Acute and chronic convection-enhanced muscimol delivery into the rat subthalamic nucleus induces antiseizure effects associated with high responder rates

Intracerebral drug delivery is an emerging treatment strategy aiming to manage seizures in patients with sys- temic drug-resistant epilepsies. In rat seizure and epilepsy models, the GABA A receptor agonist muscimol has shown powerful antiseizure potential when injected acutely into the subthalamic nucleus (STN), known for its capacity to provide remote control of different seizure types. However, chronic intrasubthalamic muscimol delivery required for long-term seizure suppression has not yet been investigated. We tested the hypothesis that chronic convection-enhanced delivery (CED) of muscimol into the STN produces long-lasting antiseizure effects in the intravenous pentylenetetrazole seizure threshold test in female rats. Acute microinjection was included to verify efficacy of intrasubthalamic muscimol delivery in this seizure model and caused significant antiseizure effects at 30 and 60 ng per hemisphere with a dose-dependent increase of responders and efficacy and only mild adverse effects compared to controls. For the chronic study, muscimol was bilaterally infused into the STN over three weeks at daily doses of 60, 300, or 600 ng per hemisphere using an implantable pump and cannula system. Chronic intrasubthalamic CED of muscimol caused significant long-lasting antiseizure effects for up to three weeks at 300 and 600 ng daily. Drug responder rate increased dose-dependently, as did drug tolerance rates. Transient ataxia and body weight loss were the main adverse effects. Drug distribution was comparable (about 2 – 3 mm) between acute and chronic delivery. This is the first study providing proof-of-concept that not only acute, but also chronic, continuous CED of muscimol into the STN raises seizure thresholds. Some of the present results have been presented as a congress poster (Feja et al., 2020).


Introduction
The mainstay of epilepsy treatment involves symptomatic seizure suppression by systemically administered antiseizure medication (ASM). However, systemic pharmacotherapies for epilepsies are facing two main challenges. First, neurological and/or peripheral adverse effects are common. Second, about one-third of patients suffering from epilepsies have seizures refractory to systemic pharmacotherapy despite the recent introduction of new ASMs (Asadi-Pooya et al., 2017;Janmohamed et al., 2020;Kalilani et al., 2018;Kehne et al., 2017).
Intracerebral drug delivery is an emerging treatment strategy aiming to overcome those challenges. The idea is to bypass blood brain barrierassociated drug resistance mechanisms and to allow high drug concentrations at promising brain target sites of the epileptic network, while Abbreviations: aCSF, artificial cerebrospinal fluid; ASM, antiseizure medication; BODIPY, boron-dipyrromethene; CED, convection-enhanced delivery; GABA, γ-aminobutyric acid; GABA-T, GABA-transaminase; ivPTZ-ST test, timed intravenous pentylenetetrazole seizure threshold test; SNr, substantia nigra pars reticulata; STN, subthalamic nucleus. reducing the risk for peripheral or neurological adverse effects . Furthermore, this targeted treatment approach enables the use of otherwise toxic substances not suitable for systemic delivery, such as the GABA A receptor agonist muscimol (3-hydroxy-5-aminomethylisoxazole), which is degraded much more slowly in the brain than GABA itself (Baraldi et al., 1979;Majchrzak and Di Scala, G., 2000). Based on a series of preclinical feasibility and safety studies in nonhuman primates (Heiss et al., 2019b(Heiss et al., , 2010(Heiss et al., , 2005, Heiss et al. (2019a) recently examined the safety and effectiveness of convection-enhanced delivery (CED) of muscimol into the seizure focus of three human patients resistant to systemically applied antiseizure medication over a period of 12-24 h. While there were no safety issues, only one patient (neocortical seizure focus) experienced reduced seizure frequency during muscimol infusion compared to vehicle infusion, while two patients with hippocampal seizure focus did not (Heiss et al., 2019a). Indeed, the high complexity of the hippocampal network might impede therapeutic interventions targeting the hippocampal area depending on the drug used (Bröer et al., 2013). Additionally, targeting the seizure focus might not always be an option, due to the existence of multiple seizure foci, undefinable focus location, or focus location in eloquent brain areas (Khoo et al., 2021). Apart from directly targeting the seizure focus, another highly promising strategy is to deliver muscimol into brain regions crucially involved in seizure propagation and remote modulation of seizure initiation. An attractive target region in this respect is the subthalamic nucleus (STN), a basal ganglia-associated region known to mediate pronounced antiseizure effects when inhibited (Bröer et al., 2012;Handreck et al., 2014) and to provide a notable nonselective seizure modulation (Gernert, 2013;Gernert and Feja, 2020). Thus, acute bilateral microinjection of muscimol into the STN of rats raised flurothyl-induced clonic seizure thresholds (Velísková et al., 1996), protected against limbic motor seizures evoked either by intravenously-injected bicuculline or by application of bicuculline into the anterior piriform cortex (Dybdal and Gale, 2000), prevented generalization of amygdala-kindled seizures (Deransart et al., 1998), and suppressed absence seizures (Deransart et al., 1996). Clinical evidence for the STN as a promising therapeutic target region comes from high-frequency deep brain stimulation trials in human patients suffering from different types of epilepsy (Chabardes et al., 2002;Klinger and Mittal, 2018;Loddenkemper et al., 2001;Ren et al., 2020).
Apart from appropriate target selection, chronic drug delivery is required for further development of intracerebral pharmacotherapy strategies aiming to achieve long-term seizure control. The putative antiseizure effects of chronic muscimol delivery into the STN have not yet been investigated. We therefore aimed to investigate if muscimol is an appropriate drug for chronic intrasubthalamic delivery.
Using an acute screening model, which provides a sensitive parametric method for assessing seizure thresholds in individual animals (Green and Murray, 1989;Löscher, 2009;Pollack and Shen, 1985), we hypothesized that chronic CED of muscimol over three weeks into the STN of rats (a) induces prolonged antiseizure effects with (b) a high responder rate, (c) only mild adverse effects, and (d) restricted drug distribution within the brain. To first verify that intrasubthalamic muscimol delivery is effective in the intravenous pentylenetetrazole seizure threshold (ivPTZ-ST) model, we included acute muscimol microinjection experiments in addition to the chronic muscimol delivery experiments.

Animals
In order to ensure better comparability with our previous studies on therapeutic targeting of the STN (Backofen-Wehrhahn et al., 2018;Bröer et al., 2012;Gey et al., 2016;Handreck et al., 2014), we used adult female Wistar Unilever rats (HsdCpb:WU), which were purchased at a body weight of 190-220 g from Envigo (Venray, Netherlands). Rats were housed without males in order to keep them acyclic or asynchronous with respect to their estrous cycle (Kücker et al., 2010). We have previously shown that the seizure threshold is not affected by the estrous cycle in rats maintained under these conditions (Bröer et al., 2012;Rattka et al., 2012). Rats were housed in groups of up to four in Mak-rolon® boxes of type IV (BIOSCAPE/Ebeco, Castrop-Rauxel, Germany) before and individually in Makrolon® boxes of type III after surgery to avoid mutual gnawing at the sutures. During both group and individual housing, animals were provided with nesting material, wooden chew sticks, and a red transparent tube (BIOSCAPE/Ebeco, Castrop-Rauxel, Germany) for environmental enrichment. Rats were housed under controlled environmental conditions (room temperature 22-24 • C, room humidity 50-60%) with a 12/12 h light/dark cycle (light on at 6:00 a. m.) for at least 2 weeks before the experiments. Handling and experiments were performed during the first half of the light period. Standard laboratory chow (Altromin 1324 standard diet) and tap water were allowed ad libitum. The weight of all rats was monitored regularly.

Ethics statement
All experiments complied with the ARRIVE guidelines, were done in compliance with the German Animal Welfare Act and the European Union (EU) council directive 2010/63/EU, and were formally approved by the animal subjects review board of the Lower Saxony State Office for Consumer Protection and Food Safety (LAVES, Oldenburg, Germany; file numbers 14/1727, 17A207, 20/3399). All surgeries were performed under general anesthesia, and all efforts were made to minimize suffering.

