Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

Avermectin and ivermectin are broad-spectrum antiparasitic agents that are in widespread use in agricultural and domestic animals, respectively (Campbell, 1989). Avermectin (so-called due to its activity against worms and ectoparasitic arthropods) was originally isolated from soil samples and shown to have a 16-membered macrocyclic lactone with a disaccharide substituent at the carbon-13 position (Albers-Schonberg et al., 1981). Avermectins designated B (such as ivermectin) have greater antiparasitic activity than those designated A. Avermectins exert their antiparasitic activity via activation of a glutamate-gated chloride channel present in the invertebrate nervous system (Cully et al., 1994) and have additional effects on invertebrate GABA receptors (Duce and Scott, 1985). In vertebrates where, to date, no glutamate-gated chloride channels have been reported, avermectins also have effects.

A number of studies have shown that ivermectin has anticonvulsant effects in a range of animal seizure models. Crichlow et al. (1986) reported that in photosensitive, genetically epileptic chickens, ivermectin provided effective protection against seizures up to 24 h after administration. In a later study, Ammendola et al. (1988) showed that ivermectin (30 and 50 mg/kg i.p.) significantly protected DBA/2 mice against sound-induced seizures. They also reported that in rats, cephazolin-induced seizures were attenuated by 10 and 20 mg/kg (i.p.) ivermectin and the protective effect of diazepam against pentylenetetrazole (PTZ)-induced tonic seizures was increased. However, ivermectin (5-30 mg/kg i.p.) was without effect against electroshock-induced seizures. Finally, Mayer and Horton (1991) induced seizures in mice with monomethylhydrazine and then administered ivermectin alone (5, 10, or 15 mg/kg) or in combination with diazepam (5 mg/kg) at doses of 5 or 10 mg/kg. Although ivermectin alone had no effect on the latency to convulse, it did increase significantly the time to death, and a combination of 10 mg/kg ivermectin and diazepam (5 mg/kg) prevented convulsions and death in all of the mice treated. Taken together, these data suggest that ivermectin is anticonvulsant, in some, but not all animal seizure models and, although the anticonvulsant activity of ivermectin is apparently relatively weak, it has a long duration of action. The current studies were designed to investigate whether the anticonvulsant properties of ivermectin are a common property of avermectins, and whether these effects are mediated through GABA_A receptors. A secondary aim of the current research was whether these acute anticonvulsant effects of avermectins were distinguishable from the mechanism responsible for their toxic effects.

The anticonvulsant effects of a range of avermectin analogs were measured in a mouse seizure model, in which the seizures were induced by PTZ. In this model, data are usually expressed as the dose that protects 50% of the animals from convulsions induced by 120 mg/kg PTZ (ED₅₀). The ED₅₀ of diazepam in this model is approximately 0.3 mg/kg. Benzodiazepines, and other compounds that interact with the GABA_A receptor complex, also have muscle relaxant and sedative effects. Consequently, all the animals were placed on a Rotarod immediately before injection of PTZ in an attempt to measure drug-induced motor impairment induced by sedation or muscle relaxant effects. It should be noted that in general avermectins do not exhibit acute toxicity in these animal models and that the animals used in the PTZ studies were humanely sacrificed in compliance with United Kingdom regulations governing the control of experiments on animals before lethal toxic effects were experienced. The LD₅₀ values quoted within were obtained after observation of mice during a 24-h period and were obtained over a period of time in the 1980s, at Merck & Co., this procedure is no longer performed.

[³H]Ivermectin binds to a distinct binding site in rat brain membranes, which is proposed to be the GABA_A receptor (Schaeffer and Haines, 1989). Avermectins have been shown to bind to sites on the GABA_A receptor (Williams and Risley, 1982; Drexler and Sieghart, 1984; Huang and Casida, 1997). Given the role of the GABA_A receptor in mediating inhibition in the central nervous system and the anticonvulsant effects of drugs that potentiate GABA receptors (e.g., benzodiazepines and barbiturates), it seemed the most plausible site through which the avermectin analogs could be mediating their anticonvulsant effect. We therefore determined the affinity of a set of structurally diverse avermectin analogs at the [³H]ivermectin binding site in rat brain and at recombinant GABA_A receptors. We have also investigated their efficacy at GABA_A receptors electrophysiologically, by potentiating GABA_A receptors on cortical neurons, on transfected cells, and in Xenopus oocytes expressing human GABA_A receptor subtypes. Avermectins have also been shown to inhibit [³H]strychnine binding to glycine receptors in spinal cord (Graham et al., 1982). Because glycine receptors are structurally related to GABA_A receptors we also investigated the effects of avermectin analogs using [³H]strychnine binding to spinal cord, and electrophysiologically on cultured primary cortical neurons to determine whether the strychnine-sensitive glycine receptors could be the source of avermectin-induced toxicity in mice.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

