Design and discovery of a high affinity, selective and β-arrestin biased 5-HT7 receptor agonist

Compound 1c, 5-chloro-2-(2-(3,4-dihydroisoquinolin-2(1H)-yl)ethyl)-2,3-dihydro-1H-inden-1-one was previously reported from our laboratory showing high affinity binding to the 5-HT7 receptor (Ki = 0.5 nM). However, compound 1c racemizes readily upon enantiomeric separation. To prevent racemization, we have redesigned and synthesized methyl and carboxyethyl analogs, compounds 2 and 3 respectively, whose binding affinities were similar to those of compound 1c. Compounds 2 and 3 cannot undergo racemization since tautomerism was no longer possible and thus, compound 2 was selected for enantiomeric separation and further evaluation. Upon enantiomeric separation, the levorotatory enantiomer, (−)2 or 2a demonstrated a higher affinity (Ki = 1.2 nM) than the (+)2 or 2b enantiomer (Ki = 93 nM) and a β-arrestin biased functional selectivity for the 5-HT7 receptor. Although 2a showed about 8 times less activity than 5-HT in the Gs pathway, it showed over 31 times higher activity than 5-HT in the β-arrestin pathway. This constitutes a significant β-arrestin pathway preference and shows 2a to be more potent and more efficacious than the recently published β-arrestin biased 3-(4-chlorophenyl)-1,4,5,6,7,8-hexahydropyrazolo[3,4-d]azepine, the N-debenzylated analog of JNJ18038683 (Compound 7).


Introduction
The serotonin-7 receptor (5-HT 7 R), the newest discovery of the 5-HTRs, is discretely localized in the thalamus, hypothalamus, cortical and limbic regions of the Central Nervous System (CNS) as well as in the periphery [1,2]. The pharmacology and therapeutic potential of the 5-HT 7 receptor and its involvement in several pathophysiological conditions have recently been reviewed [3]. Furthermore, several lines of evidence have implicated the role of 5-HT 7 R in neuropsychiatric diseases including depression, pain, migraine, schizophrenia, and disorders related to circadian rhythm [4,5]. Studies have also shown that blockade or inactivation of the 5-HT 7 R has led to antidepressant-like behavior [6]. These observations have led to the hypothesis that the 5-HT 7 R could serve as new target for the design of drugs for the treatment of depression and other CNS diseases [7]. SB-269970, JNJ18038683, and SB-258719 are 5-HT 7 R antagonists ( Fig. 1) that have served as primary tools for the pharmacological characterization of the 5-HT 7 R [8,9]. In addition, the selective 5-HT 7 R agonists AS-19 [10], E-55888, LP-12, LP-44, and Compound 7 ( Fig. 1) have been used to characterize the involvement of the 5-HT 7 R in CNS conditions including memory and cognition [11]. More recently, it has been suggested that 5-HT 7 R agonists may be useful in the treatment of alcohol and drug abuse [12]. Furthermore, there have been efforts to identify new agents that not only show selectivity for the 5-HT 7 R, but also display signaling bias to allow for further characterization of the 5-HT 7 R [13,18]. In a previous study, we reported a campaign to identify new, high affinity 5-HT 7 R ligands that might find utility in treating some of the cognitive symptoms of schizophrenia and especially identifying new ligands that display a bias for the signaling pathways at the 5-HT 7 R [14,19]. This was in response to recent research indicating noncanonical modes of G protein-coupled receptor (GPCR) signaling through β-arrestins, which trigger specific signaling pathways that are independent of G protein activation [15]. In addition, there have been discovery of ligands that block G protein activation but promoted β-arrestin binding, or vice-versa, suggesting the possibility of selectively activating intracellular signaling pathways [15]. Biased agonism is a relatively new concept which postulates that certain GPCRs can activate complex signaling networks to adopt multiple active conformations upon agonist binding [16], and in fact, biased agonists selectively stabilize only a subset of receptor conformations induced by a ligand. This thus, suggest the possibility that one can direct cellular signaling with precision and specificity and further support the notion that biased agonists may produce new classes of therapeutic agents that have fewer side effects [17]. Indeed, a 5-HT 7 R targeting β-arrestin biased ligand was reported to modulate NREM (non-rapid eye movement) and REM (rapid eye movement) sleep times indicating the potential of 5-HT 7 R biased ligands for the treatment of certain sleep disorders [18].