Intravenous pentylenetetrazole seizure threshold test
An overview of the study design is shown in Fig. 1. For determination of seizure thresholds before cannula implantation (basal seizure threshold) and at several time points in implanted rats, the timed ivPTZ-ST test (Green and Murray, 1989;Löscher, 2009;Orloff et al., 1949;Pollack and Shen, 1985;Zolkowska et al., 2021) was performed as described previously (Feja et al., 2021). The basal seizure threshold (predrug control) was determined at least 2 days before surgical cannula implantation (see below). In implanted rats, the seizure thresholds were determined 15 min after acute or once a week over three weeks during chronic intrasubthalamic muscimol or vehicle delivery (Fig. 1).
We used this antiseizure screening model, because it is particularly sensitive to interventions on the GABAergic inhibition and predicts antiseizure effects against different experimental and clinical seizure types, including difficult-to-treat focal onset seizures (Green and Murray, 1989;Mandhane et al., 2007). Furthermore, the ivPTZ-ST test can be repeatedly performed in the same rat, thereby allowing the study of alterations in seizure thresholds in individual rats (Löscher, 2009). To prevent a kindling effect, in the present study the number of ivPTZ-ST determinations was limited to a maximum of six in the same rat.
Briefly, the ivPTZ-ST was determined by continuous infusion of a 0.8% solution of PTZ via a 24 G needle into the lateral tail vein of conscious, unrestricted rats. The needle was connected to a syringe by a flexible polyethylene tubing (PE20 tubing; Instech Laboratories, Plymouth Meeting, PA, USA) and the PTZ solution was infused at a constant rate of 1.0 ml/min using an infusion pump (Pump 11 Elite, Harvard Apparatus, Holliston, MA, USA). The time until first myoclonic twitch and until clonic seizure was recorded, the latter of which also served as the end point for PTZ infusion. The ivPTZ-ST to first myoclonic twitch and to clonic seizure was calculated for each animal in mg/kg based on latency to seizure induction, animal body weight, and PTZ concentration and infusion rate.
We determined the ivPTZ-ST at the same time of the day (between 9:00 and 12:00 a.m.) to avoid intraday variance between animals. The experimenter was blinded with respect to the treatment groups of the rats.
Based on previous experience with intracerebral delivery of the ASM vigabatrin in the ivPTZ-ST model (Gey et al., 2016), we expected that not all rats would respond to muscimol with the same antiseizure effect, but that responders and nonresponders would exist. For the present study, we defined responders and nonresponders in all groups, including vehicle, as described recently (Feja et al., 2021). In responders, the increase in the respective seizure threshold, i.e., myoclonic twitch or clonic seizure, had to be at least 25% higher than the individual basal seizure threshold, while nonresponders did not reach this criterion. In contrast to our previous study, in which the responder threshold was defined in relation to the highest response induced in a rat from the vehicle group (Gey et al., 2016), our current responder definition considers interindividual differences in responsiveness to intracerebral drug delivery. In addition, this responder threshold is sufficient to consider intraindividual variations in seizure thresholds.
With regard to the chronic drug delivery, we additionally verified if loss of antiseizure efficacy (tolerance) occurred in responders over time.
Tolerance was defined as a maintained reduction in seizure threshold below the above-described 25% cut-off level during treatment following an initial increase in seizure threshold.

Behavioral and physiological test battery
The general health and putative adverse effects were assessed using a test battery (Table 1) consisting of an adapted Irwin screen (Irwin, 1968), which was reduced according to previous experience (Bröer et al., 2012;Gernert and Löscher, 2001;Gey et al., 2016), a reduced hyperexcitability test modified from (Moser et al., 1988;Rice et al., 1998), the rotarod performance test modified from (Dunham and Miya, 1957), and monitoring of body weight and rectal temperature. Any additionally observed abnormalities were noted. The Irwin screen and the hyperexcitability test were conducted in a round, black open field Overview of the study design. Muscimol was microinjected either acutely (A) or chronically (B) into the subthalamic nucleus (STN) of adult rats. For determination of the predrug control seizure threshold before cannula implantation (basal threshold) and at different time points in implanted rats, we performed the timed intravenous pentylenetetrazole (PTZ) seizure threshold test. The basal seizure threshold was determined at least 2 days (2 d) before surgical cannula implantation. Muscimol testing started not earlier than 7 days (7 d) after surgery and was performed as a crossover within-subject design (acute study) or as a parallel between-group design (chronic study). The different doses are marked with different colors. In implanted rats, the seizure thresholds were determined 15 min after acute or once a week over three weeks during chronic intrasubthalamic muscimol or vehicle delivery. Behavioral testing (not indicated in the figure) was performed 5 min after muscimol microinjection, i.e., 10 min before the PTZ test (acute study) or during muscimol delivery before the PTZ test (chronic study). BODIPYmuscimol was used to evaluate infusate delivery within the brain. For more details refer to chapter 2.
(78 cm in diameter). In animals receiving acute microinjections, the test battery was performed 5 min after termination of intrasubthalamic drug delivery, i.e., starting 10 min before we conducted the ivPTZ-ST test. In animals receiving chronic microinfusion, the test battery was performed 30-60 min prior to ivPTZ-ST testing. All animals were allowed to habituate to the laboratory conditions for 30 min prior to the start of the test battery, which was conducted as follows.
First, for the adapted and reduced Irwin screen, each rat was observed for 1 min in the open field and scored according to Table 1. Next, the response to external tactile stimuli was determined and scored in a hyperexcitability test using two simple tests (touch-response and pick-up; Table 1) to identify interindividual differences in behavioral excitability and sensory responsiveness. Motor coordination, balance, and grip strength were then assessed and scored by the rotarod test (Table 1). Here, animals had to balance on a rotating rod (8 rpm) without falling off for 1 min in one of a maximum of three trials, with a 1 min break between failed trials. Finally, body weight and rectal temperature were measured and, in implanted animals, corrected for the weight of the filled microinfusion pump (7.9 g).
Animals were habituated to the full test battery 3 days prior to the first experimental day. The test battery was always conducted between 8:00 a.m. and 12:00 p.m. to avoid a circadian influence. The open field was wiped with 0.1% acetic acid for cleaning and odor neutralization after each use.

General surgical procedures
For implantation of cannulas (acute microinjection) or cannulas and microinfusion pump (chronic microinfusion), rats were anesthetized with isoflurane (induction: 3% in an anesthesia box, maintenance: 0.8-2% via mask). In addition, the local anesthetics tetracaine (2%, Sigma-Aldrich, St. Louis, MO, USA) and bupivacaine (Carbostesin® 0.25%, AstraZeneca GmbH, London, UK) were applied to incision sites and periost, respectively, 5 min before the surgical procedures. The depth of anesthesia was tested regularly by way of pedal withdrawal reflex as a response to toe pinching, and breathing frequency was monitored. During surgery, animals were kept on a sterile surgical drape on top of a heating pad and temperature was monitored via rectal probe. To protect the animals' eyes from desiccation, Bepanthen® eye and nose ointment (Bayer Vital GmbH, Leverkusen, Germany) was applied.
After surgery, animals were kept on the heating pad and monitored until they fully regained consciousness. All animals were allowed a recovery period of about 1 week. As post-operative care, rats received 0.3 ml/kg subcutaneously (s.c.) of the analgesic L-Polamivet® (levomethadon, 2.5 mg/ml combined with fenpipramidhydrochlorid, 0.125 mg/ml, Intervet International GmbH, Unterschleissheim, Germany) for post-surgical pain control and 5 ml Sterofundin® ISO intraperitoneally (i.p.) (B. Braun Melsungen AG, Melsungen, Germany) to counterbalance fluid loss during surgery. To prevent bacterial infection, 0.1 ml/animal of the antibiotic Marbocyl® FD 1% (marbofloxacin, 10 mg/ml, Vetoquinol GmbH, Ismaning, Germany) was applied s.c. twice a day from two days before surgery until the fourth post-surgical day. The following coordinates relative to bregma (Paxinos and Watson, 2014) were used to target the STN: 3.4 mm posterior, + /-2.6 mm lateral, 7.7 mm ventral (incisor bar − 3.3 mm).

Specific procedures for acute muscimol microinjections
For acute microinjections as adapted from our previous study (Gernert and Löscher, 2001), the anesthetized rats were stereotaxically implanted with bilateral stainless steel guide cannulas (outer diameter 0.64 mm, inner diameter 0.4 mm) 2 mm above the intended injection site and closed by removable stylets of the same length before and between drug experiments. Two cylinder-head screws (M4 thread; 3 mm shaft length) and dental acrylic cement (Paladur®, Kulzer GmbH, Hanau, Germany) anchored the assembly. The first layer of acrylic cement with direct contact to the skull was prepared with marbofloxacin (50 mg marbofloxacin for every 10 g powder component of Paladur®) to reduce the risk of infection arising between the skull and the resin.
Acute microinjections were conducted in awake rats in an observer cage using a flexible, but precise injection system, which allowed to control the injection volume by the movement of a small air bubble in the tubing as described previously (Gernert and Löscher, 2001;Gey et al., 2016).
Briefly, for simultaneous bilateral microinjection of muscimol or vehicle into the STN, the stylets were removed and stainless steel injection cannulas (outer diameter 0.35 mm, inner diameter 0.15 mm) connected to 1 µl-Hamilton syringes via flexible polyethylene tubing with a volume of 1.2 µl/100 mm were inserted in both hemispheres. The injection cannulas were 2 mm longer than the guide cannulas to target the STN. One minute after inserting the injection cannulas, the drug/ vehicle solution was microinjected over a period of 4 min. The injection cannulas were left in place for 1 min after the end of injection and controlled for permeability after removal from the brain.
Dose selection for the acute experiments was based on previous experiments in other seizure/epilepsy models (Dybdal and Gale, 2000;Gernert and Löscher, 2001) and on pilot tests. Aliquots of stock solutions (120 ng/250 nl and 60 ng/250 nl) were prepared and stored at − 20 • C until use. At treatment days, the aliquots were thawed and, if necessary, further diluted to obtain the following muscimol amounts, which were acutely microinjected bilaterally into the STN, i.e., into each hemisphere: 10 ng, 30 ng, 60 ng, and 120 ng (microinjection volume, 250 nl each hemisphere). The pilot tests indicated a no-observed-effect-level of 10 ng microinjected bilaterally into the STN (n = 2; not illustrated). The high amount of 120 ng microinjected bilaterally into the STN in one of the two pilot tests caused acute non-acceptable adverse effects including sedation, paralysis of the hind limbs, and retraction of the abdominal wall as a sign of pain/distress (not illustrated). In contrast, the pilot tests indicated antiseizure effects and a good safety profile with 60 ng in 0 -no reaction or normal turn towards touch Rat is slightly touched with a pen at its flank contralateral to the implanted pump side.

