Can in vitro studies aid in the development and use of antiseizure therapies? A report of the ILAE/AES Joint Translational Task Force

In vitro preparations (defined here as cultured cells, brain slices, and isolated whole brains) offer a variety of approaches to modeling various aspects of seizures and epilepsy. Such models are particularly amenable to the application of anti‐seizure compounds, and consequently are a valuable tool to screen the mechanisms of epileptiform activity, mode of action of known anti‐seizure medications (ASMs), and the potential efficacy of putative new anti‐seizure compounds. Despite these applications, all disease models are a simplification of reality and are therefore subject to limitations. In this review, we summarize the main types of in vitro models that can be used in epilepsy research, describing key methodologies as well as notable advantages and disadvantages of each. We argue that a well‐designed battery of in vitro models can form an effective and potentially high‐throughput screening platform to predict the clinical usefulness of ASMs, and that in vitro models are particularly useful for interrogating mechanisms of ASMs. To conclude, we offer several key recommendations that maximize the potential value of in vitro models in ASM screening. This includes the use of multiple in vitro tests that can complement each other, carefully combined with in vivo studies, the use of tissues from chronically epileptic (rather than naïve wild‐type) animals, and the integration of human cell/tissue‐derived preparations.

| What is the goal of in vitro studies in epilepsy?
Epilepsy is one of the most common neurological diseases, affecting between 50 1 and 70 2 million people worldwide.Here, as part of the International League Against Epilepsy/American Epilepsy Society (ILAE/AES) Joint Translational Task Force, we discuss how in vitro preparations (defined as cultured cells, brain slices, and isolated whole brains, summarized in Table 1) can be used in the development of anti-seizure therapies.The underlying pathophysiology of epilepsy can be complex and multifaceted, and has been associated with numerous changes including cell death, 3 inflammation, 4 astrogliosis, 5 circuit re-wiring, 6 genetic variants, 7 and transcriptomic dysregulations. 8,9In many cases, the etiology of epilepsy remains unknown.These changes may not only underlie seizure genesis or enhanced seizure probability but also participate in a range of co-morbidities associated with epilepsy 10 including memory loss, sleep disruption, and moodrelated disorders, such as anxiety or depression.Frontline clinical treatment for epilepsy is with anti-seizure medications (ASMs).However, despite the availability of >30 ASMs, 11,12 approximately one third of people with epilepsy do not experience seizure freedom with these medications. 13We define this phenomenon as drug-resistant epilepsy (DRE; also termed pharmacoresistant epilepsy).Novel ASMs must be made to treat patients with DRE, which justifies using in vitro studies as they allow a high throughput when screening compounds.Box 1 summarizes the key questions that can be addressed in vitro.
In vitro approaches offer a number of significant advantages over alternative methodologies for therapeutic screening (typically direct application of putative therapeutic compounds using in vivo animal models).First, the use of cell culture-based or acute slice seizure models circumvents the blood-brain barrier (BBB), 14 a key physiological limitation that can hinder in vivo drug screening by preventing potentially efficacious compounds from physically reaching the brain. 15Because novel compounds can be applied directly to in vitro preparations, they provide a straightforward readout of efficacy that is not confounded by peripheral effects due to drug metabolism or the ability to cross the BBB.We note that this advantage can become a limitation, since potential anti-seizure therapies should cross the BBB.Moreover, the administration of novel compounds directly in vivo may raise ethical issues regarding animal welfare, 16 due to the possibility of unanticipated adverse effects.It is widely acknowledged that researchers must follow the principles of the 3Rs (reduce, refine, and replace) when conducting animal experiments. 16,17In vitro models provide a powerful alternative, which addresses all three points to the greatest degree.In fact, statistical caution should be exercised to ensure that enough animals are used (for example, defining 'N' as number of animals rather than number of cells or slices).This adheres to good experimental design and ensures sufficient statistical power to make valid inferences.
• In vitro models offer a variety of approaches for capturing various aspects of seizures and epilepsy.• These models are particularly useful in screening mechanisms of anti-seizure compounds, and also provide higher throughput tools to screen efficacy.• All models are simplifications of reality and each has inherent pros and cons.We propose, therefore, that best practice is to use a battery of in vitro models that can complement each other.• A second recommendation is that in vitro studies into anti-seizure therapies should use tissues derived from chronically epileptic, rather than wild-type, animals.• Finally, we recommend the integration of human tissue-based in vitro models into pre-clinical studies to develop anti-seizure therapies.Any model is by construction a simplification of reality.Thus every model has intrinsic limitations due to the reductionist approach used to answer a given clinical/ scientific question.In the following sections, we discuss different models and their main limitations.

