An evaluation of fluorescence polarization and lifetime discriminated polarization for high...

An evaluation of fluorescence polarization and lifetimediscriminated polarization for high throughput screening of

serine/threonine kinases

Ann Fowler,1 Denise Swift, Emma Longman, Anne Acornley, Paul Hemsley,Dave Murray, John Unitt, Ian Dale,* Elaine Sullivan, and Martin Coldwell

Lead Generation Biology, AstraZeneca R & D Charnwood, Loughborough, LE11 5RH, UK

Received 29 October 2001

Abstract

We used two kinases, c-jun N terminal kinase (JNK-1) and protein kinase C (PKC), as model enzymes to evaluate the potential

of fluorescence polarization (FP) for high-throughput screening and the susceptibility of these assays to compound interference. For

JNK-1 the enzyme kinetics in the FP assay were consistent with those found in a ½c-33P�ATP filter wash assay. Determined pIC50s

for nonfluorescent JNK-1 inhibitors were also consistent with those found in the filter wash assay. In contrast, fluorescent com-

pounds were found to interfere with the JNK-1 FP assay, appearing as false positives, defined by their lack of activity in the filter

wash assay. We also developed a second assay using a different kinase, protein kinase C, which was used to test a 5000 compound

diversity set. As for JNK-1, interference from fluorescent compounds caused a high false positive rate. The Molecular Devices

Corporation �FLARe� instrument is capable of discriminating between fluorophores on the basis of their fluorescence (excited state)

lifetime, and may assist in reducing compound interference in fluorescent assays. In both model FP kinase assays described here

some, although not complete, reduction in interference from fluorescent compounds was achieved by the use of FLARe. � 2002

Elsevier Science (USA). All rights reserved.

Keywords: Fluorescence polarization; Fluorescence lifetime; HTS; Kinases; FLARe

Existing assays for high-throughput screening (HTS)2

of kinases include scintillation proximity assay (SPA),homogeneous time resolved fluorescence (HTRF)/Lance, dissociation enhanced lanthanide fluorescenceimmunoassay (DELFIA), Alphascreen, and others.Fluorescence polarization (FP) is proposed to offerseveral advantages for HTS assays: homogeneity, suit-ability for miniaturization, precision, and resistance to

color quenching being the most significant. In order totest these claims we developed an FP assay for thethreonine kinase c-jun N-terminal kinase 1 (JNK-1),which was miniaturised and automated in both 384-wellsmall volume and 1536-well format. The enzyme kineticsand patterns of inhibition by known kinase inhibitorsdetermined in the JNK-1 FP assay were also comparedwith the results from a ½c-33P�ATP filter wash assay,which was used as a ‘‘benchmark’’ assay.

The principle of the JNK-1 FP assay is shown in Fig.1. JNK-1 transfers phosphate to threonine 71 of theprotein substrate ATF-2. Phosphorylated ATF-2 pro-tein and the fluorescein-labeled phospho-peptide tracer(which encompasses the sequence around the phospho-rylation site of ATF-2) then compete for binding to theanti-phospho-ATF-2 antibody. The higher the levels ofphospho-ATF-2 present, the more tracer is left free insolution, resulting in lower polarization of the emitted

Analytical Biochemistry 308 (2002) 223–231

www.academicpress.com

ANALYTICALBIOCHEMISTRY

* Corresponding author. Fax: +44-0-1509-645555.1 Present address: Discovery Group, Roche Vitamins Ltd., CH-4070,

Basel, Switzerland.2 Abbreviations used: HTS, high-throughout screening; SPA, scintil-

lation proximity assay; HTRF, homogenous time-resolved fluores-

cence; DELFIA, dissociation enhanced lanthanide fluorescence

immunoassay; FP, fluorescence polarization; FLARe, Fluorescence

Lifetime Assay Repertoire; LDP, lifetime discriminated polazarization;

Mops, 4-morpholinepropanesulfonic acid; DTT, dithiothreitol;

DMSO, dimethyl sulfoxide.

0003-2697/02/$ - see front matter � 2002 Elsevier Science (USA). All rights reserved.

