Abstract
Neuroactive steroids are positive allosteric modulators of γ-aminobutyric acidA (GABAA) receptor complexes. Synthetic modification generally does not increase neuroactive steroid potency beyond that of the naturally occurring progesterone metabolite, 3α-hydroxy-5α-pregnan-20-one (3α,5α-P). Recently, it has been shown that introduction of appropriately para-substituted phenylethynyl groups at the 3β-position of 5β steroids increases receptor potency. The present report presents the synthesis and pharmacological profile of an analogous series of 5α steroids. The most striking feature of this series is the further enhancement of in vitro andin vivo potency obtained. In particular, 3β-(p-acetylphenylethynyl)-3α-hydroxy-5α-pregnan-20-one (Co 152791) was 11-, 16- and 49-fold more potent than 3α,5α-P in modulating the binding of [35S]TBPS, [3H]flunitrazepam and [3H]muscimol, respectively, in rat brain membranes (Co 152791 IC50 or EC50 = 2–7.5 nM). Similarly, Co 152791 was 3- to 20-fold more potent than 3α,5α-P as an inhibitor of [35S]TBPS binding in human recombinant receptor combinations containing α1, α2, α3 or α5 and β2γ2L subunits (Co 152791 IC501.4–5.7 nM). Co 152791 displayed low efficacy and 3α,5α-P had low potency at α4/6β3γ2L GABAA receptor complexes. Interestingly, Co 152791 demonstrated remarkable potency as a potentiator of GABA-evoked currents in Xenopus oocytes expressing α1β2γ2L receptors (EC50 0.87 nM), being 184-fold more potent than 3α,5α-P. High in vitropotency was also reflected in enhanced in vivo activity in that Co 152791 exhibited exceptional anticonvulsant potency, protecting mice from pentylenetetrazol-induced seizures at a ∼5-fold lower dose than 3α,5α-P after i.p. administration (Co 152791 ED50 0.6 mg/kg). Moreover, Co 152791 was orally active (ED50 1.1 mg/kg) and exhibited a therapeutic index of 7 relative to rotorod impairment. The remarkable potency of Co 152791 as a positive allosteric modulator of GABAA receptors may be explained by its interaction with an auxiliary binding pocket in the neuroactive steroid binding site. In addition, modification at the 3β-position probably hinders metabolism of the 3α-hydroxy group contributing to the exceptional anticonvulsant potency of this compound relative to other neuroactive steroids.
Following the discovery that the anesthetic steroid alphaxalone potentiated GABA responses (Harrison and Simmonds, 1984), it soon became clear that related steroids, including metabolites of progesterone and deoxycorticosterone, are also positive allosteric modulators of GABAA receptor complexes (Gee et al., 1987; Harrison et al., 1987; Majewska et al., 1986). By analogy to other known GABA potentiators such as barbiturates and benzodiazepines, this GABAergic mechanism suggested that these novel modulators, now termed neuroactive steroids, could be useful clinically for a number of central nervous system disorders. In addition to the historical (Phillips, 1975) and recent (Andersonet al., 1997) use as intravenous anesthetics, neuroactive steroids have potential uses as antiepileptic agents (Carter et al., 1997; Gasior et al., 1997), sedative-hypnotics (Edgar et al., 1997), anxiolytics (Brot et al., 1997; Carter et al., 1995; Wieland et al., 1995,1997) and for migraine (Limmroth et al., 1996).
Although specific binding by a radiolabeled steroid has not been convincingly demonstrated, compelling evidence for a unique site on the GABAA receptor complex for neuroactive steroids has been amassed (Gee et al., 1995). The strongest single argument in favor of a unique binding site is the exquisite SAR, in particular the stereoselectivity of the 3-hydroxy group, which must be in the α configuration for potent modulation of the receptor complex (Gee et al., 1987; Harrison et al., 1987;Hawkinson et al., 1994a; Hogenkamp et al., 1997;Upasani et al., 1997). Presumably, the 3α-stereochemistry is required for the correct alignment of the hydroxy group with a hydrogen bond accepting group located in the binding site. In addition, the 20-keto function is thought to contribute to high receptor potency by interacting with a hydrogen bond donating residue in the binding site, although steroids without this group may retain moderate potency (Bolger et al., 1997; Purdy et al., 1990;Hawkinson et al., 1994a).
Substitution of the steroid nucleus at the 3β-position was initially explored to increase bioavailability by blocking metabolic oxidation of the critical 3α-hydroxy group, preventing conversion to potentially hormonally active steroid metabolites and to slow metabolic conjugation at this position (Hogenkamp et al., 1997). This approach resulted in ganaxolone (Carter et al., 1997), which is currently in phase II clinical trials for epilepsy and migraine. Recently, it was shown that substitution of the 3β-position with an ethynyl spacer unit linked to a phenyl group in the 5β steroid series results in highly potent modulators of GABAAreceptors, particularly when the phenyl group is substituted in thepara-position with hydrogen bond accepting groups such as acetyl (Upasani et al., 1997). Based on these observations, it was proposed that an auxiliary binding pocket exists adjacent to the site occupied by the steroid A-ring and that this pocket contains a hydrogen bond donating group which interacts with thep-acetyl group of the 3β-phenylethynyl substituent (Upasani et al., 1997). The present report examines the role of extended 3β-substitution in the 5α steroid series and describes the pharmacology of Co 152791 (fig. 1), the most potent known neuroactive steroid modulator of GABAA receptors.
