INTRODUCTION

Drugs of abuse, such as amphetamine, are typically administered within a complex environment containing a large number of diverse stimuli. With repeated systemic drug administration, cues come to predict the availability of the reward. This leads to the development of conditioned drug-seeking behavior in the presence of cues, thereby increasing chances of reinforcement. Interestingly, the development of associative conditioning parallels the formation of nonassociative sensitized behavioral and neurochemical responding to amphetamine (Singer et al, 2009, 2014a, b; Vezina and Leyton, 2009) and both have been implicated in escalated drug use in humans and animal models (Leyton and Vezina, 2013, 2014).

It is possible to identify mesocorticolimbic neuroadaptations resulting from sensitization, in the absence of conditioning, by administering repeated amphetamine directly into the ventral tegmental area (VTA), the site where amphetamine is known to act to initiate the development of sensitization (Vezina, 2004). Although both systemic and intra-VTA injection routes lead to the development of nonassociative sensitization to amphetamine (eg, enhanced locomotion and self-administration of drug; Vezina et al, 2002), only systemic amphetamine exposure leads to the generation of conditioning (Singer et al, 2009; Vezina and Stewart, 1990). The present experiments compared sensitization- and conditioning-related plasticity in the nucleus accumbens (NAcc) following either repeated VTA or systemic injections of amphetamine.

Cell bodies in the VTA project to and release dopamine (DA) onto medium spiny neurons (MSNs) in the NAcc. Repeated administration of DA-releasing psychostimulant drugs produces both transient and long-lasting changes in the size and density of dendritic spines on these GABAergic neurons (Dumitriu et al, 2012; Robinson and Kolb, 1997, 1999; Shen et al, 2009). These dendritic spines may be locations of memory storage in the brain (Lamprecht and LeDoux, 2004); various behavioral experiences have been shown to increase spine formation in vivo (Holtmaat et al, 2005, 2006; Lai et al, 2012; Moczulska et al, 2013; Muñoz-Cuevas et al, 2013). Importantly, by comparing the effects of intra-VTA and systemic injections of amphetamine, we recently found that spine plasticity maps onto the accrual of associative drug conditioning rather than the induction of nonassociative drug sensitization (Singer et al, 2009). Enhanced plasticity of dendritic spines in the NAcc core has also been documented during the retrieval of drug-related memories (Gipson et al, 2013). Together, these results suggest a role for dendritic plasticity in the NAcc in the expression of conditioned behaviors.

Although brain region-specific changes in dendritic architecture following repeated systemic amphetamine have been reported (Robinson and Kolb, 2004; Singer et al, 2009), not every cell within a prescribed anatomical region likely contributes to the processing of drug-related information. In the present experiments, we characterized the dendritic morphology of neurons that contribute to the expression of either sensitization or conditioning. To accomplish this, we first visualized the dendritic profiles of NAcc neurons that accumulated the transcription factor ΔFosB following repeated drug administration (Hope et al, 1994; McClung et al, 2004). These neurons show a long-lasting increase in ΔFosB expression, even without reexposure to drugs or cues, and have been shown to mediate enhanced locomotor responding to and self-administration of psychostimulant drugs (Colby et al, 2003; Kelz et al, 1999). In contrast to ΔFosB, the immediate early gene c-fos is expressed in the NAcc upon reexposure to a drug-paired stimulus (Cruz et al, 2014a). Ensembles of activated c-Fos+ neurons are believed to encode behaviorally relevant conditioning information (Cruz et al, 2014b). Indeed, we show here that rats that do not develop conditioned responding also do not show cue-elicited c-Fos expression.

The discriminative learning procedure we used promotes the formation of context-specific memories, along with localized activation of the NAcc shell (and not the NAcc core). Similar to other recent reports (Bossert et al, 2007, 2013; Everitt and Robbins, 2005; Singer et al, 2014a, b), these experiments demonstrate a role for the NAcc shell in the processing of contextual as opposed to discrete cues. Thus, by focusing on the activation of MSNs in the NAcc shell, we highlight the existence of rapid changes in the dendritic morphology of neuronal ensembles upon retrieval of conditioned contextual memories. Given the ubiquity of drug-related stimuli in a contextual stimulus complex previously associated with a drug, conditioned increases in locomotion reflect increased approach and interaction with these stimuli (Vezina and Stewart, 1987) that can promote reinstatement of drug seeking (Cruz et al, 2014a). These effects are consistent with the processing by neurons in the NAcc shell of the incentive motivational properties of psychomotor stimulant drugs (Everitt and Robbins, 2005).

