Drug substance

BI001 (see WO2017178339) was synthesized by Boehringer Ingelheim. A single batch of free base material was used throughout the studies. Purity was determined by NMR to be > 95%. BI001 shows reasonable physicochemical properties and aqueous solubility. In vitro, the compound shows good permeability and no P-glycoprotein-related efflux in the MDCK-MDR1 permeability assay (Madin Darby canine kidney (MDCK) cells overexpressing human Pglycoprotein), and no significant off-target stimulatory or inhibitory effects at 10µM in a standard off-target assay (Eurofins; Bowes et al. 2012). Plasma protein binding and brain tissue binding were 86.6% (fu, plasma 0.134) and 85.3% (fu, brain 0.148) as determined by Rapid Equilibrium Dialysis (RED device, Syngene International Ltd., India). Plasma protein binding was determined by equilibrium dialysis to be 87.4% (fu, 0.126). BI001 is stable in rat plasma (t1/2 > 130 min).

Animals

All efforts were made to minimise any animal distress and/or suffering.

Study 1: Receptor occupancy (RO%) and pharmacokinetics (PK)

Male C57BL/6NCrl mice in the weight range of 20–25 g were used for the RO% and PK studies performed at Boehringer Ingelheim GmbH (Biberach an der Riss, Germany). Mice were housed in standard wire-topped cages with ad libitum chow and water, wood shavings and nesting materials. The experiments were approved by the local German authorities (Regierungspräsidium Tübingen) and conducted in compliance with the German and European Animal Welfare Acts.

Study 2: Go/No-Go task

Male C57BL/6JArc mice were sourced from the Animal Resources Centre (Perth, WA). Mice were 1.7 months of age at the start of experimentation and group-housed of up to 3 animals per cage under a 12:12 light-dark cycle, with lights off at 3pm. Mice were housed in standard wire-topped cages with ad libitum water, wood shavings and tissue paper nesting materials.

Experiments were performed in accordance with the Prevention of Cruelty to Animal Act (2004), the guidelines of the National Health and Medical Research Council Code of Practice for the Care and Use of Animals for Experimental Purposes in Australia (2013) and approved by the Florey Animal Ethics Committee (AEC number: 19 − 010).

Formulation and dosing

BI001 was administered orally by gavage as a suspension. The vehicle was aqueous 0.5% Natrosol (Ashland Inc., Richmond, NSW, Australia) solution and 0.015% Tween80. Administration volume was 10 mL/kg. The BI001 suspension was prepared by either a rotor-stator homogenizer (Go/No-Go task) or a high energy focused Ultrasonicator (Covaris®, MA, US; RO% and PK).

For behavioural experiments, mice were acclimatised to oral gavage with water for one week prior to compound administration in the go/no-go task. For behavioural testing, BI001 (12.5 mg/kg) or Tween/Natrosol vehicle was administered 30 min prior to operant testing. This dose was selected based on the findings of Tyree and colleagues (2023), who showed that BI001 dose-dependently (2.5, 7.5 and 12.5 mg/kg) blocked the disinhibiting effects of optogenetic hypocretin neuron stimulation in a mouse go-no go task. For RO%, doses of 0.3, 1, 3 and 10 mg/kg were used, and for the PK study a dose of 4.4 mg/kg. All doses were administered orally.

Blood and tissue sampling

In Study 1, trunk blood samples were collected in the RO% study directly after decapitation. For PK, serial sampling of 20 µL of blood was performed by puncture of the saphenous vein in awake animals. In both studies, blood was directly sampled into K3-EDTA coated vials. BI001 plasma and brain concentrations were also determined in a group of satellite animals to those in Study 2 after a single dose of 12.5 mg/kg of BI001, 15 or 75 min after administration (N = 3 per time-point). Mice were dosed as described in Sec 2.3. Mice were culled by lethal pentobarbitone overdose, at which point cardiac blood samples were collected, and brains harvested and snap frozen in isopentane. Immediately after collection of blood, plasma samples were prepared by centrifugation at 4 °C and were stored at − 20 °C (Study 1) or -80 °C (Study 2) until bioanalysis.

Sample preparation and bioanalysis

Plasma protein was precipitated with acetonitrile. Brain samples were transferred to 7 mL Precellys® tubes and 4 parts of acetonitrile/methanol (1:1) solution were added. Samples were homogenized using a Precellys® homogenizer (Bertin technologies, France). After centrifugation, supernatants were stored at -20 °C until bioanalysis. Compound concentrations were determined by high performance liquid chromatography coupled with tandem mass spectrometry.

In vitro pharmacology

The in vitro potency of BI001 at human HcrtR1 and HcrtR2 was assessed using an IP1 assay, as described in WO2017178339.