| What do in vitro studies mean in the context of epilepsy?
Broadly, we define in vitro epilepsy models as those which use brain cells or networks of brain cells outside of a living organism.These systems must be capable of producing epileptiform activity, which may include seizure-like events, interictal-like patterns, status epilepticus (SE)-like activity, or cellular bursting discharges.Such activities can be evoked in a number of T A B L E 1 Overview of key advantages and disadvantages of each in vitro approach.

Type of model Main advantages Main disadvantages
Acute brain slice ways, which can include the use of acute chemoconvulsants or electrical stimulation, or certain types of activity can occur spontaneously in ex vivo tissues taken from chronically epileptic rodents or human patients (Figure 1).A full discussion of different approaches to elicit activity sits outside the scope of this article, but readers are referred to previous reports from ILAE/ AES Task Force groups. 18,19There is a wide variety of in vitro systems available in experimental epilepsy research.Here, we will define each one, describe their key advantages and limitations, and provide examples of the types of activity that can be recorded in each type of preparation.

| TYPES OF IN VITRO PREPARATION
Here we provide an overview of the main types of in vitro models used in epilepsy research, and we discuss the key advantages and limitations of each.This information is summarized in Table 1 and Figure 1.animal (typically a mouse or rat) and kept alive artificially in a short-term (acute) storage chamber. 20Brain slices typically remain viable for up to 12 h, meaning that they are amenable to roughly 1 day of experimentation (this timeframe can be extended to ~2 days when preparing slices from immature tissue under sterile conditions 21,22 ).A notable advantage of this approach over neuronal cultures is that acute slices are obtained from intact brains and can include original local connections.Moreover, with careful choice of slice orientation and thickness, acute brain slices are able to retain connections between different regions of the brain, which may be relevant to the propagation of epileptiform activity.For example, hippocampal slices can be obtained with retained connectivity to and from key related structures, including the entorhinal [23][24][25][26] and perirhinal cortices 24 and the amygdala. 24n cases where acute brain slices are obtained from naïve (non-epileptic) rodents, epileptiform activity does not occur spontaneously and is typically induced by altering the composition of cations within the artificial cerebrospinal fluid (aCSF) perfusate (e.g., zero magnesium, elevated potassium, or reduced magnesium/elevated potassium combination), [27][28][29][30][31] or using chemoconvulsants (e.g., 4-aminopyridine, picrotoxin, bicuculline, or kainate); these experimental procedures alter excitation 32 and/or inhibition [33][34][35] within the brain slice.These methods have parallels with symptomatic seizures in the clinic, where abnormal levels of cations or exposure to chemoconvulsants can also trigger acute seizures in humans.In brain slices, epileptiform activity can also be generated by electrical stimulation. 36Consideration should be given to the anatomic gradients of excitability within regions of interest (e.g., hippocampus and entorhinal cortex) and a careful correlation of seizure activity with slices obtained from along the septo-temporal or dorso-ventral axis.In vitro brain-slice studies in the entorhinal cortex 37 and hippocampus 38,39 have shown significant differences in ictal and inter-ictal initiation dependent on the precise anatomic location from which the slices were obtained.These differences in epileptiform activity across the respective axes are likely to reflect alterations in both intrinsic neuronal features and inhibitory networks.
Acute brain slices have been used to demonstrate the influence of metabolic pathways for seizure-like activity.Exposure to mitochondrial toxins (rotenone, potassium cyanide) and an aconitase inhibitor (fluorocitrate) produced repetitive inter-ictal discharges in hippocampal networks. 40n contrast, perturbation of a specific element of neuronal metabolism, mitochondrial dihydroorotate dehydrogenase, demonstrated reductions in neuronal and synaptic excitability in vitro. 41Alterations in cellular metabolism are emerging as an important driver in both acquired (for e.g., hypoxicischemic, traumatic brain injury, vascular dysfunction, SE, and brain tumor related epilepsy) and particular genetic epilepsies (for e.g., SLC135A, SLC2A1, POLG1, myoclonic epilepsy with ragged red fibers, mitochondrial encephalomyopathy lactic acidosis and stroke-like episodes, and ALDH5A1).Further work utilizing in vitro models in this specific area should indeed provide additional insights into the pathophysiology of these conditions.
Epileptiform activity recorded in brain slices is macroscopically similar to interictal discharges and seizures recorded in patients, 42 and therefore slice models have provided a rapid and realistic preparations to screen putative novel therapeutic compounds. 32By studying dynamic cellular and network changes occurring in the slice between chemoconvulsant application and the onset of seizure-like events in naïve slices, it may be possible to interrogate mechanisms of ictogenesis, seizure onset, and seizure termination (for example 43 ).However, as mentioned above, epilepsy is associated with alterations in brain circuits, which are not present in naïve control slices.For example, fast ripple oscillations, which may be a pathological biomarker of epileptic networks, can be seen in ex vivo tissues from chronically epileptic rats in response to elevated extracellular potassium, but the same phenomenon is usually not observed using control brain slices. 29Moreover, ex vivo slices obtained from rodents with epilepsy have been demonstrated to generate spontaneous interictal-like activity 44 (although ictal-like activity often still needs to be evoked with chemoconvulsants 29,44,45 and/or electrical stimulation 46 ).There is always the possibility that a compound validated in slices from non-epileptic animals may not work in slices from experimental models of epilepsy.Although more time-consuming and costly, it may be preferable to test compounds in acute brain slices from rodent in vivo models of chronic acquired 29 and genetic epilepsies. 47Using slices from experimental models of epilepsy raises further challenges.A compound may be efficient in one model/species/strain and not in another.Ideally, compounds should be tested in the seizure-onset zone.Although studies performed in acute human slices (see below) allow direct measurements in the epileptogenic zone obtained after neurosurgery, [48][49][50] the epileptogenic zone is rarely identified in experimental models before performing slice experiments.This would require using the same strategy as in patients with DRE: performing multisite recordings to estimate the onset zone and then extract the brain and make slices that include the onset zone.We are aware of only two studies in which multisite recordings have been performed, in pilocarpine-treated rats 51 and in kainic acid-treated mice 52 to identify the onset zone.Although it increases both the complexity and the cost of the experiment, this approach is technically possible.Such slices likely offer greater insight into human epilepsy, and a key recommendation of our report is the use of ex vivo brain slices obtained from epileptic rodents, where possible (Box 2).