Animals

Naive, male Swiss-Webster mice (Bantin & Kingman, Hull, England), weighing 22 to 40 g, were used 7 to 15 days after arrival. The mice were housed in groups of five on a controlled 12-h light/dark cycle, and were allowed ad libitum access to standard cube diet and tap water.

Drugs

Diazepam and pentylenetetrazole were supplied by Sigma Chemical Co. (Poole, UK). The avermectin analogs were synthesized by the Merck Research Laboratories medicinal chemistry group (Rahway, NJ). For in vitro experiments avermectins were dissolved as stock solutions of 1 or 10 mM in dimethyl sulfoxide and diluted to appropriate concentrations in saline buffer. For in vivo experiments, diazepam and all the avermectin analogs were suspended in 0.5% methylcellulose in distilled water and were administered in a volume of 10 ml/kg. All vehicle and all compounds were injected i.p.

Procedures

Mouse Model of Seizures. At the beginning of the experiment, all of the mice were trained to stay on a Rotarod, revolving at 16 rpm (model 7600; Ugo Basile, Varese, Italy), for 120 s. After this initial training period the mice were randomly divided into treatment groups (n = 8) and injected i.p. with either vehicle or one dose of an avermectin analog (25 in all). Ninety minutes later, mice were again put onto the Rotarod and the latency to drop from the Rotarod was recorded. If mice failed to drop before 120 s had elapsed, mice were removed and a latency of 120 s was recorded. The mice were then injected with PTZ (120 mg/kg) and observed for the next 30 min for tonic seizure (full extension). Once a seizure was observed the animals were euthanized. The procedure for the diazepam experiment was identical with that described above with the exception that diazepam was given 30 min before the Rotarod test. Latency data for Rotarod tests were analyzed by one-way ANOVA and post hoc Dunnett's t tests. The smallest avermectin or diazepam dose inducing a significant deficit in Rotarod performance compared with the vehicle-treated mice was designated the minimum effect dose for the compound. Because pilot experiments indicated that the potency of the avermectin analogs varied markedly, a maximal and no-effect dose was determined for each analog. Using these minimum and maximum values, full dose-response curves were determined for each analog and from these curves ED₅₀ values (the dose that protects 50% of the mice from PTZ-induced tonic seizures) were calculated by probit analysis.

Binding of [³H]Ivermectin to Rat Brain Membranes. Rat brain membranes were prepared by homogenizing previously frozen rat brain in 20 volumes of 50 mM Tris acetate, 5 mM EDTA, pH 7.4, at 4°C. The homogenate was centrifuged at 1200g for 20 min at 4°C. The pellet was resuspended in 20 volumes of 5 mM Tris acetate, pH 7.5, at 4°C and centrifuged at 1250g twice more before resuspension in HEPES, pH 7.4, at 4°C at a protein concentration of 1 mg/ml. Binding assays were carried out in 0.5-ml volumes containing 100 µg of membrane protein for 1 h at room temperature and were initiated by addition of radioligand to minimize loss of label by nonspecific adsorption. Nonspecific binding was defined with 10 µM ivermectin and, at 5 nM [³H]ivermectin, nonspecific binding was routinely 25 to 30% of total binding in rat brain and 8 to 15% in recombinant cell lines. Incubations were terminated by filtration through GF/B filters presoaked in 0.1% polyethyeneimine, followed by three washes (5 ml) with 50 mM HEPES, pH 7.5, containing 0.25% Triton X-100 to minimize nonspecific binding.