PII: S0003 -2697 (02 )00245-2

light (lower mP). The principle of the Molecular DevicesSTX-1 PKC assay kit, which was used for developmentof the PKC assay, is identical, except that a peptidesubstrate (RFARKGSLRQKNV) is used together withthe monoclonal antibody 2B9, which is specific for thissequence.

Although the ratiometric nature of the fluorescencepolarization measurement should have increased its ro-bustness to interference from color quenching, we foundthat the JNK-1 FP assay was subject to interferencefrom fluorescent test compounds, which have their ownintrinsic polarization, resulting in ‘‘false positives.’’False positives are compounds that appear active in anassay but are inactive against the target kinase whentested in other assay formats, in this case the ½c-33P�ATPfilter wash assay. In order to investigate whether thisproblem was specific to the JNK-1 assay we subse-quently developed a second FP kinase assay for proteinkinase C, which was used to test a 5000 compound di-versity set.

Elimination of false positives requires developmentand execution of secondary assays, in this case, a½c-33P�ATP filter wash assay, to identify the nongenuineactives, thus increasing the costs and timescale of thelead discovery process. An alternative approach is to usefluorescence lifetime to distinguish between light emittedby the fluorescent label and light emitted by fluorescenttest compounds (‘‘lifetime discrimination’’), thus re-ducing interference from fluorescent library compoundsin the primary HTS assay [1]. Fluorescence lifetime, themean interval between absorption and emission, is asfundamental a characteristic of a fluor as its excitationand emission wavelengths. The test compounds en-countered in HTS are assumed to have shorter fluores-cence lifetimes than commonly used labels, such as thefluorescein-labeled tracers used in the assays presentedhere. The Molecular Devices Corporation prototypeinstrument �FLARe� [2] (fluorescence lifetime assayrepertoire) was designed as a high-throughput frequencydomain lifetime fluorimeter capable of discriminatingbetween fluorescent molecules on the basis of their

fluorescence lifetime. In this study we evaluated theability of the FLARe lifetime discriminated polarization(LDP) mode, in which the FLARe software attempts toreject fluorescence contributed by short lifetime sourcesfrom the polarization measurement, to exclude fluores-cence contributed by test compounds in the kinase FPassays and hence reduce compound interference.

Materials and methods

Materials

JNK-1 assaysThe following were used for JNK-1 assays: anti-

phospho-ATF-2 STK antibody (New England Biolabs);fluorescein-labeled tracer (Molecular Devices Corpora-tion); activated human JNK-1 enzyme a1 isoform [3]and ATF-2 protein (Division of Signal TransductionTherapy, School of Life Sciences, Dundee University);384-well black HE v B plates (Molecular Devices Cor-poration); 1536-well black plates (Greiner); ½c-33P�ATP(Amersham); Microscint O (Packard); Filter PlateUnifilter 96 GF/C (Packard). Kinase inhibitors wereobtained from Calbiochem or Biomol. Other reagentswere obtained from Sigma. Enzyme reaction buffer forFP assays was TKXtra buffer (Molecular Devices Cor-poration) + 10mM MgCl2 and 0.1% b-mercaptoethanol.Enzyme reaction buffer for the filter wash assay was50mM Mops, pH 7.2, + 0.1% b-mercaptoethanol.

PKC assaysSTX-1 Kit (Molecular Devices Corporation); Human

PKC enzyme (AstraZeneca Biotechnology Laboratory);384-well black Optiplates (Packard); ½c-33P�ATP(Amersham); RFARKGSLRQKNV peptide substratefor the filter wash assay (NeoSystems); Microscint O(Packard); P81 filter plate (Whatman). Other reagentswere obtained from Sigma. Enzyme reaction buffer forPKC assays was 50mM Hepes, pH 7.6, 10mM MgCl2,1mM dithiothreitol (DTT), 0.01% Triton X-100 (+0.1%casein in FP assays).

