Journal of Pharmacological and Toxicological Methods 55 (2007) 255–264www.elsevier.com/locate/jpharmtox
Original article
Early evaluation of compound QT prolongation effects: A predictive384-well fluorescence polarization binding assay for
measuring hERG blockade
Matt Deacon a, David Singleton b, Nikki Szalkai a, Rodger Pasieczny c, Chris Peacock a,David Price a, James Boyd b, Helen Boyd a, Jill V. Steidl-Nichols c, Christine Williams a,⁎
a Pfizer Global Research and Development, Sandwich Laboratories, Ramsgate Road, Sandwich, Kent CT13 9NJ UKb Pfizer Global Research and Development, Groton Laboratories, Groton, CT 06340, USA
c Pfizer Global Research and Development, La Jolla Laboratories, San Diego, CA 92121, USA
Received 30 August 2006; accepted 30 September 2006
Keywords: Arrhythmia; Binding; Ether-a-go-go related gene; Fluorescence polarization; Methods; High throughput screening; Human; hERG; HEK 293 cells; Patchclamp
1. Introduction
Over the last few years drug-induced cardiac arrhythmia hasbecome a key safety concern for the pharmaceutical industryand its regulatory bodies, with a wide variety of pharmacolog-ical agents having been found to cause potentially lethal pro-longation of the QT phase of the electrocardiogram (Calderone,Testai, Martinotti, Del Tacca, & Breschi, 2005; De Ponti,
Poluzzi, Cavalli, Racanatini, & Montanaro, 2002; Fermini &Fossa, 2003; Haverkamp et al., 2000). Furthermore, it is wellestablished that such acquired long QT syndrome (ALQTS) ispredominantly mediated through blockade of potassiumchannels encoded by the human ether-a-go-go related gene(hERG), which comprise the delayed rectifier potassium currentinvolved in cardiac repolarization (IKr) (Curran et al., 1995;Pearlstein, Vaz, & Rampe, 2003; Sanguinetti, Jiang, Curran, &Keating, 1995; Vandenberg, Walker, & Campbell, 2001). Withrecent emphasis on the pharmaceutical industry to decreasecosts and improve efficiency by minimising attrition, it is clearly
256 M. Deacon et al. / Journal of Pharmacological and Toxicological Methods 55 (2007) 255–264
desirable to evaluate compounds for their potential to prolongQTas early as possible and ideally during early lead optimisation(Fermini & Fossa, 2003; Kola & Landis, 2004). Whilst molec-ular models that predict hERG blockade could be employed asan early filter, these models are only as effective as the biologicaldata used to define them and asmore information is known aboutnon-blockers than blockers, at present false predictions could beobtained (Aronov, 2005). Therefore, to facilitate early evalua-tion of the potential for compound QT prolongation a high-throughput, low cost assay is required. Furthermore, it is es-sential that this assay is predictive of other in vitro downstreamstudies such as electrophysiology so that, in combination withappropriate pharmacokinetic/dynamic considerations and invivo data, accurate integrated assessments can be made re-garding the safety margin of compounds entering the clinic(Redfern et al., 2003; Webster, Leishman, & Walker, 2002).
There are a number of “high-throughput” assay technologiesthat have been evaluated for profiling compound activity athERG, including radioligand binding (Chiu et al., 2004; Diazet al., 2004; Finlayson, Turnbull, January, Sharkey, & Kelly,2001), membrane potential (Dorn et al., 2005), rubidium efflux(Cheng et al., 2002; Tang et al., 2001) and automated patchclamp (Bridgland-Taylor et al., 2006; Dubin et al., 2005).However, as highlighted in recent reviews (Netzer, Bischoff, &Ebneth, 2003; Wood, Williams, & Waldron, 2004; Zheng,Spencer, & Kiss, 2004) the currently available techniques havelimitations either in terms of cost, throughput or discrepanciesin the pharmacological profile that could result in false neg-atives. Here we describe the development and characterisationof a new binding assay for the hERG channel, which is com-patible with high-throughput 384-well screening, using theinexpensive, non-radiometric fluorescence polarization (FP)technology. The assay demonstrates an excellent correlationwith historical radiometric binding and electrophysiology data,indicating that it will be a valuable drug discovery screen whichis compatible with the early elimination of compounds that havethe potential to induce QT effects in patients.
Testing was carried out in HEK 293 cells transfected with thehERG gene (Zhou et al., 1998) maintained at 37 °C in MinimumEssential Medium with Earle's Salts and L-glutamine, 10% fetalbovine serum, 1 mM sodium pyruvate, 0.1 mM nonessentialamino acids and 0.4mg/mLgeneticin.Membrane currents of cellscontinuously perfused with 35 °C extracellular recording saline(in mM: 137 NaCl, 4 KCl, 1.8 CaCl2, 1 MgCl2, 10 HEPES, 10 D-glucose, pH 7.4, 335 mOs) were measured using whole-cell patchclamp (Hamill, Marty, Neher, Sakmann, & Sigworth, 1981) witha MultiClamp 700A amplifier (Molecular Devices) and 3–5 MΩglass pipettes filled with intracellular recording saline (in mM:130 KCl, 1 MgCl2, 10 HEPES, 5 Mg-ATP, 5 EGTA, pH 7.2,320 mOs). The liquid junction potential was 4 mV. Accessresistancewas compensated by at least 80%, gainwas set to 1mV/pA, and Bessel filters were set to 3 and 10 kHz.
