Acute Striato-Cortical Synchronization Induces Focal Motor Seizures in Primates

Abstract

Objective: Whether the basal ganglia are involved in the cortical synchronization during focal seizures is still an open question. In the present study, we proposed to synchronize cortico-striatal activities acutely inducing striatal disinhibition, performing GABA-antagonist injections within the putamen in primates. Method: Experiments were performed on three fascicularis monkeys. During each experimental session, low volumes of bicuculline (0.5–4 μL) were injected at a slow rate of 1 μL/min. Spontaneous behavioral changes were classified according to Racine’s scale modified for primates. These induced motor behaviors were correlated with electromyographic, electroencephalographic, and putaminal and pallidal local field potentials changes in activity. Results: acute striatal desinhibition induced focal motor seizures. Seizures were closely linked to cortical epileptic activity synchronized with a striatal paroxysmal activity. These changes in striatal activity preceded the cortical epileptic activity and the induced myoclonia, and both cortical and subcortical activities were coherently synchronized during generalized seizures. Interpretation: Our results strongly suggest the role of the sensorimotor striatum in the regulation and synchronization of cortical excitability. These dramatic changes in the activity of this “gating” pathway might influence seizure susceptibility by modulating the threshold for the initiation of focal motor seizures.

basal ganglia, epilepsy, interneurons, seizures, striatum

Introduction

Epilepsy has long been considered as a disease of cortical origin. However, recent evidence suggests that epileptic seizures involve a widespread interaction between the cortical and subcortical structures (Deransart and Depaulis 2002; Norden and Blumenfeld 2002; Paz and Huguenard 2015; Vuong and Devergnas 2017). Among the latter, the basal ganglia (BG) constitute a network of interconnected nuclei involved in sensorimotor, cognitive, and emotional integration (DeLong and Wichmann 2010). Their input structures, the striatum and the subthalamic nucleus (STN), receive information from almost the whole cortex (Haber 2016). In turn, their output nuclei (internal globus pallidus [GPi] and substantia nigra pars reticulata [SNr]) are massively connected via inhibitory projections to the thalamo-cortical pathways (Bolam et al. 2000). Experimental studies in rodent models of generalized absence epilepsy have shown that the BG network may play a key role in the production, development, and maintenance of seizures (Deransart et al. 1998, 2000; Slaght 2004; Paz et al. 2005; Arakaki et al. 2016). However, data obtained from animal models of focal epilepsies remain sparse (Neafsey et al. 1979; Kaniff et al. 1983; Dybdal and Gale 2000; Devergnas et al. 2012). Thus, whether the BG are involved in focal seizures, either passively by propagation or actively in the synchronization of cortical activity via the striato-thalamo-cortical pathway, is still an open question.

The striatal projections neurons (SPNs) are under the control of tonic GABAergic inhibition applied by parvalbumin-positive “fast-spiking” interneurons (Koos and Tepper 1999a), which may be involved in centre-surround-type inhibitory processing (Mallet et al. 2005). This hypothesis posits that they regulate the level of synchrony of SPNs (Gittis et al. 2010) and potentially that of the striato-thalamo-cortical pathway during seizures (Aupy et al. 2019). In addition, previous experiments have shown that microinjections of GABAergic antagonists in the sensorimotor striatum can induce choreic movements (Crossman et al. 1988) and myoclonic tics (Marsden et al. 1975; Crossman et al. 1988; Tarsy et al. 2005; Darbin and Wichmann 2008; McCairn et al. 2009; Worbe et al. 2009; Gittis et al. 2011). These movement disorders presumably result from an increased cortical excitability due to a reduction in the GPi/SNr inhibitory inputs to the thalamus (McCairn et al. 2009). The latter might be caused either by decreased excitation from the STN or by increased inhibition from the striatum (Albin et al. 1989; DeLong and Wichmann 2007). However, the possibility that epilepsy might be due to BG dysfunction has never been evoked. Motor focal seizures are usually characterized by transient motor behavior that occurs simultaneously with hypersynchronous paroxystic cortical activity. Interestingly, myoclonuses and hyperkinetic behaviors can be the semiologic features of focal seizures, especially those originating from the frontal lobe (Bonini et al. 2013). Even if numerous animal models have been used over the years to try to reproduce the abnormal behavior and electroencephalographic (EEG) activity seen in patients suffering from epilepsy, the anatomy and physiology of the NHP brain are unquestionably the most similar to humans. The rich scientific literature and the availability of rodent models are compelling advantages. Nonetheless, nonhuman primate (NHP) models may be more suitable for translational therapies (e.g., deep brain stimulation) and for understanding the underlying network pathophysiology of seizures.

Considering these previous data, we questioned the potential role of the sensorimotor striatum in the cortical synchronization during focal seizure and potentially in their genesis. To this end, we performed microinjections of the GABAergic antagonist bicuculline into the sensorimotor striatum of NHPs in order to induce motor focal seizures. Thereafter, we correlated the induced motor behavior with cortical and subcortical electrophysiological changes. We demonstrate that focal disinhibition of the sensorimotor striatum can induce focal motor seizures correlated with dramatic changes in electrophysiological activity within both the BG and the motor cortices.

Materials and Method

Animals

Experiments were performed in accordance with the National Institute of Health guidelines (2010/63/EU) and the European Community Council Directive for experimental procedures in animals (86/609/EEC). The Institutional Animal Care and Use Committee of Bordeaux (CE50) approved experiments under the license number 4612 (monkeys) and 20 835 (rodents). Experiments were performed on three fascicularis monkeys (Macaca fascicularis), two males (Ze and Di) and one female (Ra) weighing between 5 and 7 kg. Two male Sprague–Dawley rats (Janvier Labs) weighing 425–450 g were also used. Animals were kept at a constant room temperature (21 ± 2 °C) and relative humidity (60%) with a 12 light/dark cycle (darkfrom 8 p.m.) and had free access to water and food. Material and method used for rodent procedure is described in detail in Supplementary material.

