AF710B, a Novel M1/σ1 Agonist with Therapeutic Efficacy in Animal Models of Alzheimer's Disease

Stimulation of the M1 muscarinic acetylcholine receptor (M1 mAChR) is an attractive therapeutic approach for Alzheimer's disease (AD), as reviewed in Fisher [1,2]. Accordingly, we recently generated compelling preclinical data, both pharmacologically and genetically, that activation of M1 mAChR restores cognition and attenuates AD-like pathology in several animal models [3,4]. Mechanistically, it was found that M1 mAChR targets crucial enzymes involved in the synthesis of β-amyloid (Aβ) and phosphorylation of tau [1,2,3,4].

The σ1 receptor (σ1R) is another potential target for drug development for AD, as it is considered to play a fundamental role in cognitive function [5,6]. Notably, decreased levels of σ1R have been detected in early-stage AD patients [7], and that stimulation of σ1R alleviates the depressive symptoms, neuronal cell death and cognitive deficits in AD mouse models [8,9]. Moreover, the mixed muscarinic/σ1R agonist ANAVEX2-73 resulted in significant inhibition of Aβ_25-35-induced tau phosphorylation and Aβ_1-42 seeding in mice [10] and was recently reported to exert positive cognitive effects in phase 2a clinical studies with AD patients (http://www.anavex.com/?post_type=news&p=1491).

σ1R is an endoplasmic reticulum (ER)-resident protein and a molecular chaperone, yet its functional role remains a mystery [5,6]. Under homeostatic conditions, σ1R is concentrated at mitochondrion-associated membrane junctions - sites of contact between ER and mitochondria - where it regulates InsP3R-mediated calcium signaling and ER lipids [11]; however, agonists or ER stress can cause σ1R to exit the mitochondrion-associated membrane domain [6], allowing it to regulate the function of plasma membrane ion channels [5] and G-protein-coupled receptors (GPCRs) [6,8,12,13]. Among other brain regions, the σ1R is expressed in hippocampal neurons with preferential localization to plasmalemmal, outer mitochondrial and ER cisternal membranes within somata, dendrites and postsynaptic elements [14]. Ultrastructural analysis also shows colocalization of σ1R with mAChR in postsynaptic densities [15]. However, the mechanistic nature of σ1R interactions with ion channels and GPCRs including mAChRs is poorly understood.

Over the past several years, our groups have focused on understanding the underlying molecular mechanism of M1 mAChR involvement in AD, and on the development of innovative therapeutics to treat this insidious disease [1,2,3,4]. Thus, we have created selective M1 mAChR orthosteric agonists in the ‘AF series' of compounds (e.g. AF102B, AF267B, AF292) that are cognitive enhancers and disease modifiers (reviewed in Fisher [1,2]; fig. 1).

In a further expansion of our rational drug discovery programs, we report here on AF710B (an active enantiomer of AF710; MW 357.5), which is a novel and potent M1 mAChR/σ1R agonist (fig. 2).

We report here results of the preclinical evaluation of AF710B in cellular and animal models of AD. Notably we found that AF710B showed exceptional efficacy in restoring cognitive decline associated with AD and with lessening BACE1, glycogen synthase kinase 3β (GSK3β) activity, p25/cyclin-dependent kinase 5 (CDK5), neuroinflammation, soluble and insoluble Aβ₄₀, Aβ₄₂, accumulation of amyloid plaques and neurofibrillary tangles. As the stabilization of postsynaptic dendritic spines in hippocampal neurons is important for learning and memory, we evaluated the potential role of M1 mAChR and σ1R in the regulation of synaptic morphology. We recently found that mushroom spines are preferentially lost in PS1-M146V-knockin (KI) mouse [16] and APP-KI mouse [17] models of AD. As mushroom spines are considered stable ‘memory spines', their loss may underlie cognitive decline in AD [16]. We therefore examined the effects of AF710B versus AF267B on hippocampal mushroom spine stability and tested the involvement of M1 mAChR and σ1R in the mechanism of action of AF710B. The results of this study indicate that AF710B, but not AF267B, efficiently rescues mushroom spines and normalizes other pathological features in AD models.

1-(2,8-Dimethyl-1-thia-3,8-diazaspiro[4.5]dec-3-yl)-3-(1H-indol-3-yl)propan-1-one (AF710; US Patent 8,673,931B2) was synthesized at the Israel Institute for Biological Research (IIBR) and was separated by Chiral Technology Europe into two enantiomers, AF710A and AF710B. AF710B [>99.5% chemical purity; >99.5% enantiomeric excess; specific rotation α = -56° (C = 0.303, methanol)], the more active enantiomer, was used in the studies reported in this paper (see US Patent 8,673,931B2).

AF267B [(2S)-2-ethyl-8-methyl-1-thia-4,8-diazaspiro(4.5)decan-3-one; >99.9% chemical purity; >99.5 enantiomeric excess] was provided by IIBR. All other compounds were of analytical grade and were purchased from commercial sources.

AF710B (10 µM) was tested in a high-throughput profiling that consists of a broad collection of 83 trans-membrane proteins (including GPCRs) and soluble receptors, sigma receptors, ion channels and monoamine transporters (CEREP, France; http://www.cerep.fr/cerep/users/pages/catalog/profiles/DetailProfile.asp? profile=2118). AF710B was further tested in binding studies at a wide concentration range on M1 mAChR (IIBR; see below), σ1R, σ2R or α₄β₂-, α₇-nicotinic receptors as well as in cellular functional assays on histamine 2, 5-HT2A, 5-HT2C, 5-HT2B and mu receptors, and at 5-HT2B receptors in an in vitro rat stomach fundus bioassay, respectively (CEREP, France).

