Identification of compounds with anti-convulsant properties in a zebrafish model of epileptic seizures

Epilepsy is a common neurological disorder that is frequently characterised by recurrent, unprovoked seizures that result from excessive and hypersynchronous electrical discharges in the brain. Depending on the location and extent of the abnormal electrical activity in the brain, epileptic seizures can manifest in many different ways, which can include temporary loss of consciousness, or abnormal motor activity that can range from minor involuntary movements to whole body convulsions. Many different epilepsy syndromes are recognised, each of which affects the nervous system in distinct ways, and in which seizures are prominent phenotypic components (Reid et al., 2009). Seizures can be either localized to specific parts of the brain, or distributed more broadly as generalized seizures. Epileptic seizures can affect people of all ages. They can occur in the absence of structural brain abnormalities or be a manifestation of an underlying brain lesion, such as a brain tumour or changes attributable to a head injury. Genetic factors also play important roles in many forms of epilepsy (Reid et al., 2009). A wide range of structurally diverse anti-epileptic drugs is currently available for treatment of this disorder (Stafstrom, 2010). These drugs act in a variety of distinct ways. For example, benzodiazepines act as direct agonists of GABA_A receptors, whereas carbamazepine and lamotrigine block sodium (Na⁺) and calcium (Ca²⁺) channels, the normal opening of which enables the firing of neuronal action potentials in response to excitatory neurotransmitters. Moreover, valproic acid (VPA) has been shown to inhibit the activities of histone deacetylases and GABA transaminase, as well as to reduce the production of phosphoinositides (Nalivaeva et al., 2009; Chang et al., 2012), so its therapeutic effects might result from a combination of these modes of action. Despite the wide variety of available treatments, approximately 30% of people with epilepsy fail to respond satisfactorily to first-line anti-epileptic drugs (Remy and Beck, 2006). Furthermore, many prescribed anti-epileptic drugs exhibit substantial side effects (Cramer et al., 2010). There is, therefore, an important unmet clinical need for new antiepileptic therapeutics with more specific mechanisms of action, fewer side effects and increased potency.

In order to develop new antiepileptics and understand the pathogenetic mechanisms underlying seizure disorders, animal models of epilepsy are invaluable. Pharmacological and genetic models of epilepsy have been developed in rodents, using known convulsant agents to induce seizures, as well as through the phenotypic analysis of mutations that cause seizures (Löscher, 2011). Rodent models have been used extensively both for elucidating seizure mechanisms and characterising the mechanisms of action of anti-epileptic drugs. However, the relatively high costs and labour-intensive nature of drug discovery work using rodents limit their usefulness as organisms in which large numbers of compounds can be efficiently screened to identify compounds with anticonvulsant activities. Consequently, seizure models have been developed in a range of non-mammalian organisms that are more amenable to high-throughput analysis (Baraban, 2007). Among these species, the zebrafish is emerging as a pre-eminent model vertebrate for the in vivo analysis of many developmental and disease mechanisms. Moreover, within the last few years, the usefulness of this organism for in vivo drug discovery has become increasingly apparent. Several recent studies have demonstrated that the zebrafish is particularly well suited to the analysis of epilepsy mechanisms and anti-epileptic drug discovery (Baraban et al., 2005; Berghmans et al., 2007; Hortopan et al., 2010a; Hortopan et al., 2010b; Stewart et al., 2012). At 7 days of age, the free-swimming, independently feeding zebrafish larva exhibits seizures when treated with chemical convulsants (Baraban et al., 2005; Winter et al., 2008), and these seizures can be ameliorated by administration of known anti-epileptic drugs (Baraban et al., 2005; Berghmans et al., 2007). Other behavioural phenotypes can also be readily analysed in embryos, larvae, or juvenile or adult zebrafish by immersion in fish water containing neuroactive compounds such as addictive, anxiolytic or anxiogenic agents (Darland and Dowling, 2001; Kokel et al., 2010; Cachat et al., 2010; Cachat et al., 2011; Steenbergen et al., 2011; Norton and Bally-Cuif, 2010). One of the key experimental advantages of the zebrafish stems from the transparency of its embryo, which facilitates three-dimensional analysis of gene expression patterns using whole-mount in situ hybridisation. We took advantage of this technique to establish a robust and sensitive gene-expression-based in vivo assay for seizures in 2-day-old [2 days post-fertilisation (dpf)] zebrafish embryos. We then utilised this assay in a medium-throughput screen of a collection of 2000 bioactive small molecules to identify a subset of compounds with previously unknown anticonvulsant activities. This group of compounds includes molecules that could represent new candidate therapeutics for treating epilepsy.

