Вы находитесь на странице: 1из 12

Chem Biol Drug Des 2007; 70: 112

Research Article

2007 The Authors Journal compilation 2007 Blackwell Munksgaard doi: 10.1111/j.1747-0285.2007.00535.x

Discovery and Design of Novel HSP90 Inhibitors Using Multiple Fragment-based Design Strategies
Jeffrey R. Huth, Chang Park, Andrew M. Petros, Aaron R. Kunzer, Michael D. Wendt, Xilu Wang, Christopher L. Lynch, Jamey C. Mack, Kerry M. Swift, Russell A. Judge, Jun Chen, Paul L. Richardson, Sha Jin, Stephen K. Tahir, Edward D. Matayoshi, Sarah A. Dorwin, Uri S. Ladror, Jean M. Severin, Karl A. Walter, Diane M. Bartley, Stephen W. Fesik, Steven W. Elmore and Philip J. Hajduk*
Global Pharmaceutical Research and Development, Abbott Laboratories, Abbott Park, IL 60064, USA *Corresponding author: Philip J. Hajduk, philip.hajduk@abbott.com
The molecular chaperone HSP90 has been shown to facilitate cancer cell survival by stabilizing key proteins responsible for a malignant phenotype. We report here the results of parallel fragmentbased drug design approaches in the design of novel HSP90 inhibitors. Initial aminopyrimidine leads were elaborated using high-throughput organic synthesis to yield nanomolar inhibitors of the enzyme. Second site leads were also identied which bound to HSP90 in two distinct conformations, an open and closed form. Intriguingly, linked fragment approaches targeting both of these conformations were successful in producing novel, micromolar inhibitors. Overall, this study shows that, with only a few fragment hits, multiple lead series can be generated for HSP90 due to the inherent exibility of the active site. Thus, ample opportunities exist to use these lead series in the development of clinically useful HSP90 inhibitors for the treatment of cancers. Key words: Drug Design, Heat Shock Protein, NMR Screening, Structure-Based Drug Design Received 14 May 2007, revised and accepted for publication 4 June 2007

prehensive yeast protein interaction study (3). This biology suggests that inhibitors of HSP90 would be broadly toxic with little selectivity between normal and malignant cells. For reasons that are not completely understood, HSP90 inhibitors do exhibit a tolerable safety margin. For instance, derivatives of the natural product geldanamycin, such as 17-allylamino-17-demethoxygeldanamycin (17-AAG), show appropriate safety in animal models and in man (2). This may be due to the dependence of cancer cells on higher levels of molecular chaperones and the subsequent concentration of highaffinity HSP90 inhibitors in these cells (4). Alternatively, 17-AAG and related inhibitors may selectively target HSP90 complexes that are specific to tumor cells (5). The structure of HSP90 is complex and affords multiple avenues for drug targeting. It is composed of three domains that assemble into a dimer (6,7), which then forms a scaffold for co-chaperones (7,8). Of particular interest has been the N-terminal ATPase domain, where the natural product inhibitors geldanamycin (9) and radicicol (10) have been shown to bind. Thus, HSP90 inhibitors could be targeted to the active site as well as numerous proteinprotein interaction sites known to stabilize the functional complex and client protein binding. In addition to these sites, relevant conformational changes are known to occur upon ATP binding (7,11), opening up the possibility to target transient conformational states of the protein. A significant limitation in the clinical use of geldanamycin and the related natural product derivatives 17-AAG and 17-DMAG is the associated liver toxicity and difficult formulation that hinders oral administration (2). Because the liver toxicity is thought to be related to the quinone ring structure rather than direct inhibition of liver HSP90 (2), it has been of high interest to find alternative small molecule inhibitors that are completely distinct from geldanamycin. This approach seems to be successful given that a number of purine- and pyrazole-based inhibitors that bind to the geldanamycin site have been developed (12). Here, we report the use of fragment-based drug design (13,14) in the discovery of aminopyrimidine derivatives that inhibit HSP90. This study shows that with only a few fragment hits, multiple distinct lead compounds can be constructed with the aid of structural studies, particularly when conformational flexibility exists in the binding site.

The codependence of many cancer-related proteins on heat shock protein 90 (HSP90) suggests that inhibitors of this molecular chaperone would have broad efficacy across many tumor types (1). Indeed, numerous oncoproteins including Cdk4, Akt, BCR-ABL, mutated P53, and v-src are known to be client proteins of HSP90 (for review see Ref. 2). In fact, the chaperone activity of HSP90 is so extensive that hundreds of proteins were found to interact with HSP90 in a com-

Results
Identication of rst site ligands In order to identify ligands that can serve as starting points in the development of HSP90 inhibitors, an NMR-based screen of the 1

Huth et al.

Table 1: HSP90 inhibitors identified from NMR fragment-based screening


Number 1 Structure NMR KD (lM) FRET Ki (lM) 0.32 (0.04) MW 316 BEIa 21

NH2 N N N Br NH2

<10b

CF3
2

N N
O

20 (10)

18 (1)

177

27

NH2

150 (20)c

189

N H

Binding efficiency index (BEI) )1000log(Ki)/(MW) (17). In slow exchange on the NMR time-scale. c Determined in the presence of compound 2. d Could not be measured due to the presence of saturating amounts of compound 2. e Not applicable due to the co-operative nature of binding observed between compounds 2 and 3.
a b

