Introduction

Top Abstract Introduction Methods Results Discussion References

The clinical efficacy of benzodiazepines as sedative/hypnotics in unquestioned. These compounds are limited, however, by negative effects on psychomotor performance, by interactions with alcohol and by dependence liability (Shader and Greenblatt, 1993). A recent advance in the hypnotic area has been the clinical introduction of zolpidem, a nonbenzodiazepine with an approximately 20-fold greater affinity for benzodiazepine receptors on GRCs containing 1 subunits than those containing 2 or 3 subunits and a >800-fold greater affinity for 1 over 5 containing GRCs (Pritchett and Seeburg, 1990). This increase in benzodiazepine receptor subtype specificity appears to result in an improved hypnotic profile over classical benzodiazepines; however, the 20-fold separation in receptor subtype affinities defines the narrow dosing range needed not to disrupt normal sleep architecture and maintain a superior clinical profile. Dosing above this range can result in disruptions of normal sleep architecture (Hoehns and Perry, 1993) and rebound insomnia on the next night's sleep (Roehrs et al., 1986).

Barbiturates, which are also allosteric modulators of the GRC, also have been used extensively as hypnotics, with their clinical use limited by their abuse potential and low therapeutic indices (Mellinger et al., 1985). Neuroactive steroids are the first class of endogenous compounds (Hu et al., 1987) known to act as positive allosteric modulators of the GRC (Cottrell et al., 1987; Gee et al., 1988; Lan et al., 1990, 1991; Puia et al., 1990; Shingai et al., 1991; McNeil et al., 1992; Paul and Purdy, 1992; Woodward et al., 1992). Consistent with their site of action, they have also been shown to possess hypnotic and anesthetic actions (Atkinson et al., 1965; Gyermek, 1967; Mendelson et al., 1987; Mok and Krieger, 1990; Steiger et al., 1993). Although significant synthetic efforts have been applied to the design of neuroactivesteroids with i.v. anesthetic activity (Phillipps, 1975), the design of orally bioavailable hypnotics has not been fully assessed. We present the pharmacological profile of CCD-3693, a synthetic orally bioavailable analog of the endogenous neuroactive steroid, pregnanolone. The overall pharmacological profile of this compound is compared with that of pregnanolone and with the approved hypnotics, triazolam (Edgar et al., 1991a and b) and zolpidem (Depoortere et al., 1986).

	Methods

Top Abstract Introduction Methods Results Discussion References

Materials. TBPS (60-100 Ci/mmol) and [³H]flunitrazepam (74-84 Ci/mmol) were obtained from NEN (Boston, MA). HPCD was obtained from American Maize Products (Indianapolis, IN). Triazolam and sterile 0.25% methylcellulose and zolpidem were the generous gifts from Upjohn (Kalamazoo, MI) and Synthelabo (Le Plessis-Robinson, France), respectively. All other reagents and drugs were obtained from Sigma Chemical Co. (St. Louis, MO).

In Vitro Pharmacology

Allosteric modulation of [³⁵S]TBPS binding to rat brain cortex. The rat neocortical membrane preparation and subsequent binding assay were performed as described in detail previously (Hawkinson et al., 1994). All values are the means of at least three replicate experiments with less than 10% S.D. Briefly, fresh rat cortical P₂ membrane preparations were washed three times and used for assay with 2 nM [³⁵S]TBPS in the presence of 5 µM GABA and using 2 µM cold TBPS to define nonspecific binding. Incubations were 90 min at room temperature and were terminated by filtration.

Allosteric modulation of [³H]flunitrazepam binding to rat brain cortex. The rat neocortical membrane preparation and subsequent binding assay were performed as previously described in detail (Hawkinson et al., 1994). All values are the means of at least three replicate experiments with less than 10% S.D. Briefly, washed P₂ membrane preparations were incubated with 1 nM [³H]flunitrazepam, in the presence of 1 µM GABA, for 90 min at room temperature and the incubations terminated by filtration. Nonspecific binding was determined with 1 µM cold clonazepam.