Results and discussion
Chemistry To obtain compound 2, the commercially available 5-chloroindanone 4a was treated with sodium hydride (NaH) and then diethyl carbonate to produce a β-keto ester 4b as reported in Scheme 1. Compound 4b was reacted with NaH and iodomethane to deliver a methyl group at the α-carbon of the β-keto ester, 4c. The methylated ester 4c was hydrolyzed and decarboxylated using acetic acid and concentrated HCl in a microwave assisted synthesis (MWAS) to afford ketone 4d. The ketone 4d was reacted with NaH and 1-bromo-2-chloro ethane to produce the alkylating agent 4e which was then coupled to 1,2,3,4tetrahydro-isoquinoline (THIQ) under a general alkylating condition to afford the final product, 2 as a racemate. The free base was converted to the oxalate salt and the NMR and elemental (CHN) analysis confirmed the product as the desired compound.
Compound 2 satisfies all the criteria for Lipinski's rule of five (Ro5) [15] except for the ClogP value of 5.2. Given the relatively high CLogP value, an additional analog was designed to reduce the CLogP value by replacing the lipophilic methyl group with a carbethoxy group to form compound 3 (CLogP = 4.42), thus ensuring that this analog fully satisfy the Ro5.

Enantiomeric separation
The enantiomers of 2 were separated using high performance liquid chromatography (HPLC) with a semi-prep scale CHIRALPAK AD-H as reported previously [16]. Reinjection of the separated enantiomers on an analytical scale CHIRALPAK AD-H HPLC column indicated 100% enantiomeric purity for both (−)2 (2a), and (+)2 (2b). No racemization was subsequently observed, and the enantiomeric purity remained at 100%. Evaluation of the enantiomers in a Jasco P-1020 polarimeter at an average cell temperature of 23

Biological evaluation
Having obtained the desired target compound, 2 and its enantiomers (−)2 or 2a and (+)2 or 2b, and 3, the compounds were sent to the Psychoactive Drug Screening Program (PDSP) to screen for binding at relevant CNS receptors. The results are reported in Tables 1 and 2. Compared to the standard drug clozapine, 2 binds with a 2-fold higher affinity at the 5-HT 7 R (Ki = 2.2 nM) and about 4-fold lower affinity than AS 19, a standard 5-HT 7 R agonist. In addition, compound 2 displays stereoselectivity with (−)2 displaying approximately twice the affinity as racemic 2 and >75 times higher affinity than (+)2 at the 5-HT 7 R. Furthermore, at the 5-HT 1A receptor, (−)2 binds with about 20fold affinity than racemic 2 (Table 1). Compound 2 also shows selectivity for the 5-HT 7 R when compared to other 5-HTRs, Dopamine (DA) receptors and the three monoamine transporters (SERT, DAT, and NET) evaluated (Tables 1  and 2). The eutomer or the (−) enantiomer, 2a, was selective by over 100-fold for all the receptors and transporters evaluated except for the 5-HT 1A R where the selectivity is greater by almost 72-fold (Tables 1 and 2). It is further noted that while the (+) enantiomer or the distomer (2b) has 25-fold less affinity for the 5-HT 7 R than the eutomer, it has in fact a higher affinity for the 5-HT 2B R (Ki = 59 nM), than for the 5-HT 7 R (Ki = 93 nM) ( Table 1). MT Missed threshold of 50% inhibition at 10 µM concentration, NR Not reported, NS Not significant a The binding affinity results are reported as mean ± SEM from a minimum of 3 independent assays, each in triplicates b Reported in ref. [14] c Reported in ref. [10] as a partial agonist with selectivity >100-fold at all 5-HTRs except for 5-HT 1D R (Ki = 6.6 nM) d Reported in ref. [18] as compound 1g MT Missed threshold of 50% inhibition at 10 µM concentration a The binding affinity results are reported as mean ± SEM from a minimum of 3 independent assays, each in triplicates