Exposure of 7 dpf zebrafish larvae to chemical convulsants such as pentylenetetrazole (PTZ) induces seizures that exhibit many of the features of epilepsy, including clonic-like convulsions and ictal-like electrographic discharges (Baraban et al., 2005; Winter et al., 2008; Berghmans et al., 2007). In both rodents and larval zebrafish, induction of seizures is accompanied by robust expression of the synaptic-activity-dependent neuroprotective gene fos in the brain (Zhang et al., 2002; Baraban et al., 2005). By 7 dpf, zebrafish larvae swim fast, respond to a wide variety of sensory stimuli and exhibit a broad range of behaviours. However, by 7 dpf the larval integument is heavily pigmented and the nervous system is highly complex, which makes the analysis of CNS-specific gene expression patterns challenging. By contrast, 2 dpf embryos have a simple, transparent integument, which makes them amenable to whole-mount analysis of gene expression patterns in the developing CNS. Furthermore, by 2 dpf, embryos exhibit locomotor responses to extrinsic mechanical stimuli, implying the existence of neural circuits that could mediate locomotor responses of embryos to treatment with seizure-inducing chemicals. Previous studies have demonstrated that exposure of zebrafish larvae to PTZ elicited concentration-dependent increases in locomotor activity (Berghmans et al., 2007). We analysed the locomotor behaviour of 4 dpf larvae exposed to concentrations of PTZ ranging from 1.25 mM to 80 mM. Robust, concentration-dependent increases in locomotor activity were observed in response to treatment with PTZ, which was greatest after exposure to 20 mM PTZ (Fig. 1A). This concentration of PTZ was selected for all subsequent experiments. Evidence suggests that convulsants such as PTZ elicit seizures by inhibiting the GABA_A receptor (Huang et al., 2001). In situ hybridisation analysis of 48 hpf embryos for expression of the GABA_A receptor subunit genes gabra1 and gabrg2 revealed widespread expression of both genes in the brain (Fig. 1B), implying the existence of large numbers of GABA-responsive neurons in the CNS at this early stage of development. Consistent with these observations, treatment of 48 hpf embryos with 20 mM PTZ also induced a robust increase in locomotor activity characteristic of seizure induction (Fig. 1C, panel 1). To characterise the PTZ-induced increase in locomotor activity further, we microinjected a muscle-specific transgene encoding the fluorescent calcium reporter GCaMP-3 into embryos at the one-cell stage. 50 hpf embryos that transiently expressed GCaMP-3 in muscle cells were then exposed to PTZ. Within 5 minutes of exposure, intense muscular contractions could be readily detected as robust, transient increases in GCaMP-3 fluorescence (Fig. 1C), each of a duration of approximately 10 seconds, with an amplitude up to five times that of baseline fluorescence and separated by longer periods of relaxation (also see supplementary material Movie 1). These spasms are characteristic features of clonic seizures, indicating that PTZ induces neuromuscular convulsions in 2-day-old zebrafish embryos that are similar to those observed in some people with epilepsy.