For initial characterization of C6–Cy3B ligand potency, thehERG current was activated with a voltage step to +20 mV for1 s followed by a ramp to −80 mV at 0.5 V/s delivered at0.25 Hz (2.8-s interpulse interval). Following 5 min of stablecontrol recording in which rundown of the tail current peak wasb2%/min, Cy3B ligand solution was continuously perfused intothe recording chamber for evaluation of current blockade, andreversal was evaluated by perfusion with saline upon reachingan apparent steady state inhibition. In order to investigate theuse-dependent nature of block by the C6–Cy3B ligand(100 nM), the hERG current was activated with a voltage stepto 0 mV for 4 s followed by a ramp to −80 mV at 0.4 V/sdelivered at 0.14 Hz (2.8-s interpulse interval). The magnitudeof block was determined and compared to data obtained withthe +20 mV voltage step for 1 s with 100 nM C6–Cy3B ligand.To further explore use-dependent block of the hERG current,the rate of onset of block by the C6–Cy3B ligand (1 μM) wasevaluated at different frequencies of channel activation. ThehERG current was activated using a voltage step to +20 mV for1 s followed by a ramp to −80 mV at 0.5 V/s, and interpulseintervals of 1.0, 2.2 and 6.8 s were investigated. Peak currents inthe presence of compound were normalized to the rundownadjusted control current amplitude, plotted as a function of time,and fit with a single exponential equation to derive a timeconstant (τ) for onset of block.
2.2. Fluorescence polarization
Membrane homogenates of HEK 293 (Cell line #15-08) cellsexpressing the hERG product supplied by (PGRD) SandwichLaboratories were prepared as follows. Cell pellets were thawedat room temperature and kept on ice. Buffer (50 mM Tris–HCl,1 mMMgCl2, 10 mMKCl, pH 7.4, 4 °C) was added to each cellpellet (10 mL of buffer per 10 g of packed cell pellet) and themixture homogenised using an Omni LabTek homogeniser(20,000 rpm for 30 s). The homogenate was centrifuged at48,000×g for 20 min between 3 and 5 °C in a Sorvall EvolutionRC centrifuge and the supernatant discarded. The pellet wasresuspended, homogenised (20,000 rpm for 10 s), and cen-trifuged as before. The resultant supernatant was discarded andthe final pellet resuspended (100 mL of the above buffer per10 g of packed cell pellet), homogenised (20,000 rpm for 10 s),dispensed in to tubes in 1, 2 and 5 mL aliquots and storedbetween −75 °C and −85 °C until use. Protein concentrationwas determined using a Coomassie Blue kit as per manufac-turer's instructions (Sigma 610A and 610-11).
The Cy3B ligand was stored in 100% DMSO and diluted to6 nM in assay buffer (50 mM Tris–HCl, 1 mM MgCl2, 10 mMKCl, 0.05% Pluronic F127, pH 7.4, 4 °C) on the day of theexperiment. Test samples and controls were diluted in 6%DMSO, 0.05% Pluronic F127. Cell membranes were removedfrom the −80 °C freezer and placed on ice after defrosting.Whenrequired the defrosted membranes were homogenised using apolytronic device for nomore than 10 s, they were then diluted inthe above assay buffer to produce a working solution of 0.3 mg/mL. The assay was compiled by adding 10 μL of test compoundor control solution, 10 μL of the Cy3B ligand and 10 μL of cell
257M. Deacon et al. / Journal of Pharmacological and Toxicological Methods 55 (2007) 255–264
membranes to a black 384-well plate (Matrix, Cat No. 4318).The plates were mixed and then incubated for a minimum of 2 hprior to reading on a Tecan Ultra (Excitation 530 nm, Emission590 nm). For association/dissociation studies the time ofincubation was modified to capture earlier readings such thatthe initial rate of change could be determined. In this case themeasurements were carried out at ∼15- to 30-s intervals for thefirst 5–10 min and 5- to 10-min intervals for the remaining time.All IC50 and Ki data were generated using Pfizer proprietarysoftware and kinetic data was analyzed using GraphPad Prism.
2.3. [3H]-Dofetilide binding SPA
Radioligand binding studies using the SPA technology wereperformed as described below. This method is covered throughpatent number WO 03/021271 A2 and was first reported at the3rd Annual Safety Pharmacology Society (SPS) meeting,Noordwijk, The Netherlands, September 29–30, 2003 (Curtis,2004). The membrane homogenate was pre-coupled withYttrium silicate polylysine beads in buffer (50 mM Tris base,10 mM KCl and 1 mMMgCl2, pH 7.4, 4 °C) at a ratio of 16 μghERG protein per 1 mg bead at 4 °C on a roller shaker forapproximately 2 h. Following this pre-coupling the mixture wascentrifuged at 1000 rpm in a Heraeus Biofuge (or similarinstrument) at 4 °C for 2 min and the supernatant discarded. Thebead/membrane mixture was then re-suspended in cold assaybuffer to give a working solution of 12.5 mg bead/mL. 10 μL oftest compound, standard compound or diluent were added fromstock plates into the assay plates. The [3H]-dofetilide ligand(10 μL of 30 nM stock) and pre-coupled bead/membrane(40 μL) were added to the assay plate which were then shakenfor 1 h at room temperature. The beads were allowed to settle fora minimum of 30 min prior to measurement. All IC50 and Ki
data were generated using Pfizer proprietary software andkinetic data was analyzed using GraphPad Prism.
