Pharmacological Profile of the Novel Antiepileptic Drug Candidate Padsevonil: Interactions with Synaptic Vesicle 2 Proteins and the GABAA Receptor

Padsevonil is an antiepileptic drug (AED) candidate synthesized in a medicinal chemistry program initiated to rationally design compounds with high affinity for synaptic vesicle 2 (SV2) proteins and low-to-moderate affinity for the benzodiazepine binding site on GABAA receptors. The pharmacological profile of padsevonil was characterized in binding and electrophysiological experiments. At recombinant SV2 proteins, padsevonil’s affinity for SV2A was greater than that of levetiracetam and brivaracetam (pKi 8.5, 5.2, and 6.6, respectively). Unlike the latter AEDs, both selective SV2A ligands, padsevonil also displayed high affinity for the SV2B and SV2C isoforms (pKi 7.9 and 8.5, respectively). Padsevonil’s interaction with SV2A differed from that of levetiracetam and brivaracetam; it exhibited slower binding kinetics: dissociation t1/2 30 minutes from the human protein at 37°C compared with <0.5 minute for levetiracetam and brivaracetam. In addition, its binding was not potentiated by the allosteric modulator UCB1244283. At recombinant GABAA receptors, padsevonil displayed low to moderate affinity (pIC50≤6.1) for the benzodiazepine site, and in electrophysiological studies, its relative efficacy compared with zolpidem (full-agonist reference drug) was 40%, indicating partial agonist properties. In in vivo (mice) receptor occupancy studies, padsevonil exhibited SV2A occupancy at low ED50 (0.2 mg/kg) and benzodiazepine site occupancy at higher doses (ED50 36 mg/kg), supporting in vitro results. Padsevonil’s selectivity for its intended targets was confirmed in profiling studies, where it lacked significant effects on a wide variety of ion channels, receptors, transporters, and enzymes. Padsevonil is a first-in-class AED candidate with a unique target profile allowing for presynaptic and postsynaptic activity. SIGNIFICANCE STATEMENT Padsevonil is an antiepileptic drug candidate developed as a single molecular entity interacting with both presynaptic and postsynaptic targets. Results of in vitro and in vivo radioligand binding assays confirmed this target profile: padsevonil displayed nanomolar affinity for the three synaptic vesicle 2 protein isoforms (SV2A, B, and C) and micromolar affinity for the benzodiazepine binding site on GABAA receptors. Furthermore, padsevonil showed greater affinity for and slower binding kinetics at SV2A than the selective SV2A ligands, levetiracetam, and brivaracetam.


