Identification of false positives in “HTS hits to lead”: The application of Bayesian models in HTS triage to rapidly deliver a series of selective TRPV4 antagonists

Sarah E. Skerratt; James E. J. Mills; Jayesh Mistry

doi:10.1039/C2MD20259J

View PDF VersionPrevious ArticleNext Article

DOI: 10.1039/C2MD20259J (Concise Article) Med. Chem. Commun., 2013, 4, 244-251

Show CompoundsShow Chemical Terms

Identification of false positives in “HTS hits to lead”: The application of Bayesian models in HTS triage to rapidly deliver a series of selective TRPV4 antagonists

Sarah E. Skerratt *^a, James E. J. Mills ^b and Jayesh Mistry ^c
^aPfizer Neusentis, The Portway Building, Granta Park, Cambridge, UK. E-mail: sarah.skerratt@pfizer.com; Tel: +44 (0)304644073
^b45 Queen St., Deal, Kent, UK. E-mail: james.mills@sandexis.co.uk; Tel: +44 (0)771 541 5402
^cElsevier Limited (Corporate Office), 32 Jamestown Road, Camden, London, UK. E-mail: j.mistry@elsevier.com; Tel: +44 (0)7584263708

Received 29th August 2012 , Accepted 22nd October 2012

First published on 23rd October 2012

Abstract

Herein, we describe the discovery and optimisation of a series of potent and selective TRPV4 antagonists. The application of a variety of computational techniques (including Bayesian modelling) at the HTS triage stage enabled the early deprioritisation of likely frequent hitters. The use of methods to positively prioritise compounds for follow-up screening allowed the rapid identification of a number of interesting TRPV4 antagonist series. The hit-to-lead efforts in one such series, the hydroxypiperidines, will be described.

Introduction

TRPV4 is a member of the Transient Receptor Potential (TRP) superfamily of mammalian cation channels. It is expressed in a wide range of tissues¹ and is thought to play a role in a number of physiological responses such as osmoregulation,^2,3 thermosensation,⁴ vascular regulation^5,6 and mechanosensation.^2,7 Indeed, it has been postulated that TRPV4 plays a crucial part in the mechanosensory pathway of the bladder by detecting changes in intravesical pressure.^8,9 TRPV4 is activated by a range of stimuli including small molecules (e.g. anandamide and 5,6-epoxyeicosatrienoic acid), temperature (27–34 °C), hypotonic solutions and mechanical stimuli.^10,11

At the outset of the TRPV4 project, there were no known selective TRPV4 antagonists with which to build confidence in mechanism. A high through-put screening (HTS) campaign was therefore initiated on the Pfizer compound file and, in order to meet the through-put requirements of this campaign, a TRPV4 calcium flux assay using FLIPR Tetra technology was utilised.

Results and discussion

HTS screening and triage

Compressed file screening generated a large number of hits (∼2.5% of compounds had >40% inhibition of TRPV4 at 10 μM) for confirmatory follow-up in a single-well, single-point TRPV4 assay. The single-point assay confirmed ∼11 k of these to be active in this assay (>40% inhibition at 10 μM), a 14% confirmation rate. It is, however, widely recognised that calcium flux assays can deliver false positives or ‘frequent hitters’ due to a compound being reactive, autofluorescent, cytotoxic or active at another target in the signal transduction pathway.^12,13 A key goal of the HTS triage effort was therefore to identify and remove false positives at the outset, allowing efforts to be focussed on true TRPV4 antagonists.

The data from a number of prior Pfizer FLIPR HTS screening campaigns were used as a means of identifying likely false positives from the ∼11 K putative actives emerging from the TRPV4 single-point screen. Firstly, if a compound was active (>75% inhibition at 10 μM) in more than one of these HTS screens, it was deemed likely to be a false positive (provided the prior HTS was not directed at a target from the same gene family as TRPV4). Secondly, for each previous FLIPR HTS dataset, a Bayesian activity model¹⁴ was built to predict the likelihood of activity of a compound in that assay. If a compound was scored highly by more than one such model (Bayesian score >5), it was deemed likely to be a false positive. This second in silico approach offers the advantage that it does not require the compound itself to have been screened in HTS assays, because evidence of activity of its near neighbours is sufficient to suggest it could be a false positive. Indeed, this in silico method filtered out ∼5 times as many compounds as the method based on in vitro data. An example of use of these Bayesian specificity models is shown in Fig. 1. Putative actives from the TRPV4 single point assay were trellised by molecular cluster, and % TRPV4 inhibition was plotted against molecular weight. Compounds were coloured by the number of times they were deemed “active” (i.e. scored >5) by any of the non-TRP FLIPR HTS Bayesian models (red = active according to ≥2 models, yellow = active according to 1 model, green = active according to 0 models). Compounds such as those found in Cluster 4 were retained, whereas compounds such as those found in Cluster 3 were discarded. A selection of compounds such as those found in Clusters 1 and 2 were also retained.