| Intact and whole brain preparations
Intact in vitro preparations are those where the whole intact structure and connections of specific regions (e.g., hippocampus) is removed and maintained in vitro.This type of preparation offers the researcher an expanded anatomic connectivity relative to slice preparations and the ability to stimulate connected regions or pathways.These intact preparations also have the advantage that they maintain anatomic gradients within the isolated structures.The in vitro intact hippocampal 53 and corticohippocampal 54 formations have been used to study the induction and propagation of epileptiform activity, and also for therapeutic screening of ASMs. 55he isolated guinea pig brain was developed initially 56 for the purposes of studying multi-synaptic connections in an intact whole brain.This preparation is advantageous over other in vitro preparations, as it maintains long-range functional connectivity, BBB function, and neurovascular interactions, and the intact vasculature can be used to ensure robust supply of oxygenated aCSF to all brain regions (this is an advantage over brain slices in which brain activity is highly influenced by the amount of oxygen provided by the perfusate 29,57 ).Therefore, robust seizure activity can be studied simultaneously in different areas and in both hemispheres using conventional electrophysiological approaches from single cells to neuronal networks.Similar to brain slices, epileptiform activity has to be evoked by pharmacological manipulations.In addition, other techniques such as ion sensitive electrodes (pH, K + ), voltage sensitive dyes, 58,59 intrinsic optical imaging, 59,60 and calcium imaging 61,62 commonly used in slices can also be applied in this preparation.Another advantage of the isolated whole brain preparation over in vitro brain slice approaches is the presence of an anatomically intact neurovascular unit. 63,64However, the effects of vasodilatation/vasoconstriction in the absence of the pulsatile action of heart beats, and how it reacts to the presence of non-physiological oxygen concentration (95% instead of 20%), remain to be assessed.Dysfunction of the neurovascular unit is associated with seizures and epilepsy 65 ; in the isolated brain the BBB can be manipulated experimentally 66 in order to probe its impact on epilepsy in the absence of peripheral influences, which may confound in vivo studies.This could be developed further by using the isolated brain from an already epileptic guinea pig, which may recapitulate BBB dysfunction seen in the human condition.In the context of drug-resistant seizures, neuroinflammatory changes and BBB alterations have been implicated in the seizure pathophysiology.The ability of this particular in vitro model to retain functionally viable and interactive neuronal, glial, neurovascular, 67 and neuroinflammatory 68 elements of the system makes it useful for testing hypotheses and novel therapeutic avenues that involve neurovascular 69 and/or neuroinflammatory aspects.Finally, the isolated brain preparation can be used to evaluate BBB permeability of ASM and brain active molecules. 70One limitation of this approach is the relative difficulty in visualizing and accessing specific cells, particularly in structures deep within the brain such as the hippocampus, which may be more readily targeted for recording in slice preparations from specific brain regions.

| Organotypic slice cultures
2][73] Slices prepared in this way can be kept for several weeks, vastly prolonging the lifetime of the tissue when compared with acute approaches.This timescale is the most notable advantage of organotypic slice cultures.Certain experimental interventions (e.g., antisense oligonucleotides [ASOs] 74 or viral vectors 75,76 ) BOX 2 Key recommendations when using in vitro models for pre-clinical research • Given that each in vitro preparation has inherent advantages and limitations, a battery of tests must be chosen carefully to assess mechanisms and efficacy in various disease models.• Although in vitro preparations serve as a higher throughput initial screening tool, subsequent use of in vivo models may help to consolidate evidence of pre-clinical efficacy.• Where possible, in vitro studies with rodent tissue should use brains from chronically epileptic animals, rather than non-diseased control tissues.This better captures the epileptic brain state, which is pre-disposed to spontaneous seizures and may provide a more valid screening tool.• Where possible, animal-based preparations should be complemented with human-based tissues in pre-clinical pipelines.This provides valuable translational evidence of efficacy in human-derived neurons or brain tissues, and provides evidence against potential speciesspecific effects.
may require timescales on the order of days to weeks to mediate their effects.This is longer than the feasible viability of acute slice preparations.A further property (and a technical necessity) of organotypic slice cultures is that they are derived from young (typically <postnatal day 10) rodents, and that the neurons and circuits within these slices continue to mature in culture.Therefore, another possible application of this technique is to probe how ASM mechanisms and efficacy change with neurodevelopment, which could have particular relevance to the treatment of childhood epilepsies.Although organotypic slice cultures retain some advantages of acute slices, there are also drawbacks to this approach.Over time in culture, organotypic slices naturally flatten as they adhere to their semi-permeable membrane support 73 and re-organize their synaptic connections. 73f interest, cultures from non-epileptic brains eventually generate spontaneous interictal-like activity and seizures-like events, 77 offering insights into the reorganization process that is considered part of in vivo epileptogenesis.This has been exploited as a model of acquired epilepsy after brain injury. 78In drug-screening studies, organotypic slice responses have been validated using in vivo acquired epilepsy models. 79