Receptor occupancy

To determine receptor occupancy, mice were dosed with BI001 at 0.3, 1, 3 and 10 mg/kg or vehicle and comparisons made versus vehicle-treated mice. At 60 min post oral dosing, BI001-treated mice were sacrificed, and brains harvested and stored at -80 °C until processing for the assessment of receptor occupancy. Coronal brain sections containing the locus coeruleus (20 μm) were prepared using a cryostat, mounted onto glass slides and stored at -80 °C. On the day of the experiment, brain sections were thawed at room temperature and incubated in 10 nM of the HcrtR1 antagonist [3H]-EX5135 (Tritec, lot. nr.15-0112-0113, molecular weight, 443.5 g/mol, specific activity 55.6 Ci/mmol, WO13068935 Example 92) in a volume of 300–400 µl. Non-specific binding was defined in the presence of 10 µM cold EX5135). After a 30-min incubation at room temperature, followed by three washing steps (Tris-NaCl buffer) and a final water wash, air dried slides were incubated with film with for 4–7 days before being analysed using Aida Image Analyzer Software.

Go/No-Go experimental proceduresFood intake and body weight

N = 24 mice were fed with standard chow ad libitum for seven consecutive days, at the end of which body weight was recorded as their original weight. After the seven days, food for each cage was then restricted so that the weights of mice were maintained at 90-95% of their original weight. Mice were habituated to handling for ten minutes per day prior to operant task training. Food restriction was continued during operant task training.

Apparatus

All operant task training and testing was conducted with 8 operant boxes (Med Associates Inc., VT, USA) in sound-attenuating chambers. Operant boxes were equipped with a grid floor, a house light (30 lx), 2 retractable levers on either side of a liquid delivery reward port, a cue light above each lever and tone generators that generated clicking sounds, variable pure tones or white noise.

Operant conditioning training protocol

After assessment of food intake under group-housed, basal conditions, mice entered operant conditioning training stages. 20 µL units of Strawberry milk (Iced Strawberry Milk, Nippy’s Ltd, Moorook, SA, Australia) was used as the reward throughout the study. Mice were first habituated for 3 days to the operant chambers for 30 min with strawberry milk in the reward port, where no levers or cues were presented. Mice were then trained to lever-press for a strawberry milk reward on the lever furthest from the door (the “back” lever) of the operant chamber. Mice were placed in boxes with the back lever extended during each trial for 1 h or 100 presses, whichever came first. Every time the lever was extended, a single lever press resulted in reward delivery, clicking reward sound and lever retraction for 5 s. Mice remained in this stage until they completed two sessions with > 60 responses made.

Following this, mice were trained to press the lever in response to the “go” signal used in the go/no-go task. Mice were placed in boxes for 40 min or the completion of 60 go trials, whichever came first. Every time the lever was extended, there was a “variable length non-responding phase” of 9 to 24 s, where lever presses did not trigger any response. This was followed by a 3 s “premature phase”, during which any lever press would immediately terminate the trial and lever was retracted for 10 s. After the premature phase, a 10 kHz, 75 dB pure tone go cue was played for 30 s to indicate a go trial. A lever press in the go trial phase resulted in reward delivery, clicking reward sound and lever retraction for 10 s. If no press was made during the go signal, the lever was retracted after 30 s and the 10 s intertrial interval would commence with no delivery of reward or clicker presentation. Mice remained in this training stage until they completed two sessions with greater than 45 correct go trials.

When mice met criteria with go-trial training they entered the final go/no-go task training. Mice were placed in boxes for 40 min or until completion of 60 trials (30 go trials and 30 no-go trials, presented randomly), whichever came first. Every time the lever extended, there was a “non-responding phase” and a “premature phase”, as described above for the go trial training. After the premature phase, either the pure tone identifying a go trial or a 75 dB white noise pulsed at 1 Hz that identified a no-go trial, was played. Lever presses in the go trial phase resulted in reward delivery and clicking reward sound, whereas lever presses in the no-go trial phase resulted in trial termination. The absence of level pressing during the entire 30 s no-go trial phase resulted in a reward delivery and clicking reward sound at the end of the no-go trial. Mice remained in this stage until they completed the criterion of two sessions with greater than 25 correct go trials and 10 correct no-go trials. This same task was then used for testing sessions with food restriction and HcrtR1 antagonist interventions.

Experimental design

A Latin square cross-over design was used with all mice undergoing all four treatment conditions: free-feeding vehicle, free-feeding HcrtR1 antagonist, food-restricted vehicle, and food-restricted HcrtR1 antagonist (Fig. 1). Mice were allocated to one of 8 groups, with each receiving a different sequence of treatments to minimise potential influences of treatment sequence. Testing in each treatment consisted of 5 consecutive days of go/no-go testing, with a three-day wash-out period between drug/vehicle transitions and seven-day transition between free-feeding/food restricted states. During washout periods, mice continued to perform daily go/no-go sessions.

