Identification and pharmacological profile of SPP1, a potent, functionally selective and brain penetrant agonist at muscarinic M₁ receptors

Abstract

Background and Purpose

We aimed to identify and develop novel, selective muscarinic M₁ receptor agonists as potential therapeutic agents for the symptomatic treatment of Alzheimer's disease.

Experimental Approach

We developed and utilized a novel M₁ receptor occupancy assay to drive a structure activity relationship in a relevant brain region while simultaneously tracking drug levels in plasma and brain to optimize for central penetration. Functional activity was tracked in relevant native in vitro assays allowing translational (rat–human) benchmarking of structure–activity relationship molecules to clinical comparators.

Key Results

Using this paradigm, we identified a series of M₁ receptor selective molecules displaying desirable in vitro and in vivo properties and optimized key features, such as central penetration while maintaining selectivity and a partial agonist profile. From these compounds, we selected spiropiperidine 1 (SPP1). In vitro, SPP1 is a potent, partial agonist of cortical and hippocampal M₁ receptors with activity conserved across species. SPP1 displays high functional selectivity for M₁ receptors over native M₂ and M₃ receptor anti‐targets and over a panel of other targets. Assessment of central target engagement by receptor occupancy reveals SPP1 significantly and dose‐dependently occupies rodent cortical M₁ receptors.

Conclusions and Implications

We report the discovery of SPP1, a novel, functionally selective, brain penetrant partial orthosteric agonist at M₁ receptors, identified by a novel receptor occupancy assay. SPP1 is amenable to in vitro and in vivo study and provides a valuable research tool to further probe the role of M₁ receptors in physiology and disease.

Abbreviations

AD

Alzheimer's disease

aCSF

artificial CSF

BGG

bovine gamma gobulin

KP,uu

unbound plasma concentration ratio

PAM

positive allosteric modulator

PEI

polyethyleneimine

SAR

structure–activity relationship

SPP

spiropiperidine

Introduction

The hallmarks of Alzheimer's disease (AD) include amyloid plaques, neurofibrillary tangles and memory loss. There are no treatments currently available to prevent disease progression, although symptomatic treatments are available to aid cognitive function. The most broadly utilized symptomatic treatments are the AChE inhibitors, which include donepezil and rivastigmine. These inhibitors confer a modest improvement on cognitive symptoms (NICE guidance, 2011) but are associated with undesired adverse effects (e.g. gastrointestinal side effects), which are dose‐dependent (Lockhart et al., 2009; FDA Website, 2012).

AD is associated with a loss of cholinergic input to forebrain regions and reduced cholinergic transmission (Douchamps and Mathis, 2017). As AChE inhibitors block the enzyme responsible for the breakdown of ACh, they act to ameliorate the ACh deficit, thereby providing cognitive benefit. The extent of AChE inhibition correlates with cognitive improvements (Rogers and Friedhoff, 1996; Bohnen et al., 2005), although in vivo PET studies performed in subjects with AD report only modest inhibition (22–27%) of cortical AChE at clinically used doses of donepezil (Kuhl et al., 2000; Bohnen et al., 2005), suggesting it should be possible to boost central cholinergic transmission and potentially therefore cognitive benefits further. While a higher dose donepezil (23 mg) has been marketed, it is associated with significant adverse effects.

ACh mediates its effects through activation of two classes of ACh receptors, muscarinic ACh receptors (M receptors) and nicotinic receptors. Clinical data demonstrate that M receptors are important targets mediating the pro‐cognitive effects of ACh released within hippocampal and cortical circuits. Firstly, non‐selective M receptor antagonists, such as scopolamine, promote cognitive deficits in man (Rasmusson and Dudar, 1979; Robbins et al., 1997; Potter et al., 2000), while a direct acting non‐selective M receptor agonist, xanomeline, provided some evidence of cognitive benefit in patients with AD and schizophrenia (Bodick et al., 1997; Shekhar et al., 2008). At the efficacious dose, xanomeline was not well tolerated in AD patients; in common with high doses of AChE inhibitors, gastrointestinal tolerability issues were a major issue (Bodick et al., 1997).

Of the five isoforms of M receptors (M_1–5), M₁ receptors are the most abundant in regions relevant for cognition (Oki et al., 2005; Scarr et al., 2016) and are strongly implicated in mediating, at least in part, the pro‐cognitive effects of ACh. The use of selective M₁ receptor agonists and positive allosteric modulators (PAMs) and M₁ receptor knockout (KO) mice revealed that these receptors modulate neuronal signalling and excitability, synaptic transmission and network function and exerts preclinical pro‐cognitive effects (e.g. Shirey et al., 2009; Uslaner et al., 2013; Lange et al., 2015; Puri et al., 2015; Vardigan et al., 2015; Butcher et al., 2016; Dennis et al., 2016; Betterton et al., 2017). A high level of expression of M₁ receptors in cortex and hippocampus is conserved across species, including in brains from human AD patients (Overk et al., 2010; Bradley et al., 2017).

As a consequence of their presynaptic and postsynaptic localization, pan activation of opposing muscarinic and nicotinic receptors by AChE inhibitors evokes a ‘net effect’, which may self‐limit pro‐cognitive potential of the cholinergic system. The promise of selectively targeting key pro‐cognitive cholinergic receptors, such as the M₁ receptor, is improved efficacy and reduced adverse effects. M₂ and M₃ receptors are the major subtypes expressed in peripheral tissues and are proposed to underlie the undesirable gastrointestinal and cardiovascular side effects of non‐selective agonists such xanomeline. However, M₁ receptors are involved along with M₃ receptors in mediating sweating and salivation (Felder et al., 2018). Thus, there is a clear need for M₁ receptor agonists that are highly brain penetrant with a partial agonist efficacy profile to minimize peripheral load and M₁ receptor‐mediated adverse effects.

Recent efforts to develop selective M₁ receptor ligands have focused on developing PAMs, allosteric agonists or orthosteric M₁ receptor agonists (Felder et al., 2018). Allosteric and orthosteric agonists have in common the ability to directly activate M₁ receptors , albeit through interaction at distinct sites. In contrast, PAMs of M₁ receptors act primarily to potentiate the action of the endogenous orthosteric agonist, ACh, at concentrations that do not activate M₁ receptors in the absence of agonist. Several of these molecules have successfully advanced to the clinic including MK‐7622, a selective M₁ receptor PAM, GSK1034702, an allosteric M1 agonist, as well as AZD‐6088 and HTL‐9936, which are orthosteric M₁ receptor agonists (Brown et al., 2013; Felder et al., 2018). These next‐generation orthosteric M₁ receptor agonists differ from first‐generation agonists, such as xanomeline, in terms of functional selectivity but lack the high degree of brain penetration desirable to optimizing the benefit/risk profile.

We report here the discovery of a novel, selective, brain penetrant spiropiperidine (SPP) partial orthosteric agonist at M₁ receptors, SPP1. This molecule was identified through the use of a novel receptor occupancy assay used to drive a structure–activity relationship (SAR) in a relevant brain region while simultaneously tracking drug levels in plasma and brain to optimize for central penetration. Functional activity at M₁ receptors was tracked in relevant native in vitro assays to allow translational (rat–human) benchmarking of SAR molecules to clinical comparators.

Methods

In vivo animal experiments

All animal care and experimental procedures were carried out in accordance with the Guide for the Care and Use of Laboratory Animals as adopted and promulgated by the US National Institutes of Health and were approved by Eli Lilly's Animal Care and Use Committee. All studies involving animals are reported in accordance with the ARRIVE guidelines for reporting experiments involving animals (Kilkenny et al., 2010; McGrath & Lilley, 2015). For occupancy experiments, male Sprague–Dawley rats (177–235 g) and wild‐type C57Bl/6J mice (17–25 g) were purchased from Harlan (Indianapolis, IN, USA,). M₁ receptor KO mice (line#1781; 15–47 g) were purchased from Taconic (Hudson, NY, USA). All animals were group‐housed and provided with food and water ad libitum. A normal 12 h/12 h light/dark cycle was applied with lights on at 7:00 h. The total number of rats used was approximately 20.

