Subthalamic Activity for Motor Execution and Cancelation in Monkeys

Abstract

The subthalamic nucleus (STN) receives cortical inputs via the hyperdirect and indirect pathways, projects to the output nuclei of the basal ganglia, and plays a critical role in the control of voluntary movements and movement disorders. STN neurons change their activity during execution of movements, while recent studies emphasize STN activity specific to cancelation of movements. To address the relationship between execution and cancelation functions, we examined STN activity in two Japanese monkeys (Macaca fuscata, both sexes) who performed a goal-directed reaching task with a delay that included Go, Cancel, and NoGo trials. We first examined responses to the stimulation of the forelimb regions in the primary motor cortex and/or supplementary motor area. STN neurons with motor cortical inputs were found in the dorsal somatomotor region of the STN. All these STN neurons showed activity changes in Go trials, suggesting their involvement in execution of movements. Part of them exhibited activity changes in Cancel trials and sustained activity during delay periods, suggesting their involvement in cancelation of planed movements and preparation of movements, respectively. The STN neurons rarely showed activity changes in NoGo trials. Go- and Cancel-related activity was selective to the direction of movements, and the selectivity was higher in Cancel trials than in Go trials. Changes in Go- and Cancel-related activity occurred early enough to initiate and cancel movements, respectively. These results suggest that the dorsal somatomotor region of the STN, which receives motor cortical inputs, is involved in preparation and execution of movements and cancelation of planned movements.

Significance Statement

The dorsal somatomotor region of the subthalamic nucleus, which receives motor cortical inputs, is important for motor control: its lesion induces abnormal involuntary movements known as hemiballism; its increased and/or oscillatory activity is observed in Parkinson's disease; and its coagulation or deep brain stimulation ameliorates parkinsonian symptoms. Here, we examined neuronal activity of this region in monkeys while they performed a reaching task with a delay including Go, Cancel, and NoGo trials and demonstrated that its activity is involved in multiple motor functions: preparation and execution of movements and cancelation of planned movements. The present results will benefit our understanding of the neuronal mechanism of STN-related movement disorders and their better therapeutics targeting the STN in the near future.

Introduction

The subthalamic nucleus (STN) receives cortical inputs via the cortico-STN hyperdirect and cortico-striato-external pallido (GPe)-STN indirect pathways (Nambu et al., 2002b; Polyakova et al., 2020) and projects to the output nuclei of the basal ganglia (BG), i.e., the internal segment of the globus pallidus (GPi) and substantia nigra pars reticulata (Jaeger and Kita, 2011). The STN plays a critical role in the control of voluntary movements and movement disorders, such as Parkinson's disease and hemiballism. Early studies reported that STN neurons in the dorsal somatomotor and ventral oculomotor regions changed their activity in relation to forelimb (Georgopoulos et al., 1983; DeLong et al., 1985; Wichmann et al., 1994) and eye movements (Matsumura et al., 1992), respectively, in monkeys.

The classical model of the BG suggests that STN activity implements excitatory influence on the BG output nuclei that inhibit the thalamus and cortex and thereby have suppressive effects on movements (Alexander and Crutcher, 1990; Mink, 1996). In fact, lesions or blockade of the dorsal region of the STN that belongs to the motor loop induces hemiballism (Hamada and DeLong, 1992; Nambu et al., 2000; Hasegawa et al., 2022) and deletion of the hyperdirect pathway from the motor cortex induces motor hyperactivity (Koketsu et al., 2021). Besides, recent single-unit electrophysiological studies of the STN in rodents (Schmidt et al., 2013; Fife et al., 2017), monkeys (Pasquereau and Turner, 2017), and humans (Bastin et al., 2014; Benis et al., 2016; Mosher et al., 2021) emphasized its activation during movement inhibition (Frank, 2006). Functional magnetic resonance imaging (Aron and Poldrack, 2006; Li et al., 2008) and local field potential studies in the human STN (Kühn et al., 2004; Ray et al., 2012; Alegre et al., 2013; Bastin et al., 2014; Zavala et al., 2017; Marmor et al., 2020) also support its activation during movement inhibition. However, such activation was found in the ventromedial region of the STN (Pasquereau and Turner, 2017; Mosher et al., 2021) that receives inputs from the prefrontal and limbic cortices and projects back to these original cortices as members of the association and limbic loops. The ventromedial region of the STN does not project to the somatomotor region in the GPi, and thus, its activation does not seem to be appropriate to cancel/stop movements in the motor loop.

The first objective of this study is to examine whether STN neurons in the dorsal somatomotor region show activity related to cancelation in addition to execution of movements. We thus examined the activity of STN neurons that belonged to the motor loop, which was confirmed by a biphasic excitatory response to electrical stimulation of the primary motor cortex (M1) and/or supplementary motor area (SMA) (Iwamuro et al., 2017; Polyakova et al., 2020). We then imposed a motor task with a delay that includes Go, Cancel, and NoGo trials, which are typically used to study neuronal activity in execution of movements, cancelation of planned movements, and withholding of movements, respectively. In contrast to a conventional Stop signal task (SST) that presents Stop signals immediately after Go signals (Aron and Poldrack, 2006; Schall and Godlove, 2012; Schmidt et al., 2013; Pasquereau and Turner, 2017), we introduced a delay period to discriminate between motor execution/cancelation and motor preparation, the latter of which is represented by the delay-related activity (Thobois et al., 2000; Cassidy et al., 2002; Fischer et al., 2017). The second objective is to compare the timing of neuronal activity in motor execution and cancelation. It has been argued that neuronal activity changes in the BG are rather late and may not contribute to the movement initiation (DeLong and Georgopoulos, 1981). In contrast, Stop-related activity changes occurred early enough to stop movements (Schmidt et al., 2013), which is mediated by the rapid hyperdirect pathway. Here, we have found that the somatomotor region of the STN exhibits activity related to execution and cancelation of planned movements, which is early enough to initiate and cancel movements.

Materials and Methods

Animals

The experimental protocols were approved by the Institutional Animal Care and Use Committee of the National Institutes of Natural Sciences, and all experiments were performed in accordance with the guidelines of the National Institutes of Health Guide for the Care and Use of Laboratory Animals. Two Japanese monkeys (Macaca fuscata; Monkey S, 6.2 kg, female; Monkey L, 7.1 kg, male) were used in this study. When the monkeys were not in active use, they were housed in individual primate cages and given ad libitum access to food and water. The monkeys’ access to water was regulated during the experimental period to increase their motivation to perform the task. Throughout the study, the monkeys were monitored daily, and their body weights were documented weekly. If a body weight <80% of the baseline was observed, the water regulation was stopped. First, each monkey was trained to sit in a primate chair quietly. After the monkeys were conditioned to handling by the experimenter and movement restrictions, training to perform a motor task was started.