Having established that convulsions could be reliably induced in 2-day-old zebrafish embryos by exposure to the seizure-inducing chemical PTZ, we sought to determine whether exposure to convulsant elicited a corresponding programme of seizure-associated gene expression. First, we investigated whether the synaptic-activity-regulated gene fos was induced by exposure of 2-day-old embryos to PTZ and another GABA_A receptor inhibitor, picrotoxin. Analysis of embryos after 30, 60 or 90 minutes of treatment with either of these convulsants resulted in the induction of fos transcripts in an intricate pattern of multiple discrete domains within the ventral telencephalon, ventral diencephalon, hindbrain, spinal cord and trunk muscle. Expression of fos was detectable in the brain within 30 minutes of administering either PTZ or picrotoxin, and had increased by 60 and 90 minutes after initial administration (Fig. 2A). Closer inspection of the expression domains of fos revealed that the robust induction of fos by PTZ in the forebrain is restricted to two distinct territories: the ventral telencephalon, including the subpallium, and the ventral diencephalon, encompassing the preoptic area, posterior tuberculum and hypothalamus (Fig. 2B). Taken together, these results confirmed that the induction of fos expression in specific regions of the developing CNS and trunk muscle of 2-day-old zebrafish embryos after treatment with convulsant compounds is a robust transcriptional response, consistent with the initiation of seizures in the CNS and the onset of convulsions in the trunk musculature. In order to determine whether PTZ- and picrotoxin-mediated induction of fos expression in these domains could be suppressed by a known anticonvulsant, we exposed embryos to the anti-epileptic drug sodium valproate (VPA) and either PTZ or picrotoxin, which resulted in the complete suppression of fos expression (Fig. 2C).

Previous studies of the transcriptional consequences of exposing in vitro cultured neurons to convulsants and other inducers of synaptic activity have demonstrated that a broad programme of gene expression is elicited, components of which might perform neuroprotective functions that limit neuronal depolarization and help prevent calcium influxes reaching excitotoxic levels (Greer and Greenberg, 2008). Among these genes, neuroprotective functions are indicated for the bHLH-PAS-domain-containing transcription factor gene Npas4 and the neurotrophic factor BDNF (Lin et al., 2008; Greenberg et al., 2009). When expression of the zebrafish orthologues of these two genes was analysed by whole-mount in situ hybridisation, both npas4 and bdnf transcripts were induced in the ventral telencephalon and diencephalon of PTZ-treated embryos, in patterns that are similar to that of fos (Fig. 3). 3 dpf zebrafish embryos that are homozygous for the recessive mind bomb mutation were previously shown to exhibit seizure behaviour and aberrantly increased expression of a close relative of the endogenous anti-convulsant and anxiolytic Neuropeptide Y gene, pyya (Hortopan et al., 2010a). Interestingly, we also observed that PTZ treatment increased pyya expression within the CNS of 50 hpf wild-type embryos, which was most readily apparent within the posterior hindbrain (Fig. 3).