Molecular modeling
To evaluate interaction of compounds with the 5-HT 7 R [13], we selected a homology model of the 5-HT 7 R published by Kim et al. [18] and the Induced Fit Docking (IFD) procedure was carried out on it, using default parameters.
The model was first validated by docking AS-19 into it and replicating the appropriate interactions of the amino acid residues at the orthosteric site (Fig. 3) as reported [18]. To explore the difference in binding affinities and to predict which enantiomer 2S or 2R has interactions consistent with the binding affinities and β-arrestin recruitment, we evaluated the docking poses of 2S and 2R enantiomers in the homology model as shown in Fig. 4. Evaluation of the interactions of 2S at the orthosteric site of the 5-HT 7 R revealed interactions with the key amino acid residues associated with β-arrestin biased signaling [18]. Thus, 2S has an ionic interaction with Asp 162, hydrogen bond interaction with Ile 233 through the oxygen of the carbonyl group but no interaction with Ser 243; these results are consistent with the interactions associated with β-arrestin recruitment as reported [18]. On the other hand, 2R had interaction with Asp 162, but no interaction with Ile 233 or Ser 243, leading to its prediction to have little or no β-arrestin recruitment capabilities.

Functional studies
The enantiomers of 2, i.e., 2a or the (−) enantiomer, and 2b, the (+) enantiomer were selected for evaluation for functional selectivity at the 5-HT 7 R. The ligands were tested using both Gs-mediated cAMP production assay and β-arrestin recruitment Tango assays in parallel and the results are reported in Table 3.
The results show that the (−) enantiomer 2a activates both G-protein and β-arrestin mediated signaling but acts more so as a β-arrestin-biased agonist at the 5-HT 7 R (EC 50 = 3.98 nM; Emax = 78.7%). In fact, 2a has a higher binding affinity (Ki = 1.2 nM), is much more potent and more efficacious at recruiting β-arrestin to the 5-HT 7 R than Compound 7 (Ki = 30 nM; EC 50 = 162 nM; Emax = 62.8%), recently disclosed in the literature (Table 3) [18]. It is interesting to further note that 2a showed 8 times less activity than 5-HT for the Gs pathway, while it is over 31 times more potent than 5-HT in the β-arrestin pathway. The (+) enantiomer, 2b on the other hand, has over 30-fold weaker binding affinity at the 5-HT 7 R and displays less activity in recruiting β-arrestin to the 5-HT 7 R by over 39fold (Table 3). Taken together, these results will predict (−) 2 or 2a to correspond to the 2S enantiomer while (+) 2 or 2b will correspond to the 2R enantiomer.
As indicated earlier, compound 3, with the carbethoxy replacement for the alpha methyl group of compound 2, was designed to improve hydrophilicity and comply with Lipinski's Ro5 [15]. Compound 3 improved CLogP by 0.80 units to 4.42 and thus, satisfy Ro5 and was synthesized for screening. As shown in Table 1, the presence of the carbethoxy group in place of the methyl group did not weaken binding affinity to the 5-HT 7 R and in fact compound 3 has a similar binding affinity as compound 2 (Ki = 3.4 nM) and Fig. 3 Validation of the homology model of 5-HT 7 R obtained from Kim et al. [18]. Compound AS-19 was docked into the model using induced-fit docking as implemented in Schrodinger's Small-Molecule Drug Discovery suite [17,[27][28][29][30]. The observed interactions with the key amino acid residues Asp 162 and Ile 233 and not Ser 243 are confirmatory Fig. 4 (Left) and (right). Induced-fit molecular docking mode of 2S and 2R-enantiomers respectively at the orthosteric site of 5-HT 7 R. In (left), the 2S enantiomer interacts with Asp 162 and Ile 233 but not Ser 243, suggesting the capacity to recruit β-arrestin to the 5-HT 7 R. In (right), the 2R enantiomer could only make an interaction with Asp 162 but not Ile 233 or Ser 243 even slightly better than that of clozapine (Ki = 5.0 nM) ( Table 1).
In summary, 5-chloro-2-(2-(3,4-dihydroisoquinolin-2 (1H)-yl)ethyl)-2-methyl-2,3-dihydro-1H-inden-1-one (2) was designed and synthesized to block the racemization of the previously reported alpha des-methyl analog, compound 1c. Following enantiomeric separation, 2a or (−) 2 has been shown not only to have retained binding affinity to the 5-HT 7 R but it has demonstrated significant β-arrestin biased functional selectivity at the receptor. In fact, while 2a showed 8 times less activity than 5-HT, for the Gs pathway, it showed over 31 times higher activity than 5-HT in the β-arrestin pathway. This β-arrestin pathway activity is more potent and more efficacious (2a; EC 50 = 3.98 nM; Emax = 78.7%) than that of the recently published β-arrestin biased Compound 7 (EC 50 = 162 nM; Emax = 62.7%) ( Table 3). The molecular modeling results revealing interactions of 2a with appropriate amino acid residues at the orthosteric site of 5-HT 7 R suggest that the eutomer 2a corresponds to the 2S isomer. To improve the drug-like characteristics of 2, an additional analog, 3 with improved hydrophilicity and yet with similar binding affinities as 2 has been synthesized and reported in the present work.

