Fragment-based drug discovery applied to Hsp90. Discovery of two lead series with high ligand efficiency

Article abstract of DOI:10.1021/jm100059d

Inhibitors of the chaperone Hsp90 are potentially useful as chemotherapeutic agents in cancer. This paper describes an application of fragment screening to Hsp90 using a combination of NMR and high throughput X-ray crystallography. The screening identified an aminopyrimidine with affinity in the high micromolar range and subsequent structure-based design allowed its optimization into a low nanomolar series with good ligand efficiency. A phenolic chemotype was also identified in fragment screening and was found to bind with affinity close to 1 mM. This fragment was optimized using structure based design into a resorcinol lead which has subnanomolar affinity for Hsp90, excellent cell potency, and good ligand efficiency. This fragment to lead campaign improved affinity for Hsp90 by over 1000000-fold with the addition of only six heavy atoms. The companion paper (DOI: 10.1021/jm100060b) describes how the resorcinol lead was optimized into a compound that is now in clinical trials for the treatment of cancer.

Full text of DOI:10.1021/jm100059d

5942 J. Med. Chem. 2010, 53, 5942–5955
DOI: 10.1021/jm100059d
Fragment-Based Drug Discovery Applied to Hsp90. Discovery of Two Lead Series with
High Ligand Efficiency†
Christopher W. Murray,* Maria G. Carr, Owen Callaghan, Gianni Chessari, Miles Congreve, Suzanna Cowan,
Joseph E. Coyle, Robert Downham, Eva Figueroa, Martyn Frederickson, Brent Graham, Rachel McMenamin,
M. Alistair O’Brien, Sahil Patel, Theresa R. Phillips, Glyn Williams, Andrew J. Woodhead, and Alison J.-A. Woolford
Astex Therapeutics, Ltd., 436 Cambridge Science Park, Milton Road, Cambridge CB4 0QA, U.K.
Received January 15, 2010
Inhibitors of the chaperone Hsp90 are potentially useful as chemotherapeutic agents in cancer. This
paper describes an application of fragment screening to Hsp90 using a combination of NMR and high
throughput X-ray crystallography. The screening identified an aminopyrimidine with affinity in the
high micromolar range and subsequent structure-based design allowed its optimization into a low
nanomolar series with good ligand efficiency. A phenolic chemotype was also identified in fragment
screening and was found to bind with affinity close to 1 mM. This fragment was optimized using
structure based design into a resorcinol lead which has subnanomolar affinity for Hsp90, excellent cell
potency, and good ligand efficiency. This fragment to lead campaign improved affinity for Hsp90 by
over 1000000-fold with the addition of only six heavy atoms. The companion paper (DOI: 10.1021/
jm100060b) describes how the resorcinol lead was optimized into a compound that is now in clinical
trials for the treatment of cancer.
Introduction
Molecular chaperones are proteins that play a role in the
conformational stability, maturation, and function of other sub-
strate proteins, which are known as clients.1 Heat shock protein
90 (Hsp90a) is a molecular chaperone, and many of its clients are
oncology targets that are known to play critical roles in cancer
progression.2,3 In fact it has been argued that interference with
Hsp90 and its associated clients can allow the simultaneous
targeting of the hallmarks of cancer, i.e., self-sufficiency in
growth signals, insensitivity to antigrowth and apoptotic signal-
ing, angiogenesis, limitless replicative potential, invasion, and
metastasis.2,4,5 It was initially thought that inhibition of the
Hsp90 chaperone cycle would lead to unacceptable toxicities
due to the ubiquitous expression of Hsp90 in normal cells.6
Figure 1. Natural product derived inhibitors of Hsp90. Note that
17-DMAG, 17AAG, and the quinol form of 17AAG (i.e., IPI-504)
have been progressed into clinical trials.
However, later work has shown that Hsp90 inhibition can lead
number of Hsp90 inhibitors have been determined.9 It has also
to reasonably specific inhibition in cancer cells because Hsp90 is
been demonstrated that inhibition leads directly to the down-
present in an activated form.6,7 Additionally, the preclinical and,
regulation of client proteins and antiproliferative activity.10 At
more recently, clinical activity of geldanamycin-based inhibitors
the start of our Hsp90 program the most advanced Hsp90 inhi-
has indicated the potential utility of Hsp90 inhibitors.2,8
bitors were derived from natural products11-15 (see Figure 1)
The chaperone cycle of Hsp90 involves the turnover of ATP
and our aim was to use fragment-based drug discovery (FBDD)
to ADP through an ATPase activity associated with the
to derive a synthetic inhibitor of Hsp90 with improved pharma-
N-terminal domain of Hsp90.9 The ATP binding site has been
ceutical properties. There are now a number of reviews describ-
characterizedcrystallographically, and the binding modes of a
ing synthetic Hsp90 inhibitors,16-19 and Figure 2 gives structures
of compounds currently in clinical development.20-25
†Coordinates for the Hsp90 complexes with compounds 1, 3, 4, 2 þ 4, 6,
Fragment-based drug discovery (FBDD) is gathering mo-
mentum in the pharmaceutical industry, and a number of
clinical candidates or advanced leads have been identified via
14, 21, 24, and 31 have been deposited in the Protein Data Bank (PDB)
under accession codes 2xdk, 2xdl, 2xdu, 2xds, 2xdx, 2xhr, 2xht, 2xhx, and
2xab, respectively, together with the corresponding structure factor files.
*To whom correspondence should be addressed. Phone: þ44 (0)1223
the approach.26-34 A recent review has listed the key concepts
that lie behind the rise in FBDD, and we recapitulate these to
226228. Fax þ44 (0)1223 226201. E-mail: c.murray@astex-therapeutics.com.
aAbbreviations: CDK4, cyclin dependent kinase 4; FBDD, fragment-
based drug discovery; Hsp70, heat shock protein 70; Hsp90, heat shock
provide some background to the field. The first concept is that
inappropriate physical properties are thought to be a major
protein 90; ITC, isothermal titration calorimetry; LE, ligand efficiency;
PDB, Protein Data Bank.
driver for attrition in drug development,35-38 and there is
r
pubs.acs.org/jmc
Published on Web 07/27/2010
2010 American Chemical Society

Article
Journal of Medicinal Chemistry, 2010, Vol. 53, No. 16 5943
Figure 2. Synthetic inhibitors of Hsp90 currently in clinical trails.
therefore a need to explore alternative methods, such as
FBDD, that might deliver candidates with improved physical
properties. It will take many FBDD applications across many
different targets before progress in this regard can be properly
assessed. Fragments are compounds of low molecular weight
(usually 100-250 Da) and typically have low binding affi-
nities (>100 μM), but a second key concept is that fragments
form high-quality interactions despite their weak potency and
therefore can provide attractive starting points for medicinal
chemistry. In our application we will show that this is true for
Hsp90, although no single application can be used to assess a
general point. A third concept important to FBDD is that
ligand efficiency (LE)39,40 can be used to track the potency of
fragment hits during lead identification and to assess whether
gains in potency are significant enough to justify increases in
molecular size. LE can be defined as
efficiently optimized into potent lead series. The results are
discussed together with any learning points that are relevant
to either Hsp90 inhibition or FBDD. A companion paper
(DOI: 10.1021/jm100060b) describes how one of these leads
was optimized into an Hsp90 inhibitor that is currently in
clinical trials for the treatment of cancer.
Results and Discussion
Fragment Screening for Hsp90. Approximately 1600 com-
pounds from our fragment library43 were screened in cock-
tails using ligand observed NMR via water LOGSY.44
Compounds showing a medium or strong LOGSY signal
(as defined in the Experimental Section) were further char-
acterized according to competition for the nucleotide site.
NMR screening was carried out in the presence of a low
concentration of the product, ADP, which binds weakly to
the ATP-ase domain of Hsp90 under the screening condi-
tions. Cocktails that contained fragments that bound to the
nucleotide site of the target could be identified immediately
by observing the displacement of ADP, estimated from the
reduction in its own LOGSY signal. Further information on
the affinity of the fragment at this site was obtained in a
second step by adding 5 mM Mg2þ to the screening solution.
This increases the affinity of ADP for the active site and leads
to the displacement of more weakly bound fragments. A
total of 125 fragments were progressed into crystallography
after analysis of the NMR data and a consideration of
chemical diversity. Of these compounds 26 fragment crystal
structures were obtained, spanning a variety of chemotypes.
The chemical structures for four of the validated hits are
shown in Figure 3.
- ΔG
HAC
- RT lnðKdÞ
LE ¼
¼
HAC
where ΔG is the free energy of binding of the ligand for a
specific protein, HAC is the number of heavy atoms in the
ligand, and Kd represents the dissociation constant for the
protein-ligand complex (IC50 is often used instead of Kd).
This paper illustrates the use of LE during fragment optimiza-
tion against Hsp90 but also illustrates that LE is not the only
criteria that should be used to select which fragments to
optimize. The fourth concept behind FBDD is that relatively
small libraries of fragments are required to sample chemical
space and some elegant models have been described in the
literature to illustrate this behavior.41,42 Here no analysis of
the sampling advantages of fragments is presented but our
application does show the identification of two distinct che-
mical leads for Hsp90 derived from screening a library of only
1600 fragments.
All affinities given in this paper are dissociation constants
determined by isothermal titration calorimetry (ITC) (see
Experimental Section for more details and for ITC values
obtained with reference compounds). All ligand efficiencies
are calculated directly from the ITC dissociation constants
and have units of kcal per heavy atom.
The motivation of this paper is to describe how biophysical
methods(NMR and X-ray crystallography) have been used to
perform fragment screening against the N-terminal domain of
Hsp90. The paper shows how two fragment hits were then