Go/Cancel/NoGo goal-directed reaching task with a delay

The monkeys were trained to perform a goal-directed reaching task with a delay using their dominant forelimb (Monkey S right, Monkey L left). The task combined three types of trials: “Go”, “Cancel”, and “NoGo” trials. Cancel and NoGo trials targeted different aspects of movement suppression, that is cancelation of planned movements and withholding of all movements, respectively. A panel was placed vertically at a distance of 30 cm in front of the monkey (Fig. 1). The panel was installed with three slots (Left, Center, Right; height 18 mm, width 6 mm, and depth 11 mm) that were aligned horizontally at 10 cm intervals and located 10 cm below the monkey's eye level. At the bottom of each slot, a two-color (green and red) light-emitting diode (LED) was installed. The task included three types of trials: Go, Cancel, and NoGo. Each trial was initiated after the monkey placed its hand at the resting position located 15 cm below the slots for at least 1,500 ms.

Figure 1.

Goal-directed reaching task with a delay. The panel with three two-color (green and red) light-emitting diode (LEDs) and resting position was set in front of the monkey. Each trial was initiated after the animal placed its hand at the resting position for at least 1,500 ms. After task initiation, there were three types of trials. In Go trials (top row, blue arrow), one of three LEDs was lit with red color for 150 ms as an instruction signal (S1), followed by a random delay period (Delay, 550–1,800 ms). During the instruction signal and delay periods, the monkey was required to keep its hand at the resting position. After a delay period, all three LEDs were lit with green color for 1,200 ms as a triggering signal (S2). During that time, the monkey was required to reach the LED inside the slot that had been instructed previously by S1. If the monkey touched the correct LED within 1,200 ms, it was rewarded with a drop of sweetened water (RW). The timing of hand release from the resting position (HR) and finger insertion into the slot (FI) was detected. In Cancel trials (middle row, red arrow), S1 and Delay were the same as in Go trials, while S2 was different, with all three LEDs lit with red color. If the monkey kept its hand at the resting position during the entire trial period, it was rewarded. In NoGo trials (bottom row, green arrow), all three LEDs were lit simultaneously with red color as S1. After a delay period, all three LEDs were lit with green color (S2). The monkey was required to keep its hand at the resting position during the entire trial period to get the reward. In case of an error trial, another trial with the same task condition was repeated.

In Go trials (Fig. 1, top row), one of three LEDs was lit with red color for 150 ms as an instruction signal (S1). After turning off the instruction signal, a random delay period (Delay, 550, 800, 1,050, 1,300, 1,550, or 1,800 ms with equal probability) was introduced. The monkey was required to keep its hand at the resting position during the S1 and delay periods. After a delay period, all three LEDs were lit with green color for 1,200 ms as a triggering signal (S2). Within 1,200 ms after S2 onset the monkey was required to put its index finger inside the slot that had been instructed by S1. If the monkey reached the correct target within 1,200 ms, it was rewarded with a drop of sweetened water (RW) at 1,200 ms after S2 onset. If the monkey released its hand before S2 onset, reached wrong slots, or reached the target after 1,200 ms, the trial is considered to be an error. The timing of hand release from the resting position (HR) and finger insertion into the slot (FI) was detected by infrared photoelectric sensors, installed in the resting position (PS-T1/T2, PS-46, Keyence) and slots on the panel (PS-T1/T2, PS-52). The reaction time (RT) and movement time (MT) were defined as the time from S2 onset to HR and the time from HR to FI, respectively. In Cancel trials (Fig. 1, middle row), S1 was presented similarly to Go trials for 150 ms. After a random delay (550–1,800 ms), all three LEDs were lit with red color as an S2 signal for 1,200 ms. The monkey was required to cancel the planned movement. If the monkey kept its hand at the resting position during the entire trial period, it was rewarded. In NoGo trials (Fig. 1, bottom row), all three LEDs were lit simultaneously with red color as S1 for 150 ms. After a random delay (550–1,800 ms), all three LEDs were lit with green color as S2 for 1,200 ms. The monkey was required to keep its hand at the resting position during the entire trial period to get the reward. Go, Cancel, and NoGo trials were randomly presented with probabilities of 60%, 30%, and 10%, respectively. The Left, Center, and Right targets were randomly presented with equal probability. In case of an error trial, another trial with the same task condition was repeated. The task was controlled online by LabVIEW Real Time software (National Instruments).

The monkeys were trained to perform the task with short RT (250–280 ms) and a high success rate (>90%) in all types of trials. Monkey S was trained for 11 months and Monkey L for 4 months before the recordings. The success rate reached ≥95% in Go and NoGo trials and ≥99% in Cancel trials. Therefore, in the present study, we analyzed only the correct trials.

Surgery

After the task training, each monkey underwent an aseptic surgical operation under general anesthesia with ketamine hydrochloride (10 mg/kg body weight, i.m.), xylazine hydrochloride (1–2 mg/kg, i.m.), and sodium thiopental (25 mg/kg, i.v.) to fix its head painlessly in a stereotaxic frame attached to a monkey chair (for details, see Nambu et al., 2000, 2002a). Antibiotics (amikacin sulfate) and analgesics (ketoprofen) were administered post-surgically. After full recovery from the first operation, the skull over the M1 and SMA was removed under anesthesia with ketamine hydrochloride (10 mg/kg, i.m.) and xylazine hydrochloride (1–2 mg/kg, i.m.) with painless head restraint. The forelimb regions of the M1 and SMA were identified by electrophysiological methods (Nambu et al., 2000, 2002a). Two pairs of bipolar stimulating electrodes (Teflon-coated stainless-steel wires, 200 µm diameter; 2 mm inter-tip distance) were implanted chronically into the distal (M1d) and proximal (M1p) forelimb regions of the M1, and one pair was implanted into the forelimb region of the SMA. Exposed areas were covered with transparent acrylic resin except for the SMA area (10-mm diameter) to access the STN. The rectangular plastic chamber covering the craniotomy was fixed onto the skull with acrylic resin. Antibiotics, steroids (dexamethasone), and analgesics were administered after the surgical procedures. Recordings of STN neuronal activity were started 5–7 d after implantation of the stimulating electrodes and performed 3 d per week (Monkey S for 29 months, Monkey L for 17 months).

Recording STN neuronal activity

Recordings were conducted while the monkey performed the task. Using a hydraulic microdrive (MO-81-S; Narishige), a glass-coated home-made Elgiloy microelectrode (0.7–1.5 MΩ at 1 kHz) or an injection electrode consisting of an Elgiloy microelectrode and two silica tubes (Polyakova et al., 2020) was inserted vertically into the STN through the dura mater with local application of lidocaine. Data regarding the drug injection through an injection electrode are reported separately. The electrode was inserted slowly (for about 2 h) to the STN to avoid pushing the brain and achieve stable recordings. Neuronal activity was recorded from the microelectrode, amplified (×10,000), filtered (100 Hz–2 kHz), and monitored using an oscilloscope. STN activity was characterized by mid-frequency discharges (20–40 Hz) with a short spike duration. The unitary activity of STN neurons was isolated, converted into digital data with a time-amplitude-window discriminator (EN-611J; Nihon Kohden), and sampled to a computer for online data analysis using LabVIEW software (National Instruments). Peri-stimulus time histograms (PSTHs; bin width, 1 ms; summed for 100 stimulus trials) were constructed to examine responses to electrical stimulation through the electrodes implanted in the M1 and SMA (bipolar stimulation, 300 µs duration, single pulse, strength of 0.5 mA, and interval of 1.4 s). When STN neurons showed a typical biphasic response consisting of early and late excitation to M1 and/or SMA stimulation (Nambu et al., 2000; Iwamuro et al., 2017; Polyakova et al., 2020), neuronal activity during task performance was recorded (bin width, 1 ms). The converted digital data were stored on a computer as well as videotapes using a PCM recorder (Cygnus Technology).