Behavioral Pharmacology

The behavioral testing of pregnanolone, CCD-3693 (fig. 1), triazolam and zolpidem was conducted in male mice and rats (Harlan Sprague-Dawley, Indianapolis, IN) according to procedures previously described in detail (Wieland et al., 1995). However, the in vivo evaluation of neuroactive steroids is complicated significantly by the difficulties associated with formulating these compounds. Initial studies were undertaken using 0.25% methylcellulose as a vehicle; however, subsequent studies used HPCD (American Maize Products, Indianapolis, IN). For i.p. or s.c. dosing 50% HPCD (w/v in 0.9% NaCl) solutions were used although oral dosing was with 10 to 20% HPCD solutions. For these studies, initial drug solutions were made in 50% HPCD (15 mg/ml of CCD-3693 and 40 mg/ml of pregnanolone), and subsequent dilutions made with 0.9% NaCl. In the rat i.p. and s.c. dosing was done at a volume of 1 ml/kg, although oral dosing was up to 10 ml/kg. With mice the corresponding numbers were 100 µl/20 g, i.p. and 400 µl/20 g, p.o.

View larger version (K): [in this window] [in a new window]   Fig. 1.   Structure of the endogenous neuroactive steroid, pregnanolone and the synthetic neuroactive steroid, CCD-3693.

Anxiolytic testing. The elevated plus-maze (Lister, 1987), was used as an ethological model in NIH Swiss-Webster mice and the Geller-Seifter conflict procedure was performed with Sprague-Dawley rats.

Anticonvulsant testing. Anti-PTZ activity was evaluated both in Sprague-Dawley rats and CF/1 mice. The PTZ dose used in mice was 85 mg/kg, s.c. and in rats was 70 mg/kg, s.c.

Motor function. Roto-rod deficits were evaluated both in the rat and mouse. LRR was also determined in mice. The ability of drugs to potentiate the motor deficits of ethanol (Frye and Breese, 1982) were evaluated in the rat (1.0 g ethanol/kg) roto-rod paradigm and in the hanging wire-mesh test for mice (1.5 g ethanol/kg).

Passive-avoidance paradigm. The acquisition test was conducted according to previously published methodologies (Holmes and Drugan, 1991). The CF/1 mice were placed in the "bright" side of a two-compartment shuttle box (GEMINI Avoidance System; San Diego Instruments, San Diego, CA) and allowed to explore the chamber for 120 sec. At the end of the exploration period, a guillotine door opened to allow access to a "dark" chamber. The mice had 120 sec to enter the dark chamber. If the mice failed to enter the dark chamber within 120 sec they were gently pushed across. Upon entering the dark chamber, the guillotine door closed and the mice received a footshock of 0.2 mA intensity and 2 sec. The mice were removed immediately and placed into their home cages. To monitor drug effects on acquisition, they were administered 10 min before behavioral testing.

Pharmacokinetics. The oral pharmacokinetic profile of pregnanolone was evaluated after a dose of 100 mg/kg. The time points evaluated were 0.5, 1, 2 and 4 hr. To evaluate the pharmacokinetic profile of CCD-3693 after acute and repeated dosing, rats were dosed orally with 15 mg/kg and plasma samples taken at 0.25, 0.5, 0.75, 1, 1.5 and 2 hr after drug administration. Animals that had been dosed daily for 3 days at 15 mg/kg, were evaluated identically on day 4. Plasma samples were extracted with hexane and the hexane extracts analyzed by GC-MS using a DB-17 column (J&W; 30 m × 0.32 mm i.d; 0.1-µm film thickness; 180°C for 2.8 min followed by a 20°C/min gradient to 230°C), for chromatography and a Finnigan ITS 40 as the detector. Total ion current chromatograms were integrated for quantitation.

