| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 Ernest Gallo Clinic and Research Center, University of California, San Francisco, Emeryville 94608; 2 Graduate Program in Neuroscience, University of California, San Francisco 94143; 3 Departments of Neurology and Physiology, and Wheeler Center for the Neurobiology of Addiction, University of California, San Francisco, California 94143
Submitted 9 July 2003; accepted in final form 20 November 2003
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
) suggested that the NAc functions as a limbic-motor interface. The evidence in favor of this idea was that the NAc receives projections from limbic structures (basolateral amygdala, prefrontal cortex, hippocampus), and projects primarily to the ventral pallidum, a major output nucleus of the basal ganglia. In addition, midbrain dopamine neurons project to the NAc, and injections of dopamine agonists into the NAc increased locomotion (Mogenson et al. 1980). Later investigations led to several more specific theories regarding the precise role played by the NAc in behavior. Among these are that the NAc neurons (and particularly dopamine acting on these neurons) contribute to an assessment of the "incentive salience" of environmental cues (Berridge and Robinson 1998)—that is, NAc neurons determine the likelihood of a motor response to a cue based on the animal's motivational state and the predictive information of the cue. Another theory suggests that NAc neurons perform a cost/benefit computation (based on environmental information, the animal's motivational state, and the amount of work required to obtain a reward) that influences the animal's choice among different possible motor behaviors (Salamone and Correa 2002). Yet another theory proposes that the NAc is essential for mediating the influence of Pavlovian conditioned stimuli (which do not require specific action by the animal to obtain reward or avoid punishment) on the animal's instrumental or other motor behavior (Cardinal et al. 2002a).
Common to all of these proposals is the idea that the activity of NAc neurons encodes sensory information relevant to the potential consequences of different behaviors and that this activity promotes actions that will maximize reward. A number of experiments have provided specific support for this hypothesis. One example comes from the study of Pavlovian-instrumental transfer. In these experiments, an animal is first trained to associate a conditioned stimulus (CS) with a food reward. The animal is then trained to perform an instrumental action (lever press) to obtain the same reward in the absence of the CS. In the test session, the animal is allowed to press the lever under extinction conditions. Intermittent presentation of the CS during the test session potentiates responding on the lever (Dickinson and Dawson 1987; Lovibond 1983). Lesions of the NAc block the potentiation of responding by the CS (Corbit et al. 2001; de Borchgrave et al. 2002; Hall et al. 2001), and injection of amphetamine (a drug that increases extracellular dopamine) into the NAc enhances the CS-induced potentiation of responding (Wyvell and Berridge 2000, 2001), suggesting a critical role for NAc neurons in promoting behavior in response to goal-associated cues.
Another line of evidence that NAc neurons facilitate the behavioral response to reward-predictive cues comes from studies of conditioned reinforcement in which animals lever-press to obtain a cue that has previously been associated with reward (Robbins 1975). NAc amphetamine injections potentiate responding for the cue by a mechanism dependent on dopamine receptors (Wolterink et al. 1993). Furthermore, approach to a CS that predicts reward is reduced by manipulations that impair NAc function (reviewed in Cardinal et al. 2002a). Also, NAc lesions disrupt the processing of predictive cues, biasing the animal toward smaller rewards that require less effort (Cardinal et al. 2001). This effect is consistent with studies showing that an action of dopamine on NAc neurons increases the effort animals will put forth to obtain reward (Salamone and Correa 2002), possibly by modulating NAc neurons that process the cues that guide the animal to reward. Thus taken together, the available behavioral evidence points strongly toward a role for NAc neurons in promoting behavioral responses to cues that possess incentive value.
Despite the growing evidence that NAc neurons contribute to the motor response to incentive cues, relatively little is known about how they encode information about environmental cues. Recordings from the primate striatum (including the ventral striatum, which includes the NAc) have revealed excitations in response to cues that predict reward (Bowman et al. 1996; Cromwell and Schultz 2003; Hassani et al. 2001; Hollerman et al. 1998; Schultz et al. 1992; Shidara et al. 1998). These excitations depend strongly on the predictive value of the cue because the type of reward predicted (Hassani et al. 2001), the magnitude of the predicted reward (Cromwell and Schultz 2003; Hollerman et al. 1998), and the temporal proximity of the reward (Shidara et al. 1998) all affect the magnitude of excitation when the behavior required to obtain reward is held constant. In subpopulations of cue-responsive neurons, the magnitude of the excitation evoked by predictive cues is also correlated with the specific motor activity required to obtain the reward (Cromwell and Schultz 2003; Hassani et al. 2001; Hollerman et al. 1998).