2.4. Automated electrophysiology studies using PatchXpress
HEK 293 cells which stably express the hERG potassiumchannel (Zhou et al., 1998) were harvested by trypsin from T25cell culture flasks and resuspended in 130 μL of external buffer(solution of composition (in mM); 130 NaCl, 4 KCl, 2 CaCl2,1 MgCl2, 10 glucose, 5 HEPES, pH 7.4±0.05 with NaOH/HCl).This cell suspension was used for experimentation on aPatchXpress (Molecular Devices) within 30 min of preparation.The intracellular solution utilized in these experiments wascomposed as follows: (in mM); 130 KCl, 1 MgCl2, 10 HEPES,5 EGTA, pH 7.2±0.05 with KOH. Membrane potential wasstepped from a holding potential of −80 mV to +30 mV for 1 s,followed by a descending voltage ramp at a rate of 0.5 mVms−1
back to holding potential of −80 mV. This protocol was evokedrepeatedly every 4 s (0.25 Hz) in the presence of increasingconcentrations of compound. Cells were accepted for experi-mentation if current rundown was b3% min−1. Dofetilide(10 μM) was added at the end of each experiment to definedofetilide-insensitive current which was digitally subtractedfrom all the preceding sweeps. The peak of the dofetilide-
sensitive current during the repolarising ramp was measured andplotted against time and the IC50 curves and values generatedusing the relevant custom written scripts within the DataXpressprogram (Molecular Devices). All experiments were performedat room temperature 26.5±1.2 °C. Compounds were solubilisedinitially in 4mMDMSO and diluted in external solution. Vehiclecontrols had no effect in time-matched experiments.
2.5. Caco-2 permeability study
Caco-2 cells were obtained from the American Type CultureCollection (ATCC), Rockville, MDUSA. All monolayers used inthis study were between passage 25 and 40. The cells were grownin T-150 flasks at 37 °C in an atmosphere of 5% CO2 using D-MEM (Invitrogen-GIBCO) supplemented with 4500 mg/L D-glucose, 584 mg/L L-glutamine, 1% non-essential amino acids,1% sodium pyruvate and 10% fetal bovine serum. The media waschanged every other day. When the flasks reached 90% con-fluence the monolayers were washed three times with HBSS–CMF and the cells were removed from the flasks by incubatingthe monolayers with 0.25% trypsin in 1 mM EDTA solution for10 min at 37 °C. The cells were collected into centrifuge tubes,pelleted at 600×g for 10 min and resuspended in D-MEM. TheCaco-2 cells were seeded into 24-well Falcon HTS Multiwell™Inserts at 60,000 cells/cm2. The cells were allowed to grow for21 days and were supplied with media every other day. About30 min before initiation of the experiment, the monolayers werewashed free of cell culture medium and allowed to incubate at37 °C in 0.1 mL of HBSS in the insert (donor compartment) and0.5 mL in the receiver compartment. The permeability experi-ments were conducted at 37 °C. At time=0, the donor solutionwas suctioned off and replaced with 0.1 mL of fresh HBSScontaining 150 μM of compound of interest. After 1 h a samplewas removed from the donor and receiver compartments andanalyzed on a Sciex API 3000™ LC/MS/MS. Permeability wascalculated from the rate of solute appearance in the receivercompartment using the following equation:
Pe ¼ VrACo
dCdt
ð1Þ
where
Pe permeability coefficient, cm/sVr volume of receiver reservoir, mLA area of exposed membrane, cm2
Co initial concentration of solute in the donor reservoirdC/dt slope of the accumulated receiver concentration with
time
3. Results
3.1. Identification of a f luorescent hERG blocker
Over the last few years a variety of fluorescent molecules andtechnologies have been employed in both academia andindustry to study protein :protein or ligand :protein interactions.
Table 1Summary of ligands synthesised and their hERG binding affinities
Type 1structuresKi (nM)
Type 2structuresKi (nM)
Type 3structuresKi (nM)
Dye label Dyestructure
Excitation/emissionproperties(nm)
n=3 carbonchain linker
n=6 carbonchain linker
NBD
485/540 39.2±3.24 7.06±0.91 50.1±1.4Fluorescein
485/535 N1000 N1000 15.2±15.1TAMRzA
530/590 Not synthesized Not synthesized 89.1±50.4
Cy3B
530/590 659±53.8 162.11±70.0 6.94±0.86
Alexa 647 Not available
590/680 Not synthesized N1000 N1000
Three types of structures were examined as potential f luorescent ligands. The templates for each type of structure are included in the f irst row of the table, with Type 1 being based on an a cisapride/astemizole template,Type 2 being based on a dofetilide template with a 3-carbon chain linker and Type 3 being based on a dofetilide template with a 6-carbon chain linker. The f irst and second columns of the table indicate which dyestructure was incorporated. Each new structure synthesised was tested for aff inity at the hERG channel via competition experiments with [3H]-dofetilide and data represented are geometric mean±S.D.
258M.Deacon
etal.
/Journal
ofPharm
acologicaland
ToxicologicalMethods
55(2007)
255–264
Fig. 1. C6–Cy3B functional blockade of hERGcurrents expressed inHEK293 cells. (a) Peak hERGcurrent (generated at 0.25Hzwith a 1-s step to +20mV followed by a rampto−80mVat 0.5V/s) is plotted as a function of time, and initiation of perfusionwith 1μMC6–Cy3B ligand and reversal of blockadewith saline perfusion is indicated. Exampletraces (inset) are shown under control conditions (♦), after 10 min perfusion with 1 μMC6–Cy3B ligand (⁎), and 20 min after washout with saline (▪). Inhibition was 18.0±3.8% and 82.5±1.0% at 0.1 and 1 μM, respectively (mean±S.E.M.,N=6–9 cells). (b) Percent inhibition of hERG current by 0.1 μMC6–Cy3B determined using a 4-s step to0 mV (33.9±2.1%, mean±S.E.M., N=4) is greater than inhibition with a 1-s step to +20 mV (18.0±3.8%, mean±S.E.M., N=6). Statistical significance is indicated by anasterisk (unpairedT-test, ⁎pb0.05). (c) Peak hERG current (generatedwith a 1-s step to +20mV followed by a ramp to−80mVat 0.5V/s) in the presence of 0.1μMC6–Cy3Bligand is normalized to the control current amplitude and plotted as a function of time for three different interpulse intervals (1.0, 2.8 and 6.8 s, mean±S.E.M.,N=4–6). Timeconstants (τ) for onset of block (right) were compared using an ANOVA followed by a Bonferroni comparison. Statistical significance is indicated by an asterisk (⁎pb0.05).