	Fig. 1 TRPV4 activity vs. molecular weight, trellised by molecular cluster. Red = predicted active by ≥2 HTS Bayesian models, yellow = predicted active by 1 model, green = predicted active by 0 models

Compounds were also removed from the screening cascade if they possessed non-drug-like properties (molecular weight >600 Da, or clogP >6), or contained undesirable or reactive substructures e.g. nitro groups, aldehydes, etc.^15,16 In addition to the application of negative filters, a number of positive filters were applied to prioritise compounds for TRPV4 IC50 follow-up. Positive filters were utilised in order to select compounds with either a higher chance of repeating their activity, or that appeared synthetically tractable for follow-up in the event that they were active. Compounds with; a high level of inhibition in the TRPV4 single point assay (>75% inhibition at 10 μM), a high local hit rate (LHR) value (proportion of near neighbours that are active) with either a LHR >10% or a LHR greater than the background hit rate to the extent that the P value, as judged from a chi-squared test, was less than 10⁻²⁰,¹⁷ or that scored highly in the TRPV4 Bayesian model built from the TRPV4 HTS data generated thus far (a Bayesian cut-off score of 15 giving a kappa value k = 0.09) were selected on the basis of their likelihood to retain activity. As can be seen in Fig. 2, the separation of active and inactive peaks in the Bayesian model is not significant; however compounds with a Bayesian score of >15 are likely to be enriched in actives.


	Fig. 2 Bayesian plot of single-point TRPV4 antagonist data

Finally, compounds with little evidence supporting either inclusion or exclusion were analysed. These moderately active compounds (40–75% inhibition of TRPV4 at 10 μM) with an unmeasurable local hit rate (0/0 active neighbours) were analysed by trellising TRPV4 activity plots by Library ID. Library compounds (i.e. compounds synthesised by parallel chemistry) were deemed synthetically tractable, and therefore prioritised for selection, along with further close-in library members not present in the original HTS screening set.

A total of 1050 compounds were selected for TRPV4 IC50 follow-up. The results from the IC50 screen were used to seed a second round of compound selection. Work focussed on close-in mining of active compounds with the aim of deriving SAR around each hit. A number of fingerprint methods (BCI,¹⁸ ECFP6¹⁹ and Atom Pair²⁰ (in-house Tanimoto similarity algorithm)) were used to identify near neighbours around each selected molecule (the probe). Each neighbour was assessed by calculating its maximum common substructure (MCSS)^21,22 with the probe and only neighbours that differed from the probe by a single moiety were retained. For each probe, the number of neighbours per point of variation was restricted to ten to permit a diverse exploration of close-in neighbours around each probe. Tracking the relative occurrence of the attachment point for each probe also gave an indication of the chemical space occupied around it, suggesting gaps that could be explored by synthesis. The triage workflow is highlighted in Scheme 1.


	Scheme 1 TRPV4 HTS triage workflow

The TRPV4 HTS and rapid follow-up screening campaign generated a number of interesting TRPV4 series. One series of particular interest was the hydroxypiperidines, exemplified by compound 1 (Fig. 3). This series demonstrated encouraging levels of hTRPV4 inhibitory activity and binding efficiency (LIPE, LE†)^15,23 and was known to be straightforward to synthesise.


	Fig. 3 Structure and properties of compound 1

Hit to lead follow-up

Hit-to-lead follow-up on the hydroxypiperidine series largely focussed on improving TRPV4 potency and efficiency (LIPE, LE). Any tool compound would need to be suitable for pre-clinical in vivo assessment so must also have an appropriate ADME profile¹⁵ and orthologue (rat) pharmacology. Initial targets sought to replace the thio-ether linker with a more polar group to lower logP²⁴ and increase the probability of achieving good metabolic stability.²⁵ In addition, replacement of the 4-pyridyl group was desired to reduce the liability for drug–drug interactions (DDI).^16,26 A small library of aryl ether hydroxypiperidines was synthesised and assessed in the hTRPV4 (human TRPV4) assay. Gratifyingly, a number of these compounds demonstrated improved TRPV4 potency and LIPE/LE, and were devoid of the pyridyl structural alert.¹⁶ Key compounds 2a–d are outlined in Table 1. Examples within the hydroxypiperidine series showed no TRPV4 agonist activity, were moderately potent inhibitors of rat TRPV4 (rTRPV4) and were selective over a number of other TRP channels. Compound 2c exemplifies the selectivity profile of this series (Table 2):

Table 1 In vitro TRPV4 data for compounds 2a–d and 3a–d


Compd	R1	R2	R3	R4	hTRPV4 IC50(nM)	LE	LIPE
2a	CN	F	Cl	Cl	12.9	0.37	4.43
2b	CN	F	Cl	CN	49.1	0.33	5.13
2c	CN	H	Cl	Cl	49.7	0.36	4.03
2d	CN	F	CN	CN	50.5	0.32	6.10
3a	CN	F	CN	Cl	3.4	0.41	5.39
3b	CN	Cl	CN	Cl	3.6	0.41	4.93
3c	CN	F	Cl	Cl	3.8	0.44	4.66
3d	CN	Cl	Cl	CN	4.0	0.41	5.18

Table 2 In vitro selectivity data for compounds 2c and 3c

	2c	3c
hTRPV4 EC50 (nM)	>1000	>1000
rTRPV4 IC50 (nM)	1150	34.1
rTRPV4 EC50 (nM)	>1000	>1000
hTRPA1 IC50 (nM)	>12500	>12500
hTRPA1 EC50 (nM)	>12500	>12500
hTRPV1 IC50 (nM)	>3900	>16000
hTRPV1 EC50 (nM)	>20000	>20000
hTRPM8 IC50 (nM)	4800	9580


	Fig. 4 Structure and properties of compound 3c

In order to further improve compound potency, a pairwise analysis was conducted on the piperidine linker.²⁷ In this analysis, a substructure search based method (named SWAP)²⁷ was used to interrogate the Pfizer database and identify pairs of molecules differing only in the replacement of a piperidine linker group. The biological activity (for a given biological assay) for each matched pair was recorded to generate a dataset that showed the performance of each isostere. Performance was assessed in terms of the likelihood of retaining, improving or losing potency with respect to the starting point. This analysis highlighted a number of potential isosteres, of which an azetidine replacement was calculated to have a ∼70% chance of retaining or improving potency (as well as lowering logP).²⁸

A number of azetidine-linked compounds, utilising the sulfonamide and aryl-ether SAR established in the hydroxypiperidine series, were synthesised. Pleasingly, a number of these analogues such as compounds 3a–d demonstrated improved hTRPV4 potency and an improved LIPE/LE profile (Table 1).