Although recordings from primate striatum are consistent with the behavioral evidence that NAc neurons encode incentive cue information, it has been difficult to relate this information directly to the large body of behavioral pharmacology, which is primarily based on rodent experiments. Primate recordings are usually made in the striatum, although the rodent literature suggests quite different roles for the dorsal striatum and NAc in behavior (Parkinson et al. 2000a; Reading et al. 1991). In addition, in rodent behavioral tasks, animals are free to locomote, whereas, during recording experiments in primates, the monkey is immobilized and free only to make arm or eye movements. Because the NAc is an important regulator of locomotion (Mogenson et al. 1993; Tzschentke and Schmidt 2000), the difference between whole body locomotor and more restricted movements could be reflected in the firing patterns of NAc neurons.
The firing of NAc neurons in rats during drug self-administration has been extensively studied using simple operant tasks such as fixed ratio (Carelli 2000, 2002; Carelli and Deadwyler 1994, 1996a,b; Carelli and Ijames 2000, 2001; Carelli et al. 1993, 1999, 2000; Chang et al. 1998, 2000; Chang et al. 1996, 1997a,b; Janak et al. 1999; Lee et al. 1999; Nicola and Deadwyler 2000; Peoples and West 1996; Peoples et al. 1997, 1998a,b, 1999a,b; Uzwiak et al. 1997). These studies report brief excitations and inhibitions just before, during, and after operant responses as well as changes in firing that appear to correlate with the level of drug in the brain. In addition, several studies have found excitations and inhibitions just before and after operant responses for natural reward (Carelli and Ijames 2001; Carelli et al. 2000; Hollander et al. 2002; Roop et al. 2002), reporting firing patterns generally similar to those found during operant responding for drug reward. CS's associated with cocaine (Carelli 2000; Carelli and Ijames 2001), and stimuli that predict cocaine (Ghitza et al. 2003) can excite and inhibit NAc neurons. However, the firing changes of rat NAc neurons in response to stimuli that predict natural reward have not been reported.
In this study, we use a discriminative stimulus (DS) task to explore how NAc neurons encode predicted outcomes and the behavioral responses required to obtain the outcome. The DS is a sensory cue that directs the animal to perform an operant response (in this case, a nose-poke into a hole equipped with a photobeam) to obtain a sucrose reward. Because the same behavior can be elicited repeatedly by presentation of the cue, the task is well suited to the correlation of behavioral events with the activity of single units. In trained animals, DSs elicit robust reward-seeking and may be involved in cue-induced relapse to drug-seeking behavior (Berridge and Robinson 1998; Kantak et al. 2002; Weiss et al. 2000; Yun and Fields 2003). Furthermore, several studies have suggested a role for the NAc in DS-controlled responding. The human NAc is activated during DS-based tasks (Breiter et al. 2001; Knutson et al. 2001), dopamine is released in the NAc of rats after DS presentation (Bassareo and Di Chiara 1999; Weiss et al. 2000), and activation of NAc dopamine receptors is essential for animals to respond to DSs at least under some circumstances (Yun et al. 2004). Here, to more fully understand how NAc neurons contribute to cue-mediated behavioral responding, we characterize the cue-evoked and operant responses of NAc neurons as rats perform a DS task.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Male Long-Evans rats (Harlan or Charles River) were used in this study (n = 21). Animals (
350 g) were individually housed on a 12-h light/dark cycle, and experiments were performed during the light phase. After receipt, rats were allowed 
1 wk of ad lib food and water, followed by 1 wk of restricted food and water prior to training. Throughout all experiments, restriction was accomplished by allowing the animals 1 h of free access to food and water per day at the end of experimental manipulations. Animal handling and experiments conformed to National Institutes of Health and Ernest Gallo Clinic and Research Center animal care and use policies.
Apparatus
Animals were trained in custom-built Lucite operant chambers contained within light- and sound-insulated boxes. Chambers were 40.6 x 40.6 cm and were equipped, on one wall, with two nose-pokes (Med Associates) and a reward receptacle located between them. Liquid reward (50 µl of a 10% sucrose solution) was delivered by a dipper (Coulbourne Instruments) in most experiments; in some experiments, a syringe pump was used to deliver the solution into a small well inside the receptacle. Receptacles were equipped with photobeams to determine the times at which the animal's head entered and exited the receptacle. Operant chambers also contained two white houselights, a white-noise speaker, and a loudspeaker for delivering auditory stimuli (Med Associates). White noise (65 dB) was present throughout all experiments. Each box was equipped with a video camera and monitor to allow experimenters to observe animals inside the behavior chambers as they performed the task.
