Bicuculline

To determine the location for epileptogenic drug infusion, we performed CT scans before the day of drug delivery while a dummy cannula was placed above the dura, within the slit of the ECoG electrode, and above the dura at the site of AAV injection. CT, MR, and PET images were aligned with PMOD® software using the “Fusion” mode to confirm that the cannula was positioned directly above the hM4Di-expressing region (Fig. 1f). On the day of the epilepsy-induction experiment, animals were placed in a monkey chair with their heads fixed. To monitor and evaluate convulsive body movements and to distinguish them from voluntary movements, monkeys were sedated with intramuscular injection of xylazine (0.8–1.2 mg kg⁻¹ initial dose, and 0.4–0.8 mg kg⁻¹ h⁻¹ for maintenance). Care was taken not to make the injections within 10 min before/after vehicle or DCZ treatment to minimize the effect of xylazine on treatment-induced changes in seizure/convulsion. A syringe with a 33-gauge needle (NanoFil Syringe, WPI) was filled with bicuculline (4 μg μL⁻¹ in PBS) that was freshly prepared on the day of drug delivery. The syringe was mounted to a motorized microinjector (Mode Legato 160, KD Scientific, Holliston, USA) on a manipulator (Model 1460, David Kopf, Ltd.) connected to a customized arm that was attached to the monkey chair. The injection needle was pierced directly through the dura and inserted into the brain to a depth of ~4.0 mm below the dura surface. After 5 min, the needle was moved 1.5 mm back and 0.4–16 μg (0.1–4 μL) of bicuculline was pressure-injected at 1 μL min⁻¹ over 1–2 injection trials until epileptic twitches were observed in the contralateral hand/arm (Supplementary Table 1). Bicuculline was delivered 10–15 min after baseline monitoring of ECoG and body movements. The microinjector was connected to the controller only during the drug injection period to reduce electrical humming noise in the electrophysiological recordings. Each recording session terminated 1–2 h after bicuculline delivery, followed by i.m. injection of Diazepam (0.2–0.8 mg kg⁻¹)—a long-lasting anti-epileptic drug for monkeys—before returning them to the home cage. Monkeys were visually inspected for abnormal behavior and were treated with additional doses of Diazepam (0.1–0.4 mg kg⁻¹) for two more days. The ECoGs and video data from 15 min before bicuculline injection until 60 min after DCZ administration were used for the analyses; data after Diazepam injection was not included.

DCZ (HY-42110, MedChemExpress) was dissolved in dimethyl sulfoxide (DMSO, FUJIFILM Wako Pure Chemical Co.), then diluted in saline to a final volume of 1 mL (2% DMSO in saline), thus achieving 0.1 mg kg⁻¹ dose. Note that this dose of DCZ yields 80–90% hM4Di occupancy and affects behavioral performance via hM4Di activation in monkeys^{12 (link)}. Vehicle control was an i.m. injection of vehicle solution (2% DMSO in saline) at the same volume. Vehicle was delivered 15–20 min after delivery of the bicuculline. The first injection of DCZ was delivered 15–20 min after bicuculline (one case) or vehicle (5 cases). The second injection of DCZ was delivered 18–30 min after the first one (five cases). The timing of vehicle and DCZ administration for each session is summarized in Supplementary Table 1.

ECoG data were recorded using a TDT System3 combined with Intan headstages. Signals were fed to headstage-amplifier/digitizers (RHD2132, Intan Technologies, Los Angeles, USA) and a preprocessing unit (PZ2, TDT), then fed into the digital signal-processing module (RZ2, TDT). The ECoG signal was band-pass filtered between 1.5 and 500 Hz digitally and stored at 3 kHz. Data were acquired using Open developer software (TDT, version 2.16) and analyzed with in-house programs running on the MATLAB R2016a environment^{42 (link)}. Signals were digitally processed with a high-pass filter (1 Hz, tenth order) and a band-stop filter (48–52 Hz, 20th order). The root-mean-square (RMS) of the waveform and its temporal average were computed as the amplitude of the seizure. Spectral power was computed using the short-time Fourier transform and normalized for each frequency using the baseline power averaged within a 3-min time window 15 to 12 min before bicuculline injection. Data were visualized either with MATLAB or GraphPad Prism 9 (GraphPad Software, San Diego, USA).