Methods

JNK-1 fluorescence polarization assayCompound in 4% DMSO (1% final in assay) was

mixed with JNK-1 enzyme (20 nM final) and incubatedfor 20min at room temperature. ATF-2 (400 nM final)/ATP solution (1000 nM final) was added and the reac-tion incubated for 40min at room temperature. EDTA(10mM final)/tracer (0.2 nM final) solution was addedto stop the reaction, followed by anti-phospho-ATF-2antibody (0.2 lg/ml final concentration), and the plateincubated for 1 h at room temperature. Total assayvolume¼ 10 ll. Plates were read on an Analyst or

Fig. 1. Principle of the JNK-1 fluorescence polarization assay.

224 A. Fowler et al. / Analytical Biochemistry 308 (2002) 223–231

Acquest fluorescence reader (Molecular Devices Cor-poration), k excitation¼ 485 nm, k emission¼ 530 nm.

JNK-1 fluorescence polarization assay optimized forFLARe

The conditions were as above, except enzyme finalconcentration¼ 200 nM, tracer final concentration¼2 nM, antibody final concentration¼ 1.5 lg/ml. Plateswere read on the FLARe instrument (Molecular De-vices Corporation), k excitation¼ 475 nm, k emission¼530 nm.

JNK-1 filter wash assayCompound in 10% DMSO (1% final in assay) was

mixed with JNK-1 enzyme (40 nM final) and incubatedfor 20min at room temperature. ATF-2 (200 nM final)/ATP (1000 nM final)/MgCl2 solution (10mM final)was added, plus 250,000 DPM/well ½c-33P�ATP, andthe reaction incubated for 30min at room temperature.A 5%(w/v final) trichloroacetic acid plus 10mM ATPwas added to stop the reaction. Total assay vol-ume¼ 75 ll. The assay was filtered through a GF/CUnifilter plate and washed with milli-Q water. Theplates were dried, scintillant added, and plates countedon a Packard Top Count NXT microplate scintillationcounter.

PKC fluorescence polarization assayCompound in 5% (1% final in assay) DMSO was

mixed with PKC enzyme (0.2 nM final) and incubatedfor 10–15min at room temperature. ATP (1000 nM fi-nal)/substrate (200 nM final)/phosphatidylserine (20 lg/ml final)/phorbol 12-myristate 13-acetate (10 nM final)solution was added and the reaction incubated for 60–65min at room temperature. EDTA (40mM final)/tracer (1 nM final) solution was added to stop the re-action, followed by the 2B9 antibody (1 in 400 final) andthe plate incubated for 4 h at room temperature. Totalassay volume¼ 50 ll. Plates were read on an Analyst orAcquest fluorescence reader (Molecular Devices Cor-poration), k excitation¼ 485 nm, k emission¼ 530 nm.

PKC fluorescence polarization assay optimized forFLARe

The conditions were as above except enzyme finalconcentration¼ 1 nM, tracer final concentra-tion¼10 nM, antibody final dilution¼ 1 in 125. Plateswere read on the FLARe instrument (Molecular DevicesCorporation), k excitation¼ 485nm, k emission¼ 530 nm.

PKC filter wash assay methodCompound in 10% DMSO (1% final in assay) was

mixed with ATP (1000 nM final)/substrate (200 nMfinal)/phosphatidylserine (20 lg/ml final)/phorbol 12-myristate 13-acetate (10 nM final) solution plus 222,000DPM/well ½c-33P�ATP was added/well, followed by PKC

enzyme (0.5 nM final) and the reaction incubated for30min at room temperature. A 0.45%(v/v final) phos-phoric acid plus 4mM ATP was added to stop the re-action. Total assay volume¼ 125 ll. The assay wasfiltered through a p81 filter plate and washed with milli-Q water. The plates were dried, and scintillant wasadded and counted on a Packard Top Count NXTmicroplate scintillation counter.

Results

JNK-1 Fluorescence polarization assay development andenzyme kinetics

Initial experiments were carried out to determinesuitable concentrations for the fluorescein-labeledtracer and anti-phospho-ATF-2 antibody in the assay.A tracer concentration of 0.2 nM gave �0.5 millioncounts per second (cps) when read on the Analystfluorescence reader and �0.25 million cps when readon the Acquest fluorescence reader in both the paralleland perpendicular planes, with an intrinsic polarizationof 60mP for the free tracer. An antibody concentrationof 0.2mg/ml resulted in a mP change [(tracerþ anti-body)� (free tracer)] of 130mP (data not shown).