Pick-up test:
0 -very easy Rat is picked up by hand.
1 -easy with vocalization 2 -difficult (startle or freeze or escape reaction) 3 -very difficult or aggressive reaction Rotarod test Each trial 1 min on rotating rod 45 cm above ground at constant speed of 8 rounds per minute (rpm).
0 -the rat passed on the first attempt (no fail) 1 -the rat passed on the second attempt (1 fail) 2 -the rat passed on the third attempt (2 fails) 3 -the rat failed all three trials 250 nl microinjected bilaterally into the STN. Therefore bilateral intrasubthalamic microinjections of 30 ng and 60 ng muscimol in 250 nl artificial cerebrospinal fluid (aCSF), and 250 nl aCSF as vehicle, were tested (n = 16) in a randomized crossover (latin square) design ( Fig. 1). Five min after microinjection, rats were tested for potential behavioral and physiological adverse effects as described above. After completion of the behavioral assessment, i.e., 15 min after microinjection, we conducted the ivPTZ-ST test. Microinjections and ivPTZ-ST tests were conducted at an interval of at least four days to allow washout of the drug (Fig. 1).
In some rats, additional tests investigating the distribution range of the infusate within the brain tissue by microinjection of borondipyrromethene (BODIPY)-muscimol (n = 6) were performed after termination of the latin square trials.
To verify if the surgical procedure of cannula implantation influenced the seizure threshold, a post-surgical seizure threshold was determined without prior drug or vehicle microinjection in 9 of the animals used for the acute experiments.

Specific procedures for chronic convection-enhanced muscimol microinfusion
For chronic CED as adapted from Gey et al. (2016), battery-powered, programmable and implantable microinfusion pumps with a reservoir volume of 900 µl and a battery life of about 6 months were used (iPRECIO®, Model SMP-200, Primetech Corporation, Tokyo, Japan). The pumps were filled with the respective muscimol or vehicle solution and maintained in an incubator at 37 • C for 24 h before surgery. For chronic CED, muscimol solutions were freshly prepared to yield daily doses of 60 ng, 300 ng, and 600 ng per hemisphere (corresponding to 2.5 ng/h, 12.5 ng/h, and 25 ng/h, respectively). The daily dose of 60 ng has been chosen because of its antiseizure efficacy in the acute experiments. Since this dose did not show sufficient efficacy in the chronic experiment (see 3.2.2), we then used the higher daily doses of 300 and 600 ng. Artificial CSF served as control. Infusion volume rate was set to 0.2 µl/h, i.e., 0.1 µl/h per hemisphere in all cases. The infusion parameters (infusion start, end, rate) were programmed with the iPRECIO® Management Software Version 1.3.
For chronic CED of muscimol and vehicle, a total of 33 rats received surgical implantation of a microinfusion pump into a subcutaneous pouch on the anesthetized rat's abdominal side. The pump was fixed to the muscle tissue in the subcutaneous pouch with sterile 5-0 nonabsorbable nylon monofilament. The catheter of the microinfusion pump was then fed through a subcutaneous passageway made from the back incision to the head incision site using a thin metal pipe. Cannulas (outer diameter 0.36 mm, inner diameter 0.18 mm; L-shaped stainless steel tube topped with a pedestal, length under pedestal 9 mm; Model 328OP, Plastics One Inc., Roanoke, USA) were attached to a 22 G Y-connector (Instech Laboratories, Plymouth Meeting, PA, USA) with additional tubing of the same material and diameter as the pump catheter (Styrene-Ethylbutylene-Styrene, outer diameter 1.2 mm, inner diameter 0.55 mm). The whole system was rinsed with muscimol solution or vehicle to confirm patency. Once filled with the appropriate substance and confirmed to contain no visible air bubbles, the cannula-Yconnector system was attached to the catheter of the microinfusion pump. The cannulas were then stereotaxically implanted (for coordinates, see 2.5). The cannulas were fixed to the skull with two screws anterior to bregma and with Paladur®. The catheter of the microinfusion pump was fixed to the neck musculature with two individual button sutures, and incisions were closed using sterile 5-0 nonabsorbable nylon monofilament with simple interrupted sutures.
In pump-implanted rats, behavioral and physiological testing was performed once a week before the ivPTZ-ST test. After three weeks of drug delivery and termination of the experiments, the microinfusion pumps were explanted under isoflurane anesthesia. The remaining solution was extracted from each pump reservoir and the pH was found to be stable compared to the initial pH adjusted before pump filling. Since the microinfusion pumps are refillable, they could be used for several experiments after cleaning and disinfection. For our experiments, each pump was used for a maximum of three rats and only for either drug or vehicle experiments.
In an additional group of 4 rats, the distribution range of the chronically delivered drug within the brain tissue was investigated by chronic microinfusion of BODIPY-muscimol (see below).

Tracking of the distribution range of the infusate with BODIPYmuscimol
For histological estimation of muscimol distribution within the brain after acute and chronic drug delivery to the STN, a fluorophoreconjugated muscimol (autofluorescent boron-dipyrromethene-labeled muscimol, BODIPY-muscimol, TMR-X conjugate; excitation = 544 nm, emission = 570 nm) was used.
Some of the rats (n = 6) used for acute microinjection received one last microinjection of BODIPY-muscimol after termination of all latin square trials ( Fig. 1) to examine the distribution range of the drug. Identical to the acute muscimol microinjection trials, rats were perfused 15 min after BODIPY-muscimol injection. Likewise, an additional group of rats (n = 4) received chronic, CED of BODIPY-muscimol over 3 weeks until perfusion and histological examination of the distribution range ( Fig. 1).
The molecular weight of BODIPY-muscimol (607.46 g/mol) is 5.32fold higher compared to muscimol (114.1 g/mol). Thus, for acute microinjection, BODIPY-muscimol was dissolved in filtered aCSF in a concentration of 320 ng/250 nl, which is equimolar to the 60 ng/250 nl muscimol solution. For chronic delivery, BODIPY-muscimol was dissolved in filtered aCSF to deliver 1.6 µg in 2.4 µl/d and hemisphere, which is equimolar to a muscimol amount of 300 ng in 2.4 µl/d and hemisphere. All steps including handling of the substance were performed under exclusion of white light. Undissolved BODIPY-muscimol was stored at − 20 • C until used. The solution was freshly prepared on the treatment day or pump filling day.

Drugs
Muscimol (Tocris Biosciences, Bristol, UK) and BODIPY-muscimol (Hello Bio, Bristol, UK) were dissolved in filtered, artificial aCSF ("mock CSF"). All solutions used for intracerebral delivery were titrated to pH 7.3 with sodium hydroxide and hydrochloric acid on the treatment day if necessary. For details on dosing of muscimol and BODIPYmuscimol please refer to the specific chapters. Filtered aCSF served as vehicle and was stored at 37 • C. Osmolarities were measured after termination of all experiments by means of osmometers (Osmomat 030 or Osmomat 3000, Gonotec, Berlin, Germany). In the acute experiments, osmolarities of muscimol solutions and aCSF were 257-307 mOsm/l and 270-282 mOsm/l, respectively. In the chronic experiments (solutions regained from microinfusion pumps), osmolarities of muscimol solutions and aCSF were 253-300 mOsm/l and 275-282 mOsm/l, respectively.
PTZ was purchased from Sigma-Aldrich (St. Louis, MO, USA) and was freshly dissolved on test days in 0.9% saline to create a 0.8% solution. Undissolved PTZ was stored at − 20 • C.

Perfusion
At termination of experiments, rats were sacrificed for verification of cannula localization and, where applicable, estimation of BODIPYmuscimol distribution. For removal of microinfusion pumps and Yconnectors, rats were again anesthetized with isoflurane (described above). Subsequently, these rats as well as the rats used for acute intrasubthalamic microinjection were transcardially perfused under deep anesthesia induced by i.p. injection of 500 mg/kg pentobarbital (Euthadorm®, CP-Pharma-Handelsgesellschaft GmbH, Burgdorf, Germany) in a volume of 1 ml/kg. Transcardial perfusion was conducted using a peristaltic pump (50 ml/min). The rats, with which no fluorescence microscopy was planned (muscimol, vehicle), were perfused with phosphate-buffered saline (PBS, 0.01 mol/l, pH 7.4) followed by 4% formaldehyde in phosphate buffer (PB, 0.2 mol/l, pH 7.4). The rats, which received BODIPY-muscimol and were planned for fluorescence microscopy, were perfused with PBS (0.01 mol/l, pH 7.4) followed by 4% paraformaldehyde (PFA) in PB (0.2 mol/l, pH 7.4).
The brains were resected from the skull and cryoprotected in sucrose. Thymol (30 mg/30 ml, Merck Millipore, Darmstadt, Germany) served as fungicide.
The acute BODIPY-muscimol microinjected animals were perfused 15 min post-injection for estimation of the distribution at the very moment of seizure threshold determination following acute muscimol microinjection. The chronic BODIPY-muscimol microinfused animals were perfused after three weeks of chronic drug delivery, i.e., on the same day or one day after the last ivPTZ-ST test.