| Isolated cells
A high-throughput in vitro approach uses isolated cultured cell lines, such as HEK293 or neuro2A cells, and automated patch clamp systems. 80At the time of writing, leading highthroughput screening systems are able to record from 384 cells simultaneously. 80By transfecting ion channels into these isolated cells, researchers can quickly record the activity of these channels in large numbers of cells.In epilepsy research, this permits the rapid phenotypic interrogation of genetic variants in relevant ion channels. 81Another application is high-throughput screening of the effects of a large number of compounds on specific ion channels. 82Isolated cells also offer technical advantages: they have superior space clamp properties compared with primary neurons 83 and they are amenable to single-channel recordings. 84The main drawback of such approaches is that they are far from being representative of realistic brain cell or circuit structures.Therefore, any findings at the channel level discovered using this method must be validated using more realistic models in order to infer their effects on epileptiform activities, which are network-driven events.
It is important to note that the presence of a genetic variant is not necessarily responsible for ictogenesis, although it may produce epilepsy. 85Furthermore, the presence of a variant does not necessarily lead to modification of firing behavior in neurons-a form of degeneracy, which in general refers to the idea that multiple component parameter values in a system can give rise to the same output. 86

| Dissociated cell cultures
A different approach to cellular models is the use of primary dissociated cultures. 87Typically, brains are dissected from rodent pups and their cells are dissociated, plated, and maintained in culture.Cultures can be purely neuronal, or they can contain mixtures of neurons and glia.A clear advantage over automated screening methods is the use of real neurons, as opposed to cell-line systems.Therefore, dissociated cultures are capable of producing realistic epileptiform bursting activity at the single-cell level. 88,89In silico modeling has been used to link activity in single neurons with different epileptiform activities. 90oreover, dissociated neurons are capable of forming synaptic connections and networks in culture, and can be plated on to microelectrode arrays (MEAs) to simultaneously monitor the activity of multiple neurons.MEA recordings have also been also used to analyze the dynamics of synchronous epileptiform activities in brain slices. 91,92lthough primary cell cultures are less realistic than the microcircuits preserved in slice preparations, they do facilitate the study of synaptic connectivity and network activity.One technical advance used a modified dynamic clamp approach to elicit realistic single-neuron activity, recorded from a seizing brain slice, into a pharmacologically isolated dissociated neuron. 93These studies exploited this model to test the effects of anti-seizure compounds on epileptiform bursts at the cellular level, revealing novel mechanisms of the ASM carbamazepine.As with other methods, dissociated cells can be prepared from in vivo genetic epilepsy models, retaining the underlying pathological gene variants for in vitro studies.Glioneuronal cultures 94,95 exposed to zero magnesium CSF and imaging techniques have been used to assess reactive oxygen species (ROS) pathways during seizure activity-an example of interrogating the impact of metabolic pathways on epileptiform events.

