The tubular striatum and nucleus accumbens distinctly represent reward-taking and reward-seeking

Abstract

The ventral striatum regulates motivated behaviors that are essential for survival. The ventral striatum contains both the nucleus accumbens (NAc), which is well established to contribute to motivated behavior, and the adjacent tubular striatum (TuS), which is poorly understood in this context. We reasoned that these ventral striatal subregions may be uniquely specialized in their neural representation of goal-directed behavior. To test this, we simultaneously examined TuS and NAc single-unit activity as male mice engaged in a sucrose self-administration task, which included extinction and cue-induced reinstatement sessions. Although background levels of activity were comparable between regions, more TuS neurons were recruited upon reward-taking, and among recruited neurons, TuS neurons displayed greater changes in their firing during reward-taking and extinction than those in the NAc. Conversely, NAc neurons displayed greater changes in their firing during cue-reinstated reward-seeking. Interestingly, at least in the context of this behavioral paradigm, TuS neural activity predicted reward-seeking, whereas NAc activity did not. Together, by directly comparing their dynamics in several behavioral contexts, this work reveals that the NAc and TuS ventral striatum subregions distinctly represent reward-taking and reward-seeking.

NEW & NOTEWORTHY The ventral striatum, considered the reward circuitry “hub,” is composed of two regions: the NAc, which is well established for its role in reward processing, and the TuS, which has been largely excluded from such studies. This study provides a first step in directly contextualizing the TuS’s activity in relation to that in the NAc and, by doing so, establishes a critical framework for future research seeking to better understand the brain basis for drug addiction.

INTRODUCTION

Goal-directed behaviors require near continuous evaluations of internal states and external stimuli to orchestrate appropriate actions. These evaluations are believed to occur in part at the level of the ventral striatum, a brain region where midbrain, corticolimbic, and thalamic inputs are integrated into the basal ganglia to influence action. Accordingly, the ventral striatum is considered the “limbic-motor interface,” (1) which imparts upon it critically important roles. For instance, some neurons in the ventral striatum’s nucleus accumbens (NAc) represent sensory information during goal-directed behaviors (2–5) in manners that are believed to underlie addiction (6, 7).

Although largely underappreciated, the ventral striatum also includes the tubular striatum (TuS; formerly known as the olfactory tubercle) (8). The TuS and NAc were, upon the conceptualization of the ventral striatum, hypothesized to contribute distinctly to behavior (9). Since then, one study reported that rats self-administer intracranial cocaine infusions into the TuS more than into the NAc (10). Furthermore, TuS neurons are recruited by stimuli conditioned to predict rewards (11–16) and during cocaine self-administration (17), and that their activity represents palatable reinforcers in manners depending upon motivational states (18). This is not surprising since the TuS receives dopamine from the ventral tegmental area (VTA), as well as glutamate from corticolimbic structures including the prefrontal cortex, hippocampus, and amygdala, in addition to input from olfactory structures (9, 19–22).

Likewise to the TuS, the representation of goal-directed behavior in the NAc may also result from its receipt of dopamine from the VTA, as well as glutamate from the aforementioned corticolimbic structures in addition to gustatory regions (23–28). This connectivity affords the NAc with its well-documented capacity to represent information needed to subserve motivated behavior, including sucrose consummatory behavior (2–5) and stimuli conditioned to predict reward (4, 29–31).

Collectively, the TuS and NAc appear to be acting in comparable manners during goal-directed behaviors, suggesting that these activities may be distributed similarly throughout the ventral striatum. However, this conclusion ignores the possibility that these regions may differ in their activity during behavior, which we reasoned is likely given the interregional differences observed in the regulation of drug reinforcement (10, 32). Differences in the dynamics of TuS and NAc neurons may indeed give rise to behavioral specializations, but only one study has come close to making this comparison (2). By monitoring neural activity throughout the striatum of rats engaged in a cue-guided sucrose self-administration task, the authors uncovered varying proportions of modulation in their firing rates across dorsal and ventral striatum subregions, including the TuS (2). However, most neurons were sampled from the dorsal striatum and NAc. More importantly, the comparison was limited by the behavioral paradigm in that rats always received a reward for instrumental responses, yet we know NAc neural activity may diverge in its representation of reward-seeking and reward-taking, particularly following a period of abstinence or extinction training (28, 29, 33, 34). In addition, previous work from our laboratory uncovered that TuS neurons modulate their firing for both instrumental responding to access the reward (viz., reward-seeking), as well as the reward itself (viz., reward-taking) (18); however, it is unknown how TuS neural dynamics evolve following extinction training.

Might there be differences in TuS and NAc neural activity during reward-taking and reward-seeking? Identifying possible interregional specializations by contextualizing the TuS alongside the NAc will provide insights into striatal dynamics that may not just subserve, but may also be modified by, behaviors such as compulsive reward-seeking. To address this question, we acquired single-unit activity simultaneously from both regions as mice engaged in a cue-guided, sucrose-reinforced operant task that included extinction training and cue-induced reinstatement (35).

MATERIALS AND METHODS

Animals

C57bl/6j male mice (2–5 mo of age, originating from stock from Jackson Labs, Bar Harbor, ME) were bred in-house in a University of Florida animal facility. Before experimentation, mice were housed up to five per cage, and after surgical procedures, they were individually housed for the remainder of the experiment. The vivarium was set to a 12-h light/dark cycle, with food and water available ad libitum except during behavioral testing and water deprivation, described in Sucrose Self-Administration Acquisition, Extinction, and Reinstatement Behavioral Task. All experimental protocols were conducted in accordance with the guidelines from the National Institute of Health and approved by the University of Florida Institutional Animal Care and Use Committee. A total of eight male mice contributed data to this study. We did not include females in our study due to well-documented, hormone-dependent sex differences in motivated behavior and ventral striatal function (36, 37) that would require additional comparisons beyond the scope of this work.

Surgical Procedures

Mice were anesthetized under 2%–4% Isoflurane (IsoFlo, Patterson Veterinary, Greeley, CO) in oxygen at 1 L/min, mounted onto a stereotaxic apparatus, and their body temperatures were maintained with a 38°C heating pad. The local anesthetic, bupivacaine hydrochloride (Marcaine, 5 mg/kg subcutaneously, sc, Patterson Veterinary), was delivered before making a 1-cm midline incision to expose the skull. A craniotomy +1.5 mm anterior from bregma and +1.1 mm medial of the midline was made (∼1.5 mm wide) and a 16-channel microwire electrode array, described previously with modifications (12), was implanted into the right hemisphere. The electrode array consisted of 102-µm-diameter perfluoroalkoxy (PFA)-insulated tungsten wires encased into groups of four with 254-µm-diameter polyimide tubing, with eight wires cut 1 mm shorter than the other eight to simultaneously target both NAc and TuS at 3.9 and 4.9 mm, respectively, ventral to the brain surface. A second craniotomy was made to allow for placement of the ground wire (127 µm stainless steel) into the contralateral cortex. The electrode arrays were fixed in place with dental cement, and animals were given meloxicam for analgesia (5 mg/kg sc, Patterson Veterinary) daily for 3 days following surgery.