Processing of brain sections from muscimol-treated animals
Coronal sections were cut at 40 µm on a freezing microtome (Leica Microsystems GmbH, Wetzlar, Germany) and mounted onto chromegelatine-coated regular glass slides, air-dried for 24 h, then Nissl stained with thionine (Sigma-Aldrich), and finally coverslipped using the Roti® Histokitt II mounting medium (Carl Roth GmbH & Co. KG, Karlsruhe, Germany). The thionine-stained sections were used for verification of microinjection sites by means of a light microscope (Leica DMLB, Leica Microsystems, Wetzlar, Germany). Accuracy of cannula implantation was compared against an anatomic atlas (Paxinos and Watson, 2014). Only rats with correct bilateral placement of cannula tips in the STN were used for data evaluation.

Processing of brain sections from BODIPY-muscimol-treated animals
Coronal sections were cut at 40 µm on a freezing microtome. One series of sections was used for verification of microinjection sites by means of thionine staining as described above. Another series was used for evaluation of the distribution range of the infusate by means of a fluorescence microscope (acute experiments: Zeiss Axioscope 50, Carl Zeiss Microscopy GmbH, Jena, Germany, equipped with a tetramethylrhodamine filter; chronic experiments: Zeiss Axio Observer 7, Carl Zeiss Microscopy GmbH, Jena, Germany, equipped with a 2.5x magnification lens, a Zeiss Colibri 7 LED light source and a DsRed and DAPI filter). For this purpose, sections were mounted on Superfrost Plus microscope slides (Thermo Fisher Scientific) in tris-buffered saline (TBS) and coverslipped using Entellan® new rapid mounting medium (Merck Millipore, Darmstadt, Germany) or ProLong™ Gold Antifade Mountant (Thermo Fisher Scientific).
In the acute experiments, the images of the second series were acquired, processed and analyzed with the Image Pro Plus 6.2 software (Media Cybernetics Inc., Abingdon, United Kingdom) and Image J (version 1.53). All sections were photographically recorded under fluorescent light for visualization of the infusate and during darkfield microscopy for visualization of the brain structures, both with 2.5x magnification. In the chronic experiments, the images from the second series were acquired, processed and analyzed with Stereo Investigator (version 11; MBF Bioscience, Williston, VT, USA) and Image J (version 1.53). All sections were photographically recorded under fluorescent light at 555 nm for visualization of the infusate and at 385 nm for visualization of the brain structures. The exposure time was 100 ms for image acquisition. Fluorescent and brightfield images (acute experiment) and fluorescent images under 385 and 555 nm excitation (chronic experiment) were joined using the image overlay function of the software. When analyzing the fluorescence of the samples and evaluating the distribution distance in the anterior-posterior plane, the first and the last section showing fluorescence were considered the first and last section containing BODIPY-muscimol. For evaluation of the distribution in the medio-lateral plane, the distance between the two fluorescent signals the furthest apart was measured on several sections and the longest distance was defined as medio-lateral expansion of BODIPYmuscimol.

Statistics
GraphPad Prism 9 (GraphPad, La Jolla, CA, USA) was used for statistical evaluations. Normally distributed data from the acute experiments (ivPTZ-ST test, body weight, and body temperature) were compared using one-way repeated measures (RM) analysis of variance (ANOVA) followed by a Sidak correction for multiple comparisons. Normally distributed data from the chronic experiments (ivPTZ-ST test, body weight, and body temperature) were compared using two-way mixed-effects model analysis with the Geisser-Greenhouse correction followed by a Sidak correction for multiple comparisons. The comparison between pre-and post-surgical seizure thresholds was done using a paired t-test.
Nonparametric statistics were used in case of ordinal data or if data did not follow a normal distribution. These data were analyzed using the Friedman test for one-way RM ANOVA by ranks, with Dunn's correction for post hoc multiple comparisons (acute experiments: behavioral parameters; chronic experiments: within-treatments comparisons between time points for behavioral parameters) or the Kruskal-Wallis test for oneway ANOVA by ranks, with Dunn's correction for post hoc multiple comparisons (chronic experiments: between-treatment comparisons within time points for behavioral parameters). Responder rate changes compared to vehicle were evaluated using the Fishers exact test with Bonferroni correction. All data were tested two-sided and the level of significance was set to α = 0.05.

Cannula localization (acute experiments)
Eleven rats used in a crossover design fulfilled the inclusion criterion of bilateral cannula placement within the STN and were used for final data evaluation. In the remaining animals, cannulas were located either unilaterally outside the STN (n = 3) or bilaterally outside the STN (n = 2), with misplaced cannula tips located in the zona incerta or internal capsule.
Although not systemically evaluated, the histological examinations of the Nissl-stained sections did not indicate any drug-induced lesions of the STN or regions adjacent to the microinjection sites apart from the tracks left by the cannulas.

Lack of surgical influence on seizure thresholds
The surgical procedure of guide cannula implantation did not alter PTZ-induced seizure thresholds (not illustrated; mean±SEM myoclonic seizure threshold was 16.87 ± 1.03 mg/kg before and 15.03 ± 0.67 mg/kg after surgery, paired t-test p = 0.08, sample size 9 animals; clonic seizure threshold was 21.00 ± 1.04 mg/kg before and 20.57 ± 0.68 mg/kg after surgery, paired t-test p = 0.63, sample size 9 animals).

Antiseizure effects (acute experiments)
Acute bilateral microinjection of muscimol into the STN induced significant antiseizure effects in the ivPTZ-ST test ( Fig. 2; sample size 11 animals). Compared to mean basal (predrug) seizure thresholds measured before microinjection, muscimol (but not aCSF as vehicle) significantly raised both the threshold for myoclonic ( Fig. 2A) and clonic seizures (Fig. 2B) when injected in amounts of 30 ng and 60 ng muscimol per hemisphere (p-values are given in the figure legend). Relative to basal seizure thresholds, the mean myoclonic seizure threshold was lowered by 11% (vehicle) and raised by 31% (30 ng muscimol) and 52% (60 ng muscimol). The mean clonus seizure threshold was raised by 5%, 39% and 59% for vehicle, 30, and 60 ng muscimol, respectively. Accordingly, mean myoclonic and clonic seizure thresholds differed significantly between vehicle-and muscimol-treated animals ( Fig. 2A,   B). The seizure threshold elevations induced by muscimol did not differ significantly between the two dosages (30 ng and 60 ng; Fig. 2A, B). However, as described in the method section, our dose finding pilot studies indicated a no-observed-effect-level at 10 ng per hemisphere and a stronger antiseizure effect with 120 ng per hemisphere, the latter of Likewise, muscimol significantly raised thresholds for the first clonic seizure (B) when compared to basal predrug threshold (##p = 0.0020 for 30 ng per side and ##p = 0.0016 for 60 ng per side) and when compared to aCSF (**p = 0.0064 for 30 ng per side and **p = 0.0010 for 60 ng per side; one-way RM-ANOVA with Sidak's multiple comparisons test, F(2.616, 26.16) = 20.71). Individual rat responses to muscimol and aCSF microinjections compared to basal thresholds for myoclonic twitch (C) and clonic seizure (D) reflect clinical observations of a heterogeneous drug responsiveness in individual patients. Drug responders were defined as showing at least 25% increase in the threshold for induction of either the myoclonic twitch (E) or the clonic seizure (F) compared to basal thresholds. With 60 ng muscimol per hemisphere, all but one animal responded to intrasubthalamic muscimol with an increase in the clonic seizure threshold. Responder rates are given below the pie charts. The strengths of antiseizure effects are color-coded in the pie charts and show a dosedependent increase, with one animal even showing a myoclonic (E) and clonic (F) seizure threshold increase of more than 100%. which was associated with unacceptable adverse effects.

Responder rate (acute experiments)
Because previous experiments with another drug (vigabatrin) have shown that not all rats respond with clear antiseizure effects to intrasubthalamic delivery at tolerable drug doses (Gey et al., 2016), we conducted an intra-individual response analysis between basal seizure thresholds and seizure thresholds after acute substance microinjection. As shown in Fig. 2C-F and Table 2, we found a dose-dependent increase in the responder rate, i.e., the proportion of rats showing at least 25% increase in ivPTZ-induced seizure thresholds in response to treatment compared to the individual basal (predrug) seizure threshold.
Concerning myoclonic twitches, none of the animals responded to intrasubthalamic microinjection of vehicle (Fig. 2C, E and Table 2). In contrast, 73% of the rats were defined as responders after 30 ng as well as after 60 ng muscimol per hemisphere (Fig. 2C, E and Table 2). The difference in responder rate between vehicle-and muscimol-treated animals was statistically significant (Table 2). Most animals (6/11) were found to be consistent responders, i.e., they responded to both the low and the high muscimol dose, while two rats only responded to the high doses, and one animal consistently did not respond at all. Only two rats responded to the low dose but not to the high dose and therefore are considered as instable responders.
Noteworthy is that although the responder:nonresponder ratio did not differ between the two muscimol doses (30 and 60 ng), the portion of responders, in which the drug showed a stronger effect, was larger with the higher muscimol dose of 60 ng (Fig. 2E). Thus, while with 30 ng muscimol, only 1 of the 8 responders showed a more than 50% increase in myoclonic twitch threshold, 5 of the 8 responders showed such an increase after 60 ng muscimol, and 1 animal even showed a more than 100% increase in myoclonic twitch threshold (Fig. 2E).
Concerning the clonic seizure thresholds (Fig. 2D, F and Table 2), the lower muscimol dose of 30 ng also increased the responder rate compared to vehicle, but the difference failed to be statistically significant (Table 2). In contrast, a significantly higher responder rate (91%) was observed after 60 ng muscimol compared to vehicle-treated animals, but not when compared to the low 30 ng dose of muscimol ( Table 2). The one animal that did not respond to the high dose (Table 2), also did not respond to the low dose, i.e., was a consistent nonresponder.
Again, the portion of responders, in which the drug showed a stronger effect, increased with increasing drug doses. While with vehicle only 1 of the 3 responders showed a more than 50% increase in clonic seizure threshold, 3 of the 6 responders showed such an increase after 30 ng muscimol. With 60 ng muscimol, 5 of the 10 responders showed a more than 50% increase, and 1 animal even showed a more than 100% increase in clonic seizure threshold compared to individual basal seizure threshold (Fig. 2F).
The responder rate is given in [%]. The numbers of animals are given in parenthesis (responders/total number of rats). While muscimol significantly raised the responder rate when compared to aCSF treatment, the two muscimol doses did not differ significantly from each other. *p < 0.05 compared to animals treated with artificial cerebrospinal fluid (aCSF), Fishers exact test with Bonferroni correction.

