Functionally biased D2R antagonists: Targeting the Î²-arrestin pathway

Subscriber access provided by - Access paid by the | UCSB Libraries

Article

Functionally biased D2R antagonists: Targeting the #arrestin pathway to improve antipsychotic treatment Michel Weiwer, Qihong Xu, Jennifer P. Gale, Michael Lewis, Arthur J. Campbell, Frederick A. Schroeder, Genevieve C. Van de Bittner, Michelle Walk, Aldo Amaya, Ping Su, Luka #or#evi#, Joshua R. Sacher, Adam Skepner, David Fei, kelly dennehy, Shannon Nguyen, Patrick W. Faloon, José Perez, Jeffrey R. Cottrell, fang liu, michelle palmer, Jen Q. Pan, Jacob M. Hooker, Yan-Ling Zhang, Edward Scolnick, Florence F. Wagner, and Edward B. Holson ACS Chem. Biol., Just Accepted Manuscript • DOI: 10.1021/acschembio.8b00168 • Publication Date (Web): 27 Feb 2018 Downloaded from http://pubs.acs.org on March 1, 2018

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

ACS Chemical Biology is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 19 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Chemical Biology

ACS Paragon Plus Environment

ACS Chemical Biology 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 2 of 19

Functionally biased D2R antagonists: Targeting the β-arrestin pathway to improve antipsychotic treatment Michel Weïwer1#*, Qihong Xu1#, Jennifer P. Gale2, Michael Lewis1, Arthur J. Campbell1, Frederick A. Schroeder3, Genevieve C. Van de Bittner3, Michelle Walk1, Aldo Amaya1, Ping Su4, Luka Dordevic1, Joshua R. Sacher1, Adam Skepner2, David Fei2, Kelly Dennehy1, Shannon Nguyen1, Patrick W. Faloon2, Jose Perez2, Jeffrey R. Cottrell1, Fang Liu4, Michelle Palmer2, Jen Q. Pan1, Jacob M. Hooker3, Yan-Ling Zhang1, Edward Scolnick1, Florence F. Wagner1*, Edward B. Holson1

1

Stanley Center for Psychiatric Research, Broad Institute of MIT and Harvard, Cambridge, MA 02142 (USA) 2

Center for the Development of Therapeutics, Broad Institute of MIT and Harvard, Cambridge, MA 02142 (USA)

3

Athinoula A. Martinos Center for Biomedical Imaging, Department of Radiology, MGH, Charlestown, MA 02129 (USA) 4

Campbell Family Mental Health Research Institute, Centre for Addiction and Mental Health, University of Toronto, Toronto, Ontario M5T1R8 (Canada) #

These authors contributed equally

*

Corresponding authors

Abstract Schizophrenia is a severe neuropsychiatric disease that lacks completely effective and safe therapies. As a polygenic disorder, genetic studies have only started to shed light on its complex etiology. To date, the positive symptoms of schizophrenia are well-managed by antipsychotic drugs, which primarily target the dopamine D2 receptor (D2R). However, these antipsychotics are often accompanied by severe side effects, including motoric symptoms. At D2R, antipsychotic drugs antagonize both G-protein dependent (Gαi/o) signaling and G-protein independent (β-arrestin) signaling. However, the relevant contributions of the distinct D2R signaling pathways to antipsychotic efficacy and on-target side effects (motoric) are still incompletely understood. Recent evidence from mouse genetic and pharmacological studies point to βarrestin signaling as the major driver of antipsychotic efficacy and suggest that a β-arrestin biased D2R antagonist could achieve an additional level of selectivity at D2R, increasing the therapeutic index of next generation antipsychotics. Here, we characterize BRD5814, a highly brain penetrant β-arrestin biased D2R antagonist. BRD5814 demonstrated good target engagement via PET imaging, achieving efficacy in an amphetamine-induced hyperlocomotion mouse model with strongly reduced motoric side effects in a rotarod performance test. This proof of concept study opens the possibility for the development of a new generation of pathway selective antipsychotics at D2R with reduced side effect profiles for the treatment of schizophrenia.