Operant Chamber

A custom-made open-top 30- × 15- × 15-cm acrylonitrile butadiene styrene (ABS) operant chamber (Fig. 1A) was housed in a noise-attenuating cabinet, with an overhead USB camera to monitor activity and a Piezo speaker (2.7 kHz, 75 dB) to provide acoustic cues. One wall consisted of a guillotine-style door, and another wall held an extruded panel with three nose-poke ports equipped with 880-nm infrared photodiodes that, when broken, signaled port entry. Above the left and right nose-poke ports were white LEDs, and the center port housed a stainless-steel lick spout attached to tubing and a 20-mL syringe, which allowed for fluid delivery via a solenoid pinch valve (Neptune Research, Inc., West Caldwell, NJ). The operant chamber was powered and controlled by an Arduino Uno microcontroller (www.arduino.cc) running custom scripts. The Arduino generated TTL pulses that were relayed into a Tucker-Davis Technologies (Alachua, FL) processor for seamless integration of the behavioral events and stimuli together with the electrophysiological recordings.

Figure 1.

Paradigm for investigating ventral striatal representation of reward-taking and reward-seeking. A: illustration of operant chamber. One wall contained three ports as well as cues. During testing the mice were tethered via a flexible cable to allow acquisition of neural data. B: trial structure during acquisition with examples of active and center port responses (pokes). Latency refers to the duration for mice to transition from active port to center port. C: cue and reinforcer outcomes of active port responses across the three phases of the task. D: operant responding across acquisition and extinction sessions. Mice demonstrated a strong preference for the active port throughout acquisition that decreased during extinction (**P < 0.01, repeated-measures ANOVA, active vs. inactive port selections within each session). E: latencies of individual mice to move from the active port to center port during acquisition. Horizontal dashed lines denote means ± SE. F: latencies of individual mice across all acquisition trials in one session. Scale bar = 5 s. G: active and inactive port pokes during the final extinction session compared with the cue-induced reinstatement session which demonstrates reinstated reward-seeking (**P < 0.01, repeated-measures ANOVA, active port selections during extinction vs. reinstatement; ns, not significant). H and I: linear regressions showing active port responses during cue-induced reinstatement was not predicted by the average latency during acquisition (H) or percent change in body wt (Δ%, I). Dashed lines indicate the 95% confidence intervals. n = 7 mice, all of which were male throughout this manuscript. Illustrated icons used in this figure and throughout the paper were created with BioRender (www.biorender.com). ITI, intertrial interval.

Sucrose Self-Administration Acquisition, Extinction, and Reinstatement Behavioral Task

Mice were placed on a 24-h water restriction schedule to promote task engagement, where their body weights were maintained at 90 ± 1.8% (interanimal range averaged across the whole experiment: 86.5%–96.5%) of their presurgery body weights, as previously described (12). Behavioral testing started following 5–8 days of this water restriction schedule. All behavioral testing occurred during the light cycle, between 1200–1700 h in successive daily 1-h sessions consisting of four phases. For all phases, mice were gently placed in the testing chamber, and their electrode arrays were connected to a flexible tether for collection of electrophysiological data. In phase 1, nose-poke selection of either the left or right ports resulted in an immediate solenoid-triggered delivery of 4 µL 10% w/v sucrose solution through the center port lick spout, and simultaneously, presentation of the light and tone together for 10 s. The 10-s cue presentation period served as a time-out period when no subsequent port entries were reinforced. A fixed-ratio 1 (FR1) schedule of reinforcement was used. Mice advanced to the next phase once they acquired at least 20 rewards per session, which was achieved by all mice in one session. During phase 2, called acquisition, the preferred port, i.e., the left or right port in which the mouse most poked, was designated as the active port, and the other port was designated as the inactive port. During acquisition, selections of the active port produced the same response-outcome contingencies as in phase 1 (i.e., reward delivery via the center port simultaneously with tone and light cue presentation for 10 s). Selections of the inactive port resulted in no light or tone presentation nor reward delivery. Mice advanced to the next phase once they completed at least three sessions of acquisition and demonstrated ≥80% preference for the active port. During phase 3, extinction, selection of either port resulted in no consequences (neither light, tone, nor reward). Mice advanced to the next phase once they demonstrated a ≥60% decrease in responses at the active port compared to their final acquisition session. Phase 4, called reinstatement, was a single-session test in which selection of the previously active port resulted in the presentation of cues (tone and light cue presentation together for 10 s) but were not reinforced with sucrose, and selection of the inactive port resulted in no consequences. One mouse did not display increased active port responding during reinstatement and therefore its behavioral data and neural data during this phase of the task were both excluded from analysis.

Histology

At the end of the behavioral experiment, mice were overdosed with sodium pentobarbital (Fatal-Plus, Patterson Veterinary) and transcardially perfused with cold 0.9% saline followed by 10% phosphate-buffered formalin. Brains were removed and stored at 4°C in 30% sucrose formalin, coronally sectioned at 40 µm on a freezing microtome, and slide-mounted with a mounting media containing a 4′,6-diamidino-2-phenylindole counterstain (DAPI-Fluoromount-G, SouthernBiotech, Birmingham, AL). Implant locations were examined using a Nikon Eclipse Ti2e fluorescent microscope. If electrode wires did not terminate in the TuS and NAc according to a standard brain atlas (38), the mouse was removed from further data analysis. One mouse had electrode wires that only terminated in the NAc and not the TuS, in which case only that mouse’s NAc units were analyzed. All other mice had correct placement of electrodes in both the NAc and TuS.

In Vivo Electrophysiology

Electrophysiological recordings of NAc and TuS units were acquired during the entirety of each session. The output of the electrode arrays was amplified with a digital headstage (Intan Technologies, Los Angeles, CA), acquired at 24.4 kHz, and monitored along with timestamped behavioral events in Synapse (Tucker-Davis Technologies). One electrode wire served as a local reference. Electrode arrays were fixed in place and no attempt was made to record from new populations of neurons on different sessions. Therefore, to account for the possibility of recording from the same neuron across sessions, only one session per phase per mouse was analyzed.

Processing and Analysis of Electrophysiology Data

We performed the following extraction and analysis steps on the TuS and NAc recording data. Importantly, all steps included mixtures of channels containing units from both brain regions, and whenever possible, semiautomated routines were incorporated to prevent experimenter influence of the outcomes.

1) Spike sorting. Offline spike sorting using Spike2 (Cambridge Electronic Design, Cambridge, UK) employed a combination of template matching and cluster cutting based on principal component analysis. Putative single neurons were defined as having <2% of their spikes occur within a 2-ms refractory period (12), otherwise they were removed from subsequent analysis.

2) Spike time extraction. Following identification of single units, trial-by-trial spike times aligned to each behavioral event (i.e., active, inactive, and center port pokes) were extracted and imported into MATLAB (MathWorks, R2017a). Spike density functions were calculated based on convolving spike trains with a function resembling a postsynaptic potential (39). Average firing rates were measured in 50-ms bins. The baseline firing rate for each neuron was averaged across a 2-s period (7 to 9 s following the active poke, which coincides with the “time-out” period in which sucrose consumption had ceased and neural activity related to planned behavior for the next trial had not yet started). Spike density functions for each analysis window were normalized to each neuron’s average background firing.

3) Measuring neural spiking variability. Fano factors (40, 41) were computed to assess variability in neural spiking across trials during each phase of the task. Within each trial during the background period, the mean spike count was divided by the variance, and all values were averaged within neurons. As there were no differences in background average Fano factors within each session (data not shown), the distributions of average Fano factors are shown combined.

4) Identifying task-modulated neurons. Two analysis windows relative to port entry (i.e., the onset of the active, center, or inactive poke) were defined: “approach” (−500 to 0 ms) and “response” (0 to 500 ms). For each neuron, the average firing rate during the background period was compared with the average firing rate during each analysis window via t test as described previously (42). The magnitude of modulation was determined by evaluating the average absolute Z-scored normalized firing rate within each analysis window. The duration of modulation was evaluated by counting the number of 50-ms bins within each analysis window that had firing rates ≥ 3 SD above or below the mean background firing rate.

