Drug seeking is significantly regulated by drug-associated cues and associative learning between environmental cues and cocaine reward is mediated by dopamine transmission within the nucleus accumbens (NAc). However, dopamine transmission during early acquisition of a cue–cocaine association has never been assessed because of the technical difficulties associated with resolving cue-evoked and cocaine-evoked dopamine release within the same conditioning trial. Here, we used fast-scan cyclic voltammetry to measure sub-second fluctuations in dopamine concentration within the NAc core and shell during the initial acquisition of a cue–cocaine Pavlovian association. Within the NAc core, cue-evoked dopamine release developed during conditioning. However, within the NAc shell, the predictive cue appeared to cause an unconditioned decrease in dopamine concentration. The pharmacological effects of cocaine also differed between sub-regions, as cocaine increased phasic dopamine release events within the NAc shell but not the core. Thus, real-time measurements not only revealed the initial development of a conditioned neurochemical response but also demonstrated differential phasic dopamine transmission patterns across NAc sub-regions during the acquisition of a cue–cocaine association.
Environmental cues paired with cocaine for just 1 day increase cocaine seeking behavior nearly 1 year later (Ciccocioppo et al., 2004). This demonstrates the importance of associations formed during initial drug exposure. However, the neural regulation of early acquisition of cue–cocaine associations is poorly understood. Although it is known that dopamine transmission within the nucleus accumbens (NAc) is critical for the acquisition of a Pavlovian association (Kelley, 2004), fluctuations in dopamine concentration during the first day of cue–cocaine conditioning have never been measured. This is primarily because, until recently, it has been technically impossible to resolve cue-evoked and cocaine-evoked dopamine release within the same conditioning trial. Specifically, the neurochemical consequences of cue presentation and the pharmacological effects of cocaine are separated by just seconds (Stuber et al., 2005a) whereas the temporal resolution of traditional measurement technology (microdialysis) is in the range of minutes (Pan et al., 1991; Watson et al., 2006).
Here, we circumvent past technical limitations by using fast-scan cyclic voltammetry (FSCV) (Wightman, 2006) to measure sub-second fluctuations in dopamine concentration ([DA]) within the NAc core and shell during the initial acquisition of a cue–cocaine association. As cue–cocaine associations are formed via classical conditioning mechanisms (Robinson & Berridge, 2008), we employed a recently described Pavlovian conditioning design in which presentation of a discrete cue was presented with non-contingent intravenous (i.v.) cocaine delivery (Uslaner et al., 2006). This paradigm ensured equal number and timing of cue–cocaine pairings on the first day of Pavlovian conditioning in drug-naive rats. Rapid dopamine measurements during this paradigm readily distinguished between the development of conditioned dopamine transmission associated with the predictive cue (Phillips et al., 2003; Stuber et al., 2005b) and increased dopamine transmission resulting from pharmacological effects of cocaine (Cheer et al., 2004; Heien et al., 2005; Aragona et al., 2008; Sombers et al., 2009).
With respect to conditioned dopamine transmission, the current study directly addressed a long-standing controversy regarding phasic dopamine signaling following the presentation of a conditioned stimulus. Prior to real-time neurochemical measurements, development of conditioned dopamine signaling has been monitored with extracellular electrophysiology measures of dopaminergic neurons (Pan et al., 2005). Such studies have suggested that conditioned stimuli cause a phasic increase in firing among the great majority of dopaminergic neurons and it is often assumed that this results in a uniform increase [DA] across sub-regions of the NAc (Schultz, 1998). Conversely, microdialysis studies suggest conditioned stimuli differentially increase [DA] across NAc sub-regions (Ito et al., 2000; Bassareo et al., 2007).
Several methodological issues may explain this discrepancy. For example, electrophysiological identification of a dopaminergic phenotype is unambiguous only in anesthetized subjects (Ungless et al., 2004; Margolis et al., 2006b) and release from dopaminergic neurons depends on firing history (Montague et al., 2004). Further, dopaminergic neurons exist in sub-populations that project to different forebrain locations (Margolis et al., 2006a; Lammel et al., 2008). Measurement of [DA] in forebrain terminal fields avoids such concerns, but microdialysis utilizes probes that sample [DA] over a rather large area and this technique lacks the temporal resolution to examine phasic dopamine signaling (Robinson et al., 2003). Conversely, FSCV employs sensors that sample from discrete microenvironments (Wightman et al., 2007) downstream from specific midbrain dopaminergic sub-populations (Ikemoto, 2007) and measures phasic dopamine communication on a similar time scale to single unit recording (Hyland et al., 2002; Schultz, 2002). Thus, FSCV was used to resolve the controversy regarding conditioned phasic dopamine communication across distinct terminal fields: in this case, the NAc core and shell.
