A fluorescence polarization assay for inhibitors of Hsp90

ANALYTICALBIOCHEMISTRY

Analytical Biochemistry 350 (2006) 202–213

www.elsevier.com/locate/yabio

A Xuorescence polarization assay for inhibitors of Hsp90

R. Howes a,¤, X. Barril a, B.W. Dymock a, K. Grant a, C.J. NorthWeld a, A.G.S. Robertson a, A. Surgenor a, J. Wayne a, L. Wright a, K. James b, T. Matthews b,

K.-M. Cheung b, E. McDonald b, P. Workman b, M.J. Drysdale a

a Vernalis (Cambridge), Granta Park, Great Abington, Cambridge CB1 6GB, UKb Cancer Research UK Centre for Cancer Therapeutics, Institute of Cancer Research, Cotswold Road, Belmont, Sutton, Surrey SM2 5NG, UK

Received 16 September 2005Available online 23 January 2006

Abstract

Hsp90 encodes a ubiquitous molecular chaperone protein conserved among species which acts on multiple substrates, many of whichare important cell-signaling proteins. Inhibition of Hsp90 function has been promoted as a mechanism to degrade client proteins involvedin tumorigenesis and disease progression. Several assays to monitor inhibition of Hsp90 function currently exist but are limited in theiruse for a drug discovery campaign. Using data from the crystal structure of an initial hit compound, we have developed a Xuorescencepolarization assay to monitor binding of compounds to the ATP-binding site of Hsp90. This assay is very robust (Z� > 0.9) and can detectaYnity of compounds with IC50s to 40 nM. We have used this assay in conjunction with cocrystal structures of small molecules to drive astructure-based design program aimed at the discovery and optimization of a novel class of potent Hsp90 inhibitors.© 2006 Elsevier Inc. All rights reserved.

Keywords: Hsp90; Fluorescence polarization; ATPase; SBDD; SAR; Resorcinol; Geldanamycin; 17AAG; Radicicol; PU3

The 90-kDa heat-shock protein, Hsp90, is a highly con-served molecular chaperone protein (for reviews on Hsp90see [1–5]). It is highly abundant in the cell and has beenshown to be essential for cell survival. There are four formspresent in humans; Hsp90� (inducible upon stress), Hsp90�(low-level constitutive expression), Grp94 (endopolasmicreticulum localized), and TRAP1/Hsp75 (mitochondrialmatrix localized). Hsp90 functions as a dimer, both homo-and heterodimers, to maintain the appropriate folding andconformation of many other proteins [6]. The ATPaseactivity of Hsp90 is located in its N-terminal domain [7,8].However, the basal ATPase activity of Hsp90 is low and itrequires two associate proteins, Aha1 and Hch1, to achievefull activity precluding direct assaying of inhibition of thisactivity [9–12]. Hsp90 is unique among chaperone proteinsin that its partner proteins are involved in many cellularpathways critical for cell growth and survival. Hsp90 regu-

* Corresponding author. Fax: +44 1223 895 556.E-mail address: [email protected] (R. Howes).

0003-2697/$ - see front matter © 2006 Elsevier Inc. All rights reserved.doi:10.1016/j.ab.2005.12.023

lates many important kinases, for example erbB2 and Raf1,and other key cellular proteins [13–15].

In many tumors Hsp90 expression is ampliWed. Theaccumulation of Hsp90 aids survival of the tumor cell byrefolding partially damaged or mutant proteins and stabi-lizing them, for example mutant p53 [8,16]. Many of the sig-naling proteins associated with Hsp90 are serine/threoninekinases which have essential roles in malignant transforma-tion. For example Raf1, a key member of the Ras-MAPKsignaling pathway, has been shown to exist in a complexwith Hsp90 and this binding is essential for the activity ofRaf1 [13].

Inhibition of Hsp90 activity has been shown to selec-tively degrade many of the client proteins involved in cellproliferation, cell cycle regulation, and apoptosis in severaltumor models [5]. The result of degradation of these pro-teins is either cell stasis or apoptosis of the tumors [17].

There are several small-molecule Hsp90 inhibitors cur-rently reported in the literature. The ansamycin, geldana-mycin, was the Wrst Hsp90 inhibitor identiWed and has been

mailto: [email protected]

mailto: [email protected]

Fluorescence polarization assay for inhibitors of Hsp90 / R. Howes et al. / Anal. Biochem. 350 (2006) 202–213 203

shown to bind to the ATP-binding pocket in the N-termi-nal domain [18]. Geldanamycin has antitumor eVects butprogress through clinical trials has been halted due to oV-target toxicity [16,19]. Several derivatives of geldanamycinhave since been developed (17-AAG and 17-DMAG) andthese are currently in clinical trials [20,21]. The antibioticradicicol inhibits Hsp90 activity in vitro but has failed toshow tumor suppression in in vivo models due to instabilityof the compound [22,23]. However, oxime derivatives ofradicicol show in vivo activity and it is expected that thesewill be progressed to the clinic [24]. PU3, a purine deriva-tive, binds to the ATP-binding pocket of Hsp90 and hasbeen shown to prevent tumor growth in vitro [25–27].

To detect Hsp90 inhibitors and generate SAR,1 we havedeveloped a high-throughput assay to monitor inhibition ofHsp90 activity. Many assays have been described to moni-tor inhibition of Hsp90 [53,54] using several diVerent tech-niques, e.g., time-resolved Xuorescence resonance energytransfer and Xuorescence polarization (FP). However allthese assays use derivatives of natural products which havebeen shown to inhibit Hsp90 function. Due to the limitedsites of modiWcation available on natural products wedecided to develop an assay using a derivatized small mole-cule. Due to the availability of a crystal structure of a leadcompound in Hsp90 we were able to rationally design aprobe for a Xuorescence polarization assay. The assay isbased upon displacement of a Xuorescently labeled mole-cule, which binds speciWcally to the ATP-binding site offull-length human Hsp90. We monitor this displacement bya decrease in Xuorescence polarization of the probe-Hsp90complex when the inhibitor binds. The assay is very robust(Z� > 0.9) and can detect inhibitor binding to IC50s of40 nM. The assay has been developed to identify inhibitorsof both full-length human Hsp90� and Hsp90�. We haveused this assay to screen known Hsp90 inhibitors and havecompared this to inhibition of the ATPase activity of yeastHsp90, an assay previously used to monitor Hsp90 inhibi-tion [28,29].

