Multiple dynamic interactions from basal ganglia direct and indirect pathways mediate action selection

Abstract

The basal ganglia are known to be essential for action selection. However, the functional role of basal ganglia direct and indirect pathways in action selection remains unresolved. Here by employing cell-type-specific neuronal recording and manipulation in mice trained in a choice task, we demonstrate that multiple dynamic interactions from the direct and indirect pathways control the action selection. While the direct pathway regulates the behavioral choice in a linear manner, the indirect pathway exerts a nonlinear inverted-U-shaped control over action selection, depending on the inputs and the network state. We propose a new center (direct) - surround (indirect) - context (indirect) "Triple-control" functional model of basal ganglia, which can replicate the physiological and behavioral experimental observations that cannot be simply explained by either the traditional "Go/No-go" or more recent "Co-activation" model. These findings have important implications on understanding the basal ganglia circuitry and action selection in health and disease.

Figure 1.. The neuronal dynamics in SNr during the 2–8 s action selection task.

(A) Schematic diagram for the design of 2–8 s task. (B) Correct rate for wild type mice across 14 days training ($\mathrm{n}=10$ mice, one-way repeated-measures ANOVA, significant effect of training days, ${\mathrm{F}}_{\mathrm{13,117}}=32.54$, $\mathrm{p}<0.0001$). (C) Movement trajectory of an example mouse in correct (left panel) and incorrect (right panel) 8-s trials (gray line: trajectory of each trials; red/black line: the average trajectory). (D) Diagram of electrode array implanted into substantia nigra pars reticulata (SNr). (E) Firing Rate Index (FRI) of neuronal activity for all task-related SNr neurons in correct 8-s trials. The magnitude of FRI is color coded and the SNr neurons are classified as four different types based on the activity dynamics. (F-I) Averaged FRI for Type 1 (F, green squares indicating activities related to left choice), Type 2 (G, green squares indicating activities related to left choice), Type 3 (H), Type 4 (I) of SNr neurons in correct (red) and incorrect 8-s trials (gray). (J) The proportion of four types of SNr neurons. Type 1 and Type 2 are major types and significantly more than Type 3 and Type 4 (Z-test, $\mathrm{p}<0.05$). (K) Integrated SNr output defined as the subtraction of averaged FRI between Type 1 and Type 2 SNr neurons. (L) Averaged psychometric curve ($\mathrm{n}=10$ mice) of choice behavior. (M) The correlation between the Type 1 and Type 2 FRI subtraction and the behavioral choice ($\mathrm{R}=0.98$, $\mathrm{p}<0.0005$). Error bars denote s.e.m., same for below unless stated otherwise.

Figure 2.. SNr neuronal dynamics reflect action selection but not interval time or reward value.

(A) Task diagram of 2–8 s control task with 10% 16-s probe trials. (B) Percentage of behavioral choice in 2-s, 8-s and 16-s trials (blue: left choice; red: right choice) ($\mathrm{n}=9$ mice, paired t-test, $\mathrm{p}<0.05$). (C) Movement trajectory of an example mouse in 16-s trials (blue: left choice; red: right choice). (D) Averaged SNr Type 1 FRI in 16-s trials (red: left choice; black: right choice). Firing rates from 8s to 16s (highlighted area) are compared between left and right choice ($\mathrm{n}=26$ neurons, two-way repeated-measures ANOVA, significant difference between left and right choices, ${\mathrm{F}}_{\mathrm{1,25}}=6.646$, $\mathrm{p}=0.016$). (E) Averaged SNr Type 2 FRI in 16-s trials (red: left choice; black: right choice). Firing rates from 8s to 16s are compared between left and right choice ($\mathrm{n}=16$ neurons, two-way repeated-measures ANOVA, significant difference between left and right choices, ${\mathrm{F}}_{\mathrm{1,15}}=5.785$, $\mathrm{p}=0.029$). (F) Subtraction of FRI for SNr Type 1 and Type 2 neurons in 16-s probe trials (red: left choice; black: right choice). (G) Task design of 2–8 s standard task. (H) Percentage of behavioral choice in 2-s and 8-s trials (blue: left choice; red: right choice) ($\mathrm{n}=6$ mice, paired t-test, $\mathrm{p}<0.05$). (I) Movement trajectory of an example mouse in 8-s trials (blue: left choice; red: right choice). (J) Task design of reversed 2–8 s task. (K) Percentage of behavioral choice in 2-s and 8-s trials in the reversed 2–8 s task (blue: left choice; red: right choice) ($\mathrm{n}=6$ mice, paired t-test, $\mathrm{p}<0.05$). (L) Movement trajectory of the same mouse as (I) in 8-s trials in the reversed 2–8 s task (blue: left choice; red: right choice). (M) Averaged FRI of the SNr Type 1 neurons in correct 8-s trials ($\mathrm{n}=14$ neurons). (N) Averaged FRI of the SNr Type 2 neurons in correct 8-s trials ($\mathrm{n}=11$ neurons). (O) Integrated SNr output as the subtraction of FRI for SNr Type 1 (M) and Type 2 neurons (N) in the standard 2–8 s task. (P) Averaged FRI of the same neurons as (M) in correct 8-s trials of the reversed 2–8 s task. (Q) Averaged FRI of the same neurons as (N) in correct 8-s trials of the reversed 2–8 s task. (R) Integrated SNr output as the subtraction of FRI for SNr Type 1 (P) and Type 2 neurons (Q) in the reversed 2–8 s task.