With respect to the pharmacological effects of cocaine, we have recently used FSCV to show that (in addition to slowing dopamine uptake; Giros et al., 1996), i.v. cocaine administration also evokes a direct increase in phasic dopamine release events within the NAc shell, but not the core (Aragona et al., 2008). However, our previous study only tested a single high-dose cocaine infusion (Aragona et al., 2008). Here, we examined if cocaine-evoked release events within the NAc shell continue to occur following multiple cocaine infusions using a lower dose consistent with those that are self-administered. The current study demonstrates regionally specific dopamine transmission patterns during the formation of a cue–cocaine association, with conditioned (cue-evoked) dopamine transmission specific to the NAc core and unconditioned (cocaine-evoked) dopamine transmission specific to the NAc shell. Thus, our data support recent evidence for anatomically and functionally separate mesolimbic dopamine pathways (Ikemoto, 2007) and suggest that these systems mediate distinct aspects of cue–cocaine associations.
Animals and surgery
Male Sprague–Dawley rats were purchased with implanted jugular vein catheters (n =24, ∼375 g, Charles River Laboratories, Wilmington, MA, USA). Rats were anesthetized with intra-muscular ketamine hydrochloride (100 mg/kg) and xylazine hydrochloride (20 mg/kg). Bipolar stimulating electrodes (Plastics One, Roanoke, VA, USA) [placed in the ventral tegmental area (VTA); 5.2 mm posterior, 1.0 mm lateral, 7.5 mm ventral relative to bregma] and Ag/AgCl reference electrodes (placed in contralateral cortex) were secured as described in detail elsewhere (Phillips et al., 2003; Wightman et al., 2007). Guide cannula (Bioanalytical Systems, West Lafayette, IN, USA) were aimed at the NAc core (1.3 mm anterior, 1.3 mm lateral, −2.5 mm ventral) or shell (1.7 mm anterior, 0.8 mm lateral, −2.5 mm ventral; relative to bregma). Following the experiment, an electrolytic lesion was made at the micro-drive setting used during the experiment and verified histologically. Experiments were in accordance with the NIH Guide for Care and Use of Animals, and approved by the Institutional Animal Care and Use Committee at the University of North Carolina at Chapel Hill.
Fast-scan cyclic voltammetry
Following 5–7 days’ recovery from surgery, glass-encased carbon-fiber electrodes were lowered using a locally constructed micro-drive (University of North Carolina at Chapel Hill, Department of Chemistry Instrument Shop) and positioned where both electrically evoked (biphasic pulses, 2 ms per phase, 24 pulses, 60 Hz, 120 μA) and transients were detected (Wightman et al., 2007). Waveform generation and processing, current transduction, and data collection and filtering have been described in detail elsewhere (Wightman et al., 2007). Background subtraction employed the cyclic voltammograms with the lowest current within the 10-s pre-cue/infusion baseline period. Current was converted to [DA] using principal component regression as previously described (Heien et al., 2005). A dopamine transient was defined as a five-fold or greater increase in [DA] relative to the root-mean-square noise value taken from the same electrode. Average transient duration is ∼1 s (Wightman et al., 2007; Aragona et al., 2008) and due to their brief duration, they are routinely described as phasic (Phillips et al., 2003). Events below 30 nm were excluded from the analysis because events below this magnitude were not reliably detected across electrodes. Transient frequency was determined with Mini Analysis (Synaptosoft, Decatur, GA, USA). Average transient frequency in the current study is similar to our recent experiments (Cheer et al., 2007; Aragona et al., 2008) but higher than our earlier work (Robinson et al., 2002; Stuber et al., 2005a) and this is due to improvements in sensitivity and conducting experiments in locations where naturally occurring dopamine transients are detected (Robinson & Wightman, 2007; Wightman et al., 2007). Assessment of acute alteration in mean [DA] during conditioning trials was restricted to a 90-s sampling period because electrode drift prevents reliable analysis for longer times for many electrodes (Heien et al., 2005). Maximal [DA] within the sampling window is referred to as ‘peak’ [DA]. In the case of cocaine-evoked increases in dopamine signaling, peak [DA] is the result of transients superimposed on gradual increases in [DA] that is due to blockade of terminal dopamine transporters (Heien et al., 2005; Cheer et al., 2007; Aragona et al., 2008). ‘Cue-evoked dopamine release’ is obviously also a transient event, but as it appears to be the result of synchronous burst firing of dopaminergic neurons (Aragona et al., 2008; Sombers et al., 2009), it is given this separate designation and the maximal [DA] within 2 s of cue onset represents its peak. Following the experiment, electrodes were calibrated as previously described (Wightman et al., 2007). All chemicals were purchased from Sigma-Aldrich (St Louis, MO, USA).