Materials and methods

Expression vector construction

pRSETA Human Hsp90� was obtained from C. Podro-mou (ICR, UK). pET19-scHsp90, encoding N-terminallyHis-tagged yeast Hsp90, was made in the following way. A2.1-kb fragment encoding Hsp90 was ampliWed by PCRusing Hotstart Pfu Turbo DNA Polymerase (Stratagene)according to the manufacturer’s instructions from genomicDNA of Saccharomyces cerevisiae strain L40coat (Invitro-gen). Primer sequences were Forward (cgc gca tat ggc tagtga aac ttt tga att tc) and Reverse (cgc gct cga gtt act aat ctacct ctt cca ttt cgg). This fragment was subcloned in pCR2.1

1 Abbreviations used: SAR, structure–activity relationships; FP, Xuores-cence polarization; DMSO, dimethyl sulfoxide; SEM, trimethylsilyleth-oxymethoxy.

using TopoTA cloning kit (Invitrogen) to create the plas-mid pCR-scHsp90. The insert was fully sequenced to con-Wrm the correct open reading frame. A 2.1-kb NdeI-XhoIfragment was subcloned into pET19b (Novagen) to pro-duce the plasmid pET19-Hsp90.

pET19-Hsp90�(9-236), encoding the His-tagged N-ter-minal ATPase domain of human Hsp90a, was constructedas follows. The region of Hsp90� encoding amino acids 9–236 was ampliWed by PCR from IMAGE clone 4026275using primers Hsp90-N-HisFor (cgc ata tgg acc aac cga tggagg ag) and Hsp90-N-Rev (gcg gat cct cat tat tca gcc tca tcatcg ct) using Hotstart Pfu Turbo DNA polymerase (Strat-agene) and subcloned into pCRII using TopoBlunt cloningkit (Invitrogen) to create plasmid pCR-Hsp90�(9-236). Theinsert was fully sequenced to conWrm that the expectedregion of Hsp90� was present. A 0.7 kb NdeI-BamHI frag-ment from this plasmid was subcloned into pET19b (Nova-gen) to create the plasmid pET19-Hsp90�(9-236).

Hsp90 protein expression and puriWcation

Expression and puriWcation of full-length Hsp90� andHis-tagged N-terminal Hsp90� was as previously described[30]. Yeast Hsp90 was overexpressed in the Escherichia colistrain BL21 and puriWed in a manner similar to that of His-tagged N-terminal Hsp90�.

Crystallization and three-dimensional structure determination

Apo Hsp90� protein was concentrated to approximately20 mg/ml using ultraWltration into a Wnal buVer containing20 mM Tris, pH 7.4, and 0.5 M NaCl. For cocrystallizationexperiments, a 20 mM stock solution of the ligand VER-00063579 in 100% DMSO was provided. At this concentra-tion the ligand is at a twofold molar excess when added tothe protein. The complex sample was left on ice for 1 h priorto setting up the crystallization trials. Cocrystals appearedovernight in conditions very similar to those for apo crys-tals, previously described [30]. These were subsequentlytransferred to cryoprotectant solution (crystallization reser-voir solution with polyethylene glycol concentrationincreased from 25 to 35%) and frozen in liquid nitrogen,and data were collected on station ID29 at the EuropeanSynchrotron Radiation Facility (Grenoble, France).

DiVraction data were processed using DENZO [31]. TheVER-00063579-bound structure was solved by molecularreplacement with AMoRe [32] using the apo Hsp90� struc-ture as the search model (PDB code [33]D1UY1). Allmodel building was carried out using the molecular graph-ics program O [34] and reWnement calculations were per-formed with Refmac5 [35]. Following structure solution,diVerence electron density maps were calculated for the ini-tial model, the ligand structure was modeled into the diVer-ence density peaks, and the coordinates were reWned beforeaddition of crystallographic water molecules using Ref-mac5 cycled with ARP/wARP [36]. The progress of the

204 Fluorescence polarization assay for inhibitors of Hsp90 / R. Howes et al. / Anal. Biochem. 350 (2006) 202–213

reWnement was assessed using Rfree and R factor and theWnal structure validated using PROCHECK [37] and theCCP4i package [38]. Full data collection and reWnementstatistics are presented in Table 1.

Chemical synthesis

Synthesis of pyrazoles (a) to (e) was achieved by themethod outlined in Scheme 1. CCT018159 (VER-00063579)was the original hit compound from a high-throughputscreen [28].

Resorcinols (1) were converted to isoXavone intermedi-ates (2) in a high-yielding one-pot procedure using theappropriate substituted phenylacetic acid in reXuxingboron triXuoride etherate. The resulting acetophenones

Table 1Data collection and reWnement statistics for Hsp90 in complex with VER-00063579

Rfree is the R factor calculated using 5% of the reXection data chosen ran-domly and omitted from the reWnement process, whereas Rcryst is calcu-lated with the remaining data used in the reWnement. Rms bond lengthsand angles are the deviations from ideal values; the rms deviation in B fac-tors is calculated between covalently bonded atoms. hrb, highest resolu-tion bin.

Data collection statisticsResolution (Å) 30–2.25Space group P21212Cell dimensions (Å)

a 98.5b 64.4c 88.6

No. molecules/asymmetric unit 2Volume (Å3) 5.6 £ 105

Solvent content (%) 57.6Measured reXections 682501Unique reXections 27767Completeness: Overall/in hrb (%) 90.1/75.0Average multiplicity/in hrb 5.5/1.5Mean I/�I: Overall/in hrb 13.3/1.5Rmerge: Overall/in hrb (%) 11.1/47.2

ReWnement statisticsRfree (%) 28.6Rcryst (%) 22.1Rms deviations:

Bonds (Å) 0.022Angles (°) 1.8B factor (Å2) 2.3

PDB Code 2CDD

Scheme 1. Reagents and conditions: (a) ArCH2CO2H, BF3.OEt2, 85 °C;(b) PCl5 / DMF rt; (c) Hydrazine hydrate, EtOH, reXux.