Statistics. ED₅₀ and TD₅₀ values were calculated for each experiment within each drug using the method of Litchfield and Wilcoxon (1949). Calculations were computed using a commercially available computer program included in the "Manual of Pharmacological Calculations with Computer Programs" (R. J. Tallarida and R. B. Murray, eds., Springer Verlag, 1987). Statistical inferences were made with the Neuman-Keuls post-hoc procedure, subsequent to an ANOVA.

Sleep-Wake Bioassay

Animal surgery. Adult, male Wistar rats (275-350 g at time of surgery, Charles River Laboratories, Wilmington, MA) were anesthetized (Nembutal, 60 mg/kg) and surgically prepared with a cranial implant that permitted chronic EEG and EMG recording. Body temperature and locomotor activity were monitored via a miniature transmitter (Minimitter) surgically placed in the abdomen. The cranial implant and telemetry surgical procedures have been described in detail elsewhere (Edgar et al., 1991a). In brief, EEG was continuously monitored using differential leads across one each of two frontal (+3.9 AP from bregma, ±2.0 ML) and two occipital (6.4 AP, ±5.5 ML). Two Teflon-coated stainless steel wires positioned under the nuchal trapezoid muscles permitted continuous EMG recording. All implants were sterilized with ethylene oxide. A minimum of 3 wk was allowed for recovery from surgery.

Recording environment. Rats were housed individually within specially modified Nalgene microisolator cages equipped with a low-torque swivel commutator and filter-top riser. These cages were located within separate, ventilated compartments of a stainless steel recording chamber. Food and water were available ad libitum. A 24-hr (LD 12:12) light-dark cycle was maintained throughout the study (32-35 lux inside the cage during lights-on, <0.05 lux during lights-off). Animals were undisturbed for 3 days both before and after treatments.

Automated data collection. Sleep and wakefulness were determined using "SCORE"a microcomputer-based sleep-wake and physiological monitoring system. A detailed description and functional validations of this system for rodents have been described elsewhere (Van Gelder et al., 1991; Edgar et al., 1991a; Seidel et al., 1995). Briefly, the system monitors amplified EEG (bandpass 1-30 Hz; digitization rate 100 Hz), integrated EMG (bandpass 10-100 Hz), T_b, nonspecific LMA and drinking activity, from up to 64 rodents simultaneously. Arousal states were classified on-line as NREM sleep, REM sleep, wake, or -dominated wake every 10 sec based on SCORE EEG feature extraction and pattern-matching algorithms. The classification algorithm used individually-taught EEG-arousal-state templates and EMG criteria to differentiate REM sleep from -dominated wakefulness, plus behavior-dependent contextual rules (e.g., if the animal was drinking, it was awake). Drinking and LMA were recorded as discrete events every 10 sec. Body temperature was recorded each minute. T_b and LMA was detected by a telemetry receiver (Datasciences, Inc., St. Paul, MN) beneath the cage. Telemetry measures (LMA and T_b) were not part of the scoring algorithm; thus, sleep-scoring and telemetry data were independent measures. The quality of data was assured by frequent on-line inspection of the signal. Graphical and statistical summaries of the 3 days before and after each animal's injection were also inspected to determine stability of the scoring. Data quality was further ensured by examining the raw EEG file (covering the first 5 hr posttreatment) for every individual treatment.

Drug administration. All treatments were administered to parallel groups under dim red illumination 6 hr after lights-out. This time-point is usually designated "CT-18" (CT = circadian time, CT-0 = lights-on). Pregnanolone, triazolam and zolpidem were suspended in sterile 0.25% methylcellulose and administered i.p. in a volume of 1 ml/kg although CCD-3693 was administered orally in 10% HPCD in a volume of 10 ml/kg. All treatment groups had a sample size of N  10, except the 30 mg/kg CCD-3693 group, for which N = 8.