Materials and methods
Melting points were determined on a Gallenkamp (UK) apparatus and are uncorrected. All NMR spectra were obtained on a Varian 300 MHz Mercury Spectrometer and the free induction decay data were processed using Mestrelab's Mnova NMR software (version 8.1) to obtain the reported NMR data. Elemental analyses were carried out by Atlantic Microlab, Inc., Norcross, GA, and are within 0.4% of theory unless otherwise noted. Flash chromatography was performed using a Teledyne CombiFlash ® with Davisil grade 634 silica gel. Starting materials and solvents were obtained from Sigma Aldrich and were used without further purification. All microwave assisted syntheses (MWAS) were carried out using a Biotage Initiator ® equipment.

Synthesis of 5-chloro-2-(2-(3,4-dihydroisoquinolin-2(1H)-yl) ethyl)-2-methyl-2,3-dihydro-1H-inden-1-one, 2
Synthesis of ethyl 5-chloro-1-oxo-2,3-dihydro-1H-indene-2carboxylate, 4b A modified method described by Paccani and co-workers [20] was followed to access the 5-substituted β-keto ester 4b. (Scheme 1). In brief, a solution of 5-chloroindanone 4a, (100 mmol) in diethyl carbonate (50 mL) was added dropwise to a stirred suspension of sodium hydride (NaH) (200 mmol, 60% in mineral oil previously washed with hexanes) in diethyl carbonate (DEC) (25 mL) at 0°C with stirring (note green coloration). When evolution of gas has ceased, the mixture was allowed to stir at room temperature overnight then diluted with dichloromethane (CH 2 Cl 2 ) and treated with aqueous acetic acid solution. The aqueous phase was separated and extracted with CH 2 Cl 2 . The combined organic extracts was dried with anhydrous sodium sulfate (Na 2 SO 4 ) and concentrated under reduced pressure to yield a thick brown crude oil. The crude oil was loaded onto a cartridge and purified on a Combiflash chromatographic equipment using a gradient elution (up to 10% EtOAc in hexanes) to afford ethyl 5-chloro-1-oxo-2,3-dihydro-1H-indene-2-carboxylate, 4b as needle-like white crystals. Yield: 89%. 1   NR Not reported, ND Experiment was conducted but no antagonism was measured, NA Experiment was not conducted a Enantiomers (2a and 2b) were tested in both Gs-cAMP and β-arrestin recruitment (Tango) assays and the results are reported as mean ± SEM from a minimum of 3 independent assays, each in triplicates b Data was obtained from ref. [18] 127 [This compound tautomerizes so the number of H and C were in some cases double. The presented data however, is consistent with the reported values cited in ref. [20].
A mixture of alkylating agent, 5a (1.66 mmol), THIQ (1.82 mmol) NaHCO 3 (1.82 mmol), and KI (catalytic) in DME or Toluene (10 mL) was placed in a 20 mL microwave vial with a stirrer and tightly sealed. The mixture was subjected to microwave heating at 120°C for 60 min, allowed to cool to room temperature and the mixture was directly purified on silica gel by flash chromatography (hexanes: EtOAc gradient up to 70% EtOAc) to afford the compound as a free base.

Receptor binding affinity studies
Binding affinities reported in were conducted by the National Institute of Mental Health Psychoactive Drug Screening Program (NIMH-PDSP). Details of the methods and radioligands used for the binding assays were previously reported [21], and a list of receptors and radioligands used in the binding affinity evaluations are reported in Table 4.

