumps) profoundly (3–4-fold) increased systemic cocaine-induced potentiation of responding for a conditioned cue previously paired with water in thirsty rats. Remarkably, this effect persisted for up to 5 weeks after cessation of BDNF administration. This seminal study was the inspiration for subsequent studies described below on the role of BDNF in cocaine self-administration, reinstatement of cocaine seeking and incubation of cocaine craving.

BDNF: cocaine self-administration and cocaine seeking

Results from a study by Self and colleagues (Graham et al., 2007) indicate that accumbens BDNF is an important modulator of cocaine's rewarding effects in the drug self-administration model. They injected BDNF or anti-BDNF into the accumbens shell of rats immediately after cocaine self-administration training for 5 days, in rats that had already learned to self-administer cocaine. Three to seven days after the last injection, BDNF-exposed rats self-administered more cocaine over a range of cocaine doses (an upward shift in the dose–response curve), suggesting increases in the rewarding effects of cocaine. In contrast, exposure to anti-BDNF had a modest effect in the opposite direction. Injections of BDNF, but not anti-BDNF, also potently increased the rats' motivation to work for cocaine when the response requirement for each successive injection exponentially increased under a progressive ratio reinforcement schedule (Richardson and Roberts, 1996). Graham et al. also locally injected an adeno-associated viral vector to knock down local BDNF protein production in mice accumbens shell. This manipulation caused a modest downward shift in the dose–response curve, suggesting a decrease in the rewarding effects of cocaine.

Results from three studies indicate that mesocorticolimbic BDNF plays an important role in cocaine craving and relapse, as assessed in rat models. In an initial study on mechanisms underlying incubation of cocaine craving, Grimm et al. (2003) reported that the time-dependent increases in cue-induced cocaine seeking after withdrawal (incubation) is associated with time-dependent increases in BDNF in mesocorticolimbic areas. On the basis of these findings, Lu et al. (2004a) injected BDNF acutely into VTA and assessed cue-induced cocaine seeking in extinction tests after 3, 10 or 30 withdrawal days. They found that these BDNF injections enhanced responding for cocaine cues for up to 30 days after cessation of cocaine use, an effect that was reversed by U0126. The finding of reversal of BDNF effects by U0126 is not surprising in light of the role of ERK signalling in BDNF-mediated synaptic plasticity (Poo, 2001). Lu et al. also found that acute VTA BDNF injections 2 h before testing had no effect on cue-induced cocaine seeking, suggesting that BDNF effects on responding to cues involve stable long-term synaptic alterations rather than acute effects on synaptic transmission or short-term plasticity.

In the same study where Graham et al. (2007) showed that BDNF in the accumbens potentiates the rewarding effects of self-administered cocaine, they also assessed extinction responding and subsequent reinstatement of cocaine seeking induced by cocaine priming injections, discrete cues previously paired with cocaine injections or a footshock stressor. Accumbens BDNF injections potently increased lever responding during both the extinction test and the subsequent reinstatement tests, whereas anti-BDNF injections had effects in the opposite direction. In apparent contrast with these findings, Berglind et al. (2007) used the same BDNF injection procedure in dorsal mPFC that was employed by Lu et al. (2004a) in the VTA. Surprisingly, they reported that these injections, which increased BDNF levels in accumbens a day later (presumably via anterograde transport; Altar et al., 1997), decreased extinction responding after 1 or 6 withdrawal days, and also decreased discrete cue- and cocaine priming-induced reinstatement after 6 withdrawal days. The reasons for the potentially discrepant findings between Berglind et al. data and those of Graham et al. are unknown. A missing piece in interpreting Berglind et al. data is that although there is evidence that endogenous BDNF expression is upregulated in the VTA and accumbens after prolonged withdrawal from cocaine (Grimm et al., 2003; Pu et al., 2006), there is no such evidence for mPFC. However, it is interesting to consider the possibility that the release of BDNF into the accumbens following microinjection into the mPFC in the study by Berglind et al. (2007) may more closely mimic spatially localized, activity-dependent release of BDNF relative to the pharmacological manipulations made by Graham et al. (2007) directly into the accumbens. Furthermore, a methodological issue in these studies is that it is unknown whether the effect of exogenous injections of local high doses of BDNF mimics cocaine-induced neuroplasticity of endogenous BDNF systems. In this regard, the data of Graham et al. of opposite effects of accumbens injections of BDNF versus anti-BDNF and the adeno-associated viral vector manipulation to knock down BDNF protein production are important for interpreting the roles of BDNF in cocaine-seeking behaviours, because the anti-BDNF and the viral vector interfere with the ability of endogenous BDNF to regulate these behaviours.