1 ACS Paragon Plus Environment

Page 3 of 19 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Introduction Schizophrenia (SCZ) is a devastating mental illness that affects about 1% of the world’s population and for which there has been limited progress in developing new therapeutics over the last few decades.1 It is marked by three major symptom classes, namely positive (including delusions and hallucinations), negative (including anhedonia, lack of motivation, and poverty of speech), and cognitive symptoms (including memory and attention deficits). Although there are no treatment for the negative and cognitive symptoms of SCZ, the positive symptoms can be managed by currently available antipsychotics. Some patients, however, do not respond well to treatment and develop a wide variety of side effects.2 In the 1980s, clozapine (Clozaril®) was identified as a unique antipsychotic drug, demonstrating unmatched efficacy in otherwise treatment resistant patients.3 Although undoubtedly a D2R antagonist, clozapine modulates a large number of receptors in the brain, resulting in a complex mechanism of action that has not been fully elucidated. Clozapine causes agranulocytosis, a form of serious bone marrow depression, and was originally approved with a requirement of weekly monitoring of blood cell counts. The pharmaceutical industry attempted to develop next generation antipsychotics with efficacy equal to or better than clozapine, but without the bone marrow depression side effects. Based upon the scientific knowledge of the time, it was thought that selectively blocking both D2R and the 5HT2A receptor simultaneously with a single compound would lead to such a drug.4,5 However, both receptors are G-protein coupled 7-transmembrane receptors that are part of an evolutionary related gene set of over 800 proteins in the human genome, making the discovery of selective modulators challenging.6 As a result, all subsequently developed antipsychotics bound to not only D2 and 5HT2A, but several other receptors from this gene family, leading to a wide range of side effects in patients.7,8 Some of the major side effects of antipsychotics, such as weight gain and somnolence, are undoubtedly due to the modulation of receptors irrelevant to their efficacy in psychosis (off-target side effects). Many others, such as increased prolactin and motor symptoms, are likely due to blockade of D2R (on-target side effects). While demonstrating overlapping polypharmacology, antipsychotic drugs all reduce function of the dopamine D2 receptor (D2R), an activity that is believed to be a critical driver of their efficacy.9,10 Recently, a genome wide association study (GWAS) found the DRD2 gene within a schizophreniaassociated locus,11 further emphasizing the potential role of dopamine signaling in disease pathogenesis and treatment options. D2R signaling occurs through G-protein dependent (Gαi/o) and G-protein independent (β-arrestin) mechanisms. D2R/β-arrestin 2 signaling is necessary for dopamine-dependent hyperactivity behaviors in rodents, including those induced by amphetamine (amphetamine induced hyperlocomotion, AIH) and apomorphine (vertical hyperactivity).12,13,14 Caron and coworkers have shown that β-arrestin 2 knock-out (KO) mice have a strongly reduced sensitivity to AIH and apomorphineinduced rearing and climbing. They also showed that antipsychotics have varying pharmacological effects on G-protein dependent signaling but have the common property of antagonizing dopamine mediated interaction of D2R with β-arrestin 2.15 More recently, a protein-protein interaction between D2R and DISC1 (disrupted in schizophrenia 1) was described as a candidate driver of hyperactivity phenotypes in mice and rats, requiring β-arrestin 2 signaling to modulate protein kinase B (Akt) phosphorylation and subsequent GSK3 kinase activation.16 These data indicate that selectively 2 ACS Paragon Plus Environment