MATERIALS AND METHODS

Subjects and Surgery

Male Sprague-Dawley rats (Harlan Sprague-Dawley, Madison, WI) weighing 250–275 g on arrival were used (Supplementary Data S1). A subset of rats (n=43) underwent stereotaxic surgery for implantation of bilateral guide cannulae aimed at the VTA (Supplementary Figure S2), according to procedures previously described (Singer et al, 2009) and outlined in Supplementary Data S1.

Behavioral Procedures

Rats were exposed to repeated amphetamine or saline injections in both locomotor sensitization (Supplementary Data S3) and discrimination learning (Supplementary Data S4) experiments. For sensitization studies, amphetamine (VTA, 2.5 μg per 0.5 μl/side; IP, 1.5 mg/kg) or 0.9% saline (VTA, 0.5 μl/side; IP, 1.0 ml/kg) were administered in rats’ home cages once every 3 days for a total of 4 VTA microinjections or 5 IP injections. Although these same doses were used for the conditioning studies, injections were given in specific contexts (Supplementary Figure S1). Briefly, Paired rats received amphetamine in an open field and saline in their home cages. In contrast, Unpaired rats received saline in the open field and amphetamine in the home cage. Control rats received saline injections in both environments. For all studies, locomotion was recorded in the open field chambers (Supplementary Data S2).

Immunoblotting

Separate rats were exposed to amphetamine or saline (IP, AMPH n=9, SAL n=11; VTA, AMPH n=8, SAL n=5; locomotor data were not collected; Supplementary Data S3) and killed via rapid decapitation 1 week later. Brains were immediately harvested and then frozen for later analysis of ΔFosB expression in the NAcc (Figure 1; Supplementary Data S5).

Figure 1
figure 1

Exposure to amphetamine IP (left) or in the VTA (right) enhances subsequent locomotor responding to amphetamine and ΔFosB expression in the NAcc. Testing was conducted 1 week following exposure. All rats were administered an IP challenge injection of amphetamine. Rats previously exposed to amphetamine (AMPH) either IP (a) or in the VTA (b) showed a sensitized locomotor response to amphetamine compared with saline (SAL)-exposed rats. Data are shown as group mean (±SEM) locomotor counts over the 2 h test as well as group mean test total locomotor counts (+SEM) in the insets. Similarly, both IP (c) and VTA (d) exposure to amphetamine led to increased ΔFosB expression in the NAcc 1 week later. Western blot results are depicted as ΔFosB/tubulin ratios and shown as % of SAL controls. Example blots are displayed below the graphs. *p<0.05; **p<0.01; ***p<0.001; amphetamine-exposed relative to saline-exposed rats. Numbers at the base of each column indicate rats/group.

PowerPoint slide

Immunohistochemistry and Cell Counting

Rats were fully anesthetized (ketamine (100 mg/kg, IP) and xylazine (10 mg/kg, IP)) and perfused intracardiacally with phosphate-buffered saline (1 × PBS) and 1.5% paraformaldehyde. Brains were then harvested (Supplementary Data S6) and 100 μm coronal sections were obtained. NAcc cell counting and DiI injections were performed in slices ~2.10±0.20 mm from bregma (Paxinos and Watson, 1998). Free-floating sections were stained for NeuN, either c-Fos or ΔFosB expression or both (cell counting). For DiI injections, free-floating sections were stained for either ΔFosB or c-Fos expression.