Methods
Synthesis
The preparation of compounds 1-5 and18 was described previously (Hogenkamp et al., 1997; Upasani et al., 1997). 3β-(Hexyn-1-yl)-3α-hydroxy-5α-pregnan-20-one (6) was prepared by adding hexyn-1-yl lithium, generated by the reaction of 1-hexyne with n-butyl lithium, to 5α-pregnane-3,20-dione 20-ketal. Similarly, addition of phenylmagnesium bromide and benzylmagnesium bromide to the same ketal afforded the 3β-phenyl (8) and 3β-benzyl (9) derivatives,respectively. 3α-Hydroxy-3β-phenylethynyl-5α-pregnan-20-one derivatives (7, 12–17) were synthesized in ∼30% to 50% yields using the previously described (Upasani et al., 1997) coupling reaction of 3β-ethynyl-3α-hydroxy-5α-pregnan-20-one (Hogenkamp et al., 1997) with the correspondingp-substituted iodobenzenes in the presence of catalytic amounts of (PPh3)2PdCl2and CuI. The 3β-phenylethyl derivative (10) was prepared by catalytic hydrogenation of the unsaturated analog (11), which was synthesized from (3R)-spiro[oxirane-2′,5α-pregnan]-20-one (Hogenkampet al., 1997) by reaction with methyl phenyl sulfoxide anion and elimination of the γ-hydroxy sulfoxide formed (Hogenkamp, 1995). All the compounds prepared were purified by column chromatography over silica gel. Purity was ascertained by thin layer chromatography and routine spectral analysis (IR and NMR spectroscopy).
Receptor Source for Binding Assays
Stable GABAA γ2Lcell line preparation.
Human α1, α2, α3 and γ2L GABAA receptor subunits were a gift from Peter Seeburg (University of Heidelberg, Germany). Human α4, α6 and α2 subunits were cloned as described (Yang et al., 1995). Human α5 was cloned from human brain by PCR utilizing oligonucleotide primers corresponding to the proposed ends of the coding region based on the human α5 genomic sequence (Knoll et al., 1993). The amino acid sequence derived from this cDNA was identical to the amino acid sequence previously reported (Wingrove et al., 1991). Human β3 (Wafford et al., 1994) was cloned from human brain by PCR utilizing oligonucleotide primers derived from the published sequences corresponding to the ends of the coding region. All plasmid DNA for transfection was prepared using two cycle cesium chloride gradient centrifugation. The transfection and stable cell line cloning of the HEK293 cells (CRL 1573; American Type Culture Collection) follows the protocol reported previously (Hawkinsonet al., 1996).
Membrane preparation.
Membranes from stable HEK293 cell lines expressing human recombinant GABAA receptor subunit combinations and well-washed rat brain cortical homogenates were prepared as described previously (Hawkinson et al., 1996).
Radioligand Binding
[35S]TBPS assay.
Steroid inhibition of 2 nM [35S]TBPS (60–100 Ci/mmol; NEN) binding was examined in 200 mM NaCl/50 mM sodium-potassium phosphate buffer (pH 7.4) as previously described (Carter et al., 1997; Hawkinson et al., 1994a, 1996). The GABA concentration was either the approximate IC50 for inhibition of TBPS binding (rat brain) or the concentration producing the peak TBPS binding from the biphasic GABA concentration-effect curve (recombinant receptors) as indicated in table 6. Incubations contained ∼350, 100, 100, 120, 140, 200, or 200 μg protein for rat brain, α1β2γ2L, α2β2γ2L, α3β2γ2L, α4β3γ2L, α5β2γ2L and α6β3γ2L membranes, respectively. The incubation and filtration were conducted as previously described (Hawkinsonet al., 1996) or in 96-well plates (2.0 ml; Beckman) followed by filtration through GF/B 96-well filter plates (Packard) and rinsed 3 times with ∼1.5 ml ice-cold assay buffer. In the latter case, Microscint scintillation cocktail (50 μl; Packard) was added to each well of the dried filter plates, which were then sealed, shaken vigorously for 5 min and counted for 5 min/well on a TopCount 6-detector scintillation counter (Packard).
[3H]Flunitrazepam assay.
Steroid enhancement of 1 nM [3H]flunitrazepam (84.5 Ci/mmol; NEN) binding in well-washed rat brain cortical P2 membranes was examined in 200 mM NaCl/50 mM sodium-potassium phosphate buffer (pH 7.4) in the presence of 1 μM GABA as previously described (Carteret al., 1997; Hawkinson et al., 1994a; Hawkinsonet al., 1996).