Binding of [³H]Strychnine to Membranes from Rat Spinal Cord. Membranes were prepared from rat spinal cord by homogenizing five spinal cords in 20 volumes of 0.32 M sucrose, 5 mM Tris acetate, pH 7.4, at 4°C. The homogenate was centrifuged at 300g for 10 min and the supernatant removed and recentrifuged at 1250g for 20 min at 4°C. The pellet was resuspended in 20 volumes of 5 mM Tris acetate, pH 7.5, at 4°C and centrifuged at 1250g a further two times before freezing until required. After thawing, the membranes were washed once by resuspension and centrifugation in 50 mM phosphate buffer, pH 7.4, at 4°C containing 200 mM NaCl and resuspended at 2.5 mg of protein/ml. Radioligand binding assays were incubated in 0.5-ml volumes containing 200 µg of spinal cord membrane protein, 2 nM [³H]strychnine, and varying concentrations of the avermectin analogs for 30 min at 4°C. Nonspecific binding was defined with 1 mM glycine. Incubations were terminated by filtration as described above for [³H]ivermectin binding.

Binding of [³H]Ivermectin to Cell Lines Expressing Recombinant GABA_A Receptors. Stable cell lines expressing 132, 232, 332, 532, 112, and 122 were grown as previously described for the 632 subtype (Hadingham et al., 1996). Cells were harvested by scraping and were washed twice by centrifugation at 1000g and resuspended in 50 mM phosphate, 120 mM NaCl, pH 7.5. Cells were either frozen as pellets or used immediately by resuspension in 10 ml of 50 mM HEPES buffer, pH 7.5. Membranes (25-75 µg of protein) were incubated with radioligand in a total volume of 0.5 ml and binding assays were carried out as for rat brain (see above).

Recombinant GABA_A Receptors Expressed in Xenopus Oocytes. Xenopus oocytes were removed from anesthetized frogs and manually defolliculated with fine forceps. After mild collagenase treatment to remove follicle cells (type IA, 0.5mg/ml for 8 min) the oocyte nuclei were then directly injected with 10 to 20 nl of injection buffer [88 mM NaCl, 1 mM KCl, 15 mM HEPES at pH 7.0 (nitrocellulose filtered)] containing different combinations of human GABA_A subunit cDNAs (20 ng/ml) engineered into the expression vector pCDM8 or pcDNAAmp. After incubation for 24 h, oocytes were placed in a 50-ml bath and perfused with modified Barth's medium consisting of 88 mM NaCl, 1 mM KCl, 10 mM HEPES, 0.82 mM MgSO₄, 0.33 mM Ca(NO₃)₂, 0.91 mM CaCl₂, 2.4 mM NaHCO₃, pH 7.5. Cells were impaled with two 1 to 3 M electrodes containing 2 M KCl and voltage clamped between 40 and 70 mV. The cell was continuously perfused with saline at 4 to 6 ml/min and drugs were applied in the perfusate. Avermectins were preapplied for 30 s before the addition of GABA. GABA was applied until the peak of the response was observed, usually 30 s or less. At least 3 min of wash time was allowed between each GABA application to prevent desensitization.

Whole-Cell Patch-Clamp Recordings. Experiments were performed on cells stably expressing human 132S receptors and on cultured rat cortical neurons. Cultures of rat cortical neurons were prepared from cerebral hemispheres of rat fetuses (16-18 days of gestation) as previously described (Priestley et al., 1990), and used after 2 to 3 weeks in culture. Glass coverslips containing the cells in a monolayer culture were transferred to a Perspex chamber on the stage of Nikon Diaphot inverted microscope. Cells were continuously perfused with a solution containing 149 mM NaCl, 3.25 mM KCl, 2 mM CaCl₂, 10 mM HEPES, 22 mM D(+)-sucrose, 11 mM D-glucose, and 0.3 µM tetrodotoxin at pH 7.2, and observed using phase-contrast optics. Patch pipettes were pulled with an approximate tip diameter of 2 µm and a resistance of 4 M with borosilicate glass and filled with 120 mM CsF, 10 mM CsCl, 10 mM HEPES, 10 mM EGTA, 4 mM NaCl, 0.5mM CaCl₂, pH adjusted to 7.25 with CsOH. Cells were patch-clamped in whole-cell mode using a List LM-EPC 7 patch-clamp amplifier. Drug solutions were applied by a double-barreled pipette assembly, controlled by a stepping motor attached to a Leitz manipulator, enabling rapid equilibration around the cell. Stable responses to GABA or glycine were obtained (2-3-s pulse) before adding avermectins.