Statistical Analysis

Statistical tests met assumptions of normality. The selection of this sample size of animals and the total numbers of units recorded from each brain region was based on previous studies in which statistically rigorous outcomes were tested/achieved using similar sample sizes of ventral striatum neurons (11, 12). Due to differences in the number of sessions required for mice to reach criteria to advance to the next phase, we evaluated the number of active and inactive port pokes during the final five acquisition sessions, final three extinction sessions, and the single cue-induced reinstatement session using repeated-measures ANOVA, with “port type” (active vs. inactive responses) and “session” as the within-subjects factors, with Bonferroni-adjusted post hoc comparisons where noted. Simple linear regressions were constructed to predict the number of each mouse’s active port responses during reinstatement based on the latency to move from the active port to the center port and body weight change as a percentage of pretesting body weight (Δ%).

Grubbs’ test was used to detect a sampling bias related to the numbers of neurons contributed by each mouse. Interregional distributions of background firing rates and Fano factors across all three sessions were compared using the Kolmogorov–Smirnov test. To determine whether there was evidence of interrelated firing between TuS and NAc neurons, we generated spike-triggered spike histograms for cell-cell pairs in 1-ms bins (total time ±0.4 s relative to the trigger spike). Cell-cell pairs were considered dependent if ≥90% of spikes occurred within 2 ms of each other.

Within each session, task-modulated units were analyzed by t tests with a false discovery rate Q of 1% using the Benjamini and Hochberg method, where each neuron’s averaged firing rate during the specified analysis window (“approach” or “response”) was compared to its average firing rate during the background period. The magnitude and duration of modulation of neurons were compared using two-way ANOVAs, with “analysis window” (approach versus response) and “brain region” (TuS versus NAc) as within-subjects factors. The χ² tests were used to compare the proportion of task-modulated units versus nonmodulated units within each structure to determine whether the ensemble sizes were significantly different. Separately, χ² tests were used to compare the proportion of TuS and NAc units modulated during one analysis window versus both analysis windows, as well as task-modulated units with increased versus decreased firing rates. Simple linear regressions were constructed to predict the number of nose-pokes based on mean firing rates during the corresponding poke as noted in each figure. The α values were set to 0.05 where appropriate. Means ± SE are reported unless otherwise indicated. Analyses and figures were generated using MATLAB and/or Prism (GraphPad, v. 8.2).

RESULTS

Our goal was to identify interregional differences in how neurons in the TuS and NAc are active during motivated behavior, to provide insights into the possible functional specializations of the ventral striatum subregions. To address this, we trained a total of eight water-restricted mice that were previously implanted with multisite electrode arrays into their TuS and NAc to perform a sucrose-reinforced operant task as illustrated in Fig. 1, A–C. This task allowed investigations into neural activity as animals 1) approached ports that triggered both sucrose delivery and presentation of discrete sensory cues, as well as 2) while animals were consuming sucrose. Furthermore, the task involved multiple phases including acquisition, extinction, and reinstatement for a thorough test of neural activity when sucrose availability, as well as the cues that predicted sucrose availability, changed. We sought to provide a simple quantification of the behavior from mice in this task to, later, most appreciate the neural data that were simultaneously acquired.

The testing phases occurred on separate days (see materials and methods) in an operant box with all cues emitting from a single wall, which also housed three extruded nose-poke ports (Fig. 1A). Neural recordings were acquired throughout all testing phases from mice connected to a flexible tether. During acquisition (Fig. 1, B and C), selection of the active port by means of a nose-poke resulted in the presentation of a light + tone cue (10-s duration) and simultaneously the delivery of a small palatable fluid reward through the center port (4 µL, 10% sucrose solution). Selection of the inactive port on the contralateral side of the chamber wall neither resulted in a cue nor reward. During the extinction phase, selections of either port resulted in neither the presentation of cues nor reward (Fig. 1C). Finally, during the cue-induced reinstatement session, active port responses triggered the same cue presentations as during acquisition, but these responses were not rewarded with sucrose (Fig. 1C). The average number of acquisition and extinction sessions was 5.29 ± 0.18 and 3.57 ± 0.43, respectively (interanimal ranges: 5–6 and 3–6, respectively).

As expected, mice displayed a strong preference for the active port throughout acquisition and at the beginning of extinction, and by the final two extinction sessions, their responses for either port were statistically indistinguishable (port type × sessions interaction, F_7,84=12.89, P < 0.0001; post hoc comparisons, P < 0.003 for each session noted in Fig. 1D). During each mouse’s final acquisition session, the average latency to select the center port after the active port was 1.17 ± 0.15 s (interanimal range: 0.74–1.89 s, Fig. 1, E and F). As shown in Fig. 1F, which plots each animal’s latency throughout a single behavioral session, the latency between active port and center port pokes varied throughout the testing session within mice, with some mice mostly engaging rapidly within trials (mouse 1, red), whereas others (mouse 7, magenta) often displayed breaks in between periods of rapid responding. The average intertrial interval (ITI) during acquisition, including the enforced 10-s time-out period was 14.12 ± 2.97 s (interanimal range: 3.8–27.1 s, data not shown).

During the cue-induced reinstatement session that occurred the day after the final extinction session, mice displayed robust reinstated reward-seeking elicited by presentation of the cues. This is evident by the increased engagement with the active port compared with their previous extinction session (port type × sessions interaction, F_1,12 = 10.19, P = 0.008; post hoc comparisons, t₇ = 4.830, P = 0.001, for active responses; t₇ = 0.316, P > 0.999, for inactive responses; Fig. 1G). Motivational drive during acquisition, assessed by the latency to transition from the active to center port, did not predict the number of active port pokes during cue-induced reinstatement (F_1,5 = 0.0006, P = 0.981, R² = 0.0001; Fig. 1H). Furthermore, changes in body weight from each mouse’s baseline weight before water restriction did not predict the numbers of active port pokes during cue-induced reinstatement (F_1,5 = 0.506, P = 0.501, R² = 0.092, Fig. 1I). Both comparisons are consistent with that fact that all mice were water-restricted within similar ranges during behavior. Taken together, and consistent with previous literature (43), this task allows assaying of reward-taking and reward-seeking. Moreover, these quantifications of response latencies inform the time windows whereby we will, as presented in detail throughout this paper, assess neural dynamics in relation to discrete behavioral events (i.e., the approach to the ports and the response at the ports).

Ventral Striatum Subregions Are Composed of Neurons with Similar Background Firing Rates and Firing Variability

Throughout the behavioral phases, we acquired single-unit activity, simultaneously, from neurons in the NAc and TuS (Fig. 2A). The implanted multiwire arrays consisted of eight tungsten electrodes arranged as a group that extended 1 mm beneath another group of eight electrodes. This design was selected to monitor TuS and NAc units in the same hemisphere while further ensuring sufficient dorsal-ventral segregation that would exclude unintentional monitoring of neurons in the ventral pallidum that, to some extent, spans between the TuS and NAc. We targeted the medial aspect of the TuS (Fig. 2B) due to several reports that this aspect of the TuS is influential in motivated behavior (10, 14, 16, 44). Further, since the electrode array was designed to sample neurons separated by ∼1 mm (dorsal-ventral), the NAc neurons were acquired from the NAc region 1 mm dorsal of the TuS. This included the NAc core and to a lesser degree, the shell (Fig. 2B).