259M. Deacon et al. / Journal of Pharmacological and Toxicological Methods 55 (2007) 255–264
Initially, such studies were performed with bright probes emit-ting light in the “green” wavelengths, such as fluorescein, but toaddress the impact of biological or compound interference, thetechnology has evolved and there are now a wide variety offluorescent probes available, covering the “green” to “red”spectrum (Gribbon & Sewing, 2003). Whilst FP assays, util-izing a variety of fluors, have become well established for theevaluation of G-protein-coupled receptor (GPCRs) and kinaseligand interactions (Allen, Reeves, & Mellor, 2000; Burke,Loniello, Beebe, & Ervin, 2003; Harris, Cox, Burns, & Norey,2003; Owicki, 2000; Zaman, Garritsen, De Boer, & VanBoeckel, 2003), this technique has not been applied widely tothe study of ion channels. Therefore, the first challenge was todevelop a fluorescent ligand for use in the hERG binding assay.
A range of the commercially available fluorescent dyes wereselected to generate a new hERG ligand, which encompassed arange of structures and spectral properties. Furthermore twodifferent strategies were employed to generate the fluorescentligand i) incorporation of a dye within a known hERG phar-macophore (based upon astemizole and cisapride), and ii)addition of a fluorescent tag, via a carbon chain linker, on to aknown hERG pharmacophore (a dofetilide derivative). In bothcases, dyes were utilized as supplied by the manufacturer andtwo activation methods were developed depending on thereactivity of the amine being coupled in the dofetilide or aste-mizole/cisapride-like core templates. For templates with a pri-
mary amine these were coupled with succinimidyl activateddyes using dimethylformamide as a solvent and eithertriethylamine or diisopropylethylamine as a base. For secondaryamino groups whose reactivity proved resistant to coupling viaconventional succinimidyl ester couplings, N[(dimethylamino)-1H-1,2,3-triazolo[4,5-b]pyridin-1-ylmethylene] hexafluoropho-sphate N-oxide (HATU) was used according to typical peptidesynthesis methods (Kates, Carpino, & Albericio, 1996). Prod-ucts were purified and isolated with preparative HPLC or flashcolumn chromatography on silica gel.
All the ligands synthesised exhibited the predicted massspectrum qualities and chromatographic purity expected. Fol-lowing quantification, which was determined spectrophotomet-rically using extinction coefficients, each ligand was evaluatedfor its fluorescence properties and ability to displace [3H]-dofetilide in competition radioligand binding studies. Whilst themajority of ligands retained binding activity at the channel, theone best suited for use in the assay upon evaluation of both itsfluorescence properties and affinity, was a dofetilide derivativelabelled via a 6-carbon chain linker to the fluorescent tag, Cy3B,which was demonstrated to have a Ki of 6.94±0.86 nM,geometric mean±S.D. (Table 1).
Although hERG binding had been established for this ligand,prior to assay development the ligand was evaluated for functionalhERG blockade via whole cell patch-clamp electrophysiology. Anapparent steady-state block of the hERG current was achieved
260 M. Deacon et al. / Journal of Pharmacological and Toxicological Methods 55 (2007) 255–264
within 10–15 min of compound exposure and the magnitude ofinhibitionwas statistically significant at both 0.1μM(18.0±3.8%),and 1 μM (82.5±1.0%) (Fig. 1a). Furthermore, this blockade wasreversed upon perfusionwith saline, confirming that the C6–Cy3Bligand was a reversible hERG blocker. Therefore, further exper-imentswere carried out to explore the state-dependent properties ofhERG current blockade by this compound. A longer duration ofdepolarization significantly increased the magnitude of blockinduced by 0.1 μM C6–Cy3B, from 18.0±3.8% to 33.9± 2.1%(Fig. 1b). Furthermore, increases in the interpulse interval (whichcorrespond to decreased frequency of channel activation)significantly reduced the rate of onset of block, as evidenced bylarger time constants at interpulse intervals of 2.8 s (5.3±1.4 min)and 6.8 s (9.0±0.9 min) compared to 1.0 s (3.7±0.6 min, Fig. 1c).
Together these studies indicated that the C6–Cy3B labelled ligandwas a functional, reversible hERG ligand that displayed state-dependent block similar to the core dofetilide template (Snyders &Chaudhary, 1996).
However, despite this similarity, one key difference notedbetween this fluorescence labelled derivative and the originaldofetilide template was a discrepancy between the potencydetermined by patch clamp (IC50∼300 nM) and that deter-mined by radioligand binding (Ki∼7 nM). One possibleexplanation for this observation is a difference in the cellpermeability of the C6–Cy3B ligand compared to dofetilide. Toevoke functional blockade of the hERG channel in whole cells,compound diffusion across the plasma membrane is likely to berequired for access to the binding site in the pore of the channel.In contrast, this may not be the case for ligand binding studiesbecause they employ homogenized cell membrane vesicles thatare usually of variable composition and form (i.e. some vesiclesexposing the intracellular surface of the channel and someexposing the extracellular surface). Therefore, in an attempt toexplore this as an explanation for the discrepancy betweenbinding and functional effects of the C6–Cy3B ligand,membrane permeability was assessed.