The hydoxyazetidines displayed similarly high levels of TRP selectivity as the hydroxypiperidines but gratifyingly delivered enhanced rTRPV antagonist activity. Compound 3c exemplifies the selectivity profile of the hydroxyazetidine series (Table 2). In vitro ADME‡ end-points for compound 3c were determined along with an assessment of hERG liability²⁹ utilising a Dofetilide binding assay³⁰ (Table 3). The in vitro ADME profile of 3c was very promising, with low turnover in human liver microsomes and only a relatively weak activity signal in the Dofetilide assay. The pharmacokinetic properties of 3c were also assessed. Compound 3c demonstrated almost complete oral absorption and a ∼2 hour half-life in the rat (Table 4). It was therefore deemed a suitable tool with which to conduct further rodent in vivo studies.

Table 3 In vitro ADME and safety data for compound 3c

Compd	HLM (μL min⁻¹ mg⁻¹)	RLM (μL min⁻¹ mg⁻¹)	DDI^aa DDI (Drug–Drug Interaction). Potential for DDI at Cytochrome P450 1A2, 2C9, 2D6 and 3A4 was assessed at 3 μM.	Log D	Dof. IC50 (nM)
3c	<8.1	87	<20%	3.2	9670

Table 4 In vivo PK parameters of 3c following IV (intravenous) and PO (oral) administration (n = 2, Sprague-Dawley rats)

3c	1 mg kg⁻¹ IV	2 mg kg⁻¹ PO
Half life (h)	2.2
Cl (mL min⁻¹ kg⁻¹)	23.5
Cl_u (mL min⁻¹ kg⁻¹)	756
Vd_ss (L kg⁻¹)	4.3
C _max (nM free)		16
T _max (h)		0.75
F%		73

Chemistry

The synthetic route towards hydroxypiperidines 2a–d is outlined in Scheme 2. Briefly, nucleophilic addition of phenol derivatives (R₁)(R₂)Ar–OH to epoxide 4 under basic conditions furnished N-Boc hydroxypiperdines 5a and b. N-Boc deprotection of compounds 5a and b was accomplished with TFA/DCM to furnish amines 6a and b. Addition of an appropriate sulfonyl chloride, ((R₁)(R₂)ArSO₂Cl), to amines 6a and b, in triethylamine and dichloromethane (DCM), gave hydroxypiperidines 2a–d.


	Scheme 2 Synthesis of hydroxypiperidines 2a–d. Reagents and conditions: (i) (R₁)(R₂)Ar–OH, Cs₂CO₃, IPA/DMSO/H₂0, 80 °C, 49–60%; (ii) TFA, DCM, 25 °C, 75–89%; (iii) (R₁)(R₂)ArSO₂Cl, NEt₃, DCM, 0–25 °C, 8–79%.

The synthesis of hydroxyazetidines 3a–d is described in Scheme 3. Ketone 7 was converted to alkene 8via addition of Ph₃PCH₂Br and potassium tert-butoxide. Alkene 8 was further converted to a regioisomeric mixture of hydroxybromides (9a and 9b) through addition of NBS in DMSO and water. Addition of sodium hydride to a solution of 9a and 9b furnished epoxide 10 in good yield. Nucleophilic ring-opening of epoxide 10 with the appropriate phenol derivative, ((R₁)(R₂)Ar–OH), afforded N-Boc-hydroxypiperdines 11a and b, which were subsequently deprotected to their respective amines (12a and b) under TFA/DCM conditions. A final sulfonylation step on amines 12a and bwith the appropriate sulfonyl chloride ((R₁)(R₂)SO₂Cl), in triethylamine and DCM, furnished hydroxyazetidines 3a–d.


	Scheme 3 Synthesis of hydroxyazetidines 3a–d. Reagents and conditions: (i) Ph₃PCH₂Br, KOt-Bu, Et₂O, RT, 61%; (ii) NBS, DMSO/H₂O, RT, 0–25 °C, 83%; (iii) NaH, THF, 0–25 °C, 62%; (iv) (R₁)(R₂)Ar–OH, Cs₂CO₃, IPA/DMSO/H₂0, 80 °C, 55–71%; (v) TFA, DCM, 25 °C, 85–97%; (vi) (R₁)(R₂)SO₂Cl, NEt₃, DCM, 0–25 °C, 16–90%.

