A non-canonical striatopallidal "Go" pathway that supports motor control

Abstract

In the classical model of the basal ganglia, direct pathway striatal projection neurons (dSPNs) send projections to the substantia nigra (SNr) and entopeduncular nucleus to regulate motor function. Recent studies have re-established that dSPNs also possess "bridging" collaterals within the globus pallidus (GPe), yet the significance of these collaterals for behavior is unknown. Here we use in vivo optical and chemogenetic tools combined with deep learning approaches to dissect the roles of bridging collaterals in motor function. We find that dSPNs projecting to the SNr send synchronous motor-related information to the GPe via axon collaterals. Inhibition of native activity in dSPN GPe terminals impairs motor activity and function via regulation of pallidostriatal Npas1 neurons. We propose a model by which dSPN GPe collaterals ("striatopallidal Go pathway") act in concert with the canonical terminals in the SNr to support motor control by inhibiting Npas1 signals going back to the striatum.

Keywords: Go pathway; Npas1; arkypallidal; axon collaterals; axonal copy; basal ganglia; bridging collaterals; direct pathway; efference copy; globus pallidus; medium spiny neurons; motor control; striatopallidal; striatum.

Fig. 1.. The density of dSPN terminals in the GPe account for more than half the density of SNr terminals.

A. Strategy for anterograde tracing of dSPN axons/terminals. B. Synaptophysin-targeted GFP is largely absent from axons (blue arrow) and enriched in terminals (white arrow) (representative images from N=5 mice). C. Synaptophysin-targeted dSPN terminals (cyan) colocalize with VGAT (red), appearing white (blue arrows). D. The density of dSPN Synaptophyin-GFP+ terminals in the GPe reaches more than half the density in the EP and SNr (ANOVA: region p<0.001; post-hocs: **p<0.01, ***<0.001) (N=5 mice). Data is mean±SEM. E. Confirmation that dSPN terminals in the GPe arise from axons projecting to the SNr. Left: Injection of retrograde herpes-simplex virus (HSV) expressing a flexed YFP into the GPe and red retrobeads into the SNr of Drd1-cre mice. Right: YFP+ cell bodies colocalized with retrobeads+ cells in the DMS, identifying dSPNs projecting to both GPe and SNr. There were also retrobeads+, YFP− cell bodies, identifying neurons projecting only to the SNr.

Fig. 2.. dSPNs send copies of motor signals to GPe and SNr axons, continuously encoding body speed

A. Strategy for dual calcium imaging of dSPN GPe and SNr terminals (N=8 mice) B. jGCaMP7s (red) colocalizes with the presynaptic marker VGAT (blue) in the GPe and SNr, appearing pink (white arrows). Optic fibers target GPe and SNr jGCaMP7s+ regions C. Mice are video-recorded in the open field and body positions obtained with DeepLabCut D. Left: Representative body trajectory. Middle: Representative trace of mouse speed over time (raw data: grey; smoothed in 2s bins: blue), showing the onset (green) and offset (orange circle) of motor bouts. Right: dSPN GPe and SNr calcium signals (Zscore of the deltaF/F; dFF) closely track mouse speed E. Average data aligned to the onset and offset of individual motor bouts. GPe and SNr dFF show increases at movement onset and decreases at movement offset F. Body speed (paired t-test p<0.001) and GPe/SNr dFF (ANOVA: main effect: p<0.001) significantly increase at movement onset vs. offset G. GPe (r= 0.58) and SNr (r= 0.53) dFF significantly correlate with mouse speed when compared to phase-shuffled data (Mann-Whitney GPe p<0.001, SNr p<0.001) H. GPe and SNr dFF are highly correlated with each other (Pearson r= 0.81) vs. phase-shuffled data (Mann-Whitney p<0.001). Data is mean±SEM.

Fig. 3.. dSPN GPe and SNr axons track the temporal boundaries of individual motor bouts.