The Caco-2 monolayer model is the most prevalent pre-clinical method employed in the pharmaceutical industry forevaluating drug absorption properties. The various mechanismsof compound transport have been well characterised in thissystem and the data obtained are routinely employed to identifycompounds with potential absorption properties, to rankcompounds with different transport properties and for predict-ing cellular uptake in vivo (Artursson, Palm, & Luthman, 2001;Shah, Jogani, Bagchi, & Misra, 2006). Therefore, the Caco-2assay was selected to study the membrane permeability of theC6–Cy3B ligand compared to dofetilide. Data generated usinghuman Caco-2 cells demonstrated that the apparent permeabil-ity (Papp) of the fluorescent ligand was very low (b1×10−6 cm/s)compared to the moderately permeable dofetilide (6.7×10−6 cm/s) (data not shown). This supported the hypothesis that the re-duced potency of the C6–Cy3B dofetilide ligand in the patch
Fig. 2. Configuration of the C6–Cy3B ligand f luorescence polarization assay.(a) Changes in polarization were monitored under different conditions todetermine whether they were a specific consequence of binding to hERG. Datapresented is the result of a single experiment where the ligand concentration was5 nM and membrane quantity was 5 μg. Each condition was represented by 32replicates. Polarization changes in the presence of hERG expressing HEK 293cell membranes (±excess of unlabelled competitor, E-4031) were compared tothe ligand alone and to those generated in the presence of HEK 293 cellmembranes expressing an alternative channel (null). Based on this data, the C6–Cy3B ligand binds specifically to hERG-expressing membranes. (b) Changes inpolarization were monitored over a range of membrane protein concentrations attwo different ligand concentrations (2 nM and 5 nM) to determine theappropriate concentrations required to maximize the assay signal. Datapresented is the result of a single experiment. Based on these results, theconditions selected were 2 nM ligand and 3 μg membrane protein. (c)Competition experiments with dofetilide were performed under different DMSOconcentrations, using 2 nM ligand and 3 μg membrane protein, to determine theimpact of this compound solvent upon the assay. Data presented is the result of asingle experiment where each response effect curve was generated with 11points, each point being the result of duplicate wells. Based on this data, theassay was considered tolerant to 3% DMSO.
Fig. 3. Determination of the C6–Cy3B ligand KD at hERG expressed in HEK 293 cells. Changes in polarization were monitored over time to determine the observedon rate (Kobs ) and dissociation rate constant (Koff ) of the ligand at the hERG channel. The data presented are the results from a single experiment using 2 nM ligandand 3 μg membrane protein, with 8 replicates per point, and are representative of the nine experiments performed. Data were analyzed using GraphPad prism todetermine the rates and these values were converted to a KD via the following equations: KD=Koff /Kon and Kon= (Kobs/[L])− (Koff /[L]). The KD value determined wassimilar to the Ki obtained from competition studies in the [3H]-dofetilide SPA assay for this ligand.
Table 2Comparison of the hERG affinities obtained for references compounds usingbinding and functional methods
261M. Deacon et al. / Journal of Pharmacological and Toxicological Methods 55 (2007) 255–264
clamp assay may be due to poor access of the compound to thereceptor site in whole cells. In contrast, it is likely that dofetilidehas sufficient permeability (although still only moderate) toachieve a level of blockade consistent with that obtained inbinding studies.
3.2. Development of a f luorescence polarization hERG bindingassay
To establish whether a specific polarization change could beobserved, the ligand was tested in a preliminary 384-well assaywith membranes produced from either human embryonickidney (HEK) 293 cells expressing the recombinant hERGchannel or from HEK 293 cells expressing an amine transporter(representing a null control). As expected, changes in therelative polarization values were observed in the presence ofhERG expressing membranes and not in the presence of the nullcontrol (Fig. 2a), indicating that the ligand was binding spec-ifically to the hERG channel. It is important to note at this point,that a polarization change could only be observed in cell mem-branes containing high levels of the hERG protein. The hERGexpression level in the cells employed was determined to beN10 pmol/mg, however using other cell lines with expressionlevels in the order of 3–5 pmol/mg (or less) yielded no signal(data not shown). This effect is not entirely unexpected as theseassays require a careful balance between the total ligand addedand the total receptors added. As both the free and the boundfluorescent ligand are contributing to the fluorescence intensity,in order to detect reproducible changes in polarization, a sig-nificant proportion of the total ligand (usually ∼20%) is re-quired to be bound to the membranes (Allen et al., 2000;Owicki, 2000). More specifically, as fluorescent ligand con-centrations are increased, assay quality is generally decreaseddue to the large proportion of free ligand and thus the smallchange in polarization (Banks & Harvey, 2002). In addition, asmembrane concentrations are increased the assay sensitivity isnot only decreased due to the effects of increased ligand deple-
tion, but the assay quality can also be decreased due to increasedlight scatter produced by the membrane vesicles (Owicki,2000). Therefore, in this assay it is likely that such a high levelof hERG expression was required to achieve an appropriateproportion of receptor/ligand complex, without any interferencefrom membrane-induced light scatter. Such high expressionlevels were not required for the scintillation proximity assay asfor this technique only bound ligand contributes to the signalgenerated and therefore light scatter interference from mem-branes does not negatively influence the data.