OHOH

R

O

O

Ar

OH

R

NNH

Ar

OH

OH R

a, b c

(1) (2)

(3)R = H, Cl, EtAr = Ph, 3,4-dimethoxy

70 - 92% 61 - 88%

were treated with a mixture of phosphorous pentachlorideand dimethylformamide that had been previously heated to60 °C. The resultant reaction mixture was then stirred atroom temperature for 1 h. Subsequent reaction of the iso-Xavone (2) with hydrazine hydrate gave the pyrazole prod-ucts (3) in generally good yields.

VER-00045864 was prepared by the route described inScheme 2. 4-Chlororesorcinol (4) was converted to ace-tophenone (5) in 52% yield by treatment with acetic acid inreXuxing boron triXuoride etherate, followed by addition ofbenzyl bromide in reXuxing acetonitrile with potassium car-bonate as base. Enamine (6) was furnished in good yield byreXuxing (5) in dimethylformamide dimethylacetal. Subse-quent treatment with hydrazine gave the pyrazole (7) inexcellent yield. Iodination at the 4-position with N-iodosuc-cinimide, followed by SEM protection of the pyrazole gaveadvanced intermediate (8) in good yield (31% overall yieldfrom 4-chlororesorcinol; six steps).

Cross-coupling reactions attempted on advanced inter-mediate (8) proved to be capricious, with variable yields.However amine (9) was prepared by reaction of (8) with 4-aminomethylboronic acid and prudent selection of palla-dium catalyst. Subsequent attachment of the Xuorescentportion of the probe compound was achieved by treat-ment of (9) with 5-carboxyXuorescein succinimidyl esterin dichloromethane using triethylamine as base. Removal

Scheme 2. Reagents and conditions: (a) AcOH, BF3.OEt2, reXux (58%); (b)BnBr, K2CO3, MeCN, reXux (90%); (c) dimethylformamide dimethylac-etal, reXux (84%); (d) NH2NH2.H2O, EtOH, reXux (92%); (e) NIS, DCM,(97%); (f) SEMCl, Cs2CO3 DMF (80%); (g) 4-aminomethylboronic acid,PdCl2(pph3)2 (cat.), NaHCO3, DMF, 100 °C (26%); (h) 5-carboxyXuores-cein succinimidyl ester, Et3N, DCM (40%); (i) BCl3, CH2Cl2 (32%).

Cl

OH

OH

Cl

O

BnO

OBn

Cl

O

BnO

OBn

N

Cl

BnO

OBn N NH

Cl

BnO

OBn N N

I

SEM

Cl

BnO

OBn N NSEM

NH2

Cl

OH

OH N NH

NH

O

O

OH

OH

OO

a, b c

d

e, f

g

h, i

(4) (5) (6)

(7)(8)

(9)

VER-00045864


of the SEM and benzyl protecting groups with boron tri-chloride in dichloromethane and subsequent reverse-phase HPLC puriWcation furnished the probe compoundVER-00045864 in moderate yield (3% from (8); threesteps).

Human Hsp90� FP assay

Compounds were assayed over a threefold dilution seriesin a 96-well plate format to determine IC50 values. Reactionmix (100 �l) (100 mM Tris.Cl, pH 7.4, 20 mM KCl, 6 mMMgCl2, 5 �g/ml bovine serum albumin, 80 nM VER-00045864, 100 nM Hsp90�) was added per well of a black96-well plate (Corning Costar No. 3915) and allowed toequilibrate for 20 min at room temperature in the dark.Control wells contained reaction mix without VER-00045864 or Hsp90�. Compounds were titrated in a three-fold dilution series in DMSO in a separate 96-well plate (V-bottomed clear plates; VWR 007/008/257) at 50 times Wnalconcentration. After incubation, plates were read on aFusion Alpha-FP (Perkin–Elmer, USA) with excitation485/20 nM and emission 535/25 nM with polarization. Then2 �l of compound was added to the reaction mix, mixedthoroughly, and allowed to equilibrate again at room tem-perature in the dark for 30 min. The Wnal DMSO concen-tration in the assay was 2%; however, the assay is tolerantof up to 5% DMSO (data not shown). IC50 values were cal-culated on the diVerence in anisotropy from the Wrst andsecond reads. The assay has also been developed for usewith yeast Hsp90 and N-terminal human Hsp90� but thiswill be presented elsewhere.

Yeast Hsp90 ATPase assay

Assay was performed essentially as described [28].Compounds were assayed over a threefold dilution seriesin a 384-well plate format to determine IC50 values. ATPwas diluted in assay buVer (100 mM Tris·Cl, pH 7.4,150 mM NaCl, and 6 mM MgCl2,) and 5 �l was added perwell of a 384-well plate (Corning Costar 3702) (Wnal[ATP] is 370 �M). Yeast full-length Hsp90 was alsodiluted in assay buVer and 5 �l was added per well (Wnal[Hsp90] is 0.42 �M). Control wells contained assay buVerwithout Hsp90. Compounds were titrated in a threefolddilution series in phosphate-free H2O in a separate 96-wellplate (V-bottomed clear plates; VWR 007/008/257) at 5times Wnal concentration and subsequently, 2.5 �l wasadded per well. This was incubated overnight at 37 °C.After incubation, 40 �l of Malachite Green reagent (ARwater: 2.325%; PVA: 5.725%; ammonium molybdate:0.0812%; Malachite Green in 2:1:1:2 ratio) was added andthe plate was put on a plate shaker for 5 min to allow thecolor to develop. The reaction was stopped by the addi-tion of 5 �l 34% sodium citrate. Plates were read on a Wal-lac Victor2 plate reader (Perkin–Elmer, USA) at 620 nm.The limit of detection for this assay is 200 nM and typicalZ� values are >0.7.

Results

Use of structure-based design to develop a Xuorescent derivative of an Hsp90 inhibitor for a FP assay

From a high-throughput screening campaign of 56,000compounds we identiWed a small molecule, VER-00063579,which inhibits the ATPase activity of yeast Hsp90 [28]. Wehave solved the crystal structure of this compound boundin the N-terminal domain of Hsp90� (Figs. 1A–C). Themolecule binds to the adenine binding site of Hsp90� pri-marily through its resorcinol unit, which overlaps exactlywith the resorcinol moiety of the natural inhibitor radicicol[39]. The hydroxyls in this ring form an extensive networkof hydrogen bonds with the carboxylate of Asp93 and threeinterstitial water molecules. The ethyl in position 4 of theresorcinol makes contacts with the apolar side chains ofLeu107, Phe138, and, to a lesser extent, Val150, as does thering itself with Met98, Val186, Asn51 (C�), Ala55, andThr184 (C�2). All these interactions make the resorcinol theanchoring point of the ligand to the active site of Hsp90�,and in fact several resorcinol-containing small moleculeshave been identiWed by NMR screening of molecular frag-ments (data not shown).