[3H]Muscimol assay.
Steroid enhancement of 5 nM [3H]muscimol (10.1 Ci/mmol; NEN) binding in well-washed rat brain cortical P2 membranes was examined in sodium-free buffer (100 mM KCl/40 mM potassium phosphate, pH 7.4) as previously described (Carter et al., 1997;Goodnough and Hawkinson, 1995; Hawkinson et al., 1996).
Data analysis
Nonlinear curve fitting of the overall data for each drug averaged for each concentration was performed using the sigmoidal equation in Prism (GraphPad). The data were fit to a two component instead of a one component model if the sum of squares was significantly lower by F-test. The concentration of test compound producing 50% inhibition (IC50) or enhancement (EC50) of specific binding, the extent of inhibition (% I) or enhancement (% E) corresponding to each component for two component modulators, and the maximal extent of inhibition (Imax) or enhancement (Emax) were determined for the individual experiments with the same model used for the overall data and then the mean ± S.E. of the individual experiments were calculated.
Electrophysiology
Receptor expression and recording in Xenopusoocytes.
RNA was prepared as previously described (Hawkinsonet al., 1996) and stored at −80°C. Preparation and microinjection of oocytes were performed as reported previously (Woodward et al., 1995). Individual oocytes were injected with ∼1 ng each of cRNA encoding the α1, β2 and γ2L subunits, and oocytes were stored in Barth’s medium containing (in mM): NaCl, 88; KCl, 1; CaCl2, 0.41; Ca(NO3)2, 0.33; MgSO4, 0.82; NaHCO3, 2.4; HEPES 5; pH 7.4, with 0.1 mg/ml gentamycin sulfate. Individual oocytes were placed on a mesh in a standard 35-mm culture dish perfused with frog Ringer’s solution containing (in mM): NaCl, 115; KCl, 2; CaCl2, 1.8; HEPES, 5; pH 7.4. Electrical recordings were made using a Dagan TEV-200 voltage clamp. Steroids were initially diluted into DMSO stocks (10 nM to 10 mM) and further diluted into Ringer just prior to experiments. The final DMSO concentration was 0.3%, which had no effect by itself. Drug solutions were applied to oocytes via a triple-barrel linear array as described in detail previously (Hawkinson et al., 1996). Modulatory effects were measured after 1- to 2-min preincubations with steroids, followed by exposure to a mixture of steroid and GABA. Maximal GABA responses were measured before and after steroid modulation experiments, and any change in the maximum current was factored in by calculating fractional currents against a linear sliding scale.
Experimental design and data analysis.
GABA concentration-response data were obtained by successive brief exposures to increasing concentrations of GABA, until an apparent maximal current was reached (1–3 mM GABA). These data were fit to the logistic equation (Eq. 1) using Origin (Microcal), where FR = I/GABAmax, n is the slope, EC50 is the concentration that produces a half-maximal response, I is the current at a given concentration of GABA (agonist) and GABAmax is the maximal current in response to GABA.
In Vivo Pharmacology
Animals.
Male NSA mice weighing between 15 and 20 g were obtained from Harlan Sprague-Dawley, Inc. Upon arrival they were housed in standard polycarbonate cages (4 per cage) containing a sterilized bedding material (Sani-Chips, P.J. Murray) in a room of constant temperature (23.0° ± 2.5°C) with a 12 hr (7:00 a.m. to 7:00 p.m.) light/dark cycle. Food (Teklad LM 485; Harlan Sprague-Dawley) and water were freely available. Animals were acclimated a minimum of 4 days prior to experimentation.
PTZ-induced seizures.
Seizures were induced by administration of 85 mg/kg, s.c. PTZ (30 min observation period). The dose of PTZ used was previously determined to be the dose producing convulsions in 97% of animals (CD97). A clonic seizure was defined as forelimb clonus of ≥ 3 sec duration. Data were treated quantally.
Motor function.
The rotorod test used a custom-built apparatus that consisted of an elevated drum of textured surface (diameter: 2.5 cm) that rotated at a constant speed (6 rpm). The height of the drum from the floor of the test apparatus was ∼30 cm. Prior to administration of test substance, animals were trained to walk continuously on the drum for a period of 2 min. During testing, animals were given 3 opportunities to remain on the apparatus continuously for 1 min. LRR was also determined in mice. Results were treated quantally.
Pharmacologic procedure.
PTZ was obtained from Sigma Chemical Co. and was dissolved in physiologic saline (0.9%). Neuroactive steroids were dissolved in hydroxypropyl-β-cyclodextrin (Amazio) 50%: distilled water 50% and were placed in solution by warming and sonication for 1–4 hrs. Solutions were prepared on a weight/volume basis on the day of, or evening prior to, use. PTZ was administered s.c.; neuroactive steroids were administered i.v., i.p. or p.o. Drugs were administered in volumes of 100, 100 and 400 μl/20 g for i.v., i.p. and p.o. dosing, respectively.