Next, the ATP and ATF-2 substrates were titrated inorder to determine Km values in the FP assay. Repre-sentative experiments are shown in Fig. 2. As the con-centration of either substrate was increased the amountof product increased; hence the proportion of tracerbound to the antibody decreased, and the degree ofpolarization correspondingly decreased. The Kms weredetermined to be 1200nM� 300 (SE, n ¼ 3) for ATPand 370nM� 30 (SE, n ¼ 3) for ATF-2. These valuesare similar to those determined in a ½c-33P�ATP filterwash assay (1700 and 160 nM for ATP and ATF-2,respectively; data not shown).

Under the final assay conditions, with both substratesat their Km concentrations and using 20 nM JNK-1

Fig. 2. Km determinations for ATP (a) and ATF-2 (b) in the JNK-1

fluorescence polarization assay. Values are means�SE of triplicate

wells in one experiment, representative of three experiments.

A. Fowler et al. / Analytical Biochemistry 308 (2002) 223–231 225

kinase, the reaction was linear for approximately 40min,which resulted in a mP change (relative to pre-stoppedcontrol) of �60mP. The assay was tolerant of 1%DMSO final concentration in the assay (data notshown).

JNK-1 FP assay automation and performance

The JNK-1 FP protocol as developed had a totalvolume of 10 ll, which is amenable to both 384-wellsmall volume and 1536-well plates. We compared thereproducibility of the assay in both formats, using MDC‘‘High Efficiency’’ (HE) 384-well plates (assembled on aBeckman Coulter Biomek 2000), and Greiner 1536-wellplates (assembled on the Cy-Bio Cybi-Well).

For the 384-well format assay read on the Analyst, a100-ms integration time resulted in coefficient of vari-ances (CVs) of 4.9 for the uninhibited (positive) controlwells and and 3.1% for the supramaximally inhibited(negative) control wells. This experiment had a windowof 55mP, giving a Z 0 factor [4] of 0.31. Increasing theintegration time to 500ms lowered the CVs to 2.8 and2.1% for the positive and negative control wells, re-spectively, and increased the Z 0 factor to 0.56. Thelonger integration time increased the counts acquiredfrom 50,000 to 250,000, but also increased the plate readtime from 4.5 to 9.5min. The 1536-well format assaywas read on the Acquest. To compensate for the lowersensitivity of the Acquest, the integration time/well wasincreased to 1000ms in order to acquire approximately250,000 counts in one integration period. The CVs forthe positive and negative control wells in this case were4.0 and 2.1, respectively. With a window of 62mP in thisassay the Z 0 factor was 0.60, and the plate read time was60min.

JNK-1 FP assay validation with compounds

A set of 50 kinase inhibitors previously described inthe literature was used to determine the reproducibilityof IC50 values determined from the JNK-1 FP assay.Fig. 3 shows composite inhibition curves from threeseparate experiments for the 2 most active compounds,staurosporine and K252a. The pIC50 values determinedfor 11 compounds found to be active in the JNK-1 FPassay are shown in Table 1, alongside the pIC50 valuesdetermined for these compounds in the ½c-33P�ATP filterwash assay. A number of the compounds, highlighted inbold, were found to be active in the homogeneous FPassay but inactive in the heterogeneous ½c-33P�ATP filterwash assay. The common characteristics of these com-pounds were that they were orange/yellow colored andappeared to be fluorescent, as shown in columns 4 and 5of Table 1. It therefore appeared that fluorescent com-pounds can interfere in FP assays, in this case appearingas false positives.

Re-optimization of the JNK-1 assay for the FLARereader

In order to evaluate whether lifetime discriminationcould reduce the false positive rate in the two FP kinaseassays, the assays required re-optimization for theFLARe reader, which has lower sensitivity than theAnalyst or Acquest. In the JNK-1 assay the tracer wasused at a final concentration of 2 nM, i.e., 10-fold higherthan in the original assay. The anti-ATF-2 antibody andJNK-1 enzyme were then titrated to give a mP change[(tracer + antibody) ) (free tracer)] of 130mP and anassay window [(tracer + antibody no enzyme)� (tracer+ antibody+ enzyme)] of 60mP, as in the original assay(data not shown). Finally, the assay was validated usingthe kinase inhibitor staurosporine, which gave a pIC50

of 6.6, compared to 6.5 in the original assay (Fig. 3).