The planes of the pyrazole and resorcinol rings form anangle of approximate 130°, allowing N2 of the pyrazole tointeract with the hydroxyl of Thr184 and an interstitialwater molecule coordinated to Asp93; interestingly thisatom overlaps exactly with the carbonyl oxygen of radici-col’s ester, suggesting that a hydrogen bond acceptor atomin this position is important for activity. The second N (N1)is only 2.7 Å apart from the carbonyl oxygen of Gly97, sug-gesting that a hydrogen bond is formed between these twoatoms, although the geometry of the interaction is not idealbecause the vector connecting the two atoms forms anangle of approximately 60° with the plane of the peptidicbond of Gly97 and, in consequence, the N–H bond of thepyrazole does not align with the lone pairs of the carbonyloxygen. Although two tautomers of the pyrazole are theo-retically possible, the chemical environment describedabove suggests that the protonation state relevant for bind-ing to Hsp90 is as drawn in Scheme 1 (compound 3). Thecarbon atoms of the pyrazole ring also contribute to bind-ing, forming apolar interactions with the side chains ofMet98 and Ala55. The methyl at position 5 is completelysolvent exposed and does not interact with the protein.Finally, the benzodioxane group is located at the entranceof the binding site and forms only a few contacts with theprotein, none of which seems particularly favorable;notwithstanding, the presence of the aryl substitution atposition 4 of the pyrazole is extremely important to guaran-tee that the conformational arrangement observed in thecrystal structure corresponds to an energy minimum.

Both the aryl ring and the C5 of the pyrazole provideappropriate vectors to reach the solvent space, and wehypothesized that these positions could be used to attach aXuorescent group that would allow creation of a suitable



probe for development of a binding assay (Fig. 2A). Fig. 2Adepicts a model of the binding mode of VER-00045864.5-CarboxyXuorescein was added at the 4-position of the pyr-azole ring of VER-00029889 via a 4-(aminomethyl)phenyllinker (Fig. 2A; see Materials and methods for schema). Posi-tion 4 was selected for its synthetic accessibility and becauseit is located very close to the surface of the protein, allowingsubstituents to reach a very wide space even without the needfor long linkers (Fig. 1A). It is interesting to note that thisposition is in the locality of position 17 of geldanamycin,where substitution with very diverse chemical groups are well

tolerated [18], which is indicative of lack of contacts with theprotein and supports our hypothesis.

The Xuorescent compound, VER-00045864, binds toHsp90 with a Kd of 80nM§16 nM (Fig. 2B). This probe canbe displaced with excess ATP and VER-00063579 (parentmolecule) (data not shown). Attempts to obtain an X-raystructure of the VER-00045864–Hsp90� complex have beenunsuccessful which is not surprising when one considers howmuch the probe protrudes out of the binding site. None ofthe crystal forms of Hsp90 known to us can accommodatethe Xuorescent group without major disruption.

Fig. 2. (A) Theoretical model and the chemical structure of the probe used in the FP assay, VER-00045864. (B) Binding of VER-00045864 to Hsp90�.Results are shown §SD from n D 8. From this, VER-00045864 has a Kd of 80 § 16 nM for Hsp90�. Variability in the assay is very low with both back-ground and signal CVs being below 4%. This results in very high Z� values (typically >0.9) and R2s of >0.98.

0

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0 500 1000 1500 2000

[Hsp90] nM

Prob

e bo

und

μM

NN

NCl

O

O

O

O

OO

O

O

VER-00045684

Signal CV 3.7%Background CV 1.8%Signal window 110mPZ' Factor > 0.9Typical R 0.972

Assay Characteristics

A

B

Fig. 1. X-ray crystallographic structure of VER-00063579 in complex with the N-terminal domain of Hsp90�. (A) Cartoon representation showing VER-00063579 (yellow ball-and-stick representation) bound to monomer A, one of two very similar monomers in the asymmetric unit. (B) Key interactionsformed between VER-00063579 and binding site residues of Hsp90�. (C) Three-dimensional stereo representation of the VER-00063579 in a Fc ¡ Fc diVer-ence map (contoured at 2�).


A Xuorescence polarization assay to measure binding to the ATPase site of Hsp90

We have used the probe VER-00045864 to develop aXuorescence polarization assay to measure compoundbinding to Hsp90 [40–44]. In this assay format, dye mole-cules (in this case Xuorescein) aligned parallel to linearlypolarized light are selectively excited. For dyes attached tosmall, rapidly rotating molecules, e.g., VER-00045864, theinitially photoselected oriented distribution becomes ran-domized prior to emission, resulting in low Xuorescencepolarization. Conversely, binding of the probe to a large,slowly rotating molecule results in high Xuorescence polari-zation. Therefore upon binding of the probe to Hsp90 weobserve an increase in Xuorescence polarization. Fluores-cence polarization therefore provides a direct readout ofthe extent of probe binding to Hsp90.

Addition of known Hsp90 inhibitors into the assay com-petes with the probe for binding to Hsp90, resulting in adecrease in FP in comparison to the probe alone (Fig. 3A).

We have used this to assay various known Hsp90 ligandsover a range of compound concentrations and hence deter-mine their IC50 values (Table 2; Fig. 3A).

Using this assay, the purine analogue PU3 has an IC50 of48.7 �M, similar to its reported aYnity of 15–20 �M [45].Radicicol has a high aYnity (IC50 60§0.2 nM), in agree-ment with reported values [46]. The ansamycin, geldanamy-cin, has an IC50 of 348 nM, similar to the reported value of0.4 �M in a Wlter binding assay [46,47]. The geldanamycinderivative, 17AAG, has a reported aYnity for Hsp90 of1 �M and our results agree with this (FP IC501.27§ 0.43�M). These results show that our assay can rap-idly determine a range of binding aYnities of compoundsfor Hsp90. The ranking of aYnities in our assay isradicicol > geldanamycin > 17AAG > PU3, which is inagreement with results from other Hsp90 assay systems [28]and validates this assay to identify and diVerentiatebetween Hsp90 inhibitors.