| Improved translation of in vitro models-Use of human-based preparations
Our discussion, so far, has focused mainly on the use of animal-derived tissues for in vitro epileptic seizure models.This raises a key concern in the context of human ASMs.Although rodent brains provide reasonable representations of human brain circuitry, there are clearly substantial differences between the two that render direct comparisons difficult to interpret.Solutions to this aspect lie in the use of human-derived tissues in vitro to model seizures and epileptiform activity.7][98][99] Briefly, human somatic cells such as skin fibroblasts or peripheral blood mononuclear cells are harvested from donated patient samples and reverted to an induced pluripotent state via overexpression of a series of four transcription factors (OCT4, SOX2, KLF4, MYC; known as Yamanaka reprogramming factors 99 ).These iPSCs can then be differentiated into the desired cell type.A key advantage of iPSCs is that they retain the human genome of the original cell (assuming no somatic mutations).This is particularly advantageous in the context of genetic diseases, where exact genetic variants are preserved.iPSC-derived cells are then grown in culture and have applications similar to dissociated rodent neuronal cultures, discussed above.These cells can also be cultured in three-dimensional scaffolds to grow realistic brain organoids.Using this method, it is possible to recapitulate complex brain anatomy, including cortical columns. 100Organoids may offer a unique model to study the impact of genetic variants on brain development, which is of particular relevance to epilepsies associated with brain malformation.Organoids develop at approximately the same rate as the human brain in utero, so the maturation of organoids to the point where synchronous network activity is detected involves several months of culture. 101Indeed, brain organoids have been derived from patients with developmental epileptic encephalopathies including Rett syndrome 102 and variants of the WW domain-containing oxidoreductase gene. 103hese organoids recapitulated key disease phenotypes such as single-cell transcriptomic dysregulation and epileptiform network activity.More recently, brain organoids were used in the preclinical development of a new gene therapy strategy, 104 emphasizing their ability to enhance translational research in epilepsy.Despite the human derivation of iPSCs and organoids, some doubts remain.Notably, it is unclear how accurately iPSCs in culture mature and thus whether they represent functional human neurons.In the case of organoids, these are devoid of external stimuli, which might impact the natural physiological development of real human brain systems.Therefore it should be considered that organoids might not fully capture all aspects of human neurodevelopment.
Another solution is offered by the use of surgically resected human brain tissue. 48,105For some patients with focal DRE, surgical removal of their seizure-onset zone is a viable clinical option to reduce or remove their seizure burden. 106With informed patient consent, this tissue can be collected and processed in the same way as rodent brain tissues.This is arguably the ideal model of DREthe exact seizure-onset zone that could not be treated with ASMs.Human brain specimens can be sectioned for acute slice recordings and typically have longer viability than rodent slices.An adapted slice storage method can be used to extend their viability for up to 72 h, 107 permitting the screening of ASO-based therapies in human brain. 108For longer term use, including the application of viral vectors, human brain slices can be maintained in organotypic culture. 109,110This approach has been extended recently to the study of the developing human brain. 