Figure 3.. Neuronal activity of striatal D1- and D2-SPNs during action selection.

(A) FRI of neuronal activity for all task-related SPNs in correct 8-s trials. SPNs were classified as Type 1 – 4. (B-E) Averaged FRI for Type 1 (B), Type 2 (C), Type 3 (D), Type 4 (E) of SPNs in correct (red) and incorrect 8-s trials (gray). (F) Diagram of simultaneous neuronal recording and optogenetic identification of D1- vs. D2-SPNs in dorsal striatum. (G) Top panel: Raster plot for a representative D1-SPN response to 100 ms optogenetic stimulation. Each row represents one trial and each black dot represents a spike. Bottom panel: Peristimulus time histogram (PETH) aligned to light onset at time zero. (H) PETH for the same neuron as shown in (G) with a finer time scale. (I) Distribution of light response latencies for D1- and D2-SPNs. (J) Action potential waveforms of the same neuron in (G) for spontaneous (black) and light-evoked (orange) spikes ($\mathrm{R}=0.998$, $\mathrm{P}<0.0001$, Pearson’s correlation). (K) Principal component analysis (PCA) of action potential waveforms showing the overlapped clusters of spontaneous (black) and light-evoked (orange) spikes. (L) Proportion of D1-SPN subtypes. Type 1 neurons are significantly more than other three types of neurons in D1-SPNs (Z-test, $\mathrm{p}<0.05$). (M) Averaged FRI for Type 1 (blue) and Type 2 (red) D1-SPNs in correct 8-s trials. (N) Proportion of D2-SPN subtypes. (O) Averaged FRI for Type 1 (blue) and Type 2 (red) D2-SPNs in correct 8-s trials.

Figure 4.. Selective genetic knockout and ablation of D1- or D2-SPNs distinctly alters action selection.