Surgical Procedure

Surgical procedures were conducted under aseptic conditions under generalized anesthesia (gaseous anesthesia, isoflurane 1.5–2%, nitrous oxide 1%, and oxygen) after induction with intramuscular ketamine HCl (10 mg/kg), atropine sulfate (0.5 mg/kg), and diazepam (0.5 mg/kg).

The site of injections was determined on presurgical MRI and compared with a stereotactic atlas. A 30-gauge stainless steel cannula guide (Plastic One Inc.) was inserted toward the sensorimotor striatum. The tip position was calculated to be 2 mm above the striatum, ~2.5 mm posterior to the anterior commissure, and 12 mm lateral to the interhemispheric line, in accordance with the coordinates of the sensorimotor putamen (Worbe et al. 2009) and M. fascicularis atlas (Szabo and Cowan 1984). The cannula guide was then fixed on the skull with titanium bone screws and methyl-methacrylate cement (Refobacin® Bone cement LV, Zimmer Biomet©). Stereotactic coordinates of the cannula guide tip are shown in supplemental material. In monkey 1, implantation was done under stereotactic conditions in relation to the anterior commissure (CA) and interhemispheric line using the M. fascicularis stereotactic atlas (Szabo and Cowan 1984). In monkeys 2 and 3, it was done under MRI-guided neuronavigation (Brainsight®, Rogue Research) and compared with stereotactic coordinates. This validated method allowed us to reach specific target areas accurately in the NHP brain including the striatum, GPi, or cortical landmarks (Frey et al. 2004).

Cortical activity was recorded chronically using EEG wires (AS-634 Cooner Wires) fixed in the skull using titanium bone screws and methyl-methacrylate cement. Electrode positions were determined on presurgical MRI and compared with the stereotactic atlas. Since we assumed that putaminal injections would trigger elementary motor phenomena and because of the steric clutter on the monkey skull, we chose to record the motor and premotor regions. Thus, the following regions were targeted and electrodes were placed above the bilateral primary motor cortex (forelimbs and hind limbs) and supplementary motor area.

Reference and Mass Were Placed on the Parieto-Occipital Cortex Contralaterally to the Site of Injection

Lenticular nucleus (including the putamen, the GPe, and the GPi) activity was recorded chronically using a 32ch Platinum Iridium (Parylene C Coating) linear microelectrode array (LMA) of 120 mm with a wire diameter of 12.5 μm and an impedance of 0.8–1 MΩ at 1KHz, implanted within the striatum and the pallidum, in its external and internal segments, ipsilateral to the injection site. The position of the LMA (Alpha Omega) was determined on presurgical MRI and then compared with the stereotactic atlas. The LMA tip targeted the mesial part of the GPi and was inserted at a 45° angle to the axis passing through the interhemispheric line in the coronal plane (see supplemental material for stereotactic coordinates of the GPi tip). The LMA was fixed in the skull using titanium bone screws and methyl-methacrylate cement. Electrode length was 11.2 mm to allow the simultaneous recording of the whole lenticular complex. Reference and mass were inserted under the subdural space.

Microinjections

During each experimental session, the animal was placed in a standard primate chair enabling head movement to be only partially restrained by a Plexiglas window to avoid the monkey manipulating the recording and injection material. A 27-gauge injection cannula (Plastic One Inc.) connected via a Delrin manifold to a 10-ml syringe (Hamilton) was inserted into the sensorimotor striatum (extending 2.5–4 mm beyond the cannula guide tip) and left in place throughout the experiment. We used sterile bicuculline methiodide (Tocris Bioscience), a potent antagonist of GABA-A receptor channels, dissolved in physiological saline at a concentration of 15 μg/μL (29.5 mmol/L). Low volumes of bicuculline (0.5–4 μL) were injected slowly at 1 μL/min. Each bicuculline injection was separated by at least 48 h in the event of behavioral modifications (either myoclonia or generalized seizures) or 24 h in the event of a previous ineffective injection. Different control sessions with saline (NaCl 0.9%) microinjections were administered. Minima of 7 NaCl injections were done in each monkey at 1, 2, or 4 μL. For each monkey, first injections were always saline injections at 2 or 4 μL (with at least 4 saline injections before any drug injections, see supplemental material).

At the end of the experimental sessions and before overdosing of anesthetic drugs, we performed injections of cholera-toxin B (CTB), controlaterally to the site of drug injection in order to estimate the volume of drug diffusion. We used the same methodology as previously described, injecting CTB in the sensorimotor striatum at a low volume (2 μL in rats and 4 μL in primate) and slow rate of 1 μL/min.

Behavioral Analysis

Spontaneous behaviors were recorded during each experimental session using a high-definition multi-channel video system (Ipela SNC, Sony©). Digital videos were stored on a hard drive for offline analysis. Video sessions lasted at least 45 min and were extended if necessary. As soon as it became obvious that drug injections induced abnormal involuntary movements, we used a modified version of Racine’s scale previously used in a kindling model of rodent seizures (Racine 1972). The scale has also been used in primates (Bachiega et al. 2008; Perez-Mendes et al. 2011). Thus, video sessions were retrospectively and blindly scored every 30 s by a neurologist specialized in epilepsy (JA). Our modified Racine’s scale (mRS) was classified as follows.