Application of the FLARe to the JNK-1 FP assay

The 11 compounds found to be active in the JNK-1FP assay were retested in the FLARe optimized assay.Two model short lifetime fluors, tartrazine and rosebengal [2,5] were also included. Data from one experi-ment, representative of three experiments, are shown inFig. 4. The FP and LDP results were obtained from thesame assay, the only difference being the attempted ex-clusion of fluorescence from short lifetime sources in theLDP measurement. In the FP measurement all the

Fig. 3. Inhibition curves for staurosporine () and K252a (j) deter-

mined in the JNK-1 fluorescence polarization assay. Values are

means�SE of three experiments. Inset: Staurosporine inhibition curve

determined in the JNK-1 FLARe optimized assay. Values are means

� SE of duplicate wells in one experiment.


compounds were apparently active, including tartrazineand rose bengal, and many of the compounds gave ap-parent percentage inhibitions of much greater than100%. The results from the LDP measurement clearlydiffered from the FP results, and are described in moredetail below.

Four classes of compounds were identified within the11 tested. The first class consisted of genuine inhibitorsof the JNK-1 enzyme (as assessed by their activity in the½c-33P�ATP filter wash assay) that did not appear to befluorescent or were weakly fluorescent. These includedSB203580, staurosporine, K252a, and apigenin. Fig. 5ashows that there was little or no change in the potencyof these compounds between the FP and LDP mea-surements. The second class consisted of genuine in-hibitors of the JNK-1 enzyme which were alsofluorescent. These included damnacanthal and ST638.Fig. 5b shows that both these compounds remainedactive in the LDP measurement. Damnacanthal showed

no decrease in pIC50 between the LDP and FP mea-surements, whereas ST638, which is more strongly in-terfering, did show some decrease in pIC50 in LDP. Thethird class contained the model short lifetime fluorstartrazine and rose bengal. Fig. 5c shows that the in-terference caused by tartrazine at high concentrationswas completely rejected by LDP, while the interferencecaused by rose bengal was partially rejected. Finally, thefourth class consisted of the fluorescent compounds thatappear as false positives in the JNK-1 FP assay. Fig. 5dshows that the interference due to one compound, tyr-phostin A9, was effectively rejected, and the interferencedue to two compounds, AG825 and PD 169316, wasconsiderably reduced, while bisindolylmaleimide I andII were only slightly affected (PD 169316 was included inthis category because, when retested at 100 lM in thethree FLARe experiments, it was clearly fluorescent,having a fluorescence ratio of 1.3 at this concentration inthis experiment).

Table 1

pIC50 values for standard inhibitors determined in the JNK-1 fluorescence polarization and 33P ATP filter wash assays

Compound Mean pIC50

(n ¼ 3)� SE

pIC50 filter-binding assay

(n ¼ 1)

Color of compound Fluorescence ratio

(n ¼ 3)�SEa

SB203580 5:5� 0:5 5.2 Colorless 1:03� 0:03

Staurosporine 6:5� 0:1 6.5 Colorless 0:97� 0:03

Damnacanthal 5:7� 0:4 5.5 Colorless 1:33� 0:08

Tyrphostin A9 5:2� 0:1 <50% Activity at 10 lM Yellow 1:35� 0:11AG825 5:5� 0:1 <50% Activity at 10 lM Yellow 1:26� 0:05

Bisindolylmaleimide II 5:7� 0:1 <50% Activity at 10 lM Orange 1:98� 0:13

K252a 7:2� 0:1 7.5 Colorless 0:96� 0:01

Bisindolylmaleimide I 5:9� 0:1 <50% Activity at 10 lM Orange 1:79� 0:03ST638 5:9� 0:1 5.7 Yellow 1:27� 0:06

Apigenin <5.02 <50% Activity at 10lM Colorless 1:08� 0:04

PD 169316 <5.05 <50% Activity at 10 lM Yellow 1:03� 0:03a Fluorescence ratio is defined as the ratio of the fluorescence intensity (cps) of a well containing 10lM compound to the fluorescence intensity

(cps) of a well containing the fluorescein-labeled tracer but no compound. Values >1 therefore may indicate the presence of a fluorescent compound.

kex ¼ 485nm; kem ¼ 535nm.

