Olanzapine: A potent agonist at the hM4D(Gi) DREADD amenable to clinical translation of chemogenetics

The atypical antipsychotic olanzapine is a potent activator of a mutated muscarinic receptor used for chemogenetic inhibition.


INTRODUCTION
Central nervous system (CNS) diseases caused by abnormal circuit function represent a major burden to society. Although many respond to conventional small-molecule treatment, some diseases such as intractable pain and refractory epilepsy account for a substantial unmet need. Drug-resistant focal epilepsy alone affects approximately 0.2% of the entire population (1,2). Although surgical resection of the epileptogenic zone is effective, it is contraindicated in the overwhelming majority of patients because of high risks of permanent disability associated with brain tissue removal (3). Several gene therapies for refractory epilepsy, based on altering the balance of excitation and inhibition, have been validated in preclinical models (4)(5)(6)(7)(8)(9). Chemogenetics using viral vector-mediated expression of the inhibitory muscarinic M4 receptor-based Gi-coupled DREADD (designer receptor exclusively activated by designer drug) hM4D(Gi) is especially promising because the therapeutic effect can be titrated by adjusting the dose of the activating ligand (10). Several recent publications have shown that hM4D(Gi) expressed in epileptogenic zones can suppress partial-onset seizures when activated (5,11,12).
A potential limitation to clinical translation of DREADD technology is that most studies to date have used clozapine N-oxide (CNO), the inactive metabolite of the atypical antipsychotic drug clozapine (CZP) (13), as the ligand. CNO is not approved for clinical use, and recent evidence shows that CNO is actively exported from the CNS and back-converted to CZP, which crosses the blood-brain barrier and subsequently acts as the ligand activating the DREADD (14,15). CZP as an activator of hM4D(Gi), however, represents major logistical and regulatory obstacles because it has an unfavorable side effect profile, including a risk of agranulocytosis and myocarditis, and can reduce seizure threshold (16)(17)(18). Related antipsychotic drugs have been proposed as potential agonists (13,14), and two other drugs activating DREADDs have recently been described: "compound 21" (C21) and perlapine (PLP) (19,20). Although PLP has previously been used as a mild sedative antihistamine drug in Japan, neither it nor C21 is approved for clinical use by the Food and Drug Administration (FDA) or the European Medicines Agency (EMA). Identification of an FDA/EMA-approved drug for repurposing as a DREADD activator would facilitate clinical translation of DREADD technology to treat CNS diseases.

hM4D(Gi) activation with drugs identified by similarity screens
We tested olanzapine (OZP; 3D rank 1), promazine (PZN; 3D rank 2), tripelennamine (TNA; 3D rank 5), diphenhydramine (DPH; 3D rank 6), chlorprothixen (CPX; 3D rank 9), and amoxapine (AXN; 2D rank 2). We also tested the first, putative active metabolite of quetiapine (2D rank 6), norquetiapine (NQN) (25) (for chemical structures of all tested molecules, see table S2). Of all the drugs tested, only OZP was able to fully activate hM4D(Gi) at a concentration between 100 and 300 nM using 1 M CNO as control as above (OZP/CNO = 1.18 ± 0.13; n = 3) (Fig. 2B). A full dose-response curve for OZP revealed an EC 50 of 5 ± 2 nM (Hill coefficient, 1.11 ± 0.25; n = 6), significantly lower than that of CZP (EC 50 = 61 ± 19 nM, n = 6; P = 0.0128, Stu-dent's t test) (Fig. 3A). To gain further insights into the observed activity differences on a molecular level, we docked OZP, CZP, and CPX into an active-state homology model of hM4D(Gi) using an induced fit procedure. In addition to the ionic interactions with D112, the docking poses of OZP suggest stacking interactions with W164 and hydrogen bonds involving Y116 and N117 (Fig. 3B). In addition, the OZP methyl group extends into a side pocket that is also occupied by the agonist iperoxo in the hM2 crystal structure complex [Protein Data Bank (PDB) entry 4MQS (26); fig. S1A]. This hydrophobic pocket is less occupied by CZP, which, in combination with lack of a hydrogen bond with the N117 amino group (Fig. 3B), could explain the lower activity of CZP compared to OZP. In contrast, the geometry of the inactive CPX and its lack of heteroatoms prevent the formation of any of the hydrogen bonds observed for OZP and position the basic moiety further away from D112 ( fig. S1B).

DISCUSSION
Although recent papers highlight the potential of PLP or C21 as potent activators of hM4D(Gi) (19), these ligands would require extensive screening to be approved for clinical use (30). PLP was marketed in Japan but was subsequently withdrawn, calling for an alternative licensed drug that can be repurposed as an activator of hM4D(Gi) for clinical translation of DREADD technology. The present study shows that OZP (ranked first in the 3D-based in silico screen) is a potent activator of hM4D(Gi). OZP is a secondgeneration atypical antipsychotic, which is approved by the FDA and EMA for treatment of schizophrenia and manic episodes in bipolar disorder. Common side effects of OZP at doses used in schizophrenia and bipolar disorder include weight gain, postural hypotension, and sedation (31). OZP is a D2 receptor antagonist, and its side effect profile therefore also includes akathisia, tardive dyskinesia, and neuroleptic malignant syndrome, although these are much less common than for first-generation antipsychotic drugs such as haloperidol and chlorpromazine. The in vitro EC 50 of OZP at hM4D(Gi) is in the range of affinities reported for its native drug targets (table S3) (32). The ability to affect performance on the rotarod with OZP (0.1 mg/kg) reported here is consistent with the principle of receptor reserve, whereby GPCR (heterotrimeric guanine nucleotide-binding protein-coupled receptor)-mediated effects can be achieved with low doses of agonist (33). Given that CZP is typically only prescribed for treatment-resistant patients because of its unfavorable side effect profile (34), CZP is much less suitable for repurposing as a DREADD activator. Nevertheless, the side effect profile of each activator must be considered and determined individually for every potential clinical application of hM4D(Gi). We therefore propose that OZP, which is widely available in oral, intramuscular, and intravenous formulations, is suited for clinical translation of hM4D (Gi)-based chemogenetics to treat CNS diseases, including refractory epilepsy.