Our results indicated that a programme of CNS-specific and muscle-specific gene expression accompanies the induction of neuromuscular convulsions by PTZ in 2-day-old embryos, and that this could be suppressed by co-administration of the known anticonvulsant VPA, thus providing a robust method of monitoring induction and suppression of convulsion-associated synaptic activity in the CNS. Of the genes analysed, fos was the most sensitive indicator of seizure onset. We therefore developed a high-throughput in situ hybridisation assay for fos transcription as the basis for identifying potential anti-convulsants, in a screen of 2000 known bioactive compounds in the Microsource Spectrum Collection. Embryos were first exposed to these compounds in 96-well plates, then PTZ was added, after which samples were fixed and analysed for expression of fos by whole-mount in situ hybridisation. A total of 46 compounds were identified that caused strong attenuation or extinction of PTZ-induced fos expression in the developing brain (Table 1 and supplementary material Table S1). These compounds included several natural and synthetic steroids, as well as steroid-related compounds and compounds known to interact with other types of nuclear hormone receptors. Other molecules with known antifungal, anti-inflammatory, antioxidant, vasodilatory, pesticide, herbicide or neuroactive properties were also identified. Three structurally related antifungal agents, sulconazole, bifonazole and oxiconazole, were identified among the 46 hits, along with three calcium channel blockers and two inhibitors of monoamine receptors. However, the mechanisms of action of many of the identified compounds are unknown. A subset of 12 compounds was then selected from the initial list of 46 putative anti-convulsants for in-depth analysis of their dose-response characteristics in the fos expression assay and the locomotor activity assay. These compounds were initially selected on the basis of their relatively high potency in suppressing fos expression, viability of embryos exposed to compound at 10μM, and their availability from suppliers other than the source of the compound library. Six of these compounds are known to act on the endocrine or nervous systems and include the steroids ethinyl estradiol and allopregnanolone, the calcium channel inhibitors nimodipine and nitrendipine, and the monoamine receptor inhibitors methiothepin and pimozide. The other six compounds, which were selected for further evaluation on the basis of their apparent potency in the initial screen and their structural diversity, were sulconazole, suloctidil, nerolidol, dioxybenzone, hexylresorcinol and retinyl acetate. Figs 4 and 5 show that each of these compounds suppressed PTZ-induced fos expression in 50 hpf embryos. Moreover, all compounds apart from retinyl acetate exhibited a graded, concentration-dependent inhibition of locomotor activity in the Viewpoint tracking assay (Figs 4, 5), which was confirmed by quantitative analysis of locomotor activity (Fig. 6). The most potent anticonvulsants identified in this screen are ethinyl estradiol, sulconazole, suloctidil, nerolidol, dioxybenzone and hexylresorcinol. Each of these compounds exhibited comparable potency at 8.2 μM to treatment of embryos with the clinically prescribed anti-epileptic drug VPA at a concentration of 1 mM. When the effect of each compound on apoptosis was evaluated, none induced appreciable cell death (supplementary material Fig. S1), confirming that the relatively brief exposure to these compounds did not elicit any significant cell toxicity. To determine whether the identified compounds might impair the general viability of larvae, cardiac function of larvae was monitored after 90 minutes of exposure to each compound at 8.2 μM and 24.7 μM (Table 2). At 24.7 μM, nimodipine, pimozide, nitrendipine, sulconazole and suloctidil caused a slowing of the heart rate, and nimodipine and pimozide elicited cardiac arrest in a proportion of treated larvae. However, only nimodipine appreciably inhibited cardiac function when assayed at a concentration of 8.2 μM, causing a slowing of heart rate (but not cardiac arrest) in treated larvae. Taken together, our results demonstrate that the use of an in situ hybridisation screen for fos expression is an efficient route to identifying compounds with potent in vivo anti-convulsant activities.

We sought to develop a zebrafish model of epileptic seizures that would be amenable to efficient screening of chemical libraries for compounds with anticonvulsant activities. We found that exposure of 2-day-old zebrafish embryos to the convulsant PTZ rapidly induced intense neuromuscular activity, increased locomotor behaviour and initiated a programme of synaptic-activity-induced gene expression in specific regions of the brain, all of which are characteristics of seizures in mammals. The widespread expression of gabra1 and gabrg2 in the brain of 2-day-old zebrafish embryos, together with our observation that a second, structurally distinct GABA pathway inhibitor, picrotoxin, induced a pattern of fos transcription in the CNS and trunk muscle that is similar to that induced by PTZ, further suggested that the effects of PTZ in the embryo are probably mediated by inhibition of the GABA pathway. Our results are consistent with previous observations of the seizure-inducing effects of PTZ on 7-day-old zebrafish larvae, which can be suppressed by a wide variety of known antiepileptic drugs (Baraban et al., 2005; Berghmans et al., 2007). The induction of fos transcription in response to excitatory neurotransmission, particularly as an indicator of seizure onset, has been documented extensively in mammals and in 7 dpf larval zebrafish, but we show here for the first time that the PTZ-induced fos expression pattern in 50 hpf embryos is complex even at this relatively early stage, and includes several spatially distinct structures within the brain. In rodents, transcription of fos is induced by convulsants such as kainic acid and in response to a range of other excitatory stimuli, including anxiogenic or addictive compounds, as well as behavioural stimuli such as fear-inducing or other stressful experiences. Because fos is such a versatile marker of synaptic activity in the mammalian CNS, it will be of interest to explore further its responsiveness to a similar range of stimuli in zebrafish. The recent identification of a large number of synaptic-activity-regulated genes raises the possibility that some of these genes are specifically activated in distinct neuronal subsets by distinct neurotransmitter subtypes (Loebrich and Nedivi, 2009). Such markers might also be useful in distinguishing different types of neuroactive drugs, e.g. psychostimulants and convulsants, through their effects on regulating the activity of distinct neuronal subtypes. Genetic analysis of fos function in the mouse indicates that this gene performs a neuroprotective role as part of the response to the onset of seizures (Zhang et al., 2002), although the relevant downstream target genes for the Fos transcription factor, and its mechanism of action, remain unidentified. The dramatic induction of fos transcription in the embryonic trunk muscle was surprising and has not previously been documented as a response to seizures. This abundant transcription of fos within the developing somites of convulsant-treated embryos could indicate a protective function for this gene in muscle cells, which might help to buffer these cells against the potentially damaging effects of the strong calcium influx that occurs during seizures. Our use of a muscle-specific GCaMP-3 transgene revealed that convulsant-induced calcium-influx-dependent increases in fluorescence could be readily detected in muscle cells within 5 minutes of PTZ administration, and suggests that measurement of GCaMP-3 fluorescence could provide the basis for a high-throughput in vivo screen for detecting anticonvulsant activities in the future. Equally, in situ hybridisation assays to detect fos and other gene expression, or quantitative analysis of fluorescence changes using a GCaMP-3 transgenic line, could be used for the efficient detection of seizure liabilities in compounds as part of safety pharmacology programmes (Winter et al., 2008).