1. Abnormal orofacial automatisms, including hypersalivation, “mouth-cleaning-like” behavior and tongue automatisms.

2. Clonus restricted to one anatomical localization either orofacial or forelimb.

3. Clonus involving two different anatomical localizations including orofacial and forelimbs.

4. Clonus involving three anatomical localizations including orofacial, forelimbs, and hind limbs or axial body parts (including straub tail or arched body posture).

5. Bilateralization or tonic-clonic generalization with or without postural impairment.

In the event of repeated generalized tonic-clonic (GTC) seizures, diazepam (1 mg/Kg, Valium®, Roche, 10 mg/2 mL) was injected i.m. 1 h after the beginning of the injection. All monkeys were monitored until their symptoms fully abated. All the above-described electrophysiological analyses were later performed on the data recorded before the diazepam injection to avoid any bias due to the effect of the injected GABA-A agonist.

Electrophysiological Recordings

During each session, we recorded simultaneously electromyographic (EMG), bipolar EEG, and intracerebral bipolar local field potential (LFP) activities. EMG recordings: EMG activity was recorded using a trignoTM Wireless system (Delsys inc.). Sensor stickers (TrignoTM Flex Sensor) were used to record the extensor and flexor carpi radialis activity contralateral to the injection site. As we were recording in awake monkeys that were only partially constrained, we had to focus our EMG recording on muscles that we believed would be predominately involved, in line with the results of previous striatal bicuculline injection studies (McCairn et al. 2009; Worbe et al. 2009). EMG signals were sampled at 1 KHz and sent wirelessly to the receiving computer. EEG recordings: Bipolar EEG recordings were performed using the multichannel system. EEG signals were sampled at 10 KHz. LFP recordings: Bipolar LFPs from the sensorimotor putamen, the GPe, and the GPi were recorded through 24 contacts of platinum-iridium linear microelectrode arrays. Intracerebral lenticular nucleus signals were sampled at 20 KHz. Neuronavigation and histopathology data allowed us to reconstruct the electrode position within the lenticular complex on the presurgical MRI. Thus, contacts were classified as follows: putamen contacts, GPe contacts, and GPi contacts. The contacts located at the edge of each structure were not taken into consideration in the analysis to avoid any anatomical localization error. Both EEG and LMA were connected to a 32-channel connector (Omnetics) and sent wirelessly to the receiving computer (advanced wireless W2100, MultiChannel System©). The injection and recording method is summarized in supplemental material.

Electrophysiological Analysis

First, EMG, EEG, and LFP signals were synchronized together as well as with the video recordings, stored, and processed offline using Matlab 2017a (Mathworks). EEG and LFP signals were analyzed using a bipolar montage (between two adjacent contacts) and band-pass filtered between 0.5 and 300 Hz. They were normalized for each experimental session and the EMG signal was rectified |$\Big(\mathrm{normalized}\ \mathrm{and}\ \mathrm{rectified}\ \mathrm{signal}=\Big|\frac{u^n-\overline{u}}{StD_u}\Big|,$| with |$u$|the signal value over time, |$\overline{\mathrm{u}}$| the average signal over the experimental session, and |${StD}_u$| its standard deviation [SD]).

Definition of Regions of Interest

We analyzed the correlation between signals locally recorded from different regions of interest (ROIs) chosen and standardized as follows.

1. As the paroxystic activity was seen mainly on the electrodes located above the primary motor cortex, we analyzed the EEG bipolar signal recorded from the electrodes above the forelimb primary motor cortex.

2. Dorsolateral striatum (Putamen).

3. Globus pallidum externus (GPe).

4. Globus pallidum internus (GPi).

Definition of Period of Interest

To quantify the interaction during seizure activity between the cortex and the subcortical regions explored and to compare them with background, three periods of interest (POIs) were chosen.

1. Background activity (BKG): several background periods without artifacts (at least 50 per monkey, all lasting 10 s) were chosen randomly during NaCl control injections and used as reference periods.

2. Myoclonic jerks: including periods of 10 artifact-free seconds corresponding to focal myoclonic activity interpreted as motor seizures (mRS = II or III).

3. GTC movements: including the first 10 s after the beginning of generalized abnormal movement interpreted as tonic-clonic seizures (mRS = IV or V).

E‌EG Analysis

To compare the number of spikes during each POI, an automatic spike-and-wave detection was applied on the EEG forelimb primary motor derivations (sampling 1 KHz, bipolar montage, band-pass filtered between 0.5 and 300 Hz) on each POI. This consisted in calculating a mean spike discrimination level during the GTC movement period. This was later applied on the other POIs. An SD of 2.0 clearly allowed the spikes to be discriminated from the baseline signal. Spike numbers for each POI were then compared using a one-way ANOVA and unpaired Student’s t-test after Bonferroni correction for multiple comparisons (IBM SPSS statistics 22®). A P-value <0.05 was considered significant.

In addition, to analyze the temporo-spectral content of EEG signals during GTC movements, we performed a time–frequency analysis (TFA). TFA assesses the changes in the involved region with a good resolution in both time and frequency. (TFA—bipolar montage following the application of a bandpass filter of 1–200 Hz. Complex Morlet wavelet analysis was performed in Matlab using a central frequency of 1 Hz and a temporal resolution [full width half maximum] of 6 s).

Back-Averaging Analysis

EMG, EEG, and LFP activities were inspected visually and muscular jerk onsets (characterized by a brief and dramatic increase in muscular activity, both on video and EMG signals) were tagged. To compare the latencies between the myoclonic activity and EEG or LFP, a back-averaging analysis was performed. The objective was to average EEG and LFP signals preceding the onset of myoclonia in order to detect the existence of a pre-myoclonic potential related to muscular activity. The latencies were calculated using a first derivative function on the signal thus obtained. An SD of 2.5 was considered significant. Latencies were then compared between ROIs using a two-way ANOVA and paired Student t-tests after Bonferroni correction for multiple comparisons (IBM SPSS statistics 22®). A P-value <0.05 was considered as significant.