Figure 2.

Similar background dynamics of tubular striatum (TuS) and nucleus accumbens (NAc) neurons. A: illustration of sagittal mouse brain indicating location of electrodes within the NAc and TuS. B: illustration of postmortem-confirmed electrode placements overlaid on a reference atlas adapted from Ref. 38 by permission from Elsevier. Ci: distribution of average background firing rates of all NAc and TuS neurons. Cii: distribution of Fano factor values estimated during background period for all NAc and TuS neurons. Background firing rates and trial-to-trial variability were similar between structures (Kolmogorov–Smirnov test). n = 8 mice, 241 neurons.

A total of 241 well-isolated neurons were acquired from the eight mice: 128 TuS neurons (18.3 ± 2.1 per mouse) and 113 NAc neurons (14.1 ± 2.2 per mouse). Table 1 shows the number of units contributed by each mouse across the three phases. There was no sampling bias in the number of units contributed by each mouse (P > 0.05, Grubbs’ test). We began our analysis into the neural activity by first performing a basic characterization of their background firing. This is important since differences in background firing may confer consequences on the identification of changes in firing throughout behavior. The distributions of background firing rates were similar between structures (D₂₄₁ = 0.085, P = 0.785, Kolmogorov–Smirnov test; Fig. 2Ci). The average background firing rates for all neurons sampled were also similar (unpaired t test, t₂₃₉ = 1.317, P = 0.189; median and range, respectively: for TuS, 1.7 Hz, 0–42.6 Hz; for NAc, 1.8 Hz, 0.1–52.7 Hz). These firing rates, which were captured 9–7 s before the active port selection and averaged across all trials, are expected, since over 90% of neurons in both the TuS and NAc are medium spiny neurons, which are reported to display firing rates less than 5 Hz (12, 15, 45, 46). Importantly, this further supports that we did not sample neurons from the neighboring ventral pallidum, since pallidum neurons typically display far greater firing rates (47). As a second analysis into background activity, we sought to characterize the trial-to-trial spiking variability during the background period. For this analysis, we quantified variability by computing each neuron’s Fano factor (40, 41), which was calculated during the background period of each trial and subsequently averaged. The background firing variability was also highly similar between the TuS and NAc (D₂₄₁ = 0.088, P = 0.745, Fig. 2Cii). Together, these results indicate that neurons in ventral striatum subregions display similar modes of background firing.

Table 1.

Numbers of neurons acquired from each mouse

	Acquisition		Extinction		Reinstatement
Mouse No.	NAc	TuS	NAc	TuS	NAc	TuS
1	3	6	1	4	2	5
2	10	9	5	11	7	7
3	6	8	9	6	9	8
4	3	5	1	3	4	3
5	4	7	3	5	5	8
6*	5		6		4
7	6	5	3	4	4	4
8#	5	8	8	12			Grand total:
Total neurons	42	48	36	45	35	35	241
Mean ± SE	5.3 ± 0.8	6.9 ± 0.6	4.5 ± 1.1	6.4 ± 1.4	5 ± 0.9	5.8 ± 0.9

^*This mouse did not have electrodes in the TuS and thus only contributed NAc neurons.

^#This mouse did not display an increase in active port poking during reinstatement and therefore did not contribute neural data toward reinstatement.

NAc, nucleus accumbens; TuS, tubular striatum.

Next, although there is no evidence supporting direct synaptic connectivity between TuS and NAc neurons (22), we nevertheless examined whether there is any functional connectivity between NAc and TuS neurons by examining spike-triggered spike histograms for cell-cell pairs (n = 156) in 1-ms bins (see methods). Cell-cell pairs were considered dependent if ≥90% of spikes occurred within 2 ms of each other. From this analysis, we did not find any evidence of dependent/interrelated neural activity across the two regions, with 0/156 units displaying coupled firing.

Neural Representation of Instrumental Responding and Reward Acquisition in Ventral Striatum Subregions

As a starting point in probing for differences among these ventral striatum regions in their activity during instrumental responding and reward acquisition, we first examined population-averaged Z-scored firing rates of all TuS and NAc units, combined, during the acquisition phase (Fig. 3A). We reasoned that this analysis would provide insights into how the ventral striatum operates as a whole during task engagement before later attempting to identify subregion differences. Although NAc activity has been carefully monitored during similar tasks as used here [e.g., (45, 46)], none have sampled equally throughout these two subregions.

Figure 3.

Ventral striatum neural dynamics during acquisition. A: peristimulus time histograms (PSTHs) of Z-scored firing rates of ventral striatal (VS) neurons (n = 90) during acquisition, aligned to responses to the active port (i), center port (ii), and inactive port (iii). Horizontal black bars above each PSTH depict interquartile range of duration of time the mice remained in each port, with the center port duration extending to 4 s (not shown in its entirety as indicated by the horizontal arrow). B and C: single-unit PSTHs computed across all trials, and raster plots from the first 10–35 trials from an example TuS neuron (B) and a NAc neuron (C) derived from different mice, aligned to responses to the active port (i) and inactive port (ii). Red rasters in i represent the mouse’s center port entry to obtain reward for each trial. Dashed vertical lines indicate onset of port entry. Horizontal purple and green bars above each PSTH depict interquartile range of poke durations. Scale bars = 2 Hz. NAc, nucleus accumbens; TuS, tubular striatum.

We observed changes in firing upon instrumental responding at the active port (Fig. 3Ai) and quite prominent increases in firing as mice approached and poked the center port to obtain the reward (Fig. 3Aii). Firing rates began to increase ∼200 ms before center poke which gradually returned to baseline levels within ∼1 s (Fig. 3Aii). Notably, a transient increase in firing was observed ∼800–400 ms before center poke, which, based on the average latency to transition from active-to-center port (Fig. 1E), would correspond to the animals’ active port poke. There was no notable change in firing during inactive port pokes (Fig. 3Aiii); however, the averaged firing rates here were derived from far fewer trials than for firing rates aligned to active and center port responses, as anticipated based upon the low salience of the inactive port. Together these results indicate, as expected (3, 18, 46, 48), that populations of ventral striatum neurons are active during discrete moments of engagement in reinforcer-motivated operant tasks, specifically upon reinforced instrumental responding and subsequent reward consumption.

Going forward, we separately analyzed TuS and NAc unit populations. In Fig. 3, B and C, we display peristimulus time histograms (PSTHs) from two example neurons. In these examples, the TuS neuron (Fig. 3Bi), but not the NAc neuron (Fig. 3Ci), displayed increased firing corresponding to the active port poke (Fig. 3Bi) that resembled the activity observed in the population-level averages in Fig. 3Ai. These example neurons in both structures began to ramp up activity ∼700 ms following the active poke, which was determined to correspond with the subsequent center port selection (Fig. 3, Bi and Ci, red rasters). Therefore, as expected, these example TuS and NAc units each displayed increased firing upon center poke when they obtained reward, which gradually returned to baseline levels. We also examined firing rates during inactive port pokes, which are composed of far fewer trials compared with active and center port pokes. As expected, the neurons did not display notable changes in their firing during the inactive poke (Fig. 3, Bii and Cii).