Variables and statistics. NREM and REM sleep were expressed as percent of time asleep per hour. The sleep-scores were derived from automated EEG/EMG scoring, described above. LMA was measured in counts per hour. This measure was independent of sleep-scoring. T_b was normalized for each animal by computing the hourly average °C change from the 24-hr base-line (pretreatment) mean.

	Results

Top Abstract Introduction Methods Results Discussion References

In vitro binding. The actions of CCD-3693 at the GABA_A receptor complex were monitored via the negative allosteric modulation of [³⁵S]TBPS binding to the chloride channel and the positive allosteric modulation of [³H]flunitrazepam binding to associated benzodiazepine receptors. Pregnanolone inhibited [³⁵S]TBPS binding to a high affinity site with an IC₅₀ of 37 nM (64%) and a low affinity site with an IC₅₀ of 8600 nM (36%); CCD-3693 possessed a slightly lower affinity (76 nM) but recognized only one apparent binding site (see fig. 2). In the case of [³H]flunitrazepam binding, pregnanolone also demonstrated biphasic actions with a 65 nM high affinity site (56% enhancement) and a 8.6 µM low affinity site (17% enhancement), with an overall enhancement of 72%. A 70% increase in [³H]flunitrazepam binding, with an EC₅₀ of 210 nM was seen with CCD-3693 (see fig. 3).

View larger version (K): [in this window] [in a new window]   Fig. 2.   Comparison of the endogenous neuroactive steroid, pregnanolone and CCD-3693 in the [35S]TBPS binding assay. Data plotted are mean ± S.E.

View larger version (K): [in this window] [in a new window]   Fig. 3.   A comparison of the actions of pregnanolone and CCD-3693 at the GABAA receptor complex, with regard to enhancement of [3H]flunitrazepam ([3H]Flu) binding. Data plotted are mean ± S.E.

In Vivo Pharmacology

The general CNS pharmacology of CCD-3693 was evaluated both in rats and mice and comparisons made with the endogenous neuroactive steroid, pregnanolone and the hypnotics, triazolam and zolpidem.

Anxiolytic activity. All compounds, except zolpidem, demonstrated parenteral activity in the mouse elevated plus maze paradigm. Similarly, in the Geller-Seifter conflict paradigm, in rats, CCD-3693 demonstrated potent activity both after parenteral and oral administration. In contrast, pregnanolone was potent parenterally but not after oral administration (table 1).

View this table: [in this window] [in a new window]   TABLE 1 Behavioral pharmacology profiles of CCD-3693, pregnanolone, triazolam and zolpidem, using HPCD as vehicle

Anticonvulsant activity. All compounds were potent in blocking PTZ-induced convulsions, both in mice and rats, after parenteral administration (see fig. 4; table 1). However, as with anxiolytic testing, CCD-3693, but not pregnanolone, was active after oral dosing.

View larger version (K): [in this window] [in a new window]   Fig. 4.   Comparison of anticonvulsant actions of pregnanolone, CCD-3693, zolpidem and triazolam in mice after i.p. administration. Seizures were induced with pentylenetetrazol (PTZ); see table 1 for ED50 values.

Motor function. In the mouse, all compounds elicited decreased performance on the rotorod, but with TD₅₀ values that were multiples of their anticonvulsant ED₅₀ values. Loss-of-righting reflex was also noted with all compounds except triazolam. In the rat, the oral TD₅₀ of CCD 3693 on the rotorod was 30 mg/kg, distinctly higher than doses needed for potent NREM-sleep-promoting effects on rat (10 mg/kg, p.o.; see below).

Interactions with ethanol. In the mouse, all compounds potentiated the actions of ethanol in the hanging wire mesh paradigm (table 1). Similarly, CCD-3693 was also active in potentiating ethanol disruption of rat performance on the rotorod after oral administration.

