Cell Type-Specific Membrane Potential Changes in Dorsolateral Striatum Accompanying Reward-Based Sensorimotor Learning

Abstract

The striatum integrates sensorimotor and motivational signals, likely playing a key role in reward-based learning of goal-directed behavior. However, cell type-specific mechanisms underlying reinforcement learning remain to be precisely determined. Here, we investigated changes in membrane potential dynamics of dorsolateral striatal neurons comparing naïve mice and expert mice trained to lick a reward spout in response to whisker deflection. We recorded from three distinct cell types: (i) direct pathway striatonigral neurons, which express type 1 dopamine receptors; (ii) indirect pathway striatopallidal neurons, which express type 2 dopamine receptors; and (iii) tonically active, putative cholinergic, striatal neurons. Task learning was accompanied by cell type-specific changes in the membrane potential dynamics evoked by the whisker deflection and licking in successfully-performed trials. Both striatonigral and striatopallidal types of striatal projection neurons showed enhanced task-related depolarization across learning. Striatonigral neurons showed a prominent increase in a short latency sensory-evoked depolarization in expert compared to naïve mice. In contrast, the putative cholinergic striatal neurons developed a hyperpolarizing response across learning, driving a pause in their firing. Our results reveal cell type-specific changes in striatal membrane potential dynamics across the learning of a simple goal-directed sensorimotor transformation, helpful for furthering the understanding of the various potential roles of different basal ganglia circuits.

Keywords: goal-directed sensorimotor transformation; reward-based learning; striatum; whole-cell membrane potential.

Graphical Abstract

Figure 1.

Whisker detection task. (A) Depiction of the whisker-based sensory detection task: mice learned to associate a brief (1 ms) downward deflection of their right C2 whisker with the availability of a water reward. (B) Whole-cell (V_m) recordings were performed in the DLS of head-restrained mice during the first training session (day 1) of this task (naïve, green) or in mice that had been trained for 7 or more days (expert, blue). (C) Trials were classified as hit if the mouse licked within the 1 s response window that followed whisker stimulus (grey area), as miss if the mouse did not lick, as false alarm if it licked when no whisker stimulus was presented (catch trials) and as correct rejection if it did not lick on catch trials. Stimulus and catch trials were randomly interleaved and separated by a randomized 4–12 s inter-trial interval. In addition, the mouse was required to not lick in the 4 s before a trial was initiated to prevent compulsive licking. (D) The probability of licking in the response window of naïve (n = 26 mice, green) and expert (n = 20 mice, blue) mice during the V_m recordings in trials with a whisker stimulus (hit rate, left) or catch trials without a whisker stimulus (false alarm rate, right) (Wilcoxon–Mann–Whitney test). Open circles indicate individual cells, closed circles with error bars indicate mean ± standard error of mean (SEM). (E) The discriminability (d') of trials with and without whisker stimuli compared between naïve and expert mice (Wilcoxon–Mann–Whitney test). Open circles indicate individual cells, closed circles with error bars indicate mean ± SEM. (F) Response time of naïve and expert mice (Wilcoxon-Mann-Whitney test). Open circles indicate individual cells, closed circles with error bars indicate mean ± SEM.

Figure 2.