Introduction
Levetiracetam (LEV) was the first antiepileptic drug (AED) shown to exert its therapeutic activity by targeting elements of the synaptic release machinery, namely, through binding to the synaptic vesicle 2A (SV2A) protein (Lynch et al., 2004). SV2A and the other two protein isoforms-SV2B and SV2Care integral membrane glycoproteins that are present in secretory vesicles of neurons and endocrine cells (Bartholome et al., 2017). The precise function of the proteins remains elusive; however, given their presence in secretory vesicles, it is most likely that they play a role in vesicle exocytosis, an observation substantiated by accumulating evidence (Mendoza-Torreblanca et al., 2013). SV2A knockout mice die in a matter of weeks; however, in vitro recordings of neurons from very young SV2A knockout mice reveal a reduction in the frequency and amplitude of spontaneous inhibitory postsynaptic currents, potentially indicating a negative effect on GABA release from presynaptic neurons (Crowder et al., 1999). Data from animals lacking both SV2A and SV2B suggest that the absence of these proteins leads to presynaptic Ca 21 accumulation during consecutive action potentials, causing abnormal increases in neurotransmitter release (Janz et al., 1999a). The overall effect is destabilization of synaptic circuits and aberrant neurotransmission (Crowder et al., 1999;Janz et al., 1999a). LEV reduces both inhibitory and excitatory postsynaptic currents in an activitydependent manner, with the largest effect seen with the highest stimulation frequency (Meehan et al., 2012). Both LEV and brivaracetam (BRV), a more potent and selective SV2A ligand , produce frequency-dependent slowing of vesicle exocytosis and recycling, and of synaptic transmission (Meehan et al., 2011(Meehan et al., , 2012Yang et al., 2015).
The substantial evidence for the role of SV2A in the pathophysiology of epilepsy (Löscher et al., 2016;Ohno and Tokudome, 2017) and the clinical utility of SV2A ligands in the treatment of patients with epilepsy ensured that drug-discovery programs focusing on SV2 ligands continued. One strand of research involved investigating the feasibility of designing a molecule with both presynaptic and postsynaptic activity via interaction with SV2 proteins and the GABA A receptor (GABA A R), respectively. The rationale for targeting these proteins was based on observations that LEV markedly potentiated the activity of AEDs acting via GABAergic transmission, notably benzodiazepines (BZDs), in several animal models resulting in an improved efficacy/safety ratio (Kaminski et al., 2009a). More recent studies have shown that SV2A dysfunction resulting from a missense mutation (L174Q) results in a selective reduction in GABAergic transmission, rendering animals carrying the mutation highly susceptible to seizures and markedly facilitating kindling (Tokudome et al., 2016a,b).
In the subsequent rational medicinal chemistry design program, the focus was to develop a single molecule that could target SV2 proteins with high affinity and postsynaptic GABA A Rs, specifically the BZD binding site, with lower affinity. Low-to-moderate affinity for this site, coupled with a partial agonist profile, could minimize the potential for the development of tolerance, a phenomenon known to occur with most BZDs (Vinkers and Olivier, 2012;Gravielle 2016). Lead optimization efforts led to the discovery of padsevonil (PSL), an imidazothiadiazole heterocycle coupled to a pyrrolidone moiety (Fig. 1). PSL constitutes a novel chemical class, as indicated by its International Nonproprietary Name, which was approved by the World Health Organization in 2017.
The objectives of the studies reported here are to characterize the interactions of PSL with its intended therapeutic targets and to determine its selectivity for these targets using an array of validated in vitro and in vivo techniques. The pharmacological profile of PSL in nonclinical models of seizures and epilepsy is reported in the accompanying article (Leclerq et al., 2019).

Animals
Experimental procedures involving animals were conducted in compliance with guidance from the local ethics committee for animal experimentation according to Belgian law. All efforts were made to minimize animal suffering.
Naïve male specific pathogen-free NMRI mice (Crl:NMRI[Han]; 24-35 g) and male Sprague-Dawley rats (200-300 g) were obtained from Charles River Laboratories (Ecully, France). All animals were housed in a holding room under a 12-hour light/dark cycle with lights on at 06:00 hours. Temperature was maintained at 20-24°C, relative humidity at 40%-70%, and the rate of air replacement at least 15 times an hour. Animals had ad libitum access to standard dry pellet food and tap water.
[   and zeocin were purchased from Life Technologies (Merelbeke, Belgium), fibroblast-like cell line COS-7 cells from European Collection of Authenticated Cell Cultures, Sigma-Aldrich (Bornem, Belgium), complete protease inhibitor cocktail from Roche (Vilvoorde, Belgium) and DNAse (deoxyribonuclease I, type II from bovine pancreas) from Sigma-Aldrich. PBS, Dulbecco's modified Eagle's medium, L-glutamine, trypsin, and fetal bovine serum were purchased from Lonza (Verviers, Belgium). All other reagents were of analytical grade and obtained from conventional commercial sources.
Human cerebral cortex was obtained from Analytical Biologic Services Inc. (Wilmington, DE).