BDNF: conclusions

The data reviewed above indicate that although BDNF contributes significantly to several behavioural effects of cocaine, only some of them likely involve cocaine-induced BDNF-mediated long-term neuroplasticity in mesocorticolimbic areas. Thus, for cocaine self-administration, extinction responding and reinstatement of cocaine seeking, either acutely increasing or decreasing BDNF levels in the accumbens during cocaine self-administration training lead to corresponding increases or decreases in cocaine reward and the motivation to seek cocaine in a drug-free state after cessation of BDNF injections. The observation that the time-dependent increases in cue-induced cocaine seeking (incubation) in mesocorticolimbic areas are associated with time-dependent increases in BDNF levels, and that acute BDNF VTA injections cause long-lasting potentiation of this incubation further support the notion that cocaine-induced BDNF-mediated long-term neuroplasticity contributes to cocaine craving and relapse. On the basis of the data of Pu et al. (2006) discussed above, it is tempting to speculate that cocaine-induced BDNF-mediated synaptic plasticity in VTA dopamine neurons causes enhanced responsiveness of these neurons to cocaine cues or other stimuli (for example, stress) that provoke cocaine craving and relapse. VTA neurons are critical for relapse to drug seeking induced by drug priming, drug cues and stress (Stewart, 1984; McFarland and Kalivas, 2001; Bossert et al., 2004; Sun et al., 2005; Wang et al., 2005).

In contrast, cocaine psychomotor sensitization likely does not involve enduring cocaine-induced BDNF-mediated neuroplasticity. Although BDNF administered into accumbens or VTA potentiate the acute locomotor-activating effects of cocaine, these manipulations had no effect on the expression of cocaine psychomotor sensitization after several weeks of withdrawal. It also has not been established that cocaine-induced BDNF-mediated neuroplasticity contributes to the rewarding effects of cocaine in the CPP model.

VTA: psychomotor sensitization

It has long been suspected that psychomotor sensitization to cocaine involves adaptations in excitatory transmission, including early work by Post and Kopanda (1976) characterizing sensitization as a ‘kindling' phenomenon. Indeed, in 1988 an entire book was devoted to seeking parallels between cocaine sensitization, electrical kindling, LTP and other forms of sensitization (Kalivas and Barnes, 1988). Although subsequent neuropharmacological studies over the last 15 years demonstrated that excitatory transmission in the VTA is necessary for the development of cocaine-induced psychomotor sensitization (Kalivas and Alesdatter, 1993; Wolf, 1998; Wolf et al., 2004), the first direct demonstration of synaptic plasticity being induced by in vivo cocaine experience at excitatory synapses on dopamine neurons in the VTA was obtained by Ungless et al. (2001). This study demonstrated that with a single cocaine injection, a long-lasting potentiation of synaptic strength (LTP) ensues, an effect that lasts at least 5, but not 10 days. These findings built upon earlier work showing increased glutamate-induced single-unit firing of dopamine neurons after repeated non-contingent cocaine exposure (White et al., 1995), and increased content of the GluR1 AMPAR subunit in the VTA at 24 h, but not 2 weeks after withdrawal from cocaine (Fitzgerald et al., 1996; Churchill et al., 1999).

Since the initial description of synaptic potentiation in the VTA, studies have focused on the molecular and cellular mechanisms of this potentiation (Wolf et al., 2004; Kauer and Malenka, 2007), and how cocaine-induced synaptic potentiation may influence behaviour. Although a clear picture of the role of cocaine-induced VTA synaptic potentiation in behaviour has yet to emerge, several studies provide data to address this issue. For example, Borgland et al. (2004) found that although cocaine-induced synaptic potentiation was correlated with acute cocaine-induced locomotor activation, after repeated cocaine treatment this correlation was no longer observed. Combining these data with the fact that VTA potentiation does not last much beyond 5 days of withdrawal from cocaine (Ungless et al., 2001), suggests that this process is more likely to be involved in early rather than long-lasting changes in reward circuits underlying addiction. This deduction is consistent not only with conclusions drawn from earlier studies showing that cocaine psychomotor sensitization could be initiated by amphetamine injections directly into the VTA and depends on local stimulation of both NMDA and D1 dopamine receptors (Kalivas and Alesdatter, 1993; Vezina, 1996), but also that the expression of this sensitization depends upon AMPA and dopamine receptor stimulation in the accumbens (Vanderschuren and Kalivas, 2000). Also, it was generally observed that adaptations measured in the VTA tended to be transient, whereas those in the accumbens were more enduring (Wolf, 1998).