Ex vivo animal experiments

For Xenopus oocyte experiments, adult female Xenopus laevis frogs were purchased from Nasco (Fort Atkinson, WI, USA). The care and use of the frogs complied with the guidelines of the UK Animals Scientific Procedures Act (1986) and associated guidelines. Frogs were kept in the laboratory in a climate‐controlled (20–23°C) and light‐regulated room with a 12 h light/12 h dark cycle. The animals were fed twice a week with trout pellets, and once a week, they were given earthworms. Eight frogs were used in this study. Frogs were anaesthetized by immersion in 0.5% 3‐aminobenzoic acid ethyl ester until the animals became unresponsive to toe pinch. Toads were then decapitated, and ovarian lobes were harvested and defolliculated by incubation in 2 mg·mL⁻¹ collagenase (Type 1 C‐0130, Sigma‐Aldrich, UK) in Ca²⁺‐free Barth's saline at room temperature. Defolliculated stage V–VI oocytes were selected and injected with 5 ng of M₁ receptor cDNA.

All animal care and experimental procedures described below were reviewed by the local ethics committee and complied with the UK Animals Scientific Procedures Act (1986). For GTPγS and radioligand binding experiments, male Sprague–Dawley rats (200–300 g) were obtained from Charles River (Harlow, UK). For electrophysiological experiments, wild‐type male C57Bl/6J and M₁ receptor KO mice (as described above) were obtained from Envigo (Loughborough, UK). For functional atrial and ileal assays, male or female Wistar rats (375–425 g) were used. All animals were group‐housed and provided with food and water ad libitum. A normal 12 h/12 h light/dark cycle was applied with lights on at 7:00 h. The total number of rats and mice (WT and M₁ receptor KO) used was approximately 20 and 20 respectively. Animals were killed by exposure to CO₂ followed by cervical dislocation. For electrophysiological experiments, mice were exposed to isoflurane followed by decapitation.

Monkey brain tissue was obtained from two rhesus monkeys (1 male/1 female) from Covance, USA and shipped to the UK under CITES (Convention on International Trade in Endangered Species). Samples of cortex were dissected from fresh frozen brain slices and homogenized as per the protocol below. Tissue from the male monkey was used for radioligand binding experiments, and tissue from the female monkey was utilized in GTPγ[³⁵S] experiments.

Human brain tissues

Human brain tissue from post mortem healthy and AD patients was provided to Eli Lilly from the Oregon Alzheimer's Disease Center with appropriate consent and utilized in experiments in the UK under the Human Tissue Act 2004. AD tissue was from subjects in Braak stage 5/6 as determined by number of amyloid plaques and neocortical tangles. Details of the demographic and histopathological status of the samples used in this study are included in Supporting Information Table S2.

Receptor occupancy

Live phase

Male Sprague–Dawley rats (N = 4 per dose group) from Charles River were used in the rat studies. Male C57Bl/6J mice from Harlan (N = 3–4 per dose group) were used in the mouse studies. Test compounds were administered by p.o. gavage at a single screening dose of 10 mg·kg⁻¹. SPP1 was administered at doses of 0.03, 0.1, 0.3, 1, 3 and 10 mg·kg⁻¹ for generation of dose‐response curves and at 3 mg·kg⁻¹ for the time‐course study. Animals received either vehicle alone (1% hydroxyethylcellulose, 0.25% polysorbate 80 and 0.05% antifoam in purified water) or test compound in a dose volume of 10 mL·kg⁻¹. In the dose–response studies, rats or mice received i.v. administration of non‐labelled tracer (LSN3172176, 10 μg·kg⁻¹, 0.5 mL·kg⁻¹ dose volume for rats and 5 mL·kg⁻¹ dose volume for mice) in the lateral tail vein 30 min after vehicle or compound administration. In the time‐course study, rats received i.v. administration of non‐labelled tracer (same conditions) in the lateral tail vein 30 min after vehicle administration and 30, 60, 120, 240, 480 or 1440 min after administration of SPP1. Animals were killed by cervical dislocation 20 min after tracer administration. Brains were removed and dissected. Frontal cortex and cerebellum were used for the tracer measurement, and the remaining brain and plasma were used for compound exposure analysis. In the time‐course study, CSF was also collected. The receptor occupancy is considered to be measured at the time of tracer administration (t), but the SPP1 exposure is measured at the time of kill (t plus 20 min). Studies were performed at Covance Alnwick or Greenfield.

Tissue preparation and tracer analysis: Cortex and cerebellar samples were weighed and placed in conical centrifuge tubes on ice. Four volumes (w/v) of acetonitrile containing 0.1% formic acid was added to each tube. Samples were then homogenized using an ultrasonic probe and centrifuged using a benchtop centrifuge at 22 000× g for 20 min. Supernatant was diluted by adding 50 to 150 μL sterile water in 96‐well plates for LC/MS/MS analysis. Analysis of LSN3172176 was carried out using an API 4000 mass spectrometer (SCIEX, Framingham, MA, USA). Chromatographic separation employed an Agilent Zorbax Eclipse XDB‐C18 column (2.1 × 50 mm; Agilent Technologies, Santa Clara, CA, USA) and a gradient mobile phase consisting of 15 to 90% acetonitrile in water with an overall 0.1% formic acid content. Detection of LSN3172176 was accomplished by monitoring the precursor to product ion transition with a mass to charge ratio (m/z) of 386.3 to 128.0. Standards were prepared by adding known quantities of the tracer to brain tissue samples from non‐treated rats or mice and processing as described above.

Receptor occupancy determinations

Receptor occupancy was calculated using the ratio method. The level of tracer was measured in each cortical and cerebellar sample. A ratio of cortical levels (total binding) to cerebellar levels (non‐specific binding) was generated for each animal. Vehicle ratios represent 0% occupancy, and a ratio of 1, where the binding in the cortex is equal to the binding in the cerebellum, represents 100% occupancy. The ratios from the SPP1 pretreated groups were interpolated linearly between the ratio in the vehicle‐treated animals (0% occupancy) and 1 (100% occupancy) in order to determine the per cent M₁ receptor occupancy. For the SPP1 dose response, a curve was fitted to a four‐parameter logistic function with the bottom and top fixed at 0% and 100%, respectively, using GraphPad Prism version 6.0, and the dose achieving 50% receptor occupancy (RecOcc₅₀) was calculated by the software. Values are given as mean ± SEM.

For conversion of total plasma or brain levels to unbound levels, the % values for SPP1 free in the plasma (4.4%) and brain (10.1%) were used. The unbound brain to unbound plasma concentration ratio is the KP,uu, where KP is the total brain to total plasma concentration ratio and uu stands for unbound brain and plasma.

Native tissue membrane preparation

All procedures were performed at 4°C. Frozen human, rhesus monkey, rat or mouse cortices or hippocampi were homogenized in sucrose buffer (10 mM HEPES, 1 mM EGTA, 1 mM DTT, 10% sucrose and 1 tablet per 50 mL Complete Protease Inhibitor Cocktail; pH 7.4) using an electric IKA RW20 homogenizer (800 r.p.m.) with glass/Teflon homogenizer. Homogenates were centrifuged at 1000× g for 10 min and supernatant collected; the pellet was re‐homogenized and centrifuged again, as above, and supernatant pooled and centrifuged at 11 000× g for 20 min. The resulting pellet was suspended in a final storage buffer (10 mM HEPES, 1 mM EGTA, 1 mM MgCl₂, 1 mM DTT; pH 7.4) and centrifuged at 27 000× g for 20 min. Supernatant was removed, and the final pellet was suspended in a quantity of final storage buffer to give a suitable protein concentration. Protein concentration was measured using the Bradford method (Coomassie Plus, Bio‐Rad protein assay kit; Biorad, Watford, UK) with bovine gamma gobulin (BGG) standards. Samples were then aliquoted and stored at −80°C.