Coherence Analysis

As we wanted to determine whether neuronal oscillations at similar frequencies between the cortex and BG were engaged in functional coupling during seizures, we performed magnitude-squared coherence analysis. This method has already been used in human and NHP to study functional coupling either during physiological behavior or pathological conditions (Buzsáki 2004; Devergnas et al. 2014; Deffains et al. 2016). Analyses were done offline using a custom-written Matlab routine (MATLAB R2016a, The Mathworks). Data were imported and POIs (GTC seizures) were tagged to create a single file per day for each. Next, coherence was estimated for 1-s sliding windows of EEG and LFP signals, overlapping by 500 ms. Coherence was then calculated for each pair of bipolar signals originating from contiguous electrodes above the primary motor cortex (M1) and each adjacent contact within the putamen, GPe, and GPi (M1-Put, M1-GPi, M1-GPe). Results were compared with sham data (baseline). The latter consisted in all nonoverlapping 10-s epochs without artifact recorded during each NaCl injection. Coherence was calculated with the same methodology. Changes in coherence values were analyzed in five frequency bands defined as follows: delta [0.5–3] Hz; theta [3–8] Hz; alpha [8–12] Hz; beta [12–25] Hz; and gamma [25–70]. Finally, to compare coherence between GTC seizure and baseline (sham injection), values of the coherence between each pair of bipolar signals in each frequency band were compared using a two-way ANOVA and paired Student t-tests after Bonferroni correction for multiple comparisons. A P-value <0.05 was considered as significant.

Histology

At the end of the experimental sessions, animals received an overdose of anesthetic drugs (pentobarbital 50 mg/kg) and were transcardially perfused with 0.9% physiological saline solution. For the monkey, the brain was removed. The two hemispheres were divided and cut in three parts each. These tissues were post-fixed in a large volume of 4% buffered paraformaldehyde solution for 1 week at 4 °C and cryoprotected in successive baths of 20% and 30% sucrose solution diluted in 0.1 M phosphate-buffered saline (PBS) at 4 °C until they sunk. For the rats, brains were removed and post-fixed for 5 days and cryoprotected in 20% sucrose-PBS only. Finally, brains were frozen by immersion in an isopentane bath at −55 °C for 5 min and stored at −80 °C. The entire striatum was cut in the coronal plane on a Leica 3050S cryostat into 50-μm serial free-floating sections (12 series for rat and 24 series for monkey) collected in PBS-azide 0.2% and stored at 4 °C.

To estimate the extent of drug diffusion, we performed immunostaining raised against CTB on serial striatal sections of two rat brains and one monkey brain (monkey 3). Briefly, sections of one series for each were blocked in a solution of BSA 2%-TritonX100®0.3%-PBS 0.1 M for 30 min before incubation overnight in a goat polyclonal anti-CTB antibody (#227040, Merck) diluted at 1:20.000 in BSA 0.2%- TritonX100®0.3%-PBS 0.1 M. After rinsing sections through PBS, the stain was revealed by 30-min incubation in an HRP-anti-goat polymer system (goat ImmPRESS Kit VECTOR), followed by DAB revelation for a few seconds (DAKO DAB Kit, #K346811–2, Agilent). Free-floating sections were mounted on slides, counterstained with 0.2% cresyl-violet solution, dehydrated, and coverslipped. Then, high-resolution whole slide images were acquired with a Panoramic SCAN (3D Histech, Hungary) at x20 magnification. The volume of the striatum and CTB staining area within the putamen was estimated with Cavalieri’s formula using Mercator software (Mercator V7.13.4, Explora Nova). In addition, the high-resolution pictures obtained were also used to visualize the microinjection sites based on the traces due to the passage of the cannula and electrodes.

Results

Injections

A total of 39 bicuculline injections were administered in three monkeys (Table 1). Histological reconstruction confirmed that neuronavigation and stereotactic implantations targeted the sensorimotor part of the putamen, posterior to the anterior commissure (Fig. 1A). The electrode targeted the tip of the GPi with a 45° angle and crossed the entire lenticular nucleus (GPi, GPe, and Putamen, Fig. 1A). Additional immunohistological studies in two rats (Fig. 1B) and one monkey (Fig. 1C) showed that the CTB did not diffuse to more than 17.8% of the entire putamen volume (supplemental data) and remained confined within it, without any cortical diffusion.

Table 1

Behavioral effect of bicuculline injection and anatomical localization of induced myoclonic jerks