Conclusions

The application of computational techniques (including Bayesian modelling) in the triage of TRPV4 HTS data enabled the early deprioritisation of putative “actives” that were, in fact, likely to be “frequent hitters”. In addition, the use of methods to positively prioritise compounds for follow-up screening allowed the rapid identification of a number of interesting TRPV4 series. Hit-to-lead work on the hydroxypiperidine series included the replacement of the sulfide linker and pyridyl group to generate potent and selective compounds such as 2a–d. Further optimisation, utilising a pairwise analysis protocol (SWAP) delivered a series of hydroxyazetine compounds. Compounds within this series, such as compound 3c, fulfil the TRPV4 potency (rat and human), selectivity and ADME criteria required for use as in vitro/in vivo tools. Pre-clinical in vivo TRPV4 pharmacology and safety data for compound 3c will be reported in due course.

Experimental section

In vitro TRPV4 assay

Receptor-evoked changes in intracellular calcium were measured using calcium-selective fluorescent Ca²⁺ NW dye (Molecular Devices, Sunnyvale, CA, USA) quantitated with a fluorometric imaging plate reader (FLIPR; Molecular Devices). Stably transfected CHOK1 cells expressing human and rat TRPV4 receptors have been used for this assay. On the day before the FLIPR experiment, frozen cryovials of cells were thawed, centrifuged and resuspended in DMEM medium (Invitrogen Cat. # 21063) containing glutamax, sodium pyruvate, non-essential amino acids and 10% FBS (Invitrogen Cat. # 10082-147). The cells were plated into black walled, clear bottom Greiner 384 poly-D-lysine plates at a density of 10 [thin space (1/6-em)]

000 cells per well and incubated overnight in 5% CO₂ at 37 °C. On the day of the experiment, media was removed and cells were loaded with dye loading buffer (0.5× Ca²⁺ NW-dye in HBSS with Ca⁺⁺, Mg⁺⁺, 2.5 mM Probenecid, 0.025% pluronic, 20 mM HEPES, pH 7.4) and incubated for 45 minutes at 37 °C. The test compounds and the reference compound (ruthenium red) were serially diluted in 100% DMSO followed by dilution in assay buffer (HBSS without Ca⁺⁺, Mg⁺⁺, 20 mM HEPES, 5.4 mM CaCl₂, 1 mM MgCl₂, 0.12% P104, pH 7.4) in order to achieve a final highest concentration of 10 μM in the assay plate. The plates were then placed in the FLIPR, and a baseline fluorescence measurement (excitation @ 488 nm and emission @ 510–570 nm) was obtained for 10 s before compound or vehicle addition. The assay was performed in dual mode: firstly, upon addition of compounds to the cells, any possible agonistic effect of the test compounds were monitored for 5 minutes after which the assay plate was incubated for further 15 minutes at room temperature. At the second step, EC75 of the agonist PF-04674114/GSK1016790A was added and fluorescence was measured for an additional 5 minutes. Change in fluorescence values in response to the addition of PF-04674114 (in the absence and presence of test compounds) was measured and inhibition curves generated using nonlinear regression (Prism v.4, GraphPad Software, San Diego, CA, USA) employing a four parameter logistic equation (Y = bottom + (top − bottom)/(1 + 10^(logEC50 − X)hillslope).

Chemistry

All commercially available chemicals and solvents were used without further purification. All temperatures are in °C. Flash column chromatography was carried out using Merck silica gel 60 (9385) or Redisep silica. NMR spectra were obtained on a Varian Mercury (400 MHz) or a Bruker Avance (400 MHz) using the residual signal of the deuterated NMR solvent as the internal reference. Chemical shifts are expressed in parts per million (ppm), multiplicity of the signals are indicated by lower-case letters (singlet s, doublet d, triplet t, quadruplet q, multiple m, broad singlet br s), and deuterated solvents are dimethylsulphoxide d6, methanol d4, and chloroform d1. Mass spectral date were obtained using Waters ZQ ESCI or Applied Biosystem's API-2000.

General procedure for synthesis of N-Boc hydroxypiperidines 5a and b and N-Boc hydroxyazetidines 11a and b. To a stirred solution of epoxide (4 or 10) (0.47 mmol) and cesium carbonate (0.56 mmol) in IPA (3 mL), DMSO (0.5 mL) and water (0.5 mL), was added (R₁)(R₂)Ar–OH (0.52 mmol) and the resulting mixture refluxed at 80 °C for 2.5 h. The reaction mixture was concentrated in vacuo and the residue partitioned between water and ethyl acetate. The organics were further washed with brine, dried over sodium sulphate, concentrated in vacuo and purified by silica-gel column chromatography.

4-(4-Cyano-3-fluoro-phenoxymethyl)-4-hydroxy-piperidine-1-carboxylic acid tert-butyl ester (5a). Following the general procedure for the synthesis of N-Boc hydroxypiperidines 5a and b and N-Boc hydroxyazetidines 11a and b, and using 2-fluoro-4-hydroxybenzonitrile and epoxide 4, the title compound was afforded as a white solid (80 mg, 49%). ¹H NMR (400 MHz, CDCl3) δ: 1.45 (s, 9H), 1.60–1.64 (m, 2H), 1.73 (m, 2H), 2.60 (m, 3H), 3.18 (m, 2H), 3.83 (s, 1H), 3.95 (m, 1H), 6.71–6.78 (m, 2H), 7.53 (t, 1H); m/z 351 [M + H]⁺.