Fig. 4. Inhibition curves determined from the fluorescence polarization and lifetime discriminated polarization measurements for the 11 compounds

which appeared active in the JNK-1 fluorescence polarization (Analyst) assay, plus two model short lifetime fluors, tested in the FLARe optimized

JNK-1 FP assay. Results from the fluorescence polarization measurement are shown in the left panel and results from the lifetime discriminated

polarization measurement are shown in the right panel. Values are means of duplicate wells in one experiment, representative of three experiments.

Compounds: tartrazine (1), SB203580 (2), staurosporine (3), damnacanthal (4), tyrphostin A9 (5), AG825 (6), bisindolylmaleimide II (7), K252a (8),

bisindolylmaleimide I (9), ST638 (10), apigenin (11), PD 169316 (12), rose bengal (13).


The results for all 13 compounds are summarized inFig. 6, which shows the percentage inhibition caused byeach compound at a test concentration of 10 lM inboth FP and LDP measurements. The genuine non-fluorescent JNK-1 inhibitors all lie on or close to theline of unity, as does the fluorescent genuine JNK-1

inhibitor damnacanthal. The model short lifetime fluorsand the fluorescent interfering compounds (the falsepositives) all gave lower apparent percentage inhibi-tions in the LDP measurement compared to FP, indi-cating that their interference was reduced to somedegree.

Fig. 5. Inhibition curves determined from the fluorescence polarization and lifetime discriminated polarization measurements for various fluorescent

and non-fluorescent compounds tested in the FLARe optimized JNK-1 FP Assay. Results from the fluorescence polarization measurement are

shown in the left panel and results from the lifetime discriminated polarization measurement are shown in the right panel. Values are means of

duplicate wells in one experiment, representative of three experiments. Compounds: SB203580 (j), staurosporine (s), K252a (N), apigenin (/),damnacanthal (+), ST638 (}), Tartrazine (r), Rose Bengal (�), tyrphostin A9 (d), AG 825 (.), bisindolylmaleimide II (—), bisindolylmaleimide I

(M), PD 169316 (�).


PKC FP assay validation with compounds

A 5000 compound diversity set was tested at 10 lMfinal concentration in the PKC assay. Fig. 7 shows thedistribution of the actives in the FP assay in two

separate experiments. Although the correlation betweenthe results obtained in the two experiments was good,the number of compounds determined to be active in theassay (the ‘‘hit rate’’) was high. More than 9% of thecompounds displayed activity when a lower percentageinhibition threshold of 50% was used, and greater than6% when a threshold of 75% was used. The apparentpercentage inhibitions caused by some of the com-pounds were also very much greater than 100%. Thesame set of compounds was also tested in a ½c-33P�ATPfilter wash assay, which gave a hit rate of 0.5% (data notshown) when a lower percentage inhibition threshold of50% was used, suggesting that many of the hits in thePKC FP assay were false positives. As for the JNK-1assay, there appeared to be a relationship between thefluorescence of a compound and its apparent activity inthe assay (Fig. 8), indicating that the high hit rate in thePKC FP assay may have been due, at least in part, tointerference from fluorescent compounds.

Application of FLARe to the PKC assay

The PKC FP assay was reoptimized for the FLARereader as described above for JNK-1. The same 5000

Fig. 7. Correlation of percentage inhibitions determined from the

fluorescence polarization measurement in two experiments for the 5000

compound diversity set tested in the analyst optimized PKC FP assay

(10lM test concentration).