DiI Injections and Dendritic Morphology

DiI injections were performed in immunostained tissue as described (Supplementary Data S6 and S7 and Supplementary Figure S3; 120 cells and 20,349 spines analyzed). For rats exposed to repeated systemic amphetamine or saline in the sensitization experiment, ΔFosB+ or unlabeled neurons (ΔFosB−) were identified and injected with DiI (Figure 2; n=3–4 rats/group). For the discrimination learning experiment (Figure 4; n=4 rats/group), c-Fos+ or unlabeled (c-Fos−) neurons were injected with DiI. For all cell injections (3.24±0.19 cell type/rat), a pipette was positioned onto the plasma membrane and DiI was deposited via an injection of positive current. Second-order dendrites from DiI-injected neurons were imaged using a confocal microscope. Dendrites underwent automatic software reconstruction (Imaris, Bitplane; Supplementary Data S7; Shen et al, 2008, 2009). Dendritic spine tip densities were higher than overall spine densities because some protrusions had multiple heads, a phenomenon associated with conditioning (Geinisman et al, 2001).

Figure 2
figure 2

Exposure to amphetamine IP had no effect on the characteristics of dendritic spines in ΔFosB+ or ΔFosB− neurons in the NAcc shell. Testing was conducted 1 week following exposure. Compared with saline (SAL) exposure, no effect of amphetamine (AMPH) exposure was detected on the distribution of spine tip diameters in (a) ΔFosB+ and (b) ΔFosB− neurons. For each reconstructed dendrite, the number of spines with tip diameters falling within 0.15 μm intervals (eg, 0.30–0.45 μm; listed as 0.45 μm on the abscissa) was counted and the frequency of spines per μm of dendrite was determined. (c) Exposure to amphetamine also produced no effects on spine or spine tip density in ΔFosB+ or ΔFosB− neurons. Data are shown as group means (+SEM). (d) DiI-labeled dendritic segment following deconvolution processing. Images were taken on a Leica SP5 confocal microscope (63 × oil objective with 4 × digital zoom). (e) Reconstruction of the dendritic segment using the Imaris software filament tracer module. (f) Surface reconstruction of the dendritic segment used to determine the quality of the DiI staining. Numbers in parentheses indicate rats, cells/group. Measures obtained from the cells of each rat (2–6/rat) were averaged and ANOVA conducted on these values. ANOVA conducted on unaveraged cell values yielded identical results, indicating homogeneity between cells for the different group measures (see Supplementary Data S7).

PowerPoint slide

Data Analysis

Sensitized locomotion, ΔFosB expression, and overall spine (tip) densities for ΔFosB+/− neurons were analyzed using independent t-tests. Tests of conditioned locomotion, c-Fos, and ΔFosB expression were analyzed using one-way ANOVAs with conditioning group as the between factor. Changes in overall dendritic spine (tip) density after the conditioning test were also analyzed using one-way ANOVAs (separately for c-Fos+ and c-Fos− neurons). Finally, spine tip diameters (for both ΔFosB and c-Fos studies) were measured across neurons and placed into bins to permit analytical comparisons to previously described results (Dumitriu et al, 2012; Shen et al, 2009). Two-way between–within ANOVAs were performed to determine the presence of conditioning group × tip diameter interactions. A significant interaction allowed for analysis of tip size frequency in specific predefined diameter bins using post hoc Scheffé tests (Kirk, 1968). Comparing frequencies of specific tip diameters to the overall tip density helped differentiate between de novo spine formation and shifts in the distribution of existing spine sizes.

RESULTS

Nonassociative Sensitization

Repeated intermittent exposure to amphetamine (1) enhances the ability of the drug to elicit psychomotor behavior (sensitization) and (2) increases expression of the transcription factor ΔFosB in the NAcc. We confirmed that these long-lasting adaptations result from either systemic (IP) or intra-VTA (microinjected) amphetamine exposure. On tests conducted 1 week following drug exposure, both routes of drug administration resulted in sensitized locomotor responding to a systemic amphetamine challenge relative to rats that previously received saline (Figures 1a and b; IP, t10=4.52, p<0.001; VTA, t11=3.75, p<0.01). Similarly, we found in a separate group of rats that exposure to either IP (Figure 1c; t18=2.14, p<0.05) or VTA (Figure 1d; t11=2.25, p<0.05) amphetamine increased ΔFosB expression in the NAcc relative to saline-exposed controls. Together, these findings demonstrate that either repeated systemic or intra-VTA amphetamine exposure is sufficient to induce behavioral sensitization and long-lasting ΔFosB expression. It is possible that elevated ΔFosB levels facilitate gene transcription needed to support the expression of sensitization (Colby et al, 2003; Hu et al, 2002; Kelz et al, 1999).

