Identification and optimization of novel Hsp90 inhibitors with tetrahydropyrido[4,3-d]pyrimidines core through shape-based screening

Article abstract of DOI:10.1016/j.ejmech.2014.03.061

Rapid Overlay of Chemical Structures (ROCS), which can rapidly identify potentially active compounds by shape comparison, is recognized as a powerful virtual screening tool. By ROCS, a class of novel Hsp90 inhibitors was identified. The calculated binding mode of the most potent hit 36 guided us to design and synthesize a series of analogs (57a-57h). Over 100-fold improvement was achieved in the target-based assay. The most potent compound 57h inhibited Hsp90 with IC₅₀ 0.10 ± 0.01 μM. It also showed much improved cell potency and ligand efficiency. Our study showed that ROCS is efficient in the identification of novel cores of Hsp90 inhibitors. 57h can be ideal leads for further optimization.

Full text of DOI:10.1016/j.ejmech.2014.03.061

European Journal of Medicinal Chemistry 79 (2014) 399e412
Contents lists available at ScienceDirect
European Journal of Medicinal Chemistry
journal homepage: http://www.elsevier.com/locate/ejmech
Original article
Identification and optimization of novel Hsp90 inhibitors with
tetrahydropyrido[4,3-d]pyrimidines core through shape-based
screening
,
b
,
c
,
,
b
,
,
,
,b
Hao-Peng Sun a 1, Jian-Min Jia a 1, Fen Jiang a b, Xiao-Li Xu a b, Fang Liu a
,
Xiao-Ke Guo a b, Bahidja Cherfaoui a b, Hao-Ze Huang a b, Yang Pan a b, Qi-Dong You a
,
,
,
,
,b,
*
a Jiangsu Key Laboratory of Drug Design and Optimization, China Pharmaceutical University, Nanjing 210009, China
b State Key Laboratory of Natural Medicines, China Pharmaceutical University, Nanjing 210009, China
c Department of Medicinal Chemistry, School of Pharmacy, China Pharmaceutical University, Nanjing 210009, China
a r t i c l e i n f o
a b s t r a c t
Article history:
Rapid Overlay of Chemical Structures (ROCS), which can rapidly identify potentially active compounds by
shape comparison, is recognized as a powerful virtual screening tool. By ROCS, a class of novel Hsp90
inhibitors was identified. The calculated binding mode of the most potent hit 36 guided us to design and
synthesize a series of analogs (57ae57h). Over 100-fold improvement was achieved in the target-based
Received 21 January 2014
Received in revised form
19 March 2014
Accepted 20 March 2014
Available online 21 March 2014
assay. The most potent compound 57h inhibited Hsp90 with IC50 0.10 ꢀ 0.01
mM. It also showed much
improved cell potency and ligand efficiency. Our study showed that ROCS is efficient in the identification
of novel cores of Hsp90 inhibitors. 57h can be ideal leads for further optimization.
Ó 2014 Published by Elsevier Masson SAS.
Keywords:
Hsp90 inhibitors
ROCS
Design and synthesis
1. Introduction
Hsp90 derived from tumor cells has particularly higher expression
level (4%e6% of the whole proteomic load of the cell) than in
Molecular chaperones are protein machines that are responsible
for conformational stability, maturation, and function of other
substrate proteins, which are known as clients [1]. Many molecular
chaperones and co-chaperones have been identified to date.
Among all the molecular chaperones, a 90 kDa heat shock protein
(Hsp90) is the most attractive member and the most extensively
studied [2e5]. Hsp90 is a molecular chaperone whose association is
required for the stability and function of multiple mutated,
chimeric and over-expressed signaling proteins that promote can-
cer cell growth and/or survival [6e8]. Hsp90 client proteins include
mutated signaling proteins (p53, Bcr-Abl, Raf-1, Akt, etc.), HER2/
normal cells (1%e2%), this makes Hsp90 to be an ideal and selective
target in the treatment of cancers [13e17]. Besides, targeting Hsp90
leads to the regulation of multiple pathways simultaneously,
therefore, it is a promising therapeutic strategy in the suppression
of tumor acquiring resistance [18].
The Hsp90 protein folding process is ATP dependent. Chaper-
oning of these client proteins is regulated through a dynamic cycle
driven by ATP binding to Hsp90 and subsequent hydrolysis [19].