Data Analysis
Dose-response functions were constructed for graphical presentation by converting the quantal response data to percentages and calculating the mean ± S.E. for each dose of 3 independent experiments. The dose of drug required to produce an anticonvulsant effect (ED50), loss-of-righting reflex (ED50), or motor impairment (TD50) in 50% of animals and its associated 95% confidence limits was calculated on the quantal sum of the data by the method of Litchfield and Wilcoxon (1949) using a commercial computer program (PHARM/PCS v4.2; MicroComputer Specialists). The TI was calculated by dividing the TD50 by the PTZ ED50.
Results
Structure-activity of 3β-substituted 3α-hydroxy-5α-pregnan-20-ones defined by [35S]TBPS binding in rat brain membranes.
Substitution of the 3β position of 3α,5α-P (compound 1) with short chain alkyl groups (compounds2 and 3) reduced potency for inhibition of [35S]TBPS binding, whereas unsaturation of the side chain (compounds 4 and 5) reversed this reduction as previously reported (Hogenkamp et al., 1997) (table 1). Further elongation of the optimal two carbon ethynyl unit (compound 5) with ann-butyl moiety (compound 6) did not alter potency. Extension of the ethynyl unit with a phenyl group resulted in compound (7), which displayed a two component, partial inhibition curve. Considering the high affinity component only, phenyl modification of the ethynyl group further increased potency.
In view of the high potency imparted by 3β-phenylethynyl substitution (compound 7), the location of the phenyl group and flexibility of the spacer group were evaluated (table2). Very low potency was observed if the phenyl group is attached directly to the steroid A-ring (compound8) or if a methylene (compound 9) or ethylene (compound 10) spacer is used. Unsaturation of the ethylene spacer resulted in active compounds with either very low efficacy in the case of the ethenyl spacer (compound 11) or two component inhibition for the ethynyl spacer (compound 7).
Para-substitution of the 3β-phenylethynyl group with methyl (compound 12), chloro (compound 13), or hydroxy (compound 14) did not appreciably alter the potency or two component profile relative to the unsubstituted compound7, except that the high affinity component of compound14 was 5-fold less potent than the high affinity component of compound 7 (table 3). In contrast, para-substitution with methoxy (15), acetyl (16; Co 152791) or carbethoxy (17) resulted in compounds displaying one component inhibition, withp-acetyl providing optimal potency. The inhibition curves for compounds substituted in the para-position with hydrogen bond donating (p-hydroxy 14), weak hydrogen bond accepting (p-chloro 13) and strong hydrogen bond accepting (p-acetyl 16; Co 152791) groups are compared to 3α,5α-P (1) and compound 18, the 5β-epimer of 16 (fig. 2). These neuroactive steroids were examined further both in vitro and in vivo.
Modulatory profile of selected neuroactive steroids in the [3H]flunitrazepam and [3H]muscimol binding assays.
In the [3H]flunitrazepam binding assay in rat brain membranes, all of the neuroactive steroids examined exhibited one component enhancement curves (fig. 3). The p-acetyl compounds 16 (Co 152791) and18 displayed the highest potency of the compounds tested, although the p-chloro compound 13 was also more potent than 3α,5α-P (1) (table4). In contrast to the [35S]TBPS result, the p-hydroxy compound 14 displayed only low affinity modulation of [3H]flunitrazepam binding. Although major differences in efficacy of modulation were not observed, thep-acetyl compounds 16 (Co 152791) and18 had higher Emax values than 3α,5α-P (1), whereas the p-chloro compound13 showed lower efficacy.
In the [3H]muscimol binding assay in rat brain membranes, the p-acetyl derivative in the 5β series (18) and 3α,5α-P (1) displayed two component enhancement, although compound 18 was significantly more potent and was the most efficacious steroid tested (table 4, fig.4). In contrast, all of the lower efficacy compounds displayed one component enhancement as noted previously for limited efficacy 20- and 21-hydroxy pregnanes (Goodnough and Hawkinson, 1995). The p-acetyl derivative in the 5α series (16; Co 152791) was the most potent modulator (EC50 2 nM) and also had relatively low efficacy, although the p-chloro compound (13) had the lowest efficacy of the compounds tested. As in the [3H]flunitrazepam assay, thep-hydroxy compound 14 displayed only low affinity modulation of [3H]muscimol binding.
Profile of compounds 13 and 16 (Co 152791) relative to 3α,5α-P in human recombinant GABAA receptors.
Compounds 13 and 16 (Co 152791) were potent inhibitors of [35S]TBPS binding in membranes prepared from stable HEK cell lines expressing human α1β2γ2L, α2β2γ2L, α3β2γ2L and α5β2γ2L subunit combinations, with IC50 values ranging from 1.4 to 12 nM (table 5, fig.5). In these receptor combinations, IC50 values for 3α,5α-P ranged from 20 to 40 nM (Hawkinson et al., 1996). In the α4β3γ2L and α6β3γ2L combinations, 3α,5α-P inhibited [35S]TBPS binding with much lower potency (IC50 1700 and 1060 nM, respectively), but retained high efficacy (Imax > 75%). In contrast, the phenylethynyl derivatives 13 and 16(Co 152791) were considerably more potent at α4β3γ2L and α6β3γ2L subunit combinations (IC50 27–210 nM), although both compounds displayed limited efficacy for inhibition of [35S]TBPS binding at these receptors (table5, fig. 5).
