Long-lasting antiseizure effects of chronic intrasubthalamic convection-enhanced delivery of valproate

Intracerebral drug delivery is an experimental approach for the treatment of drug-resistant epilepsies that allows for pharmacological intervention in targeted brain regions. Previous studies have shown that targeted pharmacological inhibition of the subthalamic nucleus (STN) via modulators of the GABAergic system produces antiseizure effects. However, with chronic treatment, antiseizure effects are lost as tolerance develops. Here, we report that chronic intrasubthalamic microinfusion of valproate (VPA), an antiseizure medication known for its wide range of mechanisms of action, can produce long-lasting antiseizure effects over three weeks in rats. In the intravenous pentylenetetrazole seizure-threshold test, seizure thresholds were determined before and during chronic VPA application (480 μ g/d, 720 μ g/d, 960 μ g/d) to the bilateral STN. Results indicate a dose-dependent variation in VPA-induced antiseizure effects with mean increases in seizure threshold of up to 33%, and individual increases of up to 150%. The lowest VPA dose showed a complete lack of tolerance development with long-lasting antiseizure effects. Behavioral testing with all doses revealed few, acceptable adverse effects. VPA concentrations were high in STN and low in plasma and liver. In vitro electrophysiology with bath applied VPA revealed a reduction in spontaneous firing rate, increased background membrane potential, decreased input resistance and a significant reduction in peak NMDA, but not AMPA, receptor currents in STN neurons. Our results suggest an advantage of VPA over purely GABAergic modulators in preventing tolerance development with chronic intrasubthalamic drug delivery and provide first mechanistic insights in intracerebral pharmaco-therapy targeting the STN.


Introduction
Systemically applied drugs currently used to suppress seizures in epilepsies are associated with two critical challenges: systemic/neurological adverse effects are common and about 30% of patients do not become seizure-free due to drug resistance (Janmohamed et al., 2020).Targeted intracerebral drug delivery is an experimental strategy of interest for its potential to circumvent drug resistance in epilepsies by bypassing the blood-brain barrier and its associated resistance mechanisms (Gernert and Feja, 2020;Löscher et al., 2020;Rogawski, 2009).Moreover, this approach achieves high drug concentrations in key brain regions of the epileptic network while maintaining low systemic concentrations, thereby reducing the risk of systemic and neurological adverse effects (Gernert and Feja, 2020;Rogawski, 2009).
Due to its role in remote modulation of seizure initiation and propagation, the subthalamic nucleus (STN) has previously been investigated in rodents as a target for intracerebral drug delivery in epilepsies.Acute (Bröer et al., 2012;Deransart et al., 1998;Deransart et al., 1996;Dybdal and Gale, 2000;Velı́šková et al., 1996) and chronic (Gernert et al., 2023;Gey et al., 2016) pharmacological inhibition of the STN via the γ-aminobutyric acid (GABA) subtype A receptor agonist, muscimol, and the GABA transaminase inhibitor, vigabatrin, revealed antiseizure effects in rodents.Chronic intrasubthalamic (intra-STN) delivery of such pure GABAergic drugs, however, has been associated with a loss of effect within 2 weeks, likely due to the development of tolerance to antiseizure effects (Gernert et al., 2023;Gey et al., 2016) via several mechanisms including GABA A receptor downregulation after prolonged occupancy (Barnes Jr., 1996).In fact, tolerance to antiseizure (and adverse) effects is known for most (systemically given) antiseizure medications (ASMs) during repeated or long-term exposure (review by Löscher and Schmidt, 2006).However, for some ASMs, including valproate (valproic acid, VPA), the experimental evidence for tolerance to antiseizure effects are debated, as it seems to have a rather complex pattern of tolerance development (Davis et al., 1994;Löscher et al., 1989;Löscher and Hönack, 1995;Löscher and Nau, 1982;Serralta et al., 2006).By selecting VPA, a drug with multiple mechanisms of action not exclusive to the GABAergic system, we aim to circumvent target-specific mechanisms of tolerance development and thereby reduce, or ideally eliminate, the development of tolerance with chronic intra-STN drug delivery.
Preclinical (Altenmüller et al., 2013;López et al., 2007;Rassner et al., 2015;Serralta et al., 2006) and clinical attempts (Cook et al., 2020) have been made to deliver VPA over longer time periods into intracranial compartments, targeting the seizure focus or lateral ventricle.These studies focused on rather short-term antiseizure effects, dose finding and safety issues, or the development of long-term release from polymer matrices.Here we investigated the antiseizure effects of VPA delivered chronically into the STN by an implantable pump/catheter system in a rat seizure model aiming to confirm an advantage of the multitarget drug over pure GABAergic drugs regarding responder rates and circumventing tolerance development over a three-week period.Behavioral testing was conducted to assess putative adverse effects of intra-STN VPA.VPA levels of plasma, liver, STN, and other brain regions were evaluated.Finally, we performed slice-electrophysiology experiments to understand VPA effects at the level of the STN, aiming to deduce a functional concept.

Animals
Wistar Unilever rats (HsdCpb:WU) were purchased from Envigo (Venray, Netherlands).Adult female animals were used in in vivo experiments to ensure better comparability to previous studies on targeted STN modulation (Gernert et al., 2023;Gey et al., 2016;Handreck et al., 2014).For in vitro electrophysiology experiments, litters of both female and male pups (P16-42) were used.
All animals were provided with nest material, wooden chew sticks, and a red transparent tube (BIOSCAPE/Ebeco, Castrop-Rauxel, Germany).Animals were housed in groups under controlled environmental conditions (room temperature: 22-24 • C; room humidity: 50-60%) with a 12 h light/dark cycle (lights on at 6 a.m.).Standard laboratory chow (Type: 1324, Altromin Spezialfutter GmbH & Co. KG, Lage, Germany) and tap water were provided ad libitum.Following pump implantation surgery, animals were housed individually to facilitate incision healing and prevent mutual gnawing at sutures.

Ethics statement
All animal experiments were conducted in compliance with the German Animal Welfare Act and the European Union (EU) council directive 2010/63/EU.They were approved by the Internal Animal Care and Use Committee (TiHo-T-2018-23) and 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, and 20/3399).Surgeries were conducted under general anesthesia combined with local anesthesia and perioperative analgesia.All efforts were made to minimize animal suffering.

Systemic study
To verify and quantify the antiseizure effects of VPA in the timed intravenous PTZ seizure threshold (ivPTZ-ST) test in our female rats, we first performed a systemic VPA study with 10 animals.Three days after basal (pre-drug control) seizure threshold and behavioral testing, rats entered the VPA testing phase of the study.Each animal received an intraperitoneal (i.p.) injection of VPA (200 mg/kg and 300 mg/kg) and sodium chloride (NaCl, vehicle control) following a crossover Latinsquare design (Fig. 1A) with a washout period of one week between consecutive injections.Doses were selected based on previous studies showing elevated seizure thresholds in kindled rats (Löscher et al., 1993;Töllner et al., 2011) and in the ivPTZ-ST test in rats (Löscher and Hönack, 1995).Behavioral and seizure threshold testing was conducted 15 and 30 min after each injection, respectively.Several weeks after the conclusion of the systemic dose response study, animals were decapitated for analysis of VPA levels in brain, plasma, and liver 30 min after a final i.p. injection of VPA (200 mg/kg or 300 mg/kg) or NaCl (Fig. 1A).