Inhibition of the ATPase activity of Hsp90 disrupts an ongoing
“folding” cycle, including multiple co-chaperone proteins, and in
turn leads to the destabilization, ubiquitination, and ultimately
proteasomal degradation of client proteins [20]. The identification
of Hsp90 inhibitors has been attractive for medicinal chemists since
the discovery of geldanamycin and radicicol (Fig. 1), which are able
to inhibit Hsp90 function through binding to the ATP binding
pocket in its N-terminal domain [21]. In 1998e1999, the first clin-
ical trials with an Hsp90 inhibitor were initiated on the geldana-
mycin derivative 17-allylamino-17-desmethoxygeldanamycin (17-
AAG, Fig. 1). Although with high in vitro activity, the poor solubil-
ity as well as the hepatotoxicity hinder the further usage of 17-AAG
as clinical candidate [22]. The water-soluble analog 17-
dimethylaminoethylgeldanamycin (17-DMAG, Fig. 1), proves to be
Neu (ErbB2), nNos, Src, Flt-3, hTert, c-Met, HIF-1a, epidermal
growth factor receptors (EGFRs) and growth factor receptors (IGF-
1Rs), Cdk4 [9]. Several Hsp90 clients are notorious oncogenes (Raf-
1, Akt, Cdk4, Src, Flt-3, hTert, c-Met, etc.), and five of them are
clinically validated cancer targets: HER-2/neu, Bcr-Abl, estrogen
receptor, androgen receptor, and VEGFR [9e12]. Additionally,
* Corresponding author. Jiangsu Key Laboratory of Drug Design and Optimization,
China Pharmaceutical University, Nanjing, 210009, China.
E-mail address: youqidong@gmail.com (Q.-D. You).
These authors contributed equally to this work.
1
http://dx.doi.org/10.1016/j.ejmech.2014.03.061
0223-5234/Ó 2014 Published by Elsevier Masson SAS.

400
H.-P. Sun et al. / European Journal of Medicinal Chemistry 79 (2014) 399e412
Fig. 1. Hsp90 inhibitors in clinical trials.
more toxic in preclinical species [23]. 17-AAG hydroquinone
(IPI-504, Fig. 1) and its major metabolite 17-amino-17-
desmethoxygeldanamycin (17-AG, Fig. 1) are now being devel-
oped by Infinity. They displayed good water solubility, much
improved oral bioavailability [24,25].
model) [41] common structure pharmacophore might be the most
successful and popular models. However, these methods are closely
depend on the training set selection during the model construction.
The resulted models always pay too much attentions on the com-
mon structural information, neglecting the unique characters of a
given target, such as spatial volume requirements and physico-
chemical properties. Additionally, the experience of the researchers
also deeply affects the accuracy and reliability of the models.
Rapid Overlay of Chemical Structures (ROCS) [42] is another
useful ligand-based method that can avoid the problems
mentioned above. It uses an alignment algorithm accounting
shape/physicochemical properties for the query molecule orienta-
tion. The hits from this model not only have similar arrangement of
critical chemical features to the reference molecule, but also
possess similar molecular shape and orientation. The complement
of these characters can help medicinal chemists to discover com-
pounds which may very different to the reference molecule, but are
actually very similar in the whole level of molecular conformation
and chemical features arrangement. As a result, ROCS shows high
success rate in the identification of molecules with novel cores for a
given target.
In parallel, many fully synthetic small molecules are reported as
potent and selective Hsp90 inhibitors by various institutions. Some
of these compounds have entered into clinical trials in different
phase [26,27]. Based on the structures of these small molecule in-
hibitors, there are two major cores, the purine and resorcinol. The
purine class is designed by structural homology with ATP [28].
Many inhibitors belong to this family and among them, PU-H71
(Memorial Sloan-Kettering Cancer Center, phase I) [29], CUDC305
(Debiopharm, phase I) [30], MPC-3100 (Myrexis, phase I) [31],
BIIB021 (Biogen Idec, phase II) (Fig. 1) are now in clinical trial [32].
The resorcinol scaffold is another important class of Hsp90 in-
hibitors, NVP-AUY922 (Novartis, phase II) [33], AT-13387 (Astex,
phase II) [34], STA-9090 (Synta, phase II) [35] and KW-2478 (Kyowa
Hakko Kirin, phase I/II) [36] are in clinical study. However, to date,
no Hsp90 inhibitors can fully satisfy the requirement of safety and
stability, thus none of them is approved to enter the market. As a
result, efficient identification of promising inhibitors with different
chemotypes is still a demanding task in this area.
We believed that the excellent activity of the resorcinol scaffold,
in concert with new synthetic methodologies, would provide a
good starting point for the synthesis of novel analogs, and as such,
our Hsp90 drug discovery program was also based on the resorcinol
class.