A. Mice for GPe/SNr dSPN axonal imaging (jGCaMP7s) are video-recorded in the rotarod set at constant speeds. Body part positions are obtained with DeepLabCut (N= 7 mice). B. Representative trace of GPe/SNr zscored deltaF/F (dFF), with zoom-in view (left inlet). IPI: interpeak interval C. Average GPe/SNr dFF traces at increasing speeds (left) and heatmaps of all individual trials at 5 rpm (3/animal) (right) showing sustained activity across the running epoch. Only the first/last 30 s of the rotarod epoch is shown (cut-off = dashed lines) D. Representative trace of mouse lower body position on the Y axis during running, showing the onset (green) and offset (black circle) of jumping bouts. dSPN GPe and SNr dFF closely track lower body Y trajectory. Dashed lines show lower (0 cm) and upper (3 cm) bounds of the rotarod. E. Left: Probability distributions of jump IPIs. Maximal probability (vertical bar) is reached at smaller IPIs as the rotarod speed increases. Right: Jump IPI averaged per animal significantly decreases as the rotarod speed increases (ANOVA: speed: p<0.01; post-hocs: *p<0.05 or **<0.01) F. Average data aligned to the onset of individual jump bouts during running. GPe and SNr dFF increases at jump onset, showing shorter and smaller transients with increasing rotarod speed G. Duration of jumping bouts (peak-to-peak) and duration of GPe/SNr dFF transients (peak-to-peak) significantly decrease with rotarod speed (all: ANOVA: speed: p<0.01, post-hocs: *p<0.05, **<0.01, ***<0.001). H. Duration of individual jumping bouts significantly correlate with duration of individual GPe/SNr dFF transients (Pearson r, all p<0.001: duration: GPe: r=0.75; SNr: r=0.67) I. Pearson correlation between GPe and SNr activity decreases in a task-dependent manner between rest (‘pre/post’: r=0.90) and running (r=0.70) (ANOVA: epoch: p<0.01, post-hoc: p<0.01). Data is mean±SEM. See also Supplementary Fig. S3–S4.

Fig. 4.. dSPN presynaptic terminal imaging in the GPe and SNr confirms task-dependent correlation in a rotarod motor task.

A. Strategy for terminal-specific calcium imaging of dSPN GPe and SNr terminals (N=7 mice) B. Synaptophysin-jGCaMP8s expression in the DMS with optic fibers targeting the GPe (inlet from other section) and SNr. C. Left: Synaptophysin-jGCaMP8s (Sy8s) is expressed in terminals (white arrows) but poorly in axons in the GPe at 10–14 days post injection. As a comparison, untargeted jGCaMP7s is detected in terminals (white arrows) and axons (blue arrows) in the GPe. Right: Quantification of optical density in boutons vs. axons in the GPe shows 6-fold bouton enrichment in Synaptophysin-jGCaMP8s and 2-fold in jGCaMP7s brains (ANOVA: epoch: p<0.001, post-hocs: p<0.01). Quantification made in unstained fixed brains (native fluorescence) to avoid potential antibody amplification artefacts. D. Mice are tested in the rotarod at accelerating speeds 10–14 days post injection. E. Representative trace of GPe/SNr zscored deltaF/F (dFF) and zoom-in inlet (left), average trace of all mice (middle) and heatmaps of all individual trials (right) showing terminal-specific dSPN GPe and SNr activity in the rotarod task. Only the first and last 30 s of the rotarod epoch is shown (cut-off = dashed lines) F. GPe and SNr terminal activity shows a significant increase in baseline and AUC in the run epoch (‘during’) vs rest (‘pre/post’) (all: ANOVA: epoch: p<0.001; post-hoc: all: p<0.001). G. Pearson correlation between GPe and SNr activity decreases in a task-dependent manner between rest (‘pre/post’: r=0.81) and running (r=0.60) (ANOVA: epoch: p<0.001, post-hoc: p<0.01). Data is mean±SEM.

Fig. 5.. dSPN bridging collaterals in the GPe support motor function as revealed with chemogenetic inhibition.

A. Strategy for chemogenetic inhibition of dSPN terminals in the GPe using GPe infusion of clozapine-N-oxide (CNO) B. Fluid cannulas target dSPN terminals in the GPe expressing hM4D-mCherry C. Representative image of local radioactive [³H]-labelled CNO infusion confirming the drug can stay restricted in the GPe at this volume (300nL) D. Left: Mice are tested in the open field after infusion of Saline (SAL) or CNO. Right: Chemogenetic inhibition of dSPN GPe terminals significantly reduces locomotion speed (ANOVA: virus × drug p<0.001; post-hoc: SAL vs CNO: hM4D p<0.001, mCherry p=0.32) (N= 10 mCherry, 7 hM4D) E. Left: Mice are tested in the rotarod at constant speed after SAL/CNO infusion. Mice are allowed to return to the rotarod if they fall. Right: Chemogenetic inhibition of dSPN GPe terminals significantly increases the number of falls (ANOVA: virus × drug p<0.05; post-hoc: SAL vs CNO: hM4D p<0.05, mCherry p=0.12) (N= 9 mCherry, 9 hM4D). Data is mean±SEM.