Adverse effects (acute experiments)
In general, acute bilateral microinjection of muscimol into the STN was well tolerated. Equivocal or moderate circling behavior, probably due to slightly asymmetrical delivery into the two brain sides, occurred in most animals in response to muscimol, being significant following microinjection of 60 ng per hemisphere compared to basal behavior and compared to vehicle control, respectively (Fig. 3A). In addition, a treatment effect was observed by the nonparametric repeated measures ANOVA by ranks (Friedman test) for the readout 'flat body posture' and for the rotarod test (Fig. 3A), but the post hoc test failed to reveal significant effects compared to controls. Preliminary investigation of the long-term development of these adverse effects in 3 rats in response to 60 ng muscimol indicated a behavioral normalization 4 h after intrasubthalamic microinjection (not illustrated). Concerning the other examined behavioral parameters, individual rats showed no, equivocal, or moderate behavior signs, but this occurred independent of the treatment group (Fig. 3A). Thus, none of these other behavioral readouts were significantly affected by acute intrasubthalamic muscimol delivery.
In addition, body weight and body temperature were neither significantly affected by acute muscimol microinjection nor by acute vehicle microinjections, when compared to basal (predrug) values (RM ANOVA, body temperature p = 0.073, body weight p = 0.519; not illustrated).

BODIPY-muscimol distribution (acute experiments)
Three of the initially used 6 animals were evaluable for acute BODIPY-muscimol distribution, while in the remaining animals technical problems impeded any evaluation. Fluorophore-conjugated muscimol showed an asymmetric spread around the injection site over a mean distance of 2.29 ± 0.07 mm and 3.12 ± 0.11 mm in the anteriorposterior and medio-lateral plane, respectively (sample size 3 animals). High-intensity fluorescence of BODIPY-muscimol was mainly observed in the STN (Fig. 4), but also detected in adjacent regions, e.g., the substantia nigra pars reticulata (SNr) and zona incerta. Dorsal spread was observed along the cannula shaft. Ventral distribution was naturally prevented by the fiber tract of the internal capsule.

Cannula localization (chronic experiments)
Just as with the acute experiments, chronic intrasubthalamic microinfusion of muscimol for three weeks did not induce any obvious lesions of the STN or regions adjacent to the microinfusion sites as verified by histological examination of Nissl-stained sections.
Twenty-five rats used in a parallel group design fulfilled the inclusion criterion of bilateral cannula placement within the STN (7 rats received aCSF, 5 rats received 60 ng, 9 rats received 300 ng, and 4 rats received 600 ng muscimol per day and hemisphere). In the remaining animals, cannulas were located either unilaterally outside the STN (n = 3) or bilaterally outside the STN (n = 3), with misplaced cannula tips located in the cerebral peduncle or internal capsule. Two further rats had to be excluded from final data evaluation because of histological signs of surgery-induced intracranial bleeding.

Antiseizure effects (chronic experiments)
While chronic delivery of vehicle into STN did not alter ivPTZ-ST, continuous bilateral infusion of muscimol elevated the threshold for both myoclonic twitches (Fig. 5A) and clonic seizures (Fig. 6A) at certain doses. When compared to basal seizure thresholds, the mean seizure threshold significantly increased for two weeks during a daily dose of 300 ng muscimol per hemisphere (myoclonic twitches, Fig. 5A), and for one week during a daily dose of 300 ng and 600 ng muscimol per hemisphere (clonic seizures, Fig. 6A). Thereafter, mean seizure threshold changes failed to differ significantly from basal seizure thresholds, indicating development of tolerance to the antiseizure effects of muscimol in at least some animals (see below). The low daily dose of 60 ng muscimol per hemisphere (as well as vehicle) did not significantly alter mean thresholds for either induction of myoclonic Table 2 Responder rates in response to acute intrasubthalamic muscimol delivery. aCSF 30 ng muscimol 60 ng muscimol Myoclonus 0% (0/11) 73% (8/11)* 73% (8/11)* Clonus 27% (3/11) 55% (6/11) 91% (10/11)* twitches or clonic seizures when compared to basal seizure thresholds ( Fig. 5A and 6A). However, with all investigated muscimol doses, subgroups of responders and nonresponders could be separated (see below). Relative to basal seizure thresholds, the maximum increases in mean seizure threshold were observed during the first week of chronic muscimol delivery. At this time-point, mean threshold for induction of myoclonic twitches was lowered by 4% (vehicle) and raised by 23% (60 ng muscimol daily), 37% (300 ng muscimol daily), and 20% (600 ng muscimol daily). The mean clonic seizure threshold was raised by 12% (vehicle), 20% (60 ng muscimol daily), 47% (300 ng muscimol daily), and 45% (600 ng muscimol daily).
Different from the acute microinjection experiments, in which the crossover design allowed comparisons between vehicle and muscimol treatment within the same animals, the chronic CED experiments could only be realized in a parallel group (between-subject) design, because the rats received a microinfusion pump containing either vehicle or a certain muscimol dose. One-way mixed-effects model analysis with posthoc Sidak correction revealed a significantly lower mean basal (predrug) Fig. 3. Heatmap illustrating the occurrence and score of behavioral adverse effects in individual rats in response to vehicle (artificial cerebrospinal fluid, aCSF) and different doses of muscimol delivered acutely (A) and chronically (B) to the subthalamic nucleus. Each column represents an individual rat. In the acute drug delivery study (A), 60 ng muscimol per hemisphere caused significant circling behavior, when compared to basal ("B") predrug values and aCSF (p = 0.0016; indicated as red frame). This circling behavior was probably due to an irregular drug delivery into the two brain sides (see 3.2.1 and 4.1). In addition, an effect of treatment was indicated by RM ANOVA on ranks for "flat body posture" (p = 0.0156) and for the performance on the rotarod (p = 0.0179), but the posthoc test failed to reveal significant differences (p > 0.05 all comparisons). In the chronic drug delivery study (B), a daily dose of 300 ng muscimol per hemisphere caused significant ataxia after two weeks (w 2) of intrasubthalamic drug delivery, when compared to basal predrug values (p = 0.0243) and aCSF (p = 0.0046; indicated as red frame). Further behavioral effects were observed in individual rats, but failed to result in significant group changes (p > 0.05 all comparisons). threshold for induction of myoclonic twitches in the 60 ng muscimol group (13.92 ± 0.58 mg/kg) compared to the vehicle group (17.68 ± 0.63 mg/kg, p = 0.004; not indicated in Fig. 5). Thus, we considered these interindividual basal group differences by calculating the ivPTZ-ST as % change to basal seizure threshold for each animal at each time-point (i.e., data normalization) to allow an appropriate statistical comparison of vehicle-and drug-treated animals (Figs. 5B and 6B). Compared to chronic vehicle infusion, a daily dose of 300 ng muscimol caused a significant mean seizure threshold elevation (% change to basal) over the complete testing period of three weeks (myoclonic twitch, Fig. 5B) and during the first week (clonic seizure, Fig. 6B). The high daily dose of 600 ng muscimol also caused mean seizure threshold elevations compared to vehicle infusions, but these differences (probably due to the low sample size) were significant only during the first week (clonic seizure, Fig. 6B). Identical to the comparisons with basal seizure thresholds, the low daily dose of 60 ng failed to significantly change mean relative seizure thresholds compared to the respective vehicle data (Figs. 5B and 6B). To follow a single animal over time, the within-subject time courses are illustrated in Fig. 5 C and 6 C.