4-(4-Cyano-phenoxymethyl)-4-hydroxy-piperidine-1-carboxylic acid tert-butyl ester (5b). Following the general procedure for the synthesis of N-Boc hydroxypiperidines 5a and b and N-Boc hydroxyazetidines 11a and b, and using 4-hydroxybenzonitrile and epoxide 4, the title compound was prepared as a white solid (9.25 g, 85%). ¹H NMR (400 MHz, DMSO-d6) δ: 1.39 (s, 9H), 1.52 (m, 4H), 3.08 (d, 2H), 3.73 (d, 2H), 3.86 (d, 2H), 7.11 (d, 2H), 7.76 (d, 2H); m/z 333 [M + H]⁺.

General procedure for Boc deprotection of N-Boc hydroxypiperidines (5a and b) and N-Boc hydroxyazetidines (11a and b). To a stirred solution of 5a and b or 11a and b (0.92 mmol) in DCM (5 mL) was added TFA (1 mL) and the resulting mixture stirred at 25 °C for 2 h. The reaction mixture was then concentrated in vacuo, triturated with pentane–ether (5 mL) and used in the next step without further purification.

2-Fluoro-4-(4-hydroxy-piperidin-4-ylmethoxy)-benzonitrile (6a). Following the general procedure for Boc deprotection of N-Boc hydroxypiperdines (5a and b) and N-Boc hydroxyazetidines (11a and b), and using N-Boc hydroxypiperdine 5a, the title compound was a white solid (TFA salt) (251 mg, 75%). ¹H NMR (400 MHz, DMSO-d6) δ: 1.74 (m, 3H), 3.09 (m, 2H), 3.17 (m, 2H), 3.95 (s, 2H), 5.20 (s, 1H), 6.90 (d, 1H), 7.15 (d, 1H), 7.84 (t, 1H), 8.23 (bs, 1H), 8.44 (bs, 1H); m/z 251 [M + H]⁺.

4-(4-Hydroxy-piperidin-4-ylmethoxy)-benzonitrile (6b). Following the general procedure for Boc deprotection of N-Boc hydroxypiperdines (5a and b) and N-Boc hydroxyazetidines (11a and b), and using N-Boc hydroxypiperdine 6a, the title compound was a white solid (TFA salt) (300 mg, 79%). m/z 233 [M + H]⁺.

General procedure for sulfonamide formation (2a–e, 3a–d). To a stirred solution of amine (6a and b or 12a and b, 0.61 mmol) and triethylamine (1.55 mmol) in DCM (5 mL) at 0 °C was added the appropriate benzenesulfonyl chloride (0.61 mmol). The reaction mixture was allowed to warm to 25 °C and stir for 18 h. The reaction mixture was partitioned between water and DCM. The organics were further washed with brine, dried over sodium sulphate, evaporated in vacuo and purified via silica-gel column chromatography.

4-[1-(2,4-Dichloro-benzenesulfonyl)-3-hydroxy-azetidin-3-ylmethoxy]-2-fluoro-benzonitrile (2a). Following the general procedure for sulfonamide formation (2a–d, 3a–d), and using amine 6a and 2,4-dichlorobenzene-1-sulfonyl chloride, the title compound was prepared as a white solid (750 mg, 79%). ¹H NMR (400 MHz, CDCl3) δ: 1.74–1.79 (m, 4H), 3.15–3.22 (m, 2H), 3.73–3.76 (m, 2H), 3.84 (s, 2H), 6.70–6.77 (m, 2H), 7.36–7.38 (dd, 1H), 7.53 (t, 1H), 7.90 (d, 1H), 7.98 (d, 1H); m/z 459 [M + H]⁺.

4-((1-((2-Chloro-4-cyanophenyl)sulfonyl)-4-hydroxypiperidin-4-yl)methoxy)-2-fluorobenzonitrile (2b). Following the general procedure for sulfonamide formation (2a–d, 3a–d), and using amine 6a and 2-chloro-4-cyano benzenesulfonyl chloride, the title compounds was prepared as a white solid (47 mg, 38% yield). ¹H NMR (400 MHz, DMSO-d6) δ: 1.64 (m, 4H), 3.06 (m, 2H), 3.58 (m, 2H), 3.91 (s, 2H), 4.92 (s, 1H), 6.95 (d, 1H), 7.13 (d, 1H), 7.81 (t, 1H), 8.04 (d, 1H), 8.11 (d, 1H), 8.35 (s, 1H); m/z 450 [M + H]⁺.

4-[1-(2,4-Dichloro-benzenesulfonyl)-4-hydroxy-piperidin-4-ylmethoxy]-benzonitrile (2c). Following the general procedure for sulfonamide formation (2a–d, 3a–d), and using amine 6b and 2,4-dichlorobenzene-1-sulfonyl chloride, the title compound was prepared as a white solid (579 mg, 56%). ¹H NMR (400 MHz, CDCl3) δ: 1.61–1.70 (m, 4H), 3.01 (m, 2H), 3.56 (m, 2H), 3.86 (s, 2H), 4.88 (s, 1H), 7.08 (d, 1H), 7.65 (d, 1H), 7.74–7.80 (m, 3H), 7.93 (d, 1H), 7.96 (dd, 1H); m/z 441 [M + H]⁺.

4-[4-(4-Cyano-3-fluoro-phenoxymethyl)-4-hydroxy-piperidine-1-sulfonyl]-isophthalonitrile (2d). Following the general procedure for sulfonamide formation (2a–d, 3a–d), and using amine 6a and 2,4-dicyano benzenesulfonyl chloride, the title compound was prepared as a white solid (22 mg, 8%). ¹H NMR (400 MHz, DMSO-d6) δ: 1.66 (m, 4H), 2.94 (m, 2H), 3.60 (m, 2H), 3.90 (s, 2H), 4.90 (s, 1H), 6.95 (d, 1H), 7.13 (d, 1H), 7.81 (t, 1H), 8.18 (d, 1H), 8.41 (d, 1H), 8.80 (s, 1H); m/z 441 [M + H]⁺.

