The duration of sleep promoting efficacy by dual orexin receptor antagonists is dependent upon receptor occupancy threshold

OX-A levels are associated with wakefulness across species

To more precisely define the timing of changes in orexin levels, their relationship with arousal and therefore the available signal window for the DORAs over the course of the circadian day, OX-A levels in cerebrospinal fluid (CSF) were determined by meso-scale immunoassay and correlated with baseline active wake as measured by polysomnography (PSG) in nocturnal and diurnal preclinical species. In nocturnal mice and rats, OX-A levels progressively drop during the inactive period, reaching a nadir in the second half of this period, and accumulate over the course of the dark (active) period (Figure 1A). Mean time in active wake generally mimics these OX-A levels. In diurnal dogs and rhesus monkeys, OX-A levels build during the light (active) period and fall during the resting period, coincident with increased arousal during the day and decreased wakefulness at night (Figure 1B). These observations predict that DORAs will have maximum efficacy in the late dark phase of nocturnal species and the late light phase of diurnal animals which correlates to the active periods of both types of species.

DORA efficacy in rats is most salient during the active phase

The sleep-promoting effects of DORAs were greatest during the active phase when orexin levels were highest in normal rats entrained to a 12:12 light:dark cycle. When administered during the mid-active phase, DORA-22 (30 mg/kg), an MK-6096 analog structurally distinct from suvorexant and DORA-12, promoted sleep characteristic of a DORA: active wake was attenuated for up to 8 h, and both delta and REM sleep were increased for up to 6 h following treatment (Figure 2A). The magnitude of wake-reducing activity was diminished when normal rats were administered DORA-12 just prior to the inactive phase onset and was accompanied by small variable increases in delta and rapid eye movement (REM) sleep (Figure 2B). Importantly, changes induced by DORA-22 administered just prior to the inactive phase trend in the same direction as that seen in vehicle-treated animals progressing from the active to the inactive phase; namely, diminished active wake, increased slow wave sleep (SWS), and small fluctuations in REM sleep.

DORA-induced efficacy occurs at OX₂R occupancies exceeding 65%

A dose response for DORA-12, a structural analog of suvorexant, was evaluated in rat PSG studies to determine the relationship between sleep-promoting efficacy as measured by active wake reduction and OX₂R occupancy. Dose-dependent attenuation of active wake was observed with DORA-12 (Figure 3A), which also resulted in corresponding increases in non-REM and REM sleep. Sleep-promoting efficacy induced by each dose of DORA-12 was quantified as a percent change in active wake relative to the vehicle-induced response over 2 h following treatment.

It was previously shown that suvorexant attenuated active wake in dogs in a dose-dependent manner with corresponding increases in non-REM and REM sleep: the effects of 1 mg/kg on active wake were observed for up to 4 h, and 3 mg/kg responses for up to 7.5 h [12]. Re-examination of the time course of plasma exposures indicated that suvorexant efficacy was observed at time points when plasma concentrations were in excess of 0.342 μM (Figure 4A), a plasma exposure corresponding to 65.7% OX₂R occupancy calculated from rat occupancy values normalized for both rat and dog unbound plasma values (1.4 and 1.0%, respectively). Time points showing active wake reduction following administration of suvorexant (1 and 3 mg/kg) coincided with plasma exposures that were calculated to result in ≥65% OX₂R occupancy (Figure 4B), consistent with the range of effective occupancy observed in rats.

OX₂R binding kinetics contribute to the timing of DORA receptor occupancy

To determine the influence of OX₂R binding kinetics on receptor occupancy and sleep-promoting efficacy, the time course of in vivo effects in response to DORA-12 was compared with that of almorexant, a compound known to exhibit exceedingly slow dissociation from OX₂R. Almorexant is well documented to have slow binding kinetics relative to other DORAs [18, 23]. Depending upon the method and competing ligand used, the T_1/2 of almorexant for association with OX₂R ranges from 28 to 162 min and from 70 to 268 min for dissociation (the later values obtained in current experiments comparing 6 different DORAs in Table 2). DORA-12 (30 mg/kg) and almorexant (100 mg/kg) were administered both during the mid-active phase and just prior to the inactive phase, a therapeutically relevant dosing time, in rats. DORA-12 significantly attenuated active wake for approximately 6 h following dosing during the active phase and up to 3.5 h after treatment at the inactive phase onset (P <0.05 through 0.001; see Figure 5A, upper panels). Almorexant-dependent reductions in active wake were seen for up to 7.5 h after treatment during the active phase and for up to 11 h following inactive-phase treatment (Figure 5B, upper panels).