	Results

Top Abstract Introduction Materials and Methods Results Discussion References

Effects of Avermectins on PTZ-Induced Tonic Seizures in Mice. Administration of ivermectin provided dose-related protection from PTZ-induced tonic seizures in mice (Fig. 1A). The ED₅₀ for this effect was 28.2 mg/kg and full protection was seen at a dose of 80.0 mg/kg. There were no significant deficits in Rotarod performance at 300.0 mg/kg (Fig. 1B). Of the remaining 24 avermectin analogs, L-676,893 (avermectin A2a) was the most potent with an ED₅₀ of 0.48 mg/kg (Fig. 1A). L-676,893 did not impair Rotarod performance up to the maximum tolerated dose of 2.5 mg/kg (Fig. 1B). However, a dose of 3.0 mg/kg L-676,893 induced tremor, ptosis, and labored breathing. The remaining compounds tested had a wide range of potencies against PTZ-induced tonic seizures. The least potent, L-685,869 (4''-epi-acetylamino-4''-deoxy-avermectin B1a 5-ketoxime), at a dose of 160 mg/kg provided only 50% protection from seizures. The results for each of the derivatives examined are summarized in Table 1 and show that the most potent avermectins were the A2a and B2a analogs, whereas the least potent were the B1 analogs.

View larger version (14K): [in this window] [in a new window]   Fig. 1.   Shown are the anticonvulsant effects of ivermectin (), diazepam (), and L-676,893 () on PTZ-induced seizures in mice (A) and the mean ± S.E. latency to fall from a Rotarod turning at 15 rpm (B). Ivermectin and L-676,893 were injected 90 min before, and diazepam 30 min before, the animals (n = 8) were placed on the Rotarod for 120 s. The mice were injected with PTZ (120 mg/kg i.p.) and observed for the next 30 min for tonic seizure (full extension). Immediately after a seizure had been observed the animals were euthanized.

Binding of Avermectins to Rat Cortical GABA_A Receptors. A set of structurally diverse avermectin analogs was tested for their affinity at [³H]ivermectin binding sites in rat brain and these are presented in Table 2. The affinity of [³H]ivermectin for the rat brain binding site (8.2 nM) agreed well with a previous study where [³H]ivermectin bound to one population of sites with a K_d of 22 nM (Schaeffer and Haines, 1989). Several of the analogs tested (L-676,893, L-676,863, and L-656,748) had equivalent affinity to ivermectin for the rat brain [³H]ivermectin binding site. However, despite their similar affinity, L-676,893 was 20- to 40-fold more potent in the anticonvulsant assay than ivermectin, L-676,863, or L-656,748. It was reasoned that the avermectin analogs may have different affinities for subtypes of the GABA_A receptor or alternatively may have differential efficacy at subtypes of the GABA_A receptors. These two possibilities were investigated further.

View this table: [in this window] [in a new window]   TABLE 2 The affinity of avermectins for the [3H]ivermectin binding site on GABAA receptors in rat brain membranes and potentiation of submaximal GABA responses on oocytes expressing human 122 receptors Data shown are mean ± S.E.M. of three experiments. Ki values were determined by competition using [3H]ivermectin at 5 to 8 nM. Potentiation of the GABA EC20 on 122s GABAA receptors expressed in oocytes are the mean ± standard error of at least three oocytes. Ki values were calculated from IC50 determinations using the Cheng-Prussof equation. [3H]Ivermectin bound to rat brain with a Kd of 8.2 ± 2.0 nM and Bmax of 2.5 ± 0.4 pmol/mg of protein.