All procedures in which animals were used were in accordance with: (i) the National Institute of Health (NIH) Guide for the Care and Use of Laboratory Animals (1996) and were approved by the IIBR Institutional Animal Care and Use Committee; (ii) the NIH guidelines and were approved by the Institutional Animal Care and Use Committee of the University of Texas Southwestern Medical Center, and (iii) the Principles of Laboratory Animal Care from NIH publication 85-23 and were approved by the University of California, Irvine, Institutional Animal Care and Use Committee.

Male Sprague-Dawley rats, 250-300 g, were sacrificed by decapitation, and brains were removed. The cerebral cortex was dissected out and placed on ice, cleaned, weighed and transferred to 20 mM Tris-HCl buffer, 2 mM EDTA, pH 7.4. The tissue was homogenized in the buffer (1:10 weight/volume) using a polytron homogenizer, and after a -70°C freeze/thaw cycle, the homogenates were centrifuged at 35,000 g at 4°C for 10 min. The supernatants were removed, and the pellets were resuspended in Tris buffer at a ratio of 1:6 (weight/volume). The homogenates were divided into aliquots of 1 ml each and then stored at -70°C till use. The binding profile in rat cortical membranes of AF710B, carbachol and the effects of AF710B on carbachol binding to M1 mAChR, respectively, were studied in triplicates for each concentration of the tested compounds in displacement studies of the M1 selective antagonist [³H]pirenzepine (86 Ci/mmol, Perkin-Elmer, Mass., USA). Each such study was replicated in 4-6 separate experiments. Competition curves, K_H, K_L and K_i values were derived using the GraphPad Prism software program, version 5.0.

Rat pheochromocytoma cells (PC12) stably transfected with rat M1 mAChR (PC12M1 cells) were seeded in 6-well plates at a density of 2 × 10⁶ cells/well. The following day, cells were washed twice with RPMI and returned to the incubator with serum-free media (starvation medium containing RPMI, 2 mM glutamine, 1% penicillin-streptomycin, 2.5 µg/ml amphotericin B, 0.1 mg/ml G418). On the next day, cells were pretreated for 3 h with AF710B at concentrations ranging from 0.1 to 100 nM, and then carbachol (10 nM) was added for 10 min. In each plate, one well served as a control (no treatment) and one well as a positive reference in which the cells were treated with carbachol (10 nM). Following this procedure, cells were collected and extracted with Ripa buffer (Sigma, R-0278, 200 μl) containing a protease inhibitor cocktail (Sigma, 1:100) and phenylmethylsulfonyl fluoride (Sigma, 1:1,000). Phospho-p44/42 mitogen-activated protein kinase [MAPK; phosphorylated extracellular signal-regulated kinase 1/2 (p-ERK1/2); 1:1,000, Cell Signaling] and phosphorylated cAMP response element-binding protein (p-CREB, Ser133; 1:1,000, Cell Signaling) were probed with anti-rabbit antibodies.

Naive Wistar rats, 225-250 g (3 months old), were trained and tested in a two-compartment box: a small illuminated compartment (21 × 23 × 22 cm³) with a lamp (60 W) at a height of 25 cm above the compartment, and a dark, large compartment (37 × 23 × 22 cm³) equipped with a grid floor. A door (7 × 5 cm²) separated the two compartments. The passive avoidance (PA) task is comprised of a training (acquisition) phase and a retention phase. In the training procedure, each rat was individually placed in the small illuminated compartment, and after 60 s of familiarization/adaptation, the door to the large compartment was opened and the latency to enter was measured (initial latency). Immediately following entry into the dark compartment, the door was closed and an inescapable foot shock (0.6 mA for 3 s) was delivered through the grid floor. A cutoff point of 180 s was used for initial latency. Animals that failed to enter the dark compartment with all four paws within 180 s were excluded from the experiment. After the acquisition trial, the rat was returned to its home cage. Retention of the PA task was measured 24 h later, by again placing the rat in the light compartment, and after a 60-second adaptation interval, the door was opened and the latency to re-enter the dark compartment was measured. A cutoff point of either 300 (experiments 1 and 2) or 600 s (experiment 3) was used for retention latency. Animals that failed to step through within 300 or 600 s were removed from the apparatus and a 300- or 600-second latency was recorded for them.

In experiment 1, we tested the effects of AF710B (100, 30 and 10 µg/kg, p.o.) and in experiment 2 lower doses (1 and 3 µg/kg, p.o.). In both experiments the rats were divided into 2 groups. One main group was treated with trihexyphenidyl (5 mg/kg, s.c.), and the second main group was treated with vehicle (double-distilled water, DDW; 1 ml/kg, s.c.), 30 min before the shock. In each main group, rats were further divided into treatment subgroups (n = 9-11): one subgroup was treated with saline (10 ml/kg, p.o.), and the other subgroups were treated with AF710B, 60 min before the shock. In experiment 3, rats were divided into 2 groups. One group was treated with trihexyphenidyl (5 mg/kg, s.c.), and the second group was treated with vehicle (DDW 1 ml/kg, s.c.), 30 min before the shock. In each group, rats were divided into treatment subgroups (n = 9-10): one subgroup was treated with saline (10 ml/kg, p.o.), and the other subgroups were treated with AF710B (10 µg/kg, p.o.), with or without the σ1R antagonist NE-100 (1 mg/kg, p.o.), 60 min before the shock.

Initial latency and retention latency were analyzed by a two-way MANOVA. Specific comparisons were made by simple main effect contrast analysis.