In neurons, the onset of synaptic activity induces transcription of a specific set of genes encoding transcription factors, intercellular signalling molecules and peptide hormones (Loebrich and Nedivi, 2009). To date, fos is perhaps the best characterised of these synaptic-activity-regulated genes. We found that genes encoding the bHLH transcription factor Npas4, the secreted intercellular signal Bdnf and the stress hormone PYYa exhibited increased transcription in the CNS of PTZ-treated embryos. Npas4 has previously been shown to promote the formation of GABAergic inhibitory synapses in mammalian cultured neurons (Lin et al., 2008), so its induction within the ventral forebrain of PTZ-treated embryos could reflect the activation of inhibitory neural circuits within this region to mitigate the excitatory effects of PTZ exposure. The neuroprotective properties of Bdnf include roles in protecting neurons against axonal injury and apoptosis, as well as in facilitating experience-dependent synaptic remodelling and promoting neurogenesis (Lipsky and Marini, 2007). We also observed that PTZ exposure led to increased expression of pyya, a close relative of NPY, which has been implicated both as an endogenous inhibitor of epileptic seizures (Baraban et al., 1997) and as a neuroprotective component of the endocrine response to stress and mood disorders, mediating anxiolytic effects on neuronal targets within the mammalian CNS (Heilig, 2004; Thorsell, 2010). The complexity of the fos expression pattern induced by PTZ treatment suggests that transcription of fos is activated in many different types of neurons within the CNS. PTZ-induced expression of fos is most prominent in the ventral telencephalon and ventral diencephalon, as is expression of npas4 and bdnf, suggesting that neurons within these regions are particularly sensitive to loss of GABA-mediated inhibitory neurotransmission. It is interesting to note that, in mammals, the ventral telencephalon and ventral diencephalon give rise to components of the limbic system and hypothalamic-pituitary-adrenal (HPA) axis, respectively, which mediate neural and endocrine responses to behavioural and pharmacological stressors, including factors that trigger the onset of seizures. Future studies will explore the possibility that GABAergic neurons lying within embryonic precursors of the limbic system and HPA axis regulate susceptibility to seizures. Phenotypic analysis of zebrafish mind bomb mutant embryos, which lack Notch signalling pathway activity, previously demonstrated that expression levels of fos, bdnf and pyya transcripts were increased in the CNS of 3-day-old mutant embryos, which also exhibit spontaneous seizures (Hortopan et al., 2010a). In light of the phenotypic similarities between PTZ-treated wild-type embryos and mind bomb mutants, it might be of interest to examine the possibility that Notch pathway activity regulates GABAergic signalling. The ease with which PTZ-induced expression of fos was detected in a specific pattern within the zebrafish embryonic brain, together with its robust inhibition by exposure of embryos to VPA, suggested that suppression of PTZ-induced fos transcription in 2-day-old embryos could provide a means for identifying putative anticonvulsants in chemical libraries. Our results demonstrate the validity of this approach, in which a hit rate of 2.3% was observed. The two largest classes of compounds that exhibited anticonvulsant activity were steroids and anti-infectives (Table 1). Among the steroids with anticonvulsant activity, progesterone and its derivatives allopregnanolone and megestrol were identified, as were the weak androgens epiandrosterone and formestane. In addition, the synthetic estrogens ethinyl estradiol and hexestrol exhibited robust anticonvulsant activities. Epiandrosterone is an isomer of the known anticonvulsant androsterone (Kaminski et al., 2005), whereas megestrol was previously shown to be neuroprotective against oxidative stress (Sarang et al., 2002). Endogenous neurosteroids such as allopregnanolone and progesterone have previously been demonstrated to exhibit potent anticonvulsant activities in rodent models, and evidence suggests that circulating neurosteroids regulate seizure frequency and severity in some individuals with epilepsy (Reddy, 2010; Pack et al., 2011). Allopregnanolone functions as a direct allosteric agonist of the GABA_A receptor, suppressing seizures by potentiating the receptor’s signalling function and thereby over-riding the inhibitory effects of PTZ (Hosie et al., 2006). Progesterone could exert its anticonvulsant effect in the zebrafish embryo as a result of its conversion to allopregnanolone, as has been observed in mice (Reddy et al., 2004). Our discovery of an anticonvulsant effect for ethinyl estradiol is intriguing because this drug is known to reduce aggression and reproductive success in zebrafish (Colman et al., 2009) and it is neuroprotective against kainic-acid-induced excitotoxicity (Picazo et al., 2010), but it has not previously been shown to stimulate inhibitory neurotransmission as a GABA_A receptor agonist or to act as an anticonvulsant. Nevertheless, ethinyl estradiol was a more potent inhibitor of PTZ-induced locomotor activity than was allopregnanolone, raising the possibility that it might have an inhibitory effect on movement that could be independent of, and in addition to, an interaction with the GABA_A receptor. Three calcium channel blockers, nimodipine, nitrendipine and suloctidil, were identified as potent inhibitors both of PTZ-induced fos expression and locomotor activity, perhaps reflecting their abilities to directly inhibit neural activity and muscular contraction by blocking calcium influx. Exposure to nimodipine (24.7 μM) caused cessation of heartbeat in a proportion (∼33–36%) of treated embryos (Table 2). This effect might be due to inhibition of calcium channel function in cardiomyocytes, rather than a non-specific effect on cell viability, because no increase in apoptosis was observed after exposure to this compound. The identification of the antipsychotics methiothepin and pimozide, both of which inhibit dopamine and serotonin receptors (Seeman, 2002; Mahé et al., 2004), as suppressors of PTZ-induced fos expression and locomotor activity, is particularly interesting and strongly suggests that, as observed in other organisms (Bozzi et al., 2011), these monoamine neurotransmitters are involved in regulating seizures in zebrafish.