Examination of the averaged firing rates from all recorded neurons revealed that TuS neurons displayed more pronounced changes in firing relative to active port selection compared with those in the NAc (Fig. 4Ai), whereas firing rates aligned to the center port selection appeared to be increased in both regions, albeit to a lesser magnitude in the NAc than TuS (Fig. 4Aii). Additionally, neither region appeared to alter their firing rates upon inactive port selection (Fig. 4Aiii). Examination of population-wide firing rates during extinction (Fig. 4B) and cue-induced reinstatement (Fig. 4C), which are composed of fewer trials than during acquisition, further suggests region-based differences in activity. For instance, TuS neurons displayed higher firing rates upon active port selection during extinction compared with NAc neurons (Fig. 4Bi).

Figure 4.

Population-averaged responses in TuS and NAc firing rates to different phases of the task. A: peristimulus time histograms (PSTHs) of Z-scored firing rates of TuS neurons (n = 48) and NAc neurons (n = 42) during acquisition, aligned to responses to the active port (i), center port (ii), and inactive port (iii). The average number of trials comprising each acquisition response is 158.1 for active port responses, 130.5 for center port responses, and 12.5 for inactive port responses. n = 8 mice. B: PSTHs of firing rates of TuS (n = 45) and NAc (n = 36) neurons during extinction, aligned to responses to the active port (i), center port (ii), and inactive port (iii). The average number of trials comprising each figure is 65.6 for active port responses, 79.4 for center port responses, and 28.9 for inactive port responses. n = 8 mice. C: PSTHs of firing rates of TuS (n = 35) and NAc (n = 35) neurons during reinstatement, aligned to responses to the active port (i), center port (ii), and inactive port (iii). The average number of trials comprising each figure is 50 for active port responses, 75.1 for center port responses, and 19.9 for inactive port responses. Horizontal black bars above each PSTH depict interquartile range of duration of time the mice remained in each port, with Aii extending to 4.01 s. n = 7 mice. Shaded region surrounding averaged firing rates represents ± SE. NAc, nucleus accumbens; TuS, tubular striatum.

Different Sizes of Ensembles Are Recruited, and Are Recruited Distinctly, in the TuS versus the NAc during Reward-Taking

Given the aforementioned findings, we hypothesized that these population-averaged differences in firing dynamics across task types were driven by ensembles that were uniquely recruited in terms of both the ensemble size and the magnitudes and durations of activity within those recruited ensembles. Related to that, we further predicted that the size of the recruited ensembles and their dynamics would evolve throughout the phases of the task due to the unique salience and motivational outcomes with which each task phase is associated. To test these hypotheses, we identified task-modulated units that were modulated during the port approach and the response periods. Neurons whose firing rates were significantly increased or decreased within 500 ms before the onset of the port poke compared with background were classified as “approach”-modulated neurons (−500 to 0 ms), and neurons whose firing rates were significantly modulated up to 500 ms after the onset of the port poke compared with background were classified as “response”-modulated neurons (0 to 500 ms). During the acquisition session, sometimes mice were still in the center port consuming the drop of fluid reward beyond this 500-ms window (Fig. 3Aii, horizontal black bar). Therefore, we selected this cutoff as a conservative time point to ensure the window was not influenced by trials where mice may have already left the port.

Among the three different task phases (acquisition, extinction, reinstatement) and the three different response types (active poke, center poke, or inactive poke), similar proportions of TuS and NAc neurons were recruited in all instances except during acquisition’s approach to the active port (Fig. 5). Indeed, a significantly greater proportion of TuS neurons were recruited compared with NAc neurons [TuS: 52.1%, 25/48 neurons, vs. NAc: 23.8%, 10/42 neurons; χ² (1, N = 90) = 7.53, P = 0.003], indicating a greater recruitment of neurons in the TuS immediately before the instrumental response to obtain the reward. Notably, this recruitment pattern was behaviorally specific since we observed similarly sized populations of TuS and NAc neurons significantly modulated during both analysis windows across all port pokes and task phases (overlapping regions of the Venn diagrams in Fig. 5). Going forward, we restricted our analyses to only neurons identified as recruited (viz., significantly modulated in their firing rates, Fig. 5) in each task and response type, including those that were modulated during both analysis windows. Additionally, we did not further analyze firing rates aligned to inactive pokes due to the low number of modulated units observed and the few trials from which these averaged firing rates were derived. Although this conservative approach resulted in a smaller population of neurons contributing data, it allowed for a more rigorous assessment of behaviorally relevant neural dynamics that were unobstructed by the activity of unmodulated neurons.

Figure 5.

Proportions of neurons recruited during each task phase. TuS and NAc neurons that were significantly modulated during the approach (−500 ms to 0 ms) to each port (active, center, and inactive), the response period (0 to 500 ms), or both analysis windows, across each task phase. There was a significantly larger proportion of modulated neurons in the TuS than NAc during the approach to the active port during acquisition (*P < 0.05, TuS vs. NAc, χ² test), whereas recruitment of TuS and NAc neurons were similar across all other task phases and port selections. Values in each Venn diagram represent the numbers of modulated neurons. NAc, nucleus accumbens; TuS, tubular striatum.

Strikingly, during acquisition, the majority of approach-modulated TuS neurons displayed increased firing rates during approach to the active port, whereas the majority of approach-modulated NAc neurons showed decreased firing rates [χ² (1, n = 35) = 3.33, P = 0.041; Fig. 6Ai pie charts]. This indicates that TuS and NAc neurons engage during reward-guided behaviors with opposite directions of responding. Opposite directions of responding were also observed for the ensembles recruited during approach to the center port [χ² (1, n = 41) = 8.88, P = 0.002; Fig. 6Bi pie charts]. Importantly, although the number of neurons that displayed increased or decreased firing differ between regions, their averaged firing rates appeared to have similar dynamics: a transient increase ∼200 ms before the poke among neurons with increased firing, and a less-pronounced decrease in firing among neurons with decreased firing in both regions (Fig. 6, Ai and Bi, PSTHs). To further compare firing dynamics, we transformed firing rates into Z-scored firing (see materials and methods). Since TuS and NAc neurons differ in their directionality of firing, as do some neurons within each structure, we then converted the Z-scored values into absolute Z-scores to allow comparisons of changes in magnitudes and durations independent of directionality. We found that regardless of their response direction (i.e., increased or decreased firing rates), approach-modulated neurons had similar magnitudes of firing across regions (active port: F_1,66 = 0.0005, P = 0.944, Fig. 6Aii; center port: F_1,78 = 2.737, P = 0.102, Fig. 6Bii). Further, approach-modulated neurons also had similar durations of significant modulation (active poke: F_1,66 = 1.273, P = 0.263, Fig. 6Aiii; center poke: F_1,78 = 0.292, P = 0.59, Fig. 6Biii).

Figure 6.

TuS and NAc neurons differentially represent reward-taking during acquisition. Peristimulus time histograms (PSTHs) of Z-scored firing rates ± SE of TuS and NAc neurons that display significantly modulated firing rates during the approach to the active port (Ai) and center port (Bi). PSTHs of Z-scored firing rates ± SE of TuS and NAc neurons that display significantly modulated firing rates during the response to the active port (Ci) and center port (Di). In Ai–Di top each, pie charts depict the proportion of significantly modulated neurons that display increased and decreased firing in each region and the percentage per group. *P < 0.05, **P < 0.01, χ² tests. In Ai–Di bottom each, 2-D histograms (50-ms bins) depict the timing of the firing rates of each neuron’s significant bins, organized in descending order by the number of significant bins detected. Aii–Dii: absolute Z-score values averaged during the approach period and the response period. ^P = 0.06, *P < 0.05, two-way ANOVA. Aiii–Diii: duration of modulated activity measured in 50-ms bins during the approach period and the response period. **P < 0.05, two-way ANOVA. n = 8 mice. NAc, nucleus accumbens; TuS, tubular striatum.