Affinity of Avermectin Analogs for GABA_A Receptors Containing Different - or -Subunits. GABA_A receptors in rodent brain are present as a family of proteins that differ in their subunit composition. We first investigated binding to one homogeneous population of recombinant GABA_A receptors, the 132 subtype. The number of binding sites for [³H]ivermectin and [³H]Ro 15-1788 on a stable cell line, expressing a single homogeneous population of GABA_A receptors, was compared. In cells expressing the 132 subtypes there were approximately twice as many ivermectin binding sites compared with benzodiazepine binding sites (B_max for [³H]ivermectin = 3833 ± 1130 fmol/mg of protein, K_d = 3.8 ± 1.5 nM, n = 3; B_max for [³H]Ro 15-1788 = 1660 ± 255 fmol/mg of protein, K_d = 1.5 ± 0.2 nM, n = 3; ratio B_max for [³H]ivermectin/B_max for [³H]Ro 15-1788 = 2.3). Our current understanding of the structure of the GABA_A receptor is that there is one BZ site per receptor monomer (Sigel and Buhr, 1997); it therefore follows that there are two avermectin sites per GABA_A receptor monomer. It was noted that the affinity of [³H]ivermectin for the recombinant 132 subtypes was higher (3.8 nM) than that observed in rat brain (8.2 nM, this study; 22 nM, Schaeffer and Haines, 1989) and that the nonspecific binding to recombinant receptors stably expressed in a fibroblast cell line was reduced compared with rat brain. The pharmacology of the [³H]ivermectin binding site on rat brain and cell lines expressing 132 GABA_A receptors was compared to determine whether the same receptor was being labeled. The same rank order of potency was observed for receptors containing the 132 subtypes and rat brain [³H]ivermectin binding sites (L-640,471 > L-676,893 = L-656,748 = L-676,863 = L-751,531 > L-669,437 > L-697,960), which confirms that [³H]ivermectin binding in rat brain is primarily to GABA_A receptors (Table 3).

View this table: [in this window] [in a new window]   TABLE 3 The affinity of avermectin analogs for GABAA receptors that vary in their -subunit structure Data shown are the mean of three determinations performed using cells stably transfected with 1, 2, and either 1, 2, or 3. Data are expressed as IC50 values.

Electrophysiological Analysis of the Effects of Avermectin Analogs on Recombinant GABA_A and Native Rat Brain GABA_A and Glycine Receptors. It is possible that various avermectin derivatives not only exhibited differences in affinity but also in their ability to potentiate GABA_A receptors (i.e., their efficacy). To investigate this further, several avermectin analogs were studied on human 112, 122, and 132 expressed in Xenopus oocytes to determine any -subunit selectivity for avermectin potentiation. In previous experiments avermectin was shown to act on receptors composed of only homomeric 1 (Arena et al., 1993), suggesting that a minimum requirement for the binding site was that of a -subunit, making this a key subunit for avermectin activity. Drugs were applied at a concentration of 1 µM together with an EC₂₀ GABA response and modulation of the amplitude of this response by avermectins was measured (Fig. 2). Ivermectin showed some subunit selectivity relative to the -subunit, potentiating 1-containing receptors by ~400%, compared with 200% on 2- or 3-containing receptors. L-676,863 was virtually inactive on 3-containing receptors, but potentiated 1- and 2-containing receptors by approximately 200%. L-676,893 potentiated to the greatest extent (~400%), but lacked subunit selectivity. Several other analogs were examined for potentiation of the GABA response in oocytes expressing human 122 receptors and data are shown in Table 2.

View larger version (27K): [in this window] [in a new window]   Fig. 2.   Potentiation of human GABAA receptors in Xenopus oocytes expressing 112, 122, and 132 recombinant subtypes by different avermectin analogs. Data are shown as percentage of potentiation of a GABA EC20 response determined on each individual oocyte, and represents the mean ± standard error from at least four oocytes.

View larger version (13K): [in this window] [in a new window]   Fig. 3.   Effects of 1 µM ivermectin (L-640,471) on GABA-induced currents in rat cortical neurons (A), and GABA-induced currents on Ltk cells stably expressing human 132 receptors (B). Application of 1 µM GABA and 0.3 µM GABA (approximate EC20 in each case) is indicated, and application of 1 µM L-640,471 is shown by the bar above the trace.