N-terminal domain of human HSP90 was conducted using a library of 11 520 compounds with an average molecular weight of 225 Da. Chemical shift changes of Leu, Val, and Ile methyl groups in the presence of compound were detected in two-dimensional 1H/13C heteronuclear single quantum correlation (2D HSQC) NMR experiments and used to confirm compound binding (15). As shown in Table 1, two related chemotypes were discovered, an aminotriazine (represented by 1) and an aminopyrimidine (represented by 2) series. The binding affinities of these compounds were measured in two ways. NMR titration studies (16) indicated that compound 2 binds with a KD of 20 lM. This is in agreement with the results of a complementary Fluorescence Resonance Energy Transfer (FRET) assay where the Ki was determined to be 18 lM. The binding of the larger aminotriazine (compound 1) was found to be 100-fold more potent. By NMR the compound was in slow exchange, consistent with the Ki of 0.32 lM measured in the FRET assay. Both of these leads were attractive starting points for affinity optimization since the potencies relative to the molecular weights are high. One method to quantify this relationship is calculate a binding efficiency index (BEI; 17), where the BEI )1000log(Ki)/(MW). As an example, a drug with a molecular weight of 350 Da and a potency of 1 nM can be described with a BEI of 25.7. By comparison, the binding efficiencies of the compounds identified in the HSP90 screen are high (21 for compound 1 and 27 for compound 2; Table 1). In order to improve the binding affinities of these lead compounds, a structure-based drug design approach 2

was taken. The structure of HSP90 when bound to the aminopyrimidine is similar to that reported for the ADP-bound state (18; Figure 1A). The exocyclic amino groups of 2 and the adenine ring of ADP both make hydrogen bounds to the side chain of D79 as well as a bound water molecule. A second water molecule also makes a key structural hydrogen bond to the N3 pyrimidine nitrogen of 2, analogous to the hydrogen bond observed to the N1 of adenine. Unexpectedly, the naphthyl-substituted aminotriazine was found to induce a conformational change that resulted in opening up a larger binding site even though the key hydrogen bonds from the triazine ring to the water molecules and D79 were unaffected (Figure 1B). This was a clear example of induced-fit binding that was difficult to predict from the structure of the ADP complex alone (18). Furthermore, this suggested that additional binding potency could be obtained by optimally filling the pocket of this open conformation of the catalytic domain. The structures in Figure 1 indicated that the aminotriazine and aminopyrimidine rings bind in a similar way to HSP90 and could likely be interchanged in second-generation compounds. The higher affinity of compound 1 and the associated structural results made it clear that aromatic substitution of the heterocycle could increase the binding affinity. The features of both leads were incorporated in a high-throughput chemical approach in which the trifluoromethyl of 2 was replaced by a variety of aromatic groups according to Scheme 1. One hundred and twenty-eight compounds with aryl substitutions at the four position of the aminopyrimidine were prepared. Chem Biol Drug Des 2007; 70: 112

Discovery and Design of Novel HSP90 Inhibitors


A

core (19). Overall, by combining features of two fragment leads using high-throughput organic synthesis (Scheme 1), a highly optimized, potent, low-molecular weight lead was discovered.

D79

Discovery of a second site ligand While elaboration of 2 into nanomolar inhibitors was successful, the chemical diversity that could be explored was severely limited by the availability of chemical reagents compatible with the chemistry shown in Scheme 1. Moreover, the aminopyrimidine compounds, including 5, were subsequently reported in a patent application (see international patent application WO 2006123165). In order to identify more novel substituents for the adjacent site, a library of 3360 compounds with an average molecular weight of 150 Da was screened by 2D NMR for the ability to bind to HSP90 in the presence of saturating amounts of compound 2. The most potent hit identified in the screen was compound 3 (containing a furanone moiety), which binds to HSP90 with a KD of 150 lM in the presence of compound 2 (Table 1). The binding of 3 was found to be co-operative, as the observed KD is >5000 lM in the absence of compound 2. The furanone hit satisfied the initial objectives of the screen as it was novel, had not been investigated in HSP90 lead optimization, and was known to bind with significant affinity.

D79

Figure 1: X-ray crystallographic structures of NMR screening hits bound to the N-terminal domain of HSP90. Grey surfaces depict the compound binding site for two of the conformations of HSP90 N-terminal domain that were observed. Atom coloring: carbons in magenta, nitrogens in blue, and fluorines in light blue. Structural waters shown as red spheres. Hydrogen bonds indicated with dotted lines. (A) Compound 2 bound to a 'closed' conformation. (B) Compound 1 bound to the 'open' conformation where a larger binding pocket is formed, sufficient to accommodate the naphthyl substituent of compound 1. Compound 4 (Table 2) was among the most potent compounds discovered with a Ki of 170 nM. Minor modification by replacing the 4-methyl with a 4-chloro substituent resulted in a threefold increase in potency (Ki of compound 5 60 nM). Interestingly, the resulting binding efficiency for compound 5 (BEI 26) is very close to that observed for the initial hit (compound 2, BEI 27; Table 1), consistent with the general observation that binding efficiencies tend to remain constant when optimal substituents are added to a fragment Chem Biol Drug Des 2007; 70: 112

Structure-based linking strategy In order to guide the linking of the furanone (3) to the aminopyrimidine (2), structures of the ternary complex were solved by both NMR and X-ray crystallography. The crystal structure (Figure 2A) showed a p-stacking interaction between the phenyl of compound 3 and the pyrimidine ring of compound 2, which is consistent with the co-operativity observed in NMR titration experiments. To accommodate the second site ligand, no conformational change in HSP90 relative to binding of compound 2 alone was observed. Unexpectedly, the NOE data from NMR experiments collected in parallel structural studies could not be reconciled with this X-ray crystallographic structure. In particular, unambiguous NOEs were observed between the furanone moiety of 3 and L103, L107, F138, and V150 of HSP90 (see Figure 2B). These NOEs can only be accommodated if the furanone induces an open conformation similar to that observed for compound 1 (Figure 1B). In fact, simple docking of compounds 2 and 3 to the protein pocket formed by compound 1 (Figure 1B) yielded a ternary complex that satisfied all of the observed intermolecular NOEs. As illustrated in Figure 2B, this alternative conformation packs the furanone ring against the hydrophobic side chain of L107, which undergoes a v1 rotation upon remodeling of the binding pocket. This binding orientation was also consistent with the co-operativity observed in NMR titration experiments because the furanone phenyl ring was calculated to be in van der Waals contact with compound 2 (Figure 2B). As it was clear from the structural studies that compounds 2 and 3 could bind to both the open and closed conformations of the HSP90 N-terminal domain, two separate linking strategies were pursued. For the stacked orientation illustrated in Figure 2A, the linking group must bend by 180. Based on this geometry, a methylsulfonamide linker was suggested to link the aminopyrimidine to the 3

Huth et al.