Among response-modulated neurons (Fig. 6, C and D), again, the majority of TuS neurons showed increased firing rates whereas the majority of NAc neurons showed decreased firing rates [active: χ² (1, n = 32) = 4.499, P = 0.019, Fig. 6Ci pie charts; center: χ² (1, n = 59) = 4.515, P = 0.019, Fig. 6Di pie charts]. Upon examination of the PSTHs, TuS neurons displayed a large increase in firing immediately before the active port poke, which was absent in NAc neurons (Fig. 6Ci). This is not surprising since 11 out of 15 of these neurons were also categorized as approach-modulated compared with five out of 17 NAc neurons (Fig. 5). Firing rates surrounding the time of the center port poke were more similar across regions (Fig. 6Di, PSTH). However, TuS response-modulated neurons had a greater magnitude of modulation than those in the NAc [active poke: main effect of region (F_1,60 = 4.978, P = 0.029), Fig. 6Cii; center poke: main effects of region (F_1,114 = 5.337, P = 0.023) and analysis window (F_1,114= 13.01, P = 0.0005), significant pairwise comparisons shown in Fig. 6Dii]. The duration of modulation was similar among these NAc and TuS populations, with longer durations during the response versus the approach period regardless of region [active poke: main effect of analysis window (F_1,60 = 13.28, P = 0.0006), Fig. 6Ciii; center poke: main effect of analysis window (F_1,114= 10.80, P = 0.001), Fig. 6Diii]. Together, these results demonstrate that TuS neuron activity during acquisition differs from that in the NAc in terms of 1) larger ensembles being recruited during approach to the active port, 2) ensembles with more neurons that primarily show increases in firing, and 3) ensembles with greater magnitudes in their responses during the active and center port pokes.

These differences noted, TuS and NAc approach-modulated neurons did display similar magnitudes and durations of firing. Overall, across several different measures, the TuS more prominently represents reward-taking than the NAc.

Interregional Differences in the Dynamics of Modulated Neurons during Extinction

During the extinction phase of the task, active port pokes no longer resulted in presentation of the cues or reward (see materials and methods). Despite no reward nor cues being available, just as observed during acquisition, we found opposite directions of responding between TuS and NAc neurons during extinction testing. All TuS neurons that were modulated during the approach to the active and center ports increased their firing rates, whereas the majority of modulated NAc neurons displayed decreased firing for the active port [χ² (1, N = 14) = 15.83, P < 0.0001, Fig. 7Ai pie charts] and increased firing (albeit a smaller proportion than TuS) for the center port [χ² (1, N = 24) = 4.8, P = 0.016, Fig. 7Bi pie charts]. TuS firing rates surrounding the time of the active poke were increased ∼500 ms before the poke and were sustained for almost 1 s after the poke, whereas NAc firing rates appeared much lower (Fig. 7Ai, PSTH). Firing rates surrounding the time of the center poke were largely more similar across regions (Fig. 7Bi, PSTH). Upon further comparison of neural dynamics, approach-modulated TuS neurons had greater magnitudes than NAc neurons surrounding the time of the active poke (main effect of brain region, F_1,24 = 11.15, P = 0.003, Fig. 7Aii) and center poke (main effect of brain region, F_1,44 = 14.81, P = 0.0004, Fig. 7Bii). Despite these differences, TuS and NAc approach-modulated neurons showed similar durations of modulation (active port: F_1,24= 0.562, P = 0.461, Fig. 7Aiii; center port: F_1,44 = 0.003, P = 0.958, Fig. 7Biii).

Figure 7.

TuS and NAc neurons differentially represent extinguished reward-seeking. Peristimulus time histograms (PSTHs) of firing rates ± SE of TuS and NAc neurons that display significantly modulated firing rates during the approach to the active port (Ai) and center port (Bi). PSTHs of firing rates ± SE of TuS and NAc neurons that display significantly modulated firing rates during the response to the active port (Ci) and center port (Di). In Ai–Di top each, pie charts depict the proportion of significantly modulated neurons that display increased and decreased firing in each region and the percentage per group. *P < 0.05, **P < 0.01, ****P < 0.0001, χ² tests. In Ai–Di bottom each, 2-D histograms (50-ms bins) depict the timing of the firing rates of each neuron’s significant bins, organized in descending order by the number of significant bins detected. Aii–Dii: absolute Z-score values averaged during the approach period and the response period. ^P = 0.06, *P < 0.05, **P < 0.01, ***P < 0.001, two-way ANOVA. Aiii–Diii: duration of modulated activity measured in 50-ms bins during the approach period and the response period. *P < 0.05, two-way ANOVA. N = 8 mice. NAc, nucleus accumbens; TuS, tubular striatum.

Among response-modulated neurons (Fig. 7, C and D), a greater majority of TuS neurons displayed increases in their firing rates compared with NAc neurons for the active port [χ² (1, n = 23) = 8.75, P = 0.002, Fig. 7Ci pie charts] as well as for the center port [χ² (1, n = 30) = 4.36, P = 0.022, Fig. 7Di pie chart]. Additionally, although modulated TuS neurons displayed greater magnitudes in firing rates than those in the NAc surrounding the active poke (Fig. 7Ci, PSTH; absolute Z-scores, main effects of region: F_1,42= 8.434, P = 0.006; main effect of analysis window: F_1,42 = 4.688, P = 0.036, pairwise comparisons trending as noted in Fig. 7Cii), they showed similar magnitudes of modulation during the center port response (Fig. 7Dii) and similar durations of modulation for both active (F_1,24 = 0.562, P = 0.461, Fig. 7Ciii) and center responses (F_1,44= 0.003, P = 0.958, Fig. 7Diii). In contrast to acquisition, where modulated neurons from both structures had longer modulated firing rates during active and center responses (Fig. 6, Ciii and Diii), extinction resulted in response-modulated neurons with 1) greater magnitudes and 2) longer durations of modulated firing during only the center response, and not the active response, in both brain regions (main effects of analysis window: magnitude: F_1,56 = 9.674, P = 0.003, Fig. 7Dii; duration: F_1,56 = 4.784, P = 0.033, Fig. 7Diii). Together, these results point toward a partial shift in the interregional firing dynamics observed during active and center port pokes in the absence of cues and reward, as the mice have learned different response-outcome contingencies in this phase.

Interregional Differences in the Dynamics of Modulated Neurons during Cue-Induced Reinstatement

As the final phase of behavioral testing, mice were allowed to poke during a single cue-induced reinstatement session in which the cue presentation, even in the absence of sucrose, induced reinstated reward-seeking (Fig. 1). Among neurons with significantly modulated firing during active port approach (Fig. 8A), as observed during acquisition and extinction, there was a greater majority of TuS neurons that had increased firing rates compared with NAc neurons [χ² (1, n = 17) = 13.82, P = 0.0001, Fig. 8Ai pie charts]. However, there were no differences between regions in the firing direction for these neurons when aligned to the center port poke [χ² (1, n = 21) = 1.53, P = 0.15, Fig. 8Bi pie charts], nor were there any differences in their magnitude or duration of modulation for these neurons during either the active poke (magnitude: F_1,30= 0.178, P = 0.676, Fig. 8Aii; duration: F_1,30 = 2.269, P = 0.142, Fig. 8Aiii) or center poke (magnitude: F_1,38 = 0.083, P = 0.775, Fig. 8Bii; duration: F_1,38 = 2.254, P = 0.142, Fig. 8Biii).

Figure 8.