Training and behavioral task
The firing patterns described in this work were observed during a DS task. Animals progressed through several stages of training before undergoing surgical implantation of electrodes in the NAc and subsequent recording of neural activity. In the first stage, food-restricted animals were introduced to the chamber. Entry into the reward receptacle triggered delivery of the sucrose reward and dimming of houselights (by turning off 1 of the 2) for 20 s, during which subsequent entries had no effect. After animals learned to obtain all 100 available rewards in <1 h (usually this took only 1 or 2 days), animals were advanced to a fixed ratio (FR) task in which a single nose-poke in either of the two nose-poke holes resulted in reward delivery, accompanied by the 20-s dimmed houselights and time-out. After animals learned to obtain 100 rewards in <2 h (1–3 days), they were advanced to a cue-response task in which a compound cue (an intermittent tone and dimmed houselights) was presented every 60 s. In this task, the left or the right nose-poke was chosen to be the "active" nose-poke, and thereafter all rewards were contingent on responses in the active nose-poke during presentation of the cue. The tone was either 6 kHz (12 rats) or 4 kHz (9 rats). This and all other tones presented in this study were intermittent, cycling between a 200-ms tone-on pulse and a 550-ms tone-off period prior to the next tone-on; all tones were 85 dB. The tone/dimmed lights lasted for 
60 s, and a single nose-poke in the active nose-poke hole during cue presentation terminated the cue, caused the delivery of the sucrose reward, and triggered a 20-s conditioned stimulus (CS) consisting of continued dimmed houselights and an 8-kHz intermittent tone. Nose-pokes in the absence of the cue and during the CS were not rewarded. Animals were advanced to the DS task when they received >60 rewards in 2 h (2–3 days).
In the DS task (Fig. 1A), the cue that, in the previous stage of training, signaled contingent reward availability was presented for 20 s with an average frequency of once every 2 min (variable interval 2-min schedule). Actual intervals between DS presentation were randomly selected by the computer from the following list: 60, 90, 104, 112, 116, 118, 119, 119.5, 120, 120.5, 121, 122, 124, 128, 136, 150, and 180 s. A response on the active nose-poke during DS presentation terminated the DS and resulted in delivery of the sucrose reward in the receptacle, accompanied by a 20-s CS (8-kHz intermittent tone/dimmed houselights). In addition to the DS, a nonrewarded stimulus (NS) was presented on an independent variable interval 2-min schedule. The NS was always 20 s long and consisted of an intermittent tone of either 6 kHz (9 rats) or 4 kHz (12 rats); the frequency that was not used for the DS was chosen to be the NS for each rat. To prevent overlap, DSs and NSs whose onset times were scheduled to occur during the other cue (or CS) were delayed by 40 s. Nose-pokes during the NS or CS were not rewarded nor were nose-pokes at any time during the session other than during DS presentation. Nose-pokes in the inactive nose-poke hole were never rewarded. Animals were run once per day, 5 days/wk. Animals usually learned to respond to >90% of DSs within one week of training. However, surgery was often not performed for 2 or more weeks after the beginning of this training phase, during which animals were run every day (5 days/wk).
|
Anesthesia was induced with ketamine/xylazine and maintained with either subsequent ketamine injections or isoflurane. One array consisting of eight 50-µm-diam Teflon-insulated wires (NB Labs, Denison, TX) was chronically implanted into the NAc of each hemisphere as previously described (Nicola and Deadwyler 2000). Target coordinates (Paxinos and Watson 1998) were (in mm) AP, +1.0 to +2.5; ML, ±0.5 to ±1.5; DV, –6.5 to 8. Electrodes were fixed to the skull with acrylic dental cement secured with stainless steel bone screws. A silver wire implanted into the cortex caudal to the NAc was used as a ground electrode, and a miniature connector wired to the electrodes was exposed at the top of the implant. Animals were allowed to recover for one week prior to beginning experiments.