Compound	Kd(nM)	T1/2ON (min)	KON(mol-1 · min-1)	T1/2OFF (min)	KOFF(min-1)
Almorexant	0.048	162a	1.58 × 107	268	2.696 × 10-3
Suvorexant	0.401	80.1	7.63 × 106	89.4	7.784 × 10-3
DORA-12	0.157	57.5	2.77 × 107	232	3.154 × 10-3
MK-6096	0.338	62.9	1.69 × 107	118	6.754 × 10-3
DORA-22	0.834	24.7	9.19 × 106	37.8	1.878 × 10-2
SB-649868	0.163	37.0	2.56 × 107	57.0	1.258 × 10-2

^aAlmorexant T_1/2 ON represents an underestimate as almorexant was applied at concentrations 3-fold above its K_d in order to measure its kinetics alongside other DORAs.

K_d, binding constant; K_OFF, dissociation constant; K_ON, association constant; T_1/2 OFF, half-life off-rates; T_1/2 ON, binding half-life.

Plasma and CSF levels of DORA-12 and almorexant, as well as OX₂R occupancy were determined in satellite rats following treatment at the onset of the inactive phase. As seen in Figure 5A (lower panels), DORA-12 plasma and CSF levels began to fall immediately after dosing, with corresponding decreases in OX₂R occupancy to 77% of maximum at 8 h, coincident with the cessation of sleep-promoting effects. OX₂R occupancy by almorexant, however, remained elevated at 10 h, even though levels within the CSF were undetectable at this time point (Figure 5B, lower panels). The persistence of OX₂R occupancy by almorexant despite disappearing levels of the compound in the plasma and CSF is consistent with its extremely slow receptor-binding kinetics relative to other DORAs including suvorexant, MK-6096, DORA-22 and SB-649868 (Table 2). Together these results indicate that binding kinetics have the capacity to influence receptor occupancy and sleep promoting efficacy, particularly for compounds with excessively slow kinetic parameters.

Next-day sleep, qEEG, and quantitative effects in rhesus monkeys

To evaluate the potential for both GABA-A receptor modulators and DORA-22 to promote sleep through the resting phase and into the subsequent active phase, rhesus monkeys were administered these compounds 2 h before lights out. The doses, diazepam at 10 mg/kg, eszopiclone at 10 mg/kg, and DORA-22 at 30 mg/kg, were selected based on their similar potential to promote sleep following active-phase dosing. Quantified over 2 h, these treatments have previously been found to attenuate active wake by 38.4, 55.0, and 40.1 min relative to vehicle, respectively [24].

Diazepam, eszopiclone, and DORA-22 produced characteristic plasma-exposure profiles (Figure 6A). The anxiolytic diazepam exhibited a peak at 1.0 h (775 nM), eszopiclone induced a more rapid C_max at 0.5 h (1680 nM), and DORA-22 a more delayed exposure profile (134 nM at 2 h). All three compounds significantly attenuated active wake in the 2 h after dosing before the onset of the inactive phase (Figure 6B). Unlike eszopiclone and DORA-22, diazepam significantly attenu-ated active wake at time points 7.5 h into the inactive phase and 3 h into the subsequent active phase. Despite the presence of measurable eszopiclone and DORA-22 plasma levels 17 h after treatment, no consistent attenuation of active wake was observed in response to either compound the morning after treatment.