3-Methylene-azetidine-1-carboxylic acid tert-butyl ester (8). To a stirred solution of Ph₃PCH₂Br (88.7 g, 248.4 mmol) in dry ether (400 mL) was added potassium tert-butoxide (27.9 g, 248.4 mmol) and the mixture stirred at 25 °C for 1 h. Ketone 7 (42.5 g, 248.4 mmol) in dry ether (400 mL) was added and the resulting mixture stirred at 25 °C for 18 h. The reaction mixture was filtered through Celite®. The filtrate was washed with water (100 mL), brine (100 mL), dried over sodium sulphate and concentrated in vacuo to afford the crude compound as a yellow oil (10.8 g, 61% yield), which was used without further purification. ¹H NMR (400 MHz, CDCl3) δ: 1.44 (s, 9H), 4.46 (s, 4H), 4.97 (s, 2H).

3-Bromomethyl-3-hydroxy-azetidine-1-carboxylic acid tert-butyl ester and 3-bromo-3-hydroxymethyl-azetidine-1-carboxylic acid tert-butyl ester (9a and 9b). To a stirred solution of alkene 8 (2.9 g, 17.13 mmol) in DMSO (15 mL) at 0 °C was added NBS (6.10 g, 34.2 mmol) and water (617 mL, 34.2 mmol). The resulting mixture was allowed to warm to 25 °C and stirred for a further 18 h. The reaction mixture was partitioned between water (25 mL) and ethyl acetate (25 mL). The organics were washed with brine (10 mL), dried over sodium sulphate, concentrated in vacuo and purified by silica-gel column chromatography (eluting with ethyl acetate [thin space (1/6-em)]

hexane, 10 [thin space (1/6-em)]

90) to afford the isomeric mixture of title compounds as a pale yellow oil (7.65 g, 83%). ¹H NMR (400 MHz, CDCl3) δ: 1.43 (s, 18H), 2.03–2.06 (t, 1H), 2.68 (s, 1H), 3.68 (s, 2H), 3.87–3.91 (m, 8H), 4.26–4.27 (dd, 2H).

1-Oxa-5-aza-spiro[2.3]hexane-5-carboxylic acid tert-butyl ester (10). To a stirred solution of isomeric mixture of 9a and 9b (4.6 g, 17.29 mmol) in THF (80 mL) at 0 °C was added sodium hydride (760 mg, 19.02 mmol, 60%) in one portion. The resulting mixture was allowed to warm to 25 °C and stirred for 7 h. The reaction mixture was quenched with saturated ammonium chloride (50 mL) and stirred at 25 °C for a further 18 h. The reaction mixture was extracted with ethyl acetate (100 mL) and the organic phase dried over sodium sulphate, concentrated in vacuo, and purified by silica-gel column chromatography (eluting with ethyl acetate [thin space (1/6-em)]

hexane, 10 [thin space (1/6-em)]

90) to afford the title compound as a pale yellow oil (2.0 g, 62%). ¹H NMR (400 MHz, CDCl3) δ: 1.44 (s, 9H), 2.84 (s, 2H), 4.17–4.26 (m, 4H).

3-(4-Cyano-3-fluoro-phenoxymethyl)-3-hydroxy-azetidine-1-carboxylic acid tert-butyl ester (11a). Following the general procedure for the synthesis of N-Boc hydroxypiperidines 5a and b and N-Boc hydroxyazetidines 11a and b, and using 2-fluoro-4-hydroxybenzonitrile and epoxide 10, the title compound was afforded as a white solid (250 mg, 57%). ¹H NMR (400 MHz, DMSO-d6) δ: 1.38 (s, 9H), 3.71 (d, 2H), 3.87 (d, 2H), 4.15 (s, 2H), 6.11 (s, 1H), 6.98 (d, 1H), 7.16 (d, 1H), 7.83 (t, 1H); m/z 323 [M + H]⁺.

3-(3-Chloro-4-cyano-phenoxymethyl)-3-hydroxy-azetidine-1-carboxylic acid tert-butyl ester (11b). Following the general procedure for the synthesis of N-Boc hydroxypiperidines 5a and b and N-Boc hydroxyazetidines 11a and b, and using 2-chloro-4-hydroxybenzonitrile and epoxide 10, the title compound was afforded as a white solid (250 mg, 55%). ¹H NMR (400 MHz, DMSO-d6) δ: 1.38 (s, 9H), 3.71 (d, 2H), 3.87 (d, 2H), 4.17 (s, 2H), 6.10 (s, 1H), 7.10 (m, 1H), 7.36 (d, 1H), 7.87 (d, 1H); m/z 339 [M + H]⁺.

2-Fluoro-4-(3-hydroxy-azetidin-3-ylmethoxy)-benzonitrile (12a). Following the general procedure for Boc deprotection of N-Boc hydroxypiperdines (5a and b) and N-Boc hydroxyazetidines (11a and b), and using N-Boc hydroxypiperdine 11a, the title compound was a white solid (TFA salt) (200 mg, 79%). ¹H NMR (400 MHz, DMSO-d6) δ: 3.94 (d, 2H), 4.03 (d, 2H), 4.19 (s, 2H), 6.59 (s, 1H), 7.05 (d, 1H), 7.22 (s, 1H), 7.88 (t, 1H), 8.74–8.86 (brs, 1H); m/z 223 [M + H]⁺.

