Drug-resistant epilepsy has remained a problem since the inception of antiepileptic drug development, despite the large variety of antiepileptic drugs available today. Moreover, the mechanism-of-action of these drugs is often unknown. This is due to the widespread screening of compounds through animal models. We have taken a different approach to antiepileptic drug discovery and have identified a biochemical pathway in Dictyostelium discoideum (a ‘slime mould’) that may relate to the mechanism-of-action of valproate, one of the most commonly used and effective antiepileptic drugs. Through screening in this pathway, we have been able to identify a whole host of fatty acids and fatty acid derivatives with potential antiepileptic activity; this was then confirmed in in vitro and in vivo mammalian seizure models. Some of these compounds are more potent than valproate and potentially lack many of the major side effects of valproate (including birth defects and liver toxicity). In addition, one of the compounds that we have identified is a major constituent of the ketogenic diet, strongly arguing that it may be the fatty acids and not the ketogenesis that are mediating the effect of this diet.
Over 50 million people worldwide have epilepsy. Approximately 30% will not become seizure-free despite optimum treatment . People with drug-resistant epilepsy have a decreased quality of life, an increased incidence of morbidities (including depression and psychosis, as well as physical diseases such as chest infections, cancer and heart disease) and an increased mortality (in some over 1% per year). New more efficacious and better-tolerated treatments are sorely needed.
The first effective antiepileptic drugs, bromides and phenobarbital, were discovered largely through serendipity, and it was not until 1937 that a cat model of seizures (the maximal electroshock model) devised by Merritt and Putnam was used to screen over 700 compounds, resulting in the discovery of the anti-seizure effect of phenytoin [2,3]. This discovery heralded the beginning of a new era in antiepileptic drug discovery and the National Institute of Neurological Diseases (U.S.A.) later started an antiepileptic drug-screening programme which has screened over 25000 compounds through animal seizure models. The problem with this approach is that there is often no understanding of the mechanisms of action of even effective compounds and, instead, it relies on observations that the compound changes the seizure threshold in animal models that are often quite divorced from the human condition. It was in this fashion that a group of French scientists identified valproic acid as a potential epilepsy treatment while using it as a solvent to dissolve a series of potential new antiepileptic drugs . Subsequently valproic acid (and its salt sodium valproate) has become one of the most widely used antiepileptic drugs with a broad spectrum of action (it is effective in most types of epilepsy). The use of valproate is, however, limited by potential adverse effects, including liver failure, cognitive slowing and, importantly, teratogenicity (fetal abnormalities including decreased IQ in children exposed to valproate in utero) . Intriguingly, despite years of research, valproate's mechanism-of-action has remained unclear, leading to a failure to produce a drug with valproate's proven efficacy, but without its side effects. With this in mind, we set out to determine the mechanism-of-action of valproate and to use this to screen for other compounds with improved efficacy and the potential to have a better side-effect profile (in particular to lack the teratogenic risks of valproate).
Cellular slime mould (social amoeba)
Although originally thought to be a part of the fungi family, cellular slime moulds belong to the Ameobozoa kingdom, where single cells multiply by binary fission, and which, under adverse conditions (in particular starvation), coalesce to form fruiting bodies that facilitate dispersal. One such social amoeba, Dictyostelium discoideum (Figure 1), has proven to be an excellent model for understanding various aspects of cellular signalling in human disease because many of its enzymes and biochemical pathways are shared with mammalian systems [6,7]. Dictyostelium has important advantages over other organisms, including rapid and easy genetic manipulation (the organism is haploid and is easily propagated) and screening of mutant libraries in pharmacogenetic approaches to understanding drug action . This has enabled researchers to determine the function of defined proteins and drugs in cell function and development. Dictyostelium can also be used as a rapid-throughput screen of drugs/chemicals, because the behaviour of cells can change rapidly following drug exposure. Moreover, this can be used to dissect the biochemical and molecular mechanisms underlying drug effects.
Dictyostelium has been used as an animal-replacement model to identify new compounds for seizure control
The wider application of Dictyostelium as a model in drug-related research has been hindered by the concern that discoveries in this model are unlikely to translate to other more complex systems (such as the mammalian brain during seizures). However, Dictyostelium has played a considerable role in the testing of anti-cancer chemotherapies, pathogenesis of bacterial infection, drugs used in osteoporosis and drugs used in bipolar disorders (including valproate) .
Valproate and Dictyostelium
Valproate, when applied to Dictyostelium at concentrations within the therapeutic range, blocked the formation of the fruiting body (i.e. development) . This effect was mediated by a rapid block in cell movement, which occurred with a similar delay to that of valproate's effect on seizure activity . However, this observation did not identify the precise biochemical action of valproate, nor did it directly relate this to valproate's effect on seizures.
Dictyostelium's movement is mediated by cell-surface receptors that detect cAMP and trigger activation of an intracellular signalling cascade involving the production of a key family of molecules called phosphoinositides. Valproate rapidly reduced phosphoinositide production in moving Dictyostelium cells [9,10]. Importantly, the phosphoinositide pathway is a phylogenically conserved pathway and is present in not only amoeba, but also humans. The next step was to determine the precise point of the pathway at which valproate acted. This problem is challenging in mammalian systems in which enzyme inhibitors are often not specific and rarely completely effective, but becomes significantly more tractable in Dictyostelium, because of the ability to rapidly delete any gene of our choice provided that the protein product is not vital. A pure strain of the knockout mutant can be grown and tested for changes in its response to valproate.
