How do we choose the appropriate animal model for antiseizure therapy development?

1. Introduction

Over 20 antiseizure drugs (ASDs) for the symptomatic management of the seizures of epilepsy have been brought to the clinic on the basis of efficacy in rodent acute and chronic seizure models. Nonetheless, over 30% of patients remain pharmacoresistant [1]. Many ASDs also carry significant adverse effects liability that can severely limit clinical use, particularly in pediatric and elderly patients. Preclinical models that better represent the patients who are still in need of effective therapies should thus be prioritized in initial ASD identification to advance potentially better tolerated, mechanistically-novel therapies.

2. Approach to drug development for symptomatic seizures

The wealth of acute and chronic seizure models available today makes the selection of the most appropriate approach to identify a new ASD challenging and, at times, even overwhelming (Figure 1). Further, ASD development has historically been performed in two phases: Identification and Differentiation, with a primary emphasis on phenotype-based (e.g. seizures), instead of target-based, drug development. For the purposes of this editorial, we will primarily focus on moderate- to high-throughput identification in a phenotype-based drug screening program; i.e. models that use behavioral seizures as the primary outcome measure to demonstrate anticonvulsant efficacy of an investigational agent. The majority of approved ASDs are effective in two models of acute (i.e. single event) seizure [2]: maximal electroshock (MES) and subcutaneous pentylenetetrazol tests (scPTZ; Table 1). These models are relatively easy to implement and amenable to moderate- to high-throughput screening of large libraries of compounds in a short timeframe. As a result, the MES and scPTZ models are positioned earlier in the development process than the resource-intensive models with spontaneous recurrent seizures (SRS; Figure 1). Kindling models with chronic evoked seizures have also been incredibly useful to identify the remaining currently available mechanistically-distinct compounds (e.g. levetiracetam [3]), as well as inform on the potential for adverse effects liability in an epileptic substrate (i.e. D-CPPene [4]), which has certainly benefited patients with epilepsy (Table 2). Nonetheless, these models have only identified therapies that treat a single symptom: seizures. Despite more than 20 clinically-available ASDs, the percentage of drug-resistant patients has remained unchanged; the approach to ASD development clearly must be revised.

Figure 1.

Approach to the identification and differentiation of promising antiseizure drugs.

Table 1.

Common preclinical models of acute and chronic seizure in use for early ASD development, as well as innovative models of pharmacoresistant chronic seizures.

Animal Model	Acute vs. Chroni c Seizur es	Induction Protocol	Seizure Endpoints	Pharmacoresi stant Model?	Translational Relevance	ASD Identified to Validate the Model	Effective ASDs at Non-Motor Impairing Doses
Maximal Electroshock (MES)	Acute	60 Hz 0.2 second alternating 50 mA current (mouse) or 150 mA (rat)	Tonic hindlimb extension seizure	No	Model for generalized tonic-clonic seizures; highly reproducible	PHT	PHT a; CBZ a; LTG f; VPA a,f; PB a; FBM a; GBP a,f; TPM f; EZG c; CBD d
Subcutaneous Pentylenetetrazol (s.c. PTZ)	Acute	Subcutaneous injection of chemoconvulsant pentylenetetrazol at dose known to elicit seizure in >97% of rodent strain (85 mg/kg in male CF-1 mice from Charles River)	Clonic spasms	No	Some translational relevance to absence seizures	ESM	ESM a; VPA a,f; BZD a; EZG f; FBM a; GBP a; PB a; TGB f; VGB f; CBD d
Corneal Kindled Mouse (50/60 Hz; traditional)	Chronic	50/60 Hz 3 mA 3 second stimulus delivered to anesthetized corneas	Presence of Racine stage seizures and/or mean seizure score	No	Model of chronic secondarily generalized focal seizures with clear behavioral and neuropathological deficits	Potentially LEV	PHT a; CBZ a; LTG f; LEV f; TGB f; EZG c,f; VGB f; GBP a; PB a,e; FBM a; VPA a,c,e,f; BDZ a,c,e,f; CBD d
6 Hz Corneal Kindled Mouse b	Chronic	6 Hz 3 sec stimulation of various current intensity (strain-dependent) delivered to anesthetized corneas for 2-4 weeks	Forelimb clonus	Yes	Model of pharmacoresistant chronic secondarily generalized focal seizures	To be determined	CLZ b; VPA b; CBZ b
Lamotrigine-Resistant Corneal Kindled Mouse e	Chronic	60 Hz 3 mA 3 second stimulus delivered to anesthetized corneas; mice receive lamotrigine pretreatment during 3-4 week kindling acquisition period	Presence of Racine stage seizures and/or mean seizure score	Yes	Model of pharmacoresistant chronic secondarily generalized focal seizures	To be determined	VPA e; DZP e; LEV e