Tissue and Membrane Preparations
Preparation of Membrane Proteins from Rat and Human Cortex. Membrane proteins from rat cortex were prepared as described previously . Briefly, after rats were sacrificed by decapitation, brains were removed rapidly and dissected on ice. All subsequent operations were performed at 4°C. Brain tissue, either rat or human cerebral cortex, was homogenized (10% w/v) in 20 mM Tris-HCl buffer (pH 7.4) containing 250 mM sucrose (buffer A). Homogenates were spun at 30,000g for 15 minutes, and the pellets were resuspended in the same buffer. After incubation at 37°C for 15 minutes, membranes were washed three times using the same centrifugation protocol. Final pellets were resuspended in buffer A at a protein concentration of 10-15 mg ml 21 and stored at 2140°C until further use.
Preparation of Membrane Proteins from HEK and COS-7 Cells. Human SV2A, B, and C were expressed in HEK cells and rat GABA A R (a1b2g2; a2b2g2; a5b2g2) in COS-7 cells. Cells were subcultured in Dulbecco's modified Eagle's medium containing 200 mM L-glutamine and 100 mg ml 21 zeocin, supplemented with 10% fetal bovine serum, and grown in a humidified atmosphere of 5% CO 2 at 37°C. Confluent cells were harvested by trypsinization and pelleted by centrifugation at 1500g for 10 minutes at 4°C. The pellet was washed with ice cold PBS using the same centrifugation protocol and homogenized in a buffer containing 15 mM Tris-HCl, 1 mM EGTA, 0.3 mM EDTA, and 2 mM MgCl 2 (pH 7.5) supplemented with complete protease inhibitor cocktail. The homogenate was freeze-thawed twice and equilibrated at 25°C followed by a 10-minute DNAse (10 U ml 21 ) treatment. The solution was centrifuged for 25 minutes at 40,000g and 4°C. Finally, the pellet was resuspended in buffer A at a protein concentration of 5-10 mg ml 21 and stored at 2140°C until further use.

Radioligand Binding Experiments
Experiments were performed as previously described . For all assays, membrane proteins (100 mg per assay for cortical membrane proteins, 70-125 mg for SV2A, 2-5 mg for SV2B, 40-60 mg for SV2C, and 75-125 mg for rat GABA A R subtypes) were incubated for 120 minutes at 4°C or 60-150 minutes at 37°C in 0.5 or 2 ml of Tris-HCl buffer (50 mM, pH 7.4) containing 2 mM MgCl 2 . All glass fiber filters (from Brandel Inc., Gaithersburg, MD) used in the experiments were presoaked in 0.1% polyethyleneimine.
Competition Binding Experiments. Increasing concentrations of unlabeled competing drugs were added in the presence of 0.9 or 9 nM [ 3 H]PSL, 1 or 4 nM [ 3 H]UCB30889 or 2 nM [ 3 H]flunitrazepam. At the end of the incubation period, membrane-bound radioligand was recovered by rapid filtration through GF/B filters. Plates were washed rapidly three times with 0.3 ml of ice-cold Tris buffer; the total washing procedure did not exceed 10 seconds.
Kinetic Experiments. Specific [ 3 H]PSL binding in association experiments was measured at the indicated times after addition of membrane proteins at 37°C. Dissociation was induced by the addition of 10 mM of unlabeled PSL to the association reaction mixture. Samples were filtered on GF/C filters and washed with 5 ml of ice-cold Tris buffer. Total filtration time per sample did not exceed 2 seconds.
Saturation Binding Experiments. Membrane proteins were incubated with [ 3 H]PSL at concentrations ranging from 0.05 to 22 nM, and samples were filtered using GF/B filters. Nonspecific binding was defined as residual binding observed in the presence of 10 mM unlabeled PSL for [ 3 H]PSL, 1 mM LEV for [ 3 H]UCB30889, or 10 mM diazepam for [ 3 H] flunitrazepam. Radioactivity was determined by liquid scintillation.
The effect of UCB1244283, a positive allosteric modulator of SV2A, on PSL binding to SV2A was also evaluated. Studies were performed at 4°C according to the protocol described by Wood and Gillard (2017).
In Vivo Receptor Occupancy. The protocol developed by Li et al. (2006) was followed using NMRI mice instead of rats. Animals received vehicle (TWEEN/saline) or PSL (10 ml/kg body weight), administered intraperitoneally, followed 27 minutes later by tail-vein injections (2 ml/g) of either [ 3 H]UCB30889 (2 mCi) or [ 3 H]flunitrazepam (1 mCi). Necks of mice were dislocated 3 minutes after injection, and whole brains were rapidly removed, weighed, and homogenized in 10 volumes of ice-cold Tris buffer. The homogenate (300 ml) was filtered over GF/C filters and washed three times with 2 ml of ice-cold Tris buffer. Retained radioactivity was counted by liquid scintillation. Studies consisted of eight treatment groups (six PSL doses, one vehicle, one UCB30889/flunitrazepam), with six mice per group.