Given the apparent role of the VTA in the development of cocaine-induced psychomotor sensitization identified in these earlier studies, recent electrophysiological studies have endeavoured to determine if the potentiation of dopamine neurons in the VTA is necessary for cocaine-induced behaviours. First, in a study examining AMPA-type glutamate receptor subtype GluR1 null mutant mice, Dong et al. (2004) found that cocaine does not increase AMPA/NMDA ratio in dopamine neurons in these mice. At the same time, however, there is a strong trend towards an increase in the basal (saline control) AMPA/NMDA ratio in these knockout mice, suggesting that perhaps AMPA receptor-mediated synaptic strength is already functioning at a high level in the absence of cocaine. GluR1 null mutants exhibit a high degree of cocaine-induced locomotion after an acute drug injection and an equivalent degree of psychomotor sensitization after repeated cocaine exposure. These results suggest that cocaine-induced potentiation in VTA dopamine neurons may not be an absolute requirement for cocaine-induced enhancement of locomotion or for psychomotor sensitization. Importantly, a limitation of this study (and related constitutive knockout studies) is that when a gene is deleted throughout development, it is unknown whether the observed changes in synaptic plasticity are indicative of the normal role of a given gene in this plasticity or some unknown compensatory changes that result from the gene deletion. Indeed, an elegant study by Borgland et al. (2006) offers another perspective on this question. These authors established a necessary role for orexin receptor signalling in VTA in the initiation of cocaine-induced synaptic potentiation and development of cocaine psychomotor sensitization. These findings suggest that cocaine-induced VTA dopamine neuron synaptic potentiation may play an important role in the development of cocaine psychomotor sensitization.

Interestingly, earlier pharmacological studies also indicated a different role for the VTA in the acute locomotor response of psychostimulants versus the development of sensitization. For example, daily injections of amphetamine or D1 agonists into the VTA did not alter locomotor activity, but elicited sensitization to a subsequent systemic amphetamine or cocaine injection (Vezina, 1993, 1996; Pierce et al., 1996b). Moreover, if synaptic potentiation is supplied via artificial means, the necessity for the cocaine-induced enhancement is bypassed and psychomotor sensitization still proceeds. For example, psychomotor sensitization can be induced by electrical stimulation of prefrontal cortical glutamatergic afferents to the VTA (Schenk and Snow, 1994), and AMPA receptor overexpression in the VTA facilitates morphine psychomotor sensitization (Carlezon et al., 1997).

In addition to the synaptic potentiation in dopamine neurons discussed above, it is interesting to consider the possibility that cocaine induces METAPLASTICITY at these neurons. In a seminal 1996 paper, Abraham and Bear (1996) defined the concept of metaplasticity to describe ‘plasticity of synaptic plasticity'. Although it has only recently been suggested that metaplasticity may contribute to addiction (Kourrich et al., 2007), it is interesting to consider whether prior findings on cocaine-induced plasticity in the VTA can be interpreted in this light. For example, in vitro electrical synaptic stimulation in slices from cocaine-exposed mice elicited smaller LTP and greater LTD. These results are consistent with an ‘OCCLUSION' interpretation of this metaplasticity. That is, having been potentiated by cocaine exposure itself, these synapses are closer to a maximal amount of potentiation (thus smaller in vitro LTP), but further from a maximal point of depression (thus larger in vitro LTD). Support for this interpretation comes from similar findings in other experimental systems such as cortical LTP and LTD (Allen et al., 2003). This interpretation suggests that at least some molecular and cellular mechanisms are shared between synaptic potentiation in the VTA induced by cocaine experience and experimenter-controlled electrical stimulation.

Consistent with the idea that cocaine-induced potentiation in VTA may induce a near-maximal potentiation that occludes further synaptic enhancement, Borgland et al. (2004) found that multiple cocaine injections do not cause any greater increase in synaptic strength than a single injection. Furthermore, the time course of the potentiation remained the same, such that the time to return to a basal level of synaptic strength was between 5 and 10 days after the last acute or repeated cocaine injection. These results demonstrate the remarkable stability of cocaine's effect on dopamine neuron excitatory synaptic transmission, especially in contrast to the dynamic changes in cocaine's effect on accumbens medium spiny neurons (MSNs; see below).

In addition to occlusion effects, other forms of cocaine-induced metaplasticity may result from changes in plasticity mechanisms themselves. For example, Liu et al. (2005) reported that repeated in vivo cocaine injections cause a decrease in inhibitory synaptic transmission in VTA dopamine neurons. This decrease in inhibition facilitates in vitro LTP induction. Because this form of metaplasticity results from a process other than changes in excitatory synaptic strength per se, it is by definition not an occlusion effect. The relationship between LTP decreases due to occlusion and the reported LTP increases due to reduced inhibition has not yet been resolved (Kauer and Malenka, 2007), and, thus, a credible analysis of the role of these forms of metaplasticity in behaviour is not yet feasible. However, it is interesting to note that if a cocaine-induced decrease in inhibition is engaged to facilitate LTP induction in vivo, as may be predicted from Liu et al. results, this process does not appear likely to be engaged by cocaine experience itself, because multiple cocaine injections do not produce additional detectable potentiation beyond a single injection (Borgland et al., 2004). Of course, this does not rule out the possibility that other types of in vivo stimuli may take advantage of reduced inhibition to induce dopamine neuron LTP in animals with repeated cocaine experience (Liu et al., 2005).