Animal	Injection no.	Volume (μL)	Onset (min)	Myoclonia anatomical distribution
Ra	1	4	16.5	Orofacial ± forelimbs and hind limbs
	2	4	17.5	Orofacial ± forelimbs
	3	2	20	Orofacial ± forelimbs
	4	2	17	Orofacial ± forelimbs
	5	2	N/A	No effect
	6	2	20	Orofacial ± forelimbs ± hind limbs ± bilatéralization and GTC
	7	1	21.5	Orofacial ± forelimbs ± hind limbs
	8	1	22	Orofacial ± forelimbs
	9	0.5	19.5	Orofacial
	10	0.5	29	Orofacial ± forelimbs
	11	0.5	N/A	No effect
	12	4	15	Orofacial ± forelimbs ± hind limbs
	13	2	N/A	No effect
	14	4	14	Orofacial ± forelimbs
	15	4	N/A	No effect
	16	4	18	Orofacial ± forelimbs
Average (SD)		2.8 (1.2)	19.4 (4.7)
Ze	1	2	13	Forelimbs ± orofacial ± hind limbs ± bilateralization and GTC
	2	2	8	Forelimbs ± orofacial ± hind limbs ± bilateralization and GTC
	3	2	11.5	Forelimbs ± orofacial
	4	1	10	Forelimbs ± orofacial
	5	1	8.5	Forelimbs ± orofacial
	6	2	8	Forelimbs ± orofacial ± hind limbs ± bilateralization and GTC
	7	2	N/A	No effect
	8	2	N/A	No effect
	9	2	40	Forelimbs
	10	2	60	Forelimbs ± orofacial ± hind limbs ± bilateralization and GTC
	11	2	N/A	No effect
	12	2	N/A	No effect
Average (SD)		1.8 (0.4)	23.2 (21.8)
Di	1	4	N/A	No effect
	2	4	4	Forelimbs
	3	4	1.5	Forelimbs
	4	4	N/A	No effect
	5	4	1.5	Forelimbs ± orofacial ± hind limbs ± bilateralization and GTC
	6	4	3	Forelimbs
	7	4	2	Forelimbs
	8	2	7.5	Forelimbs ± orofacial ± hind limbs ± bilateralization and GTC
	9	4	10.5	Forelimbs ± orofacial
	10	4	12.5	Forelimbs ± orofacial ± hind limbs ± bilateralization and GTC
	11	2	N/A	No effect
Average (SD)		3.6 (0.8)	3.0 (1.0)

Notes: N/A = not applicable.

Figure 1

Histological reconstruction and immunohistological drug diffusion analysis. Histological reconstruction confirmed that neuronavigation and stereotactic implantations of the cannula and the LMA targeted, respectively, the sensorimotor part of the putamen, posterior to the anterior commissure (^*) and the GPi tip allowing the recording of the entire lenticular nucleus (#). (A) 3D immunohistological studies in rats (rate 2) (B) and one NHP (monkey 3) at three different AP (C) showed that the volume of diffusion of CTB remained confined within the putamen, without cortical diffusion.

Behavioral Effects

Of the 39 bicuculline injections, 29 (74.3%) produced dramatic behavioral changes. Figure 2A represents the average mRS during bicuculline injections for each monkey compared with NaCl injections. These effects were reproducible and observed in the three animals studied. They were characterized by myoclonic jerks contralateral to the site of injection, clearly different from the monkeys’ normal behavior and easily detectable on video. Myoclonia could affect different muscle groups, including the orofacial musculature, the neck musculature, the proximal or the distal part of the forelimbs, the proximal part of the hind limbs, the axial musculature, and the tail. Once started, the myoclonia abated only when diazepam was injected. In monkey 1, myoclonic activity started on average 23.3 ± 14.6 min after the beginning of the injection with orofacial involvement (activation of the zygomaticus major muscle, associated with ipsilateral and abnormal blinking corresponding to eyelid myoclonia). Progressively, the myoclonia tended to become rhythmic and affected other muscular groups with a “Jacksonian march” starting with the ipsilateral proximal forelimb (activation of the deltoid and biceps brachialis), sometimes followed by the ipsilateral proximal hind limbs (activation of the ilio-psoas). In monkeys 2 and 3, myoclonia started on average 19.9 ± 19.4 min and 5.3 ± 4.3 min after the beginning of the injection, respectively. It initially involved the distal segment of the forelimbs (activation of the fingers and wrist extensors), later evolving with a “Jacksonian march,” with involvement of the proximal forelimbs and orofacial musculature, sometimes followed by myoclonic jerks of the proximal hind limbs. Interestingly, the three monkeys also presented GTC seizures (20.5% of the injections). In addition, we noticed that the symptoms seemed more severe (with a more frequent distribution to the upper and lower limbs and more secondary generalizations) with the largest volumes (2 and 4 μL). On the other hand, with the lowest volumes (0.5 and 1 μL), we did not observe any generalizations but only focal brachio-facial seizures. Decisively, NaCl sham injection of the same volume never produced any myoclonia or abnormal behavior. Figure 2B shows an example of the temporal evolution of behavioral changes during bicuculline injections for one monkey (monkey 3).

Figure 2

Racines’s scale averaged values and classical evolution during bicuculline injections. (A) A total of 39 bicuculline injections were performed and produced reproducible and dramatic behavioral changes, as confirmed by the averaged Racine’s scale for each monkey compared with NaCl injections (median and interquartile ranges). (B) Example of classical temporal evolution of behavioral changes during bicuculline injections for monkey 3 showing the onset of myoclonia in the first 10 min, not stopping until injection of benzodiazepine (60 min).

Visual Inspection and Anatomic-Electrophysiological Correlations

EMG analysis: Myoclonic jerks were clearly detectable on the EMG signal. These myoclonic jerks were characterized either by short myoclonic bursts lasting less than 300 ms (248.4 ± 45 ms), either by short myoclonic bursts immediately followed by a more tonic or prolonged muscular activity (lasting on average 430.2 ± 225.6 ms) contralateral to the injection site. We never observed any stimulus-sensitive jerks. Quantification of the myoclonic events was derived from the rectified and filtered EMG signal. As the recording session progressed, these myoclonic short EMG bursts could be directly followed by a longer and more tonic activity that was concordant with the behavior recorded with the longer abnormal tonic posturing immediately following the myoclonic jerk. GTC seizures were characterized by a disorganized, complex, EMG activity associating myoclonic EMG bursts superimposed on a permanent EMG tonic activity that lasted throughout the GTC (Fig. 3A).