Chronic intrasubthalamic microinfusion
To assess the antiseizure effects of chronic intra-STN VPA, a parallel, between-group design was used (Fig. 1B).Seizure threshold and behavioral testing was conducted for each animal at least 2 days before surgical pump and cannula implantation (basal threshold; pre-drug control), and again once per week for three weeks starting at least days post implantation surgery.Starting at the time of pump implantation, VPA (or vehicle; artificial cerebrospinal fluid, aCSF) was continuously infused bilaterally into the STN (480 μg/d, 720 μg/d, or 960 μg/d) for the duration of the chronic study.Behavioral testing was conducted one hour before each seizure threshold determination.Following the final seizure threshold determination, the microinfusion pumps were explanted, and animals were transcardially perfused for histological analysis of cannula localization or decapitated for HPLC analyses.Final group sizes included in the statistical analyses were as follows.Vehicle: n = 10, 480 μg/d VPA: n = 6, 720 μg/d VPA: n = 6, 960 μg/d VPA: n = 9.

Electrophysiology
For the electrophysiological investigation of the effect of VPA on neurons of the STN, three Wistar Unilever (HsdCpb:WU) mother rats, each with a litter of 8-10 pups, were purchased from Envigo (Venray, Netherlands).Pups of either sex were decapitated for slice preparation at postnatal day P17-28, as described below (Electrophysiology, brain slice preparation).
Current clamp experiments with VPA, and the corresponding control experiments without VPA, were conducted in a total of 38 cells from animals (4 male, 5 female; 2-7 cells per animal) across all three litters.Here, amplitude and firing rates of STN neurons were measured before and after a hyperpolarization stimulus, both before and after VPA bath application (2.5 mmol/L).Membrane potential and input resistance were also measured before and after VPA bath application.
Voltage clamp experiments in which peak NMDA and AMPA receptor currents were measured, were conducted in 5 cells from 3 animals originating from two litters (all male; 1-3 cells per animal).Behavior and intravenous pentylenetetrazole (ivPTZ) seizure threshold testing was conducted 15 and 30 min post-injection, respectively.Several weeks after conclusion of the systemic study, rats received a final i.p. injection of NaCl or VPA (200 or 300 mg/kg) and were decapitated for analysis of VPA levels in brain, liver, and plasma via higher performance liquid chromatography (HPLC).B) Chronic intra-subthalamic nucleus (intra-STN) VPA: Following basal testing, rats in the chronic intra-STN study underwent surgical implantation of microinfusion pumps and bilateral intra-STN cannulas for the continuous convection-enhanced delivery of VPA (480, 720, or 960 μg/d) or vehicle (artificial cerebrospinal fluid, aCSF).Thereafter, behavior and ivPTZ seizure threshold testing was conducted once weekly for three weeks in a parallel group design.Following the last behavior and ivPTZ-ST test, animals were either decapitated for analysis of VPA levels in brain, liver, and plasma via HPLC, or transcardially perfused with formaldehyde for histological evaluation of cannula localization.C) Slice electrophysiology: For in vitro slice electrophysiology, rat brains were resected directly following decapitation and 250 μm thick horizontal brain slices were prepared.Subsequent voltage and current clamp recordings were made before (control) and after bath application of VPA (2.5 mmol/L).Coronal and sagittal diagrams from Paxinos and Watson (1998).Created with BioRender.com.

Timed intravenous PTZ seizure threshold test
For the assessment of the antiseizure effects of systemic or chronic intra-STN VPA, the timed ivPTZ-ST test was conducted as previously described (Feja et al., 2021;Gernert et al., 2023).In short, 0.8% PTZ was continuously infused into the lateral tail vein of the conscious and freely moving animal and the time to the first myoclonic twitch and subsequent clonic seizure was recorded.Seizure thresholds were calculated in mg PTZ per kg body weight: (1) threshold to first myoclonic twitch, and (2) threshold to clonic seizure.PTZ infusion was stopped at the onset of clonic seizure or at a maximum PTZ infusion time of 90 s.All animals were subsequently monitored until fully recovered.The experimenter was blinded to the treatment group at the time of ivPTZ-ST testing.
Seizure thresholds were additionally expressed in percent change to basal threshold to account for interindividual differences, and therefore allow for comparison between treatment groups in the chronic intra-STN experiments.
Based on previous ivPTZ-ST studies conducted with vigabatrin (Gey et al., 2016), OV329, a novel GABA transaminase inhibitor (Feja et al., 2021), and muscimol (Gernert et al., 2023), animals were categorized as responders or non-responders.Responders were defined as animals demonstrating an increase in seizure threshold of at least 25% in at least one ivPTZ-ST test during VPA treatment compared to basal control (Feja et al., 2021;Gernert et al., 2023).In the chronic intra-STN VPA study, animals that no longer exhibited an increase in seizure threshold of at least 25% compared to basal in the final ivPTZ-ST test despite an initial response, were considered as having developed tolerance.

Behavioral test battery
Potential adverse effects of systemic or intra-STN VPA treatment were assessed using a behavioral test battery that included a modified Irwin screen, hyperexcitability test, and rota-rod performance test, as previously described (Gernert et al., 2023).Briefly, in a round, black open field (diameter: 78 cm), the behavioral parameters assessed in the modified Irwin screen and hyperexcitability test (see Fig. 5A-B) were scored on a scale of 0-3, with zero being the physiological score and three being the most severe score.To ensure unbiased scoring, videos of the rats' behavior were made during the Irwin screen and scored by blinded observers.Rota-rod performance was scored based on the rats' ability to stay on the rotating rod (8 rpm) for 1 min.The resulting score reflects the number of failed trials, with a maximum of 3 attempted trials.Body weight and temperature were measured directly following the behavioral test battery, and animals were returned to their home cage before seizure threshold testing.

Chronic convection-enhanced microinfusion of valproate
As previously described in detail (Gernert et al., 2023;Gey et al., 2016), a programmable implantable microinfusion pump (iPRECIO®, Model SMP-200, Primetech Corporation, Tokyo, Japan) for small laboratory animals was used for intra-STN VPA delivery.Pumps were filled with the respective solutions and maintained at 37 • C in an air incubator for at least 24 h prior to implantation.Directly before implantation, each pump was programmed to run with a continuous flow rate of 0.2 μL/h for the duration of the experiment (three weeks).Pump implantation was conducted under general anesthesia with isoflurane combined with local anesthesia and perioperative analgesia as previously described (Gernert et al., 2023).At the conclusion of the three-week testing period, microinfusion pumps were explanted under isoflurane and animals were perfused for cannula localization or decapitated under isoflurane for HPLC analysis of VPA concentrations in brain, plasma, and liver.

Histology
For the histological evaluation of cannula localization, animals were perfused transcardially with 200 mL of 0.01 M phosphate-buffered saline followed by 250 mL of 4% formaldehyde in 0.1 M phosphate buffer.Brains were removed and stored in a 30% sucrose solution in 0.1 M phosphate buffer at 4 • C before sectioning.On a freezing microtome (Leica Microsystems GmbH, Wetzlar, Germany), 40 μm coronal sections were cut and subsequently Nissl-stained with thionine.Sections were scanned via bright field microscopy (Leica Microsystems, Wetzlar, Germany) with a 25-fold magnification for STN and cannula localization.Only animals found to have cannulas located bilaterally in the STN were included in statistical analyses.