VTA: conditioned place preference

Data addressing a potential role for dopamine neuron potentiation in animal models of cocaine addiction with greater face validity than psychomotor sensitization are beginning to emerge. For example, GluR1 null mutant mice, which lack cocaine-induced potentiation in VTA, also do not exhibit robust cocaine CPP, suggesting a possible link between the two phenomena (Dong et al., 2004; Mead et al., 2005). More detailed behavioural analyses of these mice suggest that, in particular, they show deficits in the learning of certain stimulus–reward relationships (Mead et al., 2007). As discussed above, however, the results from studies using knockout mice should be interpreted with caution. For example, while GluR1 knockout mice demonstrate potentiated expression of incubation of cocaine craving (Mead et al., 2007), other evidence suggests a link between incubation of craving and increased transmission via GluR1-containing AMPAR in the accumbens (Conrad et al., 2007).

Another study provides potential evidence that dopamine neuron synaptic potentiation enhances both morphine CPP and CPA (conditioned place aversion) induced by the κ-agonist U69593 (Kim et al., 2004). In this study, cocaine exposure, known to induce dopamine neuron synaptic potentiation for 5, but not 10 days (Ungless et al., 2001), was followed at various time periods by a single session CPP procedure with either morphine or U69593. The cocaine-treated rats showed enhanced morphine CPP and U69593 CPA only during the time period (that is, up to 5 days) when dopamine neuronal potentiation is known to occur. This enhancement of CPP was absent if an NMDAR antagonist was injected into the VTA prior to cocaine exposure, a pharmacological manipulation known to disrupt dopamine neuron synaptic potentiation (Ungless et al., 2001). Moreover, CPP to morphine is dependent on orexin release into the VTA (Harris et al., 2005), which is consistent with the observations by Borgland et al. (2006) (discussed above) that synaptic potentiation of dopamine cells by cocaine depends on orexin-induced mobilization of NMDAR to the cell surface. Together, these data support the notion that dopamine neuron potentiation enhances the salience of rewarding or aversive stimuli, and thereby promotes learning about contexts associated with these stimuli.

VTA: self-administration and reinstatement

The role of dopamine neuronal potentiation in the self-administration model has not been examined. However, given the short-lived nature of measurable VTA synaptic potentiation after discontinuing cocaine administration, it is likely that its primary role is in facilitating the reward learning required to acquire drug self-administration rather than maintaining cocaine self-administration or reinstatement of cocaine seeking after withdrawal. Indeed, glutamate transmission in the VTA mediates the potentiation effect of prior exposure to systemic and VTA injections of amphetamine on subsequent acquisition of drug self-administration, and responding on a progressive ratio reinforcement schedule (Suto et al., 2003; Vezina, 2004). Perhaps consistent with the idea that synaptic potentiation in the VTA is transient and underlying the acquisition and/or sensitization of drug-seeking, short-lived upregulation of AMPAR subunits were found in the VTA after binge cocaine self-administration (Tang et al., 2004), although more enduring changes were identified in another study (Lu et al., 2003).

Although we have focused on synaptic potentiation, other aspects of excitatory transmission may also undergo neuroplasticity following cocaine. For example, Wang et al. (2005) found that cocaine self-administration induces long-lasting potentiation of CRF (corticotrophin-releasing factor)-mediated stress-induced glutamate and dopamine release in the VTA, an effect critical for stress-induced reinstatement of cocaine seeking. Of course, CRF-induced release of glutamate can be synergistic with synaptic potentiation. Indeed, like cocaine, stress exposure increases AMPA/NMDA ratio in VTA dopamine neurons (Saal et al., 2003).

Other evidence suggesting that some forms of long-lasting alterations in VTA neurons contribute to cocaine-seeking behaviours includes an intriguing form of cocaine-induced metaplasticity in dopamine neurons that involves BDNF (Pu et al., 2006). This study shows that although BDNF application by itself does not enhance dopamine neuron synaptic strength, this application facilitates the induction of in vitro LTP. In addition, in mice given repeated non-contingent exposure to cocaine, subthreshold electrical stimuli delivered in slices taken from these mice are able to generate LTP after 2 weeks of withdrawal from the drug; at this time point but not during early withdrawal, BDNF VTA levels were increased. Also, interfering with TrkB signalling blocks the enhancement of LTP, suggesting that cocaine-enhanced plasticity may be due to the increases in BDNF. Coupled with the fact that the synaptic plasticity experiments were done in the presence of GABA receptor antagonists, these findings appear to demonstrate a form of cocaine-induced metaplasticity that is distinct from that seen by reducing inhibition (Liu et al., 2005; Pu et al., 2006). This phenomenon may provide an attractive candidate mechanism for the effect of VTA BDNF injections on potentiation of cue-induced cocaine seeking after prolonged withdrawal (incubation of craving) (Lu et al., 2004a).