Electrophysiology
Prior to each session, a headstage containing unity gain field-effect transistors (NB Labs) was connected to the animal's implanted electrodes. A cable transmitted the voltage signals to a multichannel commutator (NB Labs) that allowed the rat free movement within the behavioral chamber. The signals were then amplified and spikes were sorted with a Multiunit Acquisition Processor (Plexon). To reduce noise, the signal from a reference electrode (without identifiable spike waveforms) was usually subtracted from each individual wire's signal. Templates of waveforms that appeared to be action potentials were computed by averaging together waveforms selected by the experimenter, and spikes recorded during the experiment were matched to these templates by the computer. All waveforms that exceeded an amplitude threshold were saved to disk for later analysis whether they were assigned to a template or not. Templates were adjusted by the experimenter prior to each recording session to capture waveforms that changed amplitude or shape from the previous session. After each session, spikes on each wire were re-sorted to eliminate noise and to capture waveforms not previously assigned to the appropriate template. Waveforms <75 µV peak to peak were rejected; typical noise levels were 25–50 µV. In many cases, more than one waveform shape, corresponding to more than one unit, could be isolated on a single wire. In most instances, the shapes of these waveforms could be clearly separated. When there was overlap, waveforms that could not be definitively assigned to one unit were rejected from the analysis. When spike re-sorting was complete, autocorrelograms were constructed for each unit; units without well-defined refractory periods were either rejected or re-sorted. Crosscorrelograms were constructed for units on the same wire. If two units exhibited a common refractory period, the waveforms were re-sorted again or combined if the waveforms could not be definitively distinguished.
Data analysis
Each unit was assigned, based on its firing pattern, to at least one of the subsets of neurons exhibiting the response types described in Table 1. To do this, units were first prescreened with a series of paired t-tests that compared the baseline (precue) firing rate with the firing rate during each event (DS, NS, nose-poke, receptacle entry, receptacle exit). Table 1 shows the peri-event windows from which the event-associated firing was obtained. The data used for each paired t-test was the set of all instances of the event in question (e.g., nose-pokes). A low stringency significance level (P < 0.05) was deliberately chosen so that units with peak firing changes that occurred slightly outside the time ranges shown in Table 1 would be included. Next, peri-event histograms (PEHs) from all the prescreened units with significant firing changes were examined and scored independently by two investigators as either exhibiting or not exhibiting the response type in question. If both investigators agreed that the unit exhibited the firing pattern, it was classified as such; if one or both scored the unit as not exhibiting the firing pattern, it was classified as nonphasic with respect to the event in question.
|
For Table 2 (which shows the proportion of neurons showing each response type), we limited the neuronal population to that obtained from the first recording session from each animal. For Table 3 (which shows the proportion of neurons exhibiting a given response type that also exhibited any other response type), we found the first session during which the response type listed in the first column (response type 1) was observed to arise from an electrode. This may or may not have been on the animal's first recording day. This was repeated for each electrode in each rat. We then asked whether, during the recording session, the neuron exhibited any of the other response types (response type 2), and in Table 3 we expressed the number of neurons displaying both response types 1 and 2 as the percentage of neurons with response type 1. For any electrode, this analysis uses only the first recording session during which response type 1 was observed to arise from a neuron on the electrode; subsequent recording sessions were not analyzed.
|
|
|
|
) also uses data from the experiments described in this paper, but examines firing related to other task events than those described in detail here. Histology
After completion of experiments, electrode positions were marked and the animals were perfused. Histology methods and results are reported in the companion paper (Nicola et al. 2004).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Electrophysiological recordings were taken from 21 rats that were fully trained on the DS task. In this task, a DS was presented to the animals at random intervals (mean: 2 min); performing a nose-poke response during the DS resulted in delivery of a 10% sucrose solution into a reward receptacle located next to the nose-poke hole. In addition to a DS, an NS was also randomly presented; responding to the NS was not rewarded (Fig. 1A). Typical performance on the DS task is shown in Fig. 1B. General behavioral performance was monitored by the experimenters with a video camera, and animals were observed to performed the task without superstitious learning effects. Specifically, rats responded on the active nose-poke at short latency after onset of the DS without performing other behaviors such as responding in the inactive nose-poke or checking the reward receptacle before making the nose-poke. Animals usually made a number of uncued responses (in the absence of DS, NS, and CS) in the active and, less frequently, inactive nose-poke holes. Averaged across 184 sessions, the rate of uncued responses in the active nose-poke was 0.0061 ± 0.0005 (SE) Hz, whereas the uncued response rate in the inactive nose-poke was one-tenth as fast (0.00056 ± 0.00013 Hz). These rates were significantly different (t183 = 11.3, P < 0.001), indicating that animals differentiated between inactive and active nose-poke holes in the absence of explicitly presented cues. The animals' responding was under control of the DS because the average latency to respond (2.9 s) corresponds to an instantaneous response rate of 0.35 Hz; because this is much larger than the uncued response rate in the active nose-poke (0.006 Hz), the rate of uncued responding cannot account for the high DS response ratio or the low latency to respond to the DS.