In this study, we selected the potent compound AT-13387 with
resorcinol core as the reference molecule for the construction of
ROCS model. The model was validated by a decoy set download
Ligand-based virtual screening of large compound databases
has been proved to an effective method in discovering hits with
novel chemotypes [37e39]. It is also be widely agreed that ligand-
based methods are more efficient and duplicable than structure-
based methods. Among all the Ligand-based approaches, the
qualitative (e.g. Hiphop model) [40] or quantitative (e.g. Hypogen

H.-P. Sun et al. / European Journal of Medicinal Chemistry 79 (2014) 399e412
401
from Directory of Useful Decoys (DUD, http://dud.docking.org/)
containing 979 compounds with similar physicochemical proper-
ties to AT-13387. The model can accurately recognize the active
compounds from the decoy set based on the molecular shape
similarity. After validation, the model is applied to virtually screen
the Topscience database (www.tsbiochem.com). The top 50 com-
pounds by the scoring function of ROCS were purchased from
Topscience for biological test. The inhibition effect of these 50
compounds against Hsp90 was firstly evaluated using Malachite
study suggested that the bioactive conformation of a reference
molecule is not needed for the enrichment of shape based
screening [44]. In this study, the promising Hsp90 inhibitor AT-
13387 was selected as the reference compound to generate the
ROCS model. The molecular shape of AT-13387 was depicted in gray
shadow (Fig. 2A). The 3D similarity of a given compound to AT-
13387 was ranked by two methods: a) The molecular shape simi-
larity and b) The combo score method, which was consisted of the
shape Tanimoto coefficient and the score retrieved from the ROCS
color force field. In this chemical force field, a molecule is described
by the spatial arrangement of six types: hydrogen-bond donors,
hydrogen-bond acceptors, hydrophobes, anions, cations, and rings.
ROCS color force field can be used to measure chemical features
complementarity, and to refine shape based superpositions based
on chemical similarity. As both scores vary from 0 to 1, the combo
score as the sum of both varies between 0 and 2 [45].
In order to validate the reliability of the model, an active set
including 31 compounds was collected from literatures [34,46]. The
decoy set including 979 compounds was downloaded from Hsp90
package in DUD. The multiple conformation of the active and decoy
compounds were generated by OMEGA in Openeye. The validation
was performed according to the shape similarity (Fig. 3A) and the
combo score (Fig. 3B), respectively. Both the methods can well
distinguish the active compounds from the decoy compounds. For
shape similarity, most of the active compounds were above 0.6,
while the decoy compound were lower than 0.6. For combo score,
most of the active compounds were above 1.2, while about 80% of
the compounds in decoy set were less than 1.0.
Green assay. 11 compounds with over 60% inhibition rate at 40 mM
were selected for reevaluation by the Discover RX ADP HunterÔ
Plus Assay kit (ADP assay). Most of them showed moderate inhi-
bition potency against Hsp90. Compound 36 containing the 5,6,7,8-
tetrahydropyrido[4,3-d]pyrimidine core displayed the most potent
inhibition (IC50 45.39 ꢀ 2.82
mM) in the fluorescence polarization
(FP) assay. The binding mode of 36 with Hsp90 was predicted by
molecular docking and guided by the revealed information, eight
derivatives of 36 were designed and synthesized. These derivatives
showed much improved potency both in target-based level and
cell-based level. The most potent compound 57h, exhibited over
450-fold improvement in Hsp90 inhibition (IC50 0.10 ꢀ 0.01
and over 27-fold in antiproliferative effect in MCF cells
(1.91 ꢀ 1.21 M) compared to 36. Western blot analysis confirmed
mM)
m
that 57h downregulated a series of client proteins of Hsp90,
including Her2, Erk, AKT and Raf-1 in a concentration-dependent
manner. The induced Hsp70 expression, another hallmark of
Hsp90 inhibition, was also observed. Our data prove the ligand-
based method ROCS combined with structure-based modification
can be well applied in the identification of novel Hsp90 inhibitors
and more chemotypes are expected to be discovered by this
method. Besides, compound 57h can be used as an ideal lead for
further optimization. The ligand-based method we reported here,
was meaningful to guide the researchers to efficaciously identify
novel inhibitors, especially for those targets with potent ligands.
To avoid potential bias of the validation results due to large
differences in molecular properties, we compared the basic physi-
cochemical properties between active set and decoy set, the
average values were shown in Table 1. The active set compounds
have two HBA less than the decoy set compounds, meanwhile, the
PSA of the active set is lower than the decoy set. However, the other
key physicochemical properties such as molecular weight, HBD,
rotatable bond and Log D are similar, indicating no potential bias
exists in the validation of the overlays. As a result, we believe the
overlays are reliable for the further virtual screening study.
2. Result and discussion
2.1. Generation and validation of ROCS model based on AT-13387
ROCS is a fast shape comparison application, based on the idea
that molecules have similar shape if their volumes overlay well and
any volume mismatch is a measure of dissimilarity. It uses a smooth
Gaussian function to represent the molecular volume [43], so it is
possible to routinely minimize to the best global match. A recent
2.2. Database screening for potential Hsp90 inhibitors
To identify novel Hsp90 inhibitors, the Topscience database
(over 1.5 million compounds) was used to carry out the virtual
screening using the ROCS model. The multiple conformations of
Fig. 2. (A) The ROCS model generated from AT-13387; (B) The bound conformation of AT-13387 in the co-crystal structure (PDB id: 2XJX); (C) w (F) The overlay of the hit
compounds 1, 21, 27 and 36 to the ROCS model, respectively.