Effects of Avermectins on Glycine Receptors. The ability of avermectin A2a to bind to other sites in rat brain was investigated. This compound was inactive at adenosine receptors (A1 and A2), CCK receptors (CCK_A and CCK_B), adrenergic receptors (1, 2, 1, 2), glutamate receptors (-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid, MK-801, and kainate binding sites), dopamine (D1-5), muscarinic receptors, opiate receptors (µ-, -, and -subtypes) and at the 5-hydroxytryptamine receptors and potassium channels. It did, however, have activity at the strychnine-sensitive glycine receptor of rat brain (Graham et al., 1982), and it was reasoned that this may contribute to the toxic effects observed in behavioral experiments. Initial investigations on glycine receptors from spinal cord mRNA expressed in oocytes confirmed an inhibitory effect for ivermectin (L-640,471; J. Arena, unpublished observations). The affinity of avermectins for the strychnine binding site of rat spinal cord and their functional effects on native, neuronal glycine receptors were therefore investigated.

Effects of Avermectin Analogs on Glycine Receptors Present on Cultured Rat Cortical Neurons. We have identified a neuronal glycine receptor on primary cultured cortical neurons, which is strychnine sensitive and bicuculline insensitive. Whole-cell patch-clamp techniques were used to examine the effects of eight different avermectins on this glycine receptor, for comparison with their effects on GABA receptors and toxic effects. Glycine concentration-response curves were generated in the presence of 100 µM bicuculline to inhibit glycine activation of GABA_A receptors, showing the cortical glycine receptor to have an EC₅₀ of 135 ± 6.0 µM. This receptor was sensitive to strychnine with an IC₅₀ of 64 ± 8.5 nM. A control concentration of 300 µM glycine was used to study the effects of different avermectin analogs. Avermectins inhibited glycine responses in a use-dependent manner to varying degrees (Fig. 4), and appeared to dramatically slow the off-rate of glycine from the receptor. The reasons for this are unclear at present and clearly require further investigation. To compare the extent of block with each avermectin analog, the degree of inhibition after the third glycine application in the presence of avermectin was compared. All the compounds inhibited glycine responses as shown in Table 5. The B1a analogs had the largest effect including ivermectin (L-640,471), whereas L-676,893 had a smaller effect, suggesting no correlation with its effects on GABA_A receptors.

View larger version (10K): [in this window] [in a new window]   Fig. 4.   .. Use-dependent blockade of glycine-induced currents on rat cortical neurons by avermectin analogs. Glycine (300 µM)-induced currents in the presence of 100 µM bicuculline as indicated by the short bars above each application. Ivermectin (1 µM, L-640,471) (A) and 1 µM L-676,893 (B) were applied as shown by the long bar, demonstrating a slowing of glycine off-rate and use-dependent block in the case of L-640,471.

View larger version (19K): [in this window] [in a new window]   Fig. 5.   Correlation of the PTZ ED50 and historical LD50 data for the avermectin analogs studied and the four different in vitro assays studied here, Oocyte efficacy (A), Ki versus [3H]ivermectin binding in rat brain (B), strychnine binding to rat spinal cord (C), and degree of inhibition of glycine currents in primary cortical neurons (D). The closest match for anticonvulsant activity is oocyte efficacy and for toxicity is that of [3H]ivermectin binding, suggesting the mechanism for both effects to be mediated by the GABAA receptor. *P < .05, significant correlation.

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

The present results are in agreement with previous findings that ivermectin has anticonvulsant effects in animal seizure models (Crichlow et al., 1986; Ammendola et al., 1988; Mayer and Horton, 1991). ED₅₀ values for 25 avermectin analogs ranged from 0.48 mg/kg (L-676,893) to >160 mg/kg (L-685,869). The three most potent avermectins had ED₅₀ values <1.0 mg/kg. The ED₅₀ values for the anticonvulsant effect correlated best with the efficacy of avermectins in oocytes expressing GABA_A receptors. However, with the exception of one compound the toxicity associated with avermectins also correlated with their efficacy at GABA_A receptors, suggesting that the therapeutic potential of avermectins as anticonvulsants in humans may be low.