High performance liquid chromatography (HPLC)
For analysis of VPA levels in plasma, liver, and brain, animals were decapitated under isoflurane.The brains were resected, flash frozen in liquid nitrogen, and transferred to a cold plate (Lauda Dr. R. Wobser GmbH & Co. KG, Lauda-Königshofen, Germany) for dissection.The STN, the area dorsal to the STN, substantia nigra pars reticulata (SNr), hippocampus, motor cortex, and anterior striatum were isolated, weighed, and homogenized in 0.5 mL of 80% ethanol.Liver samples were collected directly following decapitation and homogenized in distilled water on ice.For every 200 μL of brain or liver homogenate, 200 μL of ethanol and 100 μL of enanthic acid were added and samples were centrifuged first for 20 min at 15000 rpm and 4 • C, and subsequently for 30 min at 6000 rpm with a 0.2 μm filter.Blood samples were collected from the site of decapitation and centrifuged at 12000 rpm for 2.5 min with 10 μL ethylenediaminetetraacetic acid (EDTA) for every 0.5 mL collected.The resulting plasma supernatant was combined with methanol (150 μL methanol for every 50 μL of plasma) and centrifuged for min at 13200 rpm.The resulting supernatant was collected and stored at − 30 • C until HPLC analysis.
HPLC was performed using an ultraviolet detector (SPD-20 A/20AV UV-ViS Detektor, Shimadzu Deutschland GmbH, Duisburg, Germany) with a detection wavelength of 210 nm.For the stationary phase, a Nucleosil® 120-5 C18 (250 mm × 4.6 mm) reverse phase column (Macherey-Nagel GmbH & Co. KG, Düren, Germany) maintained at 35 • C was used.The mobile phase was composed of phosphate buffer and acetonitrile (65:35) adjusted to a pH of 5.6.Analysis duration was 14 min with a retention time of 10 min.

In vitro recordings
STN neurons were visualized with a 60 × 1 NA objective on a BX51WI microscope (Evident Corporation, Tokyo, Japan) equipped with gradient contrast illumination, and a TILL Photonics imaging system (FEI, Hillsboro, OR, USA) composed of a Retiga 2000 DC camera and a monochromator (Polychrome V).Recordings were performed using an EPC 10/2 amplifier (HEKA, Multi Channel Systems GmbH, Reutlingen, Germany).The pipette resistance ranged between 2.8 and MΩ.Access resistances between 4.5 and 14 MΩ were accepted for analysis.For voltage-clamp recordings, the access resistance was compensated to a residual of 3 MΩ.In current-clamp recordings, the bridge balance was set to 100% after estimating the series resistance in the voltage-clamp, which was monitored repeatedly during recordings.For voltage-clamp experiments, the internal recording solution consisted of (in mmol/L): 135 Cs-gluconate, 10 HEPES, 20 tetraethylammonium chloride, 3.3 MgCl 2 , 2 Na 2 -ATP, 0.3 Na 2 -GTP, 3 Na 2phosphocreatine, 5 CsEGTA, 10 μmol/L ZD7288 and, in some cases, 50 mol/L AlexaFluor-568 or AlexaFluor-594 (pH adjusted with CsOH to 7.4, calculated liquid junction potential: 12.7 mV).For current-clamp experiments, the internal recording solution consisted of (in mmol/L): 145 K-gluconate, 15 HEPES, 4.5 KCl, 7 Na 2 -phosphocreatine, 2 Mg-ATP, 2 K 2 -ATP, 0.3 Na 2 -GTP, and 0.5 K-EGTA, resulting in a liquid junction potential of 15.95 mV.No correction of the liquid junction potential was performed.During voltage-and current-clamp recordings, VPA was bath-applied at 2.5 mmol/L, matching the lower range of concentrations described in in vitro experiments (Armand et al., 1995;Fueta and Avoli, 1991;Zeise et al., 1991).
Voltage responses were recorded in sweeps of 5 s, consisting of a 2 s period without stimulation before and after a 1 s -50 pA hyperpolarizing stimulation.The sweep interval was 15 s and 30 repetitions were recorded.The data from the first (control) and after wash-in the last (VPA data) five sweeps of the 30 repetitions were analyzed and statistically compared (n = 33 cells).The firing rates were calculated in the first and fourth second from a sweep.The background membrane potential was defined as the average voltage value of the first second in a sweep.The input resistance (R in ) was determined from the voltage difference between the background and the voltage recorded at 200 ms during the − 50 pA hyperpolarization.In a set of control experiment (the same current-clamp protocol pattern), the same recording time was used, but without VPA bath application (n = 5 cells).
Excitatory synaptic currents were evoked by local stimulation of either medial or lateral fibers with a glass electrode (4-5 MΩ) filled with recording solution.A 200 μs long biphasic voltage pulse was delivered by an AM2100 stimulator to activate afferent inputs.During stimulation, cells were held at +40 and -70 mV to record NMDAR and α-amino-3hydroxy-5-methyl-4-isoxazolepropionic acid receptor (AMPAR) currents, respectively.Stimulus interval between NMDAR and AMPAR currents was 5 s.After the first ten NMDAR and AMPAR mediated EPSC stimulations (control data) VPA was bath-applied.The current amplitude of ten NMDAR and AMPAR mediated EPSC (VPA data) recorded after about 6-8 min of VPA wash in was chosen for statistical comparison.Excitation was pharmacologically isolated by blocking inhibitory synaptic currents with 0.5 μmol/L strychnine hydrochloride and 10 mmol/L SR 95531 hydrobromide during voltage-clamp recordings.

Drugs
For i.p. injection of VPA a 100 mg/mL injection solution (Orfiril®; Desitin Arzneimittel GmbH, Hamburg, Germany) was used.Dosing was achieved by varying injection volumes of 2 mL/kg or 3 mL/kg for a dose of 200 mL/kg or 300 mL/kg, respectively.Isotonic 0.9% NaCl injection solution (3 mL/kg; B. Braun Deutschland GmbH & Co. KG, Melsungen, Germany) was used as vehicle control.
For chronic intra-STN microinfusion, VPA sodium salt (Sigma-Aldrich Chemie GmbH, Taufkirchen, Germany) was dissolved in water for injection (B.Braun Deutschland GmbH & Co. KG, Melsungen, Germany).Freebase solutions of 200, 300, and 400 μg/μL VPA were freshly prepared and titrated to a physiological pH of 7.3 before being loaded into the appropriate microinfusion pumps.Applied with a continuous infusion rate of 0.2 μL/h, this corresponds to a daily dose of 480, 720, and 960 μg per hemisphere, respectively.At the conclusion of the chronic microinfusion experiments, the solutions from all pumps as well as aliquots of the stock solutions were collected.Osmolality of the solutions was determined via freezing point osmometer (Osmomat 3000, Gonotec, Berlin, Germany) and the concentration of VPA solutions was confirmed via HPLC.
For electrophysiology, the same VPA (Sigma-Aldrich Chemie GmbH, Taufkirchen, Germany) as was used for the chronic microinfusion study was dissolved in distilled water for a concentration of 2.5 mmol/L.

Statistical analysis
Statistical analyses for in vivo studies were conducted using Graph-Pad Prism® version 9.3.1 (GraphPad Software Inc., San Diego, CA, USA).Data collected during the electrophysiology study were analyzed using Igor Pro (Wavemetrics, Portland, OR, USA).Significance was tested two-sided with the level of significance set to α = 0.05, or with a linear correlation test implemented in IgorPro9.