GTPγ[³⁵S] binding assays

GTPγ[³⁵S] binding in human, rat and mouse (WT and M₁ receptor KO) membranes was determined in replicates of three or more using an antibody capture technique in 96‐well plate format (DeLapp et al., 1999). Membranes (15 μg per well for rat and mouse or 30 μg per well for human) were incubated with test compound and GTPγ[³⁵S] (500 pM per well) for 30 (rat), 45 (mouse) or 60 min (human). Labelled membranes were then solubilized with 0.27% Nonidet P‐40 plus Gqα antibody (E17, Santa Cruz Biotechnology, Inc., Dallas, TX, USA) at a final dilution of 1:200 and 1.25 mg per well of anti‐rabbit scintillation proximity beads. Plates were left to incubate for 3 h and then centrifuged for 10 min at 200× g. Plates were counted for 1 min per well using a Wallac MicroBeta Trilux scintillation counter (Perkin Elmer). All incubations took place at room temperature in GTP‐binding assay buffer 20 mM HEPES, 100 NaCl mM and 5 mM MgCl₂, pH 7.4. Data were converted to % response compared to oxotremorine M (Oxo‐M; 100 μM) or ACh (100 μM) before being transferred to GraphPad Prism software 6 to generate an EC₅₀ (four‐parameter curve fit, sigmoidal dose–response and variable slope).

A Furchgott analysis was performed on GTPγ[³⁵S]‐binding curves to evaluate the level of receptor reserve in rat and monkey hippocampus. For Furchgott analysis, binding was performed as above in the absence or presence of alkylating agent (60 nM propylbenzyl choline mustard), as described by Porter et al. (2002). Equally effective concentrations of Oxo‐M, SPP1 or xanomeline in alkylated and unalkylated cortical membranes were analysed by double reciprocal plot. The resulting straight line was used to calculate a K _A value using the equation: K _A = (slope − 1)/y‐intercept. Per cent occupancy at specific agonist concentrations was calculated from the K _A using the equation: % occupancy = 100·L/(L·1K _A), where L is the agonist concentration.

Functional M₂ and M₃ receptor tissue assays

Tissue studies using rat atria and ileum were run by Eurofins PanLabs, Taiwan (Lambrecht et al., 1989). Briefly, left atria and strips of ileal longitudinal muscle (1.5 cm length) were isolated from Wistar rats and were set up under 0.5 g tension in 10 mL organ baths at 32°C containing oxygenated (95% O₂–5% CO₂) McEwens (atrium) or Krebs (ileum) buffers. The tissue responses to the cumulative addition of compound were measured as changes in isometric (atrium) or isotonic (ileum) tension. A force‐displacement transducer connected to a Hellige amplifier and a multichannel recorder was used for these measurements. For experiments involving atria, agonist‐induced negative inotropy is expressed relative to a 1 μM methacholine response, and for antagonist responses, inhibition of a 1 μM methacholine‐induced negative inotropic response is calculated. For experiments involving ileum, agonist‐induced contraction is expressed relative to a 1 μM methacholine response, and for antagonist responses, inhibition of a 1 μM methacholine‐induced contraction is calculated. Assay details also available on the Eurofins website (Eurofins PanLabs, 2017a,b).

Radioligand binding competition studies

All experiments were performed in assay buffer of the following composition: 20 mM HEPES, 100 mM NaCl and 10 mM MgCl₂, pH 7.5. Experiments using recombinant M receptor membranes used 10 μg of protein per well. Experiments using native tissues used 30 μg (mouse/rat) or 50 μg (monkey/human) protein per well with all membranes prepared as described for GTPγ[³⁵S] binding above. Recombinant membranes were incubated with 0.2 nM [³H]‐N‐methylscopolamine ([³H]NMS). Native membranes were incubated with 2 nM [³H]LSN3172176 (Mogg et al., 2018). Binding was performed in the presence or absence (total binding) of 11 different concentrations of compound. Non‐specific binding was determined in the presence of 10 μM atropine. All assay incubations were initiated by the addition of membrane suspensions, and deep well blocks were shaken for 5 min to ensure complete mixing. Incubation was then carried out for 2 h at 21°C. Binding reactions were terminated by rapid filtration through GF/C filters pre‐soaked with 0.5% w/v polyethyleneimine (PEI) for 1 h. Filters were then washed three times with ice‐cold assay buffer. Dried filters were counted with Meltilex A scintillant using a Trilux 1450 scintillation counter (Perkin Elmer). The specific bound counts (dpm) were expressed as a percentage of the maximal binding observed in the absence of test compound (total) and non‐specific binding determined in the presence of 10 μM atropine. Data analysis was accomplished using Excel and GraphPad Prism 6. The concentration–effect data were curve‐fit using GraphPad Prism 6 to derive the potency (IC₅₀) of the test compound. The equilibrium dissociation constant (K_i) of the test compound was then calculated by the Cheng–Prusoff equation: K_i = IC₅₀/(1 + ([L]/K_d)) using the following K_d values: M₁ K_d = 196 pM; M₂ K_d = 769 pM; M₃ K_d = 642 pM; M₄ K_d 143 pM; and M₅ K_d = 410 pM. Mouse K_d = 0.52 nM, rat K_d = 2.66 nM, monkey K_d = 1.33 nM, and human K_d = 1.93 nM.

Xenopus oocyte recordings

The nucleotide sequence encoding the human M₁ receptor was ligated into the pcDNA3.1 plasmid vector (Invitrogen, Carlsbad, CA, USA). A volume of 18.2 nL of cDNA solution, with a cDNA concentration of 0.1 mg·mL⁻¹, was injected into the nuclei of the oocytes using a Nanoject Automatic Oocyte Injector (Drummond, Broomall, PA, USA). After injection, oocytes were incubated at 18°C for 4 days in a modified Barth's solution containing 88 mM NaCl, 1 mM KCl, 2.4 mM NaHCO₃, 0.3 mM Ca (NO₃)₂, 0.41 mM CaCl₂, 0.82 mM MgSO₄, 15 mM HEPES and 5 mg·L⁻¹ neomycin (pH 7.6). Oocytes were placed in a 0.1 mL recording chamber and perfused with an oocyte recording solution (ORS) at a rate of 10 mL/min. ORS contained: 115 mM NaCl, 2.5 mM KCl, 1.8 mM CaCl₂; MgCl₂; and 10 mM HEPES. The pH of this solution was adjusted to 7.2 with NaOH.

Four days after injection, oocytes were placed in a recording chamber and impaled by two microelectrodes filled with 3 M KCl (0.5–2.0 MΩ) and voltage‐clamped at −60 mV using a Geneclamp 500B amplifier and PCLAMP 10 software (Axon Instruments, Union City, CA, USA). Traces were filtered at 100 Hz during recording and digitized at 500 Hz using the DigiData 1200 interface (Axon Instruments). During the experiments, oocytes were continuously perfused at a rate of approximately 5 mL·min⁻¹. Drugs were applied by switching the valves of the perfusion system (BPS8; ALA Scientific Inc., Westbury, NY, USA). All experiments were carried out at room temperature. Experiments were performed on oocytes that had leakage currents of less than 100 nA at the holding potential of −60 mV.

Hippocampal slice electrophysiology

Six‐week‐old, male C57Bl/6J mice (Charles River) and M1 KO C57Bl/6NTac male mice (Harlan) were anaesthetized with isoflurane and killed by decapitation. The brain was removed and placed into ice‐cold artificial CSF (aCSF) composed of 87 mM NaCl, 75 mM sucrose, 2.5 mM KCl, 25 mM NaHCO₃, 1.25 mM NaH₂PO₄, 7 mM MgCl₂, 25 mM D‐glucose and 0.5 mM CaCl₂, continuously bubbled with 95% O₂/5% CO_2. Transverse slices (350 μm thick) were prepared from the hippocampi isolated from each cerebral hemisphere using a tissue slicer according to the manufacturer's instructions. Slices were kept in a submerged holding chamber at room temperature for 30 min and then transferred to normal aCSF, composed of 124 mM NaCl, 3 mM KCl, 26 mM NaHCO_3, 1.25 mM NaH₂PO₄, 1 mM MgSO₄, 10 mM glucose and 2 mM CaCl₂, continuously bubbled with 95% O₂/5% CO₂. Slices were allowed to recover for at least another 30 min at room temperature before being transferred to a submerged recording chamber continuously perfused at ~3.5 mL·min⁻¹ with aCSF maintained at 30–32°C by means of an inline heater (Scientifica, Uckfield, UK). Slices were weighed down with the aid of a slice anchor (made in house with a platinum‐iridium wire).