402
H.-P. Sun et al. / European Journal of Medicinal Chemistry 79 (2014) 399e412
Fig. 3. The validation of the ROCS model based on AT-13387. An active set containing 31 compounds and a decoy set containing 979 compounds were applied for the validation.
(A) The validation results based on shape similarity; (B) The validation results based on combo score.
different query molecules were calculated using OMEGA (default
settings, with a maximum of 400 conformations per molecule). In
the subsequently virtual screening, the query molecules were
ranked by the shape similarity and the combo score. The top fifty
compounds (in Supplementary data) were purchased from Top-
science for further biological evaluation.
very similar molecular shape, thus has similar biological functions.
Therefore, shape-based virtual screening is efficacious in the
identification of novel hits.
From the structures of the 11 compounds, it was striking to
observe that they possessed the same core, tetrahydropyrido[4,3-d]
pyrimidine. To confirm the results of malachite green assay, we
performed another assay using Discover RX ADP HunterÔ Plus
Assay kit. Most compounds except compound 7 showed consistent
effects to the results of malachite green assay (Table 3). The most
2.3. Biological assay of the hit compounds
potent compound 36 exhibited
a dose-dependent inhibitory
To characterize the inhibition of Hsp90 by the 50 compounds,
malachite green assay for inorganic phosphate was initially used to
measure the ATPase activity of Hsp90. Malachite green assay was a
colorimetric assay which was well suited to the automation
required for high throughput screening as a rough and fast evalu-
ation methodology [47,48]. 11 of 50 compounds (Fig. 4) showed
manner with IC50 35.18 ꢀ 3.72
m
M (Fig. 5A). Next, the four best
compounds 13, 14, 27 and 36 in this round of assay were selected
for further FP competitive binding assay to avoid false positives.
Only compound 36 showed moderate inhibition activity (IC50
45.39 ꢀ 2.82
mM, Table 3).
To further confirm the Hsp90 inhibitory activities of 36, we
prepared a luciferase reporter assay by detecting the expression
level of Hsp70, a key Hsp family member that was upregulated
when Hsp90 was inhibited. As expected, 36 showed moderate
Hsp70 induction in a dose-dependent manner, it showed over 1.5
over 60% inhibition rate at 40 mM (in Supplementary data).
We then analyzed the shape similarity and the combo score of
these compounds (Table 2). For shape similarity, the values of the
11 compounds were ranging from 0.623 to 0.735. As shown in
Fig. 3A, most of the compounds in the active set had the shape
similarity over 0.6, therefore, these compounds were recognized as
actives by ROCS model. Oppositely, for combo score, the values of
the 11 compounds were ranging from 0.935 to 1.119, which was
recognized as inactive in the model validation (Fig. 3B, most of the
actives were over 1.2). This deviation can be explained by the very
low chemical features similarity between the hits and the template
molecules (Table 2). Considering these results, shape similarity
evaluation can help us screen out more hits which might be
omitted by combo score, thus might be more proper in the evalu-
ation of the hits from virtual screening than combo score. In the
chemical database containing huge compounds, a hit compound
may have totally different spatial location and the arrangement of
chemical features compared to the template molecule, but form a
times induction of Hsp70 at the concentration of 10.0 mM (Fig. 5B).
The data further supported that 36 inhibited the activity of Hsp90
in a cell-based level. On the basis of these findings, compound 36
was selected for structural optimization with the aim to improve its
activity in not only target-based but also cell-based level.
2.4. Design of new derivatives based on hit compound 36 and
precursor evaluation
To analyze the binding mode of Hsp90 with 36, it was docked in
to active site of Hsp90 (PDB code: 2XJX). We could observe that 2,5-
dimethoxy benzyl (fragment A) of 36 inserted into the bottom of
the ATP-binding pocket comprised by the residues Asp93, Asn51
and Thr184, and 2-(1-piperazinyl)pyrimidine (fragment B) of 36
located at the edge of the pocket, pointing into the solvent acces-
sible area (Fig. 6A and B). The general molecular orientation and the
spatial location of the chemical features of 36 were similar to that of
AT-13387. However, we found the location of 2,5-dimethoxy benzyl
moiety was different from the resorcinol of AT-13387. We deduced
the variation was caused by the flexibility of the carbonenitrogen
bond in compound 36. In addition, we observed this bond was close
to the hydroxyl group of Thr184, a key residue for the recognition of
Hsp90. Therefore, the carbonenitrogen bond was replaced by the
carbonyl group. The resulted compound 57a was docked into the
active site (Fig. 6C and D). The binding mode of 57a was in good
agreement with our hypotheses: the introduced carbonyl formed a
Table 1
The comparison of the physicochemical properties of the active set, decoy set and
the hits.