Cognitive deficits. In the mouse passive avoidance paradigm, all compounds decreased acquisition; however, in the case of pregnanolone and CCD-3693, these actions only occurred at ataxic doses, although triazolam and zolpidem disrupted performance at doses significantly below ataxic doses (see fig. 5; table 1).

View larger version (K): [in this window] [in a new window]   Fig. 5.   Interference with the acquisition of passive avoidance following pregnanolone, CCD-3693, zolpidem and triazolam after i.p. administration in mice; see table 1 for MED values.

Pharmacokinetics. In the rat, oral dosing with 100 mg/kg of pregnanolone results in peak plasma levels of 40 ng/ml with rapid subsequent clearance (see fig. 6). In contrast, CCD-3693, after oral dosing with 15 mg/kg, resulted in peak blood levels of about 200 ng/ml. These data support the pharmacological data (table 1) that demonstrated improved oral activity with CCD-3693, as compared to pregnanolone. Daily dosing of rats for 3 days also did not alter the pharmacokinetic profile of CCD-3693 on day 4, suggesting that no induction of acute drug metabolism occurred. Long-term studies are still needed, however, to fully evaluate induction of drug metabolism.

View larger version (K): [in this window] [in a new window]   Fig. 6.   Blood levels of pregnanolone (left) after oral administration of 100 mg/kg in rats and of CCD-3693 (right) after 15 mg/kg, orally in naive animals (acute) and on the 4th day (chronic) after daily dosing at 15 mg/kg, orally.

Sleep-Wake Bioassay

Figures 7, left and right depict the timecourse of the effects of the higher dose of CCD-3693 (30 mg/kg p.o.) and pregnanolone (30 mg/kg i.p.) on NREM and REM sleep, LMA and T_b. The first 24 hr of each panel reveal the normal (undisturbed) circadian cycle for each variable, allowing the magnitude of the treatment effects to be appreciated relative to normal circadian variation. Active and vehicle group pretreatment baselines were closely similar.

View larger version (K): [in this window] [in a new window]   Fig. 7.   Time series plots comparing the effects of CCD-3693 (30 mg/kg; left) and pregnanolone (30 mg/kg; right) on NREM sleep, REM sleep, LMA and Tb. Light-dark cycle is indicated along the abscissa for each variable. Time of treatment (CT-18; 6 hr after lights-off) is indicated by a vertical dotted line. Data are plotted as hourly means ± S.E. (%NREM/hr; LMA counts/hr, hourly Tb difference from circadian mean). Note that both compounds markedly increase NREM for several hours posttreatment, but showed no evidence of subsequent rebound insomnia or REM sleep abnormality. Locomotor reduction was commensurate with increased sleep time, yet had little effect on the body temperature circadian rhythm.

Sleep. Pregnanolone and CCD-3693 potently promoted NREM sleep in a dose-related manner, with a rapid onset of action (fig. 8). The route of administration and vehicles for these two compounds differed (CCD-3693 p.o. vs. pregnanolone i.p.), and therefore may account for the more rapid onset of action observed for pregnanolone. Table 2 shows that 30 mg/kg of pregnanolone or CCD-3693 exerted markedly stronger NREM-promoting effects than the highest doses tested of triazolam and zolpidem, even though with respect to LMA reduction, these highest-dose treatments were roughly comparable.

View larger version (K): [in this window] [in a new window]   Fig. 8.   Soporific profiles of CCD-3693 and pregnanolone during the initial 5 hr posttreatment. CCD-3693 and pregnanolone were administered p.o. and i.p. at 10 and 30 mg/kg, respectively. Animals were treated at CT-18 (6 hr after lights-off). Data are plotted as mean ± S.E. percent time asleep per 5-min bin. Vehicle data are plotted as ±S.E. with mean line omitted. Note that oral administration of CCD-3693 dose dependently decreased sleep latency and increased sleep time relative to vehicle (shaded line).