The second largest group of functionally related compounds that we identified in our screen was a class of anti-infectives that includes sulconazole, bifonazole and oxiconazole. Each of these compounds contains an imidazole ring and is related to ketoconazole, which is used to suppress excess glucocorticoid production in people with Cushing’s syndrome, in which activity of the HPA axis is abnormally high (Feldman, 1986). The mode of action of ketoconazole involves inhibition of stress hormone synthesis, by an incompletely understood mechanism, raising the possibility that its close relatives might also act similarly, modulating neuroendocrine components of the HPA axis that regulate stress responses and thereby suppressing seizures. Interestingly, the imidazole-based anti-epileptic drugs denzimol and nafimidone inhibit sodium channels (Mishra and Ganguly, 2012), suggesting another mechanism by which sulconazole, bifonazole and oxiconazole could elicit their anticonvulsant effects. Two other anti-infective hits identified in our screen are hexylresorcinol and acrisorcin (which is a combination of 9-aminoacridine and hexylresorcinol). Hexylresorcinol has anaesthetic properties and was previously shown to bind to sodium channels, inhibiting their function (Buchholtz et al., 2009), which could account for its anticonvulsant effect in our experiments. Several other compounds identified in our screen exhibit anticonvulsant activity in rodents. Previous studies of the COX inhibitors naproxen and mefenamic acid revealed that they both possessed anticonvulsant or related activities in PTZ-treated mice (Dhir et al., 2005; Wallenstein, 1991). Moreover, mefenamic acid acts as an agonist of the GABA_A receptor (Coyne et al., 2007). Interestingly, molsidomine enhanced the anticonvulsant properties of VPA in PTZ-treated mice (Tutka et al., 2002), and peucedanin is closely related to oxypeucedanin, which exhibited anticonvulsant activity in electroshock-treated mice (Luszczki et al., 2010). Among the compounds identified in our screen, the flavonoids rotenonic acid and deguelin, as well as the organochlorines endrin, propanil and lindane, were identified as anticonvulsants. Given that each of these compounds has a history of use as a neurotoxic pesticide, it is perhaps unsurprising that they were identified in our screen for neuroactive anticonvulsants using a gene expression assay for synaptic activity. Interestingly, lindane is a direct inhibitor of GABA_A receptors and exhibits convulsant activity in mice (Tochman et al., 2000; Vale et al., 2003), so its ability to suppress fos expression might reflect neurotoxicity, rather than an intrinsic anticonvulsant activity.