Despite these findings, VTA amphetamine does not increase dendritic spine density in the NAcc (Singer et al, 2009). As both routes of drug administration result in behavioral sensitization, the observed dissociation suggests that ΔFosB expression is not sufficient to increase the dendritic spine plasticity associated with repeated IP amphetamine (Robinson and Kolb, 2004; Singer et al, 2009). Therefore, we assessed the dendritic spine morphology and density of ΔFosB+ and ΔFosB− neurons in the NAcc shell 1 week after exposure to repeated IP amphetamine or saline. To accomplish this, we immunohistochemically identified ΔFosB+ neurons and specifically targeted these cells for injection with the fluorescent neuronal tracer DiI. We then imaged these neurons using confocal microscopy to create three-dimensional reconstructions of their neuroanatomy (Figures 2d–f).

In ΔFosB+ neurons, no significant differences were detected in the sizes of dendritic spine heads (tip diameter) between rats previously exposed to amphetamine or saline (Figure 2a; repeated measures ANOVA; between-group factor (amphetamine vs saline), F1, 5=0.23, NS; within factor (diameter), F5, 25=143.36, p<0.001; interaction (diameter × group), F5, 25=0.44, NS). Similarly, ΔFosB+ neurons sampled from amphetamine- and saline-exposed rats did not differ in their overall spine density (Figure 2c top; independent t-test; t5=0.80, NS) or spine tip density (Figure 2c bottom; independent t-test; t5=0.92, NS). As a control, we also analyzed the spine characteristics of ΔFosB− neurons in the NAcc. Again, the distribution of spine tip diameters was not different between rats that received amphetamine or saline 1 week earlier (Figure 2b; repeated measures ANOVA; between-group factor (amphetamine vs saline), F1, 4=1.88, NS; within factor (diameter), F5, 20=171.13, p<0.001; interaction (diameter × group), F5, 20=1.51, NS). Overall dendritic spine density (Figure 2c top; independent t-test; t4=1.13, NS) and spine tip density (Figure 2c bottom; independent t-test; t4=0.98, NS) also did not differ between the two groups.

We draw several conclusions from these results. First, high-resolution imaging and reconstruction techniques revealed that long-lasting changes in dendritic spine properties are not present after cessation from repeated sensitizing psychostimulant exposure regimens. Instead, as indicated below, we hypothesize that repeated drug exposure allows for long-term plasticity that facilitates rapid changes in dendritic spines specifically following reexposure to conditioned stimuli predictive of the drug. Second, we hypothesize that changes in dendritic spines are not observed in ΔFosB+ neurons because these cells mediate nonassociative sensitization rather than associative conditioning. Indeed, VTA amphetamine-exposed rats show enhanced behavioral responding to amphetamine along with increased ΔFosB levels and do so in the absence of conditioned behavioral responding to drug-paired stimuli (Figure 3b; Singer et al 2009).