Figure 3

Cortical and subcortical electrophysiological modifications during myoclonia and tonic-clonic seizures. (A) EMG activity recorded from extensor carpis radialis contralateral to injection site (sampling 1 KHz, band-pass filter between 1 and 70 Hz). T0 corresponds to onset of GTC seizure (full line), while each preceding EMG burst corresponds to myoclonia (dash line). (B) EEG activity recorded from electrodes above primary motor cortex ipsilateral to injection site (bipolar montage, sampling 1 KHz, band-pass filter between 1 and 70 Hz). Each EMG myoclonic burst is concomitant with paroxysmal activity characterized by a spike-and-wave followed by polyspikes. GTC seizure begins with the same paroxysmal pattern, followed by rhythmic spiking activity within the alpha band. Spiking activity tends to increase in amplitude (neuronal recruitment), while the inter-spike interval tends to increase until the end of seizure characterized by an abrupt offset. (C) Close-up view of seizure onset (5 s before and after T0) with TFA. TFA allows the analysis of the temporospectral content of EEG signals with a good resolution in both time and frequency and assesses the clear-cut increase in frequency within the primary motor cortex region during myoclonia and seizure (see also Supplementary material). (D) Subcortical LFP activity within the putamen, the GPe and the GPi (bipolar montage, sampling 1 KHz, band-pass filter between 1 and 70 Hz), respectively. Myoclonia and seizure are characterized by paroxysmal activity similar to that in cortex within the three-recorded subcortical regions, suggesting that epileptic activity could be recorded in the BG.

EEG analysis: Visual analysis of EEG recordings revealed abnormal paroxysmal activity characterized by spikes and spike-and-waves only during effective drug injections. Their maximum amplitude was seen on the electrodes located above the primary motor cortex ipsilateral to the site of injection and with a lesser amplitude over electrodes located above the SMA. Myoclonia was always concomitant with spike-and-waves, even though not all spike-and-waves could have triggered this specific behavioral manifestation. During GTC seizures, a progressive increase in the frequency and amplitude of spike-and-wave activity was observed (Fig. 3B). During GTC, this activity diffused to other derivations including SMA, then to the contralateral primary motor cortex (supplemental material). The end of seizure was characterized by an abrupt offset followed by the diffuse flattening of EEG activity (supplemental material). These results were confirmed by the TFA (Fig. 3C). In addition, the automatic spike-and-wave detection showed a clear increase in the number of spikes-and-waves between GTC seizures and sham injections (mRS = 4 or 5, P < 0.0001, Student’s t-test), myoclonic activity and sham injections (mRS = 2 or 3, P < 0.0001), and between GTC seizures and myoclonic injection (mRS = 0, P < 0.0001) (Fig. 4A).

Figure 4

Results of spikes, back-averaging, and coherence analysis. (A) Automatic spike-and-waves detection showing a clear increase in the number of spikes-and-waves between GTC seizures and sham (NaCl) injections (P < 0.0001, Student’s t-test), myoclonic seizures and sham injections (P < 0.0001), and between GTC and myoclonic seizures (P < 0.0001). (B) Back-averaging of hundreds of myoclonia episodes, revealing the emergence of a “retro” evoked potential within the cortex and the subcortical regions preceding the onset of myoclonia. The horizontal dashed lines correspond to the averaged and normalized electrophysiological signal (EEG and LFP) and its standard deviations (2 SD). Vertical dashed lines correspond to the beginning of the evoked potential (i.e., the moment when the first derivative of the signal fate exceeds 2 SD for a duration of at least 25 ms). The vertical full line corresponds to the onset of the EMG myoclonia. Back-averaging analysis confirms that the first electrophysiological modifications appeared within the subcortical regions (putamen and GPe, followed by GPi) before the cortical electrophysiological changes and the occurrence of myoclonia. (C) Averaged values of back-averaging (4.737 myoclonia analyzed). Student t-test ^****P < 0.0001, ^***P < 0.001, ^**P < 0.01. (D) In comparison with BKG (after saline injections), cortico-subcortical functional coupling during GTC seizures showed a significant increase in putamen/cortex, GPe/cortex, and GPi/cortex magnitude-squared coherence within all the frequency bands (P < 0.0001, Student t-test).

LFP analysis: Concomitant to the EEG spikes-and-waves and myoclonic jerks, paroxysmal activity was also recorded at the subcortical level within the striatum, GPe, and GPi. Figure 3D shows an example of ictal LFP activity.

Back-averaging analysis: To compare the latencies between the occurrence of EMG myoclonic activity and EEG and LFP, a back-averaging analysis was performed. On average, electrophysiological modification appeared −28.5 ± 8.6 ms before the onset of myoclonus in the putamen; −29 ± 11.5 ms in the GPe; −23.8 ± 8.6 ms in the GPi; and −15.9 ± 12 ms in the primary motor cortex. Therefore, changes in striatal activity appeared before changes in cortical activity (P = 0.016 and P = 0.0015, Student t-test, respectively, for monkeys 1 and 2); changes in GPe activity appeared before changes in cortical activity (P = 0.019 and P = 0.0105, respectively, for monkeys 1 and 2); and changes in GPi activity began before cortical activity in monkey 2 only (P = 0.018). Figure 4B provides the averaged results of back-averaging analysis. Figure 4C shows the average results of the back-averaging latencies for each structure and each monkey. In addition, one seizure was recorded only at the subcortical level.

Cortico-Subcortical Functional Coupling during Seizures

In comparison with BKG (after saline injections), during GTC seizures (mRS = 4–5), there was a significant increase in cortex-putamen, cortex-GPe, and cortex-GPi magnitude-squared coherence within all the frequency bands (P < 0.0001, Student t-test) (Fig. 4D).

Discussion

We demonstrate for the first time that local pharmacological manipulations of the sensorimotor striatum in NHPs induce abnormal movements similar to the focal motor seizures seen in humans, with or without generalization.