Pavlovian conditioning (cocaine)
Consistent with established parameters for Pavlovian conditioning utilizing cocaine reward (Uslaner et al., 2006), a compound stimulus (cue light and tone) was simultaneously presented with the onset of a non-contingent i.v. infusion of cocaine (0.16 mg per infusion over 3 s; ∼0.5 mg/kg per infusion). Cue onset and i.v. infusion onset were simultaneous because cue onset was designed to predict central cocaine effects, not the infusion itself. The duration of the predictive cue was 20 s so that cue presentation overlapped with the previously established time of increased dopamine transmission by cocaine (Aragona et al., 2008). The conditioning session consisted of 30 cue–cocaine pairings with an average inter-trial interval of 5 min, for a total of 30 cocaine infusions in the conditioning session.
Previous studies using a similar design have shown that animals exhibit approach responses toward cocaine predictive cues (termed sign-tracking or autoshaping behavior) with no attempt to consume the conditioned stimulus (Uslaner et al., 2006). Approach behavior was analysed using offline video analysis and was defined as the subject bringing its nose to within ∼1 cm of the cue light (Uslaner et al., 2006) and the duration the subject spent sniffing the light was used to calculate cue ‘investigation’. While novel cue investigation was observed in previous studies (Uslaner et al., 2006), our subjects did not show cue investigation when i.v. saline was given instead of cocaine (n =3; data not shown). As such, our subjects did not receive environmental or cue habituation described in previous experiments (Uslaner et al., 2006).
For Pavlovian conditioning using i.v. cocaine, it is critical for the inter-trial interval (ITI) to be long enough to allow the animal to distinguish between the acute effects of the most recent cocaine infusion and the residual effects of the preceding cocaine infusions (Uslaner et al., 2006). Given that accumbal dopamine transmission does not begin to decline until 2–3 min following i.v. infusion (Aragona et al., 2008) the Pavlovian ITI must exceed this duration. However, the effects of i.v. cocaine infusion are not completely absent for ∼45 min (Pan et al., 1991) and an ITI of this length would render the current study impossible given that stable performance of carbon-fiber electrodes limits experimental duration. The ITI for self-administration of the dose of cocaine used in this present study (0.16 mg per infusion) is ∼3.3 min (Carelli & Deadwyler, 1996). To minimize stereotypy and because additional cocaine metabolism is desired in this paradigm (Uslaner et al., 2006), the ITI was increased to a mean duration of 5 min (4.75, 5.00 and 5.25 min).
Pavlovian conditioning (sucrose)
Male rats (n =7) received bilateral intra-oral catheters and voltammetry surgery on the same day. During the experimental day, the behavioral chamber was illuminated by a light on the side of the chamber. After a 2-min delay, the light was extinguished and a cue was presented. The cue consisted of a tone stimulus (65 dB, 2900 Hz) paired with illumination of a house light at the top of the chamber presented for 6 s. An infusion pump delivered 200 μL of 0.3 m sucrose intra-orally over 6 s immediately following the cue. Intra-oral delivery of sucrose by an infusion pump is slower than sucrose delivery using a solenoid (Roitman et al., 2008). Following the infusion, there was another 2-min delay before the next trial and a total of 30 trials were administered.
As described in our previous study (Aragona et al., 2008), changes in [DA] and transient probability across conditioning trials were assessed using a linear mixed model, which was chosen based on its ability to properly handle data in which observations are not independent (such as repeated-measures data), correctly model correlated error terms and incorporate random subject effects. To determine mean changes of [DA], measurements were averaged into 2.5-s time bins for the cue–cocaine conditioning study and 2.0 s for cue-sucrose conditioning. To estimate transient probability, we determined whether a transient occurred within 2.5-s bins across the conditioning trials. A transient was assigned to a specific bin depending on the time of its peak. For both [DA] and transient probability, bins were treated as the dependent variable, with time serving both as a repeated measure and a fixed effect variable. Estimation of time bins at which concentration differed from the pre-cue/infusion baseline was achieved by construction of simple slopes and comparison with the reference period (10 s prior to cue/infusion onset for cue–cocaine conditioning and 6 s prior to cue onset for cue–sucrose conditioning) to obtain t values. Significance was assigned if values crossed a critical t value (±2 for all studies), which corresponded to an α level of 0.05. For dopamine concentration data, an auto-regressive compound variance structure was used to model the data due to correlations in the variance of data points that were closer together in time. Within this model, the degrees of freedom are adjusted based on the distance between comparisons, and therefore are often assigned non-integer values. In contrast, transient probability data did not exhibit correlations in the variance of temporally related points and were therefore modeled using a compound symmetry covariance structure, which does not modify degrees of freedom. Baseline dopamine transients within the NAc core and shell as well as cue-evoked dopamine release and dopamine release evoked by a four-pulse electrical stimulation at 20 Hz were compared using independent-samples t-tests. Statistical significance was designated at α = 0.05 and all statistical analyses were carried out in SPSS version 14 for Windows (SPSS, Chicago, IL, USA).