PS1-M146V knockin (PS1-KI) [18] and APP-KI [19] homozygous mouse AD models were used in our studies. Wild-type (WT) nontransgenic (nTg) mice of the same strain (B6 strain) were used as a control. Primary hippocampal cultures from WT, PS1-KI and APP-KI mice were established from postnatal day P0-P1 pups and maintained in culture as previously described [16,17,20]. At day 7 in vitro (DIV7), hippocampal cultures were transfected with a tdTomato plasmid using a high calcium phosphate method [16]. At DIV16, neuronal cultures were treated with placebo (NBA medium), AF710B (30 nM) and AF267B (1 µM) for 16 h. Following drug treatment, the cultures were fixed (4% formaldehyde and 4% sucrose in PBS, pH 7.4) for imaging. Z stacks of optical sections were captured using a ×100 objective with a confocal microscope (Carl Zeiss Axiovert 100M with LSM510). Quantitative analysis for dendritic spines was performed by using the NeuronStudio software package (as in Sun et al. [16]). In brief, following classification of spine shapes (thin, stubby, mushroom), the percentage of mushroom spines was calculated for each neurite. For each genotype and drug combination, 12-20 cultured neurons were analyzed from at least 2 independent neuronal preparations. The data were pooled and averaged.

σ1R knockdown was achieved by lentivirus-mediated shRNA expression in hippocampal neurons. Bacteria containing plasmids encoding MISSION® shRNA targeting mouse σ1R (clone ID: NM_011014.2-1138s21c1; sequence: CCGGCCTGTAGTAATCTCTGGTGAACTCGAGTTCACCAGAGATTACTACAGGTTTTTG) or scrambled shRNA were ordered from Sigma-Aldrich (St. Louis, Mo., USA). Plasmids were purified using a Maxi-prep (NucleoBond Xtra kit; Macherey-Nagel). After replacing culture media (DMEM + 10% FBS) with NBA, plasmids for shRNA along with plasmids for Δ8.9 and VSVG for lentiviral production and packaging were transfected into HEK293 cells using PEI. Culture media were collected 48 h after transfection, centrifuged (2,000 rpm for 5 min), filtered (0.45 µm pore size) and aliquoted in cryotubes for storage at -80°C until use. 200 µl of the collected lentivirus media was added to each well of neuron cultures on DIV8. Both functional analysis in the spine loss assay and SDS-PAGE Western blot analysis of lysates from neuron cultures - using σ1R (SC-137075; Santa Cruz) and β-tubulin (E7; DSHB) antibodies and the immunoblotting protocol below - confirmed efficient knockdown by hippocampal neurons expressing small interfering RNA (siRNA) against σ1R.

Female 3xTg-AD mice harboring a presenilin-1 mutation (PS1M146V), the Swedish double mutation in APP (APPKM670/671ML) and a frontotemporal dementia mutation in tau (tauP301L) were used for all experiments [21]. Strain-matched nTg mice, 129/C57BL/6, were used as controls. AF710B was administered via intraperitoneal injection at a dose of 10 μg/kg/day. Mice were treated for 2 months, from 10 months of age to 12 months of age. No changes in body weight were found between vehicle- and AF710B-treated mice.

Behavior paradigms to assess cognition were performed for all groups of mice 2 weeks before sacrifice as described previously [4]. Mice were trained to swim to a circular clear Plexiglas platform submerged 1.5 cm beneath the surface. Four trials were performed per day, for 60 s each with 5 min between trials. Mice were trained for as many days as needed for the group to reach the training criterion of 25 s. The probe test was assessed 24 h after the last trial, with the platform removed. Performance was monitored with the EthoVision XT video-tracking system (Noldus Information Technology, Leesburg, Va., USA).

Mice were deeply anesthetized with sodium pentobarbital and killed by transcardial perfusion with 0.1 M PBS (pH 7.4) solution. The right brain hemispheres were fixed for 48 h in 4% paraformaldehyde and cryoprotected in 30% sucrose for immunohistochemical analysis. Frozen brains were sectioned coronally into 40-µm sections using a Leica SM2010R freezing microtome, serially collected in cold 0.02% sodium azide and stored at 4°C. The left hemispheres were snap-frozen on dry ice and subjected to protein extraction sequentially using T-PER tissue protein extraction reagent (Thermo Scientific, Rockford, Ill., USA) and 70% formic acid. The supernatant was aliquoted and stored at −80°C. The protein concentration in the supernatant was determined using the Bradford assay.

Equal protein amounts were separated on a 4-12% gradient SDS-PAGE, transferred to a nitrocellulose membrane and incubated overnight at 4°C with primary antibody. The following primary antibodies were used in this study: anti-Aβ_1-16 (6E10; Covance Research Products, Denver, Pa., USA), p-GSK3β (Ser9), human APP-CT20, ADAM10, ADAM17, BACE1, GSK3β, CDK5 (Calbiochem, San Diego, Calif., USA), human tau (HT7), phosphorylated tau (p-tau) AT180 (phospho-Thr231), p-tau AT270 (phospho-Thr181; Thermo Scientific), PP2A, p35, GAPDH (Sigma-Aldrich) and p-tau PHF-1 (phospho-S396/S404; Dr. Peter Davies, Albert Einstein College of Medicine, Manhasset, N.Y., USA). Following washing, the membranes were incubated with adjusted secondary antibodies coupled to horseradish peroxidase. The immunocomplexes were visualized using the SuperSignal West Pico Kit (Thermo Scientific). Band density measurements were made using ImageJ 1.36b imaging software (NIH, Bethesda, Md., USA).