To further validate this assay format we tested theabove compounds for their ability to inhibit the ATPase

Fig. 3. (A) Dose–response curves for the inhibition of Hsp90� by known inhibitors. Compounds were tested in tripling dilution series. Results are shown§SD from n 7 4. PU3 binds weakly to Hsp90 with an IC50 of 49 �M (open triangles). Geldanamycin binds tightly to Hsp90� with an IC50 of 281 nM(Wlled triangles). The 17-AAG derivative of geldanamycin binds more weakly to Hsp90� (IC50 1.27 �M; open circles). These results are in agreement withprevious results from the inhibition of the yeast Hsp90 ATPase activity and correlate with our own Wndings with this assay (Table 1). (B) Dose–responsecurves in the yeast Hsp90 ATPase assay for the original hit compound (VER-00063579/CCT018159; Wlled squares) and the probe used in the FP assayVER-00045864 (open triangles). Results are shown §SD from n 7 4.

% I

nhib

ition

of

prob

e bi

ndin

g

0

20

40

60

80

100

0.01 0.1 1 10 100

[Compound] μM

0

20

40

60

80

100

120

0.01 0.1 1 10 100 1000

[Compound] μMPU317-AAGGeldanamycin

VER-00045864VER-00063579 (CCT018159)

% I

nhib

ition

of

yeas

t H

sp90

AT

Pase

act

ivity

A

B


Table 2Activities of known Hsp90 inhibitors in yeast Hsp90 ATPase and human Hsp90� FP assays

Compound Structure Human Hsp90� FP IC50 �M (§SD)

Yeast Hsp90 ATPaseIC50 �M (§SD)

Name

VER-00045864 NA 2.25 § 0.72 FP assay probe

VER-00046793 1.27§ 0.43 17.74 § 7.7 17AAG

VER-00049032 0.060 § 0.002 0.26 § 0.01 Radicicol

VER-00063579 0.148 § 0.016 6.52 § 3.44 CCT018159 (original hit)

VER-00074533 0.281 § 0.097 2.86 § 0.31 Geldanamycin

VER-00080876 49.0§ 1.75 >200 PU3

NN

N

Cl

O

O

O

O

OO

O

O

ONH

Me

O

Me

Me

OH

ONH

MeOMeO

OCONH2

Me

OO

O

Cl

O

O

O

N

N

O

O

O

O

ONH

Me

O

Me

Me

OH

OO

MeOMeO

OCONH2

Me

N

NN

N

N

O

O

O


activity of yeast Hsp90 (Table 2; see Materials and meth-ods; [28]). Geldanamycin and 17AAG have yeast ATPaseIC50s of 2.86 and 17.7 �M, respectively, in agreementwith their previously described activities (geldanamycin4.8 § 0.8 �M, 17AAG 8.7 § 2.3 �M). These slight diVer-ences in IC50 values are due to the diVerences in Hsp90protein concentration used in the assays (Rowlands et al.used 1.69 �M; we used 0.42 �M). Radicicol has a strongerability to inhibit the ATPase activity with an IC50 of0.26 �M. This is diVerent from the reported value in thisassay (0.9 § 0.4 �M) but is in agreement with thereported binding aYnity (19 nM [46]). PU3 does notinhibit the ATPase activity of yeast Hsp90 (IC50 >200 �M) [48].

We also tested the ability of the FP assay probe com-pound (VER-00045864) and its parent compound VER-00063579 (CCT018159) to inhibit the yeast ATPase activity(Table 2; Fig. 3B). VER-00063579 (CCT018159) has anIC50 of 6.5 �M, similar to its previously reported value of8.9§0.72 �M [28]. VER-00045864 has a threefold betterability to inhibit the yeast ATPase activity (IC50 2.5 �M).

Resorcinol-containing compounds act as Hsp90 inhibitors

The initial hit compound, VER-00063579 (CCT018159),has several close analogues present in the same library [28].These analogues were not initially identiWed from the pri-mary screen due to their relatively low activities which werebelow the initial compound screening concentration (forexample, CCT018157 has a yeast ATPase IC5060.8§ 6.7�M [28]).

We have extended this analysis to other analoguesfrom the same library. Compounds were extracted on thebasis of retaining the pharmacophore of resorcinol moi-ety, pyrazole function, and bis-aryl conWguration presentin VER-00063579. Compounds identiWed were assayedfor their ability to bind to the ATPase site of humanHsp90� (Fig. 4A; Table 3). The completely unsubstitutedanalogue, VER-00063579 (CCT018159), has a humanHsp90� FP IC50 of 148 nM. The core used for thissearch, VER-00029909 (CCT066950), has a humanHsp90� FP IC50 of 7.24 �M. This low aYnity is reXectedin its yeast ATPase IC50 of 29.7 �M. Removal of the5-methyl substituent from the pyrazole ring gave no sig-niWcant change in activity (VER-00053579 148 nM com-pared to VER-00029888 381 nM) as expected for a totallysolvent-exposed moiety.

Interestingly, removal of the lipophilic substituent at the4-position of the resorcinol leads to a 10-fold loss in activityin both the FP and the ATPase assays (compare VER-00029889 (795 nM) and VER-00029907 (748 nM) withVER-00029909 (7.2�M)) which is consistent with the apo-lar contacts observed in the crystal structure. Additionally,there is no discernible diVerence observed between alkyl orhalogen substituents at this position (compare VER-00029989 (alkyl; 795 nM) with VER-00029907 (aryl;748 nM)).

Substitution of the aryl ring at the 4-position of thepyrazole had little eVect on activity, with the unsubsitut-ued phenyl and 3,4-dimethoxyphenyl substituents oVeringno advantage over the dihydrobenzodioxine ring con-tained within VER-00063579 (compare with VER-00029886 (644 nM) and VER-00029888 (381 nM)).Indeed, there appears to be no signiWcant constraint onsubstituents that can be tolerated at this position, asborne out by the probe compound (VER-00045864) hav-ing retained activity despite containing the bulky Xuores-cein moiety.