Selectivity Studies
Selectivity of PSL for its therapeutic targets was determined using radioligand binding, receptor activation, and electrophysiological studies.
Receptor activation studies were conducted at Eurofins Panlabs (St. Charles, MO) using standard protocols. The ability of PSL at concentrations of 1, 3, 10, and 30 mM to activate toll-like receptors (TLR 2 and 4) on cultured human peripheral blood mononuclear cells was evaluated by measuring concentrations of specific cytokines in the culture supernatant after stimulation. Cytokines tested included interleukin (IL)-1b, IL-6, IL-10, IL-12p40, and tumor necrosis factor a.

Functional Electrophysiological Studies
Electrophysiological recordings from GABA A channels were performed as described previously (Ghisdal et al., 2014). Briefly, Cl 2 currents were recorded from a Chinese hamster ovary-K1 cell line expressing the recombinant human a1b2g2 GABA A R subtype. Patchclamp recordings were performed on a PatchXpress system (Molecular Devices, Sunnyvale, CA). During all procedures, the holding potential was set to 260 mV. Whole-cell compensation was automatically set before each experiment. Current traces were recorded by patch-clamp amplifier (Multiclamp 700A computer-controlled patch-clamp dual headstage amplifier; Axon Instruments) at a sampling rate of 2 kHz. Recordings were performed at room temperature (∼25°C).

Data Analysis
In radioligand binding studies, data were analyzed by computerized nonlinear curve-fitting methods (Graphpad Prism 5 software, San Diego, CA), according to equations describing several binding models (Molinoff et al., 1981). IC 50 values were corrected to K i by applying the Cheng and Prusoff equation (Cheng and Prusoff, 1973).
In electrophysiological studies, data were analyzed using DataXpress 2 software (version 2.0.4.2; Molecular Devices). The potentiation of GABA A -evoked Cl 2 currents in the presence of drug was determined and compared with the maximum potentiation in the presence of 1 mM zolpidem.

Affinity for SV2A and the BZD Site
To determine the affinity of PSL for SV2A, the radioligand [ 3 H]UCB30889 was used in competition experiments in human and rat cortexes. This compound is an LEV analog with 20-fold greater affinity for SV2A than LEV, which makes it a suitable radioligand for labeling SV2A (Gillard et al., , 2006 (Table 1). Further studies on human recombinant GABA A R subtypes (a1b2g2, a2b2g2, and a5b2g2) yielded pIC 50 values of 6.1 6 0.1 (n 5 4), ,5 (n 5 3), and 6.0 6 0.1 (n 5 3), respectively, at 4°C.

In Vivo SV2A and BZD Site Occupancy
The binding profile of PSL to SV2A and the BZD site was also characterized in vivo, where results were similar to those obtained using cortical preparations: SV2A occupancy was Padsevonil Interactions with its Therapeutic Targets observed at low doses and BZD site occupancy at higher doses, indicating a greater PSL affinity for SV2A than the BZD site. In the dose-range administered intraperitoneally in mice, PSL ED 50 for SV2A occupancy was 0.4 mmol/kg (0.2 mg/kg; pED 50 6.4 mol/kg), whereas that of BZD site occupancy was markedly higher at 72 mmol/kg (36 mg/kg; pED 50 4.14 mol/kg) (Fig. 2).