A glass recording electrode (1.5 mm OD, 0.86 mm i.d. glass with filament from Harvard, USA, filled with aCSF, and a resistance of 1–4 MΩ) and a bipolar stimulating electrode (made from Ni : Cr (80:20) formvar‐coated wire (Advent Research Materials Ltd., Oxford, UK) were placed in the stratum radiatum within the CA1 area of the hippocampus. Responses were evoked by monophasic constant current stimulations of 0.1 ms duration at 30 s intervals. Data were filtered at 3 kHz and amplified 500× with an Axopatch 1D amplifier (Molecular Devices Corporation, Sunnyvale, CA, USA) and digitized at 10 kHz. Averages of four successive trials were analysed online and offline using WinLTP software (Anderson and Collingridge, 2007). The slope of the field EPSP (fEPSP) was used as a measure of synaptic strength. The stimulus intensity was set to a level which elicited about half of the maximal slope. After obtaining a stable baseline for at least 20 min, SPP1 was superfused onto the slices for 20 min at increasing concentrations (100 nM, 1 μM and 3 μM). SPP1 was prepared as a 3 mM stock solution in DMSO. Aliquots were stored at −20°C until required and diluted to their final concentration in aCSF just before use.

The % change over baseline was calculated using the pooled data average of the last 5 points (10 min) for each drug concentration versus the pooled data average of the last 5 points (10 min) of the baseline. All data are expressed as mean ± SEM. Statistical significance was assessed using unpaired t‐test, by comparing WT and M₁ receptor KO mice using the average of the last 5 points (10 min) for each concentration applied. Data analysis was performed using Excel and graphs constructed using Excel and Prism.

Cell culture

CHO cells expressing the human M₁ receptor were grown in DMEM/F12 (3:1) supplemented with 10% FBS, 10 mM HEPES, 2 mM Glutamax, 40 ug·mL⁻¹ proline, penicillin–streptomycin, 250 ug·mL⁻¹ geneticin and seeded at 50 000 cells per well in PDL coated 96‐well black/transparent bottom plates for 24 h before the experiment.

Ca²⁺ influx assays

Receptor‐mediated changes in intracellular calcium concentration were determined using a calcium‐sensitive fluorescent dye, Fluo4‐AM (Invitrogen) and a fluorimetric imaging plate reader (FLIPR; Molecular Devices Corporation). HBSS assay buffer, supplied by Invitrogen (Gibco 14025‐050), was supplemented with 20 mM HEPES and adjusted to pH 7.4. At the start of the assay, growth media were removed from the 96‐well cell plates by inversion and gentle tapping, and media were replaced with assay buffer containing 4 μM Fluo‐4‐AM/0.05% (v/v) pluronic F‐127. Cell plates were stored in the dark at room temperature for 60 min, to allow dye loading into cells. Subsequently, the dye solution was removed and replaced with assay buffer at which stage cell plates were transferred to the FLIPR for assay. Intracellular calcium levels were monitored before and after the addition of compounds, with data collection ranging from one image every second to one image every 3 s. Responses were measured as the maximal peak height in relative fluorescent units (RFUs), and the assay window was defined as the maximal response obtained by agonist‐stimulated wells. All RFU values were corrected for basal fluorescence in the absence of agonist.

pERK assays

CHO cells stably expressing the human M₁ receptor were generated in house using the Flp‐In expression system. Cells were cultured and plated as for Ca²⁺ experiments described above. Cells were serum starved overnight prior to the experiment. Phospho‐ERK1/2 (Thr²⁰²/Try²⁰⁴) levels after agonist stimulation (10 min at 37°C) were measured using a commercially available HTRF kit from CisBio (France) following the manufacturer's instructions.

Computational methods

SPP1 was initially docked using the default Schrodinger induced fit procedure in Maestro 2012.1 into a homology model of the hM₁ receptor built using the inactive state hM₂ + hM₃ X‐ray structures (3UON + 4DAJ) as templates. The pose was further refined by minimizing the ligand coordinates into a homology model of hM₁ built using the active state M₂ structure 4MQS as template, subsequent to the release of these structures. Minimization and homology modelling were done with the Amber10:EHT force field in MOE 2013.

Data and statistical analysis

The data and statistical analyses comply with the recommendations on experimental design and analysis in pharmacology (Curtis et al., 2018). For receptor occupancy experiments, animals were randomly distributed between treatment groups, and experimenters were blinded for analysis purposes. Unless otherwise stated, values reported are means ± SEM. Comparisons between groups were made using ANOVA with Bonferroni's or Dunnett's post hoc tests, as appropriate, using GraphPad Prism software version 6.0 (GraphPad Software Inc., San Diego, CA, USA). Statistical significance was taken as P < 0.05. For calculation of K_i values for the native atrial and illeal assays, the method of Cheng and Prusoff (1973) was used. EC₅₀ values of methacholine were from Clague et al. (1985). For hippocampal slice electrophysiology experiments, the following test was performed. Per animal, the averages of the measurements at the applicable time points for baseline and at each of the treatments were determined. Then, per animal, the ratio of each of the three treatments' values with the animal's applicable baseline value were calculated as well as the log₁₀ transform of this ratio. These normalized and log‐transformed measurements of the animals across treatments were used to build a repeated measurements model (using SAS® proc mixed with unstructured covariance) with genotype (WT and KO) and treatment (100 nM, 1 μM and 3 μM) as categorical factors, including a term for interaction. Type 3 tests for significance of the fixed effects were carried out as well as least‐squares means estimation for each genotype.

Materials

SPP1 (cyclopropyl 4‐[3‐methyl‐1‐(m‐tolyl)‐4‐oxo‐1,3,8‐triazaspiro[4.5]decan‐8‐yl]piperidine‐1‐carboxylate), SPP2 (cyclopropyl 4‐[1‐(3‐chloro‐4‐fluoro‐phenyl)‐3‐methyl‐4‐oxo‐1,3,8‐triazaspiro[4.5]decan‐8‐yl]piperidine‐1‐carboxylate), SPP3 (isopropyl 4‐[3‐methyl‐1‐(m‐tolyl)‐4‐oxo‐1,3,8‐triazaspiro[4.5]decan‐8‐yl]piperidine‐1‐carboxylate), propylbenzilylcholine mustard (PrBCM), SAR and reference standards (AZD6088, GSK1034702, HTL9366, LY593093, Oxo‐M, RS86, xanomeline) were prepared by Eli Lilly and Company (Lilly Research Centre, Windlesham, UK). [³H]LSN3172176 (specific activity 1.295 TBq/mmol) was synthesized by Quotient Bioscience, Cardiff, UK. [³H]‐N‐methylscopolamine ([³H]NMS;specific activity 3.16 TBq/mmol) and Chinese hamster ovary (CHO) cell membranes expressing human M₁‐M₅ receptors were obtained from Perkin Elmer (Boston, MA, USA). All other materials and compounds were obtained from standard commercial sources.

Nomenclature of targets and ligands

Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY (Harding et al., 2018), and are permanently archived in the Concise Guide to PHARMACOLOGY 2017/18 (Alexander et al., 2017a, 2017b, 2017c).

Results

Use of in vivo receptor occupancy to drive SAR for a M₁ receptor agonist

Assessment of central M₁ receptor target engagement (receptor occupancy) was performed by monitoring in vivo block of an M₁ receptor tracer molecule, LSN3172176 (Mogg et al., 2018) administered 30 min after p.o. administration of SAR molecules. This time point was selected as a suitable time point from preliminary investigations. Initially molecules were screened at a single dose of 10 mg·kg⁻¹ in rats. Molecules displaying good occupancy levels, favourable KP,uu and an appropriate in vitro functional selectivity and efficacy profile at native rat M₁‐M₃ receptors were subsequently screened to generate full dose‐occupancy curves. Compounds were screened at doses ranging from 0.03 to 30 mg·kg⁻¹ and brain and plasma collected to allow tracking of the brain unbound concentrations required to achieve RecOcc₅₀ and KP,uu, the goal being to drive down brain concentrations required to achieve RecOcc₅₀ and drive up KP,uu towards highly brain penetrant molecules (Figure 1).

Figure 1.

Driving a structure–activity relationship using M₁ receptor occupancy and unbound brain to unbound plasma concentration ratios. Compounds were screened at doses ranging from 0.03–30 mg·kg⁻¹ to generate 6‐point dose–response curves. Brain and plasma was collected from each animal at each dose level to allow calculation of unbound brain and unbound plasma concentrations. Receptor occupancy was then plotted against unbound drug concentration to generate the unbound plasma and brain concentrations generating 50% occupancy (RecOcc₅₀). This allowed tracking of the EC₅₀ nM brain unbound required to achieve RecOcc₅₀ compared to the unbound plasma concentrations and KP,uu, the goal being to decrease brain EC₅₀ and increase KP,uu towards 1 (highly brain penetrant).