Having confirmed that a change in polarization could beobserved, further studies were aimed at optimising the ligandconcentration, membrane protein concentration and the buffersystem (for example: tolerance to compound solvent). Membrane
262 M. Deacon et al. / Journal of Pharmacological and Toxicological Methods 55 (2007) 255–264
optimisation studies were performed with protein concentrationsranging from 7 to 0.5 μg using both 2 and 5 nM ligand (Fig. 2b).Changes in the relative polarization for bound and free ligandwere observed at all membrane concentrations tested for bothligand concentrations. However, whilst there was little differencebetween the signal window obtained with 5 nM or 2 nM ligand athigher protein concentrations (3–7 μg), at lower proteinconcentrations (b1 μg) the signal window was b100 mP using5 nM dye. Therefore, to minimise the impact of small day-to-dayexperimental variations and differences in membrane batches onthe data, all further studies employed 2 nM ligand and 3 μgprotein/well. In order to evaluate the dimethylsulphoxide(DMSO) tolerance and signal stability, dofetilide dose responsecurves were performed at a range of DMSO concentrations (0–5%). Data indicate that the polarization values up to 1% DMSO
Fig. 4. Correlation of the pharmacological profile of hERG generated by FPcompared to alternative binding and patch-clamp methods. (a) pKi values for arange of literature reference and Pfizer compounds were generated using either theFP or SPA technologies. Competition experiments were performed using 11 pointsin duplicate for each curve. Data presented are the average of n≥3 experiments,with confidence limits for the linear regression highlighted by the dashed lines. Ther 2 value is included, alongwith slope±S.D. These data illustrate that there is a goodcorrelation between the two data sets. (b) pKi values for a range of known andPfizer compounds were generated with the FP assay via competition experiments(11 points in duplicate for each curve). These data have been compared to the pIC50
generated in patch-clamp experiments. Data presented are the average of n≥3experiments, with confidence limits for the linear regression highlighted by thedashed lines. The r 2 value is included, along with slope±S.D. These data illustratethat there is a good correlation between the two data sets.
were very similar and only small effects on the dofetilide IC50 orsignalwindowwere observed at 3 and 5%DMSO (Fig. 2c). Thesedata remained consistent after a 12-h incubation, although thesignal window was reduced by ∼15% under all the conditions(data not shown).
Prior to a full pharmacological evaluation, the Kd for theCy3B-C6–dofetilide ligand was determined to ensure that datacould be converted from IC50 to Ki. Although saturation studiescan be performed readily to determine the Kd and Bmax valuesin radioligand binding studies, such experiments are morecomplex for FP assays. Whilst a model has been developed toensure that such experiments are analyzed and interpretedappropriately (Prystay, Gosselin, & Banks, 2001), associationand dissociation time courses were utilized here as analternative method of obtaining a direct measure of the affinityof the C6–Cy3B ligand for the hERG channel (Fig. 3). Thesestudies confirmed that the C6–Cy3B ligand bound revers-ibly to the hERG membranes. In addition, the affinity value of2.75±0.0003 nM that was calculated from the rates obtained(Kobs=0.001±0.0006 s−1, Kon=0.00023±0.0001 s−1 nM−1,Koff=0.00055±0.00047 s−1), corresponded well with the Ki
value obtained from [3H]-dofetilide competition studies(∼7 nM). The measured Kd values confirmed that the con-ditions employed in the assay (2 nM labelled ligand) werecommensurate with a sensitive assay, with the ligand concen-tration being close to Kd and therefore, the IC50 values beingclose to their affinity constants, as described by the Cheng–Prusoff IC50 to Ki relationship (Cheng & Prusoff, 1973).
3.3. Pharmacological validation of FP hERG assay
To validate the 384-well FP assay as a replacement for themore traditional hERG radioligand binding assays and the lowerthroughput automated patch assay, a selection of known hERGantagonists that encompass a wide range of affinities andstructural classes were tested. Comparison of the affinity con-stants obtained confirmed that data for these compounds werevery similar between our in-house [3H]-dofetilide SPA assay,high-throughput electrophysiology (PatchXpress) and the newFP assay (Table 2). Therefore, the evaluation was expanded toinclude an additional selection of Pfizer compounds, previouslyidentified as having affects in functional hERG assays. Thisrevealed that there was a high degree of correlation whencomparing data from the new FP binding assay to eitherhistorical binding data or functional data generated via auto-mated patch-clamp studies (Fig. 4a and b). Therefore, havingconfirmed the validity of using the new high-throughput FPbinding assay to predict for binding and functional effects at thehERG channel, additional experiments were performed toestablish the quality and reproducibility of this assay. Com-petition experiments were performed with dofetilide usingthree different batches of membranes across multiple days.From this data it could be seen that the dofetilide Ki fluctuatedby b3-fold between batches and days (varying between 4.5 and11.8 nM (data not shown). Furthermore the Z′ value, a sta-tistical parameter employed in the development of high-throughput screens to assess the signal window and noise of an
263M. Deacon et al. / Journal of Pharmacological and Toxicological Methods 55 (2007) 255–264
assay (Zhang, Chung, & Oldenburg, 1999), was also found tobe consistently above 0.6, demonstrating that this assay wasextremely robust.