To conWrm the ability of these compounds to inhibitHsp90, we tested them for their ability to inhibit the ATP-ase activity of yeast Hsp90 (Fig. 4B; Table 3). As can beseen from Table 3 there is a smaller diVerential in the com-pound activities compared to their FP results. The datafrom this assay are unable to readily distinguish betweenthe compounds, whereas we are able to make structure–activity relationships from their FP data.

Fig. 4. Using the Hsp90� FP assay we have tested several resorcinol-con-taining compounds for their ability to bind to the ATP-binding site ofHsp90�. All compounds were titrated from 10 �M in a tripling dilutionseries. Results are shown §SD from n 7 4. VER-00063579 (CCT018159)binds to Hsp90� with an IC50 of 148 nM (Wlled squares). Removal of thelipophilic substituent of the resorcinol in VER-00029907 causes a 10-foldloss in activity (IC50 748 nM; open squares). VER-00029888, which hasaltered the aryl ring of the pyrazole, has an IC50 of 381 nM (Wlled triangles).


Discussion

There is increasing evidence that inhibition of Hsp90 is avaluable tool in the Wght against cancer. The role of Hsp90as a molecular chaperone to a wide variety of importantcell-signaling molecules provides a rationale that inhibitionof its activity will arrest tumor growth [8]. Several com-pounds based on naturally occurring inhibitors of Hsp90are currently progressing through clinical trials [24,49–52].

However, one of the limitations of identifying Hsp90inhibitors has been the paucity of suitable assays to screensmall-molecule inhibitors. Previous assays are either toocomplex for such a program or are limited to relativelypoor detection of compounds. To overcome this, wedescribe a Xuorescence polarization assay which monitorsthe binding of small molecules to the ATPase site of Hsp90.We have used information gained from the crystal structure

of a small-molecule inhibitor of Hsp90, VER-00063579, todesign a suitable probe for the assay.

We have shown that this assay can be readily used todetermine the binding activities of known Hsp90 inhibitorsand that these are in agreement with published data. ToconWrm our results we have correlated this with the abilityof these compounds to inhibit the ATPase activity of yeastHsp90.

The compound identiWed previously, VER-00063579,was used to identify similar molecules from our compoundlibrary. We used the resorcinol moiety, pyrazole, and bis-aryl functions of the parent molecule as a basis for thesearch. We identiWed Wve further compounds from ourlibrary and have assayed these compounds for their abilityto bind to Hsp90. We have compared these results to theirability to inhibit the ATPase function of yeast Hsp90. Fromthe small selection of compounds we see several themes

Table 3Activities of resorcinol series compounds in yeast Hsp90 ATPase and human Hsp90� FP assays

Compound Structure Human Hsp90� FP IC50 �M (§SD) Yeast Hsp90 ATPase IC50 �M (§SD) Alternative name

VER-00063579 0.148 § 0.016 6.52§ 3.44 CCT018159

VER-00029886 0.644 § 0.30 4.84§ 2.75 CCT066963

VER-00029888 0.381 § 0.12 2.73§ 1.34 CCT066965

VER-00029889 0.795 § 0.365 4.12§ 0.63 CCT066960

VER-00029907 0.748 § 0.30 4.99§ 1.24 CCT066952

VER-00029909 7.24 § 1.80 29.7§ 20.9 CCT066950

N

N

O

O

O

O

NN

O

O

O

O

NN

O

O

O

O

Cl

NN

O

O

Cl

NN

O

O

NN

O

O


appearing. Removal of the lipophilic substituents at the 4-position of the resorcinol results in a 10-fold loss of activity.

Alteration of the aryl ring at the 4-position of the pyra-zole had little eVect on compound activity with little limita-tion on the substituents that can be tolerated at thisposition. However, we have examined only a small set ofcompounds allowing only preliminary SAR to be deter-mined. This reconWrms the decision to add the bulky Xuo-rescein moiety at this position as this compound retainsactivity to inhibit Hsp90.

Comparison of the data from the FP assay with the pre-viously described yeast ATPase assay shows the beneWts ofthis assay system. The yeast assay was unable to readily dis-tinguish between the activities of the compounds in thisstudy, mainly due to the potencies of the compoundsapproaching the lower detection limits of this assay. How-ever, the FP assay has a much lower detection limit (40 nM)and we have shown that this can readily distinguishbetween the activities of this set of related molecules.

Development of speciWc Hsp90 inhibitors will provide anew form of cancer therapy with the ability to treat a widerange of cancer types. We have described a simple, high-throughput, sensitive assay that allows prosecution of adrug discovery program targeted against Hsp90. This assaywill allow the detection and development of Hsp90 inhibi-tors that lack some of the problems associated with the Wrstin class Hsp90 inhibitors currently in clinical trials.

Acknowledgments

We thank Terry Shaw (Vernalis) for preparing yeastgenomic DNA. We also thank Richard Grant for criticalreading of the manuscript.

References

[1] A.S. Sreedhar, C. Soti, P. Csermely, Inhibition of Hsp90: a new strat-egy for inhibiting protein kinases, Biochim. Biophys. Acta. 1697 (1–2)(2004) 233–242.

[2] P. Workman, Altered states: selectively drugging the Hsp90 cancerchaperone, Trends Mol. Med. 10 (2004) 47–51.

[3] J. Buchner, Hsp90 & Co.—a holding for folding, Trends Biochem. Sci.24 (1999) 136–141.

[4] J.S. Isaacs, W. Xu, L. Neckers, Heat shock protein 90 as a moleculartarget for cancer therapeutics, Cancer Cell 3 (2003) 213–317.

[5] A. Maloney, P. Workman, HSP90 as a new therapeutic target for can-cer therapy: the story unfolds, Expert Opin. Biol. Ther. 2 (2002) 3–24.

[6] L.H. Pearl, C. Prodromou, Structure and in vivo function of Hsp90,Curr. Opin. Struct. Biol. 10 (2000) 46–51.

[7] L.H. Pearl, C. Prodromou, Structure, Function and Mechanism of theHsp90 Molecular Chaperone, Adv. Protein Chem. 59 (2002) 157–186.

[8] P. Csermely, T. Schnaider, C. Soti, Z. Prohaszka, G. Nardai, The 90-kDa molecular chaperone family: structure, function, and clinical appli-cations, A comprehensive review. Pharmacol. Ther. 79 (1998) 129–168.