Quantitative EEG analysis performed coincident with PSG revealed time-dependent and qualitative differences between the effects of these compounds that were not evident in measurements of active wake. All three compounds increased delta frequency qEEG power in advance of the normal increase in delta power observed during the resting phase after administration of vehicle (Figure 6C, bottom panels). However, both diazepam and eszopiclone increased high-frequency gamma power and substantially decreased theta power throughout the inactive period, while DORA-22 had essentially no effect on these measures (Figure 6C, top and middle panels). The effects of diazepam persisted into the subsequent active period while eszopiclone-induced qEEG changes ceased abruptly upon lights on (at which point cognitive tests were performed). The relevance of compound-induced qEEG changes occurring during cognitive testing intervals is currently unknown.

In order to better understand the consequences of compound-induced sleep and qEEG effects that persist though the inactive period, cognitive memory and attention parameters were evaluated in the same rhesus monkeys, the morning after treatment. Memory was evaluated in the delayed match to sample (DMS) task performed at lights on (14 h after dosing), while attention was evaluated in the serial choice retention time (SCRT) test administered 2 h later (16 h after dosing; see Figure 6C, bottom panels, D and C time points). Significant deficits in recognition memory, as assessed by DMS, were observed the morning after treatment with diazepam (10 mg/kg) but not following eszopiclone (10 mg/kg) or DORA-22 (30 mg/kg) relative to their respective vehicle treatments (Figure 6D). These results are consistent with diazepam-dependent active wake attenuation persisting into the active period. Despite the presence of measurable levels of this DORA-22 three hours after the DMS task was performed (Figure 6A; 61 nM), no attenuation of in performance was observed. In SCRT, both diazepam and eszopiclone elicited significant deficits in the percentage of correct responses to cues of the shortest duration relative to vehicle-treated controls (P < 0.05 for both) (Figure 6E). In contrast, deficits in next-day attention were not observed with DORA-22.

Suvorexant exposure time course in humans

Plasma samples obtained in an Phase 1 clinical evaluation of suvorexant exhibited a common exposure profile, with a time to maximum plasma concentration (T_max) of 3 h or less (Figure 7). Human free plasma levels of suvorexant sufficient for 65% OX₂R occupancy were calculated based on extensive transgenic rat occupancy data and normalized for free fraction (rat: 1.4%; human: 1.0%). The human plasma concentration of suvorexant calculated to correspond to 65% OX₂R occupancy is 0.33 μM (free + bound). Observed clinical suvorexant mean plasma concentrations fell below the levels predicted to correspond to 65% OX₂R occupancy, where sleep-promoting efficacy is predicted, by the 8 h post-dose time point for both the 10 and 20 mg therapeutic doses. Together this analysis indicates that the sleep promoting effects of suvorexant would not be expected to persist into subsequent waking hours at therapeutic doses less than 20 mg.

Detection of orexin-A in cerebrospinal fluid

CSF was collected via cisterna magna puncture similarly from male and female beagles and rhesus monkeys (Macaca mulatta) similarly. Propofol (i.v., 2–8 mg/kg, to effect ) anesthetized dogs were maintained on isofluorane anesthesia via an endotracheal tube. Using aseptic technique, a 1.5 inch spinal needle with stylet is introduced through the skin and subcutaneous tissue and then through the atlanto-occipital membrane which connects the first vertebrae with the skull and overlies the cisterna magna. The stylet is then removed from the needle and cerebrospinal fluid is allowed to flow into a collection tube. CSF was collected from conscious male and female rhesus monkeys (Macaca mulatta) using methods described previously [41], and similar to those procedures employed in dog.

CSF was collected from Male Sprague–Dawley rats (Taconic Farms, Hudson, NY) under isofluorane anesthesia (5% with oxygen at 2 L/min in an induction hood, 2% thereafter) and on a circulating water warming blanket. Heads were shaved from the posterior portion to the dorsal thoracic area exposing region just above the cisterna magna, and placed on stereotaxic apparatus to position the head of the rat at a 70-80° bend. After identifying the area by palpation, a 25 g × 5/8” butterfly needle, with 12” of tubing (Terumo Surflo Winged Infusion Set, Terumo Corporation, Somerset, NJ) is slowly inserted into the cisterna magna of the rat. A small suction is applied from a syringe connected to the end of the winged infusion set until fluid is observed in the tubing. CSF is then collected by expulsion into a microfuge tube. CSF was similarly collected from male C57BL/6 mice (Taconic Farms, Hudson, NY) except that the procedure is a non-survival surgery in mice, and animals are rapidly euthanized following CSF recovery. For mice, a stereotaxic device is fashioned from the styrofoam 15 mL conical tube holder, and skin and muscle covering the head and neck area are removed to visualize the dura mater prior to CSF collection.