4. Discussion
Achieving high-throughput hERG evaluation is key toeffectively reducing attrition early in the drug discovery pro-cess, as such, many assays have been investigated to enable this.The “high-throughput” patch assays currently available e.g.IonWorks and PatchXpress (Bridgland-Taylor et al., 2006;Dubin et al., 2005) are able to provide kinetic information andare functional in nature, thereby representing the closest ap-proximation to the gold standard assay format of manual patchclamp. However, even with these higher throughput functionalassays, pharmacological anomalies can still be encountered dueto compound handling differences or cell concentrations(Bridgland-Taylor et al., 2006; Dubin et al., 2005). Furthermore,they represent a relatively expensive solution, they can display ahigh failure rate (Dubin et al., 2005) and whilst demonstrating amodest increase in throughput versus traditional electrophysi-ology studies they are not compatible with the 1000's ofcompounds produced per week during early lead optimisation.Therefore, many groups have turned to high-throughput bindingtechniques. Whilst both [3H]-astemizole (Chiu et al., 2004) and[3H]-dofetilide assays (Diaz et al., 2004) have been reported,both assays are 96-well systems and require filtration steps,making them both expensive and limited in throughput (albeithigher than traditional electrophysiology). A membrane poten-tial assay that employs a 384-well fluorescence techniquecapable of true high throughput has also been reported (Dorn etal., 2005). However, despite being a functional system, itdisplays significant differences in the pharmacological profilefor compounds. The IC50 values obtained were found to beunderestimated by as much as 100-fold, making the correlationbetween this method and patch clamp relatively weak and notsuitable for rank ordering compounds. In contrast, the 384-well,homogenous, fluorescent assay reported here displays apharmacological profile predictive of both other hERG bindingsystems but also functional patch-clamp. Whilst the requirementfor a high expressing cell line is a factor that needs to beconsidered, the reduced membrane concentrations required inthis assay format compared to traditional radiometric methodsand the overall safety benefits of a fluorescent technology versusa radioactive one make this a wise investment. Overall, thisnovel FP assay is amenable to the evaluation of thousands ofcompounds per day, whilst maintaining both cost effectivenessand a high degree of correlation with patch-clamp data, factorsthat cannot be achieved by any of the other currently availabletechniques. As a consequence this assay has been adoptedglobally at Pfizer for the evaluation of compounds in the hit andlead optimisation stages of the drug discovery process.
Acknowledgements
The authors wish to thank Gareth Waldron, Frank Stuhmeier,Phil Gribbon, Marcel de Groot and Rob Wallis for useful
discussion during the course of this study and the preparation ofthe manuscript. Especial thanks go to all of the cell productionteam members who worked hard to provide us with a newhigher expressing hERG cell line and to Allen Hilgers for theCaco-2 permeability study.
References
Allen, M., Reeves, J., & Mellor, G. (2000). High throughput fluorescencepolarization: A homogenous alternative to radioligand binding for cellsurface receptors. Journal of Biomolecular Screening, 5, 63−69.
Aronov, A.M. (2005). Predictive in silicomodelling for hERG channel blockers.Drug Discovery Today, 10, 149−155.
Artursson, P., Palm, K., & Luthman, K. (2001). Caco-2 monolayers inexperimental and theoretical predictions of drug transport. Advances inDrug Delivery Systems, 46, 27−43.
Banks, P., & Harvey, M. (2002). Considerations for using fluorescencepolarization in the screening of G protein-coupled receptors. Journal ofBiomolecular Screening, 7, 111−117.
Bridgland-Taylor, M. H., Hargreaves, A. C., Easter, A., Orme, A., Henthorn, D. C.,Ding, M., et al. (2006). Optimisation and validation of a medium-throughputelectrophysiology-based hERG assay using IonWorks™ HT. Journal ofPharmacological and Toxicological Methods, 54, 189−199.
Burke, T. J., Loniello, K. R., Beebe, J. A., & Ervin, K. M. (2003). Developmentand application of fluorescence polarization assays in drug discovery.Combinatorial Chemistry & High Throughput Screening, 6, 183−194.
Calderone, V., Testai, L., Martinotti, E., Del Tacca, M., & Breschi, M. C. (2005).Drug-induced block of cardiac HERG potassium channels and developmentof torsade de pointes arrhythmias: The case of antipsychotics. Journal ofPharmacy and Pharmacology, 57, 151−161.
Cheng, C. S., Alderman, D., Kwash, J., Dessaint, J., Patel, R., Lescoe, M. K., et al.(2002). A high-throughput hERG potassium channel function assay: An oldassay with a new look. Drug Development and Industrial Pharmacology, 28,177−191.
Cheng, Y., & Prusoff, W. H. (1973). Relationship between the inhibitionconstant (K1) and the concentration of inhibitor which causes 50 percentinhibition (I50) of an enzymatic reaction. Biochemical Pharmacology, 22,3099−3108.
Chiu, P. J. S., Marcoe, K. F., Bounds, S. E., Lin, C. -H., Feng, J. -J., Lin, A. et al.(2004). Validation of a [3H]-astemizole binding assay in HEK 293 cellsexpressing hERG K+ channels. Journal of Pharmacological Sciences, 95,311−319.
Curran, M. E., Splawski, I., Timothy, K. W., Vincent, G. M., Green, E. D., &Keating, M. T. (1995). A molecular basis for cardiac arrhythmia: hERGmutations cause long QT syndrome. Cell, 80, 795−803.
Curtis, M. J. (2004). Abstracts from the 3rd Annual Safety Pharmacology Society(SPS) Meeting, Noordwijk, The Netherlands, September 29–30, 2003. Jour-nal of Pharmacological and Toxicological Methods, 49(3), 217−239.
De Ponti, F., Poluzzi, E., Cavalli, A., Racanatini, M., & Montanaro, N. (2002).Safety of non-anti-arrhythmic drugs that prolong the QT interval or inducetorsade de pointes, an overview. Drug Safety, 25, 263−286.
Diaz, G. J., Daniell, K., Leitza, S. T., Martin, R. L., Su, Z., McDermott, J. S. et al.(2004). The [3H]-dofetilide binding assay is a predictive screening tool forhERG blockade and proarrhythmia: Comparison of intact cell and membranepreparations and effects of altering [K+]o. Journal of Pharmacological andToxicological Methods, 50, 187−199.
Dorn, A., Hermann, F., Ebneth, A., Bothmann, H., Trube, G., Kristensen, K. et al.(2005). Evaluation of a high-throughput f luorescence assay method forhERG potassium channel inhibition. Journal of Biomolecular Screening, 10,339−344.