Electrophysiological characterization of selected neuroactive steroids at α1β2γ2L receptors expressed in Xenopusoocytes.
Oocytes showed robust expression of functional GABAA receptors 5–17 days after injection with a mixture of α1, β2 and γ2L cRNAs. Maximal current response to 10 mM GABA was 2400 ± 100 nA (n = 19) from two batches of oocytes from two separate frogs. The EC50 value for GABA in these oocytes was 24 ± 2 μM, with a slope of 1.3 ± 0.1 (n = 19), consistent with a single population of α1β2γ2L receptors. Modulatory effects of steroids on α1β2γ2L receptors were assayed using control GABA-evoked currents that were ∼5% of maximal GABA responses in each individual oocyte. Mean concentrations of GABA used to elicit 5% responses were 5.1 ± 0.6 μM (n = 17).
Compound 16 (Co 152791) was a remarkably potent potentiator of GABA-activated currents, having an EC50 of 0.87 nM and a maximal potentiation of 0.77 expressed as a fractional response (FR) of the peak current elicited by 10 mM GABA (fig.6, top; table6). Compound 13 was nearly as potent as a modulator, but had lower efficacy (FR 0.46). Other steroids in this series were less potent, with a range of efficacies (fig. 6, top; table 6). In particular, compound 14 displayed very low efficacy (FR 0.14). For comparison, the full agonist 3α,5α-P (1) evoked maximal potentiation of 0.91 with an EC50 of 160 nM (Hawkinson et al., 1996). The potency of Co 152791 for modulation of GABA currents was compared with that produced by 3α,5α-P, the benzodiazepine diazepam, and the barbiturate pentobarbital, using concentrations that resulted in an approximate doubling of the GABA control response (fig.6, bottom). The magnitude of direct activation was less than 1% of the GABAmax for the compounds evaluated as part of this study at the concentrations examined (FR < 0.01). In contrast, 10 μM 3α,5α-P (1) directly activates α1β2γ2L receptors expressed in oocytes with FR values between 0.15 and 1.2 (Hogenkamp et al., 1997).
In vivo profiles of selected neuroactive steroids.
Selected neuroactive steroids were evaluated for in vivopharmacological activity and compared to reference steroids. Dose-response data for protection against clonic seizures induced by s.c. PTZ administration in mice are summarized in table7. Consistent with the in vitro data, Co 152791 was a potent anticonvulsant, displaying an ED50 of 0.6 mg/kg, i.p. for inhibition of PTZ-induced clonic seizures (fig. 7). A comparable increase in ataxic potency relative to 3α,5α-P (1) was observed in the rotorod test after i.p. administration (TD50 4.8 mg/kg), resulting in a slightly better therapeutic index for Co 152791 (TI 8.0) than 3α,5α-P (TI 6.7). In contrast to 3α,5α-P, Co 152791 retained activity after oral administration (ED50 1.1 mg/kg). A similar shift in ataxic potency (TD507.7 mg/kg) resulted in a TI of 7.0. Conversely, compounds 13and 18 were less potent than 3α,5α-P, although compound18 retained oral activity. Compound 14 was inactive i.p. in the PTZ assay. Anesthetic activity of some of these steroids was examined by their ability to induce loss-of-righting reflex (LRR) in mice. Following i.v. administration, Co 152791 (ED50 2.4 mg/kg) was 2.3- and 4.5-fold more potent than compound 18 and 3α,5α-P, respectively, for induction of LRR.
Discussion
Substitution of the naturally occurring progesterone metabolite 3α,5α-P at the 3β-position has lead to the discovery of Co 152791 (3α-hydroxy-3β-(p-acetylphenylethynyl)-5α-pregnan-20-one; compound 16), which is the most potent known neuroactive steroid and may well be most potent GABAAreceptor modulator known. Co 152791 modulated the binding of GABAA receptor radioligands in rat brain with IC50 or EC50 values of 2–7.5 nM. This neuroactive steroid was 11-, 16- and 49-fold more potent than the endogenous neuroactive steroid 3α,5α-P (1) in the [35S]TBPS, [3H]flunitrazepam and [3H]muscimol assays, respectively. Similarly, Co 152791 inhibited [35S]TBPS binding with IC50 values of 1.4–5.7 nM in the human recombinant receptor combinations α1β2γ2L, α2β2γ2L, α3β2γ2L and α5β2γ2L, being 3- to 20-fold more potent than 3α,5α-P. This compound was remarkably potent in potentiating GABA-evoked currents in Xenopus oocytes expressing α1β2γ2L receptors (EC50 0.87 nM), being 184-fold more potent than 3α,5α-P.