Animals responded to >90% of DS presentations (DS response ratio) while responding to only half of NS presentations, a difference that was highly significant (t183 = 25.9, P < 0.001; Fig. 1C). In addition, the latency to respond to the NS was significantly higher than the DS response latency (t182 = 11.7, P < 0.001; Fig. 1D). Therefore animals clearly differentiated between the DS and NS. The response ratio for the NS was higher than would be expected given that responding to the NS did not result in reward delivery. In our hands, rats are capable of differentiating between two dissimilar cues such that response ratios are >90% for rewarded DSs and <20% for NSs (Nicola, unpublished observations), consistent with previous studies (e.g., Robbins et al. 1990). The 49% NS response ratio is therefore likely a result of generalization between the DS and NS (Hull 1943; Tarpy 1982), which were physically very similar (both were compound stimuli, with intermittent tones of slightly different frequency and dimmed houselights). The DS can be thought of as a cue that is more reward-predictive than the NS, and this is reflected in the animals' behavioral performance. Thus the high NS response ratio allowed us to compare the firing rate of NAc neurons to cues that differed in their reward-predictive value when the behavioral responses to the cues were similar.
Neurons
A total of 211 behavioral sessions were used to obtain electrophysiological recordings during the DS task; of these, 27 were the "random withholding" sessions described in the companion paper (Nicola et al. 2004). Neurons were classified according to whether there was an increase or decrease in the neuron's firing in the peri-event windows listed in Table 1, relative to the precue baseline. Subpopulations of NAc neurons responded phasically to each obvious component of the DS task (Table 2): cue presentation, operant response, entry into the reward receptacle, reward consumption, and exit from the receptacle. The proportion of neurons exhibiting each response type is shown in Table 2. To avoid the complication that the same or different neurons can be recorded on an individual microwire electrode across several sessions (see METHODS), we used only the first recording session from each animal to construct Table 2. Of the 217 neurons recorded, 105 (48.4%) exhibited at least one response type. Many neurons were phasic with respect to more than one event; the proportion of neurons displaying each response type that also displayed the other response types is shown in Table 3. Only the first session during which each response type was recorded on an electrode was used to construct Table 3. This prevents neurons recorded subsequently on the same wire (which may or may not have been the same neurons as those recorded initially) from affecting the calculated proportions. In this paper, we examine the first eight response types: phasic firing in response to one or both of the cues, and phasic firing in relation to the operant response. The remaining cell types are examined in the companion paper (Nicola et al. 2004).
Interestingly, in neurons with more than one type of phasic response, certain combinations occurred at greater than chance levels. For example, 53.5% of neurons with incentive cue excitations were also inhibited during reward consumption ("sustained receptacle inhibition" in Table 3), and 39.1% of operant-excited neurons exhibited this inhibition as well. The overall proportion of neurons showing this type of inhibition was estimated to be 18.9% (Table 2), significantly lower than the proportion in incentive-cue-excited neurons (
2 = 21.3, P < 0.001) and in operant-excited neurons (2 = 10.7, P < 0.002). Similarly, proportionally more incentive-cue-inhibited neurons exhibited excitation during reward consumption ("sustained receptacle excitation") than did neurons in the overall population (20.0 vs. 2.8%; 2 = 16.3, P < 0.001); also, more operant-inhibited neurons displayed sustained receptacle excitation than did neurons overall (9.0 vs. 2.8%; 2 = 4.6, P < 0.04). Thus there is an enriched representation of phasic firing during reward consumption in neurons active during reward-seeking behavior. These examples of overrepresentation support the validity of our system of waveform sorting. The clustering of specific firing patterns in single neurons is unlikely to be explained by the systematic misclassification of the waveforms from two or more neurons (with potentially different firing patterns) as a single neuron because such an error should result in combinations of different response types that reflect their overall proportions.
Incentive cue excitations
Six different firing responses to cue presentation were observed. Figure 2 shows four examples of the first of these (incentive cue excitations), each recorded from a different rat. The firing response of these neurons was much greater to the DS (Fig. 2, A1–D1) than to the NS (A2–D2), was often sustained throughout the interval between DS and nose-poke response and usually continued until the animal received the reward (Figs. 2, A1, C1, and D1, and 3, A and B). A smaller proportion of these neurons (the 44.2% that were not also classified as operant excited; see Table 3) exhibited firing increases that lasted for several seconds after DS presentation but were reduced just before the nose-poke (Fig. 2B1). The majority (53.5%) of incentive-cue-excited neurons were inhibited while the animal was in the reward receptacle (Table 3). These inhibitions can be clearly seen in the examples in Fig. 3, which show rasters and histograms time-locked to DS onset and receptacle entry.