5944 Journal of Medicinal Chemistry, 2010, Vol. 53, No. 16
Murray et al.
Compound 1 has a measured affinity of 250 μM and ligand
efficiency of 0.38. The binding mode for this fragment is
given in Figure 4a. The molecule lies in a deep pocket where it
forms an extensive network of hydrogen bonds with the side
chain of Asp93 and four crystallographically observed water
molecules trapped at the bottom of the pocket. The amino-
pyrimidine interactions are similar to those observed with
the adenine ring in the ADP complex (Figure 4b). Interest-
ingly, compound 1 is twisted about the bond connecting the
pyridine to the pyrimidine. As discussed in more detail
below, this is not the most stable geometry in the absence
of the protein, so it should be possible to improve the affinity
through stabilization of the protein bound conformation.
Furthermore, compound 1 is doing a poor job of filling
the proximal lipophilic pocket lined by the lipophilic side
chains of Met98, Leu107, Phe138, Val150, and Val186 (see
Figure 4c), and so improved hydrophobic fit in this region
may be advantageous.
Compound 2 is an example of an even simpler aminopyr-
imidine that was observed to bind crystallographically even
though it has very weak affinity (>1 mM). Figure 4d dis-
plays the Fo - Fc unrefined electron density map for this
complex, and despite its small size and weak affinity, the
aminopyridine can be unambiguously placed within the map
together with a number of density peaks associated with
tightly bound water molecules. Comparison of Figure 4d
with Figure 4b shows that the simpler aminopyrimidine
forms the same contacts as compound 1.
Compound 3 has an affinity of 790 μM and a ligand
efficiency (LE) of 0.26 and is a known drug (ethamivan)
with respiratory stimulant activity.45 Figure 5a shows the
binding mode in a similar frame of reference to compound 1
in Figure 4b. It can be seen that the phenolic OH of
compound 3 displaces one of the water molecules observed
Figure 3. Chemical structures of four validated hits identified by
fragment screening against Hsp90.
Figure 4. Binding mode of aminopyrimidine fragments in Hsp90. (a) Compound 1 forms multiple hydrogen bonds with water molecules and
the side chain of Asp93 at the bottom of the ATP binding site. The bond between the two aromatic rings is heavily twisted. (b) The
crystallographic overlays of compound 1 and ADP illustrate the conserved nature of the interactions. (c) Protein surface in gray and the ligand
surface in cyan for the complex of Hsp90 and compound 1. For clarity the only protein residues shown are the residues that line the proximal
lipophilic pocket (Met98, Leu107, Phe138, Val150, and Val186). The bulge on the middle-right-hand side of the protein surface represents the
proximal lipophilic pocket. It is poorly filled by the ligand surface and is not occupied by water molecules. (d) Fo - Fc electron density map
contoured at 4σ for compound 2. The view has been clipped to aid visualization.

Article
Journal of Medicinal Chemistry, 2010, Vol. 53, No. 16 5945
Figure 5. Binding mode of phenol fragment 3 in Hsp90. (a) Compound 3 displaces one of the water molecules (compare with Figure 4a and
Figure 4b) and forms hydrogen bonds with two of the conserved water molecules at the bottom of the ATP binding site. (b) Crystallographic
overlay of compound 3 with the natural product radicicol (PDB code 1BGQ). The additional hydroxyl group in radicicol makes a direct
hydrogen bond with Asp93 and with a conserved water molecule. (c) Protein surface in gray and the ligand surface in cyan for the complex of
Hsp90 and compound 3. For clarity, the only protein residues shown are the residues that line the proximal lipophilic pocket (Met98, Leu107,
Phe138, Val150, and Val186). The bulge on the lower-right-hand side of the protein surface represents the proximal lipophilic pocket. It is
poorly filled by the ligand surface and is not occupied by water molecules.
in the pyrimidine/purine structures and forms a hydrogen
bond with one of the remaining conserved water molecules.
The carbonyl of compound 3 forms a hydrogen bond with
the side chain of Thr184 and with a conserved water mole-
cule (mediating to Asp93), but there is no direct hydrogen
bond to Asp93. Despite its poor ligand efficiency, the crystal-
lographic binding mode offered the promise of a fast frag-
ment optimization campaign. Superimposition of the stru-
cture on the yeast Hsp90 complex with the natural product,
radicicol46 (Figure 5b) indicated that conversion of the
phenol to the corresponding resorcinol would allow the
formation of a direct hydrogen bond to Asp93 and an
additional water-mediated hydrogen bond. These interac-
tions are common to all known active site inhibitors of
Hsp90, and we reasoned that this would yield a large increase
in affinity and ligand efficiency. The protein-ligand com-
plex also suggested another potential route to fragment
optimization. The methoxy group of fragment 3 does not
properly fill the proximal lipophilic pocket defined by the
side chains of Met98, Leu107, Phe138, Val150, and Val186
(Figure 5c), and its replacement with other substituents
might also be beneficial.
rearrangement of residues Ile110-Gly114 which become
the central portion of a long helix. Figure 6a illustrates this
change in protein secondary structure by superimposing the
structures for fragments 2 and 4. The conformational
movement has previously been observed with larger com-
pounds in Hsp9047 and has also been reported with frag-
ments, although no X-ray structures of fragments have been
disclosed.48,49 Other conformations for this mobile region
were also observed during our Hsp90 project, but this one is
notable because it opens up a large lipophilic pocket that
can be exploited in drug design. Fragment 4 makes only
lipophilic interactions with the induced pocket formed by
the side chains of Leu107, Phe 138, Tyr139, and Trp162.
Figure 6b shows the result of an experiment where com-
pounds 2 and 4 were cosoaked into Hsp90 that resulted in
both compounds being observed to bind at the same time.
The fragment binding modes in the individual fragment
complexes are identical to the binding modes observed
in the cosoak. Superimposition of the cosoak structure of
2 þ 4 with other structures indicates that fragment 4
occupies a region that is filled by the side chain of Leu107
in the fragment 2 complex. A valid design approach would
be to attempt to link together fragments 2 and 4, but the rest
of this paper focuses on a fragment growth strategy starting
from either compound 1 or compound 3.
Compound 4 has an affinity >1 mM with an LE <0.34,
and it binds in a lipophilic pocket away from Asp93. The
formation of the pocket is driven by the substantial

5946 Journal of Medicinal Chemistry, 2010, Vol. 53, No. 16
Murray et al.
Figure 6. (a) Crystallographic complex of Hsp90 with 4 (in cyan) superimposed on the complex with 2 (in orange). The proteins are displayed
in schematic form. The upper left helix is continuous with fragment 4 but is interrupted in a more typical complex with fragment 2. This
substantial conformational movement is induced by the binding of fragment 4. (b) Binding mode of fragments 2 and 4 when cosoaked with
Hsp90. Fragment 4 binds in an induced, slot-shaped, hydrophobic pocket flanked by the lipophilic side chains of Leu107, Phe138, Tyr139,
and Tyr162.
Table 1. Substitution with Small Groups at R1 and R2 Enhances
Potency of the Aminopyrimidines
Table 2. Further SAR around the Upper Phenyl Ring of the Amino-
pyrimidines
compd
R2
R6
ITC (μM)
LE
cell IC50 (μM)
1a
5
250
2
0.38
0.56
0.55
0.54
0.6
H
H
6
OMe
OMe
Cl
H
0.35
0.068
0.036
0.083
7
OMe
OMe
H
7.9
6.4
18
8
9
Cl
0.64
a The chemical structure for 1 is shown in Figure 3
Optimization of Aminopyrimidines Starting from Fragment 1.
Tables 1 and 2 give data and chemical structures for the amino-
pyrimidines covered in this section. The tables give dissociation
constants obtained with isothermal titration calorimetry, and
the cell IC50 values reflect the inhibition of HCT116 cell
proliferation.
The crystal structure of Hsp90 with compound 1 shows
that the fragment is twisted about the bond connecting the
pyridine to the pyrimidine (torsion angle of 47.6°) despite the
presence of an aromatic nitrogen atom ortho to this bond.
Small molecule crystal structures50 and the torsion profile
reported in Figure 7a both suggest that the optimal geometry
for such ring systems is close to planarity, so it should be
possible to improve the affinity through stabilization of the
protein bound conformation.
Optimisation of fragment 1 began with virtual screening of
close analogues and resulted in the purchase of the simple
chloro analogue, compound 5, that showed an improvement
in affinity of approximately 100-fold. Initial synthesis
focused on analogues of compound 5 with the aim of stabi-
lizing the twist observed in the X-ray structure and filling the
proximal lipophilic pocket that is formed by lipophilic side
chains of Met98, Leu107, Phe138, Val 150, and Val186.
From the crystal structure of fragment 1, the elegant way
to achieve this was to substitute the upper phenyl ring at the
2 and/or 6 position with small groups (Figure 4c).
The first compound synthesized was the 2-methoxy ana-
logue 6, which showed a 5-fold improvement in affinity and a
crystal structure showed that the compound bound in the
desired conformation (coordinates have been deposited with
the PDB). However, there was a concern that this was still
not the most favored ligand conformation. The torsion
energy profile of 6 indicated that the methoxy group might
prefer to sit on the other side of ring to avoid a clash with the
pyrimidine nitrogen and that the conformation observed in
the crystal structure (torsion angle of 48.1°) is a local mini-
mum which is ∼3 kcal/mol higher in energy than the global
minimum (see Figure 7b). This reasoning was supported by
further SAR. For example, the 2,6-dimethoxy compound 7
offered a further 5-fold improvement over compound 6
despite the fact that the second methoxy group made no
direct contact with the enzyme. Additionally the 2-chloro
compound 9 was significantly more potent than compound
5. The torsion profile of 9 suggests that mutating the
methoxy group with a chloro reduces to only ∼1 kcal/mol the

Article
Journal of Medicinal Chemistry, 2010, Vol. 53, No. 16 5947
Figure 7. Energy profiles and population histograms of compounds 1, 6, 9, and 12. The continuous line on the plots gives the relative potential
energy calculated by B3LYP/6-311G* as a function of the torsion angle for each structure. The scale on the right-hand side of the plots gives the
relative potential energies in kcal/mol. The relative population of each torsion angle in the gas phase is estimated from the Boltzmann
distribution and is shown as a histogram with the percentage population shown on the left-hand scale in the plot. The torsional angle being
considered is highlighted in bold in the chemical structures.
difference in energy between local and the global minimum
(see Figure 7c). Finally, the 2-chloro-6-methoxy compound 8
only marginally improves upon this affinity and the torsional
distribution for 8 shows that all conformations with a torsion
angle between 60° and 110° are isoenergetic (see Figure 7d).
Generally the calculated torsion distributions provided use-
ful predictions of the activity changes observed in the initial
round of synthesis.
Compound 9 was the most ligand efficient molecule from
the initial synthesis iteration and was chosen as the basis for
the next iteration. Crystal structures on this series indicated
that further substitution of the upper phenyl ring with small
groups at the 4-position might facilitate additional lipophilic
interactions with the enzyme. Methoxy or chlorine substitu-
tion significantly increased affinity with compound 10 hav-
ing an affinity of 12 nM and very high ligand efficiency
(0.68).
Figure 8. Crystallographic complex of Hsp90 with 14 (in cyan)
superimposed on the complex with 1 (in orange). The original
binding mode of the fragment is maintained in the final complex.
Further work on this series was focused on trying to
improve the physical properties and the cell activity of
compound 10. The structural work suggested that substitu-
tion at the 5 position of the upper phenyl ring might be used
to allow the introduction of solubilizing groups with the aim
of improving cell activity. The 5-methoxy group 12 offered a
2-fold improvement in measured potency (although note
that the error in ITC determination is approximately 2-fold).
It also served as a growth point for the introduction of
solubilizing groups such as the morpholine group exempli-
fied in compound 14. This compound has low micromolar
activity in the cell assay and may be a potential lead molecule
versus Hsp90. The crystal structure of compound 14 is shown
in Figure 8 and is in good agreement with the design
hypotheses; the superimposition in Figure 8 also illustrates
that the final molecule forms the same interactions as the
initial starting fragment.
Optimization of Phenols Starting from Fragment 3. The
binding mode of fragment 3 indicated that the methoxy