NH2 N Cl N

NH2

Pd(PPh3)4, CsF

Ar

HO

B OH

Ar

70C, DME / MeOH

Scheme 1: Synthesis of 6-methyl-pyrimidin-2-ylamine library (example: compound 4)

Table 2: HSP90 inhibitors from fragment-based lead optimization


Number 4 Structure FRET Ki (lM) MW 255 BEIa 26

NH2 N Cl Cl NH2 N N

0.17 (0.01)

Cl Cl
O O

N Cl

0.06 (0.01)

274

26

HN

O S

O N N H N NH2

1.9 (0.2)

389

15

O O
7

N
4.0b (1.0) 321 17

N N
a b

NH2

Binding efficiency index (BEI). See legend of Table 1 for a description. Calculated from a direct binding assay where the fluorescence of 7 was detected (see Methods and Materials).

m-position of the furanone phenyl ring to generate compound 6. This design (Scheme 2) yielded a 10-fold improvement in potency relative to compound 2 alone, indicating that favorable interactions between HSP90 and the furanone in the linked state occurred. 4

Indeed, an X-ray crystallographic structure of the linked compound (Table 2, compound 6) bound to HSP90 confirmed that the two fragments bind to the protein in the same way in the linked and unlinked states (Figure 2C). Chem Biol Drug Des 2007; 70: 112

Discovery and Design of Novel HSP90 Inhibitors

L107 L103 V150

F138

Figure 2: Binding of fragment hits compared to binding of linked compounds. Atom coloring for the fragment hits: carbons in magenta, nitrogens in blue, oxygen in red, and fluorines in light blue. Atom coloring for the linked compounds is the same except that carbons are in orange. (A) X-ray crystallographic structure of a ternary complex with compounds 2 and 3. (B) NOE-based model of the same ternary complex with compounds 2 and 3 showing an alternate binding mode observed in solution. The side chains of L107, L103, F138, and V150 are colored in red to indicate key NOEs from HSP90 to 3 that were used to construct the model. (C) X-ray crystallographic structure of a binary complex with compound 6 overlaid with the ternary complex shown in A. Note the near perfect alignment of the linked and unlinked fragments. (D) X-ray crystallographic structure of a binary complex with compound 7 overlaid with the model of the ternary complex shown in B. Note that the oxazolidinone accesses the back of the pocket in an 'open' conformation of HSP90 as suggested by the NOE-based model (shown in B). The design of a furanone-based inhibitor for the open form of HSP90 (Figure 2B) was also pursued. Here, the NMR structural studies indicated that a 90 twist was required to position the furanone near L107 in the open conformation. Twelve compounds were synthesized based on this structural information. An acetylene linker from the aminopyrimidine to the o-position of the phenyl ring of 3-(benzylideneamino)oxazolidin-2-one, an analog of compound 3, satisfied this geometry (Scheme 3), and resulted in a five-fold increase in potency (Table 2, compound 7). X-ray crystallographic Chem Biol Drug Des 2007; 70: 112 studies confirmed that this inhibitor does indeed bind to the open conformation of HSP90 as designed (Figure 2D).

Discussion
Flexibility of HSP90 The compounds reported here probe various conformations of the HSP90 N-terminal domain that may be targeted by therapeutic 5

Huth et al.

O O O

NaH, EtOH O Dioxane / DMF (95%)

O E+Z

O Na

Cl N

N NH2

Zn(CN)2 Pd(PPh3)4 DMF, 80C (77%)

H2 (60 psi) 10% Pd/C


NC N NH2 N

EtOAc / AcOH (97%)

NO2 N H2N N NH2

+
Cl

O S O

NEt3 DCM (20%)

O2N

O S N H O
N

N NH2

AcOH

O O SnCl2 . H2O NMP / EtOH (49%) H2N O S N H O


N NH2 N

O Na

AcOH (40%)

HN O S N H O 6 (E only)
Scheme 2: Synthesis of N-(2-amino-6-methyl-pyrimidin-4-ylmethyl)-3-{[2-oxo-dihydro-furan-(3E)-ylidenemethyl]-amino}-benzenesulfonamide (compound 6) agents. While two states of the domain are depicted in Figure 1, a much broader range of conformations are sampled by the protein. For instance, X-ray crystallographic studies of numerous complexes identified at least six distinct conformations for amino acids A101A123 (C. Park, unpublished observations). Furthermore, most of the complexes studied required co-crystallization with the 6 compound of interest, and multiple crystal forms were observed (R. Judge, unpublished observations). Finally, using NMR spectroscopy, the backbone assignments could not be made for amino acids 105121 because of extensive line broadening, which is consistent with protein mobility (J. Huth, unpublished observations). Taken together, these data indicate a high level of protein Chem Biol Drug Des 2007; 70: 112

N N NH2

Discovery and Design of Novel HSP90 Inhibitors

Cl N N NH2 O

Pd(PhCN)2Cl2 CuI [HP(tBu)3]BF4 DIPEA Dioxane, 50C (91%)

N N NH2

O O O N NH2 O AcOH EtOH, 75C (21%) N 7


Scheme 3: Synthesis of 3-{[1-[2-(2-amino-6-methyl-pyrimidin-4-ylethynyl)-phenyl]-meth-(E)-ylidene]-amino}-oxazolidin-2-one (compound 7) flexibility that may explain how two different binding modes of compound 3 could be captured using both NMR and X-ray crystallography. illustrates the inherent challenge in linking two fragment leads in such a way that the individual binding energies are maximized. We have previously reported that the observed binding affinity for a linked compound is, on average, one log unit less than the sum of the pKD values (negative base-10 logarithm of the KD) for the two compounds (22). Using this rule of thumb, the expected affinity that could be obtained upon linking 2 (pKD 4.7) and 3 (pKD 3.8) would be approximately 30 nM (pKD 7.5). The observed Ki values of 1.9 and 4 lM (Figure 3) would suggest that the linkers used to form 6 and 7 from 2 and 3 are not yet optimal. This is likely due, at least in part, to the limited number of linkers that were evaluated for each binding mode (one for the stacked arrangement and 12 for the other). It is also interesting to consider the binding efficiency indices of the linked compounds. The values obtained for compounds 6 and 7 (BEI values of 15 and 17, respectively) are significantly lower than the binding efficiency of 27 for the core (Figure 3), suggesting that the furanone and oxazolidinone groups, as linked, contribute very inefficiently to the binding energy (19).