For the determination of the Aβ levels, T-PER soluble fractions were loaded directly onto ELISA plates, whereas the formic acid supernatants (insoluble fractions) were diluted 1:20 in a neutralization buffer (1 M Tris base, 0.5 M NaH₂PO₄) before loading. MaxiSorp immunoplates (Nunc, Rochester, N.Y., USA) were coated with mAb20.1 antibody (Dr. William E. Van Nostrand, Stony Brook University, Stony Brook, N.Y., USA) at a concentration of 25 µg/ml in coating buffer (0.1 M Na₂CO₃, pH 9.6) and blocked with 3% bovine serum albumin. Standard solutions for both Aβ₄₀ and Aβ₄₂ were made in the antigen capture buffer (20 mM NaH₂PO₄, 2 mM EDTA, 0.4 M NaCl, 0.05% 3-[(3-cholamidopropyl)dimethylammonio]propane sulfonate, and 1% bovine serum albumin, pH 7.0) and loaded onto ELISA plates in duplicate. Samples were then loaded (also in duplicate) and incubated overnight at 4°C. Plates were then washed and probed with either horseradish peroxidase-conjugated anti-Aβ₄₀ (C49) or anti-Aβ₄₂ (D32; Dr. Vitaly Vasilevko and Dr. David H. Cribbs, University of California, Irvine, Calif., USA) overnight at 4°C. The chromogen was 3,3′,5,5′-tetramethylbenzidine, and the reaction was stopped by 30% phosphoric acid. The plates were read at 450 nm using a plate reader (Molecular Dynamics, Sunnyvale, Calif., USA). The readings were then normalized to protein concentrations of the samples.

Free-floating sections were pretreated with 3% hydrogen peroxide and 10% methanol in Tris-buffered saline (TBS) for 30 min to block endogenous peroxidase activity. After a TBS wash, sections were incubated once in 0.1% Triton X-100 in TBS for 15 min and once with 2% bovine serum albumin in TBS + Triton X-100 for 30 min. Sections were then incubated overnight at 4°C with p-tau AT8 (p-Ser212/Thr214, Thermo Scientific) with 5% normal serum in TBS. After the appropriate biotinylated secondary antibody (1:200 in TBS + 2% bovine serum albumin + 5% normal serum), sections were processed using the Vectastain Elite ABC reagent and 3,3′-diaminobenzidine (Vector Laboratories, Burlingame, Calif., USA) according to the manufacturer's instructions. Sections were then mounted on gelatin-coated slides, dehydrated in graded ethanol, cleared in xylene and coverslipped with DPX mounting medium (BDH Laboratory Supplies, Poole, UK).

The immunostaining was assessed at 6 brain coronal levels. Specifically, 6 alternate 40-µm sections of the brain with an individual distance of ∼160 µm were obtained between 1.34 and 2.54 mm posterior to the bregma. Images of stained hippocampal area were acquired using an Axiocam digital camera and AxioVision 4.6 software connected to an Axioskop 50 microscope (Carl Zeiss Microimaging, Thornwood, N.Y., USA). A threshold optical density that best discriminated staining from the background was obtained using the ImageJ 1.36b imaging software (NIH). All histological assessments were made by an examiner blinded to sample identities.

Sections were first blocked with 3% normal serum, 2% bovine serum albumin and 0.1% Triton X-100 in TBS for 1 h at room temperature. Using the same buffer solution, sections were incubated overnight at 4°C with the following primary antibodies: anti-Aβ_1-16 (6E10; Covance Research Products), anti-Iba-1, anti-GFAP (Millipore, Billerica, Mass., USA), HT7 and/or p-tau AT8 (p-Ser202/Thr205; Thermo Scientific). Sections were then rinsed and incubated for 1 h with secondary Alexa Fluor-conjugated antibodies (Invitrogen, Carlsbad, Calif., USA) at room temperature. Finally, sections were mounted onto gelatin-coated slides in Fluoromount-G (Southern Biotech, Birmingham, Ala., USA) and examined under a Leica DM2500 confocal laser microscope using the Leica Application Suite Advanced Fluorescence software (Leica Microsystems, Bannockburn, Ill., USA). The immunofluorescence was assessed at the same brain coronal levels as described above. Confocal images were acquired by sequential scanning using a z-separation of 1 μm using the Leica Application Suite Advanced Fluorescence software (Leica Microsystems).

Sections were incubated in 0.5% thioflavin S in 50% ethanol for 10 min, differentiated twice in 50% ethanol, and washed in PBS solution. Staining was visualized under a confocal microscope. Volumetric imaging and image measurements were made using Imaris software (Bitplane Inc., South Windsor, Conn., USA). The thioflavin S levels represent the average value obtained by the analysis of images of the hippocampus.

All data are expressed as means ± SEM. The statistical evaluation of the results was carried out using one- or two-way analysis of variance (ANOVA). Following significant ANOVAs, multiple post hoc comparisons were performed using Bonferroni's test. Some data were analyzed using the unpaired t test. The accepted level of significance for the tests was p < 0.05. All tests were performed using the Statistica software package (StatSoft Inc., Tulsa, Okla., USA).

In high-throughput receptogram profiling, AF710B emerged as a potent ligand for M1 mAChR and σ1R, respectively, as shown by displacement of an M1 muscarinic antagonist (tritiated pirenzepine, an M1 mAChR radioligand) from rat cerebral cortex and of a σ1R agonist (tritiated pentazocine, a σ1R radioligand) from guinea pig cerebral cortex with a 6- and 2-fold order of magnitude interval between two binding sites (K_H vs. K_L), respectively (fig. 3a). Furthermore, AF710B showed both high potency and selectivity for the M1 mAChR and σ1R and very clean additional pharmacological profiles with regard to off-target activity (fig. 3a). Thus, AF710B was found to display no significant off-target activity when screened at 10 µM in a radioligand displacement screen at 83 other GPCRs, ion channels and transporters known to mediate human side effects (fig. 3a). AF710B (10 nM to 100 µM) was inactive as an agonist or an antagonist on histamine 2 and 5-HT2A and 5-HT2C receptors while on the 5-HT2B receptor it is a very weak partial agonist (17% efficacy) at 100 μM and a very weak agonist at the mu receptor, with EC₅₀ = 13 µM. Additionally, AF710B (0.1 nM to 10 µM) did not bind to human α₄β₂- and α₇-nicotinic receptors and to σ2R, while it showed agonistic effects on M1 mAChR, but not on M2-M5 mAChR (table 1). In addition, AF710B application altered the neurochemical profile to a state typified by downstream signaling of M1 mAChR and/or σ1R. This was detected at concentrations in the nanomolar range ∼3-4 orders of magnitude lower than effects on other screened GPCRs, ion channels and transporters mentioned above (detailed data not shown; table 1).