EMG recording

Electromyograms (EMGs) were recorded several times for each monkey using a biophysical amplifier (MEG-6116; Nihon Kohden) and surface electrodes from the following muscles: wrist extensor and flexor muscles, biceps brachii, triceps brachii, trapezius, and deltoid muscles. In addition, for Monkey L, the activity of lower trunk and hip muscles was also monitored. EMG signals were amplified (×2,000), filtered (100 Hz–1 kHz), rectified, and stored on a computer (sampling rate, 2 kHz).

Data analysis

Neuronal responses to cortical stimulation and task-related activity were analyzed using MATLAB R2019a. Responses of STN neurons evoked by stimulation of the M1 and SMA were evaluated based on PSTHs (for details, see Polyakova et al., 2020). The mean and standard deviation (SD) of baseline discharge rates during the 100-ms period preceding the onset of stimulation were calculated for each PSTH. Excitatory responses to stimulation were judged to be significant if firing rates during at least two consecutive bins (2 ms) reached the statistical level of the mean + 3.09 SD (corresponding to p = 0.001, one-tailed t test). If STN neurons responded to M1d and/or M1p stimulation, they were considered to be M1-recipient. If they responded to SMA stimulation, they were considered to be SMA-recipient. If they responded to both M1 and SMA stimulation, they were considered to be M1 + SMA-recipient.

Analysis of neuronal activity during the goal-directed reaching task with a delay was performed to reveal specific features of STN neuronal activity relative to different types of trials, task events, and target locations. Neuronal activity was aligned separately according to the S1 onset, S2 onset, HR, FI, and RW timings for three types of trials and target locations, i.e., Go (Left, Center, Right), Cancel (Left, Center, Right), and NoGo trials. Since Monkey S and Monkey L used different hands, we termed ipsilateral, central, and contralateral targets to the hand used instead of Left, Center, and Right targets. We analyzed only successful trials. In case of Cancel and NoGo trials, we excluded repeated trials presented after an error trial from the analysis to avoid contamination of the data with predicted behavior. We constructed raster plots and spike-density functions (SDFs) by averaging activity and smoothing with a Gaussian filter (σ = 10 ms).

The mean and SD during the 1,000 ms preceding S1 were calculated in each type of trials and target location and considered the baseline discharge rate. If a neuron demonstrated Delay-related changes (S1–S2 period) that exceeded a significant level (mean + 1.65 SD, corresponding to p = 0.05, one-tailed t test) for at least 50 ms within 300-ms interval before S2, the mean and SD during the 500-ms period before S2 were used as the baseline discharge rate. Activity changes were determined to be significant if the SDFs reached the significant level (mean ± 3.09 SD, corresponding to p = 0.001, one-tailed t test) or was zero during at least 3 ms (three consecutive bins). The starting point for the activity changes was defined as the time when the SDFs exceeded the mean ± 1.65 SD (p = 0.05, one-tailed t test) or were equal to zero during at least 3 ms. The end point was defined as the time when the amplitude crossed the mean ± 1.65 SD (p = 0.05, one-tailed t test) or once SDFs reached a nonzero value during at least 3 ms. The activity changes were considered to be related to HR and/or FI in Go trials and Cancel trials if the significant changes occurred in the following task event time windows: 400-ms interval centered at the task event, such as HR and FI (±200 ms from the task event) in Go trials, and 500-ms interval after S2 in Cancel trials. The amplitudes of significant changes related to the task event were calculated by the difference between the SDFs during significant changes (from the starting point to the end point) and the baseline activity (area over/below the baseline activity) within each task event window. Positive and negative values represent facilitatory and inhibitory responses, respectively.

The precise latency of neuronal activity changes after S2 was measured in the following way. First, SDFs with the largest activity changes among the three targets (contralateral, central, and ipsilateral) were selected. Second, the baseline activity (mean and SD) was calculated for the 200-ms period before S2. Third, the latency was defined as the time from the S2 onset to the first time when SDFs continuously crossed the mean ± 1.65 SD (p = 0.05, one-tailed t test) and exceeded the mean ± 3.09 SD (p = 0.001, one-tailed t test) for at least 10 ms (minimal duration).

For task-related heatmap plots, SDFs (σ = 10 ms) aligned at S2 were used. SDFs with the largest activity changes after S2 among three targets were selected. Task-related heatmap plots (bin width, 20 ms) were calculated from the SDF and normalized as z-scores based on mean activity within 1,000 ms before and after S2 onset (total 2,000 ms).

Activity during HR or FI (400 ms centered at HR or FI) in Go trials and after S2 (500 ms after S2 onset) in Cancel trials was modulated by target locations. The amplitudes (A) in each task event window were calculated as the areas between the SDFs (σ = 10 ms) and the baseline activity in three targets as described above. Directional selectivity (DS) of a neuron in each task event was defined as follows:

$$\mathrm{DS}=1\text{–}|(A_{\mathrm{med}}+A_{\min })/(A_{\max }\times 2)|$$

where A_max, A_med, and A_min were the values of maximum, medium, and minimum amplitudes in absolute values among three targets, respectively (i.e., │A_max│≥│A_med│≥│A_min│) (Takara et al., 2011). DS varies between 0 and 1. DS = 0 means the same amplitude among three targets.

Amplitudes of Go- and Cancel-related activity were compared. │A_max│ in each neuron was averaged, divided by the time (400 ms in Go-related activity and 500 ms in Cancel-related activity), and considered as averaged amplitude of Go- or Cancel-related activity.

EMG activity was analyzed using similar methods as applied for neuronal activity in task performance. EMG activity was aligned with task events such as S1, S2, HR, and FI, and then averaged. The mean and SD of the baseline activity were calculated during the 1,000 ms before S1. EMG activity changes were considered significant if EMG activity exceeded the mean + 3.09 SD (p = 0.001, one-tailed t test) for at least 3 ms. The latency was defined as the first time when EMG activity exceeded the mean + 1.65 SD (p = 0.05, one-tailed t test ) and lasted for at least 10 ms.

Experimental design and statistical analyses

The latencies, DSs, and amplitudes were tested using the MATLAB Lilliefors test to check whether they were normally distributed. The Mann–Whitney U test was used to establish the statistical significance of latencies, DSs, and amplitudes in Go and Cancel trials. Values of p < 0.05 or p < 0.001 were considered significant.