^{a –}Based on data available within the NINDS PANAChE database, accessed 04/2019.

^{b –}Based on Leclercq et al, Epilepsy Research 2014

^{c –}Based on Barker-Haliski et al, Neurochemical Research. 2017

^{d –}Based on Patra et al, Epilepsia 2019

^{e –}Based on Koneval et al, Epilepsia 2018

^{f –}Based on Rowley and White, Epilepsy Research 2010

Table 2.

Comparative efficacy of prototype ASDs in electrically kindled rodent models of chronic seizures.

Antiseizure Drug	Amygdala- kindled rat*	Hippocampal kindled rat	Corneal kindled mouse (60 Hz)	Corneal kindled mouse (6 Hz)
Benzodiazepines	+	+ A	+ A, B, C	+ D
Brivaracetam	+	ND	ND	ND
Cannabidiol	ND	ND	+ E	ND
Carbamazepine	+	+ F	+ F	+ D
Eslicarbazepine acetate	+	ND	+ G	ND
Ethosuximide	NE	NE A	NE A, F	ND
Felbamate	+	+	+ A, E	ND
Gabapentin	+	NEA	+ A	ND
Lacosamide	+	+ G	ND	ND
Lamotrigine	+	+ F	+ C, F	ND
Levetiracetam	+	50% effectiveF	+ C, F	+ D
Oxcarbazepine	+	ND	ND	ND
Perampanel	ND	+	ND	ND
Phenobarbital	+	+	+ C	ND
Phenytoin	NE	+	+ F	ND
Pregabalin	+	+ I	ND	ND
Retigabine (ezogabine)	+	ND F	+ C	ND
Tiagabine	+	+ F	+ F	ND
Topiramate	+	NE F	50% effective F	ND
Valproate	+	+ F	+ B, C, F	+ D
Vigabatrin	+	ND F	+ F, J	ND
Zonisamide	+	ND	ND	ND

+ indicates compounds that are effective in the kindling model. NE indicates “not effective”. ND indicates “no published data”. Bold text indicates compounds that are ineffective in the kindling model. Italic text indicates compounds that demonstrate a pharmacological profile different from the clinically validated amygdala-kindled rat.

^*Efficacy data of amygdala-kindled rats adapted from data presented in Table 2 of Klein et al, Epilepsia 2017 [18], with permission of John Wiley & Sons.

^{A –}from NINDS PANAChE database (https://panache.ninds.nih.gov), accessed June 2019.

^{B –}from Barker-Haliski et al, Neurochemical Research 2017

^{C –}from Koneval et al, Epilepsia 2018

^{D –}from Leclercq et al, Epilepsy Research 2014

^{E –}from Patra et al, Epilepsia 2019

^{F –}from Rowley and White, Epilepsy Research 2010

^{G –}from Potschka et al, Epilepsy Research 2014

^{H –}from Stohr et al, Epilepsy Research 2007

^{I –}from Vartanian et al, Epilepsy Research 2006

^{J –}from Klitgaard and Matagne, Epilepsy Research 1998

ASD identification tailored to specific patient groups may rapidly advance more mechanistically-novel pharmacotherapies for epilepsy, while we await the identification and validation of additional models. Effort is ongoing to address pharmacoresistant epilepsy with more late-stage differentiation models [5], but there is still opportunity to integrate models with pharmacoresistant chronic seizures even earlier into the ASD identification process. Until a new compound is brought forth to the clinic solely on the basis of efficacy in a preclinical model that is not MES, scPTZ, or kindling, the field will fail to validate any other preclinical model. Improved understanding of the processes and molecular targets underlying specific epilepsy syndromes clearly suggests that a new model may one day be clinically-validated. However, this validation is likely still years to decades away, while some revisions to current practice can better address the outstanding needs of patients today.