Page 4 of 19

antagonizing the D2R/β-arrestin 2 pathway may result in antipsychotic efficacy and, by sparing the Gprotein dependent pathway, may lead to an improved D2R-dependent side effect profile. Whether ontarget side effects are driven by G-protein dependent or independent (β-arrestin) pathways is not known. In this context, the development of compounds with high selectivity for the D2 receptor and functional bias toward only one signaling pathway, would represent an unprecedented level of selectivity, and help delineate the contributions of the different signaling pathways to antipsychotic efficacy and side effects. The recent understanding of both G-protein independent signaling and signaling bias has opened the way for the development of functionally selective ligands of GPCRs, with the hope of maintaining or increasing efficacy while reducing unwanted on-target side effects.17,18,19,20 Trevena, Inc., has pioneered this effort, and several biased ligands of GPCRs have now entered the clinic, showing promising results. While the development of TRV027, a β-arrestin biased agonist of the angiotensin II type 1 receptor (AT1R) was stopped in phase IIb clinical trial for the treatment of acute heart failure (AHF),21,22 TRV130 (oliceridine), a G-protein pathway selective agonist of the µ-opioid receptor, reached phase III development for the management of moderate to severe acute pain,23,24 was designated a breakthrough therapy by the U.S. FDA in February 2016 and a new drug application (NDA) was recently accepted. D2R, as the major target of antipsychotics, has also been considered as a target for biased signaling modulation to achieve efficacy while limiting on-target side effects. Allen et al. recently demonstrated that β-arrestin-biased D2R partial agonists exhibit potent antipsychotic activity in a mouse hyperlocomotion study and the effect is abolished in β-arrestin knockout mice.25 This partial agonist compound acts by normalizing D2R signaling in the presence of high dopamine levels induced by amphetamine treatment, thus partially blocking D2R/β-arrestin signaling, and leading to antipsychotic efficacy. Other D2R biased ligands have been reported recently.26,27,28 MLS1547,29 a G-protein biased agonist of D2R, antagonizes β-arrestin recruitment, albeit with weak potency. More recently, Möller et al. reported a pyrazolo[1,5-a]pyridine as G-protein biased D2R partial agonist and demonstrated an antipsychotic profile in animal models, but noted its tendency to promote motoric side effects.30 To determine if a biased pharmacological blockade of D2R/β-arrestin signaling would mimic the genetic knock-out data reported by Caron12 and translate into a potential therapeutic strategy, we undertook the design and synthesis of a β-arrestin pathway biased D2R antagonist. Our hypothesis was that such a compound could be a safer and better tolerated antipsychotic and possibly demonstrate improved efficacy if G-protein dependent pathway blockade was counterproductive. With this goal in mind, we performed a high throughput compound screen for D2R/β-arrestin signaling, and counter screened active compounds against a G-protein selective assay. To optimize these compounds, we then developed an assay system where G-protein independent (β-arrestin) as well as G-protein dependent (Gi/cAMP) signaling could be evaluated in a single cell line, in both agonist and antagonist modes. Using this unified assay system, the functional profile of compounds at D2R was carefully defined to drive the medicinal chemistry efforts. Compounds prioritized based on functional profile (β-arrestin biased D2R antagonist) were evaluated across a wide range of GPCRs using radioligand binding assays to assess their selectivity and limit off-target activity. BRD5814 was developed as a D2R/β-arrestin biased antagonist with excellent brain penetration and good target engagement. This provided the opportunity to evaluate a biased β-arrestin ligand in vivo. Critically, BRD5814 was efficacious in an AIH mouse model with strongly reduced motoric side effects in a rotarod performance test. Our data suggest that selective


Page 5 of 19 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


and biased inhibition of the D2R/β-arrestin pathway may provide an advantageous therapeutic strategy for the treatment of positive symptoms in schizophrenia.

Results and discussion A high-throughput screen was performed on 57,000 compounds comprising commercial GPCR-biased libraries, bioactives, natural products, and a diversity-oriented synthesis (DOS) library.31,32 The primary assay was developed in C2C12 cells expressing the human long form of D2R (hD2LR) and the human βarrestin 2 (hβarr2), relying on a biosensor technology that utilizes the complementation of a split enzyme β-galactosidase reporter. The hD2R was fused to the inactive alpha mutant β-galactosidase, while hβarr2 was fused to the inactive omega mutant β-galactosidase (see supporting information for details). When β-arrestin is recruited to D2R, these two inactive β-gal mutants form an active enzyme, which is used to measure activation of the receptor. 609 compounds were found to antagonize quinpirole-mediated D2R/β-arrestin 2 interaction at a single dose of 10 µM (~1% hit rate) and were retested in a concentration-response format to determine their EC50s. These 609 compounds were also evaluated in a G-protein dependent assay measuring D2R mediated modulation of cyclic AMP (cAMP) levels (CHO cells, HitHunter® cAMP assay from DiscoverX) in antagonist mode (EC80 quinpirole was used as agonist). 40 compounds were found to have the desired functional profile of antagonizing D2R/βarrestin 2 signaling without antagonizing D2R/cAMP signaling. To eliminate false positives from the primary assay, the hit compounds were further tested in a β-Gal enzymatic assay to exclude compounds directly affecting β-gal enzymatic activity. As an early measure of selectivity for D2R, counter-screens were performed against the dopamine receptors D1R, D3R and D5R, the serotonin receptor 5HT2A, the muscarinic receptor M1, and the κ-opioid receptor (KOR), using cell-based β-gal enzyme fragment complementation assays looking at β-arrestin recruitment as described for the primary assay. These cell-based counter-screens led to the prioritization of five selective D2R hit compounds from two chemical series represented by compounds BRD7640 and BRD7503 (Figure 1A and Table S1 in supporting information). These two hit compounds show micromolar binding affinity for the D2 receptor in a radioligand binding competition assay ([3H]N-methylspiperone, Figure 1B) and are orthosteric binders, as shown by a rightward shift of the quinpirole concentration-response curves with increasing concentrations of compound (Schild plot, Slope ~ 1, Figure 1C). BRD7640 and BRD7503’s selectivity was further evaluated against a panel of 41 G-protein coupled receptors present in the brain through the NIH’s psychotic drug screening program (PDSP).33 A heat map, summarizing the binding assay results, is shown on Figure 1D. Both hit compounds demonstrated a significantly decreased binding promiscuity compared with the known atypical antipsychotics clozapine and aripiprazole. Based on its high affinity for the dopamine transporter (DAT, Ki = 0.062 µM), BRD7503 was deprioritized to avoid potential confounding observations in vivo resulting from competing effects at D2R and DAT. BRD7640 thus emerged as the preferred starting point for developing a β-arrestin biased D2R antagonist.