5948 Journal of Medicinal Chemistry, 2010, Vol. 53, No. 16
Table 3. Optimization of the Phenol Series
Murray et al.
substituted compounds. Given that this 2-OH group in
radicicol forms a direct hydrogen bond to Asp93, we were
interested to see whether a phenol based on this substitution
pattern would have superior affinity. Compounds 19 and 20
were prepared, but these were about 10-fold less potent than
the corresponding 4-OH substituted compounds. This im-
plied that the 4-OH group must be a significant contributor
to the affinity of the series, presumably derived from the
favorable displacement of a mediating water molecule ob-
served in the ADP and pyrimidine complexes (compare
Figure 5a with Figure 4b).
The next synthetic iteration investigated a selection of
amides to replace the diethylamide in compound 18. The
X-ray structure indicated the importance of the carbonyl
group which forms a direct hydrogen bond with the side
chain of Thr184 and a strong water-mediated hydrogen bond
with Asp93 (see Figure 5a). The torsion between the carbonyl
and phenyl ring is highly twisted (61° in compound 18), and
the torsion is stabilized by the tertiary nature of the amide.
The design therefore focused on tertiary amides, and model-
ing indicated that they would be directed toward the side
chain of Lys58 which is well-defined in the electron density
(Lys58 is visible in the upper-left-hand side of Figure 5a and
Figure 5b). The first design idea was to form a hydrogen
bond with the amine on Lys58. Compounds 21, 22, and 23
are representative of a small number of compounds that test
this idea, and all these compounds offer an affinity improve-
ment of about 50-fold, with compound 21 being the best and
most ligand efficient. Figure 9a shows the crystal structure of
21 and confirms the formation of a good hydrogen bond
between the morpholine oxygen and Lys58. The second
design idea centered on trying to displace the side chain of
Lys58 with larger tertiary amides that might then form good
lipophilic interactions with the protein. The lysine side chain
itself forms no specific contacts with the enzyme and seemed
a plausible candidate for displacement despite its relatively
clear positioning in the electron density. Compounds 24, 25,
and 26 are representative molecules, and the isoindoline 24
delivered an affinity improvement of several-hundred-fold.
Figure 9b shows the crystal structure of 24 where it can be
seen that the side chain of Lys58 has indeed moved away and
forms a salt bridge with Glu62. The space formerly occupied
by Lys58 accommodates the phenyl ring of the isoindoline
which forms good lipophilic contact with Ala55, Lys58, and
Ile96. Interestingly and perhaps surprisingly, it was therefore
possible to increase potency significantly by either forming a
direct hydrogen bond with the side chain of Lys58 or
displacing the side chain with lipophilic groups. The best
amides from the previous synthesis iteration were then
combined with the isopropyl group. This yielded a further
2- to 4-fold improvement for compounds 27, 28, and 29 (the
error in the ITC measurement is about 2-fold, so this
probably represents a real improvement). The change to
the isopropyl group also reduced the lipophilicity by about
0.5 unit.
group could be replaced by substituents which filled the
proximal lipophilic pocket. Chloro (15), ethyl (16), isopropyl
(17), and tert-butyl (18) analogues were initially synthesized
(Table 3). The chloro group showed no improvement, while
the ethyl group showed a 20-fold improvement. The best
groups were tert-butyl and isopropyl, which were approxi-
mately 100-fold more potent. The superior filling of the
proximal lipophilic pocket was confirmed by crystallogra-
phy on compound 18 (data not shown).
The final change was to convert the best phenols into
resorcinols driven by the similarity of the binding mode of
our compounds to the natural product radicicol (Figure 5b).
The morpholine compound 30 is 40-fold more potent than
the corresponding 4-OH compound 27 and has an improved
cell activity of 140 nM. The isoindoline resorcinol 31 shows
subnanomolar affinity, excellent ligand efficiency, and good
cell activity (31 nM). For compounds 30 and 31, the addi-
tional hydroxyl group gave large improvements in affinity
Radicicol (Figure 1) is a resorcinol and contains an addi-
tional OH group at the 2-position compared with 4-OH

Article
Journal of Medicinal Chemistry, 2010, Vol. 53, No. 16 5949
Figure 9. Development of the lead compound from the phenol series. (a) Experimental binding mode of compound 21 illustrating the
formation of a new hydrogen bond with the side chain of Lys58. (b) Experimental binding mode of compound 24 in which Lys58 is displaced by
the isoindoline group. In its new position, the side chain of Lys58 forms a salt bridge with Glu62. (c) Experimental binding mode of lead
molecule 31 (in cyan) superimposed on the original fragment 3 (in orange).
proliferation in the HCT116 cells. Compound 31 was chosen
as the lead compound, and further biological characteriza-
tion of the series is described in the companion paper (DOI:
10.1021/jm100060b).
Discussion and Conclusion
We have applied the Pyramid screening technique to Hsp90
and exploited a close integration of ligand observed NMR
Figure 10. Effects of different concentrations of 31 on the levels of
screening and high throughput X-ray crystallography. We
found it necessary to use a combination of cocrystallography
and soaking into different crystal forms of Hsp90 in order
to maximize the number of high quality X-ray structures
obtained. From this process, 26 fragment structures were
obtained during the screening phase of the project, and four
of these fragments are disclosed in this paper. Perhaps most
surprising were fragments such as 4 which bind to an induced
lipophilic pocket and the successful cosoaking of this frag-
ment with 2-aminopyrimidine. Huth et al. have recently
described a similar observation and detailed an NOE-based
NMR structure determination for the tertiary arrangement of
two fragments, one of which was a simple aminopyrimidine.48
Choosing which fragments to work on is one of the parts of
fragment based discovery that requires judgment. Ligand
efficiency39,40 is one criterion that we have used extensively,
but it is not the only factor. Here fragment 3 had relatively
poor ligand efficiency (0.26) but its binding mode suggested a
CDK4 and Hsp70 in HCT116 cells as determined by Western blot.
Included as controls are the levels of GAPDH protein and the levels
when no compound is added to HCT116 cells (shown as CTL).
but also gave much better translation between the measured
affinities and the cell activity (for reasons that were not
established). Figure 9c gives the binding mode of compound
31 in Hsp90, and as anticipated, the additional hydroxyl
group forms good hydrogen bonds with Asp93 and a med-
iating water molecule. The superimposition with the original
fragment demonstrates that the binding mode was preserved
during the fragment optimization campaign.
Figure 10 shows Western immunoblotting data for
HCT116 cells treated with 31 and confirms that the com-
pound has the anticipated molecular signature of an Hsp90
inhibitor;3,10 the client protein CDK4 is depleted, while the
chaperone protein Hsp70 is induced. These effects occur at
concentrations similar to the IC50 for the inhibition of cell

5950 Journal of Medicinal Chemistry, 2010, Vol. 53, No. 16
Murray et al.
number of design ideas for improvement, i.e., by converting
the phenol into a resorcinol or by properly filling the proximal
lipophilic pocket. Experience on a variety of projects has
demonstrated that ligand efficiency can be improved where
there is a clear and correctable deficiency in the binding mode
of the fragment. However, if this cannot berapidly achieved, it
is usually better to terminate work on such a fragment.
Similarly, although fragment 1 had better ligand efficiency
(0.38), its crystal structure suggested a design strategy to
improve potency, i.e., stabilizing the active conformation
and filling the proximal lipophilic pocket. This emphasizes
that the binding mode of the fragment and the attractiveness
of the associated medicinal chemistry plan are important
determinants in judging which fragments to pursue.
Subsequent to this work Brough et al. have disclosed a
parallel virtual screening and fragment screening approach to
Hsp90.20 Aminopyrimidine based hits were identified by both
approaches, and structure based drug design was used to
hybridize hits from molecules with related binding modes.
This approach successfully led to a clinical candidate. In our
work, we have used virtual screening in a different manner to
explore SAR around very close analogues of identified frag-
ments. For example, the binding mode of fragment 3 was used
to inform a virtual fragment screen culminating in the identi-
fication of fragment 5. This more conservative approach may
allow more control over the molecular size and ligand effi-
ciency of the developing lead series. Virtual screening of
designed compounds was also heavily used to prioritize
synthesis during fragment optimization.
Relatively small libraries of compounds are required to
perform fragment screening, and if researchers rely on com-
mercially available compounds, it is likely that the same
fragments will appear in fragment screening collections from
different companies. This represents an emerging challenge
for FBDD practitioners to improve the novelty of fragments
in their libraries and to establish strategies for introducing
novelty as molecules are optimized. Hsp90 is a good example
on which to assess this because a large number of research
groups have worked on this target and some of these groups
have used FBDD. The work on aminopyrimidines by Brough
et al. illustrates how novelty can be introduced through the
fusion of a heterocycle onto an initial aminopyrimidine
fragment.20 Barker et al. identified aminopyrimidine frag-
ments containing a semisaturated ring fused onto the amino-
pyrimidine and presumably attained novelty as a result.49 We
first described our aminopyrimidine series in a patent several
years ago,51 and subsequently Huth et al. have described the
development of a similar series using FBDD.48 In the present
work, the fragment screening and optimization approach are
different, and the final compound shows activity in cells.
Resorcinol compounds have been worked on by a number
of groups.16 Both high throughput screening and fragment
screening approaches were used in the development of NYP-
AUY922 (Figure 2).21 The present work describes a pure
fragment-screening approach that originated from a frag-
ment-like drug molecule containing a phenolamide moiety;
the final compounds differ substantially in their architec-
ture and shape from resorcinols such as NVP-AUY922
which contain a five-membered aromatic heterocyle directly
attached to the resorcinol ring.22,52-56 Kung et al. have
described a resorcinol amide series that they identified inde-
pendently via high throughput screening.57,58 The present
work differs substantially in the screening and optimization
methodology and in the final clinical candidate, which is
Figure 11. Structure for lead molecule 31 together with fragment 3
from which it was derived. The lead molecule is over 1 000 000 times
more potent than the fragment, and this can be decomposed into
(i) 110-fold for replacement of the methoxy group by isopropyl as
derived from affinities for compound 3 and compound 17 (net
increase of one heavy atom), (ii) 100-fold for replacement of the
diethylamide with the isoindoline amide as derived from affinities
for compound 17 and compound 28 (net increase of four heavy
atoms), and (iii) 130-fold for the introduction of the 2-OH group as
derived from affinities for compound 28 and compound 31 (net
increase of one heavy atom).
describedinthe companionpaper(DOI:10.1021/jm100060b).
Both our series59,60 and the series of Kung and co-workers61
have previously been described in the patent literature.
The literature now contains many Hsp90 medicinal chem-
istry papers from a large number of groups. Approaches
adopted include HTS, fragment screening, virtual screening,
andnatural product optimizationoroptimization based on an
existing lead.16-18 A number of clinical compounds have also
been revealed, so it is apparent that all approaches have been
successful to some degree and that Hsp90 is a reasonably
tractabletarget from a medicinal chemistrypoint of view. This
paper shows that careful structure-based optimization based
on initially very weak fragment hits can deliver highly ligand
efficient and low molecular weight leads for Hsp90 without
the need for large amounts of synthetic chemistry. In the
case of the pyrimidine series, optimization from fragment 1
(250 μM) to compound 12 (6.3nM) improved affinity by
40000-fold while the heavy atom count increased by just five
atoms. This increase in affinity was mainly gained through the
stabilization of the active conformation of the starting frag-
ment and the improvement of hydrophobic contact with the
enzyme. Figure 11 and its caption summarize the structure-
guided optimization process that led to the resorcinol lead
compound 31 (0.54 nM) starting from the phenol fragment 3
(790 μM). Affinity was improved by over a million-fold with
the addition of only six heavy atoms, and this can be ascribed
to three changes, each contributing about 100-fold to affinity.
The first modification (methoxy to isopropyl) gave superior
filling of the proximal lipophilic pocket with a net increase of
one heavy atom; the second modification (diethylamide to
isoindolamide) was the least “efficient” with a net increase of
four heavy atoms and was driven by improved lipophilic
contact with the enzyme as a result of the conformational
movement of Lys58. The final modification (addition of a
hydroxyl group) added one additional heavy atom with the
formation of a hydrogen bond with Asp97 and a further
water-mediated hydrogen bond. In terms of the efficiency of
the added groups,62 the two fragment to lead campaigns
described in this paper are among the most efficient ever
reported. In both cases the final lead has a molecular weight of
around 300 Da which leaves a significant amount of “mole-
cular weight head room” to tune the other properties of the
molecule during lead optimization. This means that groups