Intravenous PTZ seizure threshold testing
The ivPTZ-ST data were analyzed using a mixed-effects analysis of variance (ANOVA) with a Geisser-Greenhouse correction and post-hoc Šídák's multiple comparisons test with individual variances computed for each comparison.In the systemic VPA study, each group was compared to all other groups in the post-hoc Šídák's multiple comparisons test.The proportion of animals protected from clonic seizure was compared between the acute systemic VPA dosing groups and vehicle control group via Fisher's exact tests, as was the proportion of responders to acute systemic VPA.
Due to the parallel group design required for the chronic intra-STN study, seizure threshold data expressed in mg/kg were compared only to their own basal control.For comparison of the chronic intra-STN VPA treatment groups to the vehicle (aCSF) control group, seizure thresholds were additionally expressed in percent change to basal threshold to account for interindividual differences, and therefore allow for comparison of thresholds between groups within each time point.
In both the acute systemic and chronic intra-STN VPA studies, in the case of a paravenous PTZ injection, the ivPTZ-ST data for that time point was excluded from the analysis; however, the corresponding behavioral test battery data conducted before PTZ injection was not.For the chronic intra-STN VPA study, animals were additionally excluded from statistical analyses in the case of off-target cannula localization (unilaterally off target: n = 4, bilaterally off target: n = 8), or in the case of technical issues, i.e., with the infusion pump/catheter system (n = 3), or in HPLC measurements (n = 7).One animal in the vehicle control group was excluded due to a unilateral lesion of the STN identified in the final histological evaluation.Thirty-one rats met all inclusion criteria and were included in the final ivPTZ-ST analyses.

Behavioral test battery
Scored behavioral parameters were analyzed using Friedman oneway repeated measures ANOVA for the systemic study and for intragroup analyses between time points in the chronic intra-STN study.As the Friedman test does not allow for missing values, in the case of a single missing data point, the whole data set for that animal was excluded from the Friedman test analysis.A Kruskal-Wallis test was conducted for intergroup/intra-timepoint analyses in the chronic intra-STN study.In this test, missing values are tolerated, so that single data points were excluded from the analysis when necessary.Dunn's multiple comparisons test served as post-hoc analysis.Multiple comparisons were made between all groups of the systemic VPA study, and either to basal controls within each group (Friedman test) or to vehicle control within each time point (Kruskal-Wallis test) of the intra-STN study.
Bodyweight and temperature were analyzed via repeated measures two-way ANOVA with Šídák's multiple comparisons test with a single pooled variance for post-hoc analysis.A post-hoc Spearman correlation test was conducted to evaluate a potential correlation between change in body temperature from basal and score for hypolocomotion in the Irwin screen.

Acute systemic valproate injection
Although VPA has long been in clinical use as a broad-spectrum ASM and its antiseizure effects are well established, the majority of the existing literature on the effects of VPA in rat models of PTZ-induced seizures mainly focuses on bolus i.p. or s.c.injections of PTZ (Löscher, 1999).We were therefore first interested in verifying the appropriateness of the ivPTZ-ST model for identifying changes in seizure thresholds induced by VPA.
With the results of our systemic study, we were able to reproduce antiseizure effects of systemic VPA previously described for rats (Löscher, 1989;Löscher, 1999;Löscher and Hönack, 1995;Serralta et al., 2006;Tulloch et al., 1982) and verify the ability of the ivPTZ-ST test to detect VPA-induced changes in seizure thresholds in female Wistar Unilever rats.Both myoclonic twitch and clonic seizure were clearly affected by acute, systemic VPA injection at both doses tested, but not by vehicle (0.9% NaCl).VPA treatment significantly raised seizure thresholds to first myoclonic twitch compared to both basal Fig. 2. Effects of acute intraperitoneal (i.p.) injection of valproate (VPA) on the intravenous pentylenetetrazole seizure thresholds (ivPTZ-ST) in 10 rats.Vehicle (0.9% NaCl) injections served as control.Seizure thresholds are shown in mg/kg PTZ required to induce (A) the first myoclonic twitch and (B) subsequent clonic seizure.Shown are means + SEM as well as individual values for each animal before starting in the crossover study (basal, pre-drug control; "B") and after each i.p. injection ("V" = vehicle, 200 = 200 mg/kg VPA, 300 = 300 mg/kg VPA), with post-hoc multiple comparisons between each group.Mean seizure thresholds were dose-dependently elevated in response to VPA compared to both basal thresholds and vehicle controls, while no significant change was seen between control groups.⊗ indicates rats completely protected by VPA from myoclonic twitch (A) or clonic seizure (B) up to the maximum PTZ infusion time of 90 s.Variance in sample size between time points within treatment groups is a result of paravenous PTZ injections leading to single data points removed from final analyses.Of the 10 animals tested, 8 were completely protected from clonic seizures.Responder rates and magnitude of responses to vehicle ("V"), 200 mg/kg VPA ("200"), and 300 mg/kg VPA ("300") are shown for myoclonic twitch (C) and clonic seizure (D).All rats responded to VPA with an elevation of clonic seizure thresholds.Interindividual differences in response magnitude were observed independent of tissue VPA levels (see Fig. 6).Mixed-effects one-way ANOVA with Geisser-Greenhouse correction and Šídák's multiple comparisons for post-hoc analysis; **p < 0.01, ***p < 0.005, ****p < 0.0001.
Clonic seizure thresholds were increased by 77% and 103% with 200 and 300 mg/kg VPA, respectively.Eight of 10 rats were even completely protected from clonic seizure with 300 mg/kg VPA, which was a significantly greater proportion than with vehicle (p = 0.0007) or 200 mg/kg (p = 0.0055).Again, these changes in seizure threshold were significant compared to basal (F 2.117, 18.35 = 94.35,p < 0.0001; 200 mg/ kg and 300 mg/kg: p < 0.0001; Fig. 2B) and vehicle controls (200 mg/kg and 300 mg/kg: p < 0.0001; Fig. 2B).Despite 300 mg/kg VPA having induced a larger mean increase in clonic seizure threshold compared to 200 mg/kg VPA, this difference did not reach statistical significance.It is important to note, however, that this is likely due to the maximal PTZ infusion time of 90 s we established for our PTZ experiments, and which was reached by a large proportion of animals as mentioned above.Animals showing an increase in seizure threshold of at least 25% from basal were considered responders, and responder rates were generated for each treatment (vehicle, 200 mg/kg, and 300 mg/kg VPA) and seizure type.In line with previous pharmacological studies in the ivPTZ-ST model using different seizure-suppressing substances (Bröer et al., 2012;Gernert et al., 2023;Gey et al., 2016), myoclonic and clonic seizure endpoints were differently affected with typically more pronounced effects on clonic seizures.While responder rates for myoclonic twitch ranged from 10% following vehicle injection up to 80% with 200 mg/kg VPA and 90% after 300 mg/kg VPA (Fig. 2C), all animals showed a response in clonic seizure threshold following both VPA doses tested (100% responder rate; Fig. 2D).The proportion of responders to acute hemisphere) or vehicle (artificial cerebrospinal fluid, aCSF; week 1, "W1"; week 2, "W2"; week 3, "W3") is compared to the basal (pre-drug; "B") control within the respective group.Variance in sample size between time points within treatment groups is a result of paravenous PTZ injections leading to single data points removed from final analyses.An initial non-specific increase in clonic seizure threshold after one week of continuous microinfusion was seen with both VPA and vehicle.Longlasting significant increases in clonic seizure thresholds were seen with all VPA doses tested, but not with vehicle.For direct comparison to the vehicle control group (aCSF), ivPTZ seizure thresholds (ivPTZ-ST) were additionally calculated in percent change to basal (mean + SEM; B), revealing a significant elevation of the clonic seizure threshold after three weeks of continuous VPA delivery of 480 μg/d and 960 μg/d.The defined responder threshold of 25% increase from basal is illustrated by a dashed line "R" on the y-axis.C) Responder rates and tolerance rates (a loss of response by the final ivPTZ-ST test) are given for each VPA dose and aCSF control.Rats were additionally categorized by the magnitude of the maximal response measured during chronic VPA or aCSF microinfusion compared to basal control (D).The lowest tolerance rates were observed with 480 μg/d, while the highest responder rates and strongest effect in individual rats was observed with a VPA dose of 960 μg/d.Mixed-effects model (two-way ANOVA) with Geisser-Greenhouse correction and Šídák's multiple comparison test for post-hoc analysis with individual variances computed for each comparison: *p < 0.05, **p < 0.01.systemic VPA differed significantly from vehicle for both myoclonic twitch (200 mg/kg: p = 0.0055; 300 mg/kg: p = 0.0011) and clonic seizure (200 mg/kg: p = 0.0001; 300 mg/kg: p = 0.0001), but not between doses.
Magnitudes of response for both myoclonic twitch and clonic seizure were higher with systemic VPA (Fig. 2C, D) than with intra-STN VPA (Figs. 3D and 4D), with at least 50% of animals responding with a >100% increase in seizure threshold after highest systemic VPA dose tested, but only up to 22% of animals reaching this magnitude after chronic intra-STN delivery of the highest dose tested.
The highest systemic VPA dose, 300 mg/kg, was associated with significant ataxia (300 mg/kg: p = 0.0060) and increased occurrence of wet dog shakes (p = 0.0335) compared to basal and vehicle controls.This is in line with previous observations on systemic VPA treatment in rats (Serralta et al., 2006;Töllner et al., 2011).An effect of treatment on hypolocomotion (p = 0.0074) and rota-rod performance (p = 0.0030) was indicated, however, post-hoc multiple comparisons did not reach statistical significance (Fig. 5).Acute systemic VPA did not significantly affect any of the remaining behavioral parameters evaluated (hyperlocomotion, head swaying, circling, stereotyped sniffing, hypermetric gate, flat body posture, touch response, pick-up response), nor did it significantly affect animals' body weight or temperature (p > 0.05).The observed ataxia was more pronounced and occurred in more animals after systemic VPA treatment than after intra-STN delivery (Fig. 5 and see text below).However, it is important to consider that in addition to the varying routes of application, the acute systemic and chronic intra-STN experiments additionally differed in temporal drug exposure (i.e., acute, once weekly i.p. injections vs. chronic, continuous intra-STN microinfusion).Indeed, adverse effects may decrease with chronic (or repeated) systemic application (Löscher and Fiedler, 2000).Thus, direct comparison between the effects of the two VPA application methods is limited.
For the direct comparison between vehicle and VPA doses, investigated in a parallel group design, seizure thresholds were calculated in percent change to basal.Compared to vehicle controls, bilateral daily VPA doses of 480 μg (p = 0.0103) and 960 μg (p = 0.0166) significantly increased clonic seizure thresholds in the third week of continuous VPA delivery (F 3, 27 = 4.34, p = 0.0128; Fig. 4B).Myoclonic twitch thresholds were significantly increased in response to 480 μg/d VPA (p = 0.0004) in the third week (F 3, 27 = 6.584, p = 0.0017; Fig. 3B), while none of the other doses altered myoclonic twitch thresholds, when compared to vehicle controls.
Maximum mean increases in clonic seizure thresholds were seen in the first week of aCSF and 960 μg/d VPA microinfusion, and in the second week with 480 μg/d and 720 μg/d VPA (Fig. 4B).Although the effect of VPA on thresholds to myoclonic twitch was generally less pronounced, the maximum mean increase in thresholds to myoclonic twitch were in the third week with 480 μg/d and in the second week with 960 μg/d VPA microinfusion.In comparison, mean thresholds were not increased with aCSF or 720 μg/d microinfusion (Fig. 3B).