Page 6 of 19

Figure 1. (A) Chemical structures of BRD7640 and BRD7503. (B) hD2R radioligand competition binding ([3H]N-methylspiperone) concentration curves for BRD7640 and BRD7503. (C) Quinpirole concentrationresponses in the primary D2R/β-arrestin assay in the presence of increasing concentration of BRD7640 (left) and BRD7503 (right). (D) Heat map representing radioligand binding profile across GPCRs present in the brain (data obtained from the NIH Psychotic Drug Screening Program, Full data is available in supporting information Table S2). In order to optimize this validated hit into a potent and selective β-arrestin biased D2R antagonist and carefully define the functional profile of analogs at D2R, we sought to devise a system where G-protein independent (β-arrestin) as well as G-protein dependent (Gi/cAMP) signaling could be evaluated in a single cell line, in both agonist and antagonist modes. We used CHO cells co-expressing the hD2LR fused with the ProLink β-gal peptide, and hβarr2 fused with a β-gal enzyme acceptor fragment from DiscoverX. This cell line allowed us to evaluate compound signaling activity in the β-arrestin and Gi/cAMP pathway in both agonist and antagonist modes in a single cell background (in separate experiments) and offered a greater dynamic range in both assays when compared to the C2C12 cell line. In this unified assay system, BRD7640 is a micromolar antagonist of the D2R/β-arrestin pathway (EC50 2.14 µM, Emax 80%) and a partial agonist of the Gi/cAMP pathway (EC50 0.12 µM, Emax 75%), in contrast to clozapine which fully antagonizes both signaling pathways of D2R (Figure 2). BRD7640 was also tested in these cell-based assays using dopamine (EC80 0.82 µM) as the agonist, to ensure that this chemical series does not show probe dependency (See Figure S1 in supporting information).

Figure 2. Representative curves for the functional profile of clozapine and BRD7640 in CHO cells overexpressing hD2LR. Data is shown in agonist mode (circles, dashed lines) and antagonist mode 5 ACS Paragon Plus Environment

Page 7 of 19 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


(squares, solid lines, quinpirole EC80 used as agonist) for both β-arrestin signaling (left) and Gi/cAMP signaling (right). 1.25 µM Quinpirole was used to define maximum (100%) agonist activity while 1.25 µM risperidone was used to define maximum (100%) antagonist activity.