N NH2

Implications for fragment-based drug design Particularly for fragment-based drug design, structural information is an important component for guiding the chemistry of elaboration and linking (20). However, the X-ray crystallographic structure of the pyrimidine (compound 2) alone did not reveal the binding site that could be induced and productively exploited by compound 1, nor the subsequent optimization leading to compounds 4 and 5. In addition, the two observed binding modes for the ternary complex with 2 and 3 (Figure 2A,B) were completely unanticipated and, to our knowledge, unprecedented. The fact that both binding modes could be productively exploited to develop parallel lead series that bind with higher affinity to the protein as designed is ample evidence that both modes can in fact be accessed by the ternary complex. While the structural surprises highlight the remarkable insights that structural data can provide on ligand binding, they also caution against the rigid interpretation of any single piece of structural data. It was by taking advantage of multiple structures (and even multiple structural techniques) that the two avenues for optimization were identified. The drug discovery case study described here adds to the list of examples where fragment-based drug design was used to identify novel lead compounds (20,21). However, this case study also Chem Biol Drug Des 2007; 70: 112

Concluding Remarks
We have described how multiple fragment-based drug design strategies capitalized on the same set of NMR screening hits to identify three distinct HSP90 lead inhibitors (Figure 3). In one approach, a fragment elaboration strategy, guided by X-ray crystallographic 7

Huth et al.

Fragment elaboration approach


NH2 N N N NH2
N CF3 N NH2

Methods and Materials


Protein preparation The gene for amino acids D9-E236 of human HSP90 (GenBank accession number NM_005348) was cloned into pET28a resulting in the fusion of (MGSSHHHHHHSSGLVPRGSHM) to the 5-end of the gene. For NMR screening and titration studies HSP90 was expressed in Escherichia coli BL21(DE3) cells and labeled with 13 C at the methyl groups of valine and leucine and the d-methyl of isoleucine by including [3-C-13]-a-ketobutyrate and [3,3 -C-13]a-ketoisovalerate in the medium (15). For NMR structural studies, this clone of HSP90 was expressed in E. coli BL21(DE3) with 15 N-ammonium chloride and 13C-glucose as the sole nitrogen and carbon sources. The protein was purified by nickel affinity chromatography and dialyzed into 30 mM sodium phosphate buffer, pH 7.2, 50 mM NaCl, 5 mM DTT, and 2 mM MgCl2 for all NMR studies. For X-ray crystallography, HSP90 (6His-(Thr)-[Hsp90 (9236)]) was expressed in E. coli, and purified by nickel affinity chromatography. Subsequently, the poly-HIS tag was cleaved with biotinylated thrombin, the thrombin removed with straptavidin-agarose, and the poly-HIS with a second nickel affinity purification. The cleaved HSP90 was then purified by gel filtration chromatography using an S100 column in 25 mM HEPES buffer, pH 7.5, 100 mM NaCl, and 1 mM sodium azide, and concentrated to 50 mg/mL.

1 Ki= 0.32 M BEI = 21

Br

2 Ki= 18 M BEI = 27
Cl

Cl N Cl

N NH2

5 Ki= 0.06 M BEI = 26

Open state B
O O N H

Linked fragment approaches


CF3 N N

2 BEI = 27 NH2

3 KD= 150 M
O O

O O N N

HN

6 Ki= 1.9 M BEI = 15

O O S N H

N N NH2

7 Ki= 4 M BEI = 17
N N NH2 Open state

Closed state

Figure 3: Hit-to-lead approaches using NMR screening hits. Structural studies were used to guide (A) the high-throughput organic synthesis that incorporated a fragment elaboration approach and (B) the linked fragment approaches as described in the text. Atoms are colored in blue and green to highlight components from each lead, and in red to identify linking groups (panel B only). Atoms in the NMR hits that are colored in black were not used in the final compounds. Binding efficiency indices were calculated as described in the legend for Table 1.

structures of HSP90 ligand complexes and high-throughput organic synthesis, resulted in the rapid discovery of a 60 nM inhibitor (Figure 3A). In a parallel approach, SAR by NMR and X-ray crystallography were effective in the discovery of two additional series (Figure 3B). Surprisingly, the two structural methods identified distinct binding modes for fragments in a ternary complex. In one case, binding to the open conformation of the N-terminal domain was observed. For the other, we observed binding to a closed conformation. With further optimization, one or more of these series could result in HSP90 inhibitors with clinical utility for the treatment of tumors that rely on this chaperone for proliferation and survival.

NMR-based screening Nuclear magnetic resonance-based screening (23) of HSP90 was conducted by acquiring 1H13C HSQC spectra of 2550 lM protein samples. About 11 500 compounds of diverse structure with and average molecular weight of 225 Da were solvated in 2H-DMSO to 1 M. Mixtures of 10 compounds were then randomly prepared in 96-well plates (primary library). A second library of mixtures where each mixture was prepared by combining three mixtures of 10 was prepared in 96-well plates. The final concentration of each compound in the mixture was 33.3 mM. For screening, 6 lL of each 30-compound mixture was mixed with 0.5 mL of a 25 lM protein NMR sample resulting in 400 lM of each compound and 1.2% DMSO. Subsequently, mixture hits were deconvoluted to mixtures of 10 using the primary library. A second round of deconvolutions, where individual compounds were tested at 400 lM, resulted in identification of the fragment hits. Spectra were acquired in 1530 min with 38 complex points on a Bruker (Bruker Biospin, Billerica, MA, USA) 500 MHz spectrometer equipped with a cryoprobe. For titration studies, HSQC spectra of 25 lM protein samples were acquired with 20, 40, 60, 120, 240, or 600 lM compound 2; 20, 50, or 100 lM compound 1, which was found to be in slow exchange on the NMR time-scale. For the second site screen, 3360 compounds with an average molecular weight of 150 Da were screened at 5 mM in mixtures of five compounds against 50 lM HSP90 in the presence of 200 lM compound 2 (final concentration of DMSO 3%). Mixtures that caused additional chemical shift changes in addition to those caused by compound 2 were deconvoluted by testing individual compounds at 5 mM in the presence of 200 lM compound 2. The binding affinity of compound 3 was measured by NMR titration