Fig. 6.. Confirmation that chemogenetic manipulation of dSPN GPe terminals does not affect activity in the SNr.

A. Strategy for in-vivo recordings of GPe and SNr units following optogenetic stimulation of dSPN cell bodies with ChR2, combined with chemogenetic inhibition of dSPN terminals in the GPe with hM4D using local infusion of Saline (SAL) or clozapine-N-oxide (CNO, 1mM, 300 nL) (N= 5 GPe;SAL, 5 GPe;CNO, 5 SNr;SAL, 5 SNr;CNO) B. Left: Expression of hM4D-mCherry (red) and ChR2-YFP (blue) in dSPN cell bodies (dStr) and GPe/SNr terminals. Optic fibers target the striatum. Electrodes target the GPe and SNr. Middle: Zoom-in showing colocalization (white arrow) of hM4D and ChR2 in the dStr. Right: Zoom-in showing GPe/SNr axons and terminals. C-D. Peristimulus time histograms (PSTHs) showing the mean spike frequency of all recorded GPe neurons before, during, and after the 0, 250, 500 or 1000 ms laser stimulation (1 ms bins) in animals with local GPe infusion of SAL (C) or CNO (D). E-F. Same as C-D for the SNr. G. Proportion of GPe units for which basal firing activity is significantly decreased (inhibited units) or not (non-inhibited units) after laser stimulation at different durations (Fisher’s test at 250, 500, 1000 ms: SAL vs CNO: p<0.001; at 0ms p=0.48). H. Same as G for the SNr (Fisher’s test at all stim durations: SAL vs CNO: p=0.48–0.99). I. Normalized responses in the GPe as a function of baseline activity for all units (ANOVA: stim duration × drug p<0.001; post-hocs SAL vs. CNO at 250, 500, 1000ms: all ***p<0.001; post-hocs 0 ms vs. other stim durations: SAL: #p<0.001, CNO: p=0.8–1.0). J. Same as I for significantly inhibited GPe units (ANOVA: stim duration × drug p=0.17; main effect of drug: SAL vs. CNO ***p<0.001). K. Same as I for the SNr (ANOVA: stim duration × drug p=0.16; drug p= 0.99, main effect of stim duration: ***p<0.001; post-hoc all mice pooled: 0 ms vs. other stim durations: all ***p<0.001). L. Same as J for the SNr (ANOVA: stim duration × drug p=0.40, drug p=0.82, stim duration p=0.93). Note: For J and L, 0 ms stimulation data is not included in the analysis as there were too few significantly inhibited units (0 to 1). Data is mean±SEM.

Fig. 7.. dSPN bridging collaterals in the GPe support motor function as revealed with optogenetic inhibition.

A. Strategy for optogenetic inhibition of dSPN terminals in the GPe (N= 9 eOPN3, 10 YFP) B. Left: eOPN3-mScarlet (red) colocalizes with VGAT (blue) in the GPe, appearing pink (white arrows). Right: Optic fibers target dSPN eOPN3-mScarlet+ terminals in the GPe C. Left: Optogenetic inhibition during rotarod trials at accelerating speed. Middle: Optogenetic inhibition of dSPN GPe terminals reduces latency to fall. Right: Summary data showing significant reductions in latency to fall (ANOVA: Virus × Laser p<0.01; post-hoc: on vs off: eOPN3 p<0.05, GFP p=0.10). In the optogenetic paradigm, mice were only allowed to fall once. D. Left: Optogenetic inhibition triggered in a closed loop during ongoing locomotion (see methods). ITI: intertrial interval. Body part positions are obtained with DeepLabCut. Middle: Optogenetic inhibition (30 sec) of dSPN GPe terminals reduces mouse speed, which recovers after 6 min (consistent with). Right: Summary data showing significant reductions in mouse speed in the 30-sec opto-epoch (ANOVA: Virus × Laser × Epoch p<0.05; post-hoc: on vs off in the ‘post’-epoch: eOPN3 p<0.01, GFP p=0.96) E. Left: Heatmaps showing behavioral classification of videoframes (all mice) into locomotion, fine movements or motionless. Note the mild locomotion increase 5 sec before laser onset (dashed line). Right: Percent frames in each motor classification, showing decreased locomotion and increased fine movements during dSPN GPe inhibition. F. dSPN GPe inhibition significantly (Mixed model: virus × epoch × motor-classification: p<0.001) reduces percent time spent locomoting (post-hocs: eOPN3 p<0.001, GFP p=0.99) and increases percent time spent doing fine movements (post-hocs: eOPN3 p<0.001, GFP p=0.61). Data is mean±SEM. See also Supplementary Fig. S5.