The in vitro activity of SPP1, a representative molecule from the SPP series, was assessed in GTPγ[³⁵S] binding studies using membranes prepared from hippocampus or cortex (Figure 2). In mouse hippocampal membranes, SPP1 stimulated Gq‐mediated GTPγ[³⁵S] signalling with an EC₅₀ = 12.3 ± 2.1 nM and E _max = 44 ± 1.9% (n = 3), relative to 100 μM Oxo‐M. This potent, partial agonist activity (relative to the full agonist Oxo‐M) was confirmed to be exclusively mediated by activation of M₁ receptors, as signals were absent in hippocampal membranes prepared from M₁ receptor KO mice (Figure 2A). In membranes prepared from rat cortex, SPP1 stimulated Gq‐mediated signalling with similar EC₅₀ = 21.0 ± 5.1 nM and E _max = 31 ± 1.3% (n = 4) values (data not shown). In studies using membranes prepared from human post mortem control or AD (Braak stage 5/6) frontal cortex tissue, the pharmacological profile of SPP1 was also conserved, with EC₅₀ values of 10.4 ± 3.4 and 8.1 ± 2.5 nM and E _max values of 36 ± 7.6 and 25 ± 4.1% respectively (n = 3–5) (Figure 2B).

Figure 2.

(A) Function of M₁ receptors was assessed in mouse hippocampal membranes from WT or M₁ receptor KO (M1^‐/‐) mice using GTPγ[³⁵S] Gq antibody capture binding methodology. Responses were normalized to a maximally effective concentration of Oxo‐M (100 μM). Points represent mean ± SEM from three individual experiments each containing four replicate conditions. (B) Function of M₁ receptors was assessed in native human post mortem frontal cortex from control or AD patients (Braak stage 5/6), with separate patient samples used in each individual experiment, using GTPγ[³⁵S] Gq antibody capture binding methodology. Responses were normalized to a maximally effective concentration of ACh (100 μM). Points represent mean ± SEM from three individual experiments each containing four replicate conditions.

To bench mark the in vitro potency and efficacy profile of SPP1 and other SAR molecules to reference standards and ligands that have previously been, or are currently in clinical development, additional studies in native rat and human cortical tissue preparations were performed (Figure 3). The data reveal close alignment of potency and efficacy values across species. Of note, three orthosteric reference agonists which have advanced to the clinic, xanomeline, AZD‐6088 and HTL‐9936, similar to SPP1, displayed partial agonist profiles at native rat and human receptors in cortical membranes. The potency and partial efficacy profile of xanomeline and AZD‐6088 was comparable to those of SPP1 (EC₅₀ 9–37 nM, E _max 14–29%), while HTL‐9936 was significantly less potent but with higher efficacy (EC₅₀ 2–2.7 μM, E_max 45–77%) (Figure 3; Felder et al., 2018).

Figure 3.

Correlation plots for potency and efficacy values at native rat and human cortical M₁ receptors for SAR molecules and reference standards. Native M₁ receptor function was assessed in rat and human cortex using GTPγ[³⁵S] Gq antibody capture binding methodology as shown in Figure 2. Responses were normalized to a maximally effective concentration of Oxo‐M (100 μM). Potency and efficacy values were correlated between the two species using Pearson's correlation coefficient in GraphPad Prism 6 software.

Next, the level of Gαq/11‐coupled M₁ receptor reserve in the hippocampus was measured in vitro by GTPγ[³⁵S] binding after receptor alkylation with propylbenzylcholine mustard (PrBCM) (Porter et al., 2002). The concentration of PrBCM, which resulted in a 50% decrease in the efficacy of Oxo‐M, SPP1 and xanomeline, was used to calculate the percentage of receptor occupancy at specific agonist concentrations. A maximal Oxo‐M stimulated Gαq/11 response was achieved at ~25% M₁ receptor occupancy in rat and monkey hippocampal membranes, indicating a substantial level of spare receptors. Receptor reserve was estimated to be ~75% for Oxo‐M (Figure 4; Supporting Information Figure S2). Similar experiments performed to assess receptor reserve in rat hippocampus for SPP1 and xanomeline revealed that both agonists displayed a receptor reserve of ~40%, with maximal efficacy achieved at ~60% receptor occupancy (Figure 4; Supporting Information Figure S2).

Figure 4.

Receptor reserve of agonists was assessed using Furchgott analysis in rat or monkey hippocampus using GTPγ[³⁵S] Gq antibody capture binding methodology. A dose–response curve of Oxo‐M (n = 2 monkey and n = 3 rat), SPP1 (n = 3) or xanomeline (n = 2) was assessed in the presence and absence of an alkylating agent (propylbenzyl choline mustard) at a concentration which irreversibly inhibited 50% of the response (please refer to Supporting Information Figure S2 for raw data). Data from replicate experiments were pooled, and from these curves, Furchgott plots were generated.

The functional selectivity of SPP1 was assessed in vitro in relevant native tissue assays. In rat atrial (M₂ receptor ) and rat ileum (M₃ receptor) functional assays, SPP1 was devoid of agonistic effects up to the highest concentration tested (Figure 5) but did block a subsequent challenge to a non‐selective M receptor agonist, carbachol, with IC₅₀ (K_i) values of 5.37 μM (1.74 μM) and 26.3 μM (9.89 μM) respectively (Supporting Information Figure S3). Potency and efficacy of three clinically advanced mAChR agonists, xanomeline, HTL‐9936 and AZD‐6088, were also assessed in these atrial and ileum tissue assays and have been reported (Felder et al., 2018). SPP1 also possessed high selectivity versus a panel of 29 other general cross‐reactivity targets when assessed at 10 μM (Supporting Information Table S1).

Figure 5.

Functional native rat tissue selectivity data. Rat atrium and ileum tissue response assays measured as changes in isometric (atrium) or isotonic (ileum) tension. For experiments involving atria, agonist‐induced negative inotropy is expressed relative to a 1 μM methacholine response. For experiments involving ileum, agonist‐induced contraction is expressed relative to a 1 μM methacholine response. Data points represent mean ± SEM (n = 4; SPP1, n = 2; xanomeline, each with two replicates).

The affinity and selectivity profile of SPP1 was further assessed in radioligand competition assays by assessing displacement of [³H] NMS by SPP1 from human recombinant M₁–M₅ receptors. In addition, we employed a tritiated version of a novel M₁ receptor selective ligand [³H]LSN3172176 (Mogg et al., 2018) to assess displacement from native M₁ receptors from mouse, rat, monkey and human cortex (Table 1 and Supporting Information Figure S1). SPP1 displayed high binding affinity for human M₁ receptors (K_i values of 21.3 ± 1.2 and 14.5 ± 2.5 nM on human recombinant and native M₁ receptors respectively), which was conserved across species (K_i values of 2.88 ± 0.36, 8.67 ± 0.41 and 10.7 ± 1.3 nM on mouse, rat and monkey respectively; Table 1 and Supporting Information Figure S1). SPP1 displayed modest selectivity for binding to human M₁ receptors over human M₄ > M₂ > M₅> > M₃ receptor subtypes, as did AZD‐6088. In contrast, HTL‐9936 and xanomeline did not show preferential binding to M₁ receptors (Table 2 and Supporting Information Figure S1; Mogg et al., 2018).

Table 1.

Comparison of SPP1 binding affinity to native mouse, rat, monkey or human M1 receptors

	Mouse	Rat	Monkey	Human
	Mean Ki ± SEM	Mean Ki ± SEM	Mean Ki ± SEM	Mean Ki ± SEM
SPP1	2.88 ± 0.36 nM	8.67 ± 0.41 nM	10.7 ± 1.3 nM	14.5 ± 2.5 nM

Competition radioligand binding experiments measuring displacement of [³H]LSN3172176 were used to assess affinity (K_i) of SPP1 in cortical membranes from mouse, rat, monkey and human. Values represent the mean ± SEM of data from three individual experiments each containing four replicates for each condition.

Table 2.

