Reward activity in ventral pallidum tracks satiety-sensitive preference and drives choice behavior

Abstract

A key function of the nervous system is producing adaptive behavior across changing conditions, like physiological state. Although states like thirst and hunger are known to impact decision-making, the neurobiology of this phenomenon has been studied minimally. Here, we tracked evolving preference for sucrose and water as rats proceeded from a thirsty to sated state. As rats shifted from water choices to sucrose choices across the session, the activity of a majority of neurons in the ventral pallidum, a region crucial for reward-related behaviors, closely matched the evolving behavioral preference. The timing of this signal followed the pattern of a reward prediction error, occurring at the cue or the reward depending on when reward identity was revealed. Additionally, optogenetic stimulation of ventral pallidum neurons at the time of reward was able to reverse behavioral preference. Our results suggest that ventral pallidum neurons guide reward-related decisions across changing physiological states.

Fig. 1. Dynamic preference driven by physiological state.

(A) Schematic of the “specific cues” task, where there were three trial types, each with a unique auditory cue. Correct lever presses on forced-choice trials led to delivery of the associated reward, while rats could choose to receive water or sucrose on free-choice trials. (B) Schematic of the “uncertain outcome” task. The choice trials (and cue) were the same as for the specific cues task, but the forced trials had a different auditory cue and required entry into the reward port rather than a lever press, after which either reward was delivered. (C) Example uncertain outcome session, depicting choice trials (colored, longer lines) and forced trials (black, shorter lines) for sucrose (top) and water (bottom), overlaid with preference (green). (D) Preference in each task for each of the five rats across four quarters of completed trials. (E) Mean (±SEM) lick rate relative to reward delivery across the four quarters (Q) of trials, split into forced sucrose (left) and water (right) trials. Sessions from both tasks are combined here. (F) Mean (±SEM) lick rate across 13 s, capturing nearly all of the reward-related licking (top), and within the bin used for neural analysis (0.75 to 1.95 s after delivery).

Fig. 2. Dynamic reward encoding occurs when outcome identity is revealed.

(A) Windows used for neural analysis relative to task events. (B) Schematic of the GLM used to predict individual neuron activity at time of cue and reward. (C) Proportion of neurons from each task with significant impacts of outcome, time, and outcome × time at the time of cue and time of reward. We focused our additional analysis on the outcome × time neurons at time of cue for the specific cues task and at the time of reward for the uncertain outcome task (marked in green outlines). (D) Raster from an example outcome × time neuron from the specific cues task, aligned to sucrose (left) and water (right) forced trial cues. Shading indicates window for neural analysis. (E) Mean (±SEM) sucrose (top) and water (bottom) cue-evoked firing for all outcome × time neurons from the specific cues task across the four quarters of trials. Gray shading indicates window for neural analysis. (F) Mean (±SEM) binned firing for these neurons in this window. (G) Raster from example outcome × time neuron from the uncertain outcome task, aligned to sucrose (left) and water (right) delivery on forced trials. Shading indicates window for neural analysis. (H) Mean (±SEM) sucrose-evoked (left) and water-evoked (right) firing for all outcome × time neurons from the uncertain outcome task across the four quarters of trials. Gray shading indicates window for neural analysis. (I) Mean (±SEM) binned firing for these neurons in this window.

Fig. 3. Reward-evoked activity accurately predicts behavioral preference.

(A) Normalized reward-evoked firing of an example outcome × time neuron in the uncertain outcome task on forced trials overlaid with preference for water (left) and sucrose (right). (B) Distribution of correlation coefficients between firing rate and preference for outcome × time (blue or orange) and non–outcome × time (gray) neurons on forced water trials (left) or forced sucrose trials (right). Vertical line is mean. *P < 1 × 10⁻²⁴, Wilcoxon rank sum test comparing outcome × time and non–outcome × time neurons on water trials, and P < 1 × 10⁻¹³ for sucrose trials. (C) Example fits of three models we considered to describe the activity of outcome × time neurons in the uncertain outcome task at time of reward delivery: satiety, preference, and mixed, which linearly combined satiety and preference. (D) Distribution of best-fit model for all outcome × time neurons, determined with cross-validated likelihood. (E) From three example sessions, the choices of the rats across the session and the preference estimated with a logistic function. (F) Mean (±SEM) estimate of preference from fits of the mixed model to the outcome × time neurons from these sessions. (G) Correlation between neural estimate and behavioral estimate of preference for each outcome × time neuron. (H) Estimates of the indifference point (sucrose and water equally preferred) from the neural and behavioral (±SE) models. (I) Across all uncertain outcome sessions, outcome × time neurons had preference estimates with higher correlations with the behavioral estimate than the remaining non–outcome × time neurons did (P < 1 × 10⁻²¹, Wilcoxon rank sum test). (J) Outcome × time neurons’ estimates of indifference point were closer to the behavioral estimate than the remaining non–outcome × time neurons (P < 1 × 10⁻¹⁵, Wilcoxon rank sum test).

Fig. 4. Stimulation of VP at reward delivery biases choice behavior.

(A) Task design for optogenetic stimulation experiment. Rats chose between sucrose and maltodextrin. Levers were retracted after press. (B) Optic fibers and virus containing ChR2 (or GFP control) were implanted/infused bilaterally in VP, but only the right hemisphere was stimulated in this experiment. (C) During the test session, on maltodextrin trials, VP was photostimulated unilaterally for 5 s at 40 Hz, beginning with maltodextrin delivery, or whenever the rat first entered the port thereafter, to overlap with maltodextrin consumption. (D) Preference for sucrose versus maltodextrin on choice trials at baseline (after training), on test session, and for five recovery days after without laser. There was a significant interaction between day and group across these seven sessions (F_6,126 = 10.6, P < 0.00000001). Post hoc Tukey tests (corrected for multiple comparisons) revealed a significant difference between groups on test day (P < 0.000001) and the first recovery day (P < 0.00001). (E) Preference (smoothed) on choice trials across test session for individual rats (left) and averaged for each group (right).