Fig. 8. Correlation of percentage inhibition determined from the

fluorescence polarization measurement with the fluorescence index of

the well for the 5000 compound diversity set tested in the analyst

optimized PKC FP assay (10 lM test concentration). The fluorescence

index of a well is defined as follows: For each plate, the mean and

standard deviation of the control wells, which contain no compounds,

were determined. The fluorescence intensity of an individual com-

pound-containing well is then expressed as a number of standard de-

viations above or below the mean of the controls. Positive values

therefore represent wells with higher fluorescence intensity than the

mean of the control wells. The range of variation within the control

wells was generally less than 10 standard deviations.


fluorescence polarization and lifetime discriminated polarization

measurements for all 13 compounds tested in the FLARe optimized

JNK-1 FP assay (10lM test concentration). Values are means of du-

plicate wells in one experiment, representative of three experiments.

The line of unity is shown. Compounds: genuine nonfluorescent JNK-1

inhibitors (r), genuine fluorescent JNK-1 inhibitors (d), model short

lifetime fluors (M) & interfering fluorescent compounds (�).


compound diversity set as tested in the PKC FP assay(Fig. 8) was then retested in the FLARe optimized as-say. Fig. 9 shows the distribution of the actives in boththe FP and LDP measurements. Both sets of resultswere obtained from the same plates, read on theFLARe, the only difference being the attempted exclu-sion of fluorescence from short lifetime sources in theLDP measurement. It is evident from Fig. 9 that al-though some compounds still had greater than 100%apparent inhibition in the LDP measurement, there weremany compounds which had lower apparent percentageinhibition in the LDP measurement than in FP, whilevery few had higher percentage inhibition in LDP thanFP. A staurosporine IC50 curve was also generated inthis experiment, to confirm that there was no decrease inpotency of a nonfluorescent genuine inhibitor in theLDP measurement as compared to FP. In both mea-surements the pIC50 value determined was 8.1 (Fig. 10).

The hit rates in the Analyst/Acquest optimized FP as-say, the modified FLARe assay (FP and LDP mea-surements), and the ½c-33P�ATP filter wash assay usinglower percent inhibition thresholds of 50 and 75% aresummarised in Table 2.

Summary and conclusions

In this work two model assays were used to evaluatethe suitability of FP assays for high-throughputscreening of kinases. The JNK-1 assay was used to as-sess the ease of FP assay development, miniaturization,and automation, and to compare the enzyme kineticsand pIC50s of a small number of compounds with thevalues determined in a ½c-33P�ATP filter wash assay. ThePKC assay was used to test a larger number of com-pounds. Both assays were also used to evaluate theability of FLARe to reduce interference from fluorescentcompounds in FP assays.


fluorescence polarization and lifetime discriminated polarization

measurements for the 5000 compound diversity set tested in the

FLARe optimized PKC FP assay (10lM test concentration). The line

of unity is shown. An expanded version, showing only 0–150% inhi-

bition, is inset.

Fig. 10. Inhibition curves determined in the fluorescence polarization

(d) and lifetime discriminated polarization measurements () for

staurosporine tested in the FLARe optimized PKC FP assay. Values

are means�SE of 32 wells in one experiment.

Table 2

Summary of results obtained in the fluorescence polarization and lifetime discriminated polarization measurements for the 5000 compound diversity

set tested in the PKC assays (10 lM test concentration)

Percentage

inhibition

threshold

Hit rate in 33P ATP

filter-binding assay

Hit rate in FP measurement

(original Analyst assay)

Hit rate in FP measurement

(modified FLARe assay)

Hit rate in LDP measurement

(modified FLARe assay)

50% 0.54% 9.10% 3.72% 2.84%

75% 0.15% 6.40% 2.31% 1.68%


The enzyme kinetics and sensitivity to inhibition inthe JNK-1 fluorescence polarization assay were consis-tent with those found in a ½c-33P�ATP filter wash assay.Furthermore, the Z 0 factor for the FP assay remainsacceptable when the volumes were reduced to 10 ll,which is compatible with both 384-well small volumeand 1536-well plates. Extended counting times were,however, critical to obtain low CVs in the FP mea-surement. Assays set up in a competitive format such asthe JNK-1 FP assay allow use of a physiological (i.e.,unlabeled protein) substrate. This may be useful forkinases which will not phosphorylate short peptides, andalso for kinases with high Kms for their protein sub-strate, since the competitive format avoids the necessityto label large quantities of these proteins. However,validation with a set of 50 commercially available kinaseinhibitors indicated a potential problem with interfer-ence by fluorescent compounds in the JNK-1 FP assay.Testing of a larger set of compounds in the PKC assayalso indicated a high hit rate for the FP assay, again dueto fluorescent compounds.