Previous studies using a similar injection protocol found that the drug diffused locally up to 1 mm from the Microdialysis probe (Westerink and De Vries 2001). Yoshida et al. demonstrated that bicuculline injected using the same parameter spreads as an ellipsoid of an approximate radius of 1.5 mm around the injection site (Yoshida et al. 1991). More recently, Worbe et al. investigated the local modification of neuronal activity induced by bicuculline to determine the speed of drug diffusion around the injection site. They found that during higher volume injections (3 μL), neuron modifications could be recorded up to a distance of 1000 μm from the injection site (Worbe et al. 2009). In addition, the results of our immunohistological studies performed in rats and NHPs showed that the volume of diffusion of CTB was lower than 18% of the entire putamen volume and remained confined within the latter, without cortical diffusion. Thus, it is very unlikely that the pharmacological effect of the bicuculline occurred outside the striatum surrounding the injection cannula.

Bicuculline is a well-known GABA-A inhibitor that can induce the disinhibition of the neurons surrounding the injection site. The abnormal movements we observed after intra-striatal bicuculline injections are very similar to those observed in human focal motor seizures with or without tonic-clonic generalizations.

The ability to reproduce the same effects from the same injection site in the same monkey at different volumes, together with the fact that similar effects were observed in three different monkeys, provide strong evidence for the consistency of the locally induced effect within the putamen. These manifestations were due to a specific effect of the drug, as sham injections never produced any particular behavioral changes. The explanation of why some of the injections had no behavioral effect is highly related to the injection technique used. Indeed, microinjection studies often report ineffective injections. Nevertheless our “failure” rate (25–33% of ineffectiveness) was similar to what has been shown in previous studies using bicuculline (McCairn et al. 2009; Worbe et al. 2009). These ineffective injections could potentially be related to the blockage of the injection cannula either during its intracranial placement or during the injection itself. As the striatum is a voluminous structure, it is very likely that a large population of neurons needs to be recruited to produce any behavioral effect (in line with the fact that the largest volume leads to more prominent behavioral changes). Thus, if the drug has not been properly injected or has not properly diffused within the putamen (for the abovementioned reason), it could explain the absence of any behavioral modifications.

The main effect of GABAergic antagonists was contralateral myoclonia, which involved the orofacial regions and/or the upper limb, spreading to the lower limb and then contralaterally, as in human secondary tonic-clonic generalizations. This kind of “Jacksonian march” is concordant with the available striatal somatotopic data in monkeys (Miyachi et al. 2006; Nambu 2011). The cannula tips are located within the medial part of the sensorimotor striatum, which usually corresponds to the limit between the upper limb and facial territories. If the initial ventral diffusion of bicuculline is due to the way the drug is injected, it is logical that the facial or upper limb territories were the first involved by the injected drug.

Abnormal movements (either myoclonia or GTC seizures) were well correlated to cortical EEG spikes-and-wave activity, which is pathognomonic of epilepsy. Spike-and-wave discharges result from the abnormal synchronization of a large population of cortical pyramidal cells and are considered as the EEG hallmark of epileptic activity (Wong et al. 1984). In addition, the localization of this paroxysmal activity over the primary motor cortex is concordant with the topography of myoclonia. As we never observed any bilateral or stimulus-sensitive jerks, we do not believe that the observed behavior was reticular reflex myoclonus. Likewise, GTC seizures occurred with the spreading of the EEG discharges over the different recorded regions and were associated with an increase in the frequency and amplitude of the spike-and-wave rhythmic activity. This confirms that the bicuculline-induced abnormal movements were epileptic.

Similar behavioral manifestations have been previously reported after striatal GABAergic antagonist injections with a similar protocol (Marsden et al. 1975; Crossman et al. 1988; Tarsy et al. 2005; Darbin and Wichmann 2008; McCairn et al. 2009; Worbe et al. 2009; Gittis et al. 2011) and were defined as myoclonus, chorea, tetanic episodes, or tics. Nevertheless, the possibility that these abnormal movements may be of epileptic origin has never been envisaged.

Our data demonstrate that the first electrophysiological activity (LFPs) changes were recorded within the striatum and the pallidum before spreading to the motor and premotor cortices. LFPs are classically considered as the sum of the postsynaptic activity of a limited population of neurons located around the recording electrode (Mitzdorf 1985; Buzsáki et al. 2012). Consequently, the paroxysmal activity of LFPs corresponds to the synchronized activity of the neural population around the recording site, that is, the striatum, GPe, GPi, and then the primary motor cortex. The use of a bipolar configuration for the recordings with a short distance between two adjacent contacts makes the possibility that a distant field (cortical) contributed to the BG activity very unlikely (Lalla et al. 2017; Marmor et al. 2017). Moreover, the existence of multiple phase inversions (supplemental material) between contiguous bipolar contacts closely reflects the locally generated subcortical activity.

Back-averaging analysis over hundreds of myoclonic episodes demonstrated that the first evoked potential appeared within the striatum before the primary motor cortex and before the occurrence of the myoclonic jerks. Moreover, and decisively, one seizure was recorded only at the subcortical level without any obvious behavioral modifications. This confirms that the electrophysiological modifications started within the BG and could be transmitted to the cortex, leading to behavioral changes. Apart from the recent description of the caudate nucleus generating epileptic activity during stereoelectroencephalographic recordings in humans (Aupy et al. 2018), the acute disinhibition of the putamen has never been reported to generate motor seizures.

