Insights into the interaction between time and reward prediction on the activity of striatal tonically active neurons: A pilot study in rhesus monkeys

Abstract

Prior studies have documented the role of the striatum and its dopaminergic input in time processing, but the contribution of local striatal cholinergic innervation has not been specifically investigated. To address this issue, we recorded the activity of tonically active neurons (TANs), thought to be cholinergic interneurons in the striatum, in two male macaques performing self-initiated movements after specified intervals in the seconds range have elapsed. The behavioral data showed that movement timing was adjusted according to the temporal requirements. About one-third of all recorded TANs displayed brief depressions in firing in response to the cue that indicates the interval duration, and the strength of these modulations was, in some instances, related to the timing of movement. The rewarding outcome of actions also impacted TAN activity, as reflected by stronger responses to the cue paralleled by weaker responses to reward when monkeys performed correctly timed movements over consecutive trials. It therefore appears that TAN responses may act as a start signal for keeping track of time and reward prediction could be incorporated in this signaling function. We conclude that the role of the striatal cholinergic TAN system in time processing is embedded in predicting rewarding outcomes during timing behavior.

Keywords: acetylcholine; basal ganglia; interval timing; motivation.

FIGURE 1

Self‐timed reaching task and timing performance. (a) Temporal sequence of events in the task. Monkeys were required to initiate reaching movement after a specified time interval has elapsed. At the beginning of the trial, a visual cue (yellow light), either on the left or the right side, indicated the duration of the interval (Short or Long) in the range of seconds, each time interval being associated with a particular spatial location of the stimulus. Hatched horizontal lines indicate the minimum waiting period before movement initiation assigned to each location of the cue (time threshold). Correctly timed movements were reinforced with fruit juice immediately after target contact. (b) Distribution of movement onset times produced by monkeys in the task. Red parts of the histograms represent trials where movements started before reaching the time threshold (underestimation errors). The spread of the response distribution is proportional to the length of the interval to‐be‐timed, consistent with the scalar property of interval timing. Values correspond to the mean of the distributions ± SD. Dashed lines indicate the mean. Number of trials were 1370 and 1594 for monkey C and 1219 and 1146 for monkey D, for short‐ and long‐interval trials, respectively. (c) Reward frequency during task performance. Box plots of the percentage of correctly timed movements. Each point corresponds to the proportion of correct movements from the short‐ and long‐interval trials of a block. Boxes indicate 25–75 percentile ranges of the distributions and lines through boxes correspond to medians. Isolated dots indicate outliers. Percentage of correct movements differed between time intervals for both monkeys (*p < 0.01, paired t‐test).

FIGURE 2

Electrophysiological identification of TANs. (a) Scatterplot of the baseline firing rate and spike width duration for the two main striatal neuron classes. Each dot represents data from an individual neuron. TANs displayed higher spontaneous discharge rate and longer spike duration than phasically active neurons (PANs). (b) Two examples of task‐related activity modulation for each type of neuron. Each trial is displayed as a row of spikes (dots) with perievent time histogram above each raster plot. Activity is referenced to cue onset which is marked by vertical lines. The sequence of trials is shown chronologically from top to bottom in each raster plot, irrespective of the duration of the interval. Movement onsets are indicated by red dots.

FIGURE 3

Sensitivity of TANs to the timing cue. (a) Three example neurons responding to the cue. Left, nonselective response; Middle, short selective response; Right, long selective response. Same conventions as in Figure 2b, except that data are separated by interval duration and sorted by movement onset time. Hatched horizontal lines indicate the minimum waiting period before initiating the movement (time threshold). A few underestimation errors are visible at the top of each raster (red markers before reaching the time threshold). (b) Relative proportions of neurons responding to the cue in short‐ and/or long‐interval trials. The percentages were calculated from the total number of recorded neurons (n) in each monkey. (c) Changes in TAN activity after cue onset separately for short‐ and long‐interval trials for each neuron recorded. Each line indicates the data of one neuron. Thick lines indicate significant differences in the magnitude of changes in activity between time intervals (p < 0.05, Wilcoxon signed‐rank test). Boxes indicate 25–75 percentile ranges of the distributions and lines through boxes correspond to medians. Isolated dots indicate outliers. (d) Comparison of magnitudes of changes in TAN activity after the cue separately for the short and long time intervals. Data are indicated as decreases in percentage below baseline activity and expressed as means ± SEM, *p < 0.05 (Wilcoxon signed‐rank test). (e) Modulation of TAN responses to the cue, shown as population averages separately for the short and long time intervals. The averaged population activity is aligned on the cue onset marked by the vertical line and included both correctly timed movements and underestimation errors. Colored curves represent mean activity of TANs separately for short‐ and long‐interval trials calculated in nonoverlapping time bins of 10 ms. Shading indicates SEM. n, number of neurons included for population curves.