The most active compound in the PTZ model, L-676,893, gave the highest degree of potentiation on all receptors studied, irrespective of the -subunit present. The inactive analogs L-693,752 and L-697,960 did not potentiate recombinant GABA_A receptors, strongly suggesting that the anticonvulsant mechanism is via GABA_A receptor potentiation and the potency of the avermectin analogs depends on both receptor affinity and efficacy, with high-efficacy compounds being much more potent anticonvulsants. This conclusion is supported by the correlation of anticonvulsant activity with the degree of efficacy measured in Xenopus oocytes. The compounds L-640,471 and L-676,893 were also examined using whole-cell patch-clamp recording from both cultured cortical neurons and Ltk cells stably expressing the GABA_A receptor combination 132s. The potentiation observed on cortical neurons was smaller than that elicited on Xenopus oocytes despite using an approximate GABA EC₂₀ in both assays. Additionally, not all neurons were potentiated by avermectins, suggesting possible heterogeneity. In both neurons and oocytes some small direct activation of the receptors was observed with 1 µM avermectin. When similar experiments were carried out on cells stably transfected with 132 GABA_A receptors, large direct currents in response to avermectin were consistently observed. The direct activation was not consistently observed in oocytes, and when present was smaller than that seen using the patch-clamp assay. Also, the avermectin-induced currents were not observed in untransfected cells and were sensitive to picrotoxin, suggesting an agonist-like effect on the GABA receptors present. This effect has been observed in a brain microsac preparation where avermectin induced a ³⁶Cl influx in the absence of GABA (Abalis et al., 1986) and has also been reported on rat dorsal root ganglion neurons (Robertson, 1989) and mouse hippocampal neurons (Schonrock and Bormann, 1993; Krusek and Zemkova, 1994). Prolonged exposure to ivermectin inhibited GABA currents in stably transfected cells (Fig. 3B), an effect that was also reported to occur in mouse hippocampal neurons, and in cerebellar granule neurons (Huang and Casida, 1997). These effects suggested a dual mechanism of action: allosteric potentiation followed by inhibition after prolonged exposure and the lipophilic nature of the compounds made them very difficult to wash out. It may be that this delayed inhibition of the receptor at higher concentrations could account for the relatively low LD₅₀ values noted in vivo, which were also delayed compared with the anticonvulsant effects.

Radioligand binding studies also confirm previous reports that ivermectin binds to GABA_A receptors. Furthermore, we have shown that avermectins exhibit little subtype selectivity between receptors of different - and -subunit composition, and that there are two avermectin binding sites per GABA_A receptor. The relative lack of subtype selectivity suggests that the avermectins bind to a conserved part of the receptor structure, although the exact site of action and molecular mechanism by which avermectins potentiate the GABA_A receptor are not yet known. We have not investigated all possible subtypes of the GABA_A receptor. The minor populations that contain 4, 1, 3 - or -subunits have not been considered, and it cannot be ruled out that avermectins have some binding selectivity at these subtypes. Although binding to rat brain has been of high affinity with no evidence from binding assays for more than one population of sites, it is not possible to rule out that avermectins might also bind to other small populations, which could be obscured by the high number of binding sites on GABA_A receptors in rat brain.

Avermectins have been shown to potentiate responses to GABA on cultured hippocampal neurons (Krusek and Zemkova, 1994) and also in Xenopus oocytes injected with chick brain mRNA (Sigel and Baur, 1987). The functional responses in recombinant GABA_A receptors in oocytes and rat native cortical neurons were potentiated by some, but not all avermectins. A range of levels was observed for avermectins, suggesting that not all avermectins behave as full agonists at their binding site. To date we have not observed any evidence for compounds that have inverse agonism through this site. Previous studies have suggested that the -subunit was not necessary for avermectin potentiation, and 1 subunits expressed alone could be activated by avermectin (Arena et al., 1993). Because there was little in the way of binding selectivity, we compared the degree of efficacy at recombinant receptors containing different -subunits. A degree of selective efficacy was observed in Xenopus oocytes, particularly with L-676,863, where little efficacy was seen at 132 receptors. Ivermectin also appeared to have higher efficacy at 112 receptors than at 122 or 132. The observed subunit selectivity was not however, ubiquitous to all avermectins, and it is currently not clear how these levels of subtype selectivity may relate to in vivo activity. It is also worth noting that compounds may not all be binding in an identical manner to the [³H]ivermectin binding site because L-669,437 poorly displaced [³H]ivermectin (766 nM K_i), but gave robust potentiation at 1 µM in the oocyte assay. This compound also exhibited low toxicity and exclusion of this compound from Fig. 5 would produce a significant correlation between oocyte efficacy and LD₅₀ (r² = 0.75). Nevertheless, it should be noted that although the anticonvulsant and toxic effects of avermectins appear to be mediated via the GABA_A receptor, it remains possible that the mechanism of action for these effects may not be the same. For example, the anticonvulsant effects of avermectins correlated best with their efficacy at GABA_A receptors, whereas the toxicity correlated best with their affinity. Thus, an ivermectin antagonist binding to the ivermectin site on GABA_A receptors could potentially be toxic, but not anticonvulsant.