Overall, our results describe the establishment and validation of an in vivo model of epileptic seizures in the 2-day-old zebrafish embryo and its utilisation in a re-profiling screen of a bioactive small molecule library to identify compounds with anticonvulsant properties. Some of the compounds identified in this screen, such as allopregnanolone, nitrendipine, nimodipine, pimozide and methiothepin, have known biochemical specificities that readily account for their anticonvulsant behaviour in our assays. By contrast, the mechanisms of action of sulconazole, suloctidil and hexylresorcinol are poorly understood, and those of dioxybenzone and the sesquiterpene nerolidol are completely unknown. These compounds thus represent novel starting points for the development of new anti-epileptic drugs. Future studies will aim to elucidate the mechanisms of action of these molecules and to employ the in vivo assays we have developed to identify further compounds with anticonvulsant properties.

Stocks of PTZ (200 mM, in water) or picrotoxin (100 mM in DMSO) were diluted to the required concentration in fresh E3 medium. Embryos were exposed to PTZ or picrotoxin then analysed as required. For in situ hybridisation analysis, embryos were transferred immediately to fixative containing 4% paraformaldehyde and incubated at 4°C overnight. As a positive control for anticonvulsant activity, a 250 mM stock of VPA was made up in E3 medium and administered to embryos at a final concentration of 2.5 mM before the addition of the convulsant. All procedures involving experimental animals were performed in compliance with local and national animal welfare laws, guidelines and policies.

The Spectrum Collection (Microsource Discovery Systems) of 2000 compounds was stored with each compound at a concentration of 5 mM in DMSO in 25 v-bottomed 96-well microtitre plates (Matrix) at −80°C. Assay plates contained compounds diluted to 10 μM in E3 media for drug screening. For the anticonvulsant assay, embryos were raised to 50 hpf and treated with Pronase (Sigma) to remove the chorions. Embryos were aliquoted at four embryos per well into Multiscreen mesh-bottomed plates (100 mm; Millipore) and transferred to Multiscreen 96-well culture receiver trays (Millipore) containing the Spectrum library compounds at 10 μM in columns 2–11, with control wells containing either 1 mM VPA, E3 or DMSO only in columns 1 and 12. Assay plates were incubated in the dark at 28°C for 90 minutes followed by addition of PTZ to a final concentration of 20 mM to all the compound wells and half of the control wells. Assay plates were incubated for 1 hour at 28°C and the embryos were then transferred to fixative containing 4% PFA and stored at 4°C overnight. Embryos were bleached according to the standard protocol (Thisse and Thisse, 2008) and stored at −20°C in methanol until required for in situ hybridisation. In order to facilitate screening and eliminate the need to transfer embryos between plates at any stage of the process, samples were maintained in the same 96-well mesh-bottomed Multiscreen plates (Millipore) during drug treatment, in situ hybridisation and hit detection. Hits that were identified in the primary screen were selected from the Spectrum library and rescreened using the method above. Selected compound stocks were then purchased separately from Sigma and tested in dose-response assays to confirm that the Spectrum Collection stocks had been assigned the correct identities.