2-Chloro-4-(3-hydroxy-azetidin-3-ylmethoxy)-benzonitrile (12b). Following the general procedure for Boc deprotection of N-Boc hydroxypiperdines (5a and b) and N-Boc hydroxyazetidines (11a and b), and using N-Boc hydroxypiperdine 11b, the title compound was a white solid (TFA salt) (200 mg, 77%). ¹H NMR (400 MHz, DMSO-d6) δ: 3.92 (d, 2H), 4.03 (d, 2H), 4.20 (s, 2H), 6.58 (s, 1H), 7.17 (d, 1H), 7.44 (s, 1H), 7.93 (d, 1H), 8.71–8.82 (brs, 1H); m/z 239 [M + H]⁺.

4-({1-[4-Chloro-2-cyano-phenylsulfonyl]-3-hydroxyazetidine-3-yl}methoxy)-2-fluorobenzonitrile (3a). Following the general procedure for sulfonamide formation (2a–d, 3a–d), and using amine 12a and 4-chloro-2-cyano-benzenesulfonyl chloride, the title compound was prepared as a white solid (50 mg, 16%). ¹H NMR (400 MHz, DMSO-d6) δ: 3.77 (d, 2H), 3.96 (d, 2H), 4.12 (s, 2H), 6.34 (s, 1H), 6.90 (dd, 1H), 7.11 (dd, 1H), 7.83 (t, 1H), 8.05 (s, 2H), 8.47 (s, 1H); m/z 422 [M + H]⁺.

2-Chloro-4-({1-[4-chloro-2-cyano-phenylsulfonyl]-3-hydroxyazetidine-3-yl}methoxy) benzonitrile (3b). Following the general procedure for sulfonamide formation (2a–d, 3a–d), and using amine 12b and 4-chloro-2-cyano-benzenesulfonyl chloride, the title compound was prepared as an off-white solid (52 mg, 16%). ¹H NMR (400 MHz, DMSO-d6) δ: 3.78 (d, 2H), 3.97 (d, 2H), 4.14 (s, 2H), 6.33 (s, 1H), 7.03–7.05 (dd, 1H), 7.27 (d, 1H), 7.89 (d, 1H), 8.05 (s, 2H), 8.46 (d, 1H); m/z 438 [M + H]⁺.

4-[1-(2,4-Dichloro-benzenesulfonyl)-3-hydroxy-azetidin-3-ylmethoxy]-2-fluoro-benzonitrile (3c). Following the general procedure for sulfonamide formation (2a–d, 3a–d), and using amine 12a and 4-chloro-2-chloro-benzenesulfonyl chloride, the title compound was prepared as a white solid (110 mg, 63%). ¹H NMR (400 MHz, CDCl3) δ: 4.04 (d, 2H), 4.12 (d, 2H), 4.20 (s, 2H), 6.77–6.81 (m, 2H), 7.38–7.41 (dd, 1H), 7.57 (t, 1H), 7.89 (d, 1H), 7.97 (d, 1H); m/z 431 [M + H]⁺.

2-Chloro-4-((1-((2-chloro-4-cyanophenyl)sulfonyl)-3-hydroxyazetidin-3-yl)methoxy)benzonitrile (3d). Following the general procedure for sulfonamide formation (2a–d, 3a–d), and using amine 12b and 4-cyano-2-chloro-benzenesulfonyl chloride, the title compound was prepared as a white solid (12 mg, 22%). ¹H NMR (400 MHz, DMSO-d6) δ: 3.81 (d, 2H), 3.95 (d, 2H), 4.15 (s, 2H), 6.90 (s, 1H), 7.15 (d, 1H), 7.75 (t, 1H), 8.09 (d, 1H), 8.16 (d, 1H), 8.41 (s, 1H); m/z 438 [M + H]⁺.

Acknowledgements

The authors would like to thank Francois Bertelli and Chris Chambers for generating the TRPV HTS data, Florian Wakenhut for assistance in the HTS triage and TCG Lifesciences for synthesising and generating TRPV4 data on compounds 2a–d and 3a–d.