Responder rates varied with dose and seizure type
Previous experiments with intra-STN microinfusion of vigabatrin (Gey et al., 2016) and muscimol (Gernert et al., 2023) suggested that not all animals responded equally to chronic intra-STN VPA delivery.While the responder rate increased from 40% (vehicle) to 89% (highest VPA dose) with regard to the clonic seizure threshold (Fig. 4C), the response to intra-STN VPA seems to be more complex concerning the myoclonic twitch threshold.Here, the highest responder rate (83%) was reached with the lowest VPA dose (Fig. 3C), while no robust responses were reached with the medium VPA dose, and 56% responder rate with the highest VPA dose.It can be speculated that this might be due to different mechanisms of action being involved at different STN concentrations of VPA (Biggs et al., 1992;Johannessen and Johannessen, 2003).Despite differences in responder rates, the magnitude of response within each group was found to be comparable between seizure types, with VPAinduced seizure threshold changes only exceeding a 100% increase with the highest VPA dose tested (960 μg/d; Fig. 3D and 4D).

No loss of effect with low VPA dose
In addition to high responder rates, tolerance rates as low as 0% were observed.Loss of effect could be completely avoided with 480 μg/d VPA, as no animals showed tolerance development for first myoclonic twitch or clonic seizure (Fig. 3C and 4C).In contrast, the 720 μg/d VPA group clearly induced tolerance development in almost all responders (Fig. 3C   and 4C).The highest dose of 960 μg/d yielded comparatively lower tolerance rates of 20% and 38% for myoclonic twitch and clonic seizure, respectively, indicating a non-linear relationship between VPA dose and tolerance development (Fig. 3C and 4C).With the present study we found an intra-STN dose, which is able to produce antiseizure effects without inducing tolerance development within the three weeks tested.