[9] P. Meyer, Structural basis for recruitment of the ATPase activatorAha1 to the Hsp90 chaperone machinery, EMBO J. 23 (2004) 1402–1410.

[10] G.P. Lotz, H. Lin, A. Harst, W.M. Obermann, Aha1 binds to the mid-dle domain of Hsp90, contributes to client protein activation, andstimulates the ATPase activity of the molecular chaperone, J. Biol.Chem. 278 (2003) 17228–17235.

[11] B. Panaretou, G. Siligardi, P. Meyer, A. Maloney, J.K. Sullivan, S.Singh, S.H. Millson, P.A. Clarke, S. Naaby-Hansen, R. Stein, R.Cramer, M. Mollapour, P. Workman, P.W. Piper, L.H. Pearl, C. Pro-dromou, Activation of the ATPase activity of Hsp90 by the stress-reg-ulated cochaperone Aha1, Mol. Cell 10 (2002) 1307–1318.

[12] D.F. Nathan, M.H. Vos, S. Lindquist, IdentiWcation of SSF1, CNS1,and HCH1 as multicopy suppressors of a Saccharomyces cerevisiaeHsp90 loss-of-function mutation, Proc. Natl. Acad. Sci. USA 96(1999) 1409–1414.

[13] T.W. Schulte, M.V. Blagosklonny, C. Ingui, L. Neckers, Disruption ofthe Raf-1-Hsp90 molecular complex results in destabilization of Raf-1 and loss of Raf-1-Ras association, J. Biol. Chem. 270 (1995) 24585–24588.

[14] S. Sato, N. Fujita, T. Tsuruo, Modulation of Akt kinase activity bybinding to Hsp90, Proc. Natl. Acad. Sci. USA 97 (2000) 10832–10837.

[15] W. Xu, E.G. Mimnaugh, J.S. Kim, J.B. Trepel, L.M. Neckers, Hsp90,not Grp94, regulates the intracellular traYcking and stability ofnascent ErbB2, Cell Stress Chaperones 7 (2002) 91–96.

[16] W.B. Pratt, D.O. Toft, Regulation of signaling protein function andtraYcking by the hsp90/hsp70-based chaperone machinery, Exp. Biol.Med. 228 (2003) 111–133.

[17] I. Hostein, D. Robertson, F. DiStefano, P. Workman, P.A. Clarke,Inhibition of signal transduction by the Hsp90 inhibitor 17-allyla-mino-17-demethoxygeldanamycin results in cytostasis and apoptosis,Cancer Res. 61 (2001) 4003–4009.

[18] C.E. Stebbins, A.A. Russo, C. Schneider, N. Rosen, F.U. Hartl, N.P.Pavletich, Crystal structure of an Hsp90-geldanamycin complex: tar-geting of a protein chaperone by an antitumor agent, Cell 89 (2)(1997) 239–250.

[19] J.G. Supko, R.L. Hickman, M.R. Grever, L. Malspeis, Preclinicalpharmacologic evaluation of geldanamycin as an antitumor agent,Cancer Chemother. Pharmacol. 36 (1995) 305–315.

[20] J.M. Jez, J.H. Chen, G. Rastelli, R.M. Stroud, D.V. Santi, CrystalStructure and Molecular Modelling of 17DMAG in Complex withHuman Hsp90, Chem. Biol. 10 (2003) 361–368.

[21] L. Neckers, Heat shock protein 90 inhibition by 17-allylamino-17-demethoxygeldanamycin: a novel therapeutic approach for treatinghormone-refractory prostate cancer, Clin. Cancer Res. 8 (2002) 962–966.

[22] T.W. Schulte, S. Akinaga, S. Soga, W. Sullivan, B. Stensgard, D. Toft,L.M. Neckers, Antibiotic radicicol binds to the N-terminal domain ofHsp90 and shares important biologic activities with geldanamycin,Cell Stress Chaperones 3 (1998) 100–108.

[23] S. Vogen, T. Gidalevitz, C. Biswas, B.B. Simen, E. Stein, F. Gulmen, Y.Argon, Radicicol-sensitive peptide binding to the N-terminal portionof GRP94, J. Biol. Chem. 277 (2002) 40742–40750.

[24] T. Agatsuma, H. Ogawa, K. Akasaka, A. Asai, Y. Yamashita, T.Mizukami, S. Akinaga, Y. Saitoh, Halohydrin and oxime derivativesof radicicol: synthesis and antitumor activities, Bioorg. Med. Chem.10 (2002) 3445–3454.

[25] G. Chiosis, M.N. Timaul, B. Lucas, P.N. Munster, F.F. Zheng, L.Sepp-Lorenzino, N. Rosen, A small molecule designed to bind to theadenine nucleotide pocket of Hsp90 causes HER2 degradation andthe growth arrest and diVerentiation of breast cancer cells, Chem.Biol. 8 (2001) 289–299.

[26] G. Chiosis, B. Lucas, A. Shtil, H. Huezo, N. Rosen, Development of apurine-scaVold novel class of Hsp90 binders that inhibit the prolifera-tion of cancer cells and induce the degradation of HER2 tyrosinekinase, Bioorg. Med. Chem. 10 (2002) 3555–3564.

[27] G. Chiosis, N. Rosen, Sloan-Kettering Institute For Cancer Research:WO0236075, Small molecule compositions for binding to Hsp90[Sloan-Kettering Institute For Cancer Research: WO0236075], 2002.

[28] M.G. Rowlands, Y.M. Newbatt, C. Prodromou, L.H. Pearl, P. Work-man, W. Aherne, High-throughput screening assay for inhibitors ofheat-shock protein 90 ATPase activity, Anal. Biochem. 327 (2004)176–183.

[29] G.W. Aherne, A. Maloney, C. Prodromou, M.G. Rowlands, A. Hard-castle, K. Boxall, P. Clarke, M.I. Walton, L. Pearl, P. Workman,


Assays for HSP90 and Inhibitors, in: Methods in Molecular Medi-cine: Novel Anticancer Drug Protocols, Humana Press, New Jersey,2003, pp. 149–161.

[30] L. Wright, X. Barril, B. Dymock, L. Sheridan, A. Surgenor, M. Bes-wick, M. Drysdale, A. Collier, A. Massey, N. Davies, A. Fink, C. Fro-mont, W. Aherne, K. Boxall, S. Sharp, P. Workman, R.E. Hubbard,Structure-activity relationships in purine-based inhibitor binding toHSP90 isoforms, Chem. Biol. 11 (2004) 775–785.