Article
Journal of Medicinal Chemistry, 2010, Vol. 53, No. 16 5951
can be added to the molecule to improve nonpotency related
properties without the fear that the resulting clinical candidate
lies outside druglike space.
1-benzyloxy-2,4-dichloro-5-iodobenzene as a brown oil. The
product was taken on without further purification. To 1-ben-
zyloxy-2,4-dichloro-5-iodobenzene (4.5 g, 11.9 mmol, 1.0
equiv) in tetrahydrofuran (40 mL) at -78 °C was added the
triisopropyl borate (3.4 g, 17.8 mmol, 1.5 equiv) and then the
n-BuLi (11.1 mL, 1.6 M, 1.5 equiv) slowly. The mixture was
allowed to stir for an hour, and the reaction was then quenched
with NH4Cl (sat., aq.). The mixture was allowed to warm to
room temperature and was diluted with water and ethyl
acetate. The product was extracted with ethyl acetate (ꢀ3).
The combined organic layers were washed with brine and dried
over MgSO4. The product was filtered and concentrated to
dryness under reduced pressure to leave a pale-yellow oil. The
crude reaction mixture was coupled to 2-amino-4,6-dichloro-
pyrimidine following the same procedure as for pyrimidine 7,
except the reaction was heated thermally at 50 °C for 3 h. The
title compound was isolated by preparative HPLC to yield the
product as a colorless solid. 1H NMR (400 MHz, Me-d3-OD)
δ 7.60 (s, 1H), 7.49 (d, 2H), 7.45-7.32 (m, 4H), 6.93 (s, 1H),
5.22 (s, 2H). LCMS (ESI): m/z 380 (M þ Hþ).
In conclusion we have applied a fragment screening ap-
proach to Hsp90 and developed two series of compounds
starting from fragments. In each series, structure-based drug
design led to the discovery of high potency compounds with
excellent ligand efficiency. A resorcinol lead with subnano-
molar affinity and good cell potency has been identified, and
the companion paper (DOI: 10.1021/jm100060b) describes
how this lead was converted into a compound that is currently
in clinical trials for the treatment of cancer.
Experimental Section
Chemistry. The method employed in determining the purity
of compounds was LCMS, and all compounds had purities of
>95%. Reagents and solvents were obtained from commercial
suppliers and used without further purification. Thin layer
chromatography (TLC) analytical separations were conducted
with E. Merck silica gel F-254 plates of 0.25 mm thickness and
were visualized with UV light (254 nM) and/or stained with
iodine, potassium permanganate, or phosphomolybdic acid
solutions followed by heating. Flash chromatography was
performed either manually using E. Merck silica gel or using
an automated Biotage SP4 system with prepacked disposable
SiO2 cartridges (4, 8, 19, 38, 40, and 90 g sizes) with stepped or
gradient elution at 5-40 mL/min using the indicated solvent
To 4-(5-benzyloxy-2,4-dichlorophenyl)-6-chloropyrimidin-
2-ylamine (5.0 g, 9.2 mmol, 1.0 equiv) in dichloromethane
(31 mL) was added BCl3 (10.1 mL, 10.1 mmol, 1.1 equiv) at
-78 °C under a nitrogen atmosphere. The mixture was stirred
for 2 h. The reaction was then quenched with methanol (10 mL),
and the mixture was allowed to warm to room temperature. The
mixture was then diluted with water, and the product was
extracted with ethyl acetate (ꢀ3). The combined organic layers
were washed with water, brine and dried over MgSO4. The
product was filtered and concentrated to dryness under reduced
pressure to leave a pale-yellow oil. The product was purified by
column chromatography (Rf = 0.48, 100% diethyl ether) to
yield 5-(2-amino-6-chloropyrimidin-4-yl)-2,4-dichlorophenol
as a colorless solid. 1H NMR (400 MHz, Me-d3-OD) δ 7.48 (s,
1H), 7.11 (s, 1H), 6.90 (s, 1H). LCMS (ESI): m/z 290 (M þ Hþ).
To 5-(2-amino-6-chloropyrimidin-4-yl)-2,4-dichlorophenol
(81 mg, 0.28 mmol, 1.0 equiv) in DMF (3 mL) was added
Cs2CO3 (199 mg, 0.61 mmol, 2.1 equiv) and then N-(2-chloro-
1
mixture. H nuclear magnetic resonance (NMR) spectra were
recorded in the deuterated solvents specified on a Bruker
Avance 400 spectrometer operating at 400 MHz. Chemical
shifts are reported in parts per million (δ) from the tetramethyl-
silane resonance in the indicated solvent (TMS: 0.0 ppm). Data
are reported as follows: chemical shift, multiplicity (br = broad,
s = singlet, d = doublet, t = triplet, m = multiplet), integration.
Compound purity and mass spectra were determined by a Waters
2795/Platform LC LC/MS system using the positive electrospray
ionization technique (þESI), a Waters Fractionlynx/Micromass
ZQ LC/MS system using the positive electrospray ionization
technique (þESI), or an Agilent 1200SL-6140 LC/MS system using
positive-negative switching, using a mobile phase of acetonitrile/
water with 0.1% formic acid.
4-(3-Pyridyl)pyrimidin-2-ylamine 1, 2-aminopyrimidine 2, 4,
N,N-diethyl-3-methoxy-4-hydroxybenzamide 3, 2-chloro-4-methyl-
phthalazine 4, 4-chloro-(6-phenyl)pyrimidin-2-ylamine 5, 3-chloro-
4-hydroxybenzoic acid, 3-tert-butyl-4-hydroxybenzoic acid, and
5-tert-butyl-2-methoxybenzoic acid are commercially available.
5-Isopropyl-2-methoxybenzoic acid,63 3-ethyl-4-hydroxybenzoic
acid,64 and 2,4-bis-benzyloxy-5-isopropenylbenzoic acid59 were
prepared as described previously. 4-Hydroxy-3-isopropylben-
zoic acid was prepared from 2-isopropylphenol according to the
method described64 for the synthesis of 3-ethyl-4-hydroxyben-
zoic acid. Experimental and spectroscopic details for all other
noncommercially available compounds are given below or are
available in the Supporting Information.
4-Chloro-6-[2,4-dichloro-5-(2-morpholin-4-ylethoxy)phenyl]-
pyrimidin-2-ylamine (14). Concentrated HCl (19 mL) was
added to 5-(benzyloxy)-2,4-dichloroaniline (52 g, 19.3 mmol,
1.0 equiv) in glacial acetic acid (77 mL), and the mixture was
cooled in an ice-water bath. Sodium nitrite (1.5 g, 22.2 mmol,
1.2 equiv) was added in water (64 mL) slowly, maintaining the
temperature at <5 °C. This was stirred for 30 min. This
mixture was poured into KI (6.4 g, 38.6 mmol, 2.0 equiv) and
I2 (1.7 g, 5.4 mmol, 0.3 equiv) in water (241 mL) and was left to
stir for 1.5 h at room temperature. The reaction was quenched
by diluting with water and extracting the product with dichloro-
methane (ꢀ3). The combined organic layers were washed with
brine and dried over MgSO4. The product was filtered and
concentrated to dryness under reduced pressure to leave
ethyl)morpholine HCl (57 mg, 0.31 mmol, 1.1 equiv) under a
3
nitrogen atmosphere. The mixture was heated thermally at 80 °C
for 3 h. The mixture was allowed to cool and concentrated to
dryness under reduced pressure. The title compound 14 was
crystallized from methanol to yield the product as beige crystals.
1H NMR (400 MHz, Me-d3-OD) δ 7.58 (s, 1H), 7.33 (s, 1H), 6.96
(s, 1H), 4.27 (t, 2H), 3.72 (dd, 4H), 2.88 (t, 2H), 2.66 (dd, 4H).
LCMS (ESI): m/z 403 (M þ Hþ). LCMS purity: 96.7%.
4-(2,4-Dihydroxy-5-isopropylbenzoyl)morpholine (30). Morpho-
line (66 mg, 0.75 mmol) was added to a stirred mixture of 2,4-bis-
benzyloxy-5-isopropenylbenzoic acid (187 mg, 0.5 mmol), 1-(3-di-
methylaminopropyl)-3-ethylcarbodimide hydrochloride (116 mg,
1.2 mmol), and 1-hydroxybenzotriazole (81 mg, 0.6 mmol) in
dichloromethane (10 mL), and the mixture was stirred at room
temperature for 4 h. The mixture was washed successively with 2 M
hydrochloric acid, saturated sodium bicarbonate, and water. The
organic layer was separated and concentrated under reduced
pressure to afford 4-(2,4-bis-benzyloxy-5-isopropenylbenzoyl)-
morpholine as a colorless solid. Yield: 76%. 1H NMR (DMSO-
d6) δ 7.57-7.30 (m, 10H), 7.02 (s, 1H), 6.95 (s, 1H), 5.18 (d, 4H),
5.06 (d, 2H), 3.57 (br s, 4H), 3.41 (br s, 2H), 3.16 (br s, 2H), 2.03 (s,
3H). LCMS (þESI): m/z 444 (M þ Hþ). LCMS purity: 95.8%.
4-(2,4-Bis-benzyloxy-5-isopropenylbenzoyl)morpholine (133
mg, 0.3 mmol) was dissolved in methanol (8 mL), 10% palla-
dium on carbon (50 mg) was added, and the mixture was stirred
overnight under a hydrogen atmosphere at room temperature.
The mixture was filtered, the solids were rinsed with methanol
(4 mL), and the combined fitrates were reduced to dryness under
reduced pressure to afford 30 as a colorless solid. Yield: 90%. 1H
NMR (DMSO-d6): δ 9.57 (br s, 1H), 6.85 (s, 1H), 6.36 (s, 1H),
3.56 (br m, 4H), 3.40 (br s, 4H), 3.06 (m, 1H), 1.10 (d, 6H).
LCMS (þESI): m/z 266 (M þ Hþ). LCMS purity: 97.1%.