Concentrations of OX-A within CSF were quantified by immunoassay developed using the Meso Scale Discovery (MSD) electrochemiluminescence detection technology platform (Gaithersburg, MD, USA). Purified polyclonal immunoglobulin G (G-003-36, Phoenix Peptide, Burlingame, CA, USA), raised in rabbit against OX-A (amino acids 16–33) amide, was used as the capture antibody. Goat anti-human polyclonal antibody against orexin-A (N-18) (sc-26491, Santa Cruz Biotechnology Inc., Santa Cruz, CA, USA) was used as the primary detection antibody. Sulfo-Tag-labeled anti-goat antibody, raised in donkey, was used as the secondary detection antibody and was provided by MSD (R32AG-5). Assay wash buffer (R61TX-1) was also sourced from MSD. Blocking buffer was generated by dissolving membrane blocking agent (RPN-2125, GE Healthcare, Buckinghamshire, UK) in MSD wash buffer (1.5 g per 50 mL). Antibody and orexin standard diluent was made by diluting blocking buffer three-fold in wash buffer. Standard OX-A (full length peptide) was sourced from Phoenix Peptide (003–30, Burlingame, CA, USA).

Plates were incubated overnight with 25 μL of capture antibody (3.33 μg/mL) and washed three times using MSD wash buffer (200 μL per well). Plates were then incubated with blocking buffer (150 μL per well) on a Titer Plate Shaker (Barnstead International, Dubuque, IA, USA) for 1 h at room temperature. After aspirating blocking buffer, orexin standards (reconstituted in antibody diluent) or samples (25 μL) were added to each well. The plates were sealed and incubated, shaking for 2 h at room temperature. Following incubation with standards and samples, plates were washed an additional three times with 200 μL per well of wash buffer. Next, 25 μL of primary antibody dilution (1 μg per mL) was added to each well and plates were sealed and incubated while shaking for 1 h at room temperature. Following an additional wash step, 25 μL of secondary antibody dilution (0.5 μg per mL) was added to each well, and plates were sealed and incubated, shaking, for 1 h at room temperature. As a final step, 150 μL of read buffer was added to each well and plates were read on an MSD sector imager. All measurements were performed in triplicate.

EEG

The mean time spent in sleep stages was determined in radio-telemetry-implanted mice, rats, dogs, and rhesus monkeys as first described by Renger and colleagues [42] and subsequently by Winrow and colleagues [12, 18] with minor modifications. For baseline studies, recordings were taken continuously from unperturbed mice, rats, dogs, and monkeys for 6, 6, 10, or 5 days, respectively, and the mean time spent in all sleep stages in 30-min intervals were averaged and plotted onto a single 24-h period (Figure 1). Almorexant is a DORA [43] that was previously in clinical development for the treatment of insomnia [10, 44]. The clinical development program for almorexant was terminated in early 2011. For rat studies in which DORA-22, DORA-12, or almorexant were administered, a balanced 3-day cross-over design was employed in which all animals were treated alternately with vehicle (20% Vitamin E d-alpha tocopheryl polyethylene glycol 1000 succinate [TPGS], orally [per os, p.o.]) and compound for 3 consecutive days with a 5-day intervening washout period, animals being divided randomly into either vehicle- or compound-first groups. In all cases, mean time in active wake, quiet wake, SWS, and REM sleep was determined for each 30-min interval for 3 days of each condition and averaged onto a single 24-h period. Automated sleep-stage data collection and analysis were performed as described in detail previously [12]. Statistical differences in mean time spent in active wake, quiet wake, SWS, and REM sleep between vehicle- and compound-treated groups at 30-min time points were determined by a linear mixed-effects model applied t-test.