Binding Characteristics at SV2 Isoforms
Detailed characterization of PSL binding to the three SV2 isoforms was performed using [ 3 H]PSL, the tritium-radiolabeled compound.
Kinetic Experiments. Binding kinetics of [ 3 H]PSL were determined on human recombinant SV2 isoforms expressed in HEK cells and in human and rat cortex at 37°C (Fig. 3; Table 2).
[ 3 H]PSL association kinetics were monophasic at all protein sources. Dissociation kinetics were also monophasic for the recombinant SV2 isoforms but more complex in human and rat cortex, although no clear separation of the phases was evident.
Saturation Binding Experiments. Saturation binding curves of [ 3 H]PSL on human recombinant SV2A/B/C and on human and rat cortex proteins were compatible with the labeling of a homogeneous population of binding sites at 37°C (Fig. 4). Corresponding affinities and B max values are given in Table 3. The affinity of [ 3 H]PSL was similar for all protein sources, and the K d value obtained for recombinant human SV2A corresponded with the K i obtained for the unlabeled compound using [ 3 H]UCB30889 (Table 1).
UCB1244283 is a positive allosteric modulator of SV2A that increases the binding of BRV and LEV to the protein (Daniels et al., 2013). Use of the modulator (at 30 mM and at 4°C) had no effect on the B max for [ 3 H]PSL binding to the human SV2A protein in HEK membranes, but it reduced the affinity from 2.3 6 0.2 nM (n 5 3) to 3.90 6 0.5 nM (P , 0.05, Student's t test).
Competition Experiments. Competition experiments were performed on all human recombinant SV2 isoforms expressed in HEK cells and on human and rat cortex proteins using several SV2A ligands, as well as diazepam (Fig. 5). These experiments were performed with 0.9 nM [ 3 H]PSL. Data were analyzed using sigmoidal dose-response fits with variable slope, and the resulting pIC 50 values were transformed into pK i values (Table 4). The affinity of the selective SV2A ligands-LEV, UCB30889, and BRV-was very low for both SV2B and SV2C labeled with [ 3 H]PSL, resulting in incomplete competition curves at the concentration range used. This finding is in line with previous reports, demonstrating the selectivity of LEV and BRV for SV2A (Noyer et al., 1995;Gillard et al., 2011). Competition with PSL led to a complete displacement of [ 3 H]PSL binding on all protein sources. For all human recombinant SV2 isoforms, the Hill slopes of the curve fits were not different from unity, and the obtained K i values for PSL agreed with the K d values of the radiolabeled compound.
In human and rat cortex, the competition curves of LEV, BRV, and UCB30889 fitted on the data using a sigmoidal doseresponse curve with variable slope were shallower than those for PSL. The corresponding Hill coefficients ranged from 20.8 to 20.7 and reached significant difference from unity for LEV in both human and rat cortex (P , 0.01) and for UCB30889 and BRV in rat cortex (P , 0.01) ( Table 4). The pronounced shallow profile of these competition curves allowed analysis of data with a model describing the binding to two independent populations of binding sites. At a radioligand concentration of 0.9 nM, the result fit with 85% of the binding sites having an affinity similar to that on recombinant human SV2A, and the remaining sites displaying an affinity in the low millimolar range. The latter proportion of binding sites most likely represents the SV2B component, given the very low presence of SV2C in the cortex.

Effects in Selectivity Studies
Selectivity of PSL for its therapeutic targets was determined using radioligand binding, receptor activation and electrophysiological studies.
In radioligand studies, PSL at 10 mM lacked significant effects (.50%) on a wide variety of molecular targets, including ligand-gated and G-protein-coupled receptors, ion channels, transporters, and enzymes. The only significant effect was at the BZD binding site of the GABA A R (87% inhibition), confirming the profile described already here. Approximately 50% inhibition of binding to the human neurokinin NK2 receptor was also observed and subsequently confirmed in a concentrationresponse study with a pIC 50 of 5.
In receptor activation assays, PSL at concentrations up to 30 mM failed to elicit release of any of the tested cytokines from human peripheral blood mononuclear cells, indicating a lack of activity on TLR 2 and 4. In electrophysiological studies, PSL at 10 mM had no significant effect on any of the classic AED targets such as voltage-gated ion channels or glutamatergic receptors.

Activity at Recombinant GABA A Receptors
In the recombinant human a1b2g2 GABA A R subtype, application of GABA resulted in activation of Cl 2 currents with an EC 50 of 15 mM (data not shown). To evaluate the effects of drugs on these GABA-evoked Cl 2 currents, an agonist concentration corresponding to its EC 20 (5 mM) was selected. PSL, in a concentration range of 1 nM-30 mM, when added alone for 5 minutes during preincubation, did not significantly change basal Cl 2 currents compared with control (DMSO 0.5%, data not shown) but potentiated Cl 2 currents evoked by 5 mM GABA. Potentiation of GABA-induced Cl 2 currents was dose-dependent with an EC 50 of 137 nM and a maximal effect reaching 167% (Fig. 6).
The relative efficacy of PSL in potentiating GABA A R Cl 2 currents was compared with that of zolpidem, a reference drug included in each experiment. Zolpidem is a full agonist at the BZD site and potentiates GABA [EC 20 ] with a maximal efficacy at 1 mM. The relative efficacy of PSL at 10 mM was 44% 6 16% compared with the maximal response to zolpidem defined as 100%.