|
|
|
The difference in DS- and NS-evoked excitation may have been due to the difference in latency to respond to the cues because animals were slower to respond to the NS than to the DS (Fig. 1D). However, the average excitation in response to the cues did not depend on response latencies. Figure 4F shows the excitation from 0 to 1 s after each cue for nose-poke response latencies in 1-s bins between 0 and 4 s. In this analysis, only the 25 incentive-cue-excited neurons that were recorded during sessions with at least one behavioral response to both DS and NS in each latency range were included. Two-way within-subjects ANOVA on the firing rate increases of the 25 neurons showed an overall effect of the cue [F(1,24) = 5.2, P < 0.04] but no effect of latency [F(3,72) = 0.5, P > 0.6] and no interaction between cue and latency [F(3,72) = 0.2, P > 0.8; post hoc tests were not done because of the lack of latency effects]. Therefore the firing response of incentive-cue-excited neurons was not affected by latency in the range 0–4 s but was significantly smaller across these latencies for the NS in comparison with the DS. Thus incentive cue excitation encodes information about the cue (the response is greater to the more reward-predictive cue than the less-predictive cue, even if the motor response is equivalent) and is correlated with the animal's motor response (the firing response is greater when the animal subsequently makes an operant response to the cue than when the animal does not respond).
Transient incentive cue excitations
In contrast to neurons with sustained excitations in response to the DS, some neurons exhibited brief DS-evoked excitations lasting no more than 0.5 s (Fig. 5, A1 and B). The excitation clearly did not extend until the nose-poke because firing was not increased prior to the response (P < 0.001 for overall ANOVA, P < 0.05 for SNK comparison of the post-DS firing increase vs. the preresponse increase; n = 19; Fig. 5, D and E). Furthermore, unlike the sustained incentive cue excitations that were sensitive to the information contained by the cue, firing did not differ in response to DSs and NSs (P > 0.05, SNK; Fig. 5, B, C, and E; only DSs and NSs followed by nose-pokes were used for analysis). Because few of these neurons were recorded during sessions in which animals failed to respond to at least some DSs, it was not possible to determine whether the DS-evoked excitation was different if the animal did not make an operant response. The similar response to the DS and NS, however, indicates that these neurons may encode only the information that a prominent sensory stimulus has begun, but not its predictive value.
|
Incentive-cue-inhibited neurons were similar to incentive-cue-excited neurons in most respects except for the sign of the firing change. First, similar proportions of all neurons were classified as each cell type (5.1% were incentive cue excited and 2.8% were incentive-cue-inhibited; Table 2). Second, the inhibitions usually lasted for several seconds, often continuing throughout the operant response and even reward consumption (Fig. 6, A1 and B1). This is reflected in the large proportion of incentive-cue-inhibited neurons that were also classified as operant inhibited (62.5%). Third, the NS-evoked inhibition was less than the DS-evoked inhibition (Fig. 6, A2 and B2); only 37.5% of incentive-cue-inhibited neurons were also classified as NS inhibited. Fourth, as for incentive-cue-excited neurons, the difference in the firing response to DS and NS cannot simply be attributed to the fact that the animal responded less to the NS (Fig. 7). In the 31 incentive-cue-inhibited neurons that were present in sessions in which the animal failed to respond to at least one DS and responded to at least one NS, the median DS-evoked inhibition was smaller when the animal did not respond to the cue (ANOVA P < 0.005, P < 0.05 for SNK; Fig. 7, A, C, and E). There was no significant difference in the inhibition evoked by NSs to which the animal responded and did not respond (Fig. 7, B, D, and E); however, inhibition evoked by NSs with behavioral responses was very small (median: 0.28 Hz), which would make it difficult to observe any significant reduction in this inhibition. Indeed, the inhibition evoked by NSs with behavioral responses was significantly smaller than the inhibition evoked by DSs with responses (Fig. 7, A, B, and E). These results suggest that, similar to incentive-cue-excited neurons, the inhibition in response to the DS was greater than in response to the NS, especially when the animal responded to the cues.
|
Excitations and inhibitions in response to the NS
As shown in Table 2, we found analogs of incentive-cue-excited, transient incentive-cue-excited, and incentive-cue-inhibited neurons that showed these firing patterns in response to the NS. However, it is unlikely that the NS-evoked firing patterns represent distinct populations of neurons from those exhibiting DS-evoked firing patterns. This is because the majority of NS responses occurred in neurons that also responded to the DS. Specifically, 75% of NS-excited neurons were also incentive cue excited (either transient or sustained), 71% of NS transiently excited neurons also had transient or sustained incentive cue-excitations, and 56% of NS-inhibited neurons were inhibited by the DS as well. The number of NS-responsive neurons that remained after excluding DS-responsive neurons was too small (n = 7 for NS excitation, n = 5 for transient NS excitation, and n = 12 for NS inhibition) to perform the same analyses that were done for the incentive cue-responsive neurons. Therefore although very small subpopulations of cue-responsive neurons may have larger responses to the NS than to the DS, the data support the hypothesis that NAc neuronal responses that preferentially encode reward-predictive cues are represented with much greater frequency than responses that preferentially encode less predictive stimuli.