Histology

At the end of the experiments, the recording sites were marked by current injections (cathodal DC of 20 µA for 30 s). The monkeys were deeply anesthetized with sodium thiopental (50 mg/kg, i.v.) after induction by ketamine hydrochloride (10 mg/kg, i.m.) and xylazine hydrochloride (1–2 mg/kg, i.m.) and perfused transcardially with 0.1 M phosphate buffer (PB, pH 7.3), followed by 10% formalin in 0.1 M PB, and the same buffer containing 10% sucrose and then 30% sucrose. The brains were removed and kept in 0.1 M PB containing 30% sucrose at 4°C, and then cut serially into 50-µm-thick frontal sections on a freezing microtome. These sections were then mounted onto gelatin-coated glass slides and stained with 1% Neutral Red. The recording sites were reconstructed according to the lesions made by current injections and the traces of the electrode tracks.

Results

EMG activity during task performance

We recorded EMG activity during task performance (Fig. 2) and analyzed the significant EMG activity changes and their latencies. We paid special attention to the activation of any muscles, such as the antagonist muscles, in Cancel and NoGo trials.

Figure 2.

Averaged EMG activity during task performance aligned with S2 in Monkeys S (A) and L (B). Shaded areas represent the timing of movements (from HR to FI). EMG was recorded from the wrist extensor (red), wrist flexor (blue), biceps brachii (purple), triceps brachii (light green), trapezius (brown), and deltoid (dark green) muscles in Monkeys S and L and additionally from the lower trunk (orange) and hip (light blue) muscles in Monkey L.

Figure 2A shows an example of EMG activity of Monkey S during task performance. EMG was aligned with S2 for all types of trials. In Go trials, significant muscle activity changes were observed in the wrist extensor, wrist flexor, biceps brachii, triceps brachii, trapezius, and deltoid muscles in all target locations. The RT and MT of Monkey S were 333.2 ± 51.5 and 456.7 ± 188.4 ms, respectively. The latencies of muscle activity changes after S2 onset were as follows: wrist extensor, 490 ms; wrist flexor, 433 ms; biceps brachii, 332 ms; triceps brachii, 487 ms; trapezius, 280 ms; and deltoid, 295 ms. No significant changes in muscle activity were observed during delay periods (between S1 and S2). All muscles, except the triceps brachii and deltoid muscles, showed different activity changes among three targets and may have determined the direction of reaching. In Cancel and NoGo trials, no significant changes in EMG activity were detected.

Monkey L showed similar muscle activity changes (Fig. 2B). The RT and MT of Monkey L were 302.1 ± 24.8 and 362.0 ± 41.2 ms, respectively. The latencies of muscle activity changes after S2 onset in Go trials were as follows: wrist extensor, 258 ms; wrist flexor, 309 ms; biceps brachii, 202 ms; triceps brachii, 194 ms; trapezius, 215 ms; deltoid, 215 ms; lower trunk, 424 ms; and hip, 339 ms. Significant but small changes were observed in the contralateral targets of Cancel trials in Monkey L: wrist extensor, 286 ms; wrist flexor, 247 ms; biceps brachii, 246 ms; and triceps brachii, 250 ms after S2 onset.

The results demonstrated large significant EMG activity changes in Go trials and small or no activity changes in Cancel and NoGo trials, suggesting that neuronal activity in Cancel trials is not related to muscle activity.

Overview of recorded STN neurons

A total of 308 STN neurons (139 neurons in Monkey S, 169 neurons in Monkey L) that responded to cortical stimulation were recorded during task performance (Table 1). In Monkey S, the stimulating electrode in the SMA became ineffective during experimental sessions, and only the stimulating electrodes in the M1 were used. Based on cortically evoked responses, these neurons were classified into 139 M1-recipient (Monkey S), 50 M1-recipient, 116 M1 + SMA-recipient, and 3 SMA-recipient (Monkey L) STN neurons. All STN neurons demonstrated task-related activity changes in the reaching task with a delay.

Table 1.

Number of recorded STN neurons

Cortical inputs	Monkey S	Monkey L			Subtotal	Total
	M1	M1 only	M1 + SMA	SMA only
	139	50	116	3		308
Go-Type I	59	13	48	1	121	308
Go-Type II	24	11	26	0	61
Go-Type III	44	20	31	1	96
Go-Type IV	12	6	11	1	30
Cancel-Type I	59	17	50	1	127	195
Cancel-Type II	32	9	26	1	68
Delay-related (+)	50	8	16	0	-	74

Number of STN neurons classified based on cortical inputs, response patterns during Go trials (Go-Type I-IV) and Cancel trials (Cancel-Type I, II), and Delay-related activity. Go-Type I, facilitation over the HR and FI periods; Go-Type II, facilitation during the HR period and following inhibition during the FI period; Go-Type III, weak inhibition during the HR period and following facilitation during the FI period; Go-Type IV, inhibition over the HR and FI periods; Cancel-Type I, short facilitation after S2 onset; Cancel-Type II, short inhibition after S2 onset.

Go-related activity in Go trials

All 308 STN neurons showed activity changes in relation to movements in Go trials. Although these neurons showed complex activity changes in relation to task events, i.e., facilitation/inhibition in relation to HR and/or FI, they could be classified into the following four types based on facilitation/inhibition with the largest changes among three targets in Go trials (Fig. 3A, Table 1): Go-Type I, facilitation over the HR and FI periods (121 neurons, 39%); Go-Type II, facilitation during the HR period and following inhibition during the FI period (61 neurons, 20%); Go-Type III, weak inhibition during the HR period and following facilitation during the FI period (96 neurons, 31%); and Go-Type IV, inhibition over the HR and FI periods (30 neurons, 10%).

Figure 3.

Heatmap showing z-scored SDFs (bin width, 20 ms) of the preferred target location aligned with S2 in each neuron. Significant changes were observed after S2 in all 308 STN neurons examined in Go trials (A), and in 195 neurons in Cancel trials (B), and none were observed in NoGo trials (C). The activity of each neuron was sorted according to firing patters (facilitation and inhibition) and latency from S2 onset (black vertical lines). The mean ± SD of HR and FI in Go trials are indicated by vertical dashed lines and horizontal solid lines under the heatmap. A, Activity in Go trials was classified into Go-Type I, facilitation over the HR and FI periods; Go-Type II, facilitation during the HR period and following inhibition during the FI period; Go-Type III, weak inhibition during the HR period and following facilitation during the FI period; and Go-Type IV, inhibition over the HR and FI periods, as indicated on the right side of the heatmaps (A). Activity in Cancel trials was classified into Cancel-Type I, short facilitation after S2 onset and Cancel-Type II, short inhibition after S2 onset (B).