The high potency of Co 152791 in vitro was also observedin vivo. Thus, Co 152791 exhibited exceptional anesthetic and anticonvulsant potency, inducing loss-of-righting reflex and protecting against clonic seizures induced by PTZ in mice with an ED50 of 2.4 mg/kg, i.v. and 0.6 mg/kg, i.p., respectively. In both instances, Co 152791 was ∼5 times more potent than 3α,5α-P in vivo. Although an increase in ataxic potency determined by impairment of rotorod performance was also observed after i.p. administration (TD50 4.8 mg/kg), Co 152791 exhibited a wide separation between anticonvulsant and ataxic activities, reflected in a therapeutic index of 8. Although slightly less active after oral administration, Co 152791 displayed potent anticonvulsant activity (PTZ ED50 1.1 mg/kg) with a TI of 7, superior to that for any previously reported orally active neuroactive steroid (Carter et al., 1997;Gasior et al., 1997; Kokate et al., 1994; Wielandet al., 1995).
The structure-activity relationship (SAR) for 3β-substituted derivatives of 5β-pregnane steroid modulators of the GABAA receptor indicates the presence of an auxiliary pocket in the neuroactive steroid binding site near the region occupied by the steroid A-ring (Upasani et al., 1997). The SAR in the 5α series reported here confirms this interpretation. As in the 5β series, 3β substitution of 3α,5α-P (1) with small alkyl groups reduced potency in the [35S]TBPS binding assay, although unsaturation of the side chain reversed this decrement so that the 3β-ethynyl derivative 5 had similar potency to 3α,5α-P. Although the effect of unsaturation could be due to reduction in the effective size of the substituent, this is unlikely since extension of the ethynyl group with phenyl actually increased potency. Indeed, a spacer group is required to extend the phenyl group from the steroid A-ring and this spacer must be unsaturated, with ethynyl being optimal. Compounds with spacers of 0, 1, or 2 carbon atoms were essentially inactive (IC50 > 10 μM). The key features of the ethynyl spacer are its length and rigidity, which places the phenyl group in a constrained volume in the binding pocket. These requirements are more critical in the 5α series since saturated spacers or direct attachment of the phenyl group to the 3β position in the 5β series results in compounds which retain moderate activity (IC50 100–400 nM) (Upasani et al., 1997). para-Substitution of the phenyl ring with the hydrogen bond accepting acetyl group as in Co 152791 confers optimal potency.
Substitution of the para position of the phenyl ring with groups that do not hydrogen bond or are weaker hydrogen bond acceptors than acetyl results in compounds having lower potency. Thus, thepara-unsubstituted (7) and thepara-methyl (12), -chloro (13), -methoxy (15) and -carbethoxy (17) compounds were 2.1- to 4.5-fold less potent as inhibitors of [35S]TBPS binding in rat brain (high affinity components). Compounds 7, 12, 13 and15 were also 2.2- to 13-fold less potent as potentiators of GABA-evoked currents in oocytes expressing α1β2γ2L receptors. Similar effects of para-substitution have been observed in the 5β series (Upasani et al., 1997). The corresponding 5β analog 18 of the highly potent 5α steroid Co 152791 was also consistently less potent in all assays.
Although Co 152791 (compound 16) was the most potent compound in vitro and in vivo, the correlation between in vitro and in vivo potency did not extend to all compounds examined. For example, compound 18, the 5β-epimer of 16, was ∼3-fold less potent than 3α,5α-P as an anticonvulsant, but was more potent than 3α,5α-P in vitro by a factor of 5- to 24-fold. Similarly, compound 13 was ∼2-fold less potent than 3α,5α-P as an anticonvulsant, but was consistentlymore potent than 3α,5α-P in vitro, particularly in electrophysiological assays where it was > 80-fold more potent than 3α,5α-P. Presumably, 3α,5α-P has better bioavailability than compounds 13 and 18after i.p. administration. The situation is reversed after oral administration in that compound 18 retains activity whereas 3α,5α-P is inactive, consistent with previous reports (Carteret al., 1997). Although lack of 3β substitution probably contributes to the lack of oral activity of 3α,5α-P, 3β substitution per se does not automatically confer oral activity as compound 13 was inactive orally.
In addition to potency differences, para-substitution of the 3β-phenylethynyl group in the 5α series also affects the efficacy of modulation. In the [35S]TBPS binding assay in rat brain membranes, the unsubstituted (7),p-methyl (12), p-chloro (13) and p-hydroxy (14) compounds displayed two component modulation with high and low affinity components corresponding to 27–40% and 36–57% of maximal inhibition, respectively. In contrast, para-substitution with hydrogen bond acceptors, such as methoxy (compound 15), acetyl (compound 16; Co 152791) and carbethoxy (compound17), resulted in compounds that exhibited only high affinity binding. Thus, para-substitution with strong hydrogen bond accepting groups increases the proportion of the high affinity component relative to para-substitution with groups that are hydrogen bond donors, do not hydrogen bond, or are weak hydrogen bond acceptors. These effects of hydrogen bonding groups on [35S]TBPS binding in the 5α-pregnane series are similar, but not identical, to that observed in the 5β series, where hydrogen bonding groups affect the potency but not the efficacy of modulation (Upasani et al., 1997).