Cell-type identification. (A) Fluorescent biocytin staining (cyan) of 2 example neurons recorded in the DLS of two different genetically-engineered mice expressing tdTomato in dSPNs (red) and Green Fluorescent Protein (GFP) in iSPNs (green). In the top example cell, the biocytin signal colocalizes with tdTomato revealing this neuron to be a dSPN. In the bottom example cell, biocytin signal colocalizes with GFP revealing this neuron to be an iSPN. (B) Left: example of a biocytin-labelled aspiny TAN. Right: example of a biocytin filled spiny SPN. Inset scale bar 5 µm. (C) Example coronal drawing (AP −0.22 mm) showing cell locations of dSPNs (red), iSPNs (green), and TANs (blue) recorded in naïve (open circles) or expert (filled circles) mice. (D–F) The baseline V_m (D), baseline action potential (AP) firing rate (E) and rheobase (F) of dSPNs, iSPNs, and TANs combining both naïve and expert mice (comparison between dSPN vs iSPN, dSPN vs TAN, and iSPN vs TAN, Wilcoxon–Mann–Whitney test with Bonferroni correction). (G) Grand average action potential firing rates versus the amount of current injected (F–I curves) for dSPNs, iSPNs, and TANs combining both naïve and expert mice. (H) Example V_m traces from a dSPN, an iSPN, and a TAN during a hyperpolarizing current step, demonstrating a voltage sag in the TAN. (I) Quantification of the voltage sag of dSPNs, iSPNs, and TANs combining both naïve and expert mice (comparison between dSPN vs iSPN, dSPN vs TAN, and iSPN vs TAN, Wilcoxon–Mann–Whitney test with Bonferroni correction). (J) Action potential waveforms for dSPNs, iSPNs, and TANs. (K) Action potential half-width for dSPNs, iSPNs, and TANs combining both naïve and expert mice (comparison between dSPN vs iSPN, dSPN vs TAN, and iSPN vs TAN, Wilcoxon–Mann–Whitney test with Bonferroni correction). See also Supplementary Figures 1 and 2.

Figure 3.

Learning is accompanied by cell-type specific changes in membrane potential dynamics aligned to the time of whisker stimulation. (A) Left: Whisker-stimulus triggered V_m average for hit trials in dSPNs recorded from naïve mice (green, n = 17 cells) and expert mice (blue, n = 12 cells). Right: higher temporal resolution enabling visualization of the early response. The color-coded bar indicates P-values for the difference between expert and naïve mice in 10 ms time windows (Wilcoxon–Mann–Whitney test). (B) The early slope (20–30 ms after stimulus), early ΔV_m (20–50 ms after stimulus), mid ΔV_m (50–250 ms after stimulus), and late ΔV_m (0.5–1 s after stimulus) in dSPN naïve vs expert mice (Wilcoxon–Mann–Whitney test). (C) Left: whisker stimulus triggered V_m average for hit trials in iSPNs recorded from naïve mice (n = 7 cells) and expert mice (n = 13 cells). Right: higher temporal resolution. The color-coded bar indicates P-values for the difference between expert and naïve mice in 10 ms time windows (Wilcoxon–Mann–Whitney test). (D) The early slope (20–30 ms after stimulus), early ΔV_m (20–50 ms after stimulus), mid ΔV_m (50–250 ms after stimulus), and late ΔV_m (0.5–1 s after stimulus) in iSPNs from naïve vs expert mice (Wilcoxon–Mann–Whitney test). (E) Left: whisker stimulus triggered V_m average for hit trials in TANs recorded from naïve mice (n = 5 cells) and expert mice (n = 5 cells). Right: higher temporal resolution. The color-coded bar indicates P-values for the difference between expert and naïve mice in 10 ms time windows (Wilcoxon–Mann–Whitney test). (F) The early slope (20–30 ms after stimulus), early ΔV_m (20–50 ms after stimulus), mid ΔV_m (100–400 ms after stimulus), and late ΔV_m (0.5–1 s after stimulus) in TANs from naïve vs expert mice (Wilcoxon–Mann–Whitney test). (G) Schematic summary of the learning-related changes in dSPNs. The early sensory response evoked by whisker stimulation increases in dSPNs across learning. The later component of the response likely relates to motor and premotor inputs to dSPNs, which occur earlier in expert mice, since they lick with shorter latency. These two changes could account for the overall change in V_m dynamics of dSPNs across learning shown in panels A and B. (H) Whisker deflection will drive neurons in cortex and thalamus to release glutamate onto neurons in the DLS. During task learning and execution, the mouse receives a reward in hit trials upon licking after the whisker stimulus, which likely causes a transient increase in dopamine concentration. Increased dopamine could contribute to promoting long-term potentiation of synaptic input onto the D1R-expressing dSPNs. Enhanced sensory-evoked glutamatergic responses in dSPNs from presynaptic thalamic or cortical neurons could increase the probability of licking through inhibition of neurons in substantia nigra pars reticulata, which contains tonically active inhibitory neurons. This might result in disinhibition of motor thalamus and brainstem motor nuclei, thus contributing to driving licking as a motor response to whisker deflection after reward-based learning. See also Supplementary Figures 3 and 4.