TuS and NAc neurons differentially represent reward-seeking during cue-induced reinstatement. Peristimulus time histograms (PSTHs) of firing rates ± SE of TuS and NAc neurons that display significantly modulated firing rates during the approach to the active port (Ai) and center port (Bi). PSTHs of firing rates ± SE of TuS and NAc neurons that display significantly modulated firing rates during the response to the active port (Ci) and center port (Di). In Ai–Di top each, pie charts depict the proportion of significantly modulated neurons that display increased and decreased firing in each region and the percentage per group. ***P < 0.001, χ² tests. In Ai–Di bottom each, 2-D histograms (50-ms bins) depict the timing of the firing rates of each neuron’s significant bins, organized in descending order by the number of significant bins detected. Aii–Dii: absolute Z-score values averaged during the approach period and the response period. **P < 0.01, two-way ANOVA. Aiii–Diii: duration of modulated activity measured in 50-ms bins during the approach period and the response period. *P < 0.05, ***P < 0.001, two-way ANOVA. n = 7 mice. NAc, nucleus accumbens; TuS, tubular striatum.

Among neurons modulated during the response to the active port, the TuS had a larger majority of neurons with increased firing compared with NAc neurons [χ²(1, n = 22) = 11.289, P = 0.0004, Fig. 8Ci, pie charts] and a longer duration of modulation (main effect of region, F_1,40 = 5.647, P = 0.022, Fig. 8Ciii), yet similar magnitudes (F_1,40 = 2.202, P = 0.146, Fig. 8Cii). For neurons modulated during the center port response, NAc and TuS neurons showed similar directionality in firing, with 91% of neurons in both regions increasing their firing rates (Fig. 8Di pie charts). These neurons had greater magnitudes during the response compared with the approach to the center port (main effect of analysis window, F_1,40 = 9.962, P = 0.003, Fig. 8Dii; see also Fig. 8Di, PSTH) and longer durations of firing (main effect of analysis window, F_1,40= 6.916, P = 0.012, Fig. 8Diii). Interestingly, it appears that these effects were driven by NAc neurons (for magnitude, trending toward a region effect: F_1,40 = 3.864, P = 0.056, Fig. 8Dii; for duration, main effect of region: F_1,40 = 4.092, P = 0.049, post hoc comparisons revealing significant differences for NAc neurons between the two analysis windows, P < 0.027, Fig. 8Diii). These results indicate that during reinstatement, the TuS maintains its representation of instrumental behavior (i.e., active port response) while the NAc increases its representation of the center port selection, suggesting that the NAc may be, in the presence of previously reinforced cues, representing unexpected reward omission.

TuS Firing Rates Predict Sucrose-Seeking Behavior

The results up to this point have uncovered that more TuS neurons are recruited during the reward acquisition task, and that among all recruited neurons, they display regionally unique firing dynamics during many aspects of behavior related to reward-taking, reward-seeking, and also extinction. In several instances, activity of TuS neurons reflected more pronounced representations of active and center port selections in terms of larger ensembles recruited, larger proportions of modulated neurons with increased firing rates, greater absolute Z-scored firing, and longer durations of significant firing than NAc neurons. Given these interregional differences, we hypothesized that TuS activity would predict at least some aspects of behavior. More specifically, we hypothesized that the averaged population-wide TuS firing rates during the corresponding period when the behavior was exhibited might predict the number of port pokes during each task phase since port pokes is the most direct read-out of motivation in this task. To examine this, we calculated linear regressions for each mouse’s number of active or center port selections during each task phase and the Z-scored average firing rates of all neurons from each mouse during the corresponding analysis window. As hypothesized, TuS firing could indeed predict the degree of reward-taking and seeking. First, we found a significant positive linear relationship between the number of active port selections during reinstatement and TuS firing rates during the approach to the active port (F_1,4 = 11.21, P = 0.029, R² = 0.737, Fig. 9A). This relationship was not observed when performing the same analysis on NAc firing (F_1,5 = 3.558, P = 0.118, R² = 0.416). Second, we found a significant positive linear relationship between the number of center port selections during the extinction phase and TuS firing rates during the response period, but not NAc firing rates (TuS: F_1,4 = 8.202, P = 0.046, R² = 0.672; NAc: F_1,5= 1.909, P = 0.226, R² = 0.276, Fig. 9B). These finding suggest that higher firing rates in the TuS, but not the NAc, correspond with the magnitude of reward-seeking during extinction and reinstatement. Intriguingly, we also found a significant negative linear relationship between the number of center port selections during the acquisition phase and TuS firing rates during the response period, which was not observed with NAc firing rates (TuS: F_1,4 = 30.93, P = 0.005, R² = 0.885; NAc: F_1,5 = 0.849, P = 0.399, R² = 0.145, Fig. 9C). This implies that suppression of TuS activity may impart a greater motivational drive to obtain reward. No other comparisons during other phases nor behaviors were found to be significant, including port entry durations. Together, these results suggest that TuS activity may be directly consequential to the occurrence of reward-seeking and taking in a manner that stands apart from that in the NAc.

Figure 9.

Ventral striatal firing rates’ ability to predict reward-seeking. A: TuS firing rates during the approach to the active port during reinstatement predicted the number of active port responses whereas NAc firing rates corresponding to the same time did not. B: TuS firing rates during extinction center port response predicted the number of center port responses (firing rates from NAc neurons did not). C: TuS firing rates during acquisition predicted center port responses while NAc neurons were not. Firing data are Z-scored average firing rates of all neurons from each mouse during the corresponding analysis window. *P < 0.05, simple linear regression. NAc, nucleus accumbens; ns, not significant; TuS, tubular striatum.

DISCUSSION

Goal-directed behavior requires evaluation of outcome value, internal motivational state, and environmental stimuli that may signal the availability of reinforcers. Particularly, a previously neutral stimulus gains saliency when repeatedly paired with a reinforcer. Such salient stimuli increase anticipation, or the expectation of future reward availability, and drive reward-seeking (viz., actions to obtain the reward), even after periods of abstinence and/or extinction (35, 49, 50). Although this process underlies motivation for “natural” reinforcers such as food and sex, it also is thought to underlie aberrant behaviors such as compulsive drug use, binge-eating, and/or relapse (7, 51). Here we sought to further our knowledge of the neural regulation of motivated behaviors by comparing neural dynamics within two regions of the ventral striatum—the NAc and TuS, to test the hypothesis that TuS and NAc activity is distinct during reward-taking and reward-seeking. There were several notable similarities, but also striking differences (see Fig. 10 for a summary of the main findings) that together highlight not only the TuS’s pronounced representation of sucrose-taking and sucrose-seeking that surpassed the NAc on several measures, but overarching this, that these ventral striatum subregions appear to act with some level of specialization, which may, through careful future studies, be found to critically underlie motivated behaviors including addiction.

Figure 10.

Summary of major findings. Colors of boxes indicate whether the given measurement (ensemble size, direction, magnitude, or duration of firing) was greater/more pronounced in the TuS (purple) versus the NAc (green), or the same (black). FR, firing rate.

Interregional Similarities, but also Differences in Ensemble Sizes and Neuronal Responsivity

Since no direct comparisons between TuS and NAc activity have been previously performed, here we compared three main components of activity. These included 1) the number of neurons significantly modulated in discrete time windows, 2) the direction of their modulation (excited or suppressed), and 3) their magnitude and duration of firing.