FIGURE 4

Sensitivity of TANs to reward. (a) Three example neurons responding to the cue and/or reward. Same conventions as in Figure 3a, except that activity is separately aligned on cue onset and reward delivery. A few underestimation errors are visible at the bottom of each raster aligned on reward (absence of red marker). (b) Percentages of different response types evoked by reward delivery. Same conventions as in Figure 3b. (c) Changes in TAN activity after cue onset separately for short‐ and long‐interval trials for each neuron recorded. Same conventions as in Figure 3c. (d) Comparison of magnitudes of changes in TAN activity after the delivery of reward separately for the short and long time intervals. Same conventions as in Figure 3c, except that only rewarded trials were included. **p < 0.01 (Wilcoxon signed‐rank test). (e) Relative proportions of TANs responding or not to the cue and/or reward.

FIGURE 5

Influence of movement timing on TAN responsiveness. (a) Scatter plots of the magnitude of TAN modulations after the cue and reward versus the movement onset time (MOT), separately for the short and long intervals. The red lines indicate the fit of a linear regression and shading indicates the 95% confidence interval from the regression. Spearman correlation coefficient. Vertical gray lines indicate time thresholds (monkey C: 1.3–2.3 s; monkey D: 1–2 s). (b) Magnitudes of TAN responses to the cue computed separately from trials with MOTs within consecutive 500‐ms periods. Plots show, separately for the two interval durations, mean magnitude of TAN responses to the cue and reward, according to the range of MOT values. Error bars represent SEMs. *p < 0.05, **p < 0.01 (Wilcoxon signed‐rank test).

FIGURE 6

Influence of timing accuracy on TAN responsiveness. (a) Activity of a single TAN during performance of the timing task. The sequence of trials is shown chronologically from top to bottom in the raster display (45 trials total). Green shading shows sequences of correctly timed movements performed repeatedly for more than 5 trials (trials 7–14 and 25–34). Same conventions as in Figure 2b, except that purple dots aligned on target contact indicate unrewarded trials (underestimation errors). (b) Trial‐by‐trial variation of TAN responsiveness to the cue and reward according to the trial number during sequences of correctly timed movements. Scatter plots of the magnitude of TAN modulations (ordinate) versus the number of consecutive correct trials (abscissa). Same conventions as in Figure 5a, except that each point corresponds to the average across neurons in each trial of the sequence.

FIGURE 7

Effects of interval duration on TAN activity in the Pavlovian conditioning task. (a) Licking movements after the onset of the cue. Traces of lick records are aligned on the cue onset for each interval duration. Licking started immediately after the presentation of the cue, indicating that animals have learned the reward predictive value of the stimulus. (b) Percentages of different response types evoked by the cue and reward. (c) Two example neurons displaying responses to the cue and reward. For each neuron, the top panel shows short‐interval trials and the bottom panel long‐interval trials. Same conventions as in Figure 3a. (d) Comparison of magnitudes of changes in TAN activity after the cue and reward separately for short‐ and long‐interval trials. **p < 0.01 (Wilcoxon signed‐rank test).

FIGURE 8

Sensitivity of TANs to the timing cue among striatal regions. (a) Recording sites of TANs responding to the cue in both monkeys. The location of all recorded TANs is plotted in the rostrocaudal direction on coronal sections from 5 mm anterior to 3–4 mm posterior to the anterior commissure (AC), with 1 mm intervals. Cd, caudate nucleus, Put, putamen. (b) Comparison of population activities of TANs aligned on cue onset, grouped for short and long intervals, between motor and associative striatum. Same conventions as in Figure 3d, except that curves are smoothed with a Gaussian filter (alpha = 0.04). (c) Differences in the magnitude of TAN responses to the cue between the motor and associative striatum according to interval duration (Wilcoxon signed‐rank test, p < 0.01).