Figure 4 (top) shows a typical example of Go-Type I with facilitation. Stimulation of the M1 induced a biphasic response composed of early and late excitation (ee and le in Fig. 4, bottom right), confirming cortical inputs from the M1 through the hyperdirect and indirect pathways. Raster plots and SDFs of the neuron showed the following movement-related activity changes (Fig. 4, top, Go): Firing rates were increased before HR, decreased after HR, and increased again before and after FI. This biphasic facilitation was similar among three targets, but their amplitudes varied depending on the target locations. We paid attention to activity related to HR (400 ms centered at HR) because it is related to the initiation and direction of reaching movements. HR-related activity was the largest in the contralateral target trials (preferred direction is contralateral) and showed directional selectivity (DS, 0.74). Activity during FI (400 ms centered at FI) may be related to shaping and touching the targets. FI-related activity was the largest in the ipsilateral target trials (preferred direction is ipsilateral) and showed DS (0.29).

Figure 4.

A typical example of Go-Type I and Cancel-Type I STN neurons with M1-inputs recorded from Monkey S. Raster plots demonstrate neuronal firings (blue vertical lines) during the performance of a goal-directed reaching task with a delay. Neuronal activity was aligned separately with S1 onset, S2 onset, and RW (time windows separated by thin vertical lines) in Go, Cancel, and NoGo trials, according to target locations (contralateral, central, and ipsilateral). The time of task events of S1, S2, HR, FI, and RW, is indicated by vertical thick lines. Each plot of Go trials was sorted according to the RT (between S2-HR). Continuous green traces indicate spike density functions (SDFs, σ = 10 ms) for associated raster plots, with significant facilitatory and inhibitory changes indicated by purple (mean +1.65SD, see Data analysis) and orange (mean −1.65SD) colors, respectively. In Go trials, this neuron increased its activity before HR (largest amplitude in the contralateral target trials) and around FI (largest amplitude in the ipsilateral target trials). In Cancel trials, it increased its activity in the contralateral target trials after S2 and decreased its activity in the ipsilateral target trials. Directional selectivity (DS) of Go-related activity during HR, 0.74; DS of Go-related activity during FI, 0.29; and DS of Cancel-related activity after S2 onset, 0.97. PSTHs in the bottom-right corner show the response to M1p cortical stimulation (ee, early excitation; le, late excitation). Mean, mean ± 1.65 SD (corresponding to p = 0.05, one-tailed t test), and mean + 3.09 SD (p = 0.001) are indicated by solid, dashed, and dotted gray lines, respectively.

Figure 5 (top) shows a typical example of Go-Type IV with inhibition. Stimulation of the M1 induced a biphasic response composed of early and late excitation (Fig. 5, bottom right). Firing rates were decreased before HR, recovered after HR, and decreased before FI (Fig. 5, top, Go). This biphasic inhibition was similar among three targets, but their amplitudes varied depending on the target locations. HR-related activity showed the largest changes on the central target trials (preferred direction is central) and DS (0.53).

Figure 5.

A typical example of Go-Type IV and Cancel-Type II STN neurons with M1-input recorded from Monkey S. DS of Go-related activity during HR, 0.53; DS of Go-related activity during FI, 0.77; and DS of Cancel-related activity after S2 onset, 0.54.

All STN neurons examined showed Go-related activity in all three targets with different activity among targets. During HR, more than half of neurons (62%, 191/308) showed mixed response (facilitation in one direction and inhibition in another direction), and among them absolute amplitudes of facilitation was larger than those of inhibition in 145 neurons. Other neurons showed facilitation (110 neurons) or inhibition (7 neurons) consistently among different targets in Go trials. The preferred direction of HR-related activity was evenly distributed among three targets: contralateral, 117 neurons (38%); central, 83 neurons (27%); and ipsilateral, 108 neurons (35%) (Go-Type I: 50, 32, and 39 neurons; Type II: 20, 20, and 21 neurons; Type III: 36, 27, and 33 neurons; and Type IV: 11, 4, and 15 neurons, respectively). DS of HR-related activity of all STN neurons was 0.67 ± 0.28 and similar among four types (Go-Type I: 0.68 ± 0.28; Type II: 0.65 ± 0.3; Type III: 0.68 ± 0.28; and Type IV: 0.68 ± 0.27) and between two monkeys (0.65 ± 0.25 in Monkey S, 0.69 ± 0.29 in Monkey L). FI-related activity also showed similar DS between two monkeys (total 0.63 ± 0.31; 0.60 ± 0.31 in Monkey S, 0.65 ± 0.31 in Monkey L).

Cancel-related activity in Cancel trials

In Cancel trials, there were small or no activity changes in EMG (Fig. 2), but more than half (195/308 neurons, 63%) of recorded STN neurons showed significant activity changes after S2 (Table 1). These neurons with Cancel-related activity could be classified into the following two types (Fig. 3B, Table 1): Cancel-Type I, short facilitation after S2 onset (127 neurons, 41%) and Cancel-Type II, short inhibition after S2 onset (68 neurons, 22%).

Figure 4 (middle, Cancel) shows an example of Cancel-Type I with facilitation. Firing rates were increased after S2 in Cancel trials in the contralateral target, although the amplitude of Cancel-related activity was smaller than that of Go-related activity. Cancel-related activity was the largest in the contralateral target trials (preferred direction is contralateral, the same direction as in Go trials), no activity changes were observed in the central target, and inhibition was observed in the ipsilateral target, suggesting DS of Cancel-related activity (0.97). Figure 5 (middle, Cancel) shows an example of Cancel-Type II with inhibition. Firing rates were decreased after S2 in Cancel trials in all three targets. Cancel-related activity was the largest in the ipsilateral target trials (preferred direction is ipsilateral, the same direction as in Go trials) and showed DS (0.54). Figure 6 (middle, Cancel) shows another example of Cancel-Type II with inhibition. This neuron showed different polarities between Go and Cancel trials, with the largest facilitation occurring during HR in the ipsilateral target in Go trials (top), while the largest inhibition after S2 in the contralateral target in Cancel trials (middle), which is in contrast to the neurons in Figures 4 and 5 showing the same polarity (i.e., facilitation or inhibition) between Go- and Cancel-related activity.

Figure 6.

A typical example of Go-Type I and Cancel-Type II STN neurons with M1-input recorded from Monkey S. DS of Go-related activity during HR, 0.47; DS of Go-related activity during FI, 0.42; and DS of Cancel-related activity after S2 onset, 0.86.

Among 195 STN neurons that showed Cancel-related activity, 65 neurons (33%) showed Cancel-related activity in all three targets, 61 (31%) in two targets, and 69 (36%) in one target. Majority of these neurons (76%, 149/195) showed facilitation (117 neurons) or inhibition (32 neurons) consistently among different targets in Cancel trials. The preferred direction of Cancel-related activity was mostly contralateral or ipsilateral: contralateral, 80 neurons (41%); central, 43 neurons (22%); ipsilateral, 72 neurons (37%) (chi-square test, p < 0.05) (Cancel-Type I: 53, 27, and 47 neurons; Type II: 27, 16, and 25 neurons, respectively). DS of Cancel-related activity of 195 STN neurons was 0.76 ± 0.24, and similar between Types I and II (Cancel-Type I: 0.77 ± 0.22; Type II: 0.73 ± 0.28).