The individual enantiomers of BRD7640 were evaluated and while both enantiomers were found to be micromolar antagonists of the β-arrestin pathway (0.57 µM and 1.65 µM, respectively), stereochemistry of the compound influenced activity on D2R/Gi-cAMP signaling, with one enantiomeric configuration displaying potent full agonist activity (0.018 µM) while the other enantiomer displayed weaker partial agonist activity (0.19 µM, Emax = 40%). With clear stereochemical preferences on activity, our initial design concepts focused on structure-activity relationships exploring alternative core cyclic amines. The pyrrolidine core of BRD7640 was first replaced by a 3-piperidine core (1, Table 1), leading to an increase in potency along with a dual antagonist profile (antagonist of both signaling pathways). The individual enantiomers of 1 were separated and tested, but were found to be dual antagonists, independent of stereochemistry. With stereochemical inconsistencies between the pyrrolidine and 3-piperidine cores, we decided to explore an achiral 4-piperidine core (2, Table 1). This resulted in a 10-fold increase in potency in the cell based functional assays for both signaling pathways as well as a 10-fold improvement in binding affinity for D2R (Ki 0.16 µM), leading to a potent dual agonist compound. Compound 2 was selected for further medicinal chemistry efforts aimed at maintaining cAMP partial agonist activity while reverting β-arrestin activity to an antagonist. In order to drive the functional activity of the compounds towards a β-arrestin biased antagonist, we systematically evaluated key motifs within the molecule (Table 1 and Table S3). This included removal of the methyl substituent on the 4-piperidine position (S11) or increasing the length of the linker motif extending from the piperidine nitrogen (S12), producing weak activity in both cases (See Table S3 in supporting information). Substitution on the western phenoxy moiety turned out to be critical for βarrestin antagonist activity. Indeed, introduction of increasingly more sterically demanding substituent at the ortho position of the phenyl ring led to a dramatic reduction of agonist efficacy on β-arrestin recruitment while maintaining agonist activity on G-protein dependent cAMP accumulation. This is illustrated by compounds 3-6 and BRD5814, and also highlights the absence of significant electronic effects. It is worth noting that introducing a sterically demanding substituent at the ortho position of the corresponding phenyl ring was also attempted in the 3-pyrrolidine and 3-piperidine series but led to dual antagonist profiles in both cases (See Table S3 in supporting information, S13-15). The unique combination of a 4-piperidine core with a western phenoxy moiety substituted in ortho with a sterically demanding group, such as CF3, was preferred to obtain the desired β-arrestin biased antagonist profile. CF3 substitution at the para position of the western phenyl ring afforded compound 7, with a G-protein biased D2R agonist profile. The m-pyridyl compound 8 was a very potent dual agonist, and further illustrates the steep SAR observed on the western phenyl ring. Compound 9 illustrates that the p-chloro substituent on the eastern phenyl ring of BRD5814 is not necessary for potency and functional signaling profile, but was preferred for metabolic stability (45% (BRD5814) vs 3% (9) remaining after 1 h incubation with human liver microsomes). The western ether linker, the piperidine basic nitrogen, the eastern ether linker and the methyl substituent on the 4-position of the piperidine ring were also found to be preferred for potency and functional signaling profile (See Table S3, S16-23 in supporting information). Although SAR on the eastern phenyl ring did not lead to improvement, it is also necessary



Page 8 of 19

for the desired β-arrestin biased antagonist profile as removal of the p-chlorophenyl group in compound 10 led to a potent dual antagonist. Table 1. Structure activity relationship around the cyclic amine core, western and eastern phenyl rings

Compound

Structure

D2R Ki (µM)a

Antagonist modeb

Agonist modeb

βarr2

cAMP

βarr2

cAMP

Quinpirole

-

1.68

>30

>30

0.071 (100%)

0.005 (100%)

Clozapine

-

0.26

0.45 (100%)

0.22 (100%)

>30

>30

BRD7640

1.66

2.14 (80%)

>30

>30 (15%)

0.12 (75%)

1

0.45

0.73 (100%)

0.067 (100%)

>30

>30

2

0.16

>30

>30

0.12 (75%)

0.01 (100%)

3

0.038

>30

>30

0.009 (70%)

< 0.001 (100%)

4

ND

0.078 (25%)

>30

0.058 (60%)

0.001 (100%)

5

ND

0.91 (100%)

1.36 (50%)

>30

0.50 (40%)

6

ND

0.61 (80%)

>30

0.87 (15%)

0.15 (90%)

BRD5814

0.27

0.54 (85%)

>30

0.60 (15%)

0.061 (80%)

7

1.01

>30

>30

16.3 (45%)

0.13 (90%)

8

0.83

>30

>30

0.013 (100%)

< 0.001 (100%)

9

0.12

0.18 (85%)

>30

>30 (15%)

0.17 (60%)