Carbon-fiber microelectrodes (6 × 100 μm) were secured in portions of the NAc core or shell (Fig. 1A) that supported electrically stimulated dopamine release (Phillips et al., 2003) as well as dopamine ‘transients’ (Wightman et al., 2007). A dopamine transient is defined as a naturally occurring increase in [DA] that is five times greater than the root-mean-square noise (Heien et al., 2005) and is indicative of phasic dopamine release (Aragona et al., 2008). Prior to cue–cocaine conditioning, a 10-min recording period showed that frequency of dopamine transients did not significantly differ (independent samples t-test) between the NAc core and shell (t3 = 3.16, P =0.119; 6.2 + 1.2 transients per minute; collapsed across sub-regions). Although baseline transient frequency did not significantly differ between the core (4.3 + 1.2) and shell (7.8 + 3.8), the P-value is relatively low and therefore caution must be taken with respect to affirming that these regions do not differ in baseline dopamine transmission. However, as the lack of sub-region differences in baseline transient frequency replicates our previous finding (Aragona et al., 2008), we conclude that the two sub-regions do not differ in this regard.
Cue–cocaine Pavlovian conditioning was then initiated. The first cocaine infusion (0.16 mg; ∼0.5 mg/kg) increased peak dopamine concentration (Δ[DA] = 229 + 48 nm in the NAc shell and 141 + 31 nm in the NAc core; relative to background subtraction) to magnitudes consistent with our previous study using a similar dose (Heien et al., 2005). Initial cocaine-evoked increases in dopamine transmission, in both the NAc shell and core, can be seen in mean [DA] data from the first conditioning trial (Fig. 2; dashed box). In subsequent trials, background subtraction removes the residual effects of previous cocaine infusions. Therefore, remaining analysis of increased dopamine signaling by cocaine is focused on acute increases by the cocaine infusions that occur within specific conditioning trials.
Changes in [DA] were continuously assessed over 90-s sampling periods (Heien et al., 2005) that began 10 s prior to cue/infusion onset and this portion served as the baseline to determine acute changes in [DA] by the predictive cue and cocaine infusions. As averaging FSCV data washes out dopamine transients that are not time locked to a specific event (Roitman et al., 2008), representative concentration traces (Fig. 1B–E) are required to demonstrate the phasic nature of dopamine transmission detected by FSCV.
We have recently demonstrated that cocaine directly increases phasic dopamine release events within the NAc shell, but not the core (Aragona et al., 2008). Consistent with this previous study, cocaine infusions did not acutely increase phasic dopamine signaling within the NAc core in the seconds following i.v. infusion compared with the pre-cue/infusion baseline (Fig. 1B and C). However, within the NAc shell, cocaine infusion increased the probability and magnitude of dopamine transients, beginning at ∼20 s following drug infusion both early and late within the conditioning session (Fig. 1D and E). Thus, acute phasic increases in dopamine transmission were detected even after multiple cocaine infusions (i.e. in the presence of the elevated levels of [DA] described by microdialysis studies; Pettit & Justice, 1989). Importantly, the time-point of this increase is identical to that described in our previous study that did not utilize a predictive cue (Aragona et al., 2008), indicating that cocaine-evoked release events are a pharmacological effect.
The sub-second temporal resolution of FSCV allows for cue-evoked dopamine transmission to be distinguished from cocaine-evoked increases in [DA] (Stuber et al., 2005a). Within the NAc core, cue onset did not alter dopamine transmission early in the conditioning session (Fig. 1B). However, following additional cue–cocaine pairings, cue onset evoked a robust increase in [DA] within the NAc core (Fig. 1C). These data demonstrate that cue-evoked dopamine release can emerge in one Pavlovian conditioning session using cocaine reward. Conversely, within the NAc shell, cue onset did not increase dopamine transmission either early (Fig. 1D) or late (Fig. 1E) within the conditioning session. Thus, cue-evoked dopamine release, during this Pavlovian design using cocaine reward, occurs within the NAc core but not the NAc shell.
Three-dimensional representation of mean changes in [DA] across conditioning trials provides a detailed assessment of the development of real-time dopamine transmission patterns during the establishment of this cue–cocaine association (Fig. 2A and C). Within the NAc core, cue onset did not increase [DA] early within the session (Fig. 2A). However, cue onset evoked higher [DA] values across conditioning trials and cue-evoked dopamine release was most robust toward the end of the conditioning session (Fig. 2A). For statistical analysis, change in [DA] was organized into 2.5-s bins (Fig. 2B and D) and data were averaged into the first (1–10) middle (11–20) and last (21–30) blocks of conditioning trials (the 10 s prior to cue/infusion onset served as the baseline). Within the NAc core, a linear mixed-model analysis reveals that [DA] was significantly increased during presentation of the predictive cue during the middle and last conditioning blocks (Fig. 2B; Trials 11–20, t125.4 = 7.21, P <0.001 at peak; Trials 21–30, t93.98 = 6.76, P <0.001 at peak). Importantly, this was due to cue-evoked dopamine release, as cue onset induced an immediate and significant increase in transient probability, including transients > 100 nm (Fig. 3A; Trials 11–20, t109.88 = 5.697, P <0.001 at peak; Trials 21–30, t114.8 = 5.433, P <0.001 at peak).