5952 Journal of Medicinal Chemistry, 2010, Vol. 53, No. 16
Murray et al.
2-(2,4-Dihydroxy-5-isopropylbenzoyl)-2,3-dihydro-1H-isoindole
(31). Compound 31 was prepared in a similar manner to 30 from
2,3-dihydro-1H-isoindole and 2,4-bis-benzyloxy-5-isopropenyl-
benzoic acid. 1H NMR (DMSO-d6): δ 10.04 (s, 1H), 9.64 (s,
1H), 7.31 (br m, 4H, s), 7.04 (s, 1H), 6.41 (s, 1H), 4.77 (s, 4H), 3.10
(m, 1H), 1.14 (d, 6H, d). LCMS (þESI): m/z 298 (M þ Hþ).
LCMS purity: 100.0%.
Crystallography. Two hexahistidine tagged N-terminal frag-
ments (9-236 and 9-224) of Hsp90R were cloned into a pET28
vector and expressed in BL21 (DE3). The proteins were cap-
tured using a Ni2þ affinity column followed by removal of the
tag using thrombin and final purification on a Superdex 75 gel
filtration column with a final buffer of 20 mM Tris-HCl (pH
7.4), 150 mM NaCl, and 1 mM β-ME. The proteins were
concentrated to between 20 and 25 mg mL-1 prior to crystal-
lization. For soaking experiments, two crystal forms were
exploited because of the differing loop structure around the
active site. The I222 apo crystals were generated using the
hanging drop, vapor diffusion method by combining equal
NMR Screening. NMR experiments were carried out at
500 MHz, using a Bruker DRX500 instrument equipped with a
TXIcryoprobe. The N-terminal domain of Hsp90 was expressed
and purified as described below, giving a product of molecular
mass 24 508 Da, whose mass was confirmed by time-of-flight
mass spectrometry. Samples for NMR screening were prepared
in 20 mM Tris buffer, pH7.4, containing 200 μM ADP, 150 mM
NaCl, 1 mM BME, and 10% D2O. Typical screening samples
contained 0.3 mg/mL (12.5 μM) Hsp90 and four fragments, each
at 500 μM (2% DMSO). This gave a molar ratio of fragment to
target of 40:1, a typical value for water-LOGSY experiments.44
At these protein and ligand concentrations, fragments that bind
with dissociation constants better than 1 mM are expected to be
detected. LOGSY experiments were performed using the
e-PHOGSY sequence of Dalvit et al. incorporating a 1 s delay
to allow cross-relaxation between fragments and water, and a
100 ms CPMG period during which time any protein magneti-
zation decays.65
In order to enhance the LOGSY signal, all NMR experiments
were performed at 5 °C. Lowering the temperature improves the
affinities of ligands that bind exothermically and reduces the
rates of exchange between bulk water and water of hydration,
leading to improved LOGSY intensities. In order to detect the
small positive LOGSY effects caused by binding to the protein
target, LOGSY spectra were obtained in the presence and the
absence of the protein target, and these spectra were directly
compared.
The LOGSY spectra were defined as medium or strong
signals as follows. LOGSY spectra were obtained from two
separate samples (þprotein and buffer control), and these were
subtracted (þprotein - buffer control) to give the LOGSY
difference spectrum. This spectrum shows the change in the
amount of magnetization transferred from water to the ligands,
due to protein binding. The intensity of each peak in this
spectrum depends on a number of experimental choices
(number of scans, magnetization transfer time, relaxation de-
lays, etc.), and there is no reference spectrum generated with
which to quantitate the LOGSY effects. However a standard 1D
spectrum was also run (with fewer scans and different relaxation
delays) and the LOGSY intensity was compared with this. For
each compound, the largest % LOGSY signal observed for any
detectable aromatic proton was recorded. The % LOGSY va-
lues for ligands against HSP90 (LOGSY intensity versus 1D
intensity) ranged from 0 to about 60% with strong LOGSY
signals being defined as >20% and medium signals defined as
11-20%.
volumes of protein and mother liquor (0.1 M Hepes NaOH,
3
pH 7.2, 0.2 M magnesium chloride, 20% (w/v) MPEG 2000).
Structures for compounds 2, 4, 2 þ 4, and 14 were obtained in
this system. The P21 back-soakable crystals were also generated
using the hanging drop, vapor diffusion method but 2 mM
ADP Mg2þ was added to the protein before combining with
3
mother liquor (0.1 M Bis-Tris, pH 6.0, 0.2 M ammonium
phosphate, 25% (w/v) PEG 3350). Structures for compounds
6, 18, and 21 were obtained in this system. All crystallization
experiments were performed at 4 °C and usable crystals
appeared between 1 and 3 days. Soaking was performed in
respective mother liquors containing 20% glycerol and 50 mM
compound with a final DMSO concentration of 10% (v/v) for
periods of up to 24 h. Crystals were subsequently flash frozen
and data collected on a Rigaku RU-H3R rotating anode gen-
erator with Osmic confocal blue optics and either a Jupiter CCD
or Raxis 4þþ image plate. For cocrystallization, the protein was
mixed with 5 mM compound with a final DMSO concentration
of 2.5% (v/v), followed by automated crystallization setup using
a Cartesian robot, typically at 0.2 μL final drop volumes in
sitting drop 96-well formats using a limited in house designed
matrix of crystallization conditions. Crystals grown using this
system were extracted from the low volume drops, cryopro-
tected in mother liquor with 20% glycerol, and flash frozen for
in house data collection. Compound 1 crystals were grown in
0.1 M Bis-Tris, pH 5.5, 0.2 M ammonium acetate, 25% PEG
3350. Compound 3 crystals were grown in 0.1 M Bis-Tris,
pH 5.75, 0.2 M ammonium acetate, 20% PEG 3350. Compound
24 crystals were grown in 0.1 M Bis-Tris, pH 5.25, 0.2 M
ammonium acetate, 27% PEG 3350. Compound 31 crystals
were grown in 0.1 M Tris-HCl, pH 8.0, 0.2 M magnesium
chloride, 22% PEG 4000.
Diffraction data were processed using MOSFLM66 and CCP4.67
The structures were solved using isomorphous replacement with the
coordinates previously described, 1YER68 for the I222 form and
1BYQ69 for the P21 back-soakable and cocrystallization form. The
structures were refined using REFMAC570 and Buster-TNT.71
Ligands were placed using AutoSolve72 and manual rebuilds made
in AstexViewer273 and COOT.74 Final validation checks were
performed using Procheck,75 prior to deposition of coordinates
and structure factors into the PDB.
The presence of 200 μM ADP in the buffer enabled fragment
hits, which had been identified by changes in their LOGSY
signals, to be further characterized according to their competition
at the nucleotide site, determined by a reduction in the LOGSY
signal of ADP. In the absence of Mg2þ, 200 μM corresponds to a
concentration ofADP aroundits dissociationconstant sothatthe
presence of ADP does not greatly reduce the sensitivity of the
experiment to fragment binding while providing a convenient
method of identifying cocktails that contain an active-site ligand.
LOGSY spectra were reacquired after the addition of a small
aliquot of a concentrated MgCl2 solution to the screening solu-
tions, giving a final concentration of 5 mM. This converts ADP to
its Mg2þ complex which binds more tightly to Hsp90, with a
dissociation constant measured by ITC of approximately 10 μM.
This in turn leads to complete displacement of fragments that
bind exclusively to the active site of Hsp90, enabling selective
active-site binders to be readily identified.
Isothermal Titration Calorimetry (ITC). ITC experiments
were performed on a MicroCal VP-ITC at 25 °C in a buffer
comprising 25 mM Tris, 100 mM NaCl, 1 mM MgCl2, and
1 mM TCEP at pH 7.4. The final DMSO concentration was
between 1% and 5%. The protein used was the same Hsp90
N-terminal ATPase domain construct used in both the NMR
and X-ray crystallography work. Most of ITC experiments were
set up with protein in the sample cell and compound in the
injection syringe, although in cases where compound solubility
was limiting, this was reversed. Data were fit to a single site
binding model using Origin 7.0 software. All the stoichiometry
values from the data analysis were in the range 0.8-1.3,
providing an excellent internal control for the quality, purity,
and stability of both the protein and the compounds. Replicate
experiments indicate that errors in the dissociation constants are
generally better than 2-fold. The stoichiometry parameter was
fixed at 1 in cases where the Kd value was greater than the protein