MWa
HBAb
HBDc
RBd
PSAe
Log D
Active set
Decoy set
Hits
355.80
350.49
419.40
1.45
3.19
4
2.16
2.31
0.09
3.65
4.89
4.54
69.62
108.09
66.81
2.97
2.70
2.72
a
Molecular weight.
b
c
Hydrogen-bond acceptor.
Hydrogen-bond donor.
Rotatable bond.
d
e
Polar surface area.

H.-P. Sun et al. / European Journal of Medicinal Chemistry 79 (2014) 399e412
403
Fig. 4. The structures of 11 compounds over 60% inhibition rate in the Malachite Green Assay.
direct hydrogen bond with the side chain of Thr184. 57a was sub-
sequently synthesized (see Chemistry, Scheme 1) and showed
slight improvement in affinity for Hsp90 in the FP and ADP assay,
respectively (Table 4). Similar manner was also observed in com-
pounds 57b and 57c, compared to their counterparts 13 and 14
without the carbonyl group. However, 57b and 57c displayed lower
activity compared to 57a, indicating methoxy substitution was
more favorable than F or Cl. This phenomenon could be attributed
to the H-bond recognition between methoxyl and Hsp90.
Further analysis of the binding mode of 57a indicated that
methoxyl at 5 position of 2,5-dimethoxy benzyl was too close to a
sub-pocket formed by Leu107, Phe138, and Val150 (Fig. 6C and D).
This can cause unfavorable intermolecular energy. Therefore, we
next move the methoxyl group to the 4 position of 2,5-dimethoxy
benzyl. The resulted compound 57d show obviously improved ac-
tivity compared to 57a (Table 4). Binding mode analysis suggested
that methoxyl group at R2 position can facilitate an additional in-
direct hydrogen bond with the amine on Asn51 aided by the
conserved water molecular W1 (Fig. 6E and F).
Asn51. As the two residues were recognized as the most significant
points for Hsp90 to recognize the inhibitors through forming
hydrogen bond networks with the conserved water molecules
around the two residues [33], we therefore changed the dime-
thoxyphenyl ring into the resorcinol moiety. Encouragingly, the
resulted compound 57e (see Chemistry, Scheme 1), showed much
better Hsp90 inhibitory activity than 57d (Table 4). Binding mode of
57e with Hsp90 showed direct hydrogen bonds between the hy-
droxyl of the resorcinol moiety with Asp93 and Asn51 and
conversed water molecules (Fig. 7A and B).
However, this compound missed the occupation of the sub-
pocket consist of Leu107, Phe138, and Val150, as
a result,
although the activity increased, it still belong to the micromolar
range. With the aim to enhance the Hsp90 inhibitory activity to
nanomolar range, we tried to occupy the sub-pocket by modifying
the substituents on the resorcinol moiety. As we previously
mentioned that the methoxyl was too large for the sub-pocket, we
further designed compound 57f with chloro substitution at 5 po-
sition of 2,5-dimethoxy benzyl (see Chemistry, Scheme 1). The FP
assay showed that 57f was much more potent than 57e, with the
Further work on this study was focused on the enhancement of
Hsp90 inhibitory and antiproliferative activity of 57d. The binding
mode of 57d with Hsp90 suggested that 2- and 4-methoxy groups
only interacted with the conversed water molecular W1 and W2,
rather than directly interacted with two polar residues Asp93 and
low micromolar IC50 1.67 ꢀ 0.97
mM. The ADP assay also showed
Table 3
Hsp90 ATPase activity inhibition (IC50) in the ADP and FP assay by the 11 compounds
over 60% inhibition in the Malachite Green Assay.
Compd. no.
Hsp90-ATPase activity
inhibition (
M)a
Binding Hsp90
Table 2
m
(FP, m
M)a
The shape similarity, chemical features similarity and combo score of the 11 hit
compounds.
1
7
13
14
21
23
27
30
32
36
39
17-DMAG
AT-13387
50.90 ꢀ 4.31
>100
ND
ND
>100
>100
ND
ND
>100
ND
Compd. no.
Shape similarity
Chemical features similarity
Combo score
40.24 ꢀ 3.22
47.88 ꢀ 2.91
63.20 ꢀ 1.73
55.12 ꢀ 6.11
46.13 ꢀ 0.49
74.68 ꢀ 1.12
59.28 ꢀ 3.25
35.18 ꢀ 3.72
87.96 ꢀ 1.25
1.41 ꢀ 0.05
0.35 ꢀ 0.11
1
7
0.735
0.708
0.627
0.629
0.669
0.643
0.625
0.623
0.642
0.656
0.681
0.250
0.170
0.202
0.202
0.220
0.214
0.212
0.275
0.236
0.190
0.212
1.119
1.024
0.935
0.937
0.985
0.981
0.946
1.000
0.992
0.974
0.972
13
14
21
23
27
30
32
36
39
ND
45.39 ꢀ 2.82
ND
0.09 ꢀ 0.02
0.03 ꢀ 0.01
a
Values shown are mean ꢀ SD (n ¼ 3).