EEG responses to diazepam (10 mg/kg, p.o. in 0.5% methylcellulose); eszopiclone (10 mg/kg, p.o. in 20% Vitamin E TPGS) and DORA 22 (30 mg/kg, p.o. in 20% Vitamin E TPGS) were evaluated in rhesus monkeys implanted with subcutaneous telemetric devices (D70-EEE; Data Sciences International, Arden Hills MN) housed in a 12:12 light cycle with lights-off at 16:30 and lights-on at 04:30. Animals were fed on a calorie-controlled diet of laboratory chow supplemented with fruit and vegetables. Water was available ad libitum with the exception of those monkeys trained to perform cognitive tasks, for which access was restricted for up to 4 h prior to and during cognitive testing. Treatment occurred at 14:40 (Zeitgeber time [ZT] 10:10) in a balanced 1-day cross-over design: 1 day of drug or vehicle followed by 6 days of baseline followed by 1 day of vehicle or drug. Cognition tasks were scheduled for next-day evaluation as described below. Sleep/wake architecture was scored according to vigilance state (active wake, quiet wake, delta sleep I, delta sleep II, and REM sleep) at 30-minute intervals and analyzed as previously described [24]. Spectral analysis of continuous EEG was quantified between vehicle and compound after binning continuous frequencies into canonical frequency bands (reported herein: Delta, 0.5 – 4 Hz; Theta, 4.0 – 7 Hz; Gamma, 35.0 – 100.0 Hz). qEEG values are spectral power (uV2) log transformed before analysis and averaged over 30-minute intervals, expressed as means +/− SEM, and compared with vehicle using a mixed-model ANOVA at each time point with random effects for subject and date within subject in the R statistical computing environment (cran.us.r-project.org; the R Foundation for Statistical Computing, Vienna, Austria) using a linear mixed effects model for significance testing. Significant conditional differences are indicated as tick marks at the top of the graphs (short, P ≤ 0.05; medium, P ≤ 0.01; long, P ≤ 0.001) with gray vertical bars through significantly different data points.

OX₂R occupancy

Occupancy of human OX₂R in transgenic rats expressing the receptor via the rat enolase promoter was performed ex vivo in radioligand-binding-displacement assays essentially as previously described [21] with minor modifications. To determine the relationship between OX₂R occupancy for a given compound over a full range of plasma and CSF concentrations, hOX₂R-expressing rats were dosed i.v. with compound. Immediately following 30 min of i.v. administration, animals were anesthetized, sacrificed, and CSF, blood, and brain tissue were harvested and frozen for analysis. Frozen tissue was weighed and thawed coincident with homogenization. The homogenate was pelleted at 21,000 × g for 1 min and the pellet was resuspended. A binding assay was run with brain tissue and radio ligand at ~ 10 × K_d and filtered at 12 min (in the linear association phase). Non-specific binding was determined by the addition of 1000-fold excess unlabeled ligand. Experimental (dosed)-specific binding was compared with control (vehicle-dosed) binding ([specific total counts minus specific experimental counts]/specific total counts × 100 = percent occupancy). Raw occupancy values determined in this manner over a range of plasma and CSF values were then used to construct a normalized curve where the theoretical maximum occupancy value was defined as 100% and occupancies relative to plasma or CSF values were fit to a one-sided binding (hyperbola) curve (GraphPad Prism, San Diego, CA, USA). Plasma values determined from PSG experiments in satellite animals were then used to determine corresponding occupancy values from best fit curves for a given compound at C_max. In separate experiments, plasma, CSF, and brain tissue was collected at time points following dosing to directly measure compound levels and corresponding OX₂R occupancies at these times. Resulting occupancy relative to measured plasma levels of compound was fit to a normalized binding curve as described previously [21].