Responder rate and development of tolerance (chronic experiments)
Identical to the acute experiments, we conducted an intra-individual response analysis between basal seizure thresholds and seizure thresholds during chronic CED of muscimol into the STN. Most animals defined as responders, i.e., with a seizure threshold increase of at least 25%, showed this increase during the first week of chronic muscimol delivery (myoclonus, 11/14 rats; clonus 14/15 rats), while few rats showed a delayed response (myoclonus, 3/14 rats; clonus 1/15 rats). Comparable to the findings from the acute experiments, we typically found a higher responder rate at higher muscimol doses (Figs. 5D, 6D, and Table 3).
With regard to both myoclonic twitches (Fig. 5D) and clonic seizures (Fig. 6D), the responder rate after 300 ng muscimol per day and hemisphere differed significantly from vehicle-treated rats and reached 89% for both seizure types (Table 3). Concerning clonic seizures (Fig. 6D), all animals additionally responded to the high daily dose of 600 ng per hemisphere, the responder rate thus differing significantly from vehicle controls (Table 3).
However, individual responders developed tolerance over time. The portion of animals developing tolerance to the muscimol effects on the seizure thresholds increased with increasing drug doses (Figs. 5E and 6E). Noteworthy is that not only was the daily dose of 300 ng muscimol per hemisphere a promising dose concerning the responder rate (Table 3), it was also the dose with the strongest antiseizure effect. Thus, during daily 300 ng muscimol, 3 of 8 responders showed at least 50% increase in myoclonic twitch threshold and one animal even showed a response greater than 100% above predrug control (Fig. 5E). In contrast, only one responder showed more than 50% increase in myoclonic twitch threshold during 60 ng and during 600 ng muscimol daily, respectively. The same applies to the clonic seizure threshold. The antiseizure effects were more pronounced in the 300 ng muscimol-group, since 6 of 8 responders increased the clonic seizure threshold by more than 50% above predrug threshold, while only one of 2 responders in the 60 ng muscimol-group reached this level of effect (Fig. 6E). With the high daily dose of 600 ng muscimol, 3 of 4 responders showed at least 50% increase in clonic seizure threshold, but 3 (not the same) of the 4 rats developed tolerance over time ( Fig. 6E and Table 3).
Responder rate is given in [%]. The numbers of animals are given in parenthesis (responders/total number of rats). While continuous microinfusion of muscimol significantly raised the responder rate when compared to aCSF treatment (myoclonus and clonus by 300 ng and clonus by 600 ng), the three muscimol doses did not differ significantly from each other. The numbers of responders showing loss of effect over time ("tolerance", TOL) is given in square brackets. *p < 0.05 compared to animals treated with artificial cerebrospinal fluid (aCSF), Fishers exact test with Bonferroni correction.

Adverse effects (chronic experiments)
Comparable to what we observed during the acute experiments, chronic bilateral CED of muscimol into the STN was generally welltolerated (Fig. 3B). The only significant finding was a higher incidence of equivocal or moderate ataxia two weeks following chronic microinfusion of muscimol at a daily dose of 300 ng per hemisphere into the STN, when compared to basal control values and when compared to two weeks aCSF microinfusion (Fig. 3B). In addition, the high daily dose of 600 ng muscimol increased the risk for showing ataxia as adverse effect, but probably due to the low sample size in that group failed to differ significantly from basal control.
At the daily doses of 300 ng and 600 ng muscimol, individual animals showed circling behavior, likely caused by irregular delivery of muscimol into the two brain sides (Fig. 3B). Furthermore, similar to the findings from the acute experiments, individual rats completely failed to pass the rotarod test following chronic muscimol microinfusion (300 ng and 600 ng daily) into the STN (Fig. 3B). Other motor signs, such as hypermetric gait, were only equivocally or moderately expressed in few muscimol animals, while other readouts were seen in few animals from all groups including vehicle animals (Fig. 3B). The occurrence of wet dog shakes in some animals from both vehicle and muscimol groups was attributed to the surgery and the suture in the neck. Some readouts such as head swaying and pick-up test were not affected by chronic intrasubthalamic delivery (Fig. 3B).
Mean body temperature (in • C) was reduced in all chronic treatment groups post-surgery compared to basal (predrug) values, being significant at daily muscimol doses of 60 ng (Sidak multiple comparisons, p = 0.009 after one week, p = 0.016 after two weeks, p = 0.008 after three weeks) and 300 ng (p = 0.031 after one week, p = 0.004 after two weeks; not illustrated). Likewise, body weight (in g) was reduced in all chronic treatment groups post-surgery compared to basal values, being significant at daily muscimol doses of 300 ng (p = 0.023 one week postsurgery) and 600 ng (p = 0.031 one week post-surgery). However, comparison of normalized data (calculated as % change to basal for each animal at each time-point, as was the case for the seizure thresholds in the chronic study) revealed that body temperature did not differ significantly between vehicle and muscimol-treated animals, regardless of time-point and muscimol dose (2-way ANOVA, p = 0.155 for Fig. 4. Representative images of the intracerebral distribution of borondipyrromethene (BODIPY)-conjugated muscimol along the rostrocaudal extend following acute microinjection (upper row) and chronic microinfusion (lower row) into the subthalamic nucleus (STN). The yellow fluorescence of BODIPY-muscimol is overlaid with a brightfield image (acute experiment) or a gray fluorescent image under 385 nm excitation (chronic experiment) of the same coronal section. After acute microinjection, high-intensity fluorescence of BODIPY-muscimol was mainly observed in the STN, but also detected in adjacent regions. Limited dorsal spread was observed along the cannula shaft. Ventral distribution was naturally prevented by the fiber tract of the internal capsule. In chronically microinfused animals, high-intensity fluorescence was observed around the cannula tip targeting the STN. In contrast to acute experiments, distribution of continuously infused BODIPY-muscimol exceeded the borders of the STN in a more heterogeneous fashion (for more details see 3.2.4). Capillary effects along the guiding cannulas supported dorsal spread resulting in an ellipsoidal distribution at the cannula tract. Numbers in sections indicate distance relative to bregma according to (Paxinos and Watson, 2014). Scale bar, 1 mm.
(caption on next page) M. Gernert et al. treatment factor). Body weight (in % change to basal) was significantly lower only one week after intrasubthalamic CED of muscimol at the high daily dose of 600 ng (mean±SEM, − 1.26 ± 1.24% in vehicle group versus − 12.17 ± 2.30% in 600 ng muscimol group, Sidak multiple comparisons, p = 0.028) and did not differ from vehicle-treated rats at later time-points. With 600 ng daily, not only body weight was clearly reduced within the first week CED of muscimol, but also a post-operative sedation for a couple of days became evident, the reason why we did not increase the sample size in this high dose group.

BODIPY-muscimol distribution (chronic experiments)
All 4 animals used for this part of the study were evaluable for chronic BODIPY-muscimol distribution. In chronically microinfused animals, high-intensity fluorescence of BODIPY-muscimol was observed around the cannula tip targeting the STN (Fig. 4), while lower signals were detected in more distant areas. Maximum distribution of BODIPYmuscimol exceeded the borders of the STN, showing an average spread of 2.70 ± 0.84 mm along the anterior-posterior axis and 1.25 ± 0.34 mm along the medio-lateral axis (sample size 4 animals). In contrast to acute experiments, BODIPY-muscimol was more heterogeneously distributed across the animals, which might be due to several reasons including interindividual differences in scar formation around the cannulas over the weeks. Capillary effects along the guiding cannulas supported dorsal spread yielding an ellipsoidal distribution at the cannula tract (Fig. 4).

Discussion
This is the first study investigating chronic CED of muscimol into the STN aiming to induce antiseizure effects. The main findings of the present experiments were: (a) Acute intrasubthalamic muscimol delivery caused significant antiseizure effects at 30 and 60 ng per hemisphere (b) with a dose-dependent increase of responders and efficacy, and (c) mild adverse effects. (d) Chronic CED of muscimol into the STN caused significant antiseizure effects for up to three weeks at a daily dose of 300 and 600 ng per hemisphere with (e) a dose-dependent increase in responder rate, (f) but also loss of efficacy over time in part of the responders, (g) and transient ataxia and body weight loss. (h) The distribution of BODIPY-muscimol was comparable (about 2-3 mm) between acute and chronic delivery.