Cue-evoked dopamine release within the NAc core reached a maximum concentration following 30 conditioning trials (Fig. S1, A) and this was not due to the detection capabilities of the electrode (Fig. S1, B). Rather, this appeared to be due to development of synchronous population burst firing of dopaminergic neurons, as an independent samples t-test revealed no difference between peak cue-evoked dopamine release (114 + 38 nm) and dopamine release evoked by electrical stimulation of dopaminergic neurons (122 + 14 nm; t3 = 0.072, P = 0.947; n =4, two from the NAc shell, two from the NAc core, no difference across sub-regions) that mimicked population burst firing (four-pulse stimulation at 20 Hz) (Pan et al., 2005).
After cue-evoked dopamine release was established, presentation of the predictive cue alone (i.e. in the absence of cocaine infusion) also resulted in time-locked dopamine release within the NAc core (Fig. S1, C). This is consistent with previous studies (Phillips et al., 2003) (which measured exclusively within the NAc core) and indicates that cue-evoked dopamine release was driven by conditioned sensory stimulation (Dommett et al., 2005; Pan & Hyland, 2005) and not an interoceptive signal provided by i.v. infusion (Wise et al., 2008). We have previously established that conditioning is indeed necessary for sensory input to evoke dopamine release (Phillips et al., 2003). However, to confirm that this is the same for Pavlovian conditioning, two additional subjects received 30 ‘un-paired’ cue–cocaine trials. These subjects did not show time-locked dopamine release upon cue presentation (Fig. S2). Finally, there was a significant positive correlation between the mean change in cue-evoked dopamine release and Pavlovian approach behavior (Uslaner et al., 2006) (R228 = 0.445, P <0.001; Fig. S1, D). These data confirm that cue-evoked dopamine release within the core was a conditioned effect and suggest that this signal carried motivational significance (Robinson & Berridge, 2003).
Within the NAc core, linear mixed model analysis did not show a significant increase in [DA] compared with the pre-cue/infusion baseline following multiple cocaine infusions (Fig. 2A and B; all t values < 2; P >0.05). This indicates that while multiple cocaine infusions maintain a global elevation in [DA] within the NAc core (Stuber et al., 2005a) they did not acutely increase [DA] within this region. The lack of an acute increase in [DA] following subsequent cocaine infusions is indicative of the failure of cocaine to increase transient probability in the seconds following cocaine infusion compared with the pre-cue/infusion baseline (Fig. 3A; all t values < 2; P >0.05) and this is consistent with our previous study showing that cocaine does not directly increase phasic dopamine release events in the NAc core (Aragona et al., 2008).
With respect to both cue- and cocaine-evoked dopamine transmission, dopamine signaling within the NAc shell (Fig. 2C) showed nearly the opposite pattern to that seen within the NAc core (Fig. 2A). Within the NAc shell, [DA] was lowest during cue presentation compared with the pre-cue/infusion baseline (Fig. 2C). Linear mixed-model analysis shows that cue presentation significantly decreased [DA] within the NAc shell (Fig. 2D) and that this decrease was present during the first conditioning block, suggesting that the cue-evoked decrease was unconditioned (t value < −2 for at least one comparison during the cue period for each trial block, P <0.05). Transient probability was not different during the cue period (Fig. 3B; all t values between −2 and 2; P >0.05), suggesting that the decrease was not due to detectable changes in dopamine transients. However, linear mixed-model analysis shows that the probability of high-concentration transients (defined as transients > 100 nm) was significantly increased in the last conditioning block (Fig. 3B; t128 = 3.082, P =0.003 at peak). As a result, there was an acute increase in [DA] during the final conditioning block (Fig. 2D; t47.29 = 3.654, P =0.001 at peak), which is defined as a significant increase compared with the pre-cue/infusion baseline.
We next determined if this cue-evoked decrease in [DA] within the NAc shell generalized to a similar conditioning paradigm using sucrose reward. In a separate group of subjects (n =7), a predictive cue was presented 6 s prior to non-contingent intra-oral delivery of sucrose that was readily ingested. Consistent with cue–cocaine pairings, cue presentation resulted in a time-locked decrease in mean [DA] within the NAc shell (Fig. 4A; t123.65 = −2.438, P =0.016 at trough) that was present during the first block of cue–sucrose pairings (Fig. 4B; t122.62 = −2.424, P =0.017 at trough). Together, these data demonstrate that cue-evoked decreases in [DA] within the NAc shell during early Pavlovian acquisition generalized to both drug and natural reward.