(A) Correct rate of control ($\mathrm{n}=11$ mice) and D1-NR1 KO mice ($\mathrm{n}=16$) in 2–8 s task during 14 days training (two-way repeated-measures ANOVA, significant difference between control and KO mice, ${\mathrm{F}}_{\mathrm{1,25}}=10.8$, $\mathrm{p}=0.003$). (B) Correct rate of control ($\mathrm{n}=17$) and D2-NR1 KO mice ($\mathrm{n}=10$) in 2–8 s task during 14 days training (two-way repeated-measures ANOVA, significant difference between control and KO mice, ${\mathrm{F}}_{\mathrm{1,25}}=8.728$, $\mathrm{p}=0.007$). (C) The psychometric curve for control $(\mathrm{n}=11)$ and D1-NR1 KO mice $(\mathrm{n}=16)$ (two-way repeated-measures ANOVA, significant difference between control and KO mice, ${\mathrm{F}}_{\mathrm{1,25}}=12.27$, $\mathrm{p}=0.002$). (D) The psychometric curve for control $(\mathrm{n}=17)$ and D2-NR1 KO mice $(\mathrm{n}=10)$ (two-way repeated-measures ANOVA, significant difference between control and KO mice, ${\mathrm{F}}_{\mathrm{1,25}}=9.64,\mathrm{p}=0.005$). (E) Schematic of muscimol infusion into the dorsal striatum in trained mice. (F) Correct rate for control (black: pre-muscimol, gray: post-muscimol) and mice with muscimol infusion (magenta) in dorsal striatum ($\mathrm{n}=9$ mice, paired t-test, $\mathrm{p}<0.01$). (G) The psychometric curve for control ($\mathrm{n}=9$ mice, black: pre-muscimol, gray: post-muscimol control) and mice with muscimol infusion ($\mathrm{n}=9$ mice, magenta) in dorsal striatum (two-way repeated-measures ANOVA, significant difference between control and muscimol infusion, ${\mathrm{F}}_{\mathrm{2,16}}=11.74$, $\mathrm{p}=0.0007$). (H) Timeline for selective diphtheria toxin (DT) ablation experiments. (I) Schematic of diphtheria toxin receptor (DTR) virus (AAV-FLEX-DTR-GFP) injection in dorsal striatum of D1-Cre mice. (J) Correct rate for control ($\mathrm{n}=9$ mice) and mice with dorsal striatum D1-SPNs ablation (D1-DTR, $\mathrm{n}=8$ mice) (two-sample t-test, $\mathrm{p}=0.0016$). (K) The psychometric curve for control ($\mathrm{n}=9$ mice) and D1-SPNs ablation mice ($\mathrm{n}=8$ mice) (two-way repeated-measures ANOVA, main effect of ablation, ${\mathrm{F}}_{\mathrm{1,15}}=1.84$, $\mathrm{p}=0.195$; interaction between trial intervals and ablation, $\mathrm{F}\mathrm{6,90}=4.14$, $\mathrm{p}=0.001$). (L) Movement trajectory of a control mouse in 8-s trials. (M) Movement trajectory of a D1-DTR mouse in 8-s trials. (N) Schematic of diphtheria toxin receptor (DTR) virus (AAV-FLEX-DTR-GFP) injection in dorsal striatum of A2a-Cre mice. (O) Correct rate for control ($\mathrm{n}=8$ mice) and mice with dorsal striatum D2-SPNs ablation (D2-DTR, $\mathrm{n}=8$ mice) (two-sample t-test, $\mathrm{p}=0.005$). (P) The psychometric curve for control ($\mathrm{n}=9$ mice) and D2-SPNs ablation mice ($\mathrm{n}=8$ mice) (two-way repeated-measures ANOVA, main effect of ablation, ${\mathrm{F}}_{\mathrm{1,15}}=0.477$, $\mathrm{p}=0.5$; interaction between trial intervals and ablation, ${\mathrm{F}}_{\mathrm{6,90}}=12.6$, $\mathrm{p}<0.001$). (Q) Movement trajectory of a control mouse in 8-s trials. (R) Movement trajectory of a D2-DTR mouse in 8-s trials. (S) Schematic of center-surround receptive field diagram for Go/No-Go (left) and Co-activation (right) models. ‘+’ indicates facilitating effect to selection. ‘−’ indicates inhibitory effect to selection.

Figure 5.. Optogenetic manipulation of D1- vs. D2-SPNs differently regulates action selection.

(A) Schematic of optic fiber implantation for experimentally optogenetic excitation or inhibition of D1- or D2-SPNs in the dorsal striatum. (B, C) Schematic for optogenetic excitation (B) and inhibition (C) of D1-/D2-SPNs for 1 s right before lever extension in 2–8 s task. (D) Change of correct rate for optogenetic excitation of D1-SPNs in 2-s and 8-s trials ($\mathrm{n}=11$ mice, one-sample t-test, 2-s trials: $\mathrm{p}=0.248$; 8-s trials: $\mathrm{p}<0.05$). (E) Change of correct rate for optogenetic inhibition of D1-SPNs in 2-s and 8-s trials ($\mathrm{n}=6$ mice, one-sample t-test, 2-s trials: $\mathrm{p}=0.557$; 8-s trials: $\mathrm{p}<0.05$). (F) Change of correct rate for optogenetic excitation of D2-SPNs in 2-s and 8-s trials ($\mathrm{n}=8$ mice, one-sample t-test, 2-s trials: $\mathrm{p}<0.05$; 8-s trials: $\mathrm{p}<0.05$). (G) Change of correct rate for optogenetic inhibition of D2-SPNs in 2-s and 8-s trials ($\mathrm{n}=5$ mice, one-sample t-test, 2-s trials: $\mathrm{p}<0.05$; 8-s trials: $\mathrm{p}<0.05$). (H) Schematic of center-surround receptive field diagram for Go/No-Go (left) and Co-activation (right) models. ‘+’ indicates facilitating effect to selection. ‘−’ indicates inhibitory effect to selection.