Discussion
PSL was developed in a rational drug-discovery program initiated to develop a molecule with a novel mode of action aimed at the treatment of patients with drug-resistant epilepsy. Results of experiments described here confirm that PSL is a high-affinity pan-SV2 ligand, and it acts as a lowaffinity partial agonist at the BZD site of GABA A Rs. Profiling studies also confirmed the lack of effect on typical AED targets, such as ion channels and glutamate receptors, as well Padsevonil Interactions with its Therapeutic Targets as a variety of other central nervous system targets, including ligand-gated and G-protein-coupled receptors, transporters and enzymes.
PSL is the first ligand that interacts with all three SV2 isoforms. In saturation and competition studies, [ 3 H]PSL displayed nanomolar affinity for SV2A, SV2B, and SV2C. This finding was in contrast to LEV and BRV, both displaying marked selectivity for SV2A and interacting with SV2B and SV2C only at concentrations 100-fold higher than that at which they interacted with SV2A. This finding was also consistent with the lack of binding of a LEV derivative and of [ 3 H]BRV in mouse brain tissue lacking SV2A (Lynch et al., 2004;Gillard et al., 2011). The high affinity of [ 3 H]PSL for all isoforms was also reflected in kinetic data. [ 3 H]PSL dissociation kinetics were monophasic at human recombinant SV2 isoforms but more complex in human and rat cortex, most likely owing to the labeling of a relatively low fraction of SV2B and SV2A at 0.9 nM [ 3 H]PSL. At this concentration, relatively more SV2A binding sites are labeled since [ 3 H]PSL has a 4-fold higher affinity for SV2A (K d 1.5 nM) than for SV2B (K d 6.3 nM). Labeling of only a small proportion of SV2B in both human and rat cortex is supported by results of competition experiments; SV2C labeling was not anticipated given its restricted expression in the cortex (see the following). Results with LEV and BRV demonstrated labeling of approximately 85% of the binding sites with affinities similar to that at human recombinant SV2A. Finally, both SV2A and SV2B labeling in the cortex is supported by the observation that the B max values for [ 3 H]PSL were approximately twice as high as those observed for SV2A-selective ligands (Gillard et al., , 2011. Results do not allow an accurate prediction of the SV2A/ B ratio in the cortex because differences in the labeling of both isoforms in native versus recombinant expression systems cannot be excluded. Overall, results are consistent with the interaction of PSL with all three SV2 isoforms and suggest that it may be a useful ligand in further characterization of these proteins. The roles of SV2B and SV2C in the pathophysiology of epilepsy remain unknown. Few reports have addressed the role of SV2B since its distribution in the mammalian brain partly overlaps with that of SV2A (Bajjalieh et al., 1993(Bajjalieh et al., , 1994. SV2B 2/2 mice do not show phenotypic abnormalities, and electrophysiological studies on cultured neurons from these mice do not reveal any obvious effects on neurotransmission (Janz et al., 1999a;Chang and Südhof, 2009;Venkatesan et al., 2012). In a recent study, SV2C expression was strongest in the basal ganglia and more restricted in the cortex of rodent, rhesus macaque, and human brain (Dunn et al., 2019), consistent with results from previous rodent studies (Janz and Südhof, 1999b;Dardou et al., 2011). In the human brain, SV2C was highly colocalized with the GABA transporter in the striatum, substantia nigra, and ventral tegmental area (Dunn et al., 2019). SV2C expression was weak or absent in the hippocampus of autopsy controls, but it increased in biopsies from patients with temporal lobe epilepsy owing to mesial temporal sclerosis; in contrast, SV2A and SV2B expression H]padsevonil binding on human recombinant synaptic vesicle protein 2 (SV2)A/B/C and in human and rat cortex at 37°C. Kinetic constants were calculated by nonlinear regression analysis of association and dissociation curves of [ 3 H]PSL as depicted in Fig. 3 using a model describing the interaction of a ligand with a single site showing the calculated affinity constant (K D ). Data represent mean 6 S.D.