Chem Biol Drug Des 2007; 70: 112

Discovery and Design of Novel HSP90 Inhibitors

studies using 50, 150, 400, 1500, or 5000 lM compound 3 in the presence and absence of 200 lM compound 2. Dissociation constants were obtained by monitoring the chemical shift changes as a function of ligand concentration using a single binding site model. A least-squares grid search was performed by varying the values of KD and the chemical shift of the fully saturated protein. Errors in the dissociation constants were obtained using a Monte Carlo simulation of the data assuming a Gaussian distribution for errors in chemical shifts with a standard deviation of 0.01 p.p.m. 100 Monte Carlo simulations were performed for each dissociation constant, and the reported errors are the standard deviations of the simulated values.

Fluorescence Resonance Energy Transfer assay Compounds were serially diluted into assay buffer containing 50 mM Tris (pH 7.5), 150 mM NaCl, 1 mM EDTA, 1 mM DTT, 0.05% Triton-X-100, 0.04 lM Hsp90, 0.08 lM geldanamycin-biotin, 0.08 lM streptavidin-APC (Perkin-Elmer, Catalog number R130-100, PerkinElmer, Waltham, MA, USA), and 0.001 lM Eu-W1024-labeled anti6XHis antibody (Perkin-Elmer, Catalog number AD0110) for 3 h at room temperature. In the (geldanamycin-biotin/streptavidin-APC/ HSP90/Eu-labeled anti-6XHis antibody) complex, fluorescence energy transfer occurs from europium on the anti-6XHis antibody to APC on the streptavidin. Following excitation of Eu at 340 nM, the fluorescence transfer was detected as the ratio of fluorescence intensity at 655 nm (emission of APC) to that at 615 nm (emission of Eu) using the Envision plate reader (Perkin-Elmer). A decrease in fluorescence energy transfer occurred upon displacement of geldanamycin-biotin with added compound. The resulting decrease in fluorescence energy transfer was used to calculate compound IC50 values with the GRAPHPAD software package. The average and standard deviations of replicate values are reported.

150 of HSP90 were assigned using data from a 13C-edited, 12C-filtered NOESY spectrum and a 13C-edited NOESY spectrum (80 mseconds mixing time for both) recorded on a complex of Val, Leu, Ile (d1) methyl-protonated 15N-, 13C-, and 2H-labeled HSP90(9-236) (25) with unlabeled ADP and 5 mM MgCl2, in conjunction with the crystal structure of Mg-ADP-bound HSP90 [pdb 1BYQ (18)]. Likewise, the side chain resonances of F138 were assigned from NOE data recorded on a uniformly 13C- and 15N-labeled sample of HSP0 in conjunction with the crystal structure. NMR samples of the ternary complex of HSP90 with compounds 2 and 3 were 1 mM in protein, 1.6 mM in compound 2, and 2.4 mM in compound 3. Protein-ligand NOEs were extracted from 13C-edited, 12C-filtered NOESY spectra (80 mseconds mixing time) recorded on either uniformly 13C- and 15 N-labeled protein, or Val, Leu, Ile (d1) methyl-protonated 15N-, 13 C-, and 2H-labeled protein. For the model of the ternary complex, compound 2 was first docked into the binding site of HSP90 (open conformation) based on the crystal structures of compound 1 and that of compound 2 alone (Figure 1). This was performed with the program INSIGHT II (Accelrys, San Diego, CA, USA). Compound 3 was then manually docked into the HSP90, ATP-binding site based on eight observed intermolecular NOEs. Only the open conformation of the binding site was consistent with the observed intermolecular NOEs.

Fluorescence spectroscopy for compound 7 binding An assay for the inhibition of HSP90 by compound 7 was performed using the intrinsic fluorescence of the compound in the manner like that used to monitor warfarin binding to a single binding site on albumin (24). Compound 7 has broad, weak fluorescence emission. Using an excitation wavelength of 325 nm, the fluorescence spectral maximum of the compound shifts upon binding from 415 to 395 nm, and intensifies approximately 3.5fold. Three titrations were compared, using 2, 5, and 20 lM compound and titrating the protein to 90 lM. After subtracting background fluorescence arising from the protein, the KD was calculated to be 5.1, 3.0, or 2.8 lM, respectively, using the intensity at 390 nm. The independence of the KD on compound concentration supports the assumption that 7 binds to a single site on HSP90.

Model of ternary complex based on NMR structural studies All NMR experiments were carried out at 30 C on either a Bruker (Bruker Biospin) DRX500 or DRX600 spectrometer, both with a cryoprobe accessory. The side chain resonances of L103, L107, and Val Chem Biol Drug Des 2007; 70: 112