Two component modulation of radioligand binding to the GABAA receptors present in brain membranes by neuroactive steroids has been noted in several cases (Goodnough and Hawkinson, 1995, 1994b; Zhong and Simmonds et al., 1996;Upasani et al., 1997). This phenomenon is suggestive of subtypes of GABAA receptors with differential affinities for certain neuroactive steroids, but also could be due to negative cooperativity between multiple binding sites per receptor complex, differential GABA sensitivities, partial agonism and/or complex combinations of these actions.
In an attempt to address the issue of potential neuroactive steroid subtype selectivity, the inhibition of [35S]TBPS binding by the high affinity, one component modulator 16 (Co 152791) and the two component modulator 13 was determined in membranes prepared from stable cell lines expressing six different α subunit combinations. In these six human recombinant receptors, the potency and efficacy profile for compound 13 was similar to that for Co 152791, suggesting that subtype selectivity does not explain the two component modulation of [35S]TBPS binding observed in rat brain membranes. On the other hand, two component modulation was not observed in any recombinant receptor combination examined suggesting that the two component modulation by these compounds occurs only in native receptors.
Interestingly, the profiles of the 3β-phenylethynyl substituted steroids 13 and 16 (Co 52791) differ somewhat from that for 3α,5α-P in these recombinant receptors. In α1, α2, α3 and α5-containing receptors, compounds 13 and16 (Co 152791) have higher potency (IC50 1–12 nM) and generally lower efficacy (Imax 77–93%) than 3α,5α-P (IC50 20–41 nM; Imax∼100%) as predicted from rat brain membranes. In α4 and α6-containing receptor complexes, 3α,5α-P has low potency (IC50 1–2 μM) but high efficacy (Imax 76–85%), whereas compounds 13and 16 (Co 152791) have higher potency (IC50 27–210 nM), but lower efficacy (Imax 16–40%). The low potency or efficacy of modulation of [35S]TBPS binding at α4/6β3γ2L GABAA receptor complexes suggests that these neuroactive steroids display selectivity for α1, α2, α3 and α5-containing complexes. Alternatively, these modulators may have low activity at β3 relative to β2-containing complexes. Unfortunately, direct comparisons could not be made because membranes from α4β2γ2L cells did not bind [35S]TBPS and cells expressing α6β2γ2L were not sufficiently viable.
Electrophysiological studies, although somewhat incomplete and inconsistent, do not support the finding that neuroactive steroids have low potency and/or very low efficacy at α4 and α6-containing complexes as indicated by [35S]TBPS binding. Whereas 5αTHDOC had lower efficacy for potentiation of GABA-evoked currents in α6β3γ2S than in α1β3γ2S complexes expressed in HEK 293 cells (Zhu et al., 1996), 3α,5α-P produced higher maximal potentiation in oocytes expressing α6β1γ2L complexes relative to complexes containing α1, α2, or α3 subunits (Lambert et al., 1996). In both expression systems, these neuroactive steroids had similar modulatory potency at the receptor combinations examined (Lambert et al., 1996; Zhu et al., 1996). In the case ofalpha-4-containing complexes, 3α,5α-P (30 nM) produced similar levels of potentiation at α4β2γ2L as that observed for α1β2γ2L receptors expressed in oocytes (Whittemore et al., 1996).
In addition to differences observed between native and recombinant receptors, the apparent efficacy of the neuroactive steroids examined was assay-dependent, further complicating determination of their true modulatory efficacy. In the [3H]muscimol assay in brain membranes, the p-acetyl compound in the 5β series (18) was a two-component modulator, as has previously been shown for 3α,5α-P (1) (Carter et al., 1997;Goodnough and Hawkinson, 1995). Thus, compounds displaying two component enhancement in the [3H]muscimol assay are different from those exhibiting two component inhibition in the [35S]TBPS assay. Large efficacy differences were observed in the [3H]muscimol assay, with the highest overall enhancement observed for the two component modulators 18 and 1 (overall Emax 81 and 53%, respectively). The one component [3H]muscimol modulators 13and 16 (Co 152791) displayed reduced enhancement (Emax 23% and 42%, respectively). In contrast, all of the steroids examined displayed one component modulation of [3H]flunitrazepam binding with relatively small efficacy differences, although the p-chloro compound13 had a relatively low maximal enhancement (Emax 54%) relative to the p-acetyl compounds 16 (Co 152791) and 18(Emax 73%). These different profiles may be due in part to the differential influence of GABA on steroid modulation of the binding of these three radioligands.