Chronic convection-enhanced microinfusion of valproate was well tolerated
Chronic intra-STN VPA and aCSF delivery were generally well tolerated.Chronic intra-STN VPA was mainly associated with hypolocomotion and (in some cases) flat body posture and wet dog shakes (Fig. 5), the latter of which also being typical for systemic VPA administrations as described above.However, wet dog shakes after acute systemic VPA administration are thought to be due to an effect on the serotonergic system (Löscher et al., 1988), while with intra-STN delivery we assume that it could also be due to the surgery and the suture in the neck, since it was recently also observed with aCSF infusion (Gernert et al., 2023).
In our present study, intra-group comparisons to basal controls indicated an effect of aCSF on wet dog shakes (p = 0.0293) and of 480 μg/d VPA on hypolocomotion (p = 0.0052), however post-hoc multiple comparisons did not reach statistical significance (p > 0.05).Compared to aCSF control, daily intra-STN delivery of 480 μg VPA significantly increased hypolocomotion in the first week (p = 0.0024) and flat body posture in the third week (p = 0.0257) of chronic microinfusion.Wet dog shakes occurred in single rats independent of vehicle or drug treatment and reached statistical significance with 720 μg/d VPA in the third week compared to aCSF control (p = 0.0089; Fig. 5).Motor function and coordination, as assessed by rota-rod performance, was not significantly affected by intra-STN VPA (p > 0.05).The remaining parameters evaluated in the Irwin screen and hyperexcitability tests were also not affected by intra-STN VPA, as illustrated by non-physiological scores (≥ 1) being observed only in few, or often no, animals across all groups (Fig. 5; p > 0.05).
A transient reduction in body weight occurred following surgical pump implantation with all VPA doses as well as with vehicle, though this did not reach statistical significance (p > 0.05).Within-group analyses of body weight showed subsequent normal weight gain in vehicle-   and C) rota-rod performance test.Scores are shown for each animal after acute intraperitoneal (i.p.) injection of valproate (VPA; 200 mg/kg, 300 mg/kg) or vehicle (NaCl), highlighted in green, or during chronic intrasubthalamic (intra-STN) microinfusion of VPA (480, 720, 960 μg/d and hemisphere) or vehicle (artificial cerebrospinal fluid, aCSF), highlighted in blue.In the acute systemic study (left side; green), a crossover design allowed for statistical comparisons between all groups (Friedman test with post-hoc Dunn's multiple comparisons).Rows outlined in green indicate a significant difference to both basal, "B", and vehicle, "NaCl" (p < 0.05).For the chronic intra-STN VPA data (right side; blue), comparisons were made between groups within each time point (week 1, "W1"; week 2, "W2"; week 3, "W3"); Kruskal-Wallis test with post-hoc Dunn's multiple comparisons, with significant differences to vehicle outlined in blue; p < 0.05).Within group comparisons to basal (pre-drug; "B") are not shown as statistical significance was not reached in post-hoc multiple comparisons (Friedman test with post-hoc Dunn's multiple comparisons).Gray cells indicate missing data points.Acute systemic VPA was associated mainly with ataxia and wet dog shakes, while chronic intra-STN VPA was associated with significant hypolocomotion, flat body posture, and wet dog shakes, however generally with less frequently and decreased severity compared to parameters affected by systemic VPA.The majority of the parameters assessed were not significantly affected by either systemic or intra-STN VPA.(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)Fig. 6.Valproate (VPA) concentrations in brain, plasma, and liver following acute intraperitoneal (i.p.) injection or chronic intrasubthalamic (intra-STN) microinfusion of VPA.Illustrated are total VPA concentrations of brain regions (A), plasma (B), and liver (C) collected 30 min post i.p. injection (200 mg/kg, 300 mg/kg; data illustrated in green) or after three weeks of chronic intrasubthalamic (intra-STN) microinfusion (960 μg/d bilaterally; data illustrated in blue) of VPA, as measured by high performance liquid chromatography (HPLC).VPA levels in the periphery (B and C) were dramatically reduced with intra-STN delivery compared to acute systemic administration despite the chronic nature of the intra-STN microinfusion.All animals illustrated could be defined as responders, with increases in intravenous pentylenetetrazole (ivPTZ)-induced seizure thresholds of at least 25% in response to the respective VPA treatments, though the magnitude of responses varied.This is reflected in the 960 μg/d intra-STN VPA treatment group with the rat showing the greater magnitude of response illustrated by a triangle symbol (△), and the rat with a lesser magnitude of response is represented by a square symbol (□)."SNr"substantia nigra pars reticulata, "HIP" -hippocampus, "STN" -subthalamic nucleus, "dors.STN" -area dorsal to the STN (zona incerta), "aCPU" -anterior striatum, "motCtx" -motor cortex.Note the differences in scale of the y-axis between A and B/C.(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)treated animals over time, which reached statistical significance in the third week of chronic aCSF microinfusion (p = 0.0026, not illustrated).Although it did not reach statistical significance (p > 0.05), this physiological weight gain of animals fed ad libitum was also seen in VPAtreated rats during the three-week treatment period.Accordingly, while comparison between groups indicated a main effect of time on percent change in body weight (F 1.778, 48.88 = 79.74,p < 0.0001), Šídák's multiple comparisons test did not reach statistical significance (p > 0.05).Thus, VPA-treated rats did not differ significantly in weight gain over time from vehicle-treated rats.

HPLC
At the conclusion of the systemic and intra-STN studies, a small subset of animals was sacrificed in a first approach to assessing VPA concentrations in the brain, liver, and plasma of rats following acute i.p. injection or chronic intra-STN microinfusion of VPA.Four animals from the systemic study (200 mg/kg: n = 2, 300 mg/kg: n = 2) and two from the intra-STN study (960 μg/d) could be used in the final descriptive evaluation and comparison.The results have to be confirmed with larger sample sizes.Furthermore, as mentioned in the results section 'Acute systemic valproate injection', it is important to note the difference in temporal drug exposure in addition to the varying route of administration between the acute systemic and chronic intra-STN VPA delivery.However, it has previously been shown that chronic, repeated i.p. injection of VPA yields relatively constant brain and plasma VPA concentrations throughout the chronic treatment despite changes in antiseizure and adverse effects (Löscher and Fiedler, 2000;Löscher et al., 1989).
With our highest intra-STN dose of 960 μg/d (2.77 mol/L), we reached a VPA level in the target structure of about 182 μg/g (i.e., about 1.26 mmol/L).This is comparable to the VPA level measured in the STN 30 min post i.p. injection of 300 mg/kg (about 176 μg/g; Fig. 6A).The targeted intra-STN delivery resulted in a rather restricted VPA distribution within the brain, with lower levels in areas more remote to the STN (e.g., hippocampus and motor cortex).Systemic VPA application resulted in more homogenous VPA concentrations across all brain regions measured (Fig. 6A).The lower systemic VPA dose of 200 mg/kg resulted in similarly homogenous concentrations of VPA across brain regions, however approximately 40-60% less than those measured after 300 mg/kg VPA (Fig. 6A).
VPA levels detected in liver following acute systemic doses of 200 mg/kg and 300 mg/kg VPA were 174.72 μg/g and 313.03 μg/g, respectively (Fig. 6C).This is in marked contrast to the VPA concentrations measured in the periphery after chronic intra-STN VPA delivery.Compared to systemically injected VPA at 300 mg/kg, intra-STN delivery of 960 μg/d VPA led to desirably low VPA levels in plasma (5.45 μg/mL) and liver (11.63 μg/g) (Fig. 6), thereby restricting relevant VPA levels to the targeted STN and closely surrounding structures.

In vitro electrophysiology
To match the effect of VPA treatment on the cellular physiology of STN neurons, we performed whole-cell recordings in acute brain slices.In STN neurons, 2.5 mmol/L VPA bath application led to multiple alterations in membrane properties (Fig. 7).The well documented spontaneous activity of STN neurons (Beurrier et al., 1999;Bevan and Wilson, 1999) significantly decreased from 10.48 ± 1.16 Hz to 5.68 ± 1.31 Hz before and from 14.22 ± 1.80 Hz to 6.52 ± 1.23 Hz after the hyperpolarization (p = 0.0064; n = 33 neurons) with VPA (Fig. 7B).The estimated background membrane potential significantly increased from − 49.94 ± 0.87 mV to − 43.14 ± 1.04 mV (p = 0.0031; Fig. 7C).The input resistance (R in ) significantly decreased from 56 ± 5.57 MΩ to 42.94 ± 5.87 MΩ (p = 0.002; Fig. 7D) after VPA wash-in.The change in R in was independent of the change in background membrane potential, shown by a lack of linear correlation (r = − 0.11; Fig. 7E), indicating that input resistance and background potential might be regulated separately.Finally, to ensure specificity of VPA on alteration in membrane properties, the same recording paradigm was repeated without VPA wash in (n = 5 neurons).None of the extracted parameters showed significant alteration, and the change in background potential and the R in showed no correlation (p = 0.0003; Fig. 7E).Thus, VPA appears to affect sub-and supra-threshold membrane properties of STN neurons.

