Signaling Incentive and Drive in the Primate Ventral Pallidum for Motivational Control of Goal-Directed Action

Abstract

Processing incentive and drive is essential for control of goal-directed behavior. The limbic part of the basal ganglia has been emphasized in these processes, yet the exact neuronal mechanism has remained elusive. In this study, we examined the neuronal activity of the ventral pallidum (VP) and its upstream area, the rostromedial caudate (rmCD), while two male macaque monkeys performed an instrumental lever release task in which a visual cue indicated the forthcoming reward size. We found that the activity of some neurons in VP and rmCD reflected the expected reward size transiently following the cue. Reward size coding appeared earlier and stronger in VP than in rmCD. We also found that the activity in these areas was modulated by the satiation level of monkeys, which also occurred more frequently in VP than in rmCD. The information regarding reward size and satiation level was independently signaled in the neuronal populations of these areas. The data thus highlighted the neuronal coding of key variables for goal-directed behavior in VP. Furthermore, pharmacological inactivation of VP induced more severe deficit of goal-directed behavior than inactivation of rmCD, which was indicated by abnormal error repetition and diminished satiation effect on the performance. These results suggest that VP encodes incentive value and internal drive and plays a pivotal role in the control of motivation to promote goal-directed behavior.SIGNIFICANCE STATEMENT The limbic part of the basal ganglia has been implicated in the motivational control of goal-directed action. Here, we investigated how the ventral pallidum (VP) and the rostromedial caudate (rmCD) encode incentive value and internal drive and control goal-directed behavior. Neuronal recording and subsequent pharmacological inactivation revealed that VP had stronger coding of reward size and satiation level than rmCD. Reward size and satiation level were independently encoded in the neuronal population of these areas. Furthermore, VP inactivation impaired goal-directed behavior more severely than rmCD inactivation. These results highlight the central role of VP in the motivational control of goal-directed action.

Keywords: goal-directed behavior; monkey; motivation; reward; rostromedial caudate; ventral pallidum.

Figure 1.

Reward size task and behavioral performance. a, Sequence of a trial. b, Cue stimuli. Either the stripe set (left column) or image set (right column) was used to inform the reward size (1, 2, 4, or 8 drops of liquid). c, Error rate (mean ± SEM) as a function of reward size for monkeys TA and AP. d, Mean error rate as a function of normalized cumulative reward for the two monkeys. Each color indicates reward size. Curves were best fit of Equations 1 and 2 with c = 2.1, λ = 1.9 for monkey TA and c = 4.7, λ = 2.5 for monkey AP.

Figure 2.

Recording sites in VP and rmCD. a, b, Recording sites of VP and rmCD, respectively. Left, CT/MR fusion image showing the position of an electrode. Right, Schematic diagrams representing the locations of the recorded neurons: positive reward size coding neurons (red), negative reward size coding neurons (blue), and noncoding neurons (white). Representative slices from monkey TA were used. Cd, Caudate nucleus; Put, putamen; GPe, external segment of the globus pallidus; AC, anterior commissure.

Figure 3.

Reward size coding in VP and rmCD. a, b, Example activity of VP neurons showing positive (a) and negative (b) reward size coding during cue period, respectively. Left, Raster spikes and spike density function (SDF, σ = 10 ms) were aligned at task events. The colors correspond to the respective reward sizes. Red bars above shading indicate significant linear correlation at the task period (p < 0.05, linear regression analysis); gray bars indicate no significant correlation (p > 0.05). Right, Relationship between firing rate (mean ± SEM) during cue period and reward size. Regression lines are shown in red (p < 0.01). c, d, Examples of rmCD neurons. Schema of the figures are the same as in a and b. e, f, Left, Population activities of VP neurons that were classified into positive (e) and negative (f) reward size coding neurons, respectively. Curves and shading indicate mean and SEM of normalized activity to the baseline aligned at task events. Digits in each panel indicate the number of reward size coding neurons at each task period. Right, Relationship between normalized neuronal activity during cue period (mean ± SEM) and reward size. g, h. Population activities of rmCD neurons. Schema of the figures are the same as in e and f.

Figure 4.

Time course of reward size coding. a, b, Time-dependent change of effect size (R²) depicted with heat plots for VP neurons (a) and rmCD neurons (b). Each panel shows data from each task event. Neurons are sorted by the coding latency from the cue. Top rows show positive reward size coding neurons and bottom rows show negative reward size coding neurons. c, d, Average effect size of positive (top) and negative (bottom) reward size coding neurons around task events for VP neurons and for rmCD neurons. Digits in each panel indicate the number of reward size coding neurons at each task period.

Figure 5.

Coding latency of the expected reward size. a, c, Effect size histogram aligned to cue for positive (a) and negative (c) reward size coding neurons reconstructed from Figure 4, c and d. The data from VP (red and blue) and rmCD (gray) are depicted in the same panel. Vertical lines indicate the median of coding latency. b, d, Distribution of coding latency for positive (b) and negative (d) reward size coding neurons. Asterisk indicates significant difference between VP and rmCD (p < 0.01, rank-sum test).

Figure 6.

Dual coding of reward size and satiation level in single neurons. a–d, Example VP neuron showing negative reward size coding and positive satiation level coding. a, b, Raster plots and SDF are shown for each reward condition (a) and for each session period (b). c, Relationship between firing rate during the cue period (mean ± SEM) and reward size are plotted for the session period. Linear regressions are shown as colored lines. d, Waveforms of each spike (orange) and average waveform (black) during the first minute (left) and last minute (right) in the recording session. e–h, Example rmCD neuron showing positive reward size coding and negative satiation level coding. Schema of the figures are the same as in a–d.

Figure 7.

Separate coding of reward size and satiation level in VP and rmCD. a, Scatter plot of standardized correlation coefficients (cue period) for reward size (abscissa) against satiation level (ordinate) are shown for VP neurons (left) and rmCD neurons (right), respectively. Colors indicate significant reward size coding neurons (orange), satiation level coding neurons (green), reward-and-satiation coding neurons (yellow), and noncoding neurons (gray). Yellow plots in light-yellow shading areas indicate motivational-value coding neurons. Histograms in the main panels illustrate the distribution of coefficients with significant neurons with satiation level and reward size. b, Proportion of VP neurons (solid line) and rmCD neurons (dashed line) that showed satiation level coding (left), reward size coding (center), and motivational value coding (right) for each task period (ITI, cue, prerelease, and reward periods). Asterisks indicate significant difference between VP and rmCD (p < 0.05 with Bonferroni correction, χ² test). For motivational value coding (right), the proportion of neurons with pseudo-motivational value coding by chance is shown in gray. No significant difference was observed between data and estimation (p > 0.05 with Bonferroni correction, χ² test).

Figure 8.

Behavioral change due to inactivation of VP and rmCD. a, Error rate in control (black) and rmCD inactivation (green) sessions for monkey RO (left) and monkey RI (right). Digits in panels indicate number of sessions. b, Injection sites in VP. Top, Representative CT/MR fusion image for confirmation of injection sites in monkey BI. Bottom two panels, Location of injection indicated by magenta dots. c, Error rate in control (black) and VP inactivation (magenta) sessions for monkeys BI (left) and RI (right), respectively. d, Change of early error rate (left) and length of repetitive errors (right) by rmCD inactivation (green) and VP inactivation (magenta). Asterisks indicate significant differences from control session (*p < 0.05, **p < 0.01, rank-sum test). e, Lever grip time in the first and latter half of control session (black) and of VP inactivation session (magenta) (mean ± SEM). Asterisks indicate significant differences by post hoc Tukey's test (*p < 0.05).