10

1.04

0.11 (100%)

0.035 (100%)

>30

>30

a

Data was obtained from the psychotic drug screening program (PDSP) using a radioligand binding assay ([3H]N-methylspiperone). bValues are the average of at least three experiments. EC50 are reported in µM and Emax are indicated in parenthesis. Compounds were tested in duplicate in a 9-point dose curve with 3-fold dilution starting at 33.3 µM. 7 ACS Paragon Plus Environment

Page 9 of 19 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


There are currently no publicly available crystallographic structures of D2R in its G-protein or β-arrestin signaling conformations. This lack of structural information makes interpretation of the structureactivity relationship through molecular modeling and rational design of biased ligands difficult.34 Recently, the first crystal structure of D2R was reported in complex with the atypical antipsychotic drug Risperidone (PDB: 6C38, Inactive state) and should prove very useful in understanding the binding mode of antagonist compounds.35 However, understanding the binding of agonist compounds should rely on an active state model of D2R, either bound to its G-protein or β-arrestin. Since BRD5814 is an antagonist of D2R/β-arrestin signaling but an agonist of D2R/Gαi signaling, we explored its putative binding properties in the inactive form of D2R, using the recently reported D2R-Risperidone co-crystal (PDB: 6C38), as well as the active state homology model of D2R, consisting of a ternary complex dopamineD2R-Gαi previously reported by Kling et al.36,37 to infer a potential binding model. In docking experiments using this active state D2R homology model (Figure 3A, D2upR-Gαi, average structure of the dopamine-bound active-state simulation (950-975 ns) of D2R-Gαi, generously provided by the authors),36 BRD5814 was found to bind in the dopamine binding site with its western ethoxyphenyl moiety largely overlapping with dopamine (Figure 3B). Similar to the ammonium group of dopamine (Figure 3B), the protonated piperidine nitrogen of BRD5814 makes a favorable ionic interaction with D1143.32 on transmembrane helix 3 (TM3), while its eastern phenoxy oxygen makes a strong hydrogen bond with T4127.39 (Figure 3B and C). This strong ionic interaction likely serves as a central anchor and allows the projection of the ethoxyphenyl moiety of BRD5814 towards the dopamine binding site. Indeed, the ethoxyphenyl portion of the molecule was found to consistently dock in the orthosteric dopamine binding site and is proposed to mimic the phenethyl motif of dopamine. The oCF3-phenyl moiety of BRD5814 was found to overlap with the catechol moiety of dopamine, binding in the proximity of residues Ser1935.42, Ser1945.43, Ser1975.46 on TM5 and His3936.55 on TM6. Ser1935.42 and Ser1975.46 are critical to receptor activation by dopamine, through H-bonding interactions with the catechol hydroxyl groups. Although BRD5814 projects its o-CF3-phenyl ring deep into the hydrophobic dopamine binding pocket formed by V1153.33, F3896.51, F3906.52, C1183.36 and H3936.55, it does not directly engage Ser1935.42, Ser1945.43 and Ser1975.46. These interactions formed by BRD5814 differ from dopamine and could partially explain its biased signaling property. His3936.55 has been reported as a major determinant of biased signaling in D2R, resulting from its capacity to rotate toward S1945.43 on TM5 or away from it and toward Y4087.35 on TM7 (Figure 3C).36,38 This is consistent with the steep SAR observed around the o-CF3-phenyl moiety of BRD5814, both on potency and functional signaling. Specifically, the o-CF3 group of BRD5814 is seated between F3896.51 and H3936.55 and it is reasonable to hypothesize that this hydrophobic interaction sterically stabilizes H3936.55 in a configuration that locks it toward S1945.43 on TM5, preventing its rotation towards Y4087.35 on TM7 (Figure 3C and D). It is tempting to speculate that prevention of this movement by the o-CF3 substituent favors one active conformation (G-protein signaling) over the other (hypothesized as β-arrestin signaling), driving the βarrestin biased antagonist profile of BRD5814. This is consistent with the observation that sterically demanding substituents on the ortho position of the western phenyl ring (Table 1, compounds 4-6 and BRD5814) strongly diminish D2R/β-arrestin signaling without affecting D2R/Gαi signaling while compounds with smaller substituents (Table 1, compounds 2, 3, 7 and 8) were agonists of both signaling pathways. Absence of the CF3 group would permit movement of H3936.55 towards S1945.43 on TM5 or Y4087.35 on TM7 and lead to an agonist activity on both D2R/Gαi and D2R/β-arrestin signaling.