To characterize the interaction of AF710B with the M1 mAChR, we employed a combination of binding and functional assays. This was done first utilizing brucine, an M1-positive allosteric modulator (M1 PAM) that potentiates M1 orthosteric, but not M1 allosteric agonists [22]. Notably while brucine showed a modest cooperativity on M1 mAChR with the orthosteric agonists carbachol and AF267B, it failed to potentiate AF710B (tested in binding and calcium mobilization assays) and TBPB (another M1 allosteric agonist [23]; detailed data not shown; table 1). Thus, AF710B binds to M1 mAChR on a location different from the orthosteric site, supporting an allosteric mechanism of action of AF710B.

To further characterize the M1 allosteric profile of AF710B, we tested its effects on carbachol. AF710B (0.1 nM) potentiated carbachol binding to M1 mAChR in rat cortical membranes as evidenced by a decrease by approximately 2 orders of magnitude of the K_H for carbachol. The ratio of K_L/K_H for a given agonist directly correlates with its efficacy. In this particular case, the ratio of K_L/K_H for carbachol in the presence of 0.1 nM AF710B increased from 110 to 1,650 (fig. 3b). Thus, both the affinity and efficacy of carbachol are potentiated by AF710B.

Both M1 orthosteric and allosteric agonists increase the phosphorylation of ERK1/2 [24,25]. The activated ERK/MAPK cascade has multiple targets, including CREB, which mediates its ability to induce memory consolidation and long-term memory formation [26,27]. In this context we found that AF710B did not increase p-ERK1/2, yet it potentiated ERK1/2 phosphorylation induced by 10 nM carbachol (fig. 3c). In addition, AF710B increased carbachol-induced p-CREB in starved proliferated PC12M1 cells (fig. 3d).

Taken together this indicates that AF710B is functioning like an M1 PAM mAChR. However, unlike reported M1 PAMs [23], AF710B binds to M1 mAChR with a high affinity and is also a direct M1 allosteric agonist, but different in several aspects from other reported M1 allosteric agonists (see below and table 1).

To determine whether AF710B is a cognitive enhancer, we tested its effects against cognitive impairments in a PA test induced by trihexyphenidyl in rats. Trihexyphenidyl induces cognitive deficits in rodents and humans and is a CNS-penetrating and a relatively selective M1 antagonist [28].

In general, no significant difference was found in initial latency between any of the groups tested in experiments 1-3 (data not shown). In experiment 1, a significant interaction was found between trihexyphenidyl and treatments (F_{2, 63} = 4.0, p < 0.023). Retention latency of trihexyphenidyl rats treated with DDW (54.2 ± 14.6 s) was significantly shorter than that of control rats treated with DDW (242.00 ± 30.60 s; p < 0.001). However, the retention latency of trihexyphenidyl rats treated with AF710B at 10 and 30 µg/kg, p.o., was significantly longer than that of trihexyphenidyl rats treated with DDW (p < 0.001-0.01). The retention latency of trihexyphenidyl rats treated with AF710B at 100 µg/kg, p.o., was not different from that of trihexyphenidyl rats treated with DDW. AF710B at 10 µg/kg was very potent and significantly different from the higher dosage of 100 µg/kg, p.o. (p < 0.01). In experiment 2, a significant interaction was found between trihexyphenidyl and treatments (F_{2, 52} = 5.2, p < 0.009). The retention latency of trihexyphenidyl rats treated with DDW (63.9 ± 18.8 s) was significantly shorter than that of control rats treated with DDW (274.7 ± 20.8 s; p < 0.001). However, the retention latencies of trihexyphenidyl rats treated with AF710B (1 and 3 µg/kg, p.o.) were significantly longer than that of trihexyphenidyl rats treated with DDW (p < 0.001). The antiamnesic effects of AF710B, 1 and 10 µg/kg, p.o., were replicated 2 and 4 times, respectively, and the tested compound showed a long duration of action at these doses, since the retention time was significantly different versus DDW-treated trihexyphenidyl rats, both 24 and 72 h after treatment (detailed data not shown).

As AF710B binds to both M1 mAChR and σ1R, we explored whether its antiamnesic effects are blocked by NE-100, a selective σ1R antagonist [29]. In experiment 3, we found that the retention latency of trihexyphenidyl rats treated with DDW (122.0 ± 29.2 s) was significantly shorter than that of control rats treated with DDW (525.7 ± 51.0 s; p < 0.001). However, the retention latency of trihexyphenidyl rats treated with AF710B at 10 µg/kg, p.o., was significantly longer (361.7 ± 67.1 s), and NE-100 coadministered with AF710B blocked significantly the antiamnesic effect of AF710B in this model (185.4 ± 44.0 s; p < 0.02; fig. 3f). Thus, the effects of AF710B appear to be mediated at least in part by the σ1R. Notably, NE-100 was not significantly effective in inducing deficits in the PA task and also failed to affect the retention parameters of AF710B alone (fig. 3f). As trihexyphenidyl binds 70-fold more effectively to M1 mAChR versus σ1R [30] and its amnesic effects may be attributed mainly to its M1 antagonistic effects, it can be deduced that AF710B exerts its cognitive effects in the trihexyphenidyl model via a combined activation of both M1 mAChR and σ1R.