Article
Journal of Medicinal Chemistry, 2010, Vol. 53, No. 16 5953
concentration.76 By use of the procedure outlined above, ADP
and 17-DMAG had measured dissociation constants of 9.2 and
0.21 μM. These values are in good agreement with literature
dissociation constants with the full length human Hsp90 protein
of 11 μM for ADP and 0.35 μM for 17-DMAG.77 These results
indicate that ITC on N-terminal Hsp90 is a good surrogate for
ITC on full length Hsp90, provided the compounds bind in the
ATPase active site. A competition format ITC78 was necessary
to accurately determine the affinity of compound 31. This
required preincubating the protein with one of our moderately
potent phenol compounds in the sample cell prior to initiating
the titration with compound 31. A competition binding model
was used to fit the data and obtain a Kd estimate for compound
31. More experimental detail on ITC is given in Supporting
Information together with representative data for three com-
pounds of different affinity (i.e., compounds 1, 5, and 31).
Calculation of the Torsion Energy Profile. All energies were
calculated using the ab initio quantum chemistry package
Q-Chem.79 To calculate the torsion energy profile, a torsion
scan was carried out on a Corina generated structure by varying
the desired torsion angle between 0° and 180°, in 15° intervals.
For each structure at a given torsion angle, a constrained energy
minimization (basis set, 3-21G*; method, Hartree-Fock) was
carried out with a single constraint on the torsion angle to
restrain the molecule to the specific torsion angle value while the
rest of the molecule was geometrically optimized. The energy of
the optimized molecule at the specified conformational torsion
was then calculated using B3LYP method with 6-311G* basis
set. Finally, the 12 energies were utilized to estimate approxi-
mately the relative population of each conformation using a
Boltzman distribution.
(LICOR). Antibodies were sourced as follows: CDK4 from
Santa Cruz Antibodies (Santa Cruz CA), Hsp70 from StressGen
Biotechnologies (Victoria, British Columbia, Canada), and
GAPDH from Chemicon International (Temecula, CA). Blots
were scanned to detect infrared fluorescence on the Odyssey
infrared imaging system (LICOR).
Acknowledgment. The authors thank Anne Cleasby, David
Rees, Brian Williams, Lyn Leaper, Harren Jhoti, Robin Carr,
Neil Thompson, and John Lyons for comments on the
manuscript and for providing support to the project.
Supporting Information Available: Additional synthetic
chemistry experimental information, additional ITC experi-
mental information, and representative examples of ITC data.
This material is available free of charge via the Internet at http://
pubs.acs.org.
References
(1) Bukau, B.; Weissman, J.; Horwich, A. Molecular chaperones and
protein quality control. Cell 2006, 125, 443–451.
(2) Workman, P.; Burrows, F.; Neckers, L.; Rosen, N. Drugging the
cancer chaperone HSP90: combinatorial therapeutic exploitation
of oncogene addiction and tumor stress. Ann. N.Y. Acad. Sci. 2007,
1113, 202–216.
(3) Whitesell, L.; Lindquist, S. L. HSP90 and the chaperoning of
cancer. Nat. Rev. Cancer 2005, 5, 761–772.
(4) Hanahan, D.; Weinberg, R. A. The hallmarks of cancer. Cell 2000,
100, 57–70.
(5) Workman, P. Combinatorial attack on multistep oncogenesis by
inhibiting the Hsp90 molecular chaperone. Cancer Lett. 2004, 206,
149–157.
(6) Chiosis, G.; Neckers, L. Tumor selectivity of Hsp90 inhibitors:
the explanation remains elusive. ACS Chem. Biol. 2006, 1, 279–
284.
(7) Kamal, A.; Thao, L.; Sensintaffar, J.; Zhang, L.; Boehm, M. F.;
Fritz, L. C.; Burrows, F. J. A high-affinity conformation of Hsp90
confers tumour selectivity on Hsp90 inhibitors. Nature 2003, 425,
407–410.
Virtual Screening. The X-ray crystal structure of the N-term-
inal domain of Hsp90 complexed with compound 1 was used to
virtual-screen aminopyrimidine analogues of 1. Water mole-
cules were removed from the structure with the exception of the
four conserved waters at the bottom of the ATP binding pocket.
Hydrogens were added to the protein and waters, and a binding
pocket was generated with all protein and water atoms within
(8) Solit, D. B.; Chiosis, G. Development and application of Hsp90
˚
6 A of any non-hydrogen atom in ligand 1. This binding site was
inhibitors. Drug Discovery Today 2008, 13, 38–43.
used to run all dockings and virtual screens in the presence of the
key water molecules, allowing them to toggle on/off and spin
around their three principal axes with methods and settings
previously described by Verdonk et al.80 All calculations were
run on a Linux cluster using the Astex Web-based virtual screen-
ing platform.81 A number of filters (heavy atom count, mole-
cular weight, number of donors, number of acceptors, ClogP,
and polar surface area) combined with the scoring functions
Goldscore82 and Chemscore83,84 were applied to score and to
rank the docked libraries. Of the two scoring functions, Gold-
score performed generally better in terms of reproducing the
correct binding modes observed in Hsp90 X-ray crystal struc-
tures. Finally, virtual screens were visually analyzed, and mole-
cules were selected on the basis of the binding mode, the
chemical tractability, and the calculated score.
Hsp90 Cell Assay and Western Blot. The antiproliferative
activities of the compounds were determined by measuring their
ability to inhibit the growth of HCT116 cells (obtained from
European Collection of Cell Cultures). Cell growth was mea-
sured using the Alamar blue assay as described in Squires et al.85
For Western blotting experiments, HCT116 cells were treated
with compound in six-well tissue culture plates at the stated
concentrations for 18 h before washing with PBS and lysing cells
by the addition of ice-cold lysis buffer (40 mM Tris-HCl (pH
7.5), 274 mM NaCl, 20% glycerol, 2% Triton-X-100, 1 mM
Na3VO4, 50 mM NaF, 1% protease inhibitor cocktail set III
(Calbiochem)). The cells were lysed on ice for 30 min and
clarified by centrifugation. Lysates in SDS sample buffer were
heated for 5 min at 95 °C, resolved by SDS-PAGE, and
immunoblotted with the indicated antibodies followed by
either infrared dye labeled anti-rabbit or anti-mouse antibodies
(9) Pearl, L. H.; Prodromou, C. Structure and mechanismof the Hsp90
molecular chaperone machinery. Annu. Rev. Biochem. 2006, 75,
271–294.
(10) Vilenchik, M.; Solit, D.; Basso, A.; Huezo, H.; Lucas, B.; He, H.;
Rosen, N.; Spampinato, C.; Modrich, P.; Chiosis, G. Targeting
wide-range oncogenic transformation via PU24FCl, a specific
inhibitor of tumor Hsp90. Chem. Biol. 2004, 11, 787–797.
(11) Soga, S.; Shiotsu, Y.; Akinaga, S.; Sharma, S. V. Development of
radicicol analogues. Curr. Cancer Drug Targets 2003, 3, 359–369.
(12) Neckers, L.; Schulte, T. W.; Mimnaugh, E. Geldanamycin as a
potential anti-cancer agent: its molecular target and biochemical
activity. Invest. New Drugs 1999, 17, 361–373.
(13) Ge, J.; Normant, E.; Porter, J. R.; Ali, J. A.; Dembski, M. S.; Gao,
Y.; Georges, A. T.; Grenier, L.; Pak, R. H.; Patterson, J.; Sydor,
J. R.; Tibbitts, T. T.; Tong, J. K.; Adams, J.; Palombella, V. J.
Design, synthesis, and biological evaluation of hydroquinone
derivatives of 17-amino-17-demethoxygeldanamycin as potent,
water-soluble inhibitors of Hsp90. J. Med. Chem. 2006, 49, 4606–
4615.
(14) Kaur, G.; Belotti, D.; Burger, A. M.; Fisher-Nielson, K.; Borsotti,
P.; Riccardi, E.; Thillainathan, J.; Hollingshead, M.; Sausville,
E. A.; Giavazzi, R. Antiangiogenic properties of 17-(dimethyl-
aminoethylamino)-17-demethoxygeldanamycin: an orally bioa-
vailable heat shock protein 90 modulator. Clin. Cancer Res. 2004,
10, 4813–4821.
(15) Schulte, T. W.; Neckers, L. M. The benzoquinone ansamycin 17-
allylamino-17-demethoxygeldanamycin binds to HSP90 and
shares important biologic activities with geldanamycin. Cancer
Chemother. Pharmacol. 1998, 42, 273–279.
(16) Drysdale, M. J.; Brough, P. A. Medicinal chemistry of Hsp90
inhibitors. Curr. Top. Med. Chem 2008, 8, 859–868.
(17) Janin, Y. L. Heat shock protein 90 inhibitors. A text book example
of medicinal chemistry? J. Med. Chem. 2005, 48, 7503–7512.
(18) Blagg, B. S.; Kerr, T. D. Hsp90 inhibitors: small molecules that
transform the Hsp90 protein folding machinery into a catalyst for
protein degradation. Med. Res. Rev. 2006, 26, 310–338.