Figure 3
figure 3

Exposure to amphetamine IP (left) but not in the VTA (right) produces locomotor conditioning and an amphetamine-paired stimulus-induced increase in c-Fos+ neurons in the NAcc shell. Testing was conducted 1 week following exposure. All rats were administered saline IP before placement in the open field. (a) Paired rats previously administered amphetamine IP displayed conditioned locomotion compared with Unpaired and Control rats. (b) This was not observed in Paired rats previously administered amphetamine into the VTA. Test results are shown as group mean (±SEM) locomotor counts over the 30 min test as well as group mean (+SEM) test total locomotor counts in the insets. Paralleling the locomotor results, an increase in c-Fos+ neurons was observed in the NAcc shell of Paired rats previously exposed to amphetamine IP (c; Control, 216.50±25.91 cells/rat) but not in the VTA (d; Control, 575.50±44.76 cells/rat). In contrast, all rats previously administered amphetamine IP displayed elevated ΔFosB expression in the NAcc independent of context (e), consistent with a role for ΔFosB+ neurons in the accrual of sensitization and for c-Fos+ neurons in the expression of conditioning (Control, 340.30±26.55 cells/rat). Data are shown as group mean (+SEM) number of neurons expressed as % of Control. (f) Regions sampled for cell counting in the NAcc shell. Cell counts from all six regions were combined for each rat. The representative images from a single plain of focus show staining of c-Fos, ΔFosB, and NeuN expression in the NAcc. The line drawing is from Paxinos and Watson (1998) and represents the caudal surface of a 100-μm-thick section. Number to the left indicates mm from bregma. *p<0.05; compared with Unpaired and Control (a, c); p<0.05; compared with Paired and Unpaired (e). Numbers at the base of each column indicate rats/group.

PowerPoint slide

Associative Conditioning

As neurons essential for the expression of behavioral sensitization (ΔFosB+) do not show long-lasting changes in dendritic spines (Figure 2), we next determined whether neurons participating in the expression of conditioned behavioral responses undergo dendritic alterations. To identify these neurons, rats underwent a discriminative learning procedure in which an open field was associated with the drug for some rats (Paired) and not associated with the drug for others (Unpaired). Unpaired rats received amphetamine elsewhere (Supplementary Figure S1). Control rats received saline injections in both environments. This drug conditioning regimen was conducted with IP or intra-VTA injections in separate groups of rats. Similar to what we have previously reported (Singer et al, 2009; Vezina and Stewart, 1990), when tested 1 week later, only Paired rats displayed a conditioned locomotor response after being reintroduced to the open field following a saline injection (Figure 3a; one-way ANOVA, 30-min total locomotion; F2, 12=7.41, p<0.01). The post hoc Scheffé tests demonstrated that Paired rats showed significantly enhanced conditioned locomotion relative to the Unpaired and Control rats (p<0.05) that did not differ from one another. In contrast, previous VTA amphetamine administration did not support the development of associative conditioning in Paired rats (Figure 3b; one-way ANOVA, 30-min total ambulation; F2, 14=0.92, NS). Altogether, these results indicate that drug–environment associations are formed following systemic, but not VTA, amphetamine exposure.

Based on these findings, we predicted that reexposure to the drug-paired open field would elicit conditioned cellular activation only in Paired rats demonstrating conditioned behavioral responding. We thus began by using immunohistochemistry to assess changes in expression of the immediate early gene product c-Fos following reintroduction to the drug-paired context and expression of conditioned locomotion. As expected, rats that showed conditioned behavioral responding (Paired, IP AMPH) also showed conditioned increases in c-Fos expression in the NAcc shell (Figure 3c; one-way ANOVA, c-Fos/NeuN, % Control; F2, 10=5.18, p<0.05; Scheffé post hoc, Paired (264.25±30.88 cells/rat) vs Unpaired (180.60±16.17 cells/rat) or Control (216.50±25.91 cells/rat), p<0.05). This effect was not expected nor was it observed in the NAcc core (data not shown; one-way ANOVA, c-Fos/NeuN, F2, 23=0.57, NS) as contextual stimuli are processed in the NAcc shell and not the NAcc core (Cruz et al, 2014a; Singer et al, 2014a, 2014b). Importantly, Paired rats that previously received intra-VTA amphetamine displayed neither conditioned locomotion nor increases in c-Fos expression in the NAcc shell (Figure 3d; one-way ANOVA, c-Fos/NeuN, % Control; F2, 14=0.18, NS; Paired (590.33±58.72 cells/rat), Unpaired (676.00±62.73 cells/rat), Control (575.50±44.76 cells/rat)). Thus, comparison of the results suggests that conditioned neuronal activation in the NAcc shell and behavioral responding to contextual stimuli are related.