Antiseizure effects, adverse effects, and responder rate (acute experiments)
Several animal experiments using different seizure and epilepsy models provided evidence for an antiseizure action of muscimol microinjected acutely into or close to an experimental seizure focus (Baptiste et al., 2010;Collins, 1980;Kohane et al., 2002;Ludvig et al., 2009;Piredda and Gale, 1985). In a comparative study of epidural cup delivery of different drugs at a fixed concentration of 1 mM aiming to prevent disturbance of physiological osmolarity of CSF and brain extracellular fluid, it was shown that only muscimol (5.7 µg in 50 µl), but not lidocaine, midazolam, pentobarbital, and GABA, exerted antiseizure effects in a rat model of acetylcholine-induced neocortical seizures (Baptiste et al., 2010). However, targeting the seizure focus is not always an option, for example in cases with multiple seizure foci or unknown focus localization (Khoo et al., 2021).
Basal ganglia regions such as the STN and the SNr have been preclinically investigated repeatedly over the past four decades (Iadarola and Gale, 1982) and proven to be highly promising target regions not involved in seizure initiation but being part of the epileptic network and thereby involved in seizure propagation and pronounced remote modulation of seizure-initiating areas (reviewed by Barcia and Gallego, 2009;Deransart and Depaulis, 2002;Gale et al., 2008;Gernert and Feja, 2020;Nilsen and Cock, 2004;Velísková and Moshé, 2006;Vuong and Devergnas, 2018). Direct comparison studies using the ASM vigabatrin (Bröer et al., 2012) or grafting inhibitory cells (Backofen-Wehrhahn et al., 2018) revealed that targeting the STN is more promising to suppress seizures and is better tolerated than targeting the SNr. Although not directly compared, several previous microinjection studies indicate that this might also apply to muscimol, i.e., stronger antiseizure effects combined with better tolerability when the STN instead of the SNr is targeted (Dybdal and Gale, 2000;Gernert and Löscher, 2001). Indeed, muscimol delivery into the SNr caused locomotor activation (Dybdal and Gale, 2000), and choreiform dyskinetic movements were observed after muscimol microinjection into the SNr, but not into the STN of nonhuman primates (Dybdal et al., 2013).
In addition to the antiseizure group effect, we evaluated the responder rate of acute intrasubthalamic muscimol delivery. At the high dose of 60 ng per hemisphere, the responder rate reached 73% concerning myoclonic twitch and even 91% regarding clonic seizures. Future studies have to clarify, if such high responder rates can also be achieved in models of drug-resistant epilepsies. Heiss et al. (2019b) determined the safety and behavioral effects of CED of muscimol injected over 10 min bilaterally into the STN of nonhuman primates and showed that muscimol was not toxic to primate brain tissue. Likewise, we did not observe toxic effects on rodent brain Fig. 5. Antiseizure effects and responder rates for the first myoclonic twitch after chronic bilateral convection-enhanced delivery of muscimol at different doses into the subthalamic nucleus. Basal ("B", predrug) seizure thresholds and vehicle (artificial cerebrospinal fluid, aCSF) infusions served as controls. Thresholds for induction of myoclonic twitches as determined by intravenous pentylenetetrazole (PTZ) infusion are shown as single values and means+SEM (A, B). Chronic muscimol delivery at a daily dose of 300 ng per hemisphere significantly raised myoclonic twitch thresholds for two weeks when compared to basal predrug threshold (A; twoway mixed-effects model analysis with the Geisser-Greenhouse correction followed by Sidak's multiple comparisons test, main effect of treatment F(3, 21) = 3.586 and week F(2.34, 47.57) = 3.725, #p = 0.0456 after one week [1 w] and #p = 0.0234 after two weeks [2 w]), and even for three weeks when compared to aCSF (B; two-way mixed-effects model analysis with the Geisser-Greenhouse correction followed by Sidak's multiple comparisons test, main effect of treatment F(3, 21) = 5.378, *p = 0.0410 after one week, * *p = 0.0064 after two weeks, and *p = 0.0370 after three weeks [3 w]). In (B), seizure thresholds are shown as % change to basal threshold for each animal at each time-point (i.e., data normalization to consider interindividual basal group differences) to allow an appropriate statistical comparison of vehicle-and drug-treated animals. In (C), the same data as in (B) are illustrated as within-subject time courses to follow a single animal over time. The proportion of drug responders, defined as showing at least 25% increase in seizure threshold compared to basal thresholds, is shown as pie charts (D). Responder rates are given below the pie charts. With a daily dose of 300 ng per hemisphere, all but one animal responded to the intrasubthalamic convection-enhanced muscimol delivery with a seizure threshold increase. The proportion of responders developing tolerance to the antiseizure effect over time increased with increasing muscimol doses (see tolerance rates below pie chartes). The strengths of antiseizure effects (maximal response observed during the three weeks) are color-coded in the lower pie charts (E). With a daily intrasubthalamic muscimol dose of 300 ng per hemisphere, half of the responders showed a threshold increase of more than 50% compared to individual basal thresholds, and one animal even showed a seizure threshold increase of more than 100% (E).
(caption on next page) tissue with the muscimol doses used. Heiss et al. (2019b) further showed dose-dependent motor adverse effects including hyperkinesia in response to bilateral intrasubthalamic CED of muscimol in nonhuman primates. The circling behavior we observed in most rats in response to acute muscimol delivery bilaterally into the STN is in line with the described postural asymmetry observed by (Dybdal and Gale, 2000) after unilateral microinjection of muscimol into the STN of rats and, in our present study, is probably due to irregular delivery into the two brain sides. However, bilateral intrasubthalamic vigabatrin delivery did not cause circling behavior in female rats (Bröer et al., 2012). These drug-dependent differences might be explained by the different mechanisms of action of the two drugs, i.e., continuous and nonphysiological stimulation of GABA A receptors by muscimol and a rather physiological modulation by blocking GABA degradation in active GABAergic synapses by vigabatrin.

Antiseizure effects, adverse effects, responder rate, and tolerance (chronic experiments)
For long-term seizure suppression, chronic drug delivery approaches are inevitable. To prolong antiseizure effects, intracranial delivery of different drugs by means of implantable osmotic pumps, polymer carriers, and long-term delivery by pump-driven CED has been investigated pre-clinically (reviewed by Gernert and Feja, 2020). However, while numerous previous animal studies showed promising antiseizure effects by acute intracerebral muscimol microinjections, the concept of chronic, continuous intracerebral muscimol delivery over several weeks as in our present study has not been investigated before.
In a first attempt addressing this concept, Kohane et al. (2002) directly compared acute intrafocal muscimol microinjection (5 µg) with placement of lipid-protein-sugar microparticles containing 5 µg muscimol into the hippocampus 80 min before the end of intrahippocampal pilocarpine infusion used to induce acute seizures in rats. The results indicated that encapsulation enhanced the protective antiseizure effect of muscimol and may have maintained an effective concentration for a longer time. As a proof-of-concept, we now showed that prolonged CED of muscimol into the STN over three weeks via implanted microinfusion pumps is able to induce long-lasting antiseizure effects in rats.
Although clinical findings support the concept of seizure-controlling properties of the basal ganglia (Rektor et al., 2012;Vuong and Devergnas, 2018), intrasubthalamic drug delivery has not yet been investigated clinically in patients with systemic drug-resistant epilepsy. In their clinical proof-of-principle study, Heiss et al. (2019a) showed that CED of muscimol into the seizure focus for 12-24 h was able to reduce seizure frequency in a patient with systemic drug-resistant neocortical epilepsy, but not in two patients with hippocampal epilepsy. Our findings indicate that targeting the STN might be advantageous in this respect, because the STN is known to be nonselective as to the type of seizure or seizure origin it can influence, i.e., a highly attractive feature with regard to translation into clinical application.
In line with this, we previously showed that chronic intrasubthalamic CED of the ASM vigabatrin, which irreversibly inhibits the GABA-degrading enzyme GABA-transaminase (GABA-T), has significant antiseizure effects with only mild adverse effects (Gey et al., 2016). However, similar to our present findings with muscimol, not all rats responded to chronic intrasubthalamic delivery of vigabatrin, and tolerance developed over time in half of the responders (Gey et al., 2016). In the present study, a daily dose of 300 ng muscimol infused chronically into the STN yielded a high responder rate of 89%, and mean antiseizure effects were maintained over three weeks, but nevertheless, about half of the rats developed tolerance to the antiseizure effect over time. This adaptation to prolonged agonist exposure might be due to desensitization, down-regulation, and/or allosteric uncoupling of postsynaptic receptors (Barnes, 1996;Jankovic et al., 2021;Roca et al., 1990) within the STN. Furthermore, loss of efficacy might be caused by an up-regulation of glutamate receptors in downstream neurons of the SNr, as has been described two weeks after lesioning of the STN in rats (Price et al., 1993). This would lead to an enhanced sensibility of SNr neurons to the remaining glutamatergic excitatory influence. Further network changes caused by adaptive reorganization processes may additionally counteract the targeted STN inhibition, e.g., delayed changes in firing rates of downstream superior colliculus neurons (Bressand et al., 2002).
Subsequent studies will have to show if discontinuous (intermittent) muscimol delivery into the STN by means of programmable microinfusion pumps or by on-demand drug delivery using closed-loop systems will prevent development of tolerance to the antiseizure effects. A study by Tang et al. (2011) indicated that such a discontinuous drug delivery approach might be appropriate, because periodic transmeningeal (epidural) muscimol (5.7 µg in 50 µl; once per day for four consecutive days in each week) applications over three weeks to a neocortical epileptogenic zone in rats did not induce tolerance to the antiseizure effects.
The selection of responders and nonresponders reflects the clinical observation of a heterogeneous drug responsiveness in individual patients. It is not yet known, if there is an overlap between systemic and intracerebral drug responsiveness in individual rats. Likewise, it is not yet known, which factors contribute to the heterogeneity in Fig. 6. Antiseizure effects and responder rates for the first clonic seizure after chronic bilateral convection-enhanced delivery of muscimol at different doses into the subthalamic nucleus. Basal ("B", predrug) seizure thresholds and vehicle (artificial cerebrospinal fluid, aCSF) infusions served as controls. Thresholds for induction of clonic seizures as determined by intravenous pentylenetetrazole (PTZ) infusion are shown as single values and means+SEM (A, B). Chronic muscimol delivery at daily doses of 300 ng and 600 ng per hemisphere significantly raised clonic seizure thresholds for one week when compared to basal predrug threshold (A; two-way mixed-effects model analysis with the Geisser-Greenhouse correction followed by Sidak's multiple comparisons test, main effect of treatment F(3, 21) = 3.379 and week F(1.995, 40.57) = 10.44, ##p = 0.0014 for 300 ng and ##p = 0.0097 for 600 ng) and when compared to aCSF (B; two-way mixed-effects model analysis with the Geisser-Greenhouse correction followed by Sidak's multiple comparisons test, main effect of treatment F(3, 21) = 3.125 and week F(1.424, 28.48) = 4.157, *p = 0.0117 for 300 ng and *p = 0.0122 for 600 ng). In (B), seizure thresholds are shown as % change to basal threshold for each animal at each time-point (i.e., data normalization to consider interindividual basal group differences) to allow an appropriate statistical comparison of vehicle-and drug-treated animals. In (C), the same data as in (B) are illustrated as within-subject time courses to follow a single animal over time. The proportion of drug responders, defined as showing at least 25% increase in seizure threshold compared to basal thresholds, is shown as pie charts (D). Responder rates (given below the pie charts) increased with increasing muscimol dose. With a daily dose of 600 ng per hemisphere, all rats responded to the intrasubthalamic convection-enhanced muscimol delivery with a seizure threshold increase. However, the proportion of responders developing tolerance to the antiseizure effect over time also increased with increasing muscimol doses (see tolerance rates below pie charts). The strengths of antiseizure effects (maximal response observed during the three weeks) are color-coded in the lower pie charts (E). The proportion of responders, which showed a threshold increase of more than 50% compared to individual basal thresholds, was higher with higher muscimol doses (E). responsiveness observed in our present study. Our data may indicate that the strategy of targeted intrasubthalamic drug delivery may not work for all individuals. Future studies have to clarify, if significant antiseizure effects in a subgroup of individuals (responders) can also be achieved in models of drug-resistant epilepsies and/or models that do not show a clear dependence on impaired GABAergic function. In this respect it is noteworthy that our previous direct comparison study (Gey et al., 2016) indicates that antiseizure effects observed in the PTZ seizure threshold model can well be transferred to the electrical kindling model of temporal lobe epilepsies, i.e., a model of difficult-to-treat types of seizures.