Comparison of compound affinity at human recombinant M₁–M₅ receptor subtypes

	M1	M2	M3	M4	M5
	Mean Ki ± SEM (Selectivity vs. M1)	Mean Ki ± SEM (Selectivity vs. M1)	Mean Ki ± SEM (Selectivity vs. M1)	Mean Ki ± SEM (Selectivity vs. M1)	Mean Ki ± SEM (Selectivity vs. M1)
SPP1	21.3 ± 1.2 nM	128 ± 9.0 nM (6‐fold)	7855 ± 1094 nM (370‐fold)	115 ± 0.7 nM (5.4‐fold)	195 ± 12.6 nM (9.2‐fold)
HTL‐9936	9093 ± 887 nM	6502 ± 297 nM (0.7‐fold)	150 300 ± 31 979 nM (16.5‐fold)	3804 ± 319 nM (0.4‐fold)	55 490 ± 9391 nM (6.1‐fold)
AZD‐6088	62.7 ± 19.4 nM	125 ± 8.3 nM (2‐fold)	11 871 ± 4197 nM (189‐fold)	133 ± 48 nM (2.1‐fold)	602 ± 75.4 nM (9.6 fold)

Competition radioligand binding experiments measuring displacement of [³H] NMS were used to assess affinity (K_i) of compounds in purified membranes from recombinant cells expressing M₁–M₅ receptor subtypes. Values represent the mean ± SEM of data from three individual experiments each with four replicates for each condition.

Recruitment of downstream signalling pathways was assessed in heterologous expression systems, in either CHO cells (Table 3) or Xenopus oocytes (Figure 6) expressing recombinant human M₁ receptors. In CHO‐hM1 cells, concentration response curves evoked by SPP1, AZD‐6088, HTL‐9936 or xanomeline were assessed by measuring intracellular calcium responses using FLIPR or pERK signalling using ELISA (Table 3). All agonists were highly efficacious (>70% efficacy) in evoking calcium responses with a maximal efficacy comparable to ACh. In contrast, in the same cells, the agonists showed significantly different maximal efficacies for pERK signalling, with AZD‐6088 being the least efficacious agonist (11% compared to a maximal response to ACh) and xanomeline demonstrating the highest efficacy at 55% (Table 3). In oocytes heterologously expressing M₁ receptors, application of 10 μM Oxo‐M evoked an inward current (Figure 6A) mediated by endogenous calcium‐dependent chloride channels activated by calcium release from intracellular stores on M₁ receptoractivation. The inward current shown in Figure 6A is the chloride current flowing through the activated calcium‐dependent chloride channels which occurred a few seconds after Oxo‐M was applied. Application of 10 μM xanomeline or SPP1 also resulted in the activation of calcium‐dependent chloride currents in oocytes (Figure 6B, but the amplitude of these were significantly smaller than those induced by Oxo‐M [0.129 ± 0.043 (n = 8), 1.12 ± 0.32 (n = 7) and 5.25 ± 1.09 μA (n = 9) for SPP1, xanomeline, and Oxo‐M respectively]. This is consistent with SPP1 and xanomeline being partial agonists of M₁ receptors, compared with Oxo‐M in this expression system. Control experiments were performed in order to exclude a possible contribution of endogenous M receptors in the oocytes used for these experiments. Both the application of 10 μM Oxo‐M or 10 μM SPP1 to un‐injected oocytes from the same batch as those used for the M₁ receptor cDNA injected oocytes did not evoke a chloride current at all (not shown).

Table 3.

Comparison of downstream signalling in a recombinant hM₁ receptor expressing cell line

	hM1 Ca2+ mobilization (FLIPR)	hM1 pERK
	Mean EC50 (Efficacy, % ACh Max)	Mean EC50 (Efficacy, % ACh Max)
SPP1	175 ± 62.7 nM (75 ± 4%), n = 2	16.9 ± 1.1 nM (20 ± 6%), n = 2
HTL‐9936	168.8 ± 155 nM (113 ± 3%), n = 4	1161 ± 115.3 nM (49 ± 9%), n = 2
AZD‐6088	12.2 ± 13.5 nM (92 ± 14%), n = 4	118.3 ± 60.3 nM (11 ± 5%), n = 2
Xanomeline	96.6 ± 54.4 nM (112 ± 11%), n = 4	17.8 ± 2.1 nM (55 ± 9%), n = 2

Values represent the mean ± SD of data from at least 2 independent experiments.

Figure 6.

Effects of maximally effective concentrations of Oxo‐M, xanomeline and SPP1 on Xenopus oocytes expressing recombinant human M1 receptors. (A) A 10 μM of the full agonist Oxo‐M evokes large inward currents; 10 μM xanomeline evokes much smaller inward currents, demonstrating that xanomeline is a partial agonist at M1 receptors. A 10 μM SPP1 also evokes inward currents, but these are much smaller than Oxo‐M or xanomeline‐induced inward currents. (B) Bar graph showing significant differences between the currents evoked by Oxo‐M, xanomeline (Xano) and SPP1. Bars represent the mean ± SEM of responses from n = 9 (Oxo‐M), n = 7 (xanomeline) and n = 8 (SPP1) experiments. *P < 0.05, significantly different as indicated.

SPP1 was then studied in in vitro slice electrophysiology experiments to assess effects on glutamatergic synaptic transmission in the CA1 region of the hippocampus (Figure 7). In slices from WT C57Bl/6J mice, a concentration‐dependent potentiation of synaptic transmission was observed (17, 32 and 30% over baseline at 100 nM, 1 μM and 3 μM respectively; Figure 7A). In identical experiments performed in slices from M₁ receptor KO C57Bl/6NTac mice, the effect of SPP1 on synaptic transmission was greatly reduced at all of the concentrations tested (Figure 7B). Therefore, consistent with previous findings for M₁ receptor allosteric agonist GSK‐5 (Dennis et al., 2016), SPP1 potentiates glutamatergic synaptic transmission via interaction with M₁ receptors.

Figure 7.

Effects of SPP1 on excitatory synaptic transmission in mouse hippocampal slices. (A) Pooled data showing fEPSP slope values over time after addition of SPP1 at increasing concentrations (as indicated by the coloured bars) in slices from WT and M₁ receptor KO (M1 KO ) mice. Each point represents the average of four sweeps, and the bars represent the mean ± SEM; n = 5. (B) Summary of synaptic transmission responses evoked by SPP1 in slices from WT and M₁ receptor KO mice. Points represents the mean ± SEM of the last 5 points (10 min) for the indicated concentrations taken from each individual experiment. SPP1 increased synaptic transmission in WT mice compared to M₁ receptor KO mice at each concentration tested. *P < 0.05, significantly different as indicated; F‐test. Responses in KO tissues were also significantly different from baseline; t‐statistic, P < 0.05.

Moving to in vivo studies, assessment of central M₁ receptor target engagement was performed by monitoring in vivo block of a selective receptor occupancy ligand for M₁ receptors (Mogg et al., 2018) at various doses or time points after p.o. administration of SPP1 (Figure 8). A dose‐occupancy curve was generated in rats 30 min after administration of single doses of SPP1, ranging between 0.1 and 10 mg·kg⁻¹, generating a RecOcc₅₀ value of 1.1 mg·kg⁻¹ (95% CI 0.93–1.27), n = 4 (Figure 8A). Conversion of SPP1 dose to brain unbound concentrations of SPP1 ([Br]_u) led to an RecOcc₅₀ value of 24.3 nM (Figure 8B). In a time‐course study, where the level of M₁ receptor occupancy for SPP1 was measured in rats following a single p.o. dose of 3 mg·kg⁻¹ at 30, 60, 120, 240, 480 and 1440 min after p.o. administration, a time‐dependent decrease in occupancy was observed. Occupancy fell from 64 ± 8% at 30 min to 2 ± 8% at 1440 min, n = 4 (Figure 8C). The relationship between M₁ receptor occupancy and unbound plasma, unbound brain and CSF exposures was also assessed (Figure 8C). In these experiments, the KP,uu value for SPP1 of 0.62 was significantly higher than observed in the 10 mg·kg⁻¹ screening assay (KP,uu ~0.2; Figure 1). The ratio for CSF to unbound plasma levels of SPP1 (0.83) also displayed favourable distribution to brain.

Figure 8.