Delay-related activity

We examined neuronal activity during delay periods (between S1 and S2) in Go and Cancel trials, and found that 74 neurons (24%; Monkey S, 50/139 M1-recipient; Monkey L, 8/50 M1-recipient and 16/116 M1 + SMA-recipient) increased their discharge rate (Table 1). Figure 7 shows a typical example with Delay-related activity. This neuron responded to cortical stimulation of the SMA, M1p, and M1d (Fig. 7, bottom right). It showed Delay-related activity in all three targets in both Go (top) and Cancel (middle) trials. Among 74 STN neurons that showed Delay-related activity, 54 neurons showed Delay-related activity in all three targets, nine in two targets, and 11 in one target. Among the 74 neurons, 61 showed Cancel-related activity after S2 in Cancel trials (27, 20, and 14 neurons in three, two, and one target, respectively), as exemplified in Figure 7 (middle).

Figure 7.

A typical example of Go-Type I and Cancel-Type I STN neurons with Delay-related activity recorded from Monkey L. Delay-related activity was observed in Go and Cancel trials, but not in NoGo trials. This neuron responded to SMA, M1p, and M1d stimulation.

NoGo-related activity

Among 308 STN neurons examined, only two neurons showed short activity changes after S1 in NoGo trials, in addition to Go and Cancel trials, and six neurons showed Delay-related activity in NoGo trials, in addition to Go and Cancel trials. Other neurons showed no activity changes after S1, S2, or during delay periods in NoGo trials (Figs. 3C, 4, bottom left, NoGo; Fig. 5, bottom left, NoGo; Fig. 6, bottom left, NoGo). These results suggest that STN neurons may not be involved in withholding of movements. Alternatively, the animals may have learned not to do anything except for waiting in NoGo trials.

Relation between Go- and Cancel-related activity

We compared neuronal activity between Go and Cancel trials in 195 STN neurons that showed both Go- and Cancel-related activity. First, we compared amplitudes in Go and Cancel trials. The amplitude of Go-related activity during HR in Go trials was 43,185.6 ± 62,236.2 spikes (facilitation 45,524.9 ± 64,830.4; inhibition, 31,930.9 ± 46,709.5), and significantly larger (W = 99,623, p = 1.17 × 10⁻⁴³, Mann–Whitney U test) than that of Cancel-related activity in Cancel trials (6,093.4 ± 25,315.7 spikes; facilitation, 4,901.3 ± 22,220.7; inhibition, 10,514.5 ± 34,293.5). We then compared activity patterns (Go-Type I-IV and Cancel-Type I, II) between Go and Cancel trials (Fig. 8). All Go-Type I–IV neurons showed Cancel-Type I and II activity in part, with uneven distribution. Go-Type I neurons showed a significantly higher percentage of Cancel-related activity than other Go-types neurons (97/121, 80%, chi-square test p < 0.05). Go-Type III neurons showed a significantly less percentage of Cancel-Type II activity compared to other Go-types neurons (9/96, 9%, chi-square test p < 0.05). Only 40% (12/30) of Go-Type IV neurons showed Cancel-related activity.

Figure 8.

Relation between Go-Type I–IV and Cancel-Type I, II STN neurons. Go-Type I–IV STN neurons showed Cancel-related activity (Type I or II) or no activity changes in Cancel trials. Numbers of these STN neurons showing Cancel-Type I or II activity or no Cancel-related activity are indicated.

Next, we compared preferred direction between Go-related activity during HR and FI and Cancel-related activity (Fig. 9). Majority of neurons (131/195, 67%) showed facilitation in both Go- and Cancel-related activity (Fig. 9A). Among them, around one third of neurons (37/131, 28%; the sum of diagonal elements in Fig. 9A) showed the same preferred direction between Go and Cancel trials, while the rest (94/131, 72%; the sum of off-diagonal elements in Fig. 9A) showed different preferred directions. Other neurons showed facilitation in Go-related activity and inhibition in Cancel-related activity (26/195, 13%; Fig. 9B), inhibition in Go-related activity and facilitation in Cancel-related activity (17/195%, 9%; Fig. 9C), or inhibition in both Go- and Cancel-related activity (21/195%, 11%; Fig. 9D). In each group, neurons are randomly distributed, and there is no clear relation between the preferred direction of Go-related activity and that of Cancel-related activity, suggesting that Go-related activity and Cancel-related activity are produced by different neuronal processes.

Figure 9.

Comparison of preferred direction between Go- and Cancel-related activity. Neurons that showed both Go- and Cancel-related activity (195 in total) are grouped by their polarity (facilitation or inhibition) of activity, and classified according to the preferred direction (contralateral, central, or ipsilateral) of their Go-related activity during HR and FI and Cancel-related activity. Their number and percentage are indicated, and color coded. A, Neurons showing facilitation in both Go- and Cancel-related activity. B, Neurons showing facilitation in Go- and inhibition in Cancel-related activity. C, Neurons showing inhibition in Go- and facilitation in Cancel-related activity. D, Neurons showing inhibition in both Go- and Cancel-related activity.

We then compared DS of Go-related activity during HR and Cancel-related activity in cumulative histograms (Fig. 10A1,2). DS of Cancel-related activity (0.76 ± 0.24) was significantly higher than that of Go-related activity (0.68 ± 0.27) (p = 4.39 × 10⁻⁵, z = −4.09, ranksum = 28,636) in both monkeys (Cancel, 0.76 ± 0.25 and Go, 0.65 ± 0.26 in Monkey S; Cancel, 0.76 ± 0.24 and Go, 0.69 ± 0.30 in Monkey L). This tendency was observed between Cancel-Type I, II neurons and Go-Type I–IV neurons (Cancel-Type I, 0.77 ± 0.22 and Cancel-Type II, 0.73 ± 0.28; and Go-Type I, 0.68 ± 0.28, Go-Type II, 0.65 ± 0.30, Go-Type III, 0.68 ± 0.28, and Go-Type IV, 0.68 ± 0.27). The effects of cortical inputs were compared in Monkey L (Fig. 10A3). DS of Cancel-related activity (0.80 ± 0.30) was significantly larger than that of Go-related activity (0.64 ± 0.30) in M1-recipient neurons (p = 0.0066, z = −2.72, ranksum = 1,642), whereas DS of Cancel-related activity was comparable to that of Go-related activity in M1 + SMA-recipient neurons (Cancel, 0.77 ± 0.24; Go, 0.71 ± 0.29). On the other hand, DS of Go-related activity and of Cancel-related activity was similar between different cortical inputs (M1-recipient and M1 + SMA-recipient). We also compared DS between Go- and Cancel-related activity in each neuron and found no relation (Fig. 10A4). In sum, Cancel-related activity showed higher DSs than Go-related activity, suggesting that Cancel-related activity may contribute to selective cancelation of Go-planned movements, but not to global cancelation of all movements.

Figure 10.