Digoxigenin-labelled RNA probes were prepared as recommended by the manufacturer (Roche). Details of the gabra1, gabrg2, fos, npas4, bdnf and pyya probes utilised are available on request. Whole-mount in situ hybridisation was performed using standard procedures (Thisse and Thisse, 2008) with modifications. All steps were performed using Multiscreen mesh-bottomed plates. In order to increase throughput, the Biolane HTI 16V In situ robot (Intavis) was used. Stained embryos were imaged in Multiscreen plates using a Nikon AZ100 microscope fitted with an automated stage (Prior). Extended depth of focus (EDF) images from z-sections through the embryos were compressed using the NIS-Elements software (Nikon) to show the areas of staining and the results were scored. Wells that contained embryos with no fos expression in the brain were taken forward for further analysis.

The distance moved by larvae over a 10 minute period was recorded using the Zebrabox system (Viewpoint, France). AB larvae at 4 dpf, with swim bladders inflated, were transferred to a 48-well microtitre plate, one larva per well in 445 μl of E3 medium. Dilutions of compounds over the range of 74-0.3 μM were added to embryos in triplicate, and control wells containing VPA and E3 alone were also included. Plates were incubated at 28°C in the dark for 90 minutes. Larvae were then tested for the presence of a heartbeat and motor response before addition of PTZ to wells containing compounds and half of the wells containing E3 only. After 5 minutes exposure to PTZ, larval movements were recorded with the Zebrabox for 10 minutes using a light cycle of 2 minutes:100% light; 2 minutes:0% light. Each experiment was repeated in triplicate to give nine data points for every drug concentration and the total distance moved by each larva was calculated.

For each group of embryos exposed to compounds, the mean distance moved and the standard error of mean (s.e.m.) were calculated. The effects on locomotor activity of exposing embryos to a range of PTZ concentrations were evaluated using a one-way ANOVA with Dunnett’s post-test. The effects of administering hit compounds were also evaluated using a one-way ANOVA statistical test to determine whether each compound suppressed PTZ-induced locomotor activity at 3, 8 and 24 μM concentrations.

A muscle-specific GCaMP-3 reporter plasmid was constructed by inserting the GCaMP-3 open reading frame (Tian et al., 2009) downstream of zebrafish mylz2 promoter sequences (von Hofsten et al., 2008). The circular plasmid DNA was injected into zebrafish embryos at the one-cell stage at a concentration of 0.1 ng/nl to generate mosaic embryos, which were incubated at 28.5°C until they reached 50 hpf. Embryos with detectable GFP fluorescence in muscle cells were transferred to E3 medium containing blebbistatin (50 μM) for 30–60 minutes at 28.5°C. Embryos were transferred singly to a 35 mm Wilco glass-bottomed Petri dish (22 mm glass diameter) and mounted for lateral viewing in a drop of 1% low-melting-point agarose containing 50 μM blebbistatin and 20 mM PTZ. Convulsant-induced increases in GCaMP-3 fluorescence were apparent within 3–5 minutes after mounting, and embryos were imaged with a Perkin-Elmer Spinning Disk microscope and Volocity (Improvision) software, using the GFP channel and a frame-rate of 20–40 Hz.

We are grateful to Professor Marysia Placzek for support and advice, and to Dr Claire Allen, Lisa Gleadall and Susan Surfleet for zebrafish husbandry. We are also grateful to Dr Markus Reuber for helpful comments on the manuscript.