Dubin, A. E., Nasser, N., Rohrbacher, J., Hermans, A. N., Marrannes, R.,Grantham, C. et al. (2005). Identifying modulators of hERG channel activityusing the PatchXpress Planar Patch Clamp. Journal of BiomolecularScreening, 10, 168−181.
Fermini, B., & Fossa, A. A. (2003). The impact of drug-induced QT intervalprolongation on drug discovery and development. Nature Reviews DrugDiscovery, 2, 439−447.
264 M. Deacon et al. / Journal of Pharmacological and Toxicological Methods 55 (2007) 255–264
Finlayson, K., Turnbull, L., January, C. T., Sharkey, J., & Kelly, J. S. (2001).[3H]-Dofetilide binding to hERG transfected membranes: A potential highthroughput pre clinical screen. European Journal of Pharmacology, 430,147−148.
Gribbon, P., & Sewing, A. (2003). Fluorescence readouts in HTS: No gainwithout pain? Drug Discovery Today, 8, 1035−1043.
Hamill, O. P., Marty, A., Neher, E., Sakmann, B., & Sigworth, F. J. (1981).Improved patch-clamp techniques for high-resolution current recording fromcells and cell-free membrane patches. Pf lugers Archiv: European Journal ofPhysiology, 391(2), 85−100.
Harris, A., Cox, S., Burns, D., & Norey, C. (2003). Miniaturization off luorescence polarization receptor-binding assays using CyDye-labelledligands. Journal of Biomolecular Screening, 8, 410−420.
Haverkamp, W., Breithardt, G., Camm, A. J., Janse, M. J., Rosen, M. R.,Antzelevitch, C. et al. (2000). The potential for QT prolongation and pro-arrhythmia by non-anti-arrhythmic drugs: Clinical and regulatory implica-tions. Report on a policy conference of the European Society of Cardiology.Cardiovascular Research, 47, 219−233.
Kates, S. A., Carpino, L. A., & Albericio, F. (1996). New coupling techniquesfor solid phase peptide synthesis, peptides: Chemistry, structure and biology( pp. 893). Mayf lower Scientific.
Kola, I., & Landis, J. (2004). Can the pharmaceutical industry reduce attritionrates? Nature Reviews Drug Discovery, 3, 711−715.
Netzer, R., Bischoff, U., & Ebneth, A. (2003). HTS techniques to investigate thepotential effects of compounds on cardiac ion channels at early-stages ofdrug discovery. Current Opinion in Drug Discovery & Development, 6,462−469.
Owicki, J. C. (2000). Fluorescence polarization and anisotropy in highthroughput screening: Perspectives and primer. Journal of BiomolecularScreening, 5, 297−306.
Pearlstein, R., Vaz, R., & Rampe, D. (2003). Understanding the structure–activity relationship of the human ether-a-go-go related gene cardiac K+
channel. A model for bad behaviour. Journal of Medicinal Chemistry, 46,2017−2022.
Prystay, L., Gosselin, M., & Banks, P. (2001). Determination of equilibriumdissociation constants in f luorescence polarization. Journal of BiomolecularScreening, 6, 141−150.
Redfern,W. S., Carlsson, L., Davis, A. S., Lynch,W. G.,Mackenzie, I., Palethorpe,S. et al. (2003). Relationships between preclinical cardiac electrophysiology,clinical QT interval prolongation and torsade de pointes for a broad rangeof drugs: Evidence for a provisional safety margin in drug development.Cardiovascular Research, 58, 32−45.
Sanguinetti, M. C., Jiang, C., Curran, M. E., & Keating, M. T. (1995). Amechanistic link between an inherited and an acquired cardiac arrhythmia:HERG encodes the IKr potassium channel. Cell, 81, 299−307.
Shah, P., Jogani, V., Bagchi, T., & Misra, A. (2006). Role of Caco-2 cellmonolayers in prediction of intestinal drug absorption. BiotechnologyProgress, 22(1), 186−198.
Snyders, D. J., & Chaudhary, A. (1996). High aff inity open channel block bydofetilide of hERG expressed in a human cell line. Molecular Pharmacol-ogy, 49, 949−955.
Tang, W., Kang, J., Wu, X., Rampe, D., Wang, L., Shen, H., et al. (2001).Development and evaluation of high throughput functional assay methodsfor hERG potassium channel. Journal of Biomolecular Screening, 6,325−331.
Vandenberg, J. I., Walker, B. D., & Campbell, T. J. (2001). hERG K+ channels:Friend and foe. Trends in Pharmacological Sciences, 22, 240−246.
Webster, R., Leishman, D., & Walker, D. (2002). Towards a drug concentrationeffect relationship for QT prolongation and torsade de pointes. CurrentOpinion in Drug Discovery and Development, 5, 116−126.
Wood, C., Williams, C., & Waldron, G. J. (2004). Patch clamping by numbers.Drug Discovery Today, 9, 434−441.
Zaman, G. J. R., Garritsen, A., De Boer, Th., & Van Boeckel, C. A. A. (2003).Fluorescence assays for high-throughput screening of protein kinases.Combinatorial Chemistry & High Throughput Screening, 6, 313−320.
Zhang, J., Chung, T. D. Y., & Oldenburg, K. R. (1999). A simple statisticalparameter for use in evaluation and validation of high throughput screeningassays. Journal of Biomolecular Screening, 4, 67−73.
Zheng, W., Spencer, R. H., & Kiss, L. (2004). High throughput ion channel drugdiscovery. Assay and Drug Development Technologies, 2, 543−552.
Zhou, Z., Gong, Q., Ye, B., Fan, Z., Makielski, J. C., Robertson, G. A. et al.(1998). Properties of HERG channels stably expressed in HEK 293 cellsstudied at physiological temperature. Biophysical Journal, 74(1), 230−241.