X-ray crystallography X-ray diffraction data were collected at the APS 17 BM IMCA beamline of Argonne National Laboratory. Co-crystallization was used to obtain crystal structures for each of the compounds. Compound stock solutions were made by dissolving compound in dimethyl sulfoxide (DMSO). Compound and protein were complexed in a 3:1 molar ratio, giving a final DMSO concentration of 2.42.7% (v/v) and incubated overnight at 4 C prior to crystallization. Crystallization was performed using the vapor diffusion (hanging drop) technique. The drops were made using 2 lL of protein solution combined with 2 lL of reservoir solution and suspended over a 1 mL reservoir. Crystals grew within 4 days at 4 C. For compound 1, crystals grew from 30 mg/mL protein with a reservoir solution of 22% (w/v) PEG 8000, 0.1 M sodium cacodylate pH 6.5, 0.2 M ammonium sulfate [adapted from published conditions (9)]. Crystals were cryo-protected using reservoir solution with 25% (v/v) glycerol. The crystal diffracted to 3.1 with space group P43212 (a b 118.8, c 180.8, a b c 90). Compounds 2 and 3 were obtained as a ternary complex. Protein was incubated with compound 2 overnight with a compound to protein molar ratio of 4:1. Compound 3 was then added at a compound to protein molar ratio of 6:1 and incubated for 6 h. Crystallization was performed using protein at 35 mg/mL with a reservoir of 2230% (w/v) PEG 4000, 0.1 M TrisHCl pH 8.5, 0.2 M MgCl2 [adapted from published conditions (9)]. The crystals were cryo-protected by increasing the amount of PEG 4000 to 35% (w/v) and adding 10% (v/v) glycerol. In this instance, both compounds were present in the cryo-protectant solution at 0.5 mM. The crystal diffracted to 1.7 with space group P21212 (a 64.16, b 88.76, c 98.45, a b c 90). For compound 6, crystals grew from 40 mg/mL protein with a reservoir solution of 2028% (w/v) PEG 3350, 0.1 M Bis-Tris-Propane pH 9

Huth et al.

6.5, 0.2 M sodium chloride. The crystals were cryo-protected using reservoir solution with 15% (v/v) glycerol. The crystals diffracted to 1.9 with space group C2 (a 113.35, b 89.14, c 64.46, a 90, b 121.45, c 90). For compound 7, crystals grew from 40 mg/mL protein with a reservoir solution of 818% (w/v) polyethyleneglycol monomethyl ether (PMME) 2000, 0.1 M sodium cacodylate pH 6.5, 0.2 M MgCl2 [adapted from published conditions (26)]. Crystals were cryo-protected by increasing the PMME concentration to 35% (w/v). The crystal diffracted to 1.8 with space group I222 (a 66.79, b 91.38, c 98.34, a b c 90).

Chemistry Geldanamycin-biotin Geldanamycin-biotin was synthesized with minor modifications from the method described previously (27). The product was verified pure (>95%) by reverse-phase HPLC and identified by electrospray ionization mass spectrometry ( MESI-MS), m/z 1097.7 [(M + H)+].

4-(2,4-Dichloro-phenyl)-6-methyl-pyrimidin-2ylamine (compound 4) To a mixture of 2-amino-4-chloro-6-methylpyrimidine (143 mg, 1.0 mmol) and Pd(PPh3)4 (58 mg, 0.05 mmol) in DME (4 mL) and MeOH (2 mL) was added 2,4-dichlorophenylboronic acid (228 mg, 1.2 mmol) immediately followed by CsF (456 mg, 3.0 mmol). The mixture was purged with argon, sealed and heated up to 70 C for 12 h. The mixture was diluted with EtOAc (200 mL) and washed with water then brine. After drying over Na2SO4, evaporation of the solvent gave crude product, which was dissolved in DMSO and MeOH (1:1, 3 mL) and purified by preparative HPLC on a Waters Nova-Pak (Waters, Milford, MA, USA) HR C18 6 lm 60 Prep-Pak cartridge column (40 100 mm) to give 4 as the mono TFA salt (328 mg, 89%). 1H-NMR (300 MHz, DMSO-d6) d p.p.m. 7.75 (t, 1H), 7.56 (m, 2H), 6.77 (s, 1H), 2.33 (s, 3H). MS (ESI) m/e 252/254, 254/256. Analytical calculated for C11H9Cl2N3C2HF3O2: C 42.41%, H 2.74%, N 11.41%; found: C 42.66%, H 2.44%, N 11.36%.

4-(2,4-Dichloro-phenyl)-6-chloro-pyrimidin-2ylamine (compound 5) To a mixture of 2-amino-4,6-dichloropyrimidine (163 mg, 1.0 mmol) and Pd(PPh3)4 (58 mg, 0.05 mmol) in DME (20 mL) and MeOH (10 mL) was added 2,4-dichlorophenylboronic acid (228 mg, 1.2 mmol) immediately followed by CsF (456 mg, 3.0 mmol). The mixture was purged with argon, sealed and heated up to reflux for 4 h. The reaction mixture was evaporated under vacuum, dissolved in DCM (200 mL), washed with water, brine, and dried on Na2SO4. Solvent volume was reduced on vacuum, and the residue was purified on silica gel (EtOAc:hexanes, 1:4 to 1:2) to give 5 (135 mg, 49%). 1H-NMR (300 MHz, CDCl3) d p.p.m. 7.55 (d, 1H), 7.51 (d, 1H), 7.36 (dd, 1H), 7.01 (s, 1H), 5.34 (s, 2H). MS (ESI) m/e 272/274, 274/276.