Electrophysiological assays appear to be better in quantifying the relative efficacy of neuroactive steroids compared to allosteric binding assays. In Xenopus oocytes expressing α1β2γ2L receptors, steroid potentiators of GABA-evoked currents can be grouped into high efficacy (compounds 1 and 18; FR 0.91–0.94), intermediate efficacy (compounds 12,15 and 16; FR 0.65–0.77), low efficacy (compounds 7 and 13; FR 0.39-.46) and very low efficacy (compound 14; FR 0.14). Although these low efficacy compounds may be partial agonists, this possibility was not explored in antagonism experiments. In this regard, 3α-hydroxy-3β-trifluoromethyl-5α-pregnan-20-one (Co 2–1970;Hawkinson et al., 1996) and 3α,21-dihydroxy-5β-pregnan-20-one (5βTHDOC; Xue et al., 1997) have previously been shown to be partial agonists for the neuroactive steroid site. The two component modulators in the [35S]TBPS assay showed limited efficacy for potentiation of GABA-evoked currents and the potencies of the high affinity components in the [35S]TBPS assay appear to correspond to their potencies in the electrophysiological assay. Binding assays are useful in predicting neuroactive steroid potency (Hawkinson et al., 1994b; Hogenkamp et al., 1997; Upasani et al., 1997), whereas electrophysiological measurements may be required to establish compound efficacy.
The p-hydroxy derivative 14 retained reasonable activity in both [35S]TBPS and electrophysiological assays (IC50 and EC50 110 nM), although had only micromolar activity in the [3H]flunitrazepam and [3H]muscimol assays and was inactive in vivo. Apparently, these latter assays did not detect the potent, low efficacy modulation observed in the [35S]TBPS and electrophysiological assays. Although the lack of in vivo activity may be due to the low efficacy of this compound, other possibilities include poor bioavailability and/or metabolic lability of the p-hydroxy group.
A simplified pharmacophore model is presented that describes the key features of the interactions between neuroactive steroids and their binding site on GABAA receptors (fig.8). In this model, the steroid backbone occupies a hydrophobic region in the binding site and acts as a scaffold to maintain the requisite 3α-hydroxy and facilitory 20-keto groups in appropriate positions to make hydrogen bonding interactions with amino acid residues located in the primary binding pocket (labeled A and B). In addition, an auxiliary binding pocket exists that is accessed by a rigid spacer extending from the 3β-position. In this model for 5α steroids, the hydrogen bonding interaction in the auxiliary binding pocket (labeled C) controls the potency as well as the efficacy of modulation of the receptor. In contrast, this hydrogen bonding interaction in 5β steroids affects the potency, but not efficacy, of modulation (Upasani et al., 1997). Although the reason for this difference between 5α and 5β steroids is unclear, the orientation of the steroid A-ring and its 3β substituent relative to the steroid backbone is not the same in 5α and 5β steroids, apparently resulting in a qualitatively different interaction with the binding site. It should be noted that this pharmacophore model is less useful in predicting the in vivo relative to the in vitro activity of neuroactive steroids, probably due to differences in absorption and metabolism between compounds.
In conclusion, 3α-hydroxy-3β-(p-acetylphenylethynyl)-5α-pregnan-20-one (compound 16; Co 152791) is the most potent known neuroactive steroid positive allosteric modulator of GABAA receptors. The remarkable potency of this steroid is consistent with the presence of an auxiliary binding pocket containing a hydrogen bond donating amino acid residue that interacts favorably with the p-acetyl moiety of the 3β-phenylethynyl substituent. The high receptor potency and probable enhanced bioavailability resulting from blockade of metabolism of the 3α-hydroxy group imparted by the 3β-(p-acetylphenylethynyl) substituent contribute to the exceptional anticonvulsant potency of this compound in the mouse pentylenetetrazol assay. Moreover, the high therapeutic index relative to generalized sedation, particularly after oral administration, impart Co 152791 with a favorable anticonvulsant profile relative to currently used antiepileptic agents.
Acknowledgments
The authors thank Silvia Robledo and Michael Suruki for their excellent technical assistance in the conduct of animal experiments.
Footnotes
-
Send reprint requests to: Dr. Ravindra B, Upasani, CoCensys, Inc., 201 Technology Dr., Irvine, CA 92618.
- Abbreviations:
- Co 152791
- 3β-(p-acetylphenylethynyl)-3α-hydroxy-5α-pregnan-20-one
- DMSO
- dimethylsulfoxide
- Emax
- maximal extent of enhancement or potentiation
- FR
- fractional response
- GABA
- γ-aminobutyric acid
- GABAmax
- the maximal current in response to GABA
- HEK
- human embryonic kidney
- Imax
- maximal extent of inhibition
- 3α
- 5α-P, 3α-hydroxy-5α-pregnan-20-one
- LRR
- loss-of-righting reflex
- PCR
- polymerase chain reaction
- PTZ
- pentylenetetrazol
- SAR
- structure-activity relationship
- [35S]TBPS
- [35S]t-butylbicyclophosphorothionate
- TI
- therapeutic index
- 5αTHDOC
- 3α,21-dihydroxy-5α-pregnan-20-one
- 5βTHDOC
- 3α,21-dihydroxy-5β-pregnan-20-one
- Received December 11, 1997.
- Accepted May 22, 1998.
- The American Society for Pharmacology and Experimental Therapeutics