Mean [DA] traces for the NAc core and shell during the last cue–cocaine conditioning block (final ten pairings) were superimposed to emphasize the differential dopamine transmission patterns between the NAc core and shell (Fig. 5). Despite nearly opposite directionality in transmission patterns, peak [DA] following cue-evoked dopamine release within the NAc core (114 + 38 nm) and cocaine-evoked dopamine release within the shell (131 + 18 nm) were similar in magnitude. It is possible that peak [DA] within the NAc shell resulting from cocaine-evoked dopamine release required the preceding cue-evoked decrease in [DA] because this would reduce autoreceptor activation (Sulzer & Pothos, 2000) and thus potentiate subsequent release events. Finally, peak cocaine-evoked [DA] within the NAc shell represents a 3.9-fold increase in the signal to baseline, relative to the lowest [DA] value in the conditioning trial, i.e. during the cue-evoked decrease in dopamine signaling. This is similar to the 3.7 signal to baseline achieved by cue-evoked dopamine release within the NAc core, relative to [DA] during the pre-cue/infusion baseline. Thus, the cue-evoked decrease in [DA] in the NAc shell allowed cocaine-evoked dopamine release within the shell (implicated in primary reinforcement by drugs of misuse) to achieve a similar signal to baseline increase as cue-evoked dopamine release within the core (implicated in conditioned reinforcement).
Real-time fluctuations in [DA] were measured within the NAc core and shell during the first session of a Pavlovian conditioning paradigm in which a discrete cue was paired with non-contingent i.v. infusions of cocaine. The sub-second measurements provided by FSCV unambiguously distinguished between cue-evoked and cocaine-evoked alterations in dopamine transmission, which allowed novel characterization of the initial development of dopamine transmission patterns during the acquisition of a cue–cocaine association. Within the NAc core, the predictive cue had no effect on dopamine transmission early in the conditioning session. However, cue onset evoked phasic dopamine release toward the end of the same conditioning session. Cue-evoked dopamine release developed as a function of conditioning as it was correlated with Pavlovian approach behavior and was not observed in unpaired subjects. In contrast to the NAc core, cue onset decreased [DA] within the NAc shell. Although initial cocaine infusions elevated [DA] levels in both the NAc core and shell compared with pre-drug levels, subsequent cocaine infusions acutely increased phasic dopamine release events only within the NAc shell. Together, the current data demonstrate dramatic regional specificity in phasic dopamine signaling during the formation of a Pavlovian association utilizing cocaine reward with conditioned dopamine transmission within the NAc core and unconditioned dopamine transmission within the NAc shell.
Conditioned dopamine transmission within the NAc core
In the current study, we show that cue-evoked dopamine release within the NAc core fully developed in just one session of Pavlovian conditioning utilizing cocaine reward. Consistent with our previously published self-administration study (Phillips et al., 2003), cue onset did not increase [DA] in unpaired subjects. This indicates that cue-evoked dopamine release required conditioning and was not merely the result of enhanced dopamine detection following blockade of dopamine transporters (Robinson & Wightman, 2004). Additionally, there was a significant positive correlation between cue-evoked dopamine release within the NAc core and investigation of the conditioned stimulus (Uslaner et al., 2006). This relationship confirms that the cue served as a conditioned stimulus and that cue-evoked dopamine release was a conditioned effect.
Extracellular electrophysiology studies show that presentation of a conditioned stimulus results in synchronous population bursting of dopaminergic neurons located approximately 1 mm off the midline (Pan et al., 2005) (i.e. neurons that primarily project to the NAc core; Ikemoto, 2007). In the current study, electrical stimulation of dopaminergic neurons at parameters within the physiological range of population bursting (four pulses at 20 Hz) (Pan et al., 2005) resulted in an equivalent increase in [DA] compared with cue-evoked dopamine release. This suggests that cue-evoked dopamine release was due to the development of synchronous bursting among dopaminergic neurons projecting to the NAc core (Ikemoto, 2007). Consistent with this hypothesis, dopaminergic neuron bursting is mediated by glutamate (Overton & Clark, 1997) and glutamatergic projections (Geisler et al., 2007) are activated by visual/auditory stimulation to increase dopaminergic neuron firing (Dommett et al., 2005; Pan & Hyland, 2005). Furthermore, we have recently demonstrated that cue-evoked dopamine release that predicts intracranial self-stimulation was abolished by blockade of NMDA receptors within the VTA (Sombers et al., 2009). Additionally, a recent FSCV study demonstrated that cue-evoked dopamine release within the NAc core fully developed on the third conditioning session using a Pavlovian conditioning paradigm that utilized a natural reward (sucrose pellets) and this occurred on the same day as enhancement of glutamatergic synapses on midbrain dopaminergic neurons (Stuber et al., 2008). It is known that a single cocaine exposure causes a similar alteration in glutamate synapses (Ungless et al., 2001) and thus may be related to the more rapid development of cue-evoked dopamine release with Pavlovian conditioning utilizing cocaine reward.