Accumbens: psychomotor sensitization

Although cocaine-induced changes in VTA dopamine neuron synaptic efficacy are rapidly induced and short-lived, plasticity in excitatory transmission in accumbens neurons has a slower and longer lasting time course. Also, in contrast to the robust effect of a single cocaine injection on dopamine neuron synaptic function, this treatment does not change synaptic strength in accumbens neurons in drug-naive rodents (Thomas et al., 2001; Fourgeaud et al., 2004; Kourrich et al., 2007). With repeated exposure, however, synaptic strength in these neurons appears to be depressed initially (Kourrich et al., 2007). However, 10 days after withdrawal from non-contingent cocaine exposure, this depression has reversed course to reveal robust potentiation, evident from electrophysiological measures as well as increased AMPAR surface expression (Boudreau and Wolf, 2005; Boudreau et al., 2007; Kourrich et al., 2007). These data are consistent with earlier studies demonstrating that the locomotor stimulant effect of AMPA microinjection into the accumbens is markedly potentiated after withdrawal from repeated non-contingent cocaine exposure, and that accumbens AMPA-induced psychomotor potentiation only occurs in rats that develop this sensitization (Bell and Kalivas, 1996; Pierce et al., 1996a).

Little is yet known about the molecular and cellular mechanisms for cocaine-induced accumbens synaptic plasticity. However, an enduring upregulation in GluR1 was shown in whole-cell lysates only in the accumbens of cocaine-sensitized rats (Churchill et al., 1999), and more recent studies demonstrated an increase in surface expression of GluR1/2-containing AMPAR after repeated cocaine exposure that was present only in rats developing psychomotor sensitization (Boudreau and Wolf, 2005; Boudreau et al., 2007). This effect was paralleled by an increase in ERK2 activity in the accumbens (see ERK section above). Consistent with this behavioural pharmacology and biochemical evidence, direct electrophysiological measurement of synaptic AMPAR responses 2 weeks after withdrawal from repeated non-contingent cocaine exposure demonstrates a synaptic potentiation (Kourrich et al., 2007). This increase in synaptic AMPAR responses is correlated with the development of psychomotor sensitization (MJ Thomas, unpublished data).

The current status of experimental information allows for many interpretations of the role(s) that MSNs synaptic potentiation can have in psychomotor sensitization. For example, in many areas of neuroscience, including the study of learning, epilepsy and chronic pain, a similar systems level or behavioural output often emerges from different underlying substrates with the passage of time and/or experience with additional stimuli in the environment (e.g. Squire, 1987). Thus, it is possible that sensitized psychomotor activity is supported by different underlying mechanisms at different time points after withdrawal from cocaine. In other words, during late withdrawal from repeated cocaine, MSNs synaptic potentiation may be a mediator of sensitization whereas at earlier time points it is not. In contrast, accumbens synaptic potentiation, rather than making a direct contribution to behavioural output, may actually be a compensatory response to changes in other circuit elements. As first suggested by Boudreau and Wolf (2005), it may in fact be a form of homeostasis related to the ‘SYNAPTIC SCALING' phenomenon. Synaptic scaling refers to an alteration in synaptic responses throughout a cell that compensates for chronic changes in activity (Turrigiano and Nelson, 2004). In this scenario, general decreases in accumbens neuronal excitability (Zhang et al., 1998, 2002b; Hu et al., 2004; Dong et al., 2006) or extracellular glutamate (Pierce et al., 1996a; Hotsenpiller et al., 2001; Baker et al., 2003; McKee and Meshul, 2005) known to occur after withdrawal from repeated cocaine exposure may trigger a lasting compensatory enhancement in AMPAR trafficking to excitatory synapses.

Although there are no studies yet that directly tested the scaling hypothesis, the idea fits with the fact that the observed changes in synaptic strength (Kourrich et al., 2007) and AMPAR surface expression (Boudreau and Wolf, 2005; Boudreau et al., 2007) are widespread. Between-group differences are readily detected through analysis of accumbens tissue homogenates (Boudreau and Wolf, 2005; Boudreau et al., 2007) and recording from relatively small groups of individual MSNs (Kourrich et al., 2007). From the data outlined above, it is tempting to assume that the cocaine-induced synaptic potentiation in the accumbens underlies psychomotor sensitization. However, at present it remains possible that as overall excitability to accumbens MSNs is markedly reduced after withdrawal from cocaine, the synaptic potentiation is restoring normal excitability and that interfering with this synaptic potentiation may in fact facilitate addiction-related behaviours such as psychomotor sensitization. Although this seems contradictory to the fact that injections of AMPAR antagonists in the accumbens prevent the expression of psychomotor sensitization or the reinstatement of cocaine seeking (Pierce et al., 1996a; Cornish and Kalivas, 2000; Di Ciano and Everitt, 2001; Park et al., 2002), blocking accumbens AMPARs through pharmacological means can produce undesirable side effects such as blockade of cocaine-induced dopamine release (Moghaddam and Bolinao, 1994; Pap and Bradberry, 1995).