N-(2-Amino-6-methyl-pyrimidin-4-ylmethyl)-3{[2-oxo-dihydro-furan-(3E)-ylidenemethyl]amino}-benzenesulfonamide (compound 6) A mixture of c-butyrolactone (12.00 g, 139.4 mmol) and ethyl formate (10.33 g, 139.4 mmol) was added dropwise to a solution of sodium hydride (60%, 5.85 g, 146.4 mmol) and ethanol (1 mL) in 1,4-dioxane (360 mL) and DMF (45 mL). After the addition was complete, the solution was stirred at room temperature overnight. Diethyl ether (300 mL) was then added, and the resulting precipitate was filtered on vacuum, washed with diethyl ether, and dried further on vacuum to give a mixture of E- and Z-isomers of the sodium enolate product: sodium; [2-oxo-dihydro-furan-3-ylidene]methanolate (18.06 g, 95%). Separately, 2-amino-4-chloro-6-methylpyrimidine (12.00 g, 83.6 mmol), Zn(CN)2 (5.40 g, 46.0 mmol), and Pd(PPh3)4 (4.83 g, 4.2 mmol) were heated at 80 C in DMF (220 mL) for 72 h. The solution was cooled, added to water (700 mL) and extracted with EtOAc (3 300 mL). The combined extracts were dried with brine, then Na2SO4. Solvent was removed under reduced pressure, and two recrystallizations of the crude material (EtOAc) gave 2-amino-4-cyano-6-methylpyrimidine (8.63 g, 77%). 2-Amino-4cyano-6-methylpyrimidine (8.62 g, 64.3 mmol) was dissolved in EtOAc (150 mL) and AcOH (150 mL). And, 10% Pd/C (1.72 g, 1.6 mmol) was added, and the solution was subjected to an H2 atmosphere (60 psi). The mixture was stirred at room temperature for 2 h. Solvents were removed on vacuum, and the crude material was washed with 70% EtOAc (hexanes) to give 2-amino-4-aminomethyl6-methylpyrimidine as the mono AcOH salt (12.51 g, 98%). 2-Amino-4-aminomethyl-6-methylpyrimidine mono AcOH (2379 mg, 12.0 mmol) and NEt3 (2429 mg, 24.0 mmol) were dissolved in DCM (60 mL). 3-Nitrobenzenesulfonyl chloride (2659 mg, 12.0 mmol) was added, and the mixture was stirred at room temperature for 30 min. The solution was added to water (100 mL), adjusted to a pH of 8, and extracted with EtOAc (3 50 mL). The combined extracts were dried with brine, then Na2SO4 and concentrated. The material was purified on silica gel (5% MeOH/EtOAc) to give N-(2-amino-6-methyl-pyrimidin-4-ylmethyl)-3-nitro-benzenesulfonamide (765 mg, 20%). N-(2-Amino-6-methyl-pyrimidin-4-ylmethyl)3-nitro-benzenesulfonamide (760 mg, 2.4 mmol) was dissolved in NMP (4 mL). EtOH (0.2 mL) and SnCl2H2O were subsequently added, and the mixture was heated to 70 C overnight. The solution was cooled, added to water (12 mL), adjusted to a pH of 7, and extracted with EtOAc (3 8 mL). The combined extracts were dried with brine, then Na2SO4 and concentrated. The material was purified on silica gel (5% MeOH/EtOAc) to give 3-amino-N-(2amino-6-methyl-pyrimidin-4-ylmethyl)-benzenesulfonamide (340 mg, 49%). To 3-amino-N-(2-amino-6-methyl-pyrimidin-4-ylmethyl)-benzenesulfonamide (100 mg, 0.3 mmol) and the previously synthesized [2-oxo-dihydro-furan-3-ylidene]-methanolate (46 mg, 0.3 mmol) was added AcOH (3 mL). The solution was stirred at room temperature overnight. The solution was added to water (10 mL), adjusted to a pH of 7, and extracted with EtOAc (3 5 mL). The combined extracts were dried with Na2SO4. The solution volume was reduced on vacuum to precipitate pure 6 exclusively as the E-isomer (53 mg, 40%). 1H-NMR (300 MHz, DMSO-d6) d p.p.m. 9.31 (d, 1H), 8.16 (t, 1H), 7.69 (dt, 1H), 7.57 (t, 1H), 7.517.38 (m, 2H), 7.32 (dt, 1H), 6.41 (s, 2H), 6.38 (s, 1H), 4.32 (t, 2H), 3.80 (d, 2H), 2.89 (td, 2H), 2.15 (s, 3H). Analytical calculated for C17H19N5O4S0.5H2O: C 51.25%, H

10

Chem Biol Drug Des 2007; 70: 112

Discovery and Design of Novel HSP90 Inhibitors

5.06%, N 17.58%; found: C 51.46%, H 5.11%, N 17.37%. Structure was confirmed by X-ray crystallography (Figure 2C).

3-{[1-[2-(2-Amino-6-methyl-pyrimidin-4ylethynyl)-phenyl]-meth-(E)-ylidene]-amino}oxazolidin-2-one (compound 7) 2-Amino-4-chloro-6-methylpyrimidine (459 mg, 3.2 mmol), 2-ethynylbenzaldehyde (500 mg, 3.8 mmol), and N,N-diisopropylethylamine (496 mg, 3.8 mmol) were dissolved in 1,4-dioxane (5 mL). Dichlorobis(benzonitrile)palladium(II) (37 mg, 0.10 mmol), copper(I) iodide (12 mg, 0.06 mmol), and tri-t-butylphosphonium tetrafluoroborate (61 mg, 0.21 mmol) were added, and the solution was heated at 50 C overnight. The mixture was cooled, concentrated, and purified on silica gel (50100% EtOAc/hexanes) to give 2-(2-amino-6-methylpyrimidin-4-ylethynyl)-benzaldehyde (693 mg, 91%). 2-(2-Amino-6methyl-pyrimidin-4-ylethynyl)-benzaldehyde (50 mg, 0.21 mmol) and 3-amino-oxazolidin-2-one (21 mg, 0.21 mmol) were dissolved in ethanol (1.5 mL) and acetic acid (one drop) was added. The solution was heated at 75 C for 105 min. The mixture was cooled, concentrated, and purified on silica gel (5% MeOH/EtOAc) to give 7 (14 mg, 21%). 1H-NMR (300 MHz, DMSO-d6) d p.p.m. 8.08 (s, 1H), 7.58 (dd, 1H), 7.65 (dd, 1H), 7.597.47 (m, 2H), 6.77 (s, 1H), 6.71 (s, 2H), 4.53 (t, 2H), 4.03 (t, 2H), 2.26 (s, 3H). MS (ESI) m/e 300/ 322. Structure was confirmed by X-ray crystallography (Figure 2D).

Acknowledgment
The authors would like to thank Dr Jonathan Greer for critically reading the manuscript.

Note added in proof


The PDB accession codes for the X-ray crystallography structures shown in Figures 1 and 2 are: 2QF6, 2QF0, 2QG0, 2QG2.