View larger version (K): [in this window] [in a new window]   Fig. 9.   NREM sleep accumulation profiles comparing two doses each of CCD-3693, zolpidem (ZOL), and triazolam (TRZ). Data are plotted as the difference in NREM time (with respect to corresponding vehicle) relative to baseline 24 hr earlier, summed serially over a 18-hr posttreatment interval. A positive or negative slope in the plot indicates a net gain or loss in NREM, respectively. The 18-hr time interval shown comprises the second half of the rats' subjective night (lights-off; usual activity phase) and the next subjective day (lights-on; usual rest phase) and affords optimal detection of rebound insomnia, if present. Note that the surplus NREM sleep gained from either dose of CCD-3693 was not lost to rebound insomnia, as was observed for ZOL and TRZ.

Locomotor activity. All active treatments showed significant dose-related reductions of LMA (table 2). Compared to the highest doses tested of triazolam and zolpidem, pregnanolone or CCD-3693 (30 mg/kg) promoted NREM more but reduced LMA less. Because these treatments reduced both EEG-wakefulness and LMA, it is desirable to know which variable was more strongly affected by a given treatment. Toward these ends, we defined locomotor activity intensity as the number of counts of LMA per minute of EEG-wakefulness. Plotting the high-dose treatments (fig. 10) shows that the benzodiazepine ligands reduced LMA more than wake, although the neuroactive steroids did the opposite. The main effects computed as the change from baseline for 5 hr posttreatment for CCD-3693 (F[2,23] = 4.14, P < .05), for pregnanolone (F[2,33] = 6.98, P < .005) and for triazolam and zolpidem (F[7,93] = 22.60, P < .0001) and the Dunnett contrasts for all of the high-dose treatments were statistically significant. Relative to vehicle controls, figure 10 suggests that the neuroactive steroids more specifically affect sleep-wakefulness than LMA when compared to the benzodiazepine receptor ligands.

View larger version (K): [in this window] [in a new window]   Fig. 10.   LMA intensity profiles for CCD-3693, pregnanolone, zolpidem and triazolam. Plots reflect population mean response to the highest doses used in this study. LMA intensity is a measure of locomotor activity per unit time awake. Values below the shaded line (vehicle ± S.E.M.) reflect drugs that depress motor activity during spontaneous episodes of EEG-defined waking.

Body temperature. Benzodiazepine ligands reduced T_b, zolpidem being remarkably potent in this respect even at the lowest dose, whereas, of the neuroactive steroids, only the higher dose CCD-3693 reduced T_b significantly. None of these treatments reduced T_b below the normal circadian nadir, so physiologically, these effects could be concomitants of decreased motor activity.

	Discussion

Top Abstract Introduction Methods Results Discussion References

The soporific efficacy of neuroactive steroids has long been known, but only recently considered within the framework of modern drug discovery and development. Selye (1942) first demonstrated that progesterone, and the 5-pregnan-3,20-dione precursor to the 3-hydroxylated, 5-reduced pregnanes produced anesthesia in rodents (Figdor et al., 1957; Atkinson et al., 1965). Subsequent studies established that water-soluble salts of the progesterone metabolite pregnanolone, first identified in human pregnancy urine (Marker and Kamm, 1937), could produce sedation or anesthesia in mice (Figdor et al., 1957). In addition, the 3-hydroxyl group appears essential for hypnotic activity (Atkinson et al., 1965). Our study shows that an orally bioavailable synthetic derivative of pregnanolone, CCD-3693, also has potent and dose-dependent soporific efficacy.

Neuroactive steroids are believed to affect neuron excitability at the level of the neuronal membrane by allosterically enhancing the action of GABA on chloride conductance via a unique site on the GABA_A receptor complex (Lan et al., 1990). The pharmacological profile of pregnanolone and CCD-3693 are consistent with other positive modulators of GABA action (e.g., benzodiazepines and barbiturates), exhibiting anxiolytic, anticonvulsant and sedative hypnotic properties.