PSL differs from SV2A-selective ligands not only in its affinity for SV2B and SV2C, but also in its interaction with SV2A, as suggested by the slow kinetics of [ 3 H]PSL and the effect of the modulator, UCB1244283. BRV and LEV have H]padsevonil on human recombinant synaptic vesicle protein 2 (SV2)A/B/C and on human and rat cortex at 37°C. Data represent mean 6 S.D. (n 5 3) and were obtained from the analysis of saturation curves as presented in Fig. 4  Padsevonil Interactions with its Therapeutic Targets extremely fast kinetics: BRV's dissociation t 1/2 is 20 seconds at 37°C (Gillard et al., 2011), and that of LEV is 30 seconds at 25°C (Noyer et al., 1995). [ 3 H]PSL dissociation t 1/2 from human SV2A was 30 minutes at 37°C. This slower off-rate was also seen in rat and human cortex, suggesting a predominance of SV2A in these tissues. The SV2A positive allosteric modulator, UCB1244283, increases both BRV and LEV binding, but by different mechanisms; an increase in the binding capacity for LEV and an increase in binding affinity and capacity for BRV (Wood and Gillard 2017). Allosteric modulation is typically associated with an increase in binding affinity; however, as a putative transporter of the major facilitator superfamily, SV2 can adopt different inward and outward facing conformations (Quistgaard et al., 2016), and with selective binding of some ligands to only one conformation, the presence of an allosteric modulator that stabilizes or induces that specific conformation could seemingly increase binding capacity.  , their effect on the protein remains unknown. Therefore, it is difficult to predict the functional consequences of the slow offrate and the novel mechanism of interaction of PSL with SV2A. Given the evidence that antiseizure potency correlates with SV2A binding affinity (Noyer et al., 1995;Kaminski et al., 2008Kaminski et al., , 2009bGillard et al., 2011), it is plausible that such a highaffinity and long-lasting interaction could also lead to improved antiseizure efficacy. Postsynaptically, PSL displayed low-to-moderate (micromolar) affinity for the BZD site of GABA A Rs in human and rat brain membranes with no interspecies difference in potency. Results were similar in recombinant GABA A R subtypes with a1, a2, and a5 subunits coexpressed with b2 and g2 subunits. These subunits were chosen because they have been suggested to mediate many of the BZDs' pharmacological effects (D'Hulst et al., 2009). In functional studies, PSL lacked any direct effect on recombinant a1b2g2 GABA A Rs, but it potentiated the effects of low GABA concentrations (EC 20 , giving around 20% of the maximum GABA response) with a pEC 50 of approximately 7.0 (100 nM). This observation is in keeping with the known mechanism of action of many BZDs, positive allosteric modulators of GABA A Rs that do not open the Cl 2 channel on binding, but increase the affinity for channel gating by GABA (Sigel and Steinmann, 2012). PSL's relative efficacy compared with zolpidem was 44%, indicating that it acts as a partial agonist. PSL was designed specifically to act as a partial agonist in an effort to minimize the potential for induction of tolerance. BZDs have potent antiseizure effects; however, their long-term use is limited by sedation, tolerance, and the risk of dependence (Riss et al., 2008;Rudolph and Knoflach, 2011;Ochoa and Kilgo., 2016). In epilepsy, tolerance is associated with a loss of antiseizure efficacy, a progressive increase in seizure frequency and severity, and an increased risk of withdrawal seizures, even if AEDs are kept at constant  4 pK i values and Hill slopes of selective synaptic vesicle protein (SV)2A ligands and diazepam for sites labeled with 0.9 nM [ 3 H]padsevonil on recombinant human SV2A/B/C and on human and rat cortex at 37°C. Data are presented as mean 6 S.D. (n 5 3-10) and were obtained from nonlinear regression analysis of untransformed raw data using a sigmoidal dose-response model with variable slope. Missing values could not be calculated because of incomplete competition curves as a result of compound solubility limits.
In conclusion, PSL is a first-in-class AED candidate with a presynaptic and postsynaptic mechanism of action. Studies described here have shown that PSL displays high affinity for the three SV2 isoforms, ranging from 1.5 to 6.3 nM, and low-tomoderate affinity for the GABA A R BZD site, where it acts as a partial agonist, and that this profile is maintained in vivo. Furthermore, the interaction of PSL with SV2A differs from that of LEV and BRV in that it displays a markedly slower dissociation rate, and its activity is not potentiated by an SV2A modulator. These additional properties may contribute to the highly active profile of PSL in nonclinical seizure and epilepsy models, including those where the SV2A-selective ligands, LEV and BRV, show limited or no activity (Leclerq et al., 2019).