Regarding the first component, we found similar patterns of recruitment between the TuS and NAc across the different phases, except for a larger ensemble of TuS versus NAc neurons during acquisition’s active port approach (the action that triggered sucrose delivery and cue presentation). These findings indicate that although many components of this task elicit similar recruitment patterns, suggesting a global characteristic of the ventral striatum, the TuS is more greatly recruited by instrumental responding to acquire a reward than the NAc. Neuronal ensembles are sparsely activated sets of neurons that undergo synaptic alterations thought to encode information related to stimuli and associative memories (52–54). Through c-Fos analysis, such neural ensembles are reported to be activated during different phases of goal-directed tasks, such as self-administration versus extinction (55). Here, we extend this work by taking advantage of the high temporal resolution of in vivo electrophysiology to examine dynamics of modulated neurons within discrete, behaviorally relevant windows of time. By grouping ensembles across mice to create (virtual) ensembles, we observed that the larger TuS ensemble size relative to NAc was specific to the time before (<500 ms) the active port selection. Given that similarly sized ensembles were observed across all other windows in time and task phases, the NAc and TuS may similarly participate in most aspects of motivation and reward-taking, whereas the TuS may be more specialized to process information related to the initiation of instrumental behavior, specifically when the mouse “knows” the response will be reinforced (such as during acquisition). Indeed, earlier work from our laboratory uncovered that the majority of TuS neurons start to increase their firing before an instrumental response (18). The TuS may therefore be a critical component of the network needed to initiate an internally generated (i.e., not necessarily cue-guided) sequence of reward-motivated behaviors.

Regarding the latter analyses (i.e., direction and magnitudes of firing), one of the most striking findings is that during several aspects of the task, the TuS possessed a larger proportion of modulated neurons with increased firing rates, whereas the NAc possessed a larger proportion of modulated neurons with decreased firing rates. This effect was observed across multiple phases and windows of time, including active and center port approach during both acquisition and extinction (Figs. 6 and 7), but it was not observed in the time surrounding the center port during reinstatement (Fig. 8). In several instances, particularly during extinction, a larger response magnitude was also observed in the TuS than the NAc (for a summary of findings, see Fig. 10). In other words, modulated TuS neurons are more excited than NAc neurons during reward-taking in both the presence and absence of discrete cues. During reinstatement, this effect was observed during the active port approach and response (i.e., that which triggered cue presentation) but not center port, suggesting that TuS and NAc neurons become more similar in their direction of firing when cues that predicted prior reward availability are present after extinction training. However, the data set from extinction and reinstatement phases comprised fewer trials compared with acquisition, as expected given that their behavior is no longer being reinforced with sucrose as it was during acquisition. With these low trial numbers in mind, the analyses in these two task phases may be considered somewhat preliminary.

The NAc is considered to be more strongly recruited in more complex scenarios, such as tasks that involve multiple possible outcomes or uncertainty (56). In our present study, recordings from acquisition and extinction came from later sessions when the mice were well trained, whereas reinstatement was the only session involving new response-outcome contingencies. This may explain the increased duration and magnitude of modulation observed in NAc neurons during the center port poke. Based on these observations, we reason that the NAc may be more tuned to novel, unexpected components of motivated behavior while the TuS may be specialized for acting upon well-learned behaviors. This is consistent with several papers, including work from our group, that reported changes in the TuS following cue-associated reward-learning (11–15).

An underlying assumption in this study is that changes in neural activity are either directly or indirectly contributing toward the behaviors we measured. Although we did not test that these changes in firing are necessary for behavior, we did observe that the degree of reward-seeking behavior (i.e., the number of pokes) could be predicted by the average firing rates of TuS neurons at the time surrounding each mouse’s response. Conversely, average firing rates of NAc neurons did not predict any of the behaviors examined. That this effect was observed in population-wide activity of all neurons suggests a representation of reward-seeking by the TuS that may be relevant to behavior, possibly via information flow to shared downstream structures such as the ventral pallidum (9, 57). For instance, it is intriguing to consider that the ventral pallidum may interpret a difference in firing rates between the TuS and NAc (57) as pertinent information for the orchestration of a reward-guided motor response in manners that may explain previous reports of the dramatic mediation of drug responding by the TuS (10). Importantly, we emphasize that these findings do not infer causal roles (or lack thereof) of TuS or NAc firing for mediating reward-guided behavior in this task. Future work to identify critical roles of the activity we found here is certainly needed, and this study will serve as a foundation whereby to accomplish that important goal.

Sources of Heterogeneity within the Ventral Striatum

What might underlie the interregional differences we uncovered between the TuS and NAc? First, the TuS represents odor-reward associations (11–14) and receives dense innervation from the olfactory bulb and olfactory cortex (9, 20, 21, 58). We selected discrete nonolfactory cues (i.e., tone and light) to reduce the possibility of firing that may be biased to odors. This stated, activity within both structures can be shaped by non-olfactory stimuli as well, independent of reward value (59, 60).

Second, our NAc recordings featured neurons from both the core and shell components (61), albeit likely more from the core than shell, which is intentional given that the core is more directly involved in cue-guided reward-seeking (62, 63) and core neurons are more responsive to reward-predictive cues than shell neurons (64). In the shell, neuronal activity is tuned to sucrose reward and a majority of neurons decrease firing during sucrose consumption (2, 4). A more recent study identified populations of shell neurons that encoded sucrose concentration, palatability, and licking (31). Although we did not distinguish core versus shell activity, nor did we examine dynamics relative to licking frequency (an intriguing avenue of research for future studies), it is notable that our results identify TuS as well as NAc ensembles modulated by instrumental responding and cue-guided sucrose consumption, thereby suggesting that these behaviors are represented globally, although not equally, throughout the ventral striatum.

Finally, the principal cell type in both the TuS and NAc are GABA-ergic medium spiny projection neurons, the majority of which possess either dopamine D1 or D2 receptors (D1-MSN, D2-MSN, respectively) (65). We recently reported that stimulation of TuS D1-MSNs drives reinforcement and their activity tracks the reward value of conditioned stimuli (11). Opposing roles for D1-MSNs and D2-MSNs in motivation and reward, as well as cue-induced reinstatement, have been observed in the medial NAc (66–68), but not in the ventrolateral NAc or dorsal striatum during food-guided tasks (69, 70). Although our current work does not allow for identifying these neural subtypes, given that we observe bidirectionally modulated neurons in both regions (albeit in differing proportions), it is tempting to speculate that neurons with increased firing are D1-MSNs and neurons with decreased firing are D2-MSNs. This will require careful examination in future studies.

Conclusions

Altogether, this study is a first step toward directly contextualizing the TuS’s activity in relation to that in the NAc and by doing so establishes a critical framework for future research seeking to better understand the manners whereby the ventral striatum governs motivated behaviors. Specifically, these results add to the growing literature indicating that the TuS is linked to reward-taking and reward-seeking by highlighting that not only is TuS activity modulated during reward-taking and reward-seeking, but by many measures, and in several behavioral contexts (summarized in Fig. 10), the TuS seems to be more greatly partaking in their representation.

GRANTS

This work was supported by NIH Grants R01DC014443, R01DC016519, R01DA049545, and R01DA049449 to D. W. Wesson. K. N. Wright was supported by an F32DC018452 and a seed grant from the University of Florida Center for Smell and Taste.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

K.N.W. and D.W.W. conceived and designed research; K.N.W. performed experiments; K.N.W. analyzed data; K.N.W. interpreted results of experiments; K.N.W. prepared figures; K.N.W. and D.W.W. drafted manuscript; K.N.W. and D.W.W. edited and revised manuscript; K.N.W. and D.W.W. approved final version of manuscript.