Discussion
The main findings of the present study were: (a) Chronic convectionenhanced VPA delivery targeted to the bilateral STN caused long-lasting antiseizure effects at different doses, (b) with high responder rates up to about 80-90%, and (c) eliminated tolerance development at 480 μg/d.
(d) In contrast to systemic VPA injections, chronic intra-STN VPA delivery caused a more restricted VPA distribution within the brain and only minimal VPA distribution into peripheral compartments.(e) Accordingly, adverse effects were mild to moderate during intra-STN VPA delivery, but moderate to intense after systemic VPA injections.Slice electrophysiology revealed that 2.5 mmol/L VPA bath application leads to (f) a reduction in spontaneous STN firing rate, (g) an increase of background membrane potential, and (h) a decrease of input resistance.Finally, voltage clamp recordings revealed that (i) VPA blocks NMDAR currents in STN neurons.

Antiseizure effects
Feasibility and safety of chronic intracerebral (intracerebroventricular) VPA delivery via implantable pump/catheter system had been proven in patients with pharmacoresistant epilepsy (Cook et al., 2020).Using a comparable pump/catheter approach, our preclinical study in rats for the first time investigated chronic, targeted VPA delivery into the STN and revealed long-lasting, dose-dependent antiseizure effects over the tested 3-week period.
Comparable to recent findings with intra-STN muscimol delivery (Gernert et al., 2023), our current results revealed interindividual differences in antiseizure response strength.Interindividual differences in seizure threshold elevation have also been described for systemic VPA administration in the electrical amygdala-kindling model, which was not related to differences in plasma drug concentrations (Töllner et al., 2011).Although we cannot draw a final conclusion from our exemplary HPLC results, the present data indicate that plasma (and brain) levels were comparable between animals independent from the extent of antiseizure effects induced by intra-STN VPA delivery.Independent of individual response strength, overall responder rates were high (up to 89%) with intra-STN VPA delivery.
The degree to which epilepsy-associated network alterations may impact the response profile observed in the current work remains unknown.However, for several compounds, including valproate, antiseizure activity induced by targeted intracerebral delivery in the intravenous PTZ seizure threshold model predicts antiseizure effects against other seizure types, including difficult-to-treat partial seizures (López et al., 2007;Serralta et al., 2006).Nevertheless, future studies have to clarify, whether comparable antiseizure effects are also achievable in chronic epilepsy models, particularly in models of drug resistant epilepsies.

Tolerance to antiseizure effects
Tolerance to antiseizure effects was demonstrated for most ASMs during chronic systemic exposure (Löscher and Schmidt, 2006).However, for VPA, experimental evidence for tolerance to antiseizure effects is still a matter of debate (Löscher and Schmidt, 2006), as contrasting results have been reported (Löscher, 1986;Löscher et al., 1989;Mana et al., 1992a;Mana et al., 1992b;Van Vliet et al., 2010;Wahle and Frey, 1990;Young et al., 1987).Regarding tolerance time-course, VPA seems to have a rather complex pattern of tolerance development (Löscher and Schmidt, 2006).An increasing efficacy over the first treatment days without concomitant increases in drug plasma levels was described in different epilepsy and seizure models with varying dosing protocols including constant-rate infusions and pump delivery over several days (Löscher and Hönack, 1995;Löscher and Nau, 1982;Serralta et al., 2006) and was also observed in patients with epilepsy (Davis et al., 1994).In our study, the first seizure threshold testing under drug treatment was performed one week after starting VPA delivery.Thus, it is unknown whether the above described high initial antiseizure efficacy also occurs during intrasubthalamic delivery.We occasionally observed a delayed antiseizure response during chronic delivery of 480 μg/d, which might be related to the re-increase in antiseizure efficacy after about 3-4 weeks of daily treatment described by Löscher et al. (1989).However, we cannot exclude that tolerance development would occur after further weeks of continuous delivery, although tolerance to effect of systemically applied VPA typically develops after 1 week of repeated or continuous administration (Löscher et al., 1989;Van Vliet et al., 2010;Wahle and Frey, 1990).
Among clinical strategies aiming to prevent tolerance is the use of low dosage.Accordingly, our data indicate a dose-dependent difference in the manifestation of tolerance development with a tolerance rate of 0% at the low (but therapeutically effective) intra-STN dose, 80-100% at the medium dose, and 20-38% at the higher dose, the latter of which may partly compensate tolerance development.Further studies would be necessary to elucidate whether the observed loss or attenuation of antiseizure efficacy at the medium and high dose is truly tolerance, i.e., reversible, or acquired resistance induced by VPA treatment.

Adverse effects
We found dose-limiting ataxia and wet dog shakes in response to acute systemic injection, being in line with previous observations in rats (Serralta et al., 2006;Töllner et al., 2011).With targeted intra-STN VPA delivery, we only observed mild to moderate hypolocomotion (resulting in lowered body temperature) and flat body posture, both of which were not dose-limiting.The observed uneven VPA brain distribution with high levels in the target region and gradually lower levels in brain areas distant to the STN probably prevents the occurrence of intense neurological adverse effects such as ataxia.
Our brain/plasma ratios observed 30 min after systemic VPA injections (0.21-0.25) that had antiseizure effects matched those of earlier reports in rats (Löscher and Nau, 1983) and humans (Shen et al., 1992;Vajda et al., 1981).To reach sufficient brain concentration, high systemic drug load is necessary that may cause systemic adverse effects.Indeed, long-term systemic VPA use is clinically known to bear the risk of hepatotoxicity and teratogenicity among other adverse effects (Romoli et al., 2019).The extremely low systemic drug load observed in response to intra-STN VPA delivery of VPA can be expected to prevent most of those adverse effects described for VPA-receiving patients.