Notes and references

B. Nilius, et al., TRPV4 calcium entry channel: a paradigm for gating diversity, Am. J. Physiol., 2004, 286(2, Pt. 1), C195–C205 CrossRef CAS.
W. Liedtke, et al., Mammalian TRPV4 (VR-OAC) directs behavioral responses to osmotic and mechanical stimuli in Caenorhabditis elegans, Proc. Natl. Acad. Sci. U. S. A., 2003, 100(suppl. 2), 14531–14536 CrossRef CAS.
A. Mizuno, et al., Impaired osmotic sensation in mice lacking TRPV4, Am. J. Physiol., 2003, 285(1, Pt. 1), C96–C101 CAS.
H. Todaka, et al., Warm temperature-sensitive transient receptor potential vanilloid 4 (TRPV4) plays an essential role in thermal hyperalgesia, J. Biol. Chem., 2004, 279(34), 35133–35138 CrossRef CAS.
B. Nilius, G. Droogmans and R. Wondergem, Transient receptor potential channels in endothelium: solving the calcium entry puzzle?, Endothelium, 2003, 10(1), 5–15 CrossRef CAS.
R. N. Willette, et al., Systemic activation of the transient receptor potential vanilloid subtype 4 channel causes endothelial failure and circulatory collapse: part 2, J. Pharmacol. Exp. Ther., 2008, 326(2), 443–452 CrossRef CAS.
M. Suzuki, et al., Impaired pressure sensation in mice lacking TRPV4, J. Biol. Chem., 2003, 278(25), 22664–22668 CrossRef CAS.
T. Gevaert, et al., Deletion of the transient receptor potential cation channel TRPV4 impairs murine bladder voiding, J. Clin. Invest., 2007, 117(11), 3453–3462 CAS.
K. S. Thorneloe, et al., N-((1S)-1-{[4-((2S)-2-{[(2,4-dichlorophenyl)sulfonyl]amino}-3-hydroxypropanoyl)-1-piperazinyl]carbonyl}-3-methylbutyl)-1-benzothiophene-2-carboxamide (GSK1016790A), a novel and potent transient receptor potential vanilloid 4 channel agonist induces urinary bladder contraction and hyperactivity: part I, J. Pharmacol. Exp. Ther., 2008, 326(2), 432–442 CrossRef CAS.
S. F. Pedersen, G. Owsianik and B. Nilius, TRP channels: an overview, Cell Calcium, 2005, 38(3–4), 233–252 CrossRef CAS.
Transient Receptor Potential (TRP) Channels, in Handbook of Experimental Pharmacology, ed. V. Flockerzi and B. Nilius, 2007, vol. 179, p. 606 Search PubMed.
S. D. Pratt, et al., Identification and characterization of mGlu3 ligands using a high throughput FLIPR assay for detection of agonists, antagonists, and allosteric modulators, Comb. Chem. High Throughput Screening, 2011, 14(7), 631–641 CrossRef CAS.
O. Roche, et al., Development of a virtual screening method for identification of “frequent hitters” in compound libraries, J. Med. Chem., 2002, 45(1), 137–142 CrossRef CAS.
M. Glick, et al., Enrichment of extremely noisy high-throughput screening data using a naive Bayes classifier, J. Biomol. Screening, 2004, 9(1), 32–36 CrossRef CAS.
P. D. Leeson and B. Springthorpe, The influence of drug-like concepts on decision-making in medicinal chemistry, Nat. Rev. Drug Discovery, 2007, 6(11), 881–890 CrossRef CAS.
A. F. Stepan, et al., Structural alert/reactive metabolite concept as applied in medicinal chemistry to mitigate the risk of idiosyncratic drug toxicity: a perspective based on the critical examination of trends in the top 200 drugs marketed in the united states, Chem. Res. Toxicol., 2011, 24(9), 1345–1410 CrossRef CAS.
B. A. Posner, H. Xi and J. E. J. Mills, Enhanced HTS hit selection via a local hit rate analysis, J. Chem. Inf. Model., 2009, 49(10), 2202–2210 CrossRef CAS.
Barnard Chemical Information, S., http://www.digitalchemistry.co.uk.
Pipeline Pilot, V 6.1, Accelrys Inc., San Diego, CA Search PubMed.
M. Tu, Fingerprints generated by counting frequency with which particular pairs of atoms are separated by particular through-bond distances, unpublished results.
J. E. Mills, unpublished results.
J. W. Raymond, E. J. Gardiner and P. Willett, Heuristics for similarity searching of chemical graphs using a maximum common edge subgraph algorithm, J. Chem. Inf. Comput. Sci., 2002, 42(2), 305–316 CrossRef CAS.
T. Ryckmans, et al., Rapid assessment of a novel series of selective CB(2) agonists using parallel synthesis protocols: a Lipophilic Efficiency (LipE) analysis, Bioorg. Med. Chem. Lett., 2009, 19(15), 4406–4409 CrossRef CAS.
R. Mannhold and R. F. Rekker, The hydrophobic fragmental constant approach for calculating logP in octanol/water and aliphatic hydrocarbon/water systems, Perspect. Drug Discovery Des., 2000, 18(Hydrophobicity and Solvation in Drug Design, Pt. II), 1–18 CrossRef CAS.
C. E. Keefer, G. Chang and G. W. Kauffman, Extraction of tacit knowledge from large ADME datasets via pairwise analysis, Bioorg. Med. Chem., 2011, 19(12), 3739–3749 CrossRef CAS.
J. L. Born and W. M. Hadley, Inhibition of in vitro cytochrome P-450-catalyzed reactions by substituted pyridines, J. Pharm. Sci., 1980, 69(4), 465–466 CrossRef CAS.
J. E. J. Mills, et al., SAR mining and its application to the design of TRPA1 antagonists, MedChemComm, 2012, 3(2), 174–178 RSC.
J. E. J. Mills and C. M. Barker, unpublished results.
M. C. Sanguinetti and M. Tristani-Firouzi, hERG potassium channels and cardiac arrhythmia, Nature, 2006, 440(7083), 463–469 CrossRef CAS.
M. Deacon, et al., Early evaluation of compound QT prolongation effects: a predictive 384-well fluorescence polarization binding assay for measuring hERG blockade, J. Pharmacol. Toxicol. Methods, 2007, 55(3), 238–247 CrossRef CAS.

Footnotes

† LIPE (Lipophilic Ligand Efficiency) = pK_i − clogP (or log D). LE (Ligand Efficiency) = 1.3643pK_i/number of heavy atoms.

‡ ADME (Absorption Distribution Metabolism Excretion).