GPCR Tango assays: G-protein independent β-arrestin recruitment
Recruitment of β-arrestin to agonist stimulated 5-HT7R were performed using the "Tango"-type assay described in Barnea et al. [22] with modifications, to evaluate 2a and 2b. GPCR Tango assay for measuring G-protein independent β-arrestin recruitment was carried out by the NIMH-PDSP as per their protocol with some modifications in the Tango assay system reported previously [21,22]. GPCR Tango constructs were codon optimized for better expression in mammalian cell lines and total synthesis was by Blue Heron Biotech (Bothell, WA) with independent sequencing confirmation [23]. HTLA cells (an HEK293 cell line stably expressing a tTA-dependent luciferase reporter and a ß-arrestin2-TEV fusion gene) were maintained in DMEM supplemented with 10% FBS and 2 μg/ml Puromycin and 100 μg/ml Hygromycin. The FLAG tag was designed into the GPCR Tango constructs for confirmation of surface expression and comparison of expression levels. HTLA cells were transfected using calcium phosphate transfection method [24][25][26] with GPCR tango constructs. Briefly, HTLA cells were sub-cultured into 10-cm dishes (3 × 10 6 cells per dish) and incubated overnight. For transfection in each 10-cm dish, 10 μg receptor DNA construct in 440 μl distilled water was mixed with 60 μl of 2 M CaCl 2 and the DNA/CaCl 2 solution was then added dropwise into 500 μl 2x HBS solution (50 mM HEPES, 280 mM NaCl, 10 mM KCl, 1.5 mM Na2HPO4, pH 7.00) while shaking. The mixture was incubated at room temperature for 10 min, then added to the cells dropwise and incubated overnight. Transfected HTLA cells were plated in DMEM supplemented with 1% dialyzed FBS in Poly-L-Lys-coated 384well white clear bottom cell culture plates at a density of 15,000 cells per well (40 μl). Cells were incubated for a minimum of 6 h or overnight before drug simulation treatments. Drug stimulation solutions prepared in sterile-filtered Tango assay buffer at 5× concentration were added to cells at 10 μl per well and incubated overnight. Antagonist activity was measured by pre-incubation of drug solutions (6× of the final concentration) with cells for 30 min before addition of 10 μl of a final EC80 concentration of reference agonist. The EC80 concentration was determined in separate preliminary dose-response assays. On the day of measurement, medium and drug solutions were removed and BrightGlo reagent (Promega), 20 μl per well (diluted 20-fold with Tango assay buffer) were added. Plates were incubated for 20 min at room temperature in the dark followed by chemiluminescence measurement. The chemiluminescence data was recorded using Luminesence counter and reported as relative luminescence units.

HEK-5-HT 7A /cAMP functional assay
Human embryonic kidney cells were transfected with 10 μg h5-HT 7A cDNA using PEI in unsupplemented DMEM. After 20 min, the compound was added in a final volume of 1 ml, and incubated for 20 min. For antagonists, 0.8 ml EBSS is added, cells were incubated for 10 min, the compound was added, and incubated for 10 min after which serotonin (100 nM) was added. For all conditions, after 20 min incubation with agonist, the reaction was terminated by aspiration of the buffer, and 0.1 ml trichloroacetic acid was added. Plates were incubated for 2 h on a rotator. Adenylate cyclase activity was measured using a cyclic AMP EIA kit (Cayman). Aliquots (40 µl) of each well were diluted to 200 μl with EIA buffer from the kit, and 50 µl of the dilution was added to the EIA plate. After addition of tracer and monoclonal antibody, the EIA plates were incubated for 18 h at 4°C. The reaction was aspirated, plates were washed 5 × 300 µl with wash buffer, and Ellman's reagent was added. After a 2-h incubation in the dark on a rotator, the plates were read at 410 nm. Basal cAMP is subtracted from all values. 5-HT 7 agonists stimulate cAMP formation, maximal stimulation was defined with 10 µM serotonin. The maximal drug effect is normalized to maximal serotonin effect in the tables.

Molecular modeling
The target compounds AS-19, 2a and 2b were optimized using the LigPrep tool and energy minimized using the OPLS3e force field in Schrodinger's Small-Molecule Drug discovery suite [17,[27][28][29][30]. The individual compounds in their protonated forms were docked into the homology model of the 5-HT 7 R using the IFD workflow and the results depicted in Figs. 3, 4.