Worbe et al. showed that, if the firing frequency of the SPNs located near the bicuculline injection site increased, the firing frequency of the neurons located further away in the striatum decreased (Worbe et al. 2009). This suggests that local injections of bicuculline could change the entire dynamics of the striatal network, with a complex effect on the more distant neurons which are not directly influenced by the GABAergic antagonist. Although SPNs receive axonal collaterals from neighboring SPNs (Kawaguchi 1993), their low intrinsic activity and the distal dendritic termination of collaterals on SPNs suggest that these interactions are probably weak. Thus, the main effect of bicuculline might be at the level of the somatic or dendritic terminations of the striatal GABAergic interneurons, particularly those of the fast-spiking type (FSIs) (Koos and Tepper 1999b; Koos 2004; Mallet et al. 2005). GABAergic interneurons are known to exert a powerful inhibition on the SPNs and to form extended networks throughout the entire striatum. It is likely that GABAergic antagonist mimics a decrease in the activity of these GABAergic interneurons, which could explain the changes observed outside the injection zone.

The fact that we observed a similar yet delayed paroxysmal LFP activity between the GPi and the putamen suggests that the putaminal oscillatory activity may propagate through the BG from the input to the output structure. However, the question of whether bicuculline-induced striatal activity modifications affect the direct or indirect striato-pallidal pathways (or both) remains unresolved. In addition, as we did not have the opportunity to record the motor thalamus, we cannot verify whether these abnormal striato-pallidal oscillations propagate through it or not. Nevertheless, the delay between subcortical, cortical, and muscular activity remains compatible with the propagation of motor activity through the corticospinal pathway (Ludolph et al. 1987).

How can we relate the dramatic electrophysiological changes observed within the sensorimotor striatum induced by bicuculline injection with the epileptic activity recorded from EEG? The sharp modulation of striatal inhibition after bicuculline injection could alter the level of synchrony within the striato-thalamo-cortical pathway. Indeed, the coherence between the cortex and the three BG nuclei increased dramatically in almost all the frequency bands during epileptic seizures compared with the BKG during saline control injections. This is concordant with the available data in monkeys (Darbin and Wichmann 2008) and humans, as shown by stereoelectroencephalographic recordings during focal seizures (Kuba et al. 2003; Rektor et al. 2011; Aupy et al. 2019). Sharott et al. showed that coordinated activity of fast spiking interneurons may be involved in mediating oscillatory synchronization within the striatum (Sharott et al. 2009). In addition, an abnormal striatal feedforward inhibition due to GABAergic fast spiking interneurons can promote synchronous oscillatory activity in the striato-thalamo-cortical network during absence seizures (Arakaki et al. 2016). This was corroborated in rodent GAERS experiments, as striatal output neurons stopped firing during spike-and-wave discharges, and then displayed a rebound of activity (Slaght 2004). More recently, Miyamato et al. showed that pharmacological inhibition of cortico-striatal FSI excitatory transmission could trigger absences and convulsive seizures in mice (Miyamoto et al. 2019). Thus, it is likely that injecting GABAergic antagonist within the sensorimotor striatum could dramatically modify the synaptic excitatory/inhibitory balance of the BG output structure neurons (mainly the GPi in primate) and potentially alter cortical excitability, thereby triggering seizures.

Our results suggest that the synchronization level between cortical and striatal activity might be a part of an endogenous mechanism controlling abnormal oscillations within the striato-thalamo-cortical loop. Changes in this synchronization level could potentially promote focal motor and generalized seizures. This is concordant with a recent fMRI study from He et al. demonstrating that impaired inhibitory interactions between BG and thalamus might contribute to abnormal cortico-thalamic synchronization that leads to secondary seizure generalization in temporal lobe epilepsies (He et al. 2020).

Management of focal motor seizures is particularly challenging because antiepileptic drugs often fail and because resective surgery is usually not indicated with regard to potential motor deficit. Interestingly, recent studies have shown that high-frequency STN deep brain stimulation could be an interesting alternative to treat nonsurgical pharmacoresistant focal motor seizures (Prabhu et al. 2015; Ren et al. 2020). These findings should provide a new rationale for deep brain stimulation of the output structures of the BG in patients with intractable focal motor seizures.

In conclusion, we demonstrate that focal motor epileptic seizures can be induced by focal pharmacological manipulation of the BG in NHPs. This reveals the strong impact of the sensorimotor striatum on cortical excitability. This does not mean that epileptic seizures might originate within the striatum but that this subcortical region could play a critical role in the pathophysiology of focal motor seizures through the regulation of cortical excitability. This striato-cortical pathway may not be a part of seizure propagation pathways per se. Nevertheless, striatal acute desinhibition may lead to dramatic changes in the activity of this seizure “gating” pathways influencing seizure susceptibility by modulating the threshold for the initiation and/or propagation of the seizures. Thus, the modulation of striatal GABA levels may therefore represent a pharmacological target of interest in the treatment of seizures. Furthermore, a better understanding of the role of the BG during focal seizures would help in determining which type of epilepsy (epileptogenic zone or networks) might be most likely to benefit from deep brain stimulation.

Authors’ Contributions

J.A. contributed to conception, design, and draft of the work, acquisition, analysis, and interpretation of data. B.R. contributed to acquisition, analysis, and interpretation of data. S.D. contributed to acquisition, analysis, and interpretation of data. M.D. contributed to analysis of data and substantial revision of the draft. N.B. contributed to analysis of data; T.N. contributed to acquisition of data. D.G. and P.B. contributed to conception and design of the work and substantial revision of the draft.

Funding

Fondation pour la Recherche Médicale (bourse médico-scientifique FRM: FDM20150632965 to J.A.); Servier Institute (bourse de mobilité).

Notes

The authors want to thank Hugues Orignac and François Georges for their precious help. Conflict of Interests: None reported.

Author notes

Jerome Aupy, Bastien Ribot contributed equally to the study.