404
H.-P. Sun et al. / European Journal of Medicinal Chemistry 79 (2014) 399e412
Fig. 5. The biological evaluation of the most potent compound 36. (A) The dose-dependent inhibitory manner of 36 in the ADP assay. (B) The Hsp70 induction of Hsp70 in SK-Br-3
cell after the treatment of 36.
similar manner, with
a
4-fold improvement in IC50 values
However, 57f was still much less potent than the reference
compound AT-13387 and 17-DMAG. To achieve the nanomolar goal,
we further analyzed the scaffold of 57f. As inspired by the docking,
the 2-(piperazin-1-yl)pyrimidine moiety was not inserted into the
ATP binding pocket of Hsp90, but pointed to the solvent area at the
edge of the binding groove. As this area was not important for the
compared to 57e (Table 4). The introduction of 5-chloro moiety also
increased the antiproliferative potency in the four evaluated cell
lines. The binding mode of 57f with Hsp90 showed the 5-chloro
moiety of 57f pointed to the sub-pocket as expected (Fig. 7C and
D). This is in good agreement with the observation of Roe et al. [49].
Fig. 6. The binding patterns of between the potent compounds and Hsp90. (A, C, E) 36, 57a and 57d in the active site of Hsp90 respectively, key residues were labeled in stick; (B, D,
F) 36, 57a and 57d inserted into the active site of Hsp90 with a surface colored by electrostatic state.

H.-P. Sun et al. / European Journal of Medicinal Chemistry 79 (2014) 399e412
405
Scheme 1. Synthesis of compounds 57ae57h. Reagents and conditions:(a) urea, MeONa, EtOH, reflux, 24 h, 75%; (b) POCl3, N2, 3 h, 87%; (c) activated Zn, NH4OH, EtOH, reflux, 12 h,
61%; (d) 1-(2-Primidinyl)piperazine, Na2CO3, dioxane, 85 ꢁC, 6 h, 75%; (e) 10% Pd/C, H2, MeOH, rt, 15 h, 81%; (f) substituted benzoic acid, EDCI, HOBt, Et3N, dichloromethane, rt, 12 h,
88%; (g) 10% Pd/C, H2, MeOH, rt, 2 h, 80%; (h) BBr3, dichloromethane, 0 ꢁC, 3 h, 45%; (i) N,N-dimethylformamide dimethyl acetal, reflux, 1.5 h; 79%; (j) guanidine carbonate or
acetamidine hydrochloride, sodium acetate, EtOH, reflux, 42 h, 83%; (k) TFA, 0 ꢁC, 1 h, 95%.
recognition of Hsp90 inhibitors, the extra solvation energy of this
moiety was unfavorable for the binding affinity. Besides, the mo-
lecular weight of 57f was 467, which was too high for an ideal lead
compounds for further optimization. The high molecular weight
also lead to low ligand efficiency, which was recognized as a critical
profile for a drug candidate. Taken these facts into consideration,
we next removed the 2-(piperazin-1-yl)pyrimidine (fragment B) to
obtain compound 57g and 57h (see Chemistry, Scheme 1). To our
delight, the two compounds exhibited much higher potency
compared to 57f in both target-based and cell-based level. Com-
pound 57g exhibited nanomolar activity in Hsp90 assay (IC50
0.31 ꢀ 0.13
m
M and 0.51 ꢀ 0.21
mM in FP assay and ADP assay,
respectively, Table 4), meanwhile, compound 57h exhibited com-
parable activity to the positive control 17-DMAG and AT-13387, in
not only Hsp90 assay (IC50 0.10 ꢀ 0.01
m
M and 0.42 ꢀ 0.11
mM in FP
assay and ADP assay, respectively, Table 4) but also antiproliferative
assay of the four cell lines. 57h with a methyl substituted the
fragment B, showed the best activity. This indicated that
Table 4
Antiproliferative activity in several cell lines, Hsp90 ATPase activity inhibition in the ADP and FP assay by the compounds 57ae57h.
Compd. no.