In vivo receptor occupancy dose–response and time‐course profile of SPP1. (A) The dose–response relationship for cortical M₁ receptor occupancy by SPP1 was assessed in rats (n = 5 per condition) 30 min after p.o. administration of 0.03–10 mg·kg⁻¹ SPP1. The dose which led to 50% receptor occupancy (RecOcc₅₀) was 1.1 mg·kg⁻¹. Points represent mean ± SEM from n = 4 independent experiments. (B) Unbound brain exposures ([Br]_u, nM) were calculated for each individual rat from the dose–response study and plotted against the % receptor occupancy observed. The [Br]_u concentration which led to 50% occupancy (RecOcc₅₀) was 24.3 nM. (C) The time‐course of cortical M₁ receptor occupancy by SPP1 was assessed in the rat (n = 5 per condition) 30–1440 min after p.o. administration of 3 mg·kg⁻¹ SPP1. The relationship between the time‐course of receptor occupancy and unbound brain, plasma or CSF concentration for SPP1 is shown. The functional potency of SPP1 at rat cortical M₁ receptors (21 nM) is indicated by the dashed line. Points represent mean ± SEM from n = 2 independent experiments.

Comparison of the potency of a large set of SAR molecules and reference ligands in the rat cortical receptor occupancy assay (RecOcc₅₀ values, [Br]_u, nM) and rat cortical membrane GTPγ^[35] S assay (EC₅₀ values, nM) revealed close alignment between the in vivo and in vitro assays (Figure 9). In a system with substantial receptor reserve, one may expect a non‐linear relationship. However, the tracer used to assess RecOcc50 is itself an agonist and thus only binds to a subset of the total pool of available M₁ receptors. Previously, we demonstrated that the B_max for [³H] NMS binding measured in the same tissues/membranes was significantly higher than for LSN3172176, the tracer molecule used for the receptor occupancy studies (Mogg et al., 2018).

Figure 9.

Relationship of in vitro potency to in vivo receptor occupancy. Correlation plot of in vitro to in vivo potency values for SAR molecules and reference standards at native rat cortical M₁ receptors. Native M₁ receptor function was assessed in rat cortex using GTPγ[³⁵S] Gq antibody capture binding methodology and by receptor occupancy experiments measuring block of the tracer molecule LSN3172176 (10 μg·kg⁻¹). Potency values between assays were correlated in GraphPad Prism 6 software and Pearson's correlation coefficient determined.

Finally, we sought to investigate the binding of SPP1 by docking it into the orthosteric site of an active state model of human M₁ receptors (Figure 10). SPP1 could successfully be docked into the model, and interactions with key residues are proposed to help gain an understanding as to how SPP1 is able to achieve functional selectivity despite binding to the highly conserved orthosteric site. The only residue different in the orthosteric site (defined as being within 4.5 Å of the modelled SPP1 pose) across the hM₁‐M₄ receptors is Leu¹⁸³ in hM₁ receptors, which is homologous to a phenylalanine in hM₂ receptors. The phenyl ring of the SPP scaffold is modelled to bind in the region around Leu¹⁸³ and would clash with Phe¹⁸¹ in hM₂ receptors. Therefore, an initial design hypothesis was that steric bulk in this region would maintain or enhance selectivity against M₂ receptors. The selectivity observed against M₃ receptors is harder to explain, but we note that this residue sits in the extracellular loop between helix 4 and 5 where neighbouring residues are not conserved between M receptor subtypes, and it is hypothesized that this region plays a role in the coupling between the orthosteric M₁ receptor site and receptor activation (Kruse et al., 2013; Abdul‐Ridha et al., 2014).

Figure 10.

Compound SPP1 docked into the orthosteric site of an active state model of human M₁ receptors. SPP1 (grey) docked into the orthosteric site of an active state model of the human M₁ receptor (blue), overlaid with PDB:4MQS M₂ structure iperoxo ligand + key residue (green). The basic amine in both compounds is proposed to sit in a similar place, and both are proposed to form hydrogen‐bonding interactions to Asn³⁸² and a π–cation interaction to Tyr¹⁰⁶. The hM₁ receptor residue Leu¹⁸³ is homologous to a Phe in hM₂ receptors, the only residue difference in the orthosteric site. The Trp³⁷⁸ residue is hypothesized to rotate 45° from its position in the iperoxo structure to accommodate the carbamate sidechain. No other large movements of the orthosteric site are necessary to fit the significantly larger compound.

Discussion

We report here the discovery of a novel, highly functionally selective, partial agonist at M₁ receptors that is amenable to in vivo study. SPP1 potently stimulates native M₁ receptor‐mediated Gq signalling in regions relevant for cognition, displaying partial efficacy compared to the endogenous agonist, ACh. This profile is conserved across species (mouse, rat and human) and in post‐mortem AD brain samples. SPP1 has high functional selectivity versus M₂ (>100‐fold) and M₃ receptors (>100‐fold) anti‐targets when assayed in relevant rat tissue assays. SPP1 displays high affinity for human M₁ receptors, with no significant difference in binding affinity for these receptors across species and selectivity over a panel of cross‐reactivity targets. SPP1 displays partial agonism at native and recombinant receptors in vitro and evokes changes in electrophysiological parameters in CA1 hippocampal neurons which are absent in tissue from M₁ receptor KO mice. In vivo SPP1 displays dose‐dependent occupancy of M₁ receptors in rat cortex when dosed p.o.

Use of an in vivo central M₁ receptor target engagement assay to drive the chemistry SAR proved to be an effective way to optimize desirable parameters. Traditional approaches would have relied more heavily on filtering first on in vitro parameters and subsequently collecting exposure data and finally assessing efficacy of the most promising SAR molecules in a battery of in vivo behavioural assays. Our approach decreases reliance on behavioural assays, the reliability and translatability of which have been questioned. Receptor occupancy assays may also be less time consuming, requiring fewer animal numbers to generate highly reproducible results, aiding efficient drug discovery. The close correlation between the in vivo brain unbound exposures achieved at the RecOcc₅₀ and in vitro EC₅₀ values in the rat cortex GTPγ[³⁵S]‐Gq assay for a range of M₁ receptor agonists, as well as the overlapping pharmacology observed at rat and human M₁ receptors gives confidence in the robustness of the data and highlights the translational value of the assays.

Benchmarking of SPP1 to M₁ receptor ligands that have advanced to the clinic, within the same in vitro and in vivo assays allowed for relative comparisons to be drawn. Xanomeline displayed potent partial M₁ receptor agonist properties, similar to SPP1 in the GTPγ[³⁵S] assay. Xanomeline also displayed high brain penetration but displayed little functional selectivity for M₁ over M₃ receptors and clinically was poorly tolerated at the efficacy dose in AD patients (Bodick et al., 1997). Selectivity over M₂ and M₃ receptors has been a key differentiator for the next‐generation of drug discovery, and AZD‐6088, as observed with SPP1, achieved the desired high functional selectivity for M₁ receptors in spite of binding to the highly conserved orthosteric site (Felder et al., 2018). AZD‐6088 however is a good substrate for the PGP transporter, displaying relatively low brain penetration and development of AZD‐6088 was terminated clinically due to adverse events at peripheral sites (Felder et al., 2018). Hence, minimizing peripheral load and optimization of central penetration appear to be an important goal even for highly selective M₁ receptor agonists. SPP1 displayed a KP,uu of between 0.2 and 0.8 (including CSF partitioning), which is a significant improvement over AZD‐6088 (KP,uu 0.13). Ideally KP,uu values should be ~1, and a number of molecules with values in this range were identified in the course of SAR screening. Restricting agonist efficacy (E _max) may also be an important feature for minimizing unwanted M₁ receptor‐mediated adverse effects, such as salivation or sweating, and was a key goal for the current work. Compared to xanomeline, AZD‐6088 and SPP1, HTL‐9936 is a less potent but more efficacious M₁ receptor agonist, displaying relatively low brain penetration (Felder et al., 2018). Clinically, HTL‐9936 was reported to evoke statistically significant changes in brain electrical activity measured using many EEG biomarkers, relevant to cognition (Heptares Therapeutics Website, 2016; Felder et al., 2018).