One aspect of plasticity in accumbens MSNs that has emerged as a potential contributor is the increase in dendritic spine density and ACTIN CYCLING produced by repeated cocaine exposure (Robinson and Kolb, 2004; Toda et al., 2006). Actin cycling regulates the growth, retraction and shape of dendritic spines, and is controlled by a number of binding proteins that determine the rate and direction of actin polymerization and depolymerization (Kasai et al., 2003; Carlisle and Kennedy, 2005; Lippman and Dunaevsky, 2005). Moreover, actin cycling is thought to be a necessary component in many forms of synaptic plasticity, including LTP (Malinow and Malenka, 2002; Lisman, 2003). Interestingly, the actin-binding proteins associated with the increase in actin cycling after repeated cocaine exposure predict that the majority of actin polymerization is in the form of filopodia rather than lamellipodia (Toda et al., 2006). Filopodia are generally thought to be less stable, and more amenable to neuroplastic changes than lamellipodia are (Carlisle and Kennedy, 2005). Thus, the increase in actin cycling and spine density may predict that more rapid or quantitatively larger alterations in POSTSYNAPTIC DENSITY (PSD) proteins, and, consequently, spine structure can be induced by subsequent experimental or environmental stimuli. In other words, this biochemical and morphological plasticity in dendritic spines may be one means by which metaplasticity is achieved.

Accordingly, and akin to the situation in the VTA, metaplasticity is produced in the accumbens. For example, in vivo cocaine exposure increases LTP in accumbens MSNs when induced by experimenter-controlled electrical stimulation (Yao et al., 2004; Goto and Grace, 2005), while decreasing LTD induced via these means (Thomas et al., 2001; Fourgeaud et al., 2004; Martin et al., 2006). In contrast, recent studies show that when cocaine experience itself is used as a plasticity-generating stimulus, it can become a potent initiator of accumbens synaptic depression in mice with a history of cocaine exposure and subsequent withdrawal (Kourrich et al., 2007). Two studies provide evidence that repeated cocaine exposure modifies the induction of tetanus-induced LTP in vitro (Yao et al., 2004) and in vivo (Goto and Grace, 2005). Interestingly, although LTP at PFC–accumbens synapses appears to be enhanced (Yao et al., 2004; Goto and Grace, 2005), LTP at hippocampal–accumbens synapses is reduced (Goto and Grace, 2005). Given that these two populations of afferents are thought to carry distinct types of information and serve distinct functions in modulating accumbens MSNs firing patterns (O'Donnell and Grace, 1995), the cocaine-induced imbalance between plasticity at PFC and hippocampal inputs is likely to have a dramatic, yet complex role in modulating accumbens function. Goto and Grace proposed that this plasticity imbalance may help to explain why cocaine-exposed rats have difficulty in set shifting during goal-directed learning tasks, and thus produce excessive perseveration errors—potentially a useful model for certain aspects of addiction.

In addition to enhanced experimenter-induced accumbens LTP, other data demonstrate decreased LTD induction after cocaine exposure. For example, a single exposure to cocaine is sufficient to block the in vitro induction of a presynaptic endocannabinoid-mediated form of LTD in accumbens MSNs (Fourgeaud et al., 2004). This blockade occurs in the absence of any detectable changes in excitatory synaptic strength induced by a single cocaine injection (Thomas et al., 2001; Fourgeaud et al., 2004; Kourrich et al., 2007), and thus provides another example of metaplasticity that cannot be explained by a simple occlusion effect.

The investigation of metaplasticity discussed thus far has involved testing for cocaine-induced changes in LTP or LTD by inducing these phenomena with experimenter-controlled electrical stimulation. One difficulty in extrapolating from these metaplastic effects to their potential roles in behaviour is that it is unclear under what circumstances these forms of plasticity are actually engaged during in vivo experience. In contrast, a form of metaplasticity was recently demonstrated by using in vivo cocaine itself (rather than electrical stimulation) as the plasticity-inducing stimulus (Boudreau et al., 2007; Kourrich et al., 2007). These studies demonstrated that a single cocaine injection in cocaine-experienced mice during an extended drug-free period (10–14 days) triggered a synaptic depression (Thomas et al., 2001; Kourrich et al., 2007) and AMPAR internalization (Boudreau et al., 2007), whereas the same treatment in drug-naive animals did not (Thomas et al., 2001; Fourgeaud et al., 2004; Boudreau et al., 2007; Kourrich et al., 2007). Cocaine re-exposure after withdrawal from the drug also provides a trigger for the expression of psychomotor sensitization.