Comparison of DS (A) and latency (B) between Go- and Cancel-related activity. A1, Cumulative histograms showing DS of Go-related activity during HR (blue) and of Cancel-related activity (dark red) in Monkey S. A2, DS in Monkey L. A3, DS of STN neurons with different cortical inputs (i.e., M1 + SMA, M1) in Monkey L. A4, Scatter plot showing the relation between the DS of Go-related activity during HR (abscissa) and Cancel-related activity (ordinate) in each neuron. B1, Cumulative histograms showing latency of Go-related activity during HR (blue) and of Cancel-related activity (dark red) in Monkey S. Dashed green line and shaded area represent mean RT (333.2 ms) and SD (51.5 ms), respectively. Vertical violet line represents the earliest EMG changes responsible for forelimb movements (deltoid, 295 ms). B2, Latency in Monkey L. Mean RT and SD, 302.1 ± 24.8 ms. The earliest EMG changes, triceps brachii, 194 ms. B3, Latency of STN neurons with different cortical inputs in Monkey L. B4, Scatter plot showing the relation between the latency of Go-related activity during HR (abscissa) and Cancel-related activity (ordinate) in each neuron.

Latency of Go- and Cancel-related activity

We examined the latency of STN neuronal activity in reference to HR. In Go trials, most HR-related activity (259/308, 84%) preceded movement onset defined as HR (Go-Type I, 114/121, 94%; Type II, 50/61, 82%; Type III, 67/96, 70%; Type IV, 28/30, 93%). The mean preceding time before HR was 138.5 ± 68.7 ms (Monkey S) and 155.6 ± 59.3 ms (Monkey L).

We then compared the latencies of Go-related activity during HR and Cancel-related activity in reference to S2 with the earliest EMG changes and mean RT (Fig. 10B1,2). In Go trials, among 308 neurons, activity of 270 neurons (88%) (Monkey S, 126/139, 91%; Monkey L, 144/169, 85%) preceded the mean RT, and that of 231 neurons (75%) (Monkey S, 120/139, 86%; Monkey L, 111/169, 66%) preceded the earliest EMG changes. The latency in Go-related STN activity (183.5 ± 119.0 ms) and that in Cancel-related activity (177.3 ± 110.4 ms) showed a similar distribution in both monkeys. We also compared latencies among response patterns, and found a similar distribution (Go-Type I, 150.0 ± 76.3 ms; Go-Type II, 183.9 ± 136.0 ms; Go-Type III, 229.0 ± 124.8 ms; and Go-Type IV, 175.1 ± 157.9 ms; Cancel-Type I, 167.8 ± 100.6 ms and Cancel-Type II, 191.2 ± 122.0 ms). In Monkey L, we compared latencies between M1- and M1 + SMA-recipient neurons (Fig. 10B3). In Cancel trials, the latency of M1-recipient neurons (249.3 ± 140.8 ms) was significantly longer than that of M1 + SMA-recipient neurons (171.2 ± 106.6 ms) (W = 1,494, p = 1.16 × 10⁻², Mann–Whitney U test), whereas in Go trials, they were comparable (M1-recipient, 190.7 ± 115.1 ms; M1 + SMA-recipient, 194.6 ± 120.1 ms). We then compared the latency in each neuron between Go- and Cancel-related activity (Fig. 10B4), and found a positive correlation: Neurons with earlier changes in Go trials tended to show earlier changes in Cancel trials. These results suggest that neuronal discharges in the STN precede EMG activity and limb movements in Go trials, and seem to have sufficient time to initiate or cancel movements in Go and Cancel trials, respectively.

Location of recording sites

The locations of recorded STN neurons during task performance are shown in the frontal sections of two monkeys (Fig. 11). Most neurons are located in the dorsal region of the STN, corresponding to the somatomotor region that receives inputs from the M1 and SMA (Nambu et al., 1996; Haynes and Haber, 2013; Iwamuro et al., 2017). Special attention was paid to the location of neurons exhibiting Cancel- and Delay-related activity. Neurons with Cancel-related activity were located in the central part of this region (Fig. 11, red triangles or circles). Neurons with Delay-related activity were sparsely found in this region (Fig. 11, red or blue triangles). We calculated the center of mass and its SD of neurons associated with/without Cancel-related activity and with/without Delay-related activity (Fig. 11, corresponding closed symbols and light-colored ellipses), and found they were closely located each other (within 0.5 mm) in both monkeys. We also examined the distribution of response patterns (Go-Type I–IV and Cancel-Type I, II), and found no preference (data not shown). In Monkey L, we plotted locations of STN neurons based on cortical inputs (M1 + SMA-recipient vs M1-recipient) and observed the dominance of M1-recipient neurons in the lateral STN although there was a substantial overlap (data not shown), which was in line with the distribution of cortical inputs we reported previously (Nambu et al., 1996; Iwamuro et al., 2017).

Figure 11.

Locations of recorded neurons in the STN. Recording locations are shown in frontal sections of left STN in Monkey S (flipped horizontally) and right STN in Monkey L. Neurons with/without Cancel-related activity are marked in red/blue, respectively. Neurons with/without Delay-related activity are represented by triangle/circle, respectively. Neurons that did not respond to cortical stimulation are indicated by dots. The center of mass and its SD of each group are indicated by a corresponding closed symbol and light-colored ellipse.

Discussion

The findings of the present study were as follows. (1) M1- and/or SMA-recipient neurons in the dorsal somatomotor region of the STN all showed Go-related activity, suggesting their involvement in execution of movements. (2) Some of these neurons exhibited Cancel- (63%) and Delay-related activity (24%), suggesting their involvement in cancelation of planed movements and preparation of movements, respectively. These STN neurons rarely showed NoGo-related activity, suggesting less involvement in withholding of movements. (3) Go- and Cancel-related activity was selective to the target locations, and DS of Cancel-related activity was higher than that of Go-related activity. (4) Changes in Go- and Cancel-related activity were early enough to initiate and cancel movements, respectively. To our knowledge, this is the first report describing Cancel/Stop-related activity in the dorsal somatomotor region of the STN with motor cortical inputs. These results support the idea that the somatomotor STN plays an important role in preparation and execution of movements and cancelation of planned movements.

Consideration of the task

In the present study, to clarify the role of the STN in preparation and execution of movements and cancelation of planned movements, we used a goal-directed reaching task with a delay that included Go/Cancel/NoGo trials (Fig. 1). Cancel trials in the present study are slightly different from the conventional SST that presents Stop signals immediately after Go signals. Go and Stop signals trigger independent processes that compete with each other in the horse race model (Logan et al., 1984; Verbruggen and Logan, 2009). The Stop process must overtake and stop the Go process that has already been initiated in the SST. On the other hand, Cancel trials in the present study cancel movements that have been planned during delay periods but not yet been initiated, and therefore may not need rapid and strong STN activation as in the SST. However, Cancel trials and the SST may share the similar brain systems, which are involved in motor response inhibition (Meyer and Bucci, 2016; Tewari et al., 2016).