Fig. 8.. dSPN axons inhibit ongoing motor-related calcium dynamics in their GPe Npas1 target neurons.

A. Left: Opto-stimulation of dSPN GPe axons and simultaneous recording of Npas1 calcium activity (N= 5 mCherry, N=6 ChrimsonR). Right: All-optical setup for opto-stimulation (595nm) and 405nm/465nm photometry B. ChrimsonR-TdTomato+ dSPN terminals in the GPe (red) closely apposed to (white arrows) GCaMP6s+ Npas1 cell bodies (cyan). Optic fibers in the same GPe region C. Opto-stimulation is triggered in a closed loop when Npas1 dFF surpasses a defined threshold (see methods) D. 10 s, 20 Hz stimulation of dSPN GPe axons leads to a power-dependent reduction in Npas1 activity. Left: Average traces, Middle: Heatmaps of all trials sorted by degree of inhibition, Right: Amplitude change in Npas1 dFF in the opto-window (ANOVA: virus × power p<0.001; post-hoc: Chrimson **p<0.01, ***<0.001, mCherry p>0.8) E. As expected, there is no effect of unilateral stimulation on mouse speed (ANOVA: power p=0.50) F. 10 s, 2 mW stimulation of dSPN GPe axons leads to a frequency-dependent reduction in Npas1 activity. Left: Average traces, Middle: Heatmaps of all trials sorted by degree inhibition, Right: Amplitude change in Npas1 dFF in the opto-window (ANOVA: virus × power p<0.01; post-hocs: Chrimson: 0 vs 20 Hz p<0.05, 0 vs 10 Hz p=0.99, mCherry p=0.28 and 0.51) H. No effect of stimulation on speed (ANOVA: power p=0.56) I. Opto-stimulation triggered in a closed loop during ongoing locomotion when mouse speed reaches a defined threshold (see methods) J. As expected, mouse speed increases prior to the opto-trigger (baseline vs. ‘threshold’ and stim periods), but is not affected by opto-stimulation (0 vs. 0.2mW) (ANOVA: epoch p<0.001, epoch × LED p=0.79; post-hocs all p<0.001) K. 5 s, 20 Hz stimulation of dSPN GPe axons at ultra-low power (0.2 mW) reduces motor-related Npas1 calcium activity. Left: Average and summary data showing significant increase in dFF prior to the opto-trigger (ANOVA: epoch p<0.05). Middle: Heatmaps of all trials sorted by degree inhibition, Right: Amplitude change in Npas1 dFF in the opto-window (ANOVA: virus × LED p<0.01; post-hocs: Chrimson p<0.01, mCherry p=0.84). Heatmaps: straight line = LED onset, dashed line = off. Data is mean±SEM. See also Supplementary Fig. S6.

Fig. 9.. GPe Npas1 but not ChAT neurons mediate the effects of bridging collaterals on motor function.

A. Optogenetic stimulation of GPe Npas1 neurons B. Optic fibers target Npas1 ChR2-YFP+ neurons in the GPe. C. Left: 10 trials optogenetic stimulation during open field locomotion. Right: 20Hz stimulation of Npas1 neurons significantly reduces mouse speed (ANOVA: virus × epoch p<0.01; post-hocs ChR2 **p<0.01, ***<0.001, GFP p>0.9) (N= 12 ChR2, 13 YFP). D. Left: Optogenetic stimulation (20Hz) during rotarod trials at accelerating speed. Right: Stimulation of Npas1 neurons significantly reduces latency to fall (ANOVA: virus × laser p<0.05; post-hocs: ChR2 p<0.01, GFP p=0.97) (N= 8 ChR2, 9 YFP). E. Optogenetic stimulation of GPe ChAT neurons (N= 8 ChR2, 8 YFP) F. Optic fibers target ChAT ChR2-YFP+ neurons in the caudal GPe. G. Neither 10 Hz (ANOVA: virus × epoch p=0.11) or 20Hz (p=0.06) ChAT stimulation affects mouse speed H. 20Hz stimulation of ChAT neurons does not affect latency to fall in the rotarod (ANOVA: virus × laser p=0.63). Data is mean±SEM.