Stable representation of sounds in the posterior striatum during flexible auditory decisions

Abstract

The neuronal pathways that link sounds to rewarded actions remain elusive. For instance, it is unclear whether neurons in the posterior tail of the dorsal striatum (which receive direct input from the auditory system) mediate action selection, as other striatal circuits do. Here, we examine the role of posterior striatal neurons in auditory decisions in mice. We find that, in contrast to the anterior dorsal striatum, activation of the posterior striatum does not elicit systematic movement. However, activation of posterior striatal neurons during sound presentation in an auditory discrimination task biases the animals' choices, and transient inactivation of these neurons largely impairs sound discrimination. Moreover, the activity of these neurons during sound presentation reliably encodes stimulus features, but is only minimally influenced by the animals' choices. Our results suggest that posterior striatal neurons play an essential role in auditory decisions, and provides a stable representation of sounds during auditory tasks.

Fig. 1

Activation of distinct subregions of the dorsal striatum produced different effects on movement. a Top: experimental setup. Optogenetic stimulation in freely moving mice of direct-pathway neurons from one of four different sites in the dorsal striatum: anterior striatum (left or right) and posterior striatum (left or right). Middle: Coronal brain slice. Green dots indicate the tip of fixed optical fibers implanted in the anterior striatum (gray) confirmed postmortem. Bottom: purple lines indicate the stimulation sites by movable optical fibers implanted in the posterior striatum. b Representative head angle trace over one trial of unilateral stimulation at each site. The blue bar represents the laser pulse (1.5 s) delivered in each trial. Positive angles correspond to left rotation. c Average change in head angle by optogenetic stimulation in each mouse tested. Each gray circle is one trial, each filled circle is the average for one hemisphere of one mouse. Stimulation of anterior striatum (green) produced a significantly larger change in head angle compared to stimulation of the posterior striatum (purple) in either hemisphere (p = 0.034, Wilcoxon rank-sum test)

Fig. 2

Activation of direct-pathway posterior striatal neurons biased sound-driven decisions. a Schematic of the two-alternative choice sound frequency discrimination task. Mice initiated each trial by entering a center port and had to choose one of two side reward ports depending on the sound presented: low-frequency = left, high-frequency = right. b Psychometric performance for one behavioral session that included optogenetic activation of direct-pathway neurons in the left posterior striatum on 20% of trials. Error bars indicate 95% confidence intervals. c Psychometric performance for one behavioral session that included optogenetic activation of direct-pathway neurons in the right posterior striatum. d Change in the percentage of rightward choices during optogenetic stimulation for each hemisphere in each mouse tested. Each open dot represents one session, each filled dot represents the average bias for that hemisphere in one mouse (N = 3 mice, 10 sessions each hemisphere per mouse). Horizontal bars represent averages across all sessions for all mice. Stimulation produced significantly different biases in the left vs. the right hemisphere (p < 0.001, Wilcoxon rank-sum test)

Fig. 3

Inactivation of posterior striatal neurons impaired sound-driven decisions. a Mice were injected bilaterally with muscimol in the posterior striatum 30 min before they performed the two-alternative choice sound discrimination task. b Average psychometric performance for one mouse on sessions with injection of muscimol (four sessions) or saline control (four sessions). Error bars indicate 95% confidence intervals. c Average percentage of correct trials on each saline session (black) and each muscimol session (brown) for each mouse. Bars indicate average across sessions for each mouse. Muscimol inactivation significantly reduced the percentage of correct trials on each mouse (p = 0.021, Wilcoxon rank-sum test)

Fig. 4

Posterior striatal neurons displayed frequency-selective sound-evoked responses. a Example of sound responses from a posterior striatal neuron during the sound discrimination task. Yellow bar indicates the duration of the sound (100 ms). A box plot above the spike raster shows the distribution of center-port exit times. b Sound responses of a different posterior striatal neuron which showed only suppression of activity. Both example cells showed clear frequency selectivity. c Magnitude of sound-evoked response estimated from the 100 ms period during sound presentation for each neuron (all trials pooled together). A positive response index indicates an increase in activity compared to baseline spontaneous activity. A negative index, a decrease in activity. Neurons with statistically significant evoked responses (p < 0.05, Wilcoxon rank-sum test) are shown in black. d Sound selectivity index calculated by comparing neural responses to high- vs. low-frequency sounds based on the categorization boundary from the task. Neurons with statistically significant differences (p < 0.05, Wilcoxon rank-sum test) are shown in black

Fig. 5

Firing rate of posterior striatal neurons during movement depends on movement direction. a Activity from a posterior striatal neurons aligned to the moment the mouse left the center port and moved toward a reward port. Activity differed between right and left choices. Plot includes all trials (any stimulus frequency). b Activity from a different posterior striatal neuron showing the opposite movement selectivity. Box plots above spike rasters in a and b show the range of onset times for the sound stimuli. c Movement selectivity index: (I − C)/(I + C), where I and C are the average firing rates in the period 50–150 ms after leaving the center port for choices ipsilateral and contralateral to the recording hemisphere, respectively. N = 520 cells recorded from five mice. Cells with a statistically significant difference in activity during contralateral and ipsilateral choices (p < 0.05, Wilcoxon rank-sum test) are shown in black

Fig. 6

Choice influenced sound-evoked responses in a subset of posterior striatal neurons. a Response of one neuron to a stimulus near the categorization boundary, grouped according to the animal’s choice. Sound-evoked response for this neuron was modulated by the animal’s choice. The box plot above the spike raster shows the distribution of center-port exit times. b Activity of a different neuron showing no influence of choice on the sound-evoked response. c Influence of choice on sound-evoked activity (from the 100 ms period during sound) for all neurons that showed a response to stimuli near the categorization boundary (N = 67 cells from five mice). Less than 12% of neurons showed a significant modulation by choice (p < 0.05, Wilcoxon rank-sum test), shown in black

Fig. 7

Effect of rapid changes in sound-action associations on activity of posterior striatal neurons. a Switching task: the rewarded action associated with a sound of intermediate frequency changed from one block of trials to the next. b Example performance of one mouse (one session) as the contingency changes. Colors match those in a. c Influence of changing sound-action association on sound-evoked activity (in the 100 ms period during sound) for all neurons that showed a response to the switching stimulus (N = 155 cells from four mice). Less than 13% of neurons showed a significant change across sound-action contingencies (p < 0.05, Wilcoxon rank-sum test), shown in black. d Responses of one posterior striatal neuron to the stimulus of intermediate frequency for three blocks of trials. Only correct trials are included. Sound-evoked responses for this neuron changed systematically depending on the rewarded action associated with the stimulus. e Activity of a different neuron showing no change in sound-evoked responses across sound-action contingencies. Box plots above spike rasters show the distribution of center-port exit times