Functional concepts
The exact mechanisms underlying the antiseizure effects of STN manipulations are not fully understood. The high responder rates we observed upon acute and chronic muscimol delivery into the STN are likely related to its local cellular effects resulting in long-range network activity changes without exerting alterations in global brain function. This network-specific modulation is able to prevent adverse effects observed after systemic drug treatments. Muscimol has a 10-fold greater affinity than GABA for the GABA A receptor and selectively and potently inhibits neural activity (Baufreton et al., 2001;Ebert et al., 1999;Enna et al., 1977). Because GABA is the primary inhibitory neurotransmitter in the brain and is present in up to 70% of all brain synapses, a regionally selective neuronal suppression by CED of muscimol into the STN avoids systemic and reduces neurological adverse effects.
The STN is composed of glutamatergic projection neurons and receives strong GABAergic input from the external globus pallidus and fast cortical glutamatergic input via the hyperdirect pathway. The STN in turn monosynaptically excites downstream neurons in several areas including the SNr, which is composed of GABAergic projection neurons and shows secondarily decreased activity by intrasubthalamic muscimol (Feger and Robledo, 1991).
Numerous studies suggest that among other mechanisms, direct or indirect (via STN) inhibition of GABAergic SNr neurons provides disinhibition of downstream antiseizure zones within the dorsal midbrain (Deransart and Depaulis, 2002;Gale, 1992;Gale et al., 2008;Gernert and Feja, 2020;Redgrave et al., 2010;Soper et al., 2016). Furthermore, the STN is not only highly interconnected with basal ganglia and cortical regions, but also modulates the activity of limbic seizure circuits (via internal globus pallidus and lateral habenula), thalamic regions, and of brainstem nuclei such as the pedunculopontine nucleus (Bonnevie and Zaghloul, 2019;Hamani et al., 2004). Thus, the STN has a key position within the epileptic network and seems to utilize multiple pathways in addition to simply modulating nigral activity during seizure suppression . This is supported by the more pronounced antiseizure effects induced by intrasubthalamic compared to intranigral vigabatrin delivery (Bröer et al., 2012). However, it should be noted that the microinfused muscimol was not restricted to the STN but also reached surrounding tissue including the SNr and thalamic areas. Thus, direct effects on regions surrounding the STN most probably contribute to the observed antiseizure effects.
Likely because of its widespread connectivity within motor and limbic basal ganglia and cortical circuitry (Bonnevie and Zaghloul, 2019;Hamani et al., 2004), the STN is impressively nonselective as to the seizure types it can influence and has the attractive ability to modulate seizures with different origins.

Myoclonic versus clonic seizure modulation
The ivPTZ-ST test is a sensitive model, which allows to determine the effect of drugs on different components of seizure behavior (Mandhane et al., 2007). In general, our experiments showed that in the ivPTZ-ST test, both myoclonic and clonic seizure thresholds were comparably sensitive to modulation by intrasubthalamic muscimol delivery, with a slightly more favorable sensitivity of myoclonic twitches concerning long-term antiseizure effects of intrasubthalamic muscimol. Nevertheless, our data emphasize the ability of the STN for nonselective seizure modulation. However, apart from putative differences that may arise from using male or female animals, this is different from previous findings in female rats showing that threshold increases for clonic seizures were clearly more marked than respective increases for myoclonic twitches in response to therapeutic modulations targeting the STN. These previous studies targeting the STN in the ivPTZ-ST test included acute (Bröer et al., 2012) and chronic vigabatrin microinfusions (Gey et al., 2016), and transplantations of genetically engineered GABA-releasing cells (Handreck et al., 2014) and inhibitory interneuron precursor cells (Backofen-Wehrhahn et al., 2018). The reasons underlying this discrepancy between previous studies and the current study are unknown, but may be related to the specific treatment type and/or the exact location of the targeted treatment within the STN. It is rather unlikely that it is due to the different distribution ranges of the therapeutic agent, because a rather wide-spread distribution was found by Gey et al. (2016) and Backofen-Wehrhahn et al. (2018), while a more restricted distribution was described by Handreck et al. (2014) and in the present study.

Drug distribution
With CED, a drug is pumped through the extracellular space of the brain parenchyma and is distributed by a combination of bulk flow and diffusion . Compared to a purely diffusion-based distribution along a concentration gradient, CED results in a more even and widespread drug distribution along a pressure gradient (Heiss et al., 2005;Lonser et al., 2015;Rogawski, 2009). Accordingly, we previously showed that chronic intrasubthalamic delivery of vigabatrin in effective seizure-suppressing amounts was associated with long-distance drug distribution with corresponding irreversible GABA-T-inhibition over a wide range of the rat brain (Gey et al., 2016). However, targeted pharmacotherapy requires a more restricted drug distribution . This is challenging in rats, because of the small STN volume of about 0.8 mm 3 bilateral (Hardman et al., 2002). Compared to our previous vigabatrin study (Gey et al., 2016), we now achieved a drug delivery more limited to the STN, although we observed a dorsal spread along the cannulas upon chronic muscimol delivery (and to a lesser extent following acute microinfusion) and part of the drug exceeded the STN borders, albeit in lower concentrations as indicated by the BODIPY-muscimol data. It therefore can be assumed that even low concentrations of muscimol in regions adjacent to the surgically targeted STN contributed to the antiseizure (and adverse) effects. However, a substantial part of the observed antiseizure effects is likely due to STN inhibition, as indicated by a neural transplantation study showing antiseizure effects only when grafts of GABA-releasing neurons were located in the STN, but not in adjacent regions outside the STN (Handreck et al., 2014). In addition, Dybdal and Gale (2000) showed that the anticonvulsant effects of muscimol in STN were site-specific, in as much as bilateral muscimol injections, positioned 2 mm dorsal to the STN in rats, did not result in seizure protection in their experiments. Factors affecting muscimol distribution include binding to GABA A receptors and other structures, uptake into cells, drug metabolism, and capillary permeability (Heiss et al., 2005). Heiss et al., (2019bHeiss et al., ( , 2010 showed that co-infusion of surrogate imaging tracers and 3 H-muscimol allows monitoring of muscimol distribution by MRI and autoradiography during and after CED in the brain of nunhuman primates. They further showed a linear relationship between volume of muscimol infusion and volume of distribution in the nonhuman primate brain and a relatively homogenous "square-shaped" drug distribution profile (Heiss et al., 2010). One could argue that due to the larger molecular weight of BODIPY-muscimol used in our study, the distribution range of pure muscimol might be broader than that of BODIPY-muscimol. However, while passive diffusion range depends on several drug-related characteristics including molecular weight, the distribution range achieved by CED using a pressure gradient is determined by other variables including tissue characteristics, cannula size, and volume of infusion . Thus, CED is suitable not only for smaller molecules, but also for slowly distributing drugs of high molecular weight (Bobo et al., 1994;Lonser et al., 2015).
Our BODIPY-muscimol experiments provide a first approximation of the muscimol distribution after CED into the STN. Thus, we cannot exclude that with the used method, we did not capture the full extent of muscimol spread after chronic delivery. Nevertheless, the detection of intense BODIPY fluorescence in the target area upon completion of the experiment further substantiates chronic intrasubthalamic CED of muscimol as a reliable concept for prolonged antiseizure effects.

Conclusions
We provided proof-of-concept that not only acute, but also chronic, continuous CED of muscimol into the STN raises thresholds for different seizure types. Muscimol is an attractive compound for such a targeted approach, because muscimol acts as a specific and reversible agonist at GABA A receptors and acute muscimol infusions were delivered safely to the STN of nonhuman primates (Heiss et al., 2019b) and to the STN of humans with Parkinson's disease (Levy et al., 2001). Our data strongly supports the concept of the STN as a key basal ganglia structure exerting a widespread influence on seizure circuitry through numerous other brain regions. We showed that antiseizure responder rates were increased at higher muscimol doses. Because tolerance rate also increased with increasing muscimol doses, the study emphasizes the importance of additional preclinical studies including intermittent instead of continuous dosing regimens to further develop the intracerebral drug delivery approach for epilepsies and further neurological diseases. Thus, adjusting the drugs and drug application regimes implemented may allow for sustained antiseizure effects in a chronic treatment plan.

Declaration of Competing Interest
None