Two lines of evidence suggest that this cocaine experience-induced form of accumbens metaplasticity may contribute to the expression of psychomotor sensitization. First, Boudreau et al. demonstrated that cocaine-exposed rats that do not exhibit psychomotor sensitization do not exhibit cocaine-induced accumbens AMPAR internalization after withdrawal from the drug. Second, after withdrawal from amphetamine, infusion of an AMPAR endocytosis-blocking peptide into the accumbens 90 min prior to a drug challenge interferes with the expression of amphetamine psychomotor sensitization (Brebner et al., 2005). Because this peptide selectively interferes with the induction of LTD in vitro without decreasing basal AMPAR responses, one reasonable interpretation of these results is that an amphetamine challenge induces an acute synaptic depression in the accumbens in vivo and that this depression is necessary for the expression of psychomotor sensitization. However, an issue to consider in the interpretation of these data is that there are no electrophysiological data to confirm that acute exposure to amphetamine (or cocaine) induces synaptic depression during tests for the expression of psychomotor sensitization. Thus, although these data are suggestive, the relationship between accumbens synaptic depression and the expression of cocaine sensitization will require further exploration.

Accumbens: self-administration and reinstatement of drug seeking

A variety of data strongly suggest that the prefrontal glutamatergic afferents to the accumbens are critical for the reinstatement of cocaine seeking, as recently reviewed by Kalivas and O'Brien (2008). This includes increased synaptic release of glutamate in the accumbens during drug seeking (McFarland et al., 2003, 2004) and blockade of drug seeking by inhibiting AMPAR (Cornish and Kalivas, 2000; Di Ciano and Everitt, 2001; Park et al., 2002) or stimulating mGluR2/3 receptors, which reduces synaptic glutamate release (Bossert et al., 2006; Peters and Kalivas, 2006). These findings are by and large consistent with the synaptic potentiation of glutamatergic synapses after withdrawal from non-contingent cocaine exposure outlined above. In fact, recent evidence indicates that a profound increase in GluR1 surface and total expression in accumbens also occurs after withdrawal from cocaine self-administration, a phenomenon that may mediate the incubation of cocaine craving (Conrad et al., 2007).

However, on the surface these collective data would seem to be at odds with the electrophysiological demonstrations of an enduring decrease in LTD in the accumbens of mice withdrawn from daily sessions of cocaine self-administration (Martin et al., 2006). Importantly, these investigators demonstrated that enduring loss of LTD was not observed in several control conditions (yoked cocaine, food self-administering or naive). Given the synaptic potentiation apparent from other electrophysiological and biochemical studies, one might have expected an occlusion of LTP rather than diminished LTD. Similarly, another study reports decreased excitatory field potential responses in response to stimulation of glutamatergic afferents after withdrawal from daily self-administered cocaine (Schramm-Sapyta et al., 2006). Finally, another finding, which on the surface appears inconsistent with the possibility that synaptic potentiation in accumbens MSNs mediates cocaine-seeking behaviour, is that inhibition of actin cycling by accumbens injections of latrunculin (which prevents actin polymerization) or a Lim-kinase inhibitor (which stabilizes actin structure) potentiates cocaine-induced reinstatement of drug seeking (Toda et al., 2006). Thus, the increased potential for metaplasticity inherent in high actin cycling may be protective and limiting to the behavioural effect of an acute cocaine injection. However, understanding the relationship between actin cycling, loss of LTD and synaptic potentiation requires significant additional experimentation, and conclusions drawn from the available data should be considered speculative.

Transient plasticity

The potentiation of synaptic strength in VTA dopamine cells by a single injection of cocaine has emerged as a key observation regarding a potential transient mechanism for initiating enduring neuroplasticity (Ungless et al., 2001; Borgland et al., 2004, 2006; Wolf et al., 2004; Bellone and Luscher, 2006; Dong et al., 2006). This potentiation of synaptic AMPARs appears mediated in part by an increase in NMDAR signalling that involves orexin (Borgland et al., 2006). Importantly, the facilitation of dopamine cell firing by potentiating synaptic strength at glutamate synapses would increase dopamine signalling in terminal fields known to harbour enduring forms of cocaine-induced neuroplasticity. Increased ERK activity in these dopamine terminal fields by acute cocaine may also initiate cellular events that mediate the enduring changes in gene expression and protein function that underlie cocaine psychomotor sensitization and cocaine seeking (Lu et al., 2006; Girault et al., 2007). Non-transcriptional ERK-mediated alterations in dendritic plasticity (Wu et al., 2001) may also account for the role of ERK in cocaine psychomotor sensitization, and consolidation and reconsolidation of memory for cocaine cues. In this regard, morphological changes in dendritic spines of neurons in dopamine axon terminal fields (Robinson and Kolb, 2004) and associated increases in actin cycling (Toda et al., 2006) are induced by drug exposure regimens that induce ERK-dependent psychomotor sensitization (Ferguson et al., 2006), posing a potential role for ERK-stimulated signalling in these forms of enduring plasticity. Furthermore, a recent study directly linked ERK2 activation to increased AMPAR surface expression in the accumbens of cocaine-sensitized rats (Boudreau et al., 2007).