Synaptic loss is a main feature of many neurodegenerative disorders, including AD [31]. In previous studies we determined that mushroom synaptic spines are lost in primary hippocampal neuronal cultures from PS1-KI and APP-KI mice [16,17]. We confirmed these findings in the present experiments (fig. 4a-d). Application of AF267B (1 µM, an effective concentration on M1 mAChR-mediated readouts [2]) for 16 h starting on DIV16 partially rescued mushroom spines in PS1-KI cultures, but not in APP-KI cultures (fig. 4a, b), indicating potential beneficial effects, but only for certain forms of AD. AF267B partially reduced the prevalence of WT mushroom spines, which could be related to a slight toxic effect or depression of synaptic strength. By contrast, 30 nM AF710B did not change the proportion of mushroom spines in hippocampal cultures from nTg animals when compared with vehicle (fig. 4a, 3b). Notably, AF710B markedly rescued mushroom spine loss in both AD transgenic mouse models, nearly to the level of nTg animals (fig. 4a, b), indicating that this compound exerts a disease-modifying effect on AD neurons.

To explore the underlying molecular mechanism responsible for AF710B's protective effect, similar studies with primary hippocampal cultures from APP-KI mice were performed in the presence of the M1 mAChR antagonist pirenzepine (1 μM) and/or the σ1R antagonist NE-100 (1 μM; fig. 4c). While the effect of AF710B on spine morphology was not affected by the blockage of σ1R, it was completely abolished in the presence of the M1 mAChR antagonist (fig. 4c, d). In addition, the combination of pirenzepine and NE-100 did not show any additive or synergistic effects in the presence of AF710B (fig. 4c, d). These results suggest that AF710B prevents mushroom spine loss in hippocampal neuron cultures from APP-KI mice by stimulating M1 mAChR activity.

We further examined the possible involvement of σ1R in mushroom spine stability in WT mouse hippocampal neuronal cultures by using lentivirus-mediated shRNA interference delivery. σ1R knockdown decreased the mushroom spine density from 38.4 ± 1.1% in hippocampal neurons transfected with scrambled siRNA to 22.1 ± 1%, which is significantly lower (p < 0.0001; fig. 4e, f). As before, treatment of cultures with 30 nM AF710B for 16 h had no effect on the mushroom spine density in control siRNA-expressing neurons (fig. 4e, f). However, AF710B treatment failed to rescue mushroom spine loss in hippocampal neurons expressing siRNA against σ1R (fig. 4e, f), indicating that the ability of AF710B to stabilize hippocampal mushroom spines requires not only M1 mAChR function, but also σ1R expression. To explain these results, we propose that AF710B may act by stimulating an M1 mAChR/σ1R complex in the synaptic spines. Further biochemical experiments will be needed to test this hypothesis.

Given the marked effects induced by AF710B in vitro and in vivo, we sought to further investigate its effects on the cognitive decline induced by AD using the 3xTg-AD mouse model. As previously observed, 12-month-old 3xTg-AD mice performed significantly worse relative to age-matched nTg mice in the reference spatial memory version of the Morris water maze. Vehicle-treated 3xTg-AD mice exhibited longer latencies to find the platform in the training session (fig. 5a) and reduced target quadrant preference during the probe trial (fig. 5b) compared with age-matched nTg mice treated with vehicle. Notably, treatment with AF710B (10 μg/kg/day, i.p., for 2 months) improved the cognitive function of 3xTg-AD mice during the training and probe sessions of the Morris water maze (fig. 5a, b). Importantly, the improvement in the cognitive performance in the AF710B-treated 3xTg-AD mice is not directly related to changes in motor function, since no significant alterations of the swimming speed and total distance traveled in the water maze were observed when compared with vehicle-treated nTg and 3xTg-AD mice (data not shown).

Accumulation of Aβ and tau pathology coincides with the synaptic loss and cognitive decline in the 3xTg-AD mice. Therefore, we next examined for changes in Aβ pathology. We found that AF710B treatment significantly reduced levels of Aβ₄₀ and Aβ₄₂, in both detergent soluble and insoluble fractions (fig. 5c). Moreover, we found significantly less amyloid deposition in the AF710B-treated animals versus vehicle-treated animals, as indicated by the significant decrease in thioflavin S-positive plaques (fig. 5d). To elucidate the mechanism by which AF710B reduces Aβ levels, we determined whether APP-processing pathways were modified (fig. 5e). Steady-state levels of APP and α-APP-cleaving enzymes ADAM10 and ADAM17 were unaffected by AF710B treatment. Conversely, AF710B significantly diminished the expression of the putative β-secretase enzyme BACE1. Given the decline in BACE1 levels, we used the antibody 6E10 to detect full-length APP and its proteolytic fragments CTFβ [99-amino-acid-long C-terminal APP fragment produced by β-secretase cleavage (C99)] and CTFα [83-amino-acid-long C-terminal APP fragment produced by α-secretase cleavage (C83)]. As expected, the levels of C99, but not C83, were significantly diminished in the AF710B-treated 3xTg-AD mice versus vehicle-treated 3xTg-AD mice.

To examine the effect of AF710B administration on tau pathology, we first tested for changes in total p-tau levels using the anti-AT100 antibody. Notably, p-tau was found reduced in AF710B-treated 3xTg-AD mice versus vehicle-treated 3xTg-AD mice (fig. 5f). This data was further confirmed through confocal studies, which showed a reduction in AT8-positive neurons in AF710B-treated 3xTg-AD mice (fig. 5g). Biochemical analysis of p-tau revealed significant decreases in the p-tau epitopes AT180, AT270 and PHF-1 (fig. 5h). To determine if the reductions in p-tau levels were due to modulations in tau kinase function, we measured for changes in several kinases involved in tau phosphorylation. We found that AF710B administration significantly increased phosphorylation in GSK3β, at the inhibitory Ser9 residue (fig. 5h). This was also associated with an effect on p35 cleavage, since the brains of AF710B-treated 3xTg-AD mice presented diminished levels of p25 fragment. However, AF710B did not affect the steady-state levels of CDK5, GSK3β or PP2A. Overall, our data indicates that AF710B modulates tau pathology via the modulation of GSK3β and CDK5 activities.