111Resected human slices do suffer some disadvantages.Patient donors and tissue samples are highly heterogeneous with key variables including patient age, sex, ASM history as well as the anatomic location of the seizure-onset zone and the viability of the resected tissue.Moreover, heterogeneity of slices from within samples in terms of their pharmacological response to anti-seizure drug application and effect on electrophysiological recordings has been reported. 112There is also no direct control for such tissues, although non-epileptic cortical regions, resected for access to deeper structures, are often used.In light of the questions associated with the interpretation of the epileptic human brain slice models, it should be mentioned that such slices often do not exhibit spontaneous seizures but do in some cases generate spontaneous inter-ictal activity, [112][113][114][115][116][117] which may mainly mirror γaminobutyric acid (GABA)ergic processes.

| How well do in vitro assays predict in vivo and clinical usefulness of ASMs?
Several in vitro models have been used to test ASM effects with a view to translating findings from preclinical testing to human clinical trials.In addition to reducing animal use, this approach could also aid with guiding drug dosing in trials.Testing ASM efficacy in vitro also allows a clean dissection of seizure modifications in the absence of potential pharmacokinetic complications that arise from in vivo systems approaches.Furthermore, the use of in vitro models facilitates quick screening of drugs individually or in combination. 118This combinational approach may be useful to help gain better a quantitative understanding of "rational polytherapy," which would be more complex in vivo.
One in vitro model that has been used for ASM screening is the "Mg 2+ -free" or "zero Mg 2+ " model.Removal of magnesium from the bathing medium results in epileptiform events in cortical and hippocampal slices. 30However, these events evolve over time to become short recurrent discharges (termed late recurrent discharges [LRDs]).The evolution of washout of Mg 2+ has been shown to reduce GABA-mediated inhibition 119 and the LRDs are likely to reflect disinhibition within neuronal networks.Of interest in the context of drug screening, the LRDs are frequently pharmacoresistant to standard ASMs 120 such as phenytoin, phenobarbital, and midazolam.The insensitivity to benzodiazepines has led to the suggestion that zero Mg 2+ LRDs may be considered an in vitro model of SE.Recent work using organotypic hippocampal slices 121 has demonstrated that persistent SE-like activity is associated with a reduction in GABA A receptor conductance and chloride extrusion capability.Alternatively, the LRDs may reflect interictal activity, which can also persist in the presence of ASMs. 122In contrast, the potassium channel blocker 4-aminopyridine (4-AP) produces stable occurrence of seizure-like (ictal) events and associated inter-ictal events in several cortical (entorhinal, perirhinal, amygdala) and hippocampal regions of brain slices. 122,123The 4-AP model has been used to capture the efficacy of newer-generation ASMs including lacosamide, zonisamide, and levetiracetam, 124 as well as classic ASMs such as CBZ, topiramate (TPM), and valproate (VPA). 122The elevated potassium (high K + ) hippocampal slice model 28 has been used to screen the efficacy of the ASMs lamotrigine, phenytoin, and valproate, both individually and in combination. 118his exemplifies the use of in vitro models to interrogate polytherapy.Many people with drug-resistant seizures experience polypharmacy, and such experiments are therefore key to fully reflect the setting in the epilepsy clinic.Finally, it should be noted that the simplicity of in vitro models may also be a drawback.Putative ASMs may be efficacious in vitro but not in vivo, 125 presumably due to peripheral metabolism and the BBB in vivo.They may also have different effects in ex vivo tissue from chronically epileptic animals, which likely have different mechanisms of ictogenesis when compared with naïve tissues in which we induce seizures typically by modifying the extracellular environment.