The effect of the host system on the functional response profile of GPCR agonists is well documented (Kenakin, 2014a), which accounted for our emphasis on assessing functional pharmacology in relevant native tissues for M₁ receptors ( cortex and hippocampus) and M₂/M₃ receptor anti‐targets (atrium and ileum). A general experimental observation is that cellular responses do not require 100% receptor occupancy to produce a maximal response, leading to a phenomenon called “receptor reserve” (spare receptors not required for production of the maximal response). Previously, ~85% receptor reserve was reported for Oxo‐M in mouse hippocampus, with activation of ~15% of the receptors generating a maximal response (Porter et al., 2002). Our data with Oxo‐M in rat and monkey hippocampus closely replicate the findings of Porter and colleagues, confirming high receptor reserve in this brain region across species. Receptor reserve is a property of the tissue, the assay and the agonist (Kenakin, 2014a) and, as expected, the partial agonists SPP1 and xanomeline required higher fractional occupancy to achieve a maximal response compared to Oxo‐M. However, these agonists still displayed significant receptor reserve (~40%). This data support the hypothesis that in vivo, pharmacodynamic responses for SPP1 might be observed at low occupancy of M₁ receptors. Although understanding the occupancy required to achieve an in vitro pharmacodynamic response in a relevant tissue is useful, understanding the overall efficacy/receptor occupancy relationship in a biological system is the real goal. This relationship is important to define in order to set an appropriate clinical target efficacy dose range and to avoid overdosing. Such definition is not a simple task, as the translational value of building an efficacy/receptor occupancy relationship using preclinical behavioural assays has its limitations. The use of translational pharmacodynamic biomarkers relevant for cognition might be another option for further refining dose selection using a PET‐based strategy (Felder et al., 2018), and aligned to this, LSN3172176 has the potential to be used as a PET tracer in man (Jesudason et al., 2017; Nabulsi et al., 2017).

Our data showing that the pharmacological profile of SPP1 is conserved in AD cortical tissue is consistent with other data indicating that postsynaptic M₁ receptors are still present in such samples, and function is retained (Overk et al., 2010; Nathan et al., 2013; Bradley et al., 2017). These data confirm therefore that targeting these receptors in AD is a valid approach to increasing cholinergic signalling.

Activation of Gq (assessed in the functional GTPγS binding assays) is one of the most proximal events to activation of M₁ receptors. These receptors can also signal via non‐Gq‐dependent means, such as β‐arrestin recruitment, and recruit diverse downstream signalling cascades, demonstrating significant signal amplification and in some cases agonist dependent‐signalling bias, whereby certain cellular pathways are activated to a greater extent than others (Kenakin, 2014b). Our data revealed considerable variability in the potency and efficacy profile of xanomeline, SPP1, HTL‐9936 and AZD‐6088 across the various in vitro cellular assays studied. Such variability is likely to reflect cell and assay specific factors (e.g. receptor reserve, effector coupling and proximity to M₁ receptor activation), but a degree of signalling bias is suggested as the agonists do not retain the same rank order of potency or maximal efficacy across assays. For example, xanomeline and SPP1 displayed similar efficacy in Gq‐GTPγS functional binding but markedly different efficacies in pERK signalling and oocyte studies. These studies highlight the complexity of GPCR and M₁ receptor signalling and our lack of understanding of which signalling profile is desirable for maximizing efficacy and minimizing adverse effects. A role for β‐arrestin recruitment in M₁ receptor PAM‐mediated cognitive benefits has been suggested by Ma et al. (2009) and, in some elegant studies of the M₃ receptors using genetically engineered mice to map physiological pathways, a compelling case has been put forward for pursuit of next‐generation GPCR therapies based on understanding and exploiting this complexity (Bradley et al., 2014, 2016).

Confirmation that SPP1 could alter hippocampal neuronal excitability and that the effects are largely M₁ receptor‐dependent is a relevant finding for a potential therapeutic designed to alter hippocampal function in vivo. M₁ receptors are abundant in human hippocampus (Levey et al., 1995), and physiologically, activation of these receptors leads to the regulation of a range of ion channels that are co‐expressed in the same neurons, including NMDA receptors (Markram and Segal, 1990; Calabresi et al., 1998; Marino et al., 1998; Zwart et al., 2017) and potassium channels (Shapiro et al., 2000; Buchanan et al., 2010; Giessel and Sabatini, 2010; Zwart et al., 2016). Currents through NMDA receptors are enhanced upon activation of M₁ receptors, whereas currents through potassium channels are reduced. Both effects result in an increase in excitability of the postsynaptic neuron.

For many years, it was believed that selectivity could not be achieved with ligands that bound to the orthosteric site of M receptors, due to the high sequence identity between M₁–₅ receptors at this site. However, the new highly functionally selective M₁ receptor agonists reported by us and others suggest that functional selectivity is achievable, even if binding selectivity for these molecules is more modest. We suggest that this may stem from the larger size of these next‐generation lignads compared to ACh and several first‐generation non‐selective agonists, as well as the orientation in which they occupy the orthosteric site and affect allosteric receptor activation.

In summary, we report the discovery of SPP1, a novel, functionally selective, brain penetrant partial orthosteric agonist at M₁ receptors, identified through the use of a novel receptor occupancy assay. SPP1 is amenable to both in vitro and in vivo study and should provide a valuable research tool to further probe the role of M₁ receptors in physiology and pathophysiology.

Authorship contributions

L.M.B., P.J.G., H.E.S., A.J.M., E.S., V.B., C.C. and G.N.W. participated in research design. H.E.S., A.J.M., E.M.C., D.A.E., F.P. and R.Z. conducted the experiments. P.J.G. contributed new reagents or analytic tools. P.J.G., H.E.S., A.J.M., E.M.C., D.E., F.P., R.Z., G.N.W. performed data analysis. L.M.B., A.J.M., H.E.S., D.E., R.Z. and P.J.G. wrote or contributed to the writing of the manuscript.

Conflict of interest

The authors declare no conflicts of interest. Eli Lilly does not sell any of the compounds or devices mentioned in this article.

Declaration of transparency and scientific rigour

This Declaration acknowledges that this paper adheres to the principles for transparent reporting and scientific rigour of preclinical research recommended by funding agencies, publishers and other organisations engaged with supporting research.

Supporting information

Figure S1 Muscarinic Receptor Subtype Selectivity and Species Selectivity. (A, C, D) Recombinant mAChR in vitro binding profile. Displacement of [³H] NMS radioligand binding by increasing concentrations of SPP1 (A), HTL‐9936 (C) and AZD‐6088 (D) across mAChR receptor subtypes M1‐M5. (B) Displacement of [³H]LSN3172176 by SPP1 in native cortical tissues across species (mouse, rat, monkey, human). Data points represent mean specific binding ± SEM from 3 independent experiments each containing 3 replicates.

Figure S2 Receptor reserve of agonists was assessed using Furchgott analysis in rat (A‐C) or monkey (D) hippocampus using GTPγ[³⁵S] Gq antibody capture binding methodology. A dose response curve of oxotremorine‐M (n = 2 monkey and n = 3 rat) , SPP1 (n = 3) or xanomeline (n = 2) was assessed in the presence and absence of an alkylating agent (propylbenzyl choline mustard) at a concentration which blocked irreversibly inhibited 50% of the response. Data points represent mean response ± SEM from the indicated number of experiments each containing 3 replicates.

Figure S3 Functional native rat tissue selectivity data for SPP1. Rat atrium and ileum tissue response assays measured as changes in isometric (atrium) or isotonic (ileum) tension. For experiments involving atria, inhibition of a 1 uM methacholine‐induced negative inotropy response is shown. For experiments involving ileum, inhibition of a 1 uM methacholine‐induced negative inotropy contraction is shown. Data points represent mean ± SEM (n = 4).

Table S1 Cross‐reactivity Data for SPP1.

Table S2 Human Tissue Samples ‐ Demographic and Histopathological Information.

Broad, L. M. , Sanger, H. E. , Mogg, A. J. , Colvin, E. M. , Zwart, R. , Evans, D. A. , Pasqui, F. , Sher, E. , Wishart, G. N. , Barth, V. N. , Felder, C. C. , and Goldsmith, P. J. (2019) Identification and pharmacological profile of SPP1, a potent, functionally selective and brain penetrant agonist at muscarinic M₁ receptors. British Journal of Pharmacology, 176: 110–126. 10.1111/bph.14510.