Inflammation is another critical component of AD, and it is associated with increased numbers and/or size of microglia and astrocytes. We found significantly fewer reactive astrocytes and activated microglia in AF710B-treated animals, as indicated by the lower detection of GFAP and Iba-1, respectively. This anti-inflammatory effect was most prominent in the immediate vicinity of plaques. The immunofluorescence studies indicated lower amyloid load and glial reactivity (fig. 5i).

We report here on AF710B, which is a novel and highly potent mixed M1 mAChR/σ1R agonist with a unique mechanism of action leading to a concomitant activation of both M1 mAChR and σ1R. Notably, AF710B showed both high potency and selectivity for the M1 mAChR and σ1R and very clean additional pharmacological profiles with regard to off-target activity (fig. 3a; table 1). AF710B exhibits an allosteric agonistic profile on M1 mAChR since very low concentrations of AF710B significantly potentiated the binding and efficacy of carbachol on M1 mAChR and some of the carbachol/M1-mediated downstream effects (e.g. ERK1/2 and CREB phosphorylation; fig. 3b-d). Furthermore, AF710B showed exceptional efficacy in vitro on downstream effects mediated by M1 mAChR and σ1R, including the ability to preserve synaptic elements (fig. 4), and in vivo in restoring cognitive deficits in animal models (trihexyphenidyl-treated rats and 3xTg-AD mice) and with lessening major hallmarks associated with AD such as BACE1, GSK3β activity, p25/CDK5, neuroinflammation, soluble and insoluble Aβ₄₀, Aβ₄₂, accumulation of amyloid plaques and neurofibrillary tangles (in 3xTg-AD mice; fig. 3, 5; table 1).

While pK studies have not yet been performed on AF710B, its exceptional brain penetration and target engagement can be implied from the very high potency of the compound in in vivo studies. Notably, AF710B functions as a cognitive enhancer in the PA test involving trihexyphenidyl-treated rats at unprecedented low doses of 1-30 µg/kg, p.o. (fig. 3e). The effects of AF710B on cognition are enantioselective since the enantiomer AF710A is not active at 1 and 10 µg/kg, p.o. (table 1, 2a). Strong enantioselectivity of AF710B versus AF710A was consistently observed in vitro and in vivo (table 1). How does AF710B pharmacodynamics overcome existing compounds? We found that AF710B is the most potent cognitive enhancer and with the highest safety margin versus a variety of compounds tested by us in the trihexyphenidyl model under the same experimental protocols as tested for AF710B, including, inter alia, AF710A, AF102B, AF267B, AF292 and donepezil (table 2a).

To explore the mechanism of action in vivo of AF710B in ameliorating PA impairments induced by trihexyphenidyl, we used NE-100, a selective σ1R antagonist [29]. We found that NE-100 coadministered with AF710B significantly blocked the antiamnesic effect of AF710B in this model (fig. 3f). Thus, the effects of AF710B appear to be mediated in part by the σ1R. However, NE-100 was not significantly effective in inducing deficits in the PA task and also failed to affect the retention parameters of AF710B alone (fig. 3f). As trihexyphenidyl binds to M1 mAChR > σ1R [30] and its amnesic effects may be attributed mainly to its M1 antagonistic effects, it can be implied that AF710B exerts its cognitive effects in the trihexyphenidyl model via a unique concomitant activation of both M1 mAChR and σ1R.

Can such effects of AF710B be recapitulated with a mixture of separate agents acting on M1 mAChR and σ1R, respectively? We tested this in the trihexyphenidyl model via a combined treatment of the M1 orthosteric agonist (AF267B) with donepezil. Notably the acetylcholinesterase inhibitor donepezil, the current major treatment in AD patients, is also a potent σ1R agonist, and its effects may arise from cholinergic mechanisms, as well as an interaction with the σ1R [32]. Importantly, we found that the effects of AF710B could not be recapitulated by the combination of AF267B + donepezil as no significant potentiation was detected and in fact the beneficial effects of the M1 agonist were unchanged by donepezil (table 2a). Thus, it can be deduced that the mechanism of action of AF710B differs from a combination of a directly acting M1 orthosteric agonist and σ1R agonist.

As AF710B potentiated the effects of carbachol on 3 tested readouts (e.g. binding, p-ERK1/2, p-CREB; fig. 3b-d), it is possible that the affinity and efficacy of the endogenous agonist acetylcholine for the M1 mAChR are also potentiated by AF710B, which may explain the improved memory performance in trihexyphenidyl-treated rat and 3xTg-AD mouse models. These findings already separate AF710B from M1 orthosteric agonists (e.g. AF267B, AF292, AF102B and others) or from σ1R agonists. Supersensitive M1 receptors are induced by positive M1 PAMs that are currently under examination by several major drug companies [23]. These M1 PAMs potentiate the whole repertoire of acetylcholine-induced M1 mAChR-mediated downstream responses, including release of intracellular Ca²⁺. In contrast, AF710B diverges mechanistically from other M1 PAMs and M1 agonists (allosteric or orthosteric) in in vitro studies, and in spite of inducing a supersensitive M1 mAChR, it does not potentiate carbachol-mediated release of intracellular Ca²⁺ (detailed data not shown; table 1). Notably, orthosteric, allosteric or bitopic M1 agonists increase intracellular Ca²⁺ at the same concentration range as that these compounds induce other downstream signaling. This is not the case for AF710B, and therefore this compound diverges mechanistically from M1 orthosteric agonists (e.g. AF267B, AF102B, AF292 and more [1,2]), M1 allosteric agonists (e.g. TBPB and more), or M1 PAMs (benzyl quinolone carboxylic acid, brucine and more [23]; detailed data not shown; table 1). Interestingly, unlike AF710B, M1 PAMs and M1 agonists (allosteric or orthosteric) were not reported to have σ1R activity. In addition, reported M1 PAMs or M1 allosteric agonists (or bitopic M1 agonists) are effective as cognitive enhancers at doses >1 mg/kg, unlike AF710B (effective at the microgram per kilogram range). Thus, AF710B exhibits a remarkably different mechanism of action and superior profile versus all these compounds.