3. Expert Opinion

3.1. Pharmacoresistant chronic seizure models should be prioritized

Chronic seizure models offer the most etiologically-relevant platform on which to accurately replicate clinical epilepsy and are thus deserving of more use earlier in ASD identification. Unlike the acute MES/scPTZ tests that are most often conducted in young adult, neurologically-intact rodents, chronic seizure models reproduce the neuropathology and behavioral comorbidities associated with epilepsy. Chronic seizure models may also identify promising compounds that exert an anticonvulsant effect through novel mechanisms (e.g. inflammation [6]). Refinements in computational power have certainly improved our ability to implement preclinical models with SRS detected by EEG, but the fact remains that initial identification of treatments for symptomatic seizures – i.e. ASDs – must rely on models suited to moderate- to high-throughput testing to accommodate large libraries of compounds. Models with SRS detected by EEG frankly cannot support lead selection and optimization necessary to advance the most promising compounds and as such, implies that more “low-tech” models will still comprise a large proportion of the initial phases of ASD development (Figure 1). Pharmacoresistant chronic seizure models that are amenable to moderate- to high-throughput screening should be further prioritized for ASD identification.

The 50/60 Hz corneal kindled mouse (CKM) is a well-established model of kindling suitable to moderate-throughput drug screening [7]. Unlike the amygdala-kindled rat, CKM have minimal technical requirements to quickly generate ample numbers of animals to screen large libraries of compounds [7,8]. CKM demonstrate a pharmacological profile similar to amygdala-kindled rats, and may, in fact, more closely align with amygdala-kindled rats than previously reported alignment with hippocampal-kindled rats [8] on the basis of documented sensitivity to gabapentin, levetiracetam, and topiramate (Table 2). While the amygdala-kindled rat was validated with the identification of levetiracetam [9], the 50/60 Hz CKM represents a model of chronic seizures that exhibits behavioral [10] and neuropathological [11] changes of epilepsy, and is suitable for moderate-throughput screening unlike the amygdala-kindled rat. However much like the amygdala-kindled rat, CKM are not innately pharmacoresistant [8]. Modifications have produced a 6 Hz corneal kindling protocol that may be more pharmacoresistant [12], albeit the published pharmacological profile of this model is quite limited (Table 1). More recently, a lamotrigine-resistant CKM has demonstrated a broad and highly pharmacoresistant profile, including resistance to retigabine and carbamazepine (Table 1 [10]). Specifically, this model is elicited by the chronic administration of lamotrigine early in the epileptogenesis process, similar to a patient receiving lamotrigine monotherapy upon epilepsy diagnosis. Seizures are still evoked in the presence of lamotrigine, which mimics poor seizure control in man, and the animals subsequently develop a highly-drug resistant kindled seizure. The lamotrigine-resistant CKM [10], instead of naïve CKM, should thus be positioned in parallel to frontline validated acute seizure tests for ASD Identification. Addressing drug-resistant epilepsy earliest carries the potential to advance ASDs with pharmacological profiles distinct from current standards-of-care.

3.2. Models for ASD identification should align with clinical need.

Historical ASD development has also not adequately addressed the needs of all patients with epilepsy primarily because development is often conducted in neurologically-intact male rodents equivalent to a 20- to 40-year-old human adult [2]. Better representation of patient demographics in preclinical models is thus an underexploited strategy that can be deployed today to identify impactful ASDs. The very young (pediatric patients) and very old (geriatric patients) themselves require more effective and better tolerated therapies. Use of established seizure models in rodents harboring specific genotypes associated with age-related neurological disorders could identify beneficial therapies. Further, such an approach would shift our primary reliance on phenotype-based screening to a hybridized target-based drug screening platform, which may even then identify therapies that are broadly applicable to other epilepsy syndromes, in general.