Fig. 1

Experimental design. Mice were first allocated to 1 of 8 different treatment sequence groups. All mice then underwent all 4 treatment conditions. Free feeding (FF) and food-restricted (FR) animals underwent an initial 5-day block of drug (BI001) or vehicle (Veh) administration, followed by a 3-day drug washout period, then a 5-day block of testing with the alternate drug or Veh treatment. Next, they underwent a 7-day transition period between feeding states, followed by two additional 5-day blocks of GNG testing with BI001 or vehicle, with another 3-day drug washout period in between

To analyse performance in the go/no-go task, we measured go trial responding, no-go trial responding, non-responding presses (presses during the non-responding phase), premature presses (presses during the premature phase) and response time.

Fig. 2

Dose-normalized plasma PK profile (closed circles, N = 3, ±SD) and exposure data from Study 1 RO% (open circle 0.3 mg/kg; open downward triangle 1 mg/kg; open diamond 3 mg/kg; open triangle 10 mg/kg) Study 2 (satellite) Go/NoGo animals (closed diamond). Dotted lines show 2-fold lower and higher mean PK data and illustrate that dose-normalized RO% and Go/NoGo satellite animal exposure data were, with a single exception, within a 2-fold range from the PK study mean

Fig. 3

HcrtR1 receptor occupancy in the locus coeruleus measured 1 h after oral administration of BI001 at doses 0.3, 1, 3 and 10 mg/kg (solid circles. N = 3, ±SD). Solid line; non-linear regression with variable slope of RO data, IC50 94 nM. The dotted line shows calculated Go/NoGo mean RO (66%) from 0.25 to 1.25 h post dose of 12.5 mg/kg (N = 3, ±SD)

Statistical analyses

Prior to further analyses, averaged go trial accuracy, no-go trial accuracy, premature presses, and non-responding presses from the final 10 days of no-go trial training were calculated for each mouse. Each parameter was then scaled to a standard normal distribution and a Euclidean distance matrix between animals constructed across these parameters was used to generate a hierarchical clustering dendrogram. This clustered mice into 2 major similarity clusters, one with less impulsive mice, the other with more impulsive mice (one animal was clustered by itself as distinct from either of the other clusters; Fig. 4).

Differences between cluster go accuracy, no-go accuracy, and non-responding presses were compared with unpaired t-tests. This clustering categorisation was used as a factor in subsequent statistical models to analyse go/no-go data.

Go accuracy (go correct /(go correct + go incorrect)*100), no-go accuracy (no-go correct /(no-go correct + no-go incorrect)*100), and pretrial responding ((premature + non-responding)/total *100) were analysed using logistic mixed effect regression models with feeding status (free-feeding (FF) vs. food-restricted (FR)), treatment (vehicle vs. BI001), cluster (less impulsive vs. more impulsive) and all interaction terms as fixed effects and subject random effects on the intercept.

The time taken to correctly respond by pressing a lever on a go trial (go correct response time), time to fail at withholding from pressing on a no-go trial (no-go incorrect response time) and signal detection theory parameters for sensitivity (d’) and response bias (c), calculated as d’ = Z(hit rate) – Z(false alarm rate) and. c = – 0.5 x [Z(hit rate) + Z(false alarm rate)] where Z is the inverse of the cumulative distribution function of the standard normal distribution, hit rate is the proportion of go trials at which a response was correctly made, and false alarm rate is the proportion of responses incorrectly made during no-go trials (McVay and Kane 2009) were analysed using linear mixed-effect regression models with feeding status (free-feeding (FF) vs. food-restricted (FR), treatment (vehicle vs. BI001), cluster (less impulsive vs. more impulsive), treatment sequence and all interaction terms as fixed effects and subject random effects on the intercept.

The most complex go/no go performance, response time, and signal detection theory parameter statistical models and all simpler sub-models were compared and the final model selection for analyses were made based on the Akaike Information Criteria (AIC), a metric balancing goodness of fit with model complexity (Portet 2020). AIC was calculated for all models and simpler sub-models, and the model with the lowest AIC selected as the best fitting model for these data. Only analyses from the final AIC-selected models are reported. Estimated marginal means were calculated for each combination of factors and post-hoc comparisons performed using z-tests with Tukey’s correction for multiple comparisons (go/no-go performance) or t-tests with Kenward-Roger degrees of freedom and Tukey’s correction for multiple comparisons (response times).

Linear and logistic regression analyses were performed using R version 4.2.2 with Rstudio (Allaire 2012) using the tidyverse, (Wickham 2016) lme4 (Bates et al. 2015), MuMIn, (Barton 2023) and emmeans (Russell et al. 2023) packages and decision tree dendrograms generated with Python 3.7.4 (Van Rossum and Drake 2012) using the scikit-learn (Pedregosa et al. 2011) and the SciPy libraries (Virtanen et al. 2020).