OX₂R occupancy in dogs and humans was calculated based on the plasma/occupancy relation determined in rats and corrected for the free fraction of suvorexant in each species. In dogs, suvorexant plasma levels determined in satellite animals coincident with PSG studies reported by Winrow and colleagues [12] were used to calculate OX₂R occupancy levels based on standard curves of OX₂R occupancy in rats versus the free fraction of plasma suvorexant levels. Human free plasma levels of suvorexant sufficient for 80% OX₂R occupancy were calculated based on extensive rat data and normalized for free fraction (rat: 1.4%; human: 1%). The plasma level of suvorexant calculated to correspond to 80% OX₂R occupancy in humans is 0.71 μM (free + bound) [17]. Human plasma levels of suvorexant collected in Phase 1 (10, 20, 50, and 76 mg doses) and Phase 2/3 (steady state, 40 mg) clinical trials were plotted against the suvorexant plasma value calculated to correspond to the exposure required for the OX₂R occupancy determined from standard OX₂R occupancy versus free fraction exposure curves of preclinical species.

Monkey Delayed Match to Sample (DMS) task

All cognition testing occurred in the animal’s home cage equipped with a touchscreen. Each DMS trial was initiated with a single ‘sample’ image (150 × 200 pixels: 1.8 inches × 2.3 inches) in one of eight colors presented at the center of a touchscreen. Evaluation began when the monkey touched the sample image, at which point the screen became blank for a retention period. Following the retention period, four choice images (150 × 200 pixels) were presented, one in each corner of the screen. One of the four choice images matched the color of the sample image, whereas the three remaining ‘distractor’ images were drawn from the pool of remaining colors. A reinforcer was provided when the monkey touched the choice image that matched the color of the sample image. Incorrect choices were not reinforced and resulted in a 5 s timeout. On completion of the trial, an inter-trial interval of 5 s was presented prior to the next trial, and failure to respond to the samples or choice images within 30 s resulted in the screen turning blank for a 5 s timeout prior to initiating another trial. Test sessions comprised 96 trials or lasted for up to 45 min. Three discrete retention intervals were titrated for each subject’s baseline performance in the absence of compound to yield performances of approximately 80–100%, 55–65%, and 35–45% of responses being correct at short, medium, and long retention intervals, and ranged from 0.25–0.5 s, 2.5–14 s, and 9–39 s, respectively. Response latencies to sample and choice images were also recorded. Values reported reflect the proportion of correct responses collapsed over all retention intervals. Differences from vehicle were examined using repeated measures ANOVA.

Monkey Serial Choice Reaction Time (SCRT) task

Each SCRT trial was initiated with the appearance of 10 blue square ‘target’ images evenly distributed at 3-inch intervals around the perimeter of a touchscreen presented together with a centrally located circular ‘centering’ image (1.4 inch diameter circle). The task began with the monkey touching the centering image, which then turned gray. The monkey was then required to touch the centering image continuously throughout a variable pre-cue interval of 0.25-7.5 s. On completion of the pre-cue interval, one of the blue target images was cued by turning red for one of four cue durations before turning back to blue. To obtain a reward, the monkey was required to touch the target image that had been ‘cued’ red within a 5 s limited hold period. Inaccurate choices were not reinforced and resulted in a 3 s timeout. On completion of the trial, an inter-trial interval of 2 s was initiated prior to the next trial. A failure to respond to the centering image within 60 s resulted in the screen becoming blank for a 2 s timeout prior to the start of the next trial. If the monkey initiated a trial but removed its hand from the touchscreen prior to the appearance of the cue, the trial was terminated and the screen was blanked during a 2 s inter-trial interval. Sessions were terminated after 20 min. Cue location and duration were pseudo-randomly assigned. Cue duration and the target size were titrated for each subject on the basis of performance during previous baseline sessions to yield a performance of 60–80% correct responses for the trials of the shortest cue duration. The briefest cue duration varied from 0.04–0.1 s and the cues varied in size from 0.2–0.7 inches, although each cue had the same touch-sensitive area of 1.75 inches. Values reported in the current study reflect the proportion of correct responses to targets after cues of the shortest duration only.