Early studies of neuroactive steroid efficacy relied on loss-of-righting reflex in rodentsa crude index of sedation and central nervous system depression that is inadequate for assessment of sleep parameters normally measured by electroencephalography. In contrast, this study focused on EEG sleep-stage assessments and parallel physiological and behavioral measures necessary to identify an ideal sedative hypnotic. This approach indicates the superior sensitivity of the EEG measures in the case of CCD-3693, where full hypnotic activity in the EEG assay was seen at 10 to 30 mg/kg, p.o. although loss-of-righting reflex TD₅₀ for CCD-3693 in the rat was 45.1 mg/kg, p.o. Indeed, the combination of standardized physiological and behavioral measures in our study offered useful preclinical measures of drug onset of action (based on EEG sleep), rebound wakefulness after drug-induced sleep and locomotor activity inhibition. In addition, continuous body temperature measures were obtained that can provide an early indication of potential cardiovascular side-effects (because rat body temperature is highly sensitive to change in vasomotor tone).

Timing of drug administration in the circadian cycle. An important consideration in the design of this preclinical sleep-wake assay was the timing of drug treatment. Although one could postulate that the most relevant time to administer a novel soporific agent is at the beginning of the rat's circadian rest phase (akin to bedtime drug administration in humans), we have found that, with the exception of REM sleep inhibition measures (see below), preclinical sleep-wake assessments of sedative hypnotics in nocturnal rodents are both sensitive and reliable (e.g., offer predictive use) when treatments are performed in the middle of the rat's activity phase (e.g., CT-18). There are specific features of rodent sleep that serve to validate this otherwise empirically derived study design as well. In addition to existing reports that rats are sensitive to benzodiazepines and other soporific agents at CT-18 (Edgar et al., 1991a; Seidel et al., 1995), treatment at this time of day offers the advantage of a 5-hr window posttreatment in which sleep and wake levels are fairly stable (and reproducible) under control conditions. The day-to-day variability in NREM sleep levels at the daily cusp of the circadian activity to rest transition, confounded further by the polyphasic nature of sleep-wake in rodents, and potential NREM ceiling effects during the circadian rest phase (Mistlberger et al., 1983) can undermine the sensitivity of sleep-wake assessments in rodents when treated too close to CT-0. This is a particularly important consideration when drug effects are calculated as a function of base-line measures 24 hr earlier. Finally, treatment at CT-18 facilitates the detection of rebound wakefulness during the animals rest phase as a drug's soporific effects subside. As noted in our study, soporific drug effects were not detected beyond 5 hr posttreatment for both neuroactive steroids and the benzodiazepine receptor ligands. Any immediate compensatory mechanisms responding to the drug-induced sleep, or wakefulness secondary to drug withdrawal is, by this study design, coincidentally timed with the rat's sleep phase (when such effects are most readily detected).

NREM-sleep-promoting (hypnotic) efficacy. Both benzodiazepine receptor ligands and neuroactive steroids showed rapid onset of action (5-25 min to reach peak NREM-promoting activity). Differences in the onset of action between CCD-3693, pregnanolone and benzodiazepine ligands were likely a function of route of administration (e.g., rate of absorption), and therefore precluded meaningful sleep latency comparisons between these compounds. NREM sleep measures in the first 5 hr posttreatment, however, suggest that the neuroactive steroids were intrinsically more efficacious than triazolam and zolpidem (table 2). Although sleep induced by each of the compounds tested was reversible (e.g., the animals were readily awakened by somatosensory stimuli), the benzodiazepine receptor ligands do not show prolonged periods occupied by 90%+ NREM-sleep (Edgar et al., 1991b) as was observed after neuroactive steroid treatments (see fig. 8). The latter reflects increased sleep continuity (increased sleep bout lengths) which, in humans, is an important determinant of sleep quality (Levine et al., 1987; Roehrs et al., 1994). Edgar and colleagues have proposed that the NREM-promoting efficacy of benzodiazepine receptor ligands is "gated" by the underlying physiological (homeostatic) need for sleep, whereas other classes of compounds may directly invoke NREM sleep (Edgar et al., 1991a and b, Trachsel et al., 1992). For example, ₂-adrenergic agonists such as clonidine and dexmedetomidine invoke NREM sleep to high levels at any time in the circadian cycle (Seidel et al., 1995). Given their efficacy, it seems likely that this could also be true for the neuroactive steroids in our study, although additional treatments at different circadian times are needed to confirm this. All of the active compounds in this study appeared to have a relatively short duration of NREM-promoting action which can be estimated from figure 8. The detailed time course of reduced LMA (not presented here) closely coincided with the automated EEG-scoring data in figures 7 and 8.