Operant excitations
Operant excitations were defined by a significant increase in firing either just before or surrounding the DS-evoked nose-poke response (Table 1). Many incentive-cue-excited neurons were also classified as operant excited (56%) because incentive cue excitations tended to last throughout the nose-poke response. Representative examples of operant-excited neurons are shown in Fig. 8, demonstrating the tight coupling between nose-poke response and the peak of the excitation. These examples also show substantial inhibition during reward consumption (that is, after reward receptacle entry), an attribute shared by 39% of operant-excited neurons.
|
The substantial overlap between incentive-cue- and operant-excited subpopulations raises the question of how different is the information encoded by these two firing patterns. To address this issue, we first divided neurons with incentive cue excitations and operant excitations into three nonoverlapping classes: neurons with incentive cue excitations only (IC+; example shown in Fig. 2B; n = 21), those with both incentive cue and operant excitations (IC+Op+; Fig. 2, A, C, and D; n = 27), and those with operant excitations only (Op+; Fig. 8, A and B; n = 48). We then asked whether the operant-related firing of these neurons was affected by the cue being presented when the animal responded: DS, NS, or none. The latter category was comprised of nose-pokes made in the absence of stimuli; these were present in almost every session and were not rewarded. Often, animals made a series of such nose-pokes in rapid succession; therefore to avoid including data from overlapping time windows, uncued nose-pokes that followed a previous uncued nose-poke by <10 s were excluded from the analysis. In addition, neurons that were recorded only in sessions in which animals failed to make uncued nose-pokes or to respond to the NS were excluded.
The operant-related firing of IC+Op+ neurons was greater in the presence of the cue, but the firing of IC+ and Op+ neurons was not. The excitation of IC+ neurons in the 0.5 s just prior to the nose-poke was not significantly different for nose-pokes elicited by the DS than for nose-pokes elicited by the NS or uncued nose-pokes (P > 0.1 for ANOVA, n = 19; Fig. 9, A and D). The baseline firing rate also did not differ (P > 0.1). In contrast, the operant-related firing of IC+Op+ neurons was greatest for the nose-pokes during the DS, smaller for nose-pokes during the NS, and smallest for uncued nose-pokes (P < 0.001, P < 0.05 for SNKs, n = 26; Fig. 9, B and D). The baseline firing rates also differed slightly but significantly under the three conditions (median: 3.7 Hz for uncued, 3.9 Hz for NS, and 4.1 Hz for DS; P < 0.02 for ANOVA, with all rates significantly different by SNK test). However, these differences are small and unlikely to account for the differences in operant-associated firing. In contrast to the differences in operant firing in IC+Op+ neurons, Op+ cells exhibited the same degree of excitation no matter whether the nose-poke was a response to the DS, NS, or uncued (P > 0.06, n = 45; Fig. 9, C and D), as suggested by the examples shown in Fig. 8. Baseline firing rates did not differ (P > 0.1). Therefore IC+Op+ differ from IC+ and Op+ neurons in that their operant-related firing depends on whether the nose-poke is a response to a cue and whether the cue is reward-predictive. One reason why differences in the operant-related firing of IC+ neurons were not obviously greater than differences in baseline rate may be that IC+ neurons by definition do not have large operant-related responses. In contrast, Op+ neurons are defined by their operant excitation. These results suggest that Op+ neurons encode information about the subsequent or ongoing behavior but not about the predictive information contained by the preceding sensory cue, whereas IC+Op+ neurons encode predictive information immediately after cue presentation and continue to encode this information at least until the animal makes the operant response.
|
The fact that many incentive-cue-excited neurons also show operant excitation (IC+Op+) raises the question of whether DS-evoked excitation is sustained at the same level throughout the DS-nose-poke interval or if, instead, transient changes occur just before the nose-poke or just before reward receptacle entry. To answer this question, we first constructed PEHs for IC+, IC+Op+, and Op+ neurons, aligned with each of three different events: the DS (only DSs followed by a nose-poke response were used), the nose-poke response to the DS, and the reward receptacle entry after a successful nose-poke response to the DS (Fig. 10). The histogram bars are median firing rate across all neurons in the class. Particularly for IC+Op+ neurons, two peaks are present: one immediately after the DS (Fig. 10D) and another immediately prior to the nose-poke response (Fig. 10E). For IC+ neurons, the operant-associated peak is somewhat more diffuse than the DS-evoked peak (Fig. 10, A and B).