Enduring basal plasticity and metaplasticity

These forms of cocaine-induced plasticity are probably the most widely described, and are manifested more in dopamine axon terminal fields such as the accumbens, mPFC and amygdala than in the VTA (but see Pu et al., 2006 regarding enduring effects of BDNF in the VTA to facilitate the excitability of dopamine cells). Although recent reviews provide an overview of plasticity in dopamine terminal fields in many cellular domains ranging from gene expression and protein function to neurotransmission (Hyman et al., 2006; Kauer and Malenka, 2007; Kalivas and O'Brien, 2008), the topics in the present review focus on the accumbens and enduring baseline changes in ERK, BDNF and glutamatergic synaptic activity. In this regard, the concept of glutamatergic synaptic potentiation in MSNs is probably most relevant. Thus, a collision of findings from behavioural pharmacology with glutamate receptor agonists and antagonists (Pierce et al., 1996a; Suto et al., 2004), neurochemical measures of glutamate receptor subunit surface expression (Churchill et al., 1999; Boudreau and Wolf, 2005; Boudreau et al., 2007; Conrad et al., 2007) and electrophysiological analysis of synaptic currents (Kourrich et al., 2007) all support the perspective that withdrawal from repeated cocaine exposure causes increased responsiveness of glutamatergic synapses to AMPAR stimulation in the accumbens.

Determining the role of synaptic potentiation in mediating the augmented behavioural responding associated with repeated cocaine exposure, such as locomotor sensitization or cocaine seeking, will require further experimentation employing pharmacological or behavioural treatments that reverse synaptic potentiation and also antagonize or potentiate behavioural responding. The fact that blocking accumbens AMPAR inhibits both the expression of cocaine psychomotor sensitization and cocaine seeking (Vanderschuren and Kalivas, 2000; Kalivas and McFarland, 2003; Bossert et al., 2005) can be viewed as a first step towards such a proof. However, eliminating excitatory transmission is far too general a manipulation to be considered as evidence directly supporting a role for synaptic potentiation. Similarly, BDNF treatments in the accumbens potentiated cocaine seeking, while inhibiting BDNF reduced subsequent cocaine seeking (Graham et al., 2007). BDNF treatment is known to promote manifestations of synaptic potentiation, such as LTP (Poo, 2001). Moreover, the incubation of cocaine craving is associated with accumulating BDNF in the accumbens (Grimm et al., 2003). However, these observations fall short of directly linking synaptic potentiation with BDNF and augmented behavioural responding after cocaine self-administration.

Another aspect of dendritic physiology that shows enduring basal alterations in the accumbens after chronic cocaine exposure is the increase in dendritic spine density and actin cycling (Robinson and Kolb, 2004; Toda et al., 2006). Given the important role of spine morphology and actin cycling regulation of PSD proteins in synaptic strength (Kasai et al., 2003; Lisman, 2003; Carlisle and Kennedy, 2005), it is tempting to link these changes to synaptic potentiation and increased AMPAR surface expression in accumbens (Boudreau and Wolf, 2005; Boudreau et al., 2007). However, in contrast to predictions of this hypothesis, accumbens injections of latrunculin (a compound that inhibits actin cycling and has been shown to change spine structure in vitro) potentiated cocaine seeking (Toda et al., 2006) at a dose that caused a rapid loss of GluR1 from the PSD subfraction (S Toda and PW Kalivas, unpublished observations). Thus, if these findings are indeed related to synaptic potentiation, they argue that synaptic potentiation may be compensatory, because reducing GluR1 expression in the PSD (and presumably reducing synaptic potentiation as well) by disrupting actin cycling promoted cocaine seeking. One possible account of the data of Toda et al. is that they reflect a distinction between experimenter-administered and self-administered cocaine with regards to synaptic potentiation. Thus, while synaptic potentiation may be positively linked to psychomotor sensitization, it is negatively linked to self-administration behaviour. Indeed, a study examining LTD after self-administered cocaine found a loss of LTD in the accumbens (Martin et al., 2006). Assuming the validity of the occlusion hypothesis, which predicts that a loss of LTD occurs in synapses that are already depressed (Allen et al., 2003), these data argue that in the basal state after withdrawal from self-administered cocaine, glutamate synapses in the accumbens are not potentiated. However, more research is needed to address this question, as recent data of Conrad et al. (2007) indicate that extended withdrawal (45 days) from cocaine self-administration is accompanied by increased surface expression of the AMPAR subunit GluR1 in accumbens, and that this increased expression is associated with a progressive increase in cue-induced cocaine seeking (see above).