We first examined a family of enzymes that are most commonly associated with the production of phosphoinositide, called PI3Ks (phosphoinositide 3-kinases) . In humans, there are 14 PI3K (catalytic and regulatory) proteins present in cells, thus it is challenging to determine which, if any, of these proteins are the target of valproate. In Dictyostelium, there are six genes encoding these proteins, and all of these genes were deleted in a single cell line, which was then tested with valproate. Valproate blocked phosphoinositide production in these cells in the absence of the PI3K enzymes, indicating that valproate did not work through changing the activity of these enzymes. This has helped us to narrow down the possible targets of valproate, but its precise target, as yet, remains elusive.
Despite not having a precise mechanism, the observation that valproate reduces phosphoinositide production provided a method of developing a rapid-throughput assay to assess whether there are compounds with a similar mode of action [10,11]. Restricting this assay to compounds that are chemically similar to valproate (a branched-chain fatty acid) enabled us to explore the correlation of the structure of branched- and straight-chain fatty acids in this assay to establish the structure–activity relationship of these compounds. The observation that only small changes in the structure (e.g. the addition of an extra carbon) has a profound effect on the compound's activity indicated that these compounds were not working through some non-specific effect of fatty acids on this pathway, but that a more precise interaction was taking place.
We thus identified a group of compounds that are effective in a non-mammalian assay, which relates to a mechanism-of-action of valproate . Moreover, we had established a structure–function relationship for efficacy in this assay. The next task was to determine whether these compounds are effective against seizure activity. There are number of reasons that there may not be a direct correlation between activity in the Dictyostelium assay and anti-seizure activity; not least, some of the compounds may have additional mechanisms that are pro- or anti-epileptic. Rather than move directly to animal seizure/epilepsy models, we wished to reduce animal usage and potential suffering by using a high-throughput in vitro model of seizure activity. We therefore turned to ex vivo hippocampal slice models of seizure-like activity. These models involve the induction of continuous seizure-like activity by increasing excitability through the administration of a convulsant to, or by changing ion concentrations in, 300–400-μm-thick hippocampal slices kept ‘alive’ through perfusion with oxygenated artificial cerebrospinal fluid . This seizure-like activity is resistant to a number of established antiepileptic drugs and has therefore been proposed to be a possible screen for drugs that may be effective in drug-resistant epilepsies. Importantly valproate shows some efficacy in these models at clinically relevant concentrations . Using these models, we were able to reduce the number of potential compounds that would be tested in vivo. Since one of the primary drives of our research was not only improved efficacy but also reduced side effects compared with valproate, we also screened our potential compounds in in vitro assays of liver toxicity and histone deacetylation inhibition (a possible contributor to the teratogenic effects of valproate) . We then confirmed efficacy for a few of these compounds in an in vivo model of prolonged seizure activity (status epilepticus) [11,12].
Epilepsy and coconuts
One of the compounds that we identified in our Dictyostelium and in vitro seizure assays that had a particularly marked antiepileptic effect was decanoic acid. Decanoic acid and octanoic acid (which was not effective in our seizure models) are the main constituents of coconut oil, which is used as the basis for medium-chain fatty acids in the ketogenic diet used to treat epilepsy. The ketogenic diet, a low-carbohydrate and high-fat diet, came to prominence in the treatment of epilepsy in the 1920s. It had been proposed that the production of ketones through the metabolism of fats was the underlying mechanism of this diet ; however, ketone production correlates poorly with the antiepileptic effect of the diet, and ketones have variable efficacy in seizure models. Indeed, in their initial experiments in which they first identified phenytoin as an effective antiepileptic drug, Putnam and Merritt  found that ketones only had antiepileptic effects at toxic doses. The discovery of phenytoin resulted in a decrease in the use of the ketogenic diet and the death knell for the diet was sounded by the discovery of valproate. It was not until the 1990s that there was a resurgence of interest in the diet for use in children with refractory epilepsy, and the first randomized controlled trial in children with epilepsy was reported in 2008 . However, the mechanism-of-action of the diet has remained unclear. Moreover, the diet is often poorly tolerated and so rarely used in adults. The observation that decanoic acid is frequently used as a constituent of the diet and that there are high levels of decanoic acid in the blood of children on the diet  indicate that decanoic acid's antiepileptic effect may be one of the main mechanisms underlying this diet, and may be a way to make a more palatable and better-tolerated diet by avoiding the ketogenesis.
As a tool for biomedical research, the social amoeba Dictyostelium provides an animal replacement model that has great advantages over traditional systems, including the ease of genetic manipulations and of developing rapid throughput drug screens. In the present review we have shown how Dictyostelium can be used to identify the mechanisms of action of antiepileptic drugs, and to use these observations to identify a new family of antiepileptic drugs with improved potency and reduced side effects compared with current treatments (i.e. valproate). This family of compounds shares a common chemical structure with fats provided in the medium-chain triacylglycerol ketogenic diet used to treat children with drug-resistant epilepsy, and implicates these fatty acids as playing a direct role in seizure control, rather than through the ketogenesis that been previously hypothesized. These discoveries pave the way to the development of new treatments and treatment approaches not only to drug-resistant epilepsy, but also, through improved side-effect profiles, to the treatment of all epilepsies.
5th Conference on Advances in Molecular Mechanisms Underlying Neurological Disorders: A joint Biochemical Society/European Society for Neurochemistry Focused Meeting held at the University of Bath, U.K., 23–26 June 2013. Organized and Edited by Marcus Rattray (University of Bradford, U.K.) and Rob Williams (University of Bath, U.K.).
This work was supported by the National Centre for the Replacement, Refinement and Reduction of Animals in Research (NC3Rs) [grant number G0900775 (to M.C.W. and R.S.B.W.)].
Compounds described in this paper have been registered under patent number WO2012069790 A1 filed by Matthew Walker and Robin Williams.