In contrast, assessment of ΔFosB in rats subjected to the IP injection conditioning procedure showed that, despite the selective expression of locomotor conditioning in Paired rats (Figure 3a), ΔFosB expression in the NAcc shell was elevated in both Paired and Unpaired rats relative to Controls (Figure 3e; one-way ANOVA, ΔFosB/NeuN, % Control; F2, 23=5.49, p<0.05; Scheffé post hoc, Paired (452.06±30.78 cells/rat) and Unpaired (429.17±17.73 cells/rat) vs Control (340.30±26.55 cells/rat), p<0.05). A small number of c-Fos+ neurons were also found to be ΔFosB+ after reexposure to the amphetamine-paired environment (<10%; 40.73±5.82 cells/rat; 3-dimensional confocal analysis). This number is far lower than that observed in a previous report (>60%; Mattson et al, 2008) and >10-fold lower than the number of c-Fos+ neurons observed in the present experiments, indicating that context-dependent expression of c-Fos itself reflects activity in an ensemble of neurons that is critical for the expression of conditioned responding to contextual stimuli.

It has been documented that amphetamine conditioning is associated with increases in dendritic spine density in the NAcc (Singer et al, 2009). As c-Fos+ cells may be essential to the expression of conditioned responding (Figure 3), we investigated whether these neurons undergo dendritic changes upon reexposure to the drug-paired context. At 1 week after the induction of conditioning (IP injections), rats were reexposed to the open field for 30 min, killed, and their brains rapidly removed for subsequent c-Fos immunohistochemistry and DiI injections. Dendritic reconstructions revealed strikingly different results from those observed in ΔFosB+ neurons. Paired rats showed a significantly different distribution of dendritic spine head sizes relative to both Unpaired and Control rats (Figure 4a; repeated measures ANOVA; between-group factor (Paired, Unpaired, Control), F2, 9=3.45, p<0.05; within factor (diameter), F5, 45=285.48, p<0.001; interaction (diameter × group), F10, 45=7.39, p<0.001). Closer examination revealed that Paired rats showed a significant increase in the frequency of medium-sized spine tips (0.45–0.60 μm diameters) compared with Unpaired and Control rats (post hoc Scheffé tests, p<0.001). In contrast, no group differences were observed in spine tip diameters in c-Fos- neurons (Figure 4b; repeated measures ANOVA; between-group factor (Paired, Unpaired, Control), F2, 9=0.35, NS; within factor (diameter), F5, 45=218.45, p<0.001; interaction (diameter × group), F10, 45=0.88, NS). These findings obtained in NAcc shell MSNs are consistent with adaptations processing information necessary for contextual conditioning and do not argue against similar changes occurring in the NAcc core. Indeed, both changes in c-Fos expression (Kufahl et al, 2009) and in spine morphology (Gipson et al, 2013) have been reported in the NAcc core following the presentation of discrete drug-paired cues.

Figure 4
figure 4

Reexposure to amphetamine-paired contextual stimuli rapidly increases spine tip diameter and spine density in c-Fos+ but not c-Fos− neurons in the NAcc shell. One week following exposure, rats were placed in the open fields for the 30 min conditioning test. Brains were harvested immediately after the test. (a) The number of medium-sized (0.45–0.60 μm) spine tip diameters was significantly increased in the c-Fos+ neurons of Paired relative to Unpaired and Control rats. (b) This was not observed in c-Fos− neurons. Similarly, a significant increase in spine and spine tip density was observed in the c-Fos+ but not c-Fos− neurons of Paired relative to Unpaired and Control rats (c). Together, these findings indicate a rapid amphetamine-paired stimulus-induced increase in de novo spines with spine tip characteristics capable of enabling enhanced transmission. (d) DiI-labeled dendritic segment following deconvolution processing. Images were taken on a Leica SP5 confocal microscope (63 × oil with 4 × digital zoom). (e) Reconstruction of the dendritic segment using the Imaris software filament tracer module. (f) Surface reconstruction of the dendritic segment using Imaris. *p<0.05; ***p<0.001; relative to Unpaired and Control. Numbers in parentheses indicate rats, cells/group. Measures obtained from the cells of each rat (2–6/rat) were averaged and ANOVA conducted on these values. ANOVA conducted on unaveraged cell values yielded identical results, indicating homogeneity between cells for the different group measures (see Supplementary Data S7).

PowerPoint slide

Although changes in spine tip diameter are suggestive of rapid plasticity, these could reflect either the altered morphology of existing spines or the de novo formation of new spines. We, therefore, investigated whether reexposure to the open field resulted in altered overall spine density and tip density in c-Fos+ and c-Fos− neurons (Figure 4c). Relative to Unpaired and Control groups, c-Fos+ neurons sampled from Paired rats displayed conditioned increases in both dendritic spine density (one-way ANOVA, spines/μm; F2, 9=6.33, p<0.05; post hoc Paired vs Unpaired and Control, p<0.05) and tip density (one-way ANOVA, tips/μm; F2, 9=4.67, p<0.05; post hoc Paired vs Unpaired and Control, p<0.05). Finally, neither spine density (one-way ANOVA, spines/μm; F2, 9=0.07, NS) nor spine tip density (one-way ANOVA, spine tips/μm; F2, 9=0.51, NS) modifications were observed in c-Fos− neurons. Together, the selective increase in both spine and tip density in c-Fos+ neurons from Paired rats suggests that these are newly formed spines responding to conditioned stimuli. This effect was clearly not due to simple exposure to amphetamine as it was absent in Unpaired rats that received the same number of amphetamine injections as Paired rats during conditioning.

DISCUSSION

Repeated intermittent exposure to drugs of abuse, a pattern often associated with initial casual experimentation, can lead to long-lasting neuroadaptations capable of regulating drug-induced behaviors and escalating drug use in humans and animal models (Leyton and Vezina, 2013, 2014). Intermittent exposure, especially to psychostimulants like amphetamine and cocaine, is known to produce sensitization (Vezina, 2004), but as drugs are typically administered in the presence of a vast number of disparate environmental stimuli, the formation of multiple conditioned drug–stimulus associations is also ensured (Vezina and Leyton, 2009). Therefore, a major challenge for characterizing the neuroadaptations underlying associative conditioning is to study this form of plasticity in isolation. By comparing the effects of repeated systemic and intracranial VTA injections of amphetamine in a discrimination learning procedure, we were able to isolate neuroadaptations specific to amphetamine-induced associative conditioning.

Here we show that reexposure to a context previously associated with systemic amphetamine increases the number of c-Fos+ neurons in the medial NAcc shell (Figure 3) and increases spine density and the frequency of medium-sized spine tip diameters (Figure 4) selectively in these cells. As spine number and volume are proportional to the area of the postsynaptic density and to the synaptic content of AMPA receptors (Holtmaat and Svoboda, 2009), these adaptations observed in NAcc shell c-Fos+ MSNs are well positioned to enable heightened conditioned responding to a drug-paired context. Similar c-Fos-associated recruitment of AMPA receptors to mushroom-type spines was reported after fear conditioning (Matsuo et al, 2008).

As with repeated systemic amphetamine injections, repeated exposure to amphetamine in the VTA leads to enhanced ΔFosB expression in the NAcc and behavioral sensitization (Figure 1). However, unlike the former, VTA microinjections of the drug do not lead to associative memory formation as detected by the lack of conditioned locomotor responding to previously drug-paired stimuli and the failure of these stimuli to increase c-Fos expression in the NAcc (Figure 3). In addition, the number of ΔFosB+ neurons is increased in both Paired and Unpaired rats regardless of the environment in which they were exposed to systemic amphetamine (Figure 3e) and, unlike c-Fos+ neurons, they do not exhibit changes in dendritic morphology (Figure 2) even though a large number of drug-associated stimuli (eg, odor, tactile, visual experimenter, and procedural cues) were present in the period leading to tissue harvest. Notably, unlike c-Fos, ΔFosB levels are slow to accumulate (Hope et al, 1994) and do not match the fast time course of conditioned locomotion, making it difficult for the two to be associated. These findings are consistent with a role for c-Fos expression in associative learning (Cruz et al, 2014a; Matsuo et al, 2008; Ostrander et al, 2003) and further support a strong relationship between c-Fos-associated changes in dendritic morphology in the NAcc and the expression of conditioning. These effects are not observed in c-Fos- neurons that do not participate in the expression of conditioning. Indeed, as IP but not VTA amphetamine leads to conditioning, the observed changes in spine morphology likely reflect associative memories manifesting as the stimulus-induced recruitment of newly formed spines favoring medium-sized tip diameters. In contrast, the increase in ΔFosB observed here and by others (Colby et al, 2003; Kelz et al, 1999) likely reflects the encoding of information mediating nonassociative sensitization.

Importantly, manipulations, such as inhibiting cyclin-dependent kinase 5 (cdk5), which are known to prevent drug-induced spine proliferation in the NAcc (Norrholm et al, 2003), also prevent the development of conditioning but spare sensitization (Singer et al, 2014b). Although we did not observe spine alterations in ΔFosB+ neurons in the present study, it remains possible under some conditions for changes to occur in these neurons that enable the development of conditioned associations, as ΔFosB is a transcription factor for the cdk5 gene. For example, morphological changes may occur in some ΔFosB+ neurons coexpressing c-Fos or at a time point different from that examined in the present experiments. Indeed, genetic overexpression of ΔFosB does increase spine density and the existence of immature spines in the NAcc shortly after a cocaine injection (Grueter et al, 2013; Maze et al, 2010). However, the viral-mediated nature of these changes, potentially elevating ΔFosB expression above endogenous levels and in neurons not participating in drug cue-encoding ensembles, make these findings difficult to interpret in light of the present results. The goal of the present experiments was to assess the ability of drug-paired stimuli to induce morphological changes in the dendritic spines of select neurons in the NAcc. Clearly, future experiments will need to characterize further the time course of these effects and the sequence of signaling events underlying them. Our current findings support a strong relationship between rapid cue-induced changes in the dendritic spine morphology of c-Fos+ neurons in the NAcc and the expression of conditioned responding to drug-paired stimuli.

The selective increase in the frequency of spines with medium tip diameters (0.45–0.60 μm) in c-Fos+ neurons in the NAcc shell of Paired rats is similar to changes observed in this dendritic spine size in randomly selected neurons in the NAcc core following cocaine or exposure to discrete cues (Dumitriu et al, 2012; Gipson et al, 2013). Interestingly, Golgi–Cox staining used in previous studies is also most sensitive for detecting changes in these medium-sized spines (Shen et al, 2009). Together, these findings suggest the importance of this particular spine subtype in mediating behavioral responses. It is unknown why these changes in spines were restricted to medium-sized tip diameters, but it may be that these spine head sizes are volatile and susceptible to rapid changes based on increased or decreased synaptic input (Matsuo et al, 2008).

A common view in models of learning is that changes in the morphology of dendritic spines and the long-term stabilization of these changes embody the neural representation of persistent memories (Koleske, 2013; Lamprecht and LeDoux, 2004; Muñoz-Cuevas et al, 2013; Roberts et al, 2010; Xu et al, 2009). The present findings, as well as the transient effects reported elsewhere (Gipson et al, 2013), are not consistent with these models according to which repeated drug exposure leads to long-lasting, static changes in spine morphology. In the present experiments, both Paired and Unpaired rats were exposed to the same number of systemic amphetamine injections during conditioning, yet only Paired rats showed a conditioned response when tested in the drug-paired context. Thus, the increases in spine density and spine tip diameter observed in the NAcc of Paired rats following reexposure to the drug-paired stimuli were rapid (~30 min), again, as they were not observed in Unpaired rats tested in the absence of the drug-associated stimuli. Rather than static changes in spine morphology, these findings suggest that associative conditioning requires neuroadaptations in NAcc MSNs that can allow drug-paired stimuli, when present, to evoke rapid conditioned changes in dendritic spines capable of enabling the expression of a behavioral conditioned response.

FUNDING AND DISCLOSURE

This research was funded by the US National Institutes of Health (NIH) Grants R01 DA09397 (to PV), T32 DA07255 (to BFS), and F31 DA030021-01A1 (to BFS). Work by BFS was partially funded by the Chicago Biomedical Consortium with support from The Searle Funds at The Chicago Community Trust. The authors declare no conflict of interest.