Molecular biology
The hM4D(Gi) plasmid was purchased from Addgene (#45548). Standard molecular biology techniques were used to clone GFP-T2A into an AAV2 transfer plasmid (GeneOptimizer, GeneArt; Thermo Fisher Scientific). The codon-optimized version of HA-hM4D (Gi) full sequence is available upon request) was linked to GFP via a viral 2A peptide, and contained a woodchuck hepatitis posttranscriptional regulatory element (WPRE). For all in vivo experiments, the CMV promoter was replaced with a 1.3 kb CamKII promoter to allow expression in excitatory neurons (27), the antibiotic resistance was changed from ampicillin to kanamycin, and a restriction site after the 2A peptide was removed. A stop codon and restriction site after the hM4D(Gi) reading frame was inserted using polymerase chain reaction methods to facilitate excision of a nontagged hM4D(Gi) from an hM4D(Gi) mCherry plasmid (Addgene; #50477). The untagged hM4D(Gi) fragment was cloned into an AAV expression plasmid under the control of the human CamKII promoter and containing a WPRE and bovine growth hormone polyA and flanked by AAV2-inverted terminal repeats. AAV8 or AAV5 serotype vectors were packaged using methods described previously (36).

In silico screening
One low-energy conformation of C21 calculated with Omega 2.3.2 (37,38) was used as the query for the generation of the shape-based model. The default model was modified, and the final model only contained the color features shown in Fig. 2A. A maximum number of 200 conformers were generated for DrugBank version 5.0.7 (39) with Omega 2.3.2 (37,38). The default settings of vROCS 3.0.0 (22,23) were used for screening, and hits were ranked according to the TanimotoCombo score. The ChEMBL (24) web service (www. ebi.ac.uk/chembl/; access date 21 June 2017) was used to find FDA/ EMA-approved drugs with similar 2D structure to C21.

Homology modeling and docking
The crystal structure of hM2 in the active state in complex with the agonist iperoxo [PDB entry 4MQS (26)] was used as a template to create a homology model of active hM4 in MOE 2018.0101 (40). The hM4 sequence used for homology modeling already contained the Y113C and A203G mutations. The default settings were applied, except that iperoxo was considered during model generation and refinement. Iperoxo from the hM2 crystal structure was copied into the hM4 model, and the complex was prepared using the Protein Preparation Wizard (41,42) in Maestro release 2017-2 (43). The ligands were docked into the hM4 model using induced fit docking in Maestro release 2017-2. Redocking was performed with XP settings; otherwise, the default parameters were applied. Figures of the structures and docking poses were created with PyMOL (44).

Behavioral analysis
At P35, mice were trained on a mouse rotarod (Ugo Basile). Mice were initially acclimatized for 10 min on the rotarod turning at 5 rpm, replacing them each time they fell off, followed by acceleration over a 5-min period from 5 to 40 rpm. The sequence was repeated four times. The latency to fall or to three consecutive cartwheels was recorded. Training was repeated daily until every mouse's performance reached a plateau, taking approximately 2 weeks.
On the day of DREADD agonist testing, the mice had a further acclimatization session (5 min at 5 rpm, followed by four accelerations), followed by a break of at least 30 min. They were then tested twice, with the same protocol as the acclimatization session, before, and 20 min after intraperitoneal injection of OZP (0.1 mg/kg). OZP and CZP were dissolved in 0.5% DMSO/0.9% NaCl to a concentration of 0.01 mg/kg before injection. The latency to falling off or cartwheeling was recorded.
To test rotarod performance in rats, animals were initially trained 4 weeks after virus injection, on four consecutive days at a fixed rotation speed of 10 rpm (three trials on each day), and allowed to remain on the rotarod for up to 300 s. All six control rats and five of six rats in the AAV2/5-hCamKII-hM4D(Gi) group learned to remain on the rotarod for 300 s in the final training session. On day 5, rats were administered OZP (0.1 mg/kg, i.p.) and tested on an accelerating rotarod 10 min later (with a start speed of 4 rpm, reaching a final speed of 20 rpm after 30 s). Behavioral tests were performed by a researcher blinded to viral treatment.

Statistical analysis
Statistical analysis was performed with GraphPad Prism 5.01 or IBM SPSS 22.0.0.0. Student's unpaired/paired t test, one-way ANOVA with Bonferroni post hoc test, or repeated-measures ANOVA with LSD post hoc test was used as indicated. Data are shown as mean ± SEM, and the significance level was set to an  of 0.05.

SUPPLEMENTARY MATERIALS
Supplementary material for this article is available at http://advances.sciencemag.org/cgi/ content/full/5/4/eaaw1567/DC1 Fig. S1. Docking poses of OZP and CPX. Table S1. Hit list of the 3D-and 2D-based screens. Table S2. Structures of all tested molecules (note that NQN, which has not been tested, is also shown). Table S3. Ki values for CZP and OZP at different receptors and hM4D(Gi) EC50 (bold and indicated with an arrow).