Much effort is currently dedicated to the identification of therapies for pediatric epileptic encephalopathies. These syndromes are devastating but incredibly rare; e.g. Dravet’s syndrome occurs in an estimated 1:15,700 live births [13]. Nonetheless, numerous preclinical models of Scn1a^+/− variants (the gene implicated in Dravet’s syndrome) have been developed to replicate the age-dependent clinical course of Dravet’s syndrome, but these preclinical models are not clinically-validated. There has yet to be an ASD brought forth on the basis of efficacy in a Scn1a^+/− model that was otherwise missed in an established seizure model in adult rodents. As an example, cannabidiol (CBD) gained FDA approval for Dravet’s syndrome in 2018. While CBD attenuates the incidence of SRS in Scn1a^+/− mice [14], CBD also demonstrates broad antiseizure efficacy in the MES, scPTZ, and kindling models in adult, wild-type rodents [15]. This one example suggests that an investigational compound that is intended for Dravet’s syndrome may just as easily advance to clinic on the basis of efficacy in the MES or kindling models in wild-type rodents at significant cost-savings. Until one of these pediatric genetic epilepsy models alone identifies a compound that is effective in patients with epilepsy, these models will remain late-stage differentiation tools that are primarily useful to increase confidence in an otherwise promising ASD.

At the opposite end of life, the prevalence of epilepsy is 1.5% among patients 65 or older; twice as frequent as in younger adults [16]. Patients with the most common neurological disorder of aging, Alzheimer’s disease (AD), are also increasingly appreciated to experience nonconvulsive focal seizures [17]. As a result of the near exclusive use of adult, neurologically-intact rodents for historical ASD development, our preclinical understanding of the efficacy and tolerability of ASDs to guide the management of seizures in elderly patients, and those with AD, is under informed. There is now compelling justification to extend ASD development into aged preclinical models because of the combined recognition that seizures arise in AD and the worldwide growth of individuals living well-beyond their 70’s.

Very little effort is presently devoted to ASD development in aged rodents. Fortunately, the available rodent models that are often used for ASD identification (MES, kindling) can also be readily applied to aged rodent models of aging disorders (e.g. AD). Seizure threshold increases in an age-dependent manner; adjusting the electroconvulsive stimulus for the MES test in aged rodents in merely requires modifications of the stimulation intensity. Further, unlike many pediatric epilepsies that arise due to genetic variants, age-related epilepsy is more frequently a result of factors that do not arise due to ion-channel variants, which may also provide valuable insight into the mechanisms underlying ictogenesis and epileptogenesis. Application of well-established seizure models to aged rodents with appropriate genetic mutations associated with aging disorders (e.g. AD) could even identify therapeutic targets or treatments that may carry broad applicability to other seizure disorders, in general. Elderly patients with epilepsy represent the fastest growing population with this condition; any effort to implement aged rodents for ASD development would meaningfully impact the pharmacotherapeutic management of seizures in the elderly. Altogether, more strategic integration of appropriately-aged preclinical seizure and epilepsy models into ASD development practices thus represents an untapped opportunity to advance more beneficial therapies for differentiation studies.

ASD development practice should increasingly emphasize special populations, including pharmacoresistant patients, and the very young and very old. Lamotrigine-resistant CKM is a frontline screening model that exhibits the pharmacological profile and throughput capacity to support screening of large compound libraries in an etiologically-relevant preclinical model of pharmacoresistant chronic seizures. There is also significant opportunity to apply the standard models that have been instrumental to the historical identification of ASDs to aged rodents. It should be clear that this editorial has, however, only discussed strategies to improve the initial identification of potential ASDs (Figure 1), nor has it addressed the need for disease-modifying agents. This is not to underscore the value of EEG monitoring of SRS in the late-stage differentiation of promising agents. Differentiation studies improve confidence that an agent may ultimately be efficacious for the symptomatic treatment of seizures in the patient with epilepsy, but no SRS model has yet to be clinically-validated and these models are not currently suited to moderate- to high-throughput ASD identification. As a result, use of the MES, scPTZ, or kindling models in selected genetic models of age-related neurological conditions represents a presently untapped opportunity for the field to integrate more etiologically-relevant models earlier into the ASD development process, which may also ultimately inform on the processes underlying ictogenesis and epileptogenesis, in general.