Unconditioned dopamine transmission within the NAc shell
In contrast to the NAc core, cue onset appeared to cause a brief unconditioned decrease in [DA] within the NAc shell during Pavlovian conditioning utilizing both non-contingent cocaine and sucrose reward. This regional specificity is in contrast to electrophysiology studies suggesting that conditioned stimuli increase firing in the majority of dopaminergic neurons and thus increase [DA] across all striatal regions (Schultz, 2002). However, caution must be taken when attempting to infer terminal dopamine communication solely from neuronal activity for several reasons. First, identification of a dopaminergic phenotype is reliable only in anesthetized preparations (Margolis et al., 2006b). Second, while certain sub-populations of dopaminergic neurons are disproportionately sampled in freely moving electrophysiology studies, recent electrophysiology studies support functionally distinct sub-populations of dopaminergic neurons (Brischoux et al., 2009; Matsumoto & Hikosaka, 2009). Behavioral studies have focused primarily on ‘conventional’ dopaminergic neurons (Lammel et al., 2008) that preferentially project to the NAc core (Hyland et al., 2002; Pan et al., 2005), whereas dopaminergic neurons that project to the NAc shell (Ikemoto, 2007) and have not been examined in freely moving electrophysiology studies. Finally, there is not a one-to-one correspondence between action potential generation and terminal dopamine release (Montague et al., 2004) and [DA] can be significantly modulated at the terminal level (Cragg, 2006; Britt & McGehee, 2008).
By directly measuring changes in terminal [DA] with FSCV, the current study avoided these concerns and revealed that cue onset decreases [DA] within the NAc shell in this paradigm. This finding supports recent descriptions of functionally distinct dopamine projection pathways between the NAc core and shell (Ito et al., 2004; Aragona et al., 2006; Di Chiara & Bassareo, 2007; Ikemoto, 2007; Liu et al., 2008). It is only recently that we have demonstrated that it is possible to detect decreases in [DA] using FSCV (Roitman et al., 2008). FSCV is a differential technique in which raw data show phasic increases in [DA] relative to a background-subtracted time point (Wightman, 2006). However, averaging across trials can reveal time-locked decreases in [DA] under certain behavioral situations, such as intra-oral infusion of an aversive tastant (Roitman et al., 2008). Future studies are needed to address if this shell-specific decrease is primarily due to decreased dopaminergic neuronal firing (Ungless et al., 2004) or to differences in terminal modulation among regionally specific afferents (Zahm, 2000).
Consistent with our previous study, which used a single high-dose infusion of cocaine (Aragona et al., 2008), the lower dose infusions used in the current study also caused unconditioned increases in phasic dopamine release events within the NAc shell. Sucrose did not significantly increase [DA] above baseline values and this is most likely due to the relatively slow intra-oral infusion rate by pump infusion (current study) compared with the rapid solenoid delivery used in our previous study (Roitman et al., 2008), which did result in a significant increase in [DA]. Acute increases in [DA] by cocaine are in addition to the global elevation in [DA] described by microdialysis studies (Pettit & Justice, 1989) and are critical for the timing of drug intake in rats self-administering cocaine (Stuber et al., 2005a). Previous studies from our laboratory show that cocaine-evoked increases in transient frequency within the NAc shell are due to a true increase in the number of phasic dopamine release events and are not merely the result of enhanced dopamine detection (Aragona et al., 2008; Sombers et al., 2009). The magnitude of increased transient frequency is equal to that of autoreceptor blockade (Aragona et al., 2008), a manipulation known to increase burst firing of dopaminergic neurons (Andersson et al., 1995). Cocaine-evoked dopamine transients within the shell were eliminated by infusion of GABA agonists (Aragona et al., 2008) and lidocaine directly into the VTA (Sombers et al., 2009) demonstrating that this increase is indeed due to increased firing of dopaminergic neurons. Finally, burst firing is mediated by glutamate activation of NMDA receptors (Overton & Clark, 1997), peripheral cocaine administration increases glutamate within the VTA (Kalivas & Duffy, 1995; Wise et al., 2008; You et al., 2008), and blockade of NMDA receptors within the VTA reduces the frequency of dopamine transients within the NAc shell (Sombers et al., 2009). Together these data strongly suggest that the unconditioned increase in phasic dopamine release events within the NAc shell following cocaine infusion is due to a brief increase in burst firing among dopaminergic neurons projecting to the NAc shell.