[31] Z. Otwinowski, W. Minor, in: C.W. Carter, R.M. Sweet (Eds.), Macro-molecular Crystallography, part A, 276th ed., Academic Press, NewYork, 1997, pp. 307–326.

[32] J. Navaza, Acta Crystallogr. D Biol. Crystallogr. (Pt 10) (2001) 1367–1372.

[33] H.M. Berman, J. Westbrook, Z. Feng, G. Gilliland, T.N. Bhat, H.Weissig, I.N. Shindyalov, P.E. Bourne, The protein data bank, NucleicAcids Res. 28 (2000) 235–242.

[34] T.A. Jones, J.Y. Zou, S.W. Cowan, Kjeldgaard Improved methodsfor building protein models in electron density maps and the loca-tion of errors in these models, Acta Crystallogr. A 47 (Pt 2) (1991)110–119.

[35] G.N. Murshudov, A.A. Vagin, E.J. Dodson, ReWnement of macromo-lecular structures by the maximum-likelihood method, Acta Crystal-logr. D 53 (1997) 240–255.

[36] A. Perrakis, R. Morris, V.S. Lamzin, Automated protein model build-ing combined with iterative structure reWnement, Nat. Struct. Biol. 6(1999) 458–463.

[37] R.A. Laskowski, M.W. MacArthur, D.S. Moss, J.M. Thronton, J.Appl. Crystallogr. 26 (1993) 283.

[38] Collaborative Computational Project Number 4. The CCP4 Suite:Programs for Protein Crystallography, Acta Crystallogr. D 50 (1994)760–763.

[39] Y. Ikuina, N. Amishiro, M. Miyata, H. Narumi, H. Ogawa, T. Akiy-ama, Y. Shiotsu, S. Akinaga, C. Murakata, Synthesis and antitumoractivity of novel O-carbamoylmethyloxime derivatives of radicicol, J.Med. Chem. 46 (2003) 2534–2541.

[40] J.C. Owicki, Fluorescence polarization and anisotropy in highthroughput screening: perspectives and primer, J. Biomol. Screen. 5(2000) 297–306.

[41] G.J. Parker, T.L. Law, F.J. Lenoch, R.E. Bolger, Development of highthroughput screening assays using Xuorescence polarization: nuclearreceptor-ligand-binding and kinase/phosphatase assays, J. Biomol.Screen. 5 (2000) 77–88.

[42] U. Haupts, M. Rudiger, S. Ashman, S. Turconi, R. Bingham, C. Whar-ton, J. Hutchinson, C. Carey, K.J. Moore, A.J. Pope, Single-moleculedetection technologies in miniaturized high-throughput screening:

Xuorescence intensity distribution analysis, J. Biomol. Screen. 8 (2003)19–33.

[43] A.J. Pope, U.M. Haupts, K.J. Moore, Homogeneous Xuorescencereadouts for miniaturized high-throughput screening: theory andpractice, Drug Discov. Today 4 (1999) 350–362.

[44] R.B. Thompson, I. Gryczynski, J. Malicka, Fluorescence polarizationstandards for high-throughput screening and imaging, Biotechniques32 (2002) 34, 37–38, 40, 42.

[45] G. Chiosis, B. Lucas, H. Huezo, D. Solit, A. Basso, N. Rosen, Devel-opment of purine-scaVold small molecule inhibitors of Hsp90, Curr.Cancer Drug Targets 3 (2003) 371–376.

[46] S.M. Roe, C. Prodromou, R. O’Brien, J.E. Ladbury, P.W. Piper, L.H.Pearl, Structural basis for inhibition of the Hsp90 molecular chaper-one by the antitumour antibiotics radicicol and geldanamycin, J. Med.Chem. 42 (1999) 260–266.

[47] C.W. Carreras, A. Schirmer, Z. Zhong, D.V. Santi, Filter bindingassay for the geldanamycin-heat shock protein 90 interaction, Anal.Biochem. 317 (2003) 40–46.

[48] B. Dymock, X. Barril, M. Beswick, A. Collier, N. Davies, M. Drysdale,A. Fink, C. Fromont, R.E. Hubbard, A. Massey, A. Surgenor, L.Wright, Adenine derived inhibitors of the molecular chaperoneHsp90-SAR explained through multiple X-ray structures, Bioorg.Med. Chem. Lett. 14 (2004) 325–328.

[49] G. Chiosis, H. Huezo, N. Rosen, E. Mimnaugh, L. Whitesell, L. Nec-kers, 17AAG: low target binding aYnity and potent cell activity—Wnding an explanation, Mol. Cancer Ther. 2 (2003) 123–129.

[50] P. Workman, Auditing the pharmacological accounts for Hsp90molecular chaperone inhibitors: unfolding the relationship betweenpharmacokinetics and pharmacodynamcs, Mol. Cancer Ther. 2 (2003)131–138.

[51] M.J. Egorin, D.M. Rosen, J.H. WolV, P.S. Callery, S.M. Musser, J.L.Eiseman, Metabolism of 17-(allylamino)-17-demethoxygeldanamy-cin (NSC 330507) by murine and human hepatic preparations, CancerRes. 58 (1998) 2385–2396.

[52] V.G. Brunton, G. Steele, A.D. Lewis, P. Workman, Geldanamycin-induced cytotoxicity in human colon-cancer cell lines: evidenceagainst the involvement of c-Src or DT-diaphorase, Cancer Chemo-ther. Pharmacol. 41 (1998) 417–422.

[53] J. Kim, S. Felts, L. Llauger, H. He, H. Huezo, N. Rosen, G. Chiosis,Development of a Xuorescence polarization assay for the molecularchaperone Hsp90, J. Biomol. Screen. 9 (2004) 375–381.

[54] V. Zhou, S. Han, A. Brinker, H. Klock, J. Caldwell, X.-J. Gu, A time-resolved Xuorescence resonance energy transfer-based HTS assay anda surface plasmon resonance-based binding assay for heat sock pro-tein 90 inhibitors, Anal. Biochem. 331 (2004) 349–357.

A fluorescence polarization assay for inhibitors of Hsp90

Documents