Functional considerations
Antiseizure effects induced by targeted intra-STN VPA most likely can be attributed to modulation of local neuronal activity within the STN and surrounding tissue, subsequently being propagated to other brain regions within the epileptic network.This is supported by our observation of high VPA levels in the STN and surrounding areas, such as SNr and zona incerta.In contrast, VPA levels were much lower in more distant regions such as hippocampus, motor cortex, and striatum during chronic intra-STN delivery compared to systemic injections.Simultaneous VPA infusion to regions directly surrounding the STN might be avoidable by multi-bolus targeting, shown to improve coverage of irregularly shaped brain structures while reducing off-target dosing (Ramadi et al., 2020).However, in our study SNr and zona incerta are themselves regions of interest for targeted seizure modulation.
Indeed, a direct effect on SNr neurons can be assumed for the present intra-STN VPA study as we reached a VPA concentration of about 130 μg/g within the SNr.This is in accordance with the nigral VPA concentration described to reduce firing rates of GABAergic SNr neurons after acute systemic VPA injection (Rohlfs et al., 1996).Additionally, our electrophysiological data revealed reduced firing rates of glutamatergic STN neurons upon VPA bath application.A previous study showed that lowering STN activity consequently lowers monosynaptically connected GABAergic SNr single unit activity in rats (Feger and Robledo, 1991), thereby disinhibiting downstream anticonvulsant zones such as superior colliculus neurons (Deransart and Depaulis, 2002;Gale et al., 1993;Hyder et al., 2023;Redgrave et al., 2010;Soper et al., 2016;Wicker et al., 2019).Thus, we assume that direct and indirect lowering of SNr activity contributes to the network effects we induced by intra-STN delivery of VPA.
Zona incerta, located dorsally to the STN, also showed high VPA levels and might have additionally contributed to the observed antiseizure effects induced by chronic intra-STN VPA.Clinically, deep brain stimulation not only of the STN (reviewed by : Benabid et al., 2001;Yan et al., 2022), but also of the zona incerta reduced seizure frequency in two patients with refractory sensorimotor focal seizures (Franzini et al., 2008).
Within the STN, VPA application might affect multiple neuronal targets that govern different aspects of membrane biophysics.Moreover, the NMDAR component of glutamatergic transmission was blocked within the STN.Selective reduction in peak NMDAR, but not AMPAR, currents is in line with findings from rat neocortex (Zeise et al., 1991), amygdala (Gean et al., 1994), and hippocampus (Ko et al., 1997).Likely, VPA leads to reduced feedforward mediated excitation by reducing input from glutamatergic synaptic transmission and intrinsic spontaneous activity in these glutamatergic neurons (Hammond et al., 1983;Kitai and Kita, 1987;Prensa and Parent, 2001).Although the VPA concentration used in the present in vitro experiments were higher than those achieved in the brain following chronic intra-STN VPA delivery, concentrations even lower than the brain concentrations reached in our in vivo experiments have previously been shown to suppress NMDAinduced transient depolarizations (0.1-1 mmol/L) in rat cortical slices (Zeise et al., 1991) and block ictal epileptiform discharges (0.5 mmol/L) in hippocampal slices obtained from rats at a similar age (P16-27) compared to the current study (Fueta and Avoli, 1991).Concentrations of 1 mM have additionally been shown to depress sodium reactivation in cultured rat hippocampal neurons (Van den Berg et al., 1993) and strongly decrease the frequency of, and spike number per, paroxysmal depolarizing shifts in mouse hippocampal slices (Lin et al., 2017).Future experiments should additionally comparably investigate STN slices from VPA nonresponders and responders, the latter of which with and without development of tolerance.

Conclusions
Our results show that chronic intra-STN convection-enhanced VPA delivery produces antiseizure effects that can be maintained at least over a three-week period in rats.We observed high responder rates as well as low tolerance rates at discrete VPA dosages.Electrophysiological data D. MacKeigan et al. indicate that a lowered STN activity and reduction in peak NMDAR current may be functionally involved in the observed antiseizure effects.Our present work is an important step towards further development of intracerebral drug delivery in epilepsies and is relevant for further neurological diseases such as movement disorders as well.

D
.MacKeigan et al.

Fig. 1 .
Fig. 1.Overview of the study design for in vivo (A and B) and in vitro (C) valproate (VPA) experiments.A) Acute systemic VPA: Following basal (pre-drug) behavior and seizure threshold testing, rats in the systemic VPA study received once weekly i.p. injections in a cross-over design with two VPA doses (200 and 300 mg/kg) and vehicle (NaCl).Behavior and intravenous pentylenetetrazole (ivPTZ) seizure threshold testing was conducted 15 and 30 min post-injection, respectively.Several weeks after conclusion of the systemic study, rats received a final i.p. injection of NaCl or VPA (200 or 300 mg/kg) and were decapitated for analysis of VPA levels in brain, liver, and plasma via higher performance liquid chromatography (HPLC).B) Chronic intra-subthalamic nucleus (intra-STN) VPA: Following basal testing, rats in the chronic intra-STN study underwent surgical implantation of microinfusion pumps and bilateral intra-STN cannulas for the continuous convection-enhanced

Fig. 3 .
Fig. 3. Effects of chronic intrasubthalamic convection-enhanced delivery of valproate (VPA) on thresholds to the first intravenous pentylenetetrazole (ivPTZ)induced myoclonic twitch.A) Thresholds to first myoclonic twitch (mean + SEM) during chronic intra-STN microinfusion of VPA (480, 720, or 960 μg/d, perhemisphere) or vehicle (artificial cerebrospinal fluid, aCSF; week 1, "W1"; week 2, "W2"; week 3, "W3") is compared to the basal (pre-drug; "B") control within the respective treatment group.Variance in sample size between time points within treatment groups is a result of paravenous PTZ injections leading to single data points removed from final analyses.Compared to basal control, no significant changes were observed in response to VPA.For direct comparison to the vehicle control group, ivPTZ seizure thresholds (ivPTZ-ST) were additionally calculated in percent change to basal (mean + SEM; B), revealing a significant elevation of the myoclonic twitch threshold after three weeks of continuous 480 μg/d VPA delivery.The defined responder threshold of 25% increase from basal is illustrated by a dashed line "R" on the y-axis.C) Responder rates and tolerance rates (a loss of response by the final ivPTZ-ST test) are given for each VPA dose and aCSF control.Rats were additionally categorized by the magnitude of the maximal response measured during chronic VPA or aCSF microinfusion compared to basal control (D).The highest responder and lowest tolerance rates were observed at 480 μg/d, while the strongest effect in individual rats was observed at a VPA dose of 960 μg/d.Mixed-effects model (two-way ANOVA) with Geisser-Greenhouse correction and Šídák's multiple comparison test for post-hoc analysis with individual variances computed for each comparison: ***p < 0.001.

D
. MacKeigan et al. (caption on next page) D. MacKeigan et al.

Fig. 5 .
Fig. 5. Occurrence and severity of behavioral adverse effects assessed via A) Irwin screen, B) hyperexcitability tests, and C) rota-rod performance test.Scores are shown for each animal after acute intraperitoneal (i.p.) injection of valproate (VPA; 200 mg/kg, 300 mg/kg) or vehicle (NaCl), highlighted in green, or during chronic

Fig. 7 .
Fig. 7. Valproate (VPA) induced alterations in membrane and synaptic properties of neurons in the subthalamic nucleus (STN).A) Spontaneous activity and response to a hyperpolarizing current step (− 50 pA) recorded under control conditions (left) and during wash-in of 2.5 mmol/L VPA (right).B) Spontaneous firing rate extracted for 1 s is greater in control compared to VPA conditions before and after the hyperpolarizing current injection (p = 0.0064, n = 33).Data were averaged from five control and five VPA trials per cell.Open symbols represent single cells and closed symbols average values.Averages are presented as mean ± SEM.C) VPA significantly increased the background membrane potential (E background , p = 0.0031).Symbols as in B. D) VPA significantly decreased the input resistance (R in , p = 0.0020).Symbols as in B. E) VPA-induced changes in membrane potential and input resistance do not correlate (p = 0.0003).Gray data points in the center are control data (n = 5) of the same recording time without VPA wash-in.F) Synaptic α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor (AMPAR)-and Nmethyl-D-aspartate (NMDAR)-currents recorded at − 70 mV and + 40 mV, respectively, before (gray) and during VPA wash-in (black).G) AMPAR-currents are unaffected by VPA (p = 0.1532; n = 6).Symbols as in B. H) NMDAR-currents are significantly reduced by VPA (p = 0 0.0010).Symbols as in B. I) NMDAR/AMPAR current ratio is significantly decreased by VPA (p = 0.0075).Symbols as in B.