| Does the use of in vitro models in experimental studies aid in identifying ASM mechanisms?
A key advantage of in vitro preparations for mechanistic drug discovery is the ability to visualize brain activity at the cellular level.A fundamental technique in this regard is patch-clamp electrophysiology, 84 allowing the researcher to precisely record the activity of individual identified cells and comparing their activity in the presence and absence of ASMs.Current clamp recordings can reveal changes in passive biophysical properties such as resting membrane potential and input resistance, as well as active firing properties including action potential waveforms and firing rates.Voltage clamp recordings can also be used to either monitor inputs to the individual neuron or to isolate the currents from specific voltage-gated ion channels.These approaches can be combined to demonstrate the effects of ASMs on individual biophysical processes, which ultimately change the properties of individual neurons and in turn neuronal networks.Using extracellular recording of slice seizure models combined with simultaneous patch clamp of individual neurons in the network, it is possible to resolve both single cell and network level activity at the same time, and so to infer the relationship between the two. 29Another advantage of in vitro approaches for understanding drug mechanisms is the ease and speed with which ASMs can be applied to preparations.This allows the researcher to record a baseline measurement (without ASM), rapidly apply the ASM to observe its effect, and then wash it out.This is less straightforward with highly lipid-soluble agents such as cannabinoids, which progressively partition into cellular membranes during exposure and wash out very gradually. 126However, for most drugs, this means that it is straightforward to assess the activity of the same cells or networks in the presence and absence of a well-controlled concentration of ASM.This is extremely challenging in vivo, where ASMs must typically be applied systemically.This means that ASMs take a long time to reach the brain, and that the researcher has much less control over the active concentration of the ASM reaching the brain.
It is important to understand the mechanisms of ASMs in order to identify their possible indications or contraindications in the epilepsy clinic.For example, in vitro studies show that ASMs including CBZ and phenytoin inhibit voltage-gated sodium channel function as their main antiseizure mechanism. 127This mechanism means that they are strong frontline ASMs, but also contraindicated their use in epilepsies associated with impaired firing of inhibitory neurons, for example Dravet syndrome. 128There are several notable examples in drug discovery where in vitro interrogation has revealed unexpected mechanisms of ASMs, often different from the mechanism intended in their design.For example, gabapentin and pregabalin are rationally designed analogues of GABA, which were intended to treat epilepsy by increasing GABA availability and so network inhibition in the brain.However, it was later discovered that despite their design, these ASMs have no clear impact on the GABAergic system, instead mediating their anti-seizure effects by targeting voltagegated calcium channels.