References
1. Solit D.B., Rosen N. (2006) Hsp90: a novel target for cancer therapy. Curr Top Med Chem;6:12051214. 2. Cullinan S.B., Whitesell L. (2006) Heat shock protein 90: a unique chemotherapeutic target. Semin Oncol;33:457465. 3. Zhao R., Davey M., Hsu Y.C., Kaplanek P., Tong A., Parsons A.B., Krogan N., Cagney G., Mai D., Greenblatt J., Boone C., Emili A., Houry W.A. (2005) Navigating the chaperone network: an integrative map of physical and genetic interactions mediated by the hsp90 chaperone. Cell;120:715727. 4. Gooljarsingh L.T., Fernandes C., Yan K., Zhang H., Grooms M., Johanson K., Sinnamon R.H., Kirkpatrick R.B., Kerrigan J., Lewis T., Arnone M., King A.J., Lai Z., Copeland R.A., Tummino P.J. (2006) A biochemical rationale for the anticancer effects of Hsp90 inhibitors: slow, tight binding inhibition by geldanamycin and its analogues. Proc Natl Acad Sci U S A;103:7625 7630.

5. Kamal A., Thao L., Sensintaffar J., Zhang L., Boehm M.F., Fritz L.C., Burrows F.J. (2003) A high-affinity conformation of Hsp90 confers tumour selectivity on Hsp90 inhibitors. Nature;425:407410. 6. Chadli A., Bouhouche I., Sullivan W., Stensgard B., McMahon N., Catelli M.G., Toft D.O. (2000) Dimerization and N-terminal domain proximity underlie the function of the molecular chaperone heat shock protein 90. Proc Natl Acad Sci USA;97: 1252412529. 7. Ali M.M., Roe S.M., Vaughan C.K., Meyer P., Panaretou B., Piper P.W., Prodromou C., Pearl L.H. (2006) Crystal structure of an Hsp90-nucleotide-p23/Sba1 closed chaperone complex. Nature;440:10131017. 8. Meyer P., Prodromou C., Hu B., Vaughan C., Roe S.M., Panaretou B., Piper P.W., Pearl L.H. (2003) Structural and functional analysis of the middle segment of hsp90: implications for ATP hydrolysis and client protein and cochaperone interactions. Mol Cell;11:647658. 9. Stebbins C.E., Russo A.A., Schneider C., Rosen N., Hartl F.U., Pavletich N.P. (1997) Crystal structure of an Hsp90-geldanamycin complex: targeting of a protein chaperone by an antitumor agent. Cell;89:239250. 10. Roe S.M., Prodromou C., O'Brien R., Ladbury J.E., Piper P.W., Pearl L.H. (1999) Structural basis for inhibition of the Hsp90 molecular chaperone by the antitumor antibiotics radicicol and geldanamycin. J Med Chem;42:260266. 11. Huai Q., Wang H., Liu Y., Kim H.Y., Toft D., Ke H. (2005) Structures of the N-terminal and middle domains of E. coli Hsp90 and conformation changes upon ADP binding. Structure;13:579590. 12. McDonald E., Workman P., Jones K. (2006) Inhibitors of the HSP90 molecular chaperone: attacking the master regulator in cancer. Curr Top Med Chem;6:10911107. 13. Shuker S.B., Hajduk P.J., Meadows R.P., Fesik S.W. (1996) Discovering high-affinity ligands for proteins SAR by NMR. Science;274:15311534. 14. Hajduk P.J., Meadows R.P., Fesik S.W. (1997) Drug design discovering high-affinity ligands for proteins. Science;278:497499. 15. Hajduk P.J., Augeri D.J., Mack J., Mendoza R., Yang J.G., Betz S.F., Fesik S.W. (2000) NMR-based screening of proteins containing C-13-labeled methyl groups. J Am Chem Soc;122: 78987904. 16. Hajduk P.J., Sheppard G., Nettesheim D.G., Olejniczak E.T., Shuker S.B., Meadows R.P., Steinman D.H. et al. (1997) Discovery of potent nonpeptide inhibitors of stromelysin using SAR by NMR. J Am Chem Soc;119:58185827. 17. Cele A.Z., Metz J.T. (2005) Ligand efficiency indices as guideposts for drug discovery. Drug Discov Today;10:464469. 18. Prodromou C., Roe S.M., O'Brien R., Ladbury J.E., Piper P.W., Pearl L.H. (1997) Identification and structural characterization of the ATP/ADP-binding site in the Hsp90 molecular chaperone. Cell;90:6575. 19. Hajduk P.J. (2006) Fragment-based drug design: how big is too big? J Med Chem;49:69726976. 20. Hajduk P.J., Greer J. (2007) A decade of fragment-based drug design: strategic advances and lessons learned. Nat Rev Drug Discov;6:211219. 21. Erlanson D.A., McDowell R.S., O'Brien T. (2004) Fragment-based drug discovery. J Med Chem;47:34633482.

Chem Biol Drug Des 2007; 70: 112

11

Huth et al.

22. Hajduk P.J., Huth J.R., Sun C. (2006) SAR by NMR: an analysis of potency gains realized through fragment-linking and fragment-elaboration strategies for lead generation. In: Jahnke W., Erlanson D.A., editors. Fragment-based Approaches in Drug Discovery. Weinheim: Wiley-VCH;p. 181192. 23. Huth J.R., Sun C., Sauer D.R., Hajduk P.J. (2005) Utilization of NMR-derived fragment leads in drug design. Methods Enzymol;394:549571. 24. Maes V., Engelborghs Y., Hoebeke J., Maras Y., Vercruysse A. (1982) Fluorimetric analysis of the binding of warfarin to human serum albumin. Equilibrium and kinetic study. Mol Pharmacol;21:100107.

25. Goto N.K., Gardner K.H., Mueller G.A., Willis R.C., Kay L.E. (1999) A robust and cost-effective method for the production of Val, Leu, Ile (delta 1) methyl-protonated 15N-, 13C-, 2H-labeled proteins. J Biomol NMR;13:369374. 26. Wright L. et al. (2004) Structure-activity relationships in purinebased inhibitor binding to HSP90 isoforms. Chem Biol;11:775785. 27. Zhou V., Han S., Brinker A., Klock H., Caldwell J., Gu X.J. (2004) A time-resolved fluorescence resonance energy transfer-based HTS assay and a surface plasmon resonance-based binding assay for heat shock protein 90 inhibitors. Anal Biochem;331:349357.

12

Chem Biol Drug Des 2007; 70: 112

Вам также может понравиться