Figure 6.. A triple-control computational model of basal ganglia direct and indirect pathways for action selection.

(A) Network structure of the cortico-basal ganglia model based on realistic anatomy and synaptic connectivity. (B) Schematic of center-surround-context receptive field diagram for ‘Triple-control’ model. ‘+’ indicates facilitating effect to selection. ‘−’ indicates inhibitory effect to selection. (C) The psychometric curves of behavioral output in control (black) and D1-SPNs ablation condition (blue) in ‘Triple-control’ model ($\mathrm{n}=10$, two-way repeated-measures ANOVA, main effect of ablation, ${\mathrm{F}}_{\mathrm{1,18}}=98.72$, $\mathrm{p}<0.0001$; interaction between trial intervals and ablation, ${\mathrm{F}}_{\mathrm{6,108}}=7.799$, $\mathrm{p}<0.0001$). (D) The psychometric curves of behavioral output in control (black) and D2-SPNs ablation condition (red) in ‘Triple-control’ model ($\mathrm{n}=10$, two-way repeated-measures ANOVA, main effect of ablation, ${\mathrm{F}}_{\mathrm{1,18}}=99.54$, $\mathrm{p}<0.0001$; interaction between trial intervals and ablation, ${\mathrm{F}}_{\mathrm{6,108}}=177.6$, $\mathrm{p}<0.0001$). (E) Change of correct rate for optogenetic excitation of D1-SPNs in 2-s and 8-s trials ($\mathrm{n}=10$, one-sample t-test, 2-s trials: $\mathrm{p}=0.407$; 8-s trials: $\mathrm{p}<0.05$). (F) Change of correct rate for optogenetic excitation of D2-SPNs in 2-s and 8-s trials ($\mathrm{n}=10$, one-sample t-test, 2-s trials: $\mathrm{p}<0.05$; 8-s trials: $\mathrm{p}<0.05$). (G) Change of correct rate for optogenetic inhibition of D1-SPNs in 2-s and 8-s trials ($\mathrm{n}=10$, one-sample t-test, 2-s trials: $\mathrm{p}=0.28$; 8-s trials: $\mathrm{p}<0.05$). (H) Change of correct rate for optogenetic inhibition of D2-SPNs in 2-s and 8-s trials ($\mathrm{n}=10$, one-sample t-test, 2-s trials: $\mathrm{p}<0.05$; 8-s trials: $\mathrm{p}<0.05$).

Figure 7.. Computational modeling reveals direct and indirect pathways regulating action selection in a distinct manner.

(A) Schematic for manipulation of D1-SPNs in ‘Triple-control’ model. (B) Schematic of manipulation of D1-SPNs in the center-surround-context receptive field diagram for ‘Triple-control’ model. ‘+’ indicates facilitating effect to selection. ‘−’ indicates inhibitory effect to selection. (C) Correct rate change in 2s trials when manipulating D1-SPNs with different manipulation strengths ($\mathrm{n}=10$, one-way repeated-measures ANOVA, effect of manipulation strength, ${\mathrm{F}}_{\mathrm{36,324}}=1.171$, $\mathrm{p}=0.238$). (D) Correct rate change in 8s trials when manipulating D1-SPNs with different manipulation strengths ($\mathrm{n}=10$, one-way repeated-measures ANOVA, effect of manipulation strength, ${\mathrm{F}}_{\mathrm{36,324}}=13.71$, $\mathrm{p}<0.0001$). (E) Schematic for optogenetic manipulation of D2-SPNs in ‘Triple-control’ model. (F) Schematic of manipulation of D2-SPNs in the center-surround-context receptive field diagram for ‘Triple-control’ model. ‘+’ indicates facilitating effect to selection. ‘−’ indicates inhibitory effect to selection. (G) Correct rate change in 2s trials when manipulating D2-SPNs with different manipulation strengths ($\mathrm{n}=10$, one-way repeated-measures ANOVA, effect of manipulation strength, ${\mathrm{F}}_{\mathrm{36,324}}=59.13$, $\mathrm{p}<0.0001$). (H) Correct rate change in 8s trials when manipulating D2-SPNs with different manipulation strengths ($\mathrm{n}=10$, one-way repeated-measures ANOVA, effect of manipulation strength, ${\mathrm{F}}_{\mathrm{36,324}}=40.75$, $\mathrm{p}<0.0001$). (I) Diagram of linear modulation of direct pathway. (J) Diagram of nonlinear modulation of indirect pathway.