Role of cancel-related activity

Activity in the STN excites the GPi, thereby inhibits thalamocortical activity, and has inhibitory effects on movements (Alexander and Crutcher, 1990; Mink, 1996; Fife et al., 2017). Stop-related activity has been reported previously in mice (Schmidt et al., 2013), monkeys (Isoda and Hikosaka, 2008; Pasquereau and Turner, 2017), and humans (Bastin et al., 2014; Benis et al., 2016; Mosher et al., 2021). However, such activity was located in the ventromedial region of the STN (Pasquereau and Turner, 2017; Mosher et al., 2021). The ventromedial STN belongs to the associative and limbic cortico-BG loops and does not project to the somatomotor region of the GPi, and therefore, its activation does not seem to be appropriate to cancel/stop movements directly. In the present study, more than half of the dorsal STN neurons examined showed Cancel-related activity (Table 1). These neurons belong to the somatomotor cortico-BG loop, and project to the somatomotor region of the GPi. The latency of their Cancel-related activity was early enough to cancel/stop movements (Fig. 10B). There were no activity changes in NoGo trials in dorsal STN neurons, which differed from putaminal neurons (Takara et al., 2011). One possible origin of Cancel-related activity, especially Cancel-Type I, in the STN is the Stop-related activity increase in the SMA (Chen et al., 2010) that can be transferred to the STN by the hyperdirect and indirect pathways (Polyakova et al., 2020). STN neurons with M1 + SMA inputs showed shorter latency than those with M1 inputs in Cancel-related activity (Fig. 10B3), supporting its origin in the SMA. Cancel-Type II activity may reflect decreased excitability in the M1 in Cancel/Stop trials (van den Wildenberg et al., 2010). Cancel/Stop-related activity in the STN can be the origin of similar activity observed in the GPe (Mallet et al., 2016).

Role of go-related activity

Recorded neurons receiving inputs from the M1 and/or SMA were found in the dorsal region of the STN (Fig. 11), in line with our previous studies (Nambu et al., 1996; Iwamuro et al., 2017). Neurons in the somatomotor region of the STN shows forelimb movement-related activity (Georgopoulos et al., 1983; DeLong et al., 1985; Wichmann et al., 1994), and those in the oculomotor region of the ventral STN shows activity related to saccadic eye movements (Matsumura et al., 1992). Thus, such Go-related activity in the STN seems to reflect cortical activity in the M1/SMA and frontal/supplementary eye fields, respectively. In the present study, STN neurons showed complex task-related activity patterns. The observed activity pattern with multiple peaks of facilitatory and inhibitory phases and with target selectivity may be caused by complexity of the tasks (Greenhouse et al., 2015; Fischer et al., 2016). The facilitatory phase could be mediated by the net excitatory hyperdirect and indirect pathways, while the inhibitory phase could be mediated by the inhibitory GPe-STN projection, part of the indirect pathway (Polyakova et al., 2020).

When we compared the timing of STN activity and EMG changes, activity changes of around 75% of STN neurons preceded EMG changes, and Go- and Cancel-related activity showed similar latencies (Fig. 10B). Thus, STN activity changes have sufficient time to initiate or cancel movements. This early activity changes seem to be transmitted from the cortex to the STN through the extremely rapid hyperdirect pathway. Earlier studies reported that neuronal activity changes in the BG were rather late, and might not contribute to the initiation of movements (DeLong and Georgopoulos, 1981), although another study reported earlier activity changes in the STN compared to those in the GPi (Georgopoulos et al., 1983). Therefore, we urge re-examination of temporal distribution of BG activity in relation to the initiation of movements. The role of Go-related activity in the STN, such as suppressing competing motor actions by a surround inhibition, will be discussed later.

Role of delay-related activity

Our results also showed Delay-related activity after S1 onset in a quarter of recorded neurons (Table 1), which may support their involvement in motor preparation (Thobois et al., 2000; Fischer et al., 2017). These neurons received both M1 and M1 + SMA inputs. The Delay-related activity in the SMA and premotor cortex (Kurata and Wise, 1988) could be transmitted to the STN through the hyperdirect and indirect pathways (Polyakova et al., 2020).

Clinical significance

STN lesions in humans and monkeys induce involuntary movements known as hemiballism (Hamada and DeLong, 1992; Nambu et al., 2000; Hasegawa et al., 2022). According to our present results, Cancel-related activity in the STN will be lost after STN lesions, and the STN will not be able to cancel unwanted movements effectively, resulting in hemiballism. On the other hand, the dorsal somatomotor region of the STN, which is supposed to be hyperactive in Parkinson's disease, is the target for stereotactic surgery, such as coagulation and deep brain stimulation for the treatment of advanced Parkinson's disease (Chiken and Nambu, 2015; Chiken et al., 2021; Nambu et al., 2023). Our present results suggest that these procedures would interrupt Cancel-related activity in the STN, enable the release of movements, and thus ameliorate parkinsonian symptoms.

Functions of the dorsal STN in motor execution and cancelation

Recent studies have emphasized STN activity during motor suppression (Schmidt et al., 2013; Bastin et al., 2014; Benis et al., 2016; Fife et al., 2017; Pasquereau and Turner, 2017; Mosher et al., 2021). On the other hand, the present study revealed that the STN neurons that showed Cancel-related activity always showed Go-related activity, suggesting that the same STN neurons could be involved in both cancelation/stop and execution of movements. Several hypotheses have been presented on the function of the STN, and we would like to examine how they can explain the current results.

1. The main function of the STN could be to cancel/stop movements, as observed in Stop trials (Schmidt et al., 2013; Pasquereau and Turner, 2017; Mosher et al., 2021), and decreased STN activity could facilitate movements to the target. The STN could also cancel/stop unnecessary movements in Go trials by surround inhibition (Mink, 1996; Nambu et al., 2000, 2002b, 2023). For example, while reaching one of the three targets, an increase in STN activity could suppress reaching movements to other targets. If it were the case, Cancel-related activity should suppress all movements, and show similar but larger activity than Go-related activity, which contradicts the present results (Figs. 4–9).

2. The hyperdirect pathway is considered to reset on-going cortical activity to prepare for the next coming signals through the direct pathway that release movements, and the following indirect pathway to stop movements (Nambu et al., 2000, 2002b, 2023). Based on this hypothesis, STN activity seems to contribute to the initiation and termination of movements, i.e., motor execution in Go trials, and also to motor plan cancelation in Cancel trials. Indeed, STN neurons with late facilitation in Go trials (Go-Type I and III), which may be related to the termination of movements, tended to show facilitation in Cancel trials (Cancel-Type I) (Fig. 8), which is consistent with this hypothesis.

3. The main function of STN activity could be to decrease the variability of interspike intervals in GPi neurons, and may not simply excite/inhibit GPi activity (Hasegawa et al., 2022). According to this hypothesis, it is rational that there was indeed no clear relationship between Go- and Cancel-related activity (Figs. 8, 9).

Another point we should consider is the STN projects not only to the GPi, but also to the pedunculopontine tegmental nucleus, through which it may decrease muscles tone to allow smooth limb movements during Go/Cancel trials (Takakusaki et al., 2004).

So far, there is no simple hypothesis that fully accounts for Go- and Cancel-related activity in the STN, and we need further studies.