Page 10 of 19

In the inactive form of D2R (PDB: 6C38), docking models show that BRD5814 adopts an inverted binding pose, compared to what is observed in the active state model, but conserves a favorable ionic interaction between its piperidine nitrogen and D1143.32 on TM3 (Figure 3E). Interpretation of this result remains elusive as BRD5814 stabilizes a receptor conformation that allows Gαi signaling while preventing β-arrestin signaling, which is likely to be significantly different from the purely inactive state of D2R in 6C38. All together, our computational models provide a potential rationale, supported by the SAR, to explain the functional profile of our biased ligand BRD5814.

Figure 3. (A) Active state dopamine bound D2R-Gαi homology model reported by Kling et al.36 D2R is shown as grey ribbon, Gαi is shown as cyan ribbon, and dopamine is shown as magenta sticks. (B) Topview of the orthosteric site of the D2R-Gαi homology model bound to BRD5814 (yellow sticks) and dopamine (magenta sticks). The ammonium group of dopamine is anchored by an ionic interaction with D1143.32, while both hydroxyl groups hydrogen bond with S1935.42 and S1975.46. Protons as well as part of the receptor have been omitted for clarity. (C) Zoomed top-view of the orthosteric site of the D2R-Gαi homology model bound to BRD5814. The protonated piperidine nitrogen of BRD5814 makes an ionic interaction with D1143.32 and projects the hydrophobic o-CF3-phenoxyethyl moiety deep into the highly hydrophobic dopamine binding pocket formed by V1153.33, F3896.51, F3906.52, C1183.36 and H3936.55. BRD5814 CF3 substituent seats between H3936.55 and F3896.51 (green spheres) and prevents the rotation of H3936.55 towards Y4087.35 on TM7. (D) Similar view with BRD5814 represented as yellow spheres, highlighting the steric effect of its o-CF3 substituent on the rotation movement of H3936.55. (E) D2RRisperidone co-crystal (6C38). BRD5814 (yellow) docked in 6C38 adopt an inverted binding pose and overlap with Risperidone (Magenta). The full D2R functional profile of BRD5814 is shown in Figure 4A. BRD5814 demonstrates potent antagonist activity versus the β-arrestin pathway and no antagonist activity versus the Gi/cAMP pathway. To further confirm BRD5814 antagonist activity on D2R/β-arrestin signaling, modulation of pGSK3α/β was measured in HEK293T cells co-expressing D2R and DISC1 as previously reported.16 Upon D2R activation by an agonist, β-arrestin is recruited and serves as a scaffolding protein to bring together protein phosphatase 2 (PP2A) and protein kinase B (Akt) as a complex, which leads to Akt dephosphorylation and subsequent deactivation. This, in turn leads to a decrease in phosphorylated 9 ACS Paragon Plus Environment