The reported activity of avermectin B1a at the strychnine binding site (Graham et al., 1982) suggests another possible mechanism of action, particularly for the toxic effects. Several compounds studied were found to inhibit [³H]strychnine binding to spinal cord with a range of affinities, and all the analogs tested at 1 µM inhibited glycine responses on cortical neurons to some degree. The inhibition appeared to be use-dependent because the first post-avermectin response was not reduced, but subsequent responses were inhibited (Fig. 4). The compounds also appeared to slow the recovery from glycine dramatically, suggesting a decrease in glycine off-rate. The most potent inhibitory compounds appeared to be the B1a analogs, which interestingly are the most anthelmintic of the analogs; indeed the most active compound on GABA receptors, L-676,893, gave little inhibition of glycine receptors, suggesting that their structural requirements for interaction at the two receptors are different. The inactive compound L-697,960 also gave little inhibition of glycine receptors. With the exception of L-697,960, all the analogs inhibited [³H]strychnine binding in spinal cord, again in agreement with their inhibitory activity at glycine receptors. Interestingly, the B1a analogs are also the most potent on invertebrate glutamate receptors (Arena et al., 1995), which are more closely related to vertebrate glycine receptors than vertebrate glutamate or GABA_A receptors (Vassilatis et al., 1997).

Although a reduction in convulsant episodes among humans under ivermectin treatment for onchocerciasis (river blindness) has been reported (Kipp et al., 1992) it is unlikely that this effect is mediated by its action at human GABA_A receptors. After anecdotal reports that ivermectin reduced grand mal seizures in an African population, Kipp et al. (1992) selected 91 known epileptics from the population receiving ivermectin treatment. Grand mal seizures were present in 69 patients and petite mal in 22. Although Kipp et al. (1992) did not report the dose given to the epileptics, ivermectin is normally given as a once-yearly oral dose of 150 µg/kg. At this dose ivermectin suppresses microfilariae in the skin and eyes and in most people it prevents the progression of onchocerciasis. Of the 91 patients, 34 reported some reduction in seizure frequency and 13 reported no seizures at all during the ivermectin treatment. These data and those described above might suggest that the antiepileptic effects observed in humans were due to a central action of ivermectin. However, a subsequent report by Kipp et al. (1994) showed that in two African villages in which the presence of microfilariae of Onchocerca volvulus differed (68 and 19%, respectively), the number of epileptics in each village also significantly differed (8 and 0.02%, respectively). These data suggest that there is a causal link between onchocerciasis and epilepsy because reducing microfilariae also reduces the incidence of seizures. In addition, the half-life of ivermectin is 1.8 days, whereas seizures were reduced for the 6- to 12-month period in which the parasites were under control. Thus, it is very unlikely that the reduction in seizures was due to a central nervous system-mediated effect of ivermectin, but rather it is likely that the reduction in seizures is linked to the reduction in parasites.

In conclusion, avermectins have anticonvulsant effects in mice treated with PTZ. It is likely that the anticonvulsant effects are mediated via the GABA_A receptor because the efficacy of avermectins at the GABA_A receptor correlated most closely with their anticonvulsant potency. However, it is also likely that the lethal action of avermectins in mice after doses that have anticonvulsant effects are also mediated via the GABA_A receptor. These data suggest that the therapeutic potential of avermectins as anticonvulsant agents is therefore limited.