Antiproliferative activity cell lines (IC50
,
m
M)a
Hsp90-ATPase activity inhibition (IC50
,
m
M)a
Binding Hsp90 (FP, IC50, m
M)a
SK-Br-3
MCF-7
HCT116
A231
36
57a
57b
57c
57d
57e
57f
57g
10.73 ꢀ 1.21
23.54 ꢀ 1.42
24.5 ꢀ 1.30
11.23 ꢀ 1.71
16.01 ꢀ 1.63
12.23 ꢀ 1.81
5.17 ꢀ 0.43
2.30 ꢀ 0.11
4.32 ꢀ 0.74
3.11 ꢀ 0.21
0.14 ꢀ 0.01
53.25 ꢀ 3.13
43.31 ꢀ 1.91
46.17 ꢀ 4.54
42.16 ꢀ 3.75
20.16 ꢀ 2.84
17.52 ꢀ 0.34
10.22 ꢀ 1.25
4.41 ꢀ 1.03
1.91 ꢀ 1.21
0.8 ꢀ 0.32
43.20 ꢀ 1.45
34.56 ꢀ 1.02
67.89 ꢀ 2.75
21.68 ꢀ 6.20
11.48 ꢀ 1.44
23.14 ꢀ 2.82
2.74 ꢀ 1.57
1.21 ꢀ 0.23
2.82 ꢀ 0.92
1.21 ꢀ 0.21
0.08 ꢀ 0.02
27.53 ꢀ 2.71
51.32 ꢀ 1.93
54.44 ꢀ 1.61
24.60 ꢀ 4.11
37.60 ꢀ 3.26
21.32 ꢀ 1.11
6.65 ꢀ 0.39
5.22 ꢀ 1.08
3.76 ꢀ 0.56
0.17 ꢀ 0.07
1.01 ꢀ 0.42
35.18 ꢀ 1.75
16.44 ꢀ 1.51
27.21 ꢀ 2.17
24.48 ꢀ 1.92
7.88 ꢀ 0.93
3.23 ꢀ 0.62
0.93 ꢀ 0.21
0.51 ꢀ 0.21
0.42 ꢀ 0.11
1.41 ꢀ 0.05
0.34 ꢀ 0.11
45.39 ꢀ 2.83
40.36 ꢀ 0.34
>100
>100
30.54 ꢀ 1.68
10.23 ꢀ 2.55
1.67 ꢀ 0.97
0.31 ꢀ 0.13
0.10 ꢀ 0.01
0.09 ꢀ 0.02
0.03 ꢀ 0.01
57h
17-DMAG
AT-13387
0.28 ꢀ 0.07
a
Values shown are mean ꢀ SD (n ¼ 3).

406
H.-P. Sun et al. / European Journal of Medicinal Chemistry 79 (2014) 399e412
Fig. 7. The binding patterns of between the potent compounds and Hsp90. (A, C, E) 57e, 57f and 57h in the active site of Hsp90 respectively, key residues were labeled in stick; (B, D,
F) 57e, 57f and 57h inserted into the active site of Hsp90 with a surface colored by electrostatic state.
hydrophobic side chains at this position were preferable to the
polar groups. Compared to AT-13387, the novel tetrahydropyrido
[4,3-d]pyrimidines core of 57h may stabilize the molecule in the
binding groove by the extra polar contacts through the N atoms in
the ring (Fig. 7E and F).
Hsp70 induction in a dose-dependent manner, it showed over 3
times induction of Hsp70 at the concentration of 1 M (Fig. 8A).
m
Encouraged by these results, the ability of 57h to affect turnover of
several Hsp90 client proteins in SK-Br-3 cells was also assessed. As
shown in Fig. 8B, 57h dose-dependently induced the degradation of
Hsp90 client proteins Her2, Akt, Erk and Raf-1, which was in a
similar manner with that in the anti-proliferative assay. At the same
Subsequently, we conducted a luciferase reporter assay by
detecting the expression level of Hsp70. 57h showed the obvious
Fig. 8. The biological evaluation of the most potent compound 57h. (A) The Hsp70 induction of Hsp70 in SK-Br-3 cell after the treatment of 57h. (B) Western-Blot analysis of the
expression level of Hsp90, Hsp70 and several client proteins of Hsp90 after incubation with 57h in SK-Br-3 cells. 17-DMAG and AT-13387 were used as positive control at 1.0 M. 57h
M) for 36 h. Cell extracts were prepared and equivalent amounts of protein were separated by SDS-PAGE and
m
was incubated with SK-Br-3 cancer cells at indicated concentrations (
subsequently western-blotted for the indicated proteins.
m

H.-P. Sun et al. / European Journal of Medicinal Chemistry 79 (2014) 399e412
407
time, 57h dose-dependently up-regulated Hsp70. All the data
confirmed 57h as potent Hsp90 inhibitor.