5954 Journal of Medicinal Chemistry, 2010, Vol. 53, No. 16
Murray et al.
(19) Sgobba, M.; Rastelli, G. Structure-based and in silico design of
Hsp90 inhibitors. ChemMedChem 2009, 4, 1399–1409.
(37) Vieth, M.; Siegel, M. G.; Higgs, R. E.; Watson, I. A.; Robertson,
D. H.; Savin, K. A.; Durst, G. L.; Hipskind, P. A. Characteristic
physical properties and structural fragments of marketed oral
drugs. J. Med. Chem. 2004, 47, 224–232.
(38) Wenlock, M. C.; Austin, R. P.; Barton, P.; Davis, A. M.; Leeson,
P. D. A comparison of physiochemical property profiles of devel-
opment and marketed oral drugs. J. Med. Chem. 2003, 46, 1250–
1256.
(39) Hopkins, A. L.; Groom, C. R.; Alex, A. Ligand efficiency: a useful
metric for lead selection. Drug Discovery Today 2004, 9, 430–431.
(40) Kuntz, I. D.; Chen, K.; Sharp, K. A.; Kollman, P. A. The maximal
affinity of ligands. Proc. Natl. Acad. Sci. U.S.A. 1999, 96, 9997–
10002.
(20) Brough, P. A.; Barril, X.; Borgognoni, J.; Chene, P.; Davies,
N. G.; Davis, B.; Drysdale, M. J.; Dymock, B.; Eccles, S. A.;
Garcia-Echeverria, C.; Fromont, C.; Hayes, A.; Hubbard, R. E.;
Jordan, A. M.; Jensen, M. R.; Massey, A.; Merrett, A.; Padfield,
A.; Parsons, R.; Radimerski, T.; Raynaud, F. I.; Robertson,
A.; Roughley, S. D.; Schoepfer, J.; Simmonite, H.; Sharp,
S. Y.; Surgenor, A.; Valenti, M.; Walls, S.; Webb, P.; Wood, M.;
Workman, P.; Wright, L. Combining hit identification strategies:
fragment-based and in silico approaches to orally active 2-amino-
thieno[2,3-d]pyrimidine inhibitors of the Hsp90 molecular chaper-
one. J. Med. Chem. 2009, 52, 4794–4809.
(21) Eccles, S. A.; Massey, A.; Raynaud, F. I.; Sharp, S. Y.; Box, G.;
Valenti, M.; Patterson, L.; de Haven, B. A.; Gowan, S.; Boxall, F.;
Aherne, W.; Rowlands, M.; Hayes, A.; Martins, V.; Urban, F.;
Boxall, K.; Prodromou, C.; Pearl, L.; James, K.; Matthews, T. P.;
Cheung, K. M.; Kalusa, A.; Jones, K.; McDonald, E.; Barril, X.;
Brough, P. A.; Cansfield, J. E.; Dymock, B.; Drysdale, M. J.;
Finch, H.; Howes, R.; Hubbard, R. E.; Surgenor, A.; Webb, P.;
Wood, M.; Wright, L.; Workman, P. NVP-AUY922: a novel
heat shock protein 90 inhibitor active against xenograft tumor
growth, angiogenesis, and metastasis. Cancer Res. 2008, 68, 2850–
2860.
(41) Leach, A. R.; Hann, M. M.; Burrows, J. N.; Griffen, E. J. Fragment
screening: an introduction. Mol. BioSyst. 2006, 2, 430–446.
(42) Hann, M. M.; Leach, A. R.; Harper, G. Molecular complexity and
its impact on the probability of finding leads for drug discovery.
J. Chem. Inf. Comput. Sci. 2001, 41, 856–864.
(43) Hartshorn, M. J.; Murray, C. W.; Cleasby, A.; Frederickson, M.;
Tickle, I. J.; Jhoti, H. Fragment-based lead discovery using X-ray
crystallography. J. Med. Chem. 2005, 48, 403–413.
(44) Dalvit, C.; Fogliatto, G.; Stewart, A.; Veronesi, M.; Stockman, B.
WaterLOGSY as a method for primary NMR screening: practical
aspects and range of applicability. J Biomol. NMR 2001, 21, 349–
359.
(22) Brough, P. A.; Aherne, W.; Barril, X.; Borgognoni, J.; Boxall, K.;
Cansfield, J. E.; Cheung, K. M.; Collins, I.; Davies, N. G.; Drys-
dale, M. J.; Dymock, B.; Eccles, S. A.; Finch, H.; Fink, A.; Hayes,
A.; Howes, R.; Hubbard, R. E.; James, K.; Jordan, A. M.; Lockie,
A.; Martins, V.; Massey, A.; Matthews, T. P.; McDonald, E.;
Northfield, C. J.; Pearl, L. H.; Prodromou, C.; Ray, S.; Raynaud,
F. I.; Roughley, S. D.; Sharp, S. Y.; Surgenor, A.; Walmsley, D. L.;
Webb, P.; Wood, M.; Workman, P.; Wright, L. 4,5-Diarylisox-
azole Hsp90 chaperone inhibitors: potential therapeutic agents for
the treatment of cancer. J. Med. Chem. 2008, 51, 196–218.
(23) Bao, R.; Lai, C. J.; Qu, H.; Wang, D.; Yin, L.; Zifcak, B.; Atoyan,
R.; Wang, J.; Samson, M.; Forrester, J.; DellaRocca, S.; Xu, G. X.;
Tao, X.; Zhai, H. X.; Cai, X.; Qian, C. CUDC-305, a novel
synthetic HSP90 inhibitor with unique pharmacologic properties
for cancer therapy. Clin. Cancer Res. 2009, 15, 4046–4057.
(24) Huang, K. H.; Veal, J. M.; Fadden, R. P.; Rice, J. W.; Eaves, J.;
Strachan, J. P.; Barabasz, A. F.; Foley, B. E.; Barta, T. E.; Ma, W.;
Silinski, M. A.; Hu, M.; Partridge, J. M.; Scott, A.; DuBois, L. G.;
Freed, T.; Steed, P. M.; Ommen, A. J.; Smith, E. D.; Hughes, P. F.;
Woodward, A. R.; Hanson, G. J.; McCall, W. S.; Markworth,
C. J.; Hinkley, L.; Jenks, M.; Geng, L.; Lewis, M.; Otto, J.; Pronk,
B.; Verleysen, K.; Hall, S. E. Discovery of novel 2-aminobenzamide
inhibitors of heat shock protein 90 as potent, selective and orally
active antitumor agents. J. Med. Chem. 2009, 52, 4288–4305.
(25) Lundgren, K.; Zhang, H.; Brekken, J.; Huser, N.; Powell, R. E.;
Timple, N.; Busch, D. J.; Neely, L.; Sensintaffar, J. L.; Yang, Y. C.;
McKenzie, A.; Friedman, J.; Scannevin, R.; Kamal, A.; Hong, K.;
Kasibhatla, S. R.; Boehm, M. F.; Burrows, F. J. BIIB021, an orally
available, fully synthetic small-molecule inhibitor of the heat shock
protein Hsp90. Mol. Cancer Ther. 2009, 8, 921–929.
(45) Wolfson, B.; Siker, E. S.; Ciccarelli, H. E. A double blind compar-
ison of doxapram, ethamivan and methylphenidate. Am. J. Med.
Sci. 1965, 249, 391–398.
(46) Roe, S. M.; Prodromou, C.; O’Brien, R.; Ladbury, J. E.; Piper,
P. W.; Pearl, L. H. Structural basis for inhibition of the Hsp90
molecular chaperone by the antitumor antibiotics radicicol and
geldanamycin. J. Med. Chem. 1999, 42, 260–266.
(47) Wright, L.; Barril, X.; Dymock, B.; Sheridan, L.; Surgenor, A.;
Beswick, M.; Drysdale, M.; Collier, A.; Massey, A.; Davies,
N.; Fink, A.; Fromont, C.; Aherne, W.; Boxall, K.; Sharp, S.;
Workman, P.; Hubbard, R. E. Structure-activity relationships in
purine-based inhibitor binding to HSP90 isoforms. Chem. Biol.
2004, 11, 775–785.
(48) Huth, J. R.; Park, C.; Petros, A. M.; Kunzer, A. R.; Wendt, M. D.;
Wang, X.; Lynch, C. L.; Mack, J. C.; Swift, K. M.; Judge, R. A.;
Chen, J.; Richardson, P. L.; Jin, S.; Tahir, S. K.; Matayoshi, E. D.;
Dorwin, S. A.; Ladror, U. S.; Severin, J. M.; Walter, K. A.; Bartley,
D. M.; Fesik, S. W.; Elmore, S. W.; Hajduk, P. J. Discovery and
design of novel HSP90 inhibitors using multiple fragment-based
design strategies. Chem. Biol. Drug Des. 2007, 70, 1–12.
(49) Barker, J. J.; Barker, O.; Boggio, R.; Chauhan, V.; Cheng, R. K.;
Corden, V.; Courtney, S. M.; Edwards, N.; Falque, V. M.; Fusar,
F.; Gardiner, M.; Hamelin, E. M.; Hesterkamp, T.; Ichihara, O.;
Jones, R. S.; Mather, O.; Mercurio, C.; Minucci, S.; Montalbetti,
C. A.; Muller, A.; Patel, D.; Phillips, B. G.; Varasi, M.; Whittaker,
M.; Winkler, D.; Yarnold, C. J. Fragment-based identification of
Hsp90 inhibitors. ChemMedChem 2009, 4, 963–966.
(50) Allen, F. H. The Cambridge Structural Database: a quarter of a
million crystal structures and rising. Acta Crystallogr. B 2002, 58,
380–388.
(51) Chessari, G.; Congreve, M.; Callaghan, O.; Cowan, S.; Murray,
C. W.; Woolford, A. J.; O’Brien, M. A.; Woodhead, A. J. Prepara-
tion of Azinamines as Modulators of Heat Shock Protein 90
(Hsp90). WO2006123165, 2006.
(52) Feldman, R. I.; Mintzer, B.; Zhu, D.; Wu, J. M.; Biroc, S. L.; Yuan,
S.; Emayan, K.; Chang, Z.; Chen, D.; Arnaiz, D. O.; Bryant, J.; Ge,
X. S.; Whitlow, M.; Adler, M.; Polokoff, M. A.; Li, W. W.; Ferrer,
M.; Sato, T.; Gu, J. M.; Shen, J.; Tseng, J. L.; Dinter, H.; Buckman,
B. Potent triazolothione inhibitor of heat-shock protein-90. Chem.
Biol. Drug Des. 2009, 74, 43–50.
(26) Erlanson, D. A.; McDowell, R. S.; O’Brien, T. Fragment-based
drug discovery. J. Med. Chem. 2004, 47, 3463–3482.
(27) Rees, D. C.; Congreve, M.; Murray, C. W.; Carr, R. Fragment-
based lead discovery. Nat. Rev. Drug Discovery 2004, 3, 660–672.
(28) Erlanson, D. A. Fragment-based lead discovery: a chemical up-
date. Curr. Opin. Biotechnol. 2006, 17, 643–652.
(29) Fattori, D. Molecular recognition: the fragment approach in lead
generation. Drug Discovery Today 2004, 9, 229–238.
(30) Hajduk, P. J.; Greer, J. A decade of fragment-based drug design:
strategic advances and lessons learned. Nat. Rev. Drug Discovery
2007, 6, 211–219.
(31) Alex, A. A.; Flocco, M. M. Fragment-based drug discovery: what
has it achieved so far? Curr. Top. Med. Chem. 2007, 7, 1544–1567.
(32) Congreve, M.; Chessari, G.; Tisi, D.; Woodhead, A. J. Recent
developments in fragment-based drug discovery. J. Med. Chem.
2008, 51, 3661–3680.
(33) Chessari, G.;Woodhead, A. J. From fragment to clinical candidate;a
historical perspective. Drug Discovery Today 2009, 14, 668–
675.
(34) Schulz, M. N.; Hubbard, R. E. Recent progress in fragment-based
lead discovery. Curr. Opin. Pharmacol. 2009, 9, 615–621.
(35) Leeson, P. D.; Springthorpe, B. The influence of drug-like concepts
on decision-making in medicinal chemistry. Nat. Rev. Drug Dis-
covery 2007, 6, 881–890.
(36) Lipinski, C. A.; Lombardo, F.; Dominy, B. W.; Feeney, P. J.
Experimental and computational approaches to estimate solubility
and permeability in drug discovery and development settings. Adv.
Drug Delivery Rev. 2001, 46, 3–26.
(53) Cikotiene, I.; Kazlauskas, E.; Matuliene, J.; Michailoviene, V.;
Torresan, J.; Jachno, J.; Matulis, D. 5-Aryl-4-(5-substituted-2,4-
dihydroxyphenyl)-1,2,3-thiadiazoles as inhibitors of Hsp90 chape-
rone. Bioorg. Med. Chem. Lett. 2009, 19, 1089–1092.
(54) Cheung, K. M.; Matthews, T. P.; James, K.; Rowlands, M. G.;
Boxall, K. J.; Sharp, S. Y.; Maloney, A.; Roe, S. M.; Prodromou,
C.; Pearl, L. H.; Aherne, G. W.; McDonald, E.; Workman, P.
The identification, synthesis, protein crystal structure and in
vitro biochemical evaluation of a new 3,4-diarylpyrazole class
of Hsp90 inhibitors. Bioorg. Med. Chem. Lett. 2005, 15, 3338–
3343.
(55) Dymock, B. W.; Barril, X.; Brough, P. A.; Cansfield, J. E.; Massey,
A.; McDonald, E.; Hubbard, R. E.; Surgenor, A.; Roughley, S. D.;
Webb, P.; Workman, P.; Wright, L.; Drysdale, M. J. Novel, potent
small-molecule inhibitors of the molecular chaperone Hsp90 dis-
covered through structure-based design. J. Med. Chem. 2005, 48,
4212–4215.