In the second part of this paper we used both kinaseFP assays to evaluate the ability of lifetime discrimina-tion by the FLARe to reduce interference from fluo-rescent compounds in FP assays. In the JNK-1 assayknown model short lifetime fluors and fluorescent in-terfering compounds (i.e., compounds which appearedactive in the JNK-1 FP assay but which were inactive inthe ½c-33P�ATP filter wash assay) interfered less with theLDP measurement than the FP measurement. In thesame assay there was no decrease in potency of standardnonfluorescent inhibitors.

In the case of PKC, increasing the fluorescein con-centration alone appeared to reduce the false positiverate in the FLARe optimized assay, as the hit ratesfound in the FP measurement using this assay werelower than in the original assay optimized for the An-alyst/Acquest. Use of the lifetime discriminated polar-ization measurement as opposed to FP on the FLARereduced the hit rates still further. However, even usingthe LDP measurement in the FLARe optimized assaystill gave a considerably higher hit rate than determinedin the ½c33P�-ATP filter wash assay, and a large numberof compounds still exhibited greater than 100% appar-ent inhibition in the LDP measurement. Possible rea-sons for this include incomplete exclusion of compound

fluorescence by FLARe (either because the compoundsdid not, as assumed, have fluorescence lifetimes shorterthan that of fluorescein (4 ns [2,6]) or because theirlifetimes were not sufficiently well separated from that offluorescein) or other types of interference, e.g., disrup-tion of the antibody–epitope interaction.

Problems we encountered in using the FLARe in-cluded higher assay costs, due to the higher concentra-tions of reagents used in the FLARe optimized assays,and longer plate read times than on the Analyst/Ac-quest, both due to the lower sensitivity of the FLAReinstrument.

Overall the data suggest that FLARe may have someutility in reducing interference from fluorescent com-pounds in FP assays, although complete rejection ofcompound interference was not achieved.

Acknowledgments

We thank Molecular Devices Corporation for theiradvice and support, and Dr Chris Armstrong (Divisionof Signal Transduction Therapy, School of Life Sci-ences, Dundee University) for providing activated hu-man JNK-1 a1 enzyme.

References

[1] T. French, J.C. Owicki, D.N. Modlin, S.S. Deshpande, I. Mineyev,

K. Crawford, W. Burton, Fluorescence-lifetime technologies for

high-throughput screening, Proc. SPIE 3259 (1998) 209–218.

[2] T. French, B. Bailey, D.P. Stumbo, D.N. Modlin, A time-resolved

fluorometer for high-throughput screening, Proc. SPIE 3603 (1999)

272–280.

[3] Y. Fleming, C.G. Armstrong, N. Morrice, A. Paterson, M.

Goedert, P. Cohen, Synergistic activation of stress-activated

protein kinase 1/c-Jun N-terminal kinase (SAPK1/JNK) isoforms

by mitogen-activated protein kinase kinase 4 (MKK4) and MKK7,

Biochem. 352 (2000) 145–154.

[4] J.H. Zhang, T.D.Y. Chung, K.R. Oldenburg, A simple statistical

parameter for use in evaluation and validation of high throughput

screening assays, J. Biomol. Screening 4 (1999) 67–73.

[5] G.R. Fleming, A.W.E. Knight, J.M. Morris, R.J.S. Morrison,

G.W. Robinson, Picosecond fluorescence studies of xanthene dyes,

J. Am. Chem. Soc. 99 (1977) 4306–4311.

[6] N. Periasamy, A.S. Verkman, Subtraction of background fluores-

cence in multiharmonic frequency-domain fluorimetry, Anal.

Biochem. 201 (1992) 107–113.


An evaluation of fluorescence polarization and lifetime discriminated polarization for high...

Documents

Transcript of An evaluation of fluorescence polarization and lifetime discriminated polarization for high...