"Rebound" wakefulness. After all but the lowest-dose treatments, rats slept more than usual for that time of day. This drug-induced "surplus" of NREM sleep was quantified by calculating the posttreatment amount minus the usual amount for that time of day (e.g., baseline 24 hr earlier), and then viewed as a running-sum in figure 9. After the NREM-promoting effect of benzodiazepine receptor ligands had subsided, a distinct compensatory decrease in the abundance of NREM was observed, which persisted until the total accumulated minutes of NREM approached values normal for a 24-hr period. This compensatory decrease in NREM may be related to "rebound insomnia" after benzodiazepine receptor ligands that has been reported in humans (Mitler et al., 1984; Roehrs et al., 1986, 1990), and, as such, can have considerable impact on the clinical use of hypnotics. As with the human response, it remains unclear whether this compensation in rats is a function of drug withdrawal (e.g., secondary to drug-induced receptor-sensitivity changes), is driven by the normal sleep-homeostatic process or both. Remarkably, the neuroactive steroids showed no evidence of compensatory wakefulness after the initial NREM-promoting effect had subsided, and therefore allows the possibility that "rebound insomnia" may be significantly less of a problem for neuroactive steroids than has been the case for some benzodiazepine receptor ligands.

Interference with REM sleep. At the time-of-day for these treatments (CT-18), REM sleep abundance had reached its normal circadian nadir; thus, acute reductions of REM sleep due to drug effects may have been limited to some extent by a floor effect. Nonetheless, we found that the neuroactive steroids did not measurably interfere with REM sleep at this time of day, whereas the highest doses of triazolam and zolpidem resulted in statistically significant reductions of REM sleep. Although these data are consistent with the action of neuroactive steroids in cats (Heuser, 1967), a more accurate assessment of the relative REM-interfering properties of these compounds in rats may be obtained by drug administration at a time-of-day when REM sleep was more abundant (e.g., CT-5).

Behavioral and physiological specificity of actions. Relative to vehicle controls, figure 10 suggests that neuroactive steroids more specifically affected sleep-wakefulness than LMA. In contrast, triazolam and zolpidem disproportionately reduced locomotor activity during waking episodes in the first 3 hr posttreatment, suggesting the benzodiazepine receptor ligands may not be as specific for sleep induction as the neuroactive steroids. Benzodiazepine receptor ligands generally show "myorelaxant" effects. Within a clinical context, this effect may be related to (or the same as) motor impairment, which could constrain the usefulness of benzodiazepine receptor ligands as hypnotics in some segments of the general population. For example, nocturnal motor impairment and disorientation could contribute to falls and hip fractures in the geriatric population. The relative specificity for the neuroactive steroids in promoting NREM sleep rather than reducing LMA could, therefore, indicate that motor impairment may be significantly less of a problem for neuroactive steroids than for benzodiazepine receptor ligands. Consistent with the foregoing, decreased acquisition of a passive avoidance paradigm in mice only occurred at ataxic doses of pregnanolone and CCD-3693, although triazolam and zolpidem disrupted performance at doses significantly below ataxic doses.