| Developments and opportunities using human-derived in vitro models
A key limitation of most classical in vitro models is that they are typically derived from animal tissues and do not reflect the complexity of human neurons and brain networks (for example 129,130 ).Therefore, antiseizure molecules that are efficacious in animal-based models may not address the full molecular, biophysical, genetic, and anatomic complexity of human seizures and, ultimately, may not work in humans the same way they do in animalbased models.Typically, this may not be evident until a time-consuming and costly pre-clinical pipeline has been completed.This may include rodent in vivo studies, followed by Good Laboratory Practice (GLP) grade studies in another species (possibly dog or non-human primate) and then even first-in-human trials, before lack of translation between animal and human-derived tissues becomes apparent.This is a hugely costly strategy in terms of time, money, and also ethically, owing to the animals used.Recent advances in the use of human-derived tissues for translational epilepsy research are invaluable to bridge this gap 105,108,109,131 and to provide relatively cheap and fast human-based screens of new therapeutics, before they begin further pre-clinical validation.For small molecules, acutely resected human epileptic tissue can be used as an in vitro seizure screen.For approaches with slower mechanisms of action, such as nucleic acid or virally-delivered therapies, these slices can be maintained in organotypic cultures. 110For approaches targeting the genetic epilepsies, advances in iPSC and gene-editing technology, as well as brain organoids, provide strong platforms that clearly mirror the human genetic pathology.Of course these strategies have limitations (they are not whole brain preparations) and they must be used strategically in parallel with in vivo animal models, or with potentially higher throughput animal-based in vitro models.Nevertheless, human-derived in vitro models can provide an early screen of drug efficacy in human brain and should be integrated into pre-clinical pipelines at an early stage.

| In vitro optical approaches to translational epilepsy research
Optical technologies are rapidly advancing and provide key new tools in translational drug discovery for epilepsy.Broadly, optical approaches can be used to modulate neuronal activity (optogenetics 132 ) and/or as a readout of neural activity. 133The combination of these techniques is referred to as "all-optical" neurophysiology. 134This has certain advantages over electrophysiological techniques.For example, when using light as a readout it is straightforward to measure the cellular activity of many neurons simultaneously.Optical tools to read out neuronal activity can include voltage indicators, 133 calcium sensors, 135 or fluorescent reporters, which are sensitive to neurotransmitters such as glutamate 136 and GABA. 137This is much more laborious using electrophysiology and, in the case of patch-clamp approaches, it is more technically challenging and limited by the number of micropipettes that can be physically placed into the preparation.Indeed, closedloop systems have been developed that use optogenetic mediators to manipulate circuit-level excitability in response to input from electrically recorded network activity. 138,139The authors of the study 138 argue that such an approach provides benefit for mechanistic studies and also offers translational benefit, although others have noted that challenges may remain with delivery of light to deep brain structures in a clinical setting. 140This concept was developed initially using in vitro in brain slice preparations, 138,139 demonstrating the utility of in vitro approaches for developing optogenetic-based therapies.
We believe that the application of these rapidly advancing technologies to human tissue-based in vitro preparations will offer substantial translational and mechanistic insights into epilepsy.Typically the use of these optical technologies requires genetic transduction of fluorescent sensors into neurons.Therefore, transduction into human brain preparations would either require the use of fastacting viral vectors (which have translational limitations), or human brain organotypic slice cultures, which permit viral expression using adeno-associated virus (AAV). 109

| CONCLUSIONS
A variety of in vitro preparations can be used to study epilepsy and anti-seizure compounds.These preparations range from individual cells through to brain slices and whole brain preparations.In vitro preparations are amenable to a number of recording techniques, both electrophysiological and optical, at the single cell and network levels.The increasing use of human cell-and tissue-based in vitro techniques can enhance the clinical relevance of the approaches.Given the relative convenience of these preparations, different in vitro techniques can be combined into a battery of complementary approaches to enhance their utility in pre-clinical epilepsy research.In summary, we propose a series of recommendations (Box 2) to guide the careful use of in vitro seizure

2. 1 |
Acute brain slicesAcute brain slice preparations refer to sections of brain tissue obtained immediately following euthanasia of an F I G U R E 1 Rodent and human-based in vitro models offer a variety of approaches for capturing various aspects of seizures and epilepsy.Preparations range in scale may effectively capture whole networks and/or isolated neuronal systems.A well-designed combination of in vitro models can form effective and high-throughput screening platforms, facilitating the prediction of the clinical usefulness of ASMs and their mechanisms of action.4-AP, 4-aminopyridine; ASM, anti-seizure medication; BIC, bicuculline; iPSC, induced pluripotent stem cell; KA, kainate; NMDA, N-methyl-d-aspartate; PTX, picrotoxin; PTZ, pentylenetetrazole.