Results from drug screening and the trihexyphenidyl model implied that AF710B might improve memory by a combined activation of M1 mAChR and σ1R. As hippocampal mushroom spine loss may underlie memory deficits in AD [16], we tested a potential role for AF710B in stabilizing synaptic connections and found that AF710B rescues mushroom spine loss in vitro in PS-KI and APP-KI hippocampal cell culture models of AD. Considering the established role of σ1R as a sensor of ER Ca²⁺ dysregulation [11] and the deranged ER Ca²⁺ signaling in AD neurons [16,20], it is likely that σ1R activity is abnormal in spines of AD neurons. Indeed, σ1R agonists affect learning and memory, cognition and mood and show potent antiamnesic properties [8], and σ1R knockdown causes hippocampal spine loss [33]. Additionally, σ1R agonists may have therapeutic potential in protein conformation diseases (e.g. AD, Parkinson's disease and more) and several neuropsychiatric disorders [8]. Conversely, σ1R KO mice are viable, but have a late-onset neurodegeneration phenotype [15]. Thus, it appears that partial loss of σ1R function or reduction in its expression may contribute to synaptic impairment and neuronal loss in these disorders. We therefore tested the potential involvement of σ1R in mediating the neuroprotective effects of AF710B. Although the σ1R antagonist NE-100 did not block the rescue of mushroom spines by AF710B in hippocampal neurons from APP-KI mice, inhibition of M1 mAChR prevented the rescue by AF710B. However, σ1R knockdown in WT hippocampal neuron cultures compromised mushroom spine stability, and AF710B treatment was unable to prevent this, indicating requirements for both σ1R and M1 mAChR activity in hippocampal mushroom spine stability. To explain these results, we propose that AF710B may act by stimulating an M1 mAChR/σ1R complex in the synaptic spines. Notably, heteromers constituted by σ1Rs with D1Rs or D2Rs (both GPCRs) were reported [12,13], and it is reasonable to presume that other GPCRs, including the M1 mAChR, may also form heteromers with σ1R. Further biochemical experiments will be needed to test this hypothesis.

The hypothesis that AF710B targets a putative M1 mAChR-σ1R complex may be tied with the procognitive effects of AF710B at unprecedented low doses in memory-based behavioral tasks in both trihexyphenidyl-treated rats (fig. 3e, f; table 2a) and 3xTg-AD mice (fig. 5a, b; table 2b [38,39,40]). In this context, in the 3xTg-AD mice AF710B mitigated cognitive deficits, and decreased BACE1 levels, GSK3β and p25/CDK5 activity, neuroinflammation, soluble and insoluble Aβ₄₀ and Aβ₄₂ plaques, and tau pathology (fig. 5). Furthermore, unlike the remarkable disease-modifying properties found with AF710B in 3xTg-AD mice at very low doses (10 µg/kg/day i.p., for 2 months), no in vivo disease-modifying properties were reported for M1 PAMs or M1 allosteric agonists and M1 dualsteric agonists in Tg mice or in other in vivo models [2,23].

GSK3β and CDK5 contribute to hyperphosphorylation of tau protein, one of the pathological hallmarks of the disease (reviewed in Su and Tsai [34]). CDK5 activity is increased in AD patients, probably due to the conversion of the CDK5 activator p35 to the truncated p25 protein (reviewed by Su and Tsai [34]). Elevated p25 levels are observed in AD patients [35]. The CDK5 activator protein p25 preferentially binds and activates GSK3β [36]. All these data supported the hypothesis that CDK5/p25 acts as a master regulator of neuronal cell death in AD, Parkinson's disease and several neurodegenerative diseases [34]. Interestingly, activation of the M1 mAChR decreased several p-tau isoforms via GSK3β inhibition but not the p-tau isoform detected with PHF-1 [3], while activation of σ1R helps maintain proper tau phosphorylation by potentially circumventing the formation of overreactive CDK5/p25 [37]. Effective therapies for tauopathies may require inhibition of both CDK5/p25 and GSK3β. AF710B fulfills this condition, and because it decreased all tested p-tau isoforms including PHF-1, it may be effective as a disease modifier even in more advanced stages of AD and also for other tauopathies. Overall, our data indicates that AF710B decreases tau pathology and may restore synaptic homeostasis by decreasing p25 and inhibiting GSK3β and CDK5 activities.

In summary, we identified AF710B as a novel comprehensive therapeutic agent to ameliorate cognitive deficits, synaptic loss, amyloid and tau pathologies, neuroinflammation, and neurodegeneration in AD and perhaps additional related protein aggregation diseases. AF710B appears to be a highly effective treatment for AD, when compared with competing drugs (table 2a, b). A drug like AF710B that works in extremely low doses provides far more dosing and testing options for future AD therapeutics.

This study was supported by grants from the NIH: National Institute on Aging P50-AG016573 (to F.M.L.) and AG027544 (to F.M.L. and R.M.) and National Institute of Neurological Disorders and Stroke R01NS080152 (to I.B.) and F32NS093786 (to D.A.R.). Dr. Abraham Fisher is an academic adviser at the Weizmann Institute of Science, Rehovot, Israel. Dr. Ilya Bezprozvanny is a holder of the Carl J. and Hortense M. Thomsen Chair in Alzheimer's Disease Research. We thank Dr. Takaomi Saido (RIKEN Brain Science Institute, Wako-shi, Saitama, Japan) for providing APP-KI mice. We thank Dr. Menahem Segal (Weizmann Institute of Science, Rehovot, Israel) for his advice and suggestions.

Abraham Fisher is member of the scientific advisory board of Anavex, USA.