Article
Journal of Medicinal Chemistry, 2010, Vol. 53, No. 16 5955
(56) Gopalsamy, A.; Shi, M.; Golas, J.; Vogan, E.; Jacob, J.; Johnson,
M.; Lee, F.; Nilakantan, R.; Petersen, R.; Svenson, K.; Chopra, R.;
Tam, M. S.; Wen, Y.; Ellingboe, J.; Arndt, K.; Boschelli, F.
Discovery of benzisoxazoles as potent inhibitors of chaperone heat
shock protein 90. J. Med. Chem. 2008, 51, 373–375.
(57) Kung, P. P.; Funk, L.; Meng, J.; Collins, M.; Zhou, J. Z.; Johnson,
M. C.; Ekker, A.; Wang, J.; Mehta, P.; Yin, M. J.; Rodgers, C.;
Davies, J. F.; Bayman, E.; Smeal, T.; Maegley, K. A.; Gehring,
M. R. Dihydroxylphenyl amides as inhibitors of the Hsp90
molecular chaperone. Bioorg. Med. Chem. Lett. 2008, 18, 6273–
6278.
(58) Kung, P. P.; Huang, B.; Zhang, G.; Zhou, J. Z.; Wang, J.; Digits,
J. A.; Skaptason, J.; Yamazaki, S.; Neul, D.; Zientek, M.; Elleraas,
J.; Mehta, P.; Yin, M. J.; Hickey, M. J.; Gajiwala, K. S.; Rodgers,
C.; Davies, J. F.; Gehring, M. R. Dihydroxyphenylisoindoline
amides as orally bioavailable inhibitors of the heat shock protein
90 (Hsp90) molecular chaperone. J. Med. Chem. 2010, 53, 499–503.
(59) Chessari, G.; Congreve, M.; Figueroa, E.; Frederickson, M.;
Murray, C. W.; Woolford, A. J.; Carr, M. G.; O’Brien, M. A.;
Woodhead, A. J. Preparation of Benzamides as Hsp90 Inhibitors.
WO2006109075, 2006.
(72) Mooij, W. T.; Hartshorn, M. J.; Tickle, I. J.; Sharff, A. J.; Verdonk,
M. L.; Jhoti, H. Automated protein-ligand crystallography for
structure-based drug design. ChemMedChem 2006, 1, 827–838.
(73) Hartshorn, M. J. AstexViewer: a visualisation aid for structure-
based drug design. J. Comput.-Aided Mol. Des 2002, 16, 871–881.
(74) Emsley, P.; Cowtan, K. Coot: model-building tools for molecular
graphics. Acta Crystallogr., Sect. D: Biol. Crystallogr. 2004, 60,
2126–2132.
(75) Laskowski, R. A. PDBsum: summaries and analyses of PDB
structures. Nucleic Acids Res. 2001, 29, 221–222.
(76) Turnbull, W. B.; Daranas, A. H. On the value of c: can low affinity
systems be studied by isothermal titration calorimetry? J. Am.
Chem. Soc. 2003, 125, 14859–14866.
(77) Onuoha, S. C.; Mukund, S. R.; Coulstock, E. T.; Sengerova, B.;
Shaw, J.; McLaughlin, S. H.; Jackson, S. E. Mechanistic studies on
Hsp90 inhibition by ansamycin derivatives. J. Mol. Biol. 2007, 372,
287–297.
(78) Sigurskjold, B. W. Exact analysis of competition ligand binding by
displacement isothermal titration calorimetry. Anal. Biochem.
2000, 277, 260–266.
(79) Shao, Y.; Molnar, L. F.; Jung, Y.; Kussmann, J.; Ochsenfeld, C.;
Brown, S. T.; Gilbert, A. T.; Slipchenko, L. V.; Levchenko, S. V.;
O’Neill, D. P.; DiStasio, R. A., Jr.; Lochan, R. C.; Wang, T.; Beran,
G. J.; Besley, N. A.; Herbert, J. M.; Lin, C. Y.; Van Voorhis, T.;
Chien, S. H.; Sodt, A.; Steele, R. P.; Rassolov, V. A.; Maslen, P. E.;
Korambath, P. P.; Adamson, R. D.; Austin, B.; Baker, J.; Byrd,
E. F.; Dachsel, H.; Doerksen, R. J.; Dreuw, A.; Dunietz, B. D.;
Dutoi, A. D.; Furlani, T. R.; Gwaltney, S. R.; Heyden, A.; Hirata,
S.; Hsu, C. P.; Kedziora, G.; Khalliulin, R. Z.; Klunzinger, P.; Lee,
A. M.; Lee, M. S.; Liang, W.; Lotan, I.; Nair, N.; Peters, B.;
Proynov, E. I.; Pieniazek, P. A.; Rhee, Y. M.; Ritchie, J.; Rosta, E.;
Sherrill, C. D.; Simmonett, A. C.; Subotnik, J. E.; Woodcock,
H. L., III; Zhang, W.; Bell, A. T.; Chakraborty, A. K.; Chipman,
D. M.; Keil, F. J.; Warshel, A.; Hehre, W. J.; Schaefer, H. F., III;
Kong, J.; Krylov, A. I.; Gill, P. M.; Head-Gordon, M. Advances in
methods and algorithms in a modern quantum chemistry program
package. Phys. Chem. Chem. Phys. 2006, 8, 3172–3191.
(80) Verdonk, M. L.; Chessari, G.; Cole, J. C.; Hartshorn, M. J.;
Murray, C. W.; Nissink, J. W.; Taylor, R. D.; Taylor, R. Modeling
water molecules in protein-ligand docking using GOLD. J. Med.
Chem. 2005, 48, 6504–6515.
(81) Watson, P.; Verdonk, M. L.; Hartshorn, M. J. A Web-based
platform for virtual screening. J. Mol. Graph. Modell. 2003, 22,
71–82.
(82) Jones, G.; Willett, P.; Glen, R. C. Molecular recognition of
receptor sites using a genetic algorithm with a description of
desolvation. J. Mol. Biol. 1995, 245, 43–53.
(83) Eldridge, M. D.; Murray, C. W.; Auton, T. R.; Paolini, G. V.; Mee,
R. P. Empirical scoring functions. 1. The development of a fast
empirical scoring function to estimate the binding affinity of
ligands in receptor complexes. J. Comput.-Aided Mol. Des. 1997,
11, 425–445.
(84) Verdonk, M. L.; Cole, J. C.; Hartshorn, M.; Murray, C. W.;
Taylor, R. D. Improved protein-ligand docking using GOLD.
Proteins 2003, 52, 609–623.
(60) Chessari, G.; Congreve, M.; Figueroa, E.; Frederickson, M.;
Murray, C. W.; Woolford, A. J.; Carr, M. G.; Downham, R.;
O’Brien, M. A.; Phillips, T. R.; Woodhead, A. J. Preparation
of Hydroxybenzamides as Hsp90 Inhibitors. WO2006109085,
2006.
(61) Funk, L. A.; Johnson, M. C.; Kung, P. P.; Meng, J. J.; Zhou, J. Z.
Preparation of Hydroxyarylcarboxamide Derivatives for Treating
Cancer. WO2006117669, 2006.
(62) Verdonk, M. L.; Rees, D. C. Group efficiency: a guideline for hits-
to-leads chemistry. ChemMedChem 2008, 3, 1179–1180.
(63) Taber, D. F.; Korsmeyer, R. W. Simple stereoselective synthesis of
(()-oplopanone. J. Org. Chem. 1978, 43, 4925–4927.
(64) Mutsukado, M.; Tanikawa, K.; Shikada, K.; Sakoda, R. 3(2H)-
Pyridazinones and Antiallergic Agents Containing Them.
EP193853, 1986.
(65) Dalvit, C.; Pevarello, P.; Tato, M.; Veronesi, M.; Vulpetti, A.;
Sundstrom, M. Identification of compounds with binding affinity
to proteins via magnetization transfer from bulk water. J Biomol.
NMR 2000, 18, 65–68.
(66) Leslie, A. G. W.; Brick, P.; Wonacott, A. MOSFLM. Daresbury
Lab. Inf. Q. Protein Crystallogr. 2004, 18, 33–39.
(67) Collaborative Computational Project, No. 4. The CCP4 suite:
programs for protein crystallography. Acta Crystallogr. 1994,
D50, 760-763.
(68) Stebbins, C. E.; Russo, A. A.; Schneider, C.; Rosen, N.; Hartl,
F. U.; Pavletich, N. P. Crystal structure of an Hsp90-geldanamycin
complex: targeting of a protein chaperone by an antitumor agent.
Cell 1997, 89, 239–250.
(69) Obermann, W. M.; Sondermann, H.; Russo, A. A.; Pavletich,
N. P.; Hartl, F. U. In vivo function of Hsp90 is dependent
on ATP binding and ATP hydrolysis. J. Cell Biol. 1998, 143,
901–910.
(70) Murshudov, G. N.; Vagin, A. A.; Dodson, E. J. Refinement of
macromolecular structures by the maximum-likelihood method.
Acta Crystallogr. D 2004, D53, 240–255.
(85) Squires, M. S.; Feltell, R. E.; Wallis, N. G.; Lewis, E. J.; Smith,
D. M.; Cross, D. M.; Lyons, J. F.; Thompson, N. T. Biological
characterization of AT7519, a small-molecule inhibitor of cyclin-
dependent kinases, in human tumor cell lines. Mol. Cancer Ther.
2009, 8, 324–332.
(71) Roversi, P.; Blanc, E.; Vonrhein, C.; Evans, G.; Bricogne, G.
Modelling prior distributions of atoms for macromolecular refine-
ment and completion. Acta Crystallogr., Sect. D: Biol. Crystallogr.
2000, 56, 1316–1323.