and FP assay, respectively). Anti-proliferative assay results also
showed that compound 36 had the potential to be developed as an
anti-proliferative agent against cancer cells. To improve potency of
the hit 36, we designed and synthesized eight analogs aided by
structure-based design using docking simulation. As we expected,
most derivatives showed improvement potency both in target-
based and cell-based assay, and the most potent derivative 57h
showed over 450-fold improvement in Hsp90 inhibition compared
2.5. Chemistry
The synthesis routes used to prepare compounds 57ae57h were
depicted in Scheme 1. Compounds 57ae57f were prepared in an
analogous manner starting from the commercially available ethyl
1-benzyl-4-oxopiperidine-3-carboxylate (compound 51, Scheme
1). 52 was prepared by amination of 51 with urea under basic
conditions. Then, 52 was treated with POCl3 to provide the desired
intermediate 53. 53 was reduced by Zn to give compound 54. The
intermediate 55 was obtained by reaction of 53 and commercially
available 1-(2-Pyrimidinyl)piperazine. The benzyl group of 55 was
removed by Pd/C to afford the key intermediate 56. Finally, 57ae57f
were obtained by coupling the key intermediate 56 with the
appropriate substituted benzoic acid using standard EDCI/HOBt
coupling conditions. Compounds 57g and 57h were prepared in an
analogous manner starting from the commercially available com-
pound 58. Reaction of the compound 58 with N,N-
dimethylformamide dimethyl acetal at reflux afforded 59. The in-
termediate 59 was treated by guanidine carbonate or acetamidine
hydrochloride in ethyl alcohol in the presence of sodium acetate to
provide intermediates 60a and 60b. Compounds 57g and 57h were
obtained from 60a and 60b in the same way as 57a - 57f.
to the original hit 36, with IC50 0.42 ꢀ 0.11
m
M and 0.10 ꢀ 0.01
mM in
the ADP and FP assay respectively. In addition, 57h displayed much
better potency in affecting the degradation of client proteins and
cell proliferation in Sk-Br-3 cell. These results provided important
information for further structural modifications of 57h to develop
novel potent Hsp90 inhibitors.
By analyzing the biological data, we summarize the SAR as fol-
lows (Fig. 9): i) the fragment A inserted into the active site
comprised by the residues Asp93, Asn51 and Thr184, and the
fragment B made no direct contact with the protein and was
directed towards solvent; ii) the introduction of the carbonyl, 2-OH
and 4-OH improved the potency by formed H-bond with the resi-
dues Thr184, Asp93 and Asn51; iii) the chloro at 5-position was
important for cytotoxicity; iv) the fragment B could be replaced by
the amino or methyl group, and the methyl group in this position
was favorable for the Hsp90 inhibition activities.
The ligand-based method we reported here, was also efficacious
and meaningful to guide the researchers to identify novel in-
hibitors, especially for those targets with potent ligands.
3. Conclusion
In this study, a ROCS model on the basis of AT-13387 was
established and validated to screen Topscience database, from
which 11 compounds containing the similar scaffold were identi-
fied as potent inhibitors of molecular chaperone Hsp90. Among all
of the 11 compounds, compound 36 demonstrated the most potent
4. Experimental
4.1. Chemistry
Melting points were determined on a Mel-TEMP II melting point
apparatus and are uncorrected. 1H NMR spectra were recorded with
inhibitory activity (IC50 of 35.18 ꢀ 3.72
mM, 45.39 ꢀ 2.82
mM for ADP
Fig. 9. Structureeactivity relationships of the optimized compounds based on the screening hit 36.

408
H.-P. Sun et al. / European Journal of Medicinal Chemistry 79 (2014) 399e412
a Bruker Avance 300 MHz spectrometer at 300 K, using TMS as an
internal standard. MS spectra were recorded on a Shimadzu GCe
MS 2010 (EI) or a Mariner Mass Spectrum (ESI), or a LC/MSD TOF
HR-MS Spectrum. All compounds were routinely checked by TLC
and 1H NMR. TLCs and preparative thin-layer chromatography were
performed on silica gel GF/UV 254 supported by glass plate, and the
chromatograms were performed on silica gel (200e300 mesh)
visualized under UV light at 254 and 365 nm. Purity for final
compounds was greater than 95% and was measured by HPLC with
Agilent Technologies 1260 infinity C18 4.60 mm ꢂ 150 mm column
using a mixture of solvent methanol/water at the flow rate of
0.5 mL/min and peak detection at 245 nm under UV. All solvents
were reagent grade and, when necessary, were purified and dried
by standards methods. Concentration of solutions after reactions
and extractions involved the use of a rotary evaporator operating at
a reduced pressure of ca. 20 Torr. Organic solutions were dried over
anhydrous sodium sulfate. Analytical results are within (0.40% of
the theoretical values).
afford 4.3 g (61%) of 54 as a white solid. Rf ¼ 0.2 (PE:EA ¼ 8:1); Mp:
253e254 ꢁC; 1H NMR (300 MHz, CDCl3,
ppm): 2.33e2.35 (m, 2H),
2.52e2.53 (m, 2H), 3.11 (s, 2H), 3.53 (s, 2H), 7.13e7.39 (m, 5H), 8.53
d
(s, 1H).
4.1.4. 6-Benzyl-2-(4-(pyrimidin-2-yl)piperazin-1-yl)-5,6,7,8-
tetrahydropyrido[4,3-d]pyrimidine 55
54 (4.0 g, 15.4 mmol) was dissolved in dioxane (100.0 mL),
then 1-(2-Pyrimidinyl)piperazine (5.1 g, 30.8 mmol) and sodium
carbonate (16.3 g, 154.0