|
0.5 s prior to the nose-poke response, whereas the time course of the IC+Op+ peak prior to the nose-poke was much slower (Fig. 11B) and IC+ neurons showed no excitation tightly coupled to the operant response (Fig. 11B). These results indicate that the firing of IC+ neurons remains elevated at a low constant level until just before the nose-poke. They also support the idea that IC+Op+ neurons may have a second peak (operant excitation) just prior to the nose-poke.
|
|
1–1.5 s before the response, no matter whether the DS was presented a short (e.g., <2 s) or long time (>6 s) before the operant response. Accordingly, when the firing of these neurons was aligned to the DS onset, the excitation was delayed until just before the nose-poke response (Fig. 12C, left). Substantially different results were obtained with IC+ neurons (Fig. 12A). For these neurons, the firing rate was increased immediately after the DS and remained significantly increased until the nose-poke for DS-nose-poke latencies that were <2 and 2–4 s (Fig. 12A, left). At longer latencies (4–6 and >6 s), IC+ neurons were not significantly excited above baseline at any time point. At short latencies, the excitation remained significant and pronounced until the operant response (Fig. 12A, right). The excitation of IC+ neurons is therefore tightly locked to the DS and is sustained throughout DS-nose-poke latencies <4 s. These results differ from Op+ neurons in that Op+ firing is not tightly coupled to DS onset and is not sustained throughout the DS-nose-poke intervals of any length.
IC+Op+ neurons exhibited time courses of excitation that appeared to be hybrid between IC+ and Op+ neurons (Fig. 12B). Like IC+ neurons, these neurons were almost always significantly excited immediately after DS presentation (Fig. 12B, left), but the excitation was clearly smaller at longer (>4 s) than shorter latencies (<4 s). The smaller excitation at longer latencies can be seen in the examples in Fig. 2, C1 and D1. Also like IC+ neurons, the excitation of IC+Op+ neurons remained sustained throughout DS-nose-poke intervals <4 s, until the animal performed the nose-poke (Fig. 12B, right). However, unlike IC+ neurons and like Op+ neurons, at longer latencies (>4 s), IC+Op+ neurons exhibited a rapid increase in excitation beginning 1–2 s before the nose-poke (Fig. 12B, right) which was significant for the longest latencies (>6 s). Therefore at shorter DS-nose-poke latencies, IC+Op+ neurons exhibit the properties of IC+ neurons, whereas at longer latencies, IC+Op+ neurons exhibit the properties of Op+ neurons.
In summary, the excitation of Op+ neurons is tightly coupled to the operant response, whereas IC+ and IC+Op+ neurons show abrupt increases in firing time-locked to DS onset that are sustained throughout DS-response intervals that are <4 s. IC+Op+ neurons also show excitation time-locked to the operant response when the DS-nose-poke latency is >4 s. Therefore the operant-associated firing peak of IC+ and IC+Op+ neurons in the histograms in Figs. 10B and 9E (which included all DS-nose-poke latencies) can be explained by two factors. For IC+ neurons, the peak is due only to the sustained firing at shorter latencies. For IC+Op+ neurons, the peak is due to both sustained firing between the DS and nose-poke at shorter latencies, and a prominent, independent operant-associated peak at longer latencies. In contrast, the time course of Op+ neuron firing is independent of the DS-nose-poke latency. The excitation always begins 1–2 s before the operant response, whether the latency is long or short. This means that the apparent DS-evoked peak in Op+ neuron firing (Fig. 10G) is not a true DS-evoked peak but rather a consequence of many excitations tightly time-locked to the operant response, which occurs at variable times after DS presentation.
According to Fig. 4F, the incentive cue excitation is not dependent on the behavioral response latency. Although this result appears to conflict with the data shown in Fig. 12, A and B, showing that sustained excitation is less pronounced at longer latencies, Fig. 12 shows a greater range of latencies and time points. In
indicates the median, and the lower and upper edges of the box show the 1st and 3rd quartiles, respectively, of the distribution of firing rate increases. *, P < 0.05. F: the mean firing rate increase in response to cues does not depend on the latency to respond. The firing rate increase in the 1st 1 s after the DS is significantly greater than after the NS, with no effect of latency. Post hoc comparisons were not performed because of the absence of latency effects. Error bars represent SE.