Page 11 of 19 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


GSK3α/β (p-GSK3α/β (Ser21/9)), a direct substrate of Akt, which can be measured and quantified by western blot. As expected, treatment with the non-biased D2 agonist quinpirole (10 µM) led to a decrease in GSK3α/β (Ser 21/9) phosphorylation, while pre-treatment with the non-biased D2 antagonist haloperidol (10 µM) completely abolished quinpirole’s effect (Figure 4B and C). Similar to haloperidol, BRD5814 dose-dependently blocked quinpirole’s effect and fully reestablished GSK3 phosphorylation status at 10 µM. This orthogonal functional assay demonstrates that the biased antagonist BRD5814 can effectively block D2R/β-arrestin dependent GSK3 signaling cascade. These results are consistent with the Gi/cAMP independence of GSK3 function and demonstrate, for the first time, that biased ligands can effectively modulate this downstream functional response. With a β-arrestin biased D2R antagonist in hand and having shown antagonist activity in both direct βarrestin recruitment as well as downstream GSK3 signaling, we turned to radioligand binding assays to evaluate the selectivity of our biased ligands across a panel of GPCRs present in the brain. This is important to anticipate off-target toxicities as well as potential confounding activities. Because we included selectivity across GPCRs present in the brain as part of our counter-screen and prioritized our hit compound BRD7640 based on its relatively clean binding profile (see Figure 1D), we anticipated to obtain relatively selective D2R lead compounds. Indeed, BRD5814 shows good selectivity for the D2-D3 receptors and a significantly cleaner binding profile than clozapine and aripiprazole (Figure 4D and Table S2 in supporting information). Progression from the hit compound BRD7640 to the lead BRD5814 led to an increase in binding affinity for the D2-D3 receptors while decreasing binding affinity for all other receptors, with the exception of 5HT2B/C and Alpha 1A. Although strong binding of 5HT2B was observed, BRD5814 was found to function as an antagonist of 5HT2B (Table S4). This is important as 5HT2B agonist activity has been linked to severe valvulopathy.39 In order to move toward in-vivo studies and evaluate the potential of BRD5814 to recapitulate the genetic KO data reported by Caron, its pharmacokinetic properties were evaluated (see Table S4). Sufficient aqueous solubility (22 µM), plasma stability (100% in human, mouse, and rat), and liver microsomes stability (45%, 25%, and 2% remaining after 1h incubation in human, mouse, and rat, respectively), as well as limited hERG inhibition (IC50 > 10 µM), encouraged us to evaluate BRD5814 for brain penetration. Permeability assessed using the MDR1-MDCKII cell line, which overexpresses Pglycoprotein (P-gp), one of the major efflux pump present at the blood-brain barrier, proved BRD5814 is not a P-gp substrate (efflux ratio = 1.1) and led us to anticipate good brain penetration. Brain pharmacokinetic studies in mice and rats established the outstanding brain penetration of BRD5814, with AUC brain/plasma ratios of 7.7 and 11.0, and half-life in the brain of 4.2 h and 4.8 h, respectively (Figure 5A and B). The surprisingly long half-life in the brain and plasma, considering the relatively low liver microsomes stability in rodents, can be explained by the high protein binding of BRD5814 in both plasma (99.1% (h), 100% (m), 99.4% (r)) and brain (99.9% (m)) but could be a disadvantage when considering target engagement. In vivo target engagement data in specific tissue is critical to properly assess pharmacological rescue of observed phenotypes and is often lacking in pharmacological studies. We thus evaluated BRD5814 target engagement at D2R through positron emission tomography (PET) studies using [11C]-raclopride as radiotracer, a highly selective PET ligand for the D2-D3 receptors. Rats were pre-treated intravenously (i.v.) with ascending doses of BRD5814 five minutes before administration of [11C]-raclopride, and tracer uptake was measured over an hour. Dose dependent blockade of [11C]-raclopride was observed, with a promising 50% D2R occupancy at a dose of 8 mg/kg BRD5814, which is in line with what is observed 10 ACS Paragon Plus Environment


Page 12 of 19

with a therapeutically relevant dose of clozapine (1-5 mg/kg, Figure 5C). The sustained presence of BRD5814 in the rat brain demonstrated via brain PK also led to a sustained target engagement, as shown by the 58% D2R occupancy when rats were pre-treated 1 hour before [11C]-raclopride administration. PET studies not only allowed us to demonstrate sustained target engagement in the rat brain, but also allowed us to estimate a potential efficacious dose in subsequent mouse behavioral studies, to achieve >50% D2R occupancy, which is required to observe antipsychotic efficacy in humans and rodents.40

Figure 4. (A) Representative curves showing the functional profile of BRD5814 in CHO cells overexpressing the human D2 receptor. Data is shown in agonist mode (circles) and antagonist mode (triangles, quinpirole EC80 was used as agonist) for both Gi/cAMP signaling (green) and β-arrestin signaling (blue). 1.25 µM Quinpirole was used to define maximum (100%) agonist activity while 1.25 µM risperidone was used to define maximum (100%) antagonist activity. (B) Western blot analysis of GSK3α and GSK3β phosphorylation status in HEK293T cells expressing D2R and DISC1 following compound treatment (30 min). (C) Densitometric analysis of phosphorylated GSK3β (Ser9, left) and GSK3α (Ser21, right). The level of phosphorylated GSK3α/β were normalized to total GSK3α/β. Results for each sample are presented as the percentage of the control sample (control). * p

Functionally biased D2R antagonists: Targeting the Î²-arrestin pathway

Recommend Documents