While conditioning is not required for cocaine-evoked dopamine release (Aragona et al., 2008; Sombers et al., 2009), acute increases in [DA] by cocaine were greatest toward the end of the conditioning session (current study). The most parsimonious explanation for this is that cocaine accumulated within the brain because it was delivered every ∼5 min and it takes longer for cocaine to be fully metabolized following an i.v. infusion (Pan et al., 1991). As such, cocaine concentration increases in the brain with each infusion causing later infusions to be the functional equivalent of higher dose infusions (see figure 3A in Stuber et al., 2005a). Due to the differential nature of FSCV, only acute increases in [DA] during each conditioning trial were assessed. Future studies are needed that measure long-term changes in [DA] across conditioning trials (Hermans et al., 2008). Although these data suggest that increased cocaine-evoked [DA] across the conditioning session is a pharmacological effect, we cannot exclude the possibility that progressively higher [DA] were due to sensitization of the mesolimbic system by early cocaine infusions (Robinson & Berridge, 2003) or to the development of an association between interoceptive signals related to drug infusion and central cocaine action resulting in later infusions causing greater dopamine release (Wise et al., 2008).
Regionally specific dopamine function in reward processing
The function of cue-evoked alterations in dopamine transmission within the NAc shell is less apparent. Previous microdialysis studies using cocaine self-administration have shown that conditioned stimuli do not increase [DA] within the NAc shell (Ito et al., 2000). Therefore, the brief pause in dopamine transmission described in the current study (which could not be detected with microdialysis) is not inconsistent with previous cocaine studies. However, pharmacological enhancement of dopamine transmission in the shell improves the ability of conditioned stimuli to increase reward seeking (Wyvell & Berridge, 2000). Therefore, dopamine transmission within the shell can enhance the incentive impact of a conditioned stimulus, even if the stimulus itself does not directly increase dopamine release. However, presentation of a non-acute conditioned stimulus (10-min object presentation) that precedes morphine reward has been shown to increase [DA] in the NAc shell but not the core (Bassareo et al., 2007), suggesting that conditioned stimuli may evoke dopamine release in a reward-specific manner. Alternatively, differential dopaminergic processing of conditioned information within the NAc may depend on the nature of the information provided by the cue. For instance, the NAc shell (but not the core) is critical for context conditioning (achieved, in part, via interaction with the hippocampus) (Ito et al., 2008), which may explain why discriminative cues that signal the availability to self-administer electrical brain stimulation evoke phasic dopamine release within the NAc shell (Owesson-White et al., 2008). Additionally, the NAc shell may be more important for regulation of drug seeking by discriminative stimuli following drug abstinence (Ghitza et al., 2003), while the NAc core may be important for drug seeking related to conditioned stimuli following abstinence (Hollander & Carelli, 2007). Together, these studies suggest that the function of conditioned dopamine signaling within the NAc shell is context specific.
Conversely, there is convergent evidence across behavioral paradigms indicating that dopamine transmission within the NAc shell is involved in primary reward processing associated with drugs of misuse. Lesions of the NAc shell attenuate the unconditioned potentiating effects of psychostimulants (Cardinal & Everitt, 2004) and these drugs are self-administered directly into the NAc shell (and olfactory tubercle) but not the NAc core (Ikemoto, 2003). Psychostimulants preferentially increase [DA] within the NAc shell (Pontieri et al., 1995; Aragona et al., 2008; Frank et al., 2008) and dopamine receptor blockade within the NAc shell prevents the acquisition of drug-induced conditioned place preferences (Fenu et al., 2006). Importantly, there is evidence that activation of D1-like receptors is especially important for mediating the unconditioned aspects of associative learning (Di Chiara & Bassareo, 2007) and cocaine directly increases the probability of high-concentration phasic release events within the NAc shell which likely increases the proportion of low-affinity D1-like receptors that are activated (Richfield et al., 1989). Thus, the current data are relevant to the hypothesis that rapid and robust increases in [DA] mediate reward associated with drugs of misuse (Di Chiara & Bassareo, 2007; Volkow et al., 2007).
In conclusion, the current data suggest that dopamine transmission within the NAc core and shell mediates distinct aspects of motivational and associative processes. These regions regulate different components of a particular theoretical framework describing dopamine function, or perhaps separate theories are required to best describe dopamine transmission in a given brain region. Regardless, it is clear that all theories of dopamine function must account for regional specificity regarding phasic dopamine signaling during the acquisition of a cue–cocaine association.
We thank Kate Fuhrman and Mark Stuntz for technical assistance, and Joshua L. Jones, and Robert A. Wheeler for critical reading of the manuscript. John Peterson, and Collin and Larry George helped with instrumentation. This work was supported by F32 21489 (B.J.A.), DA 17318 (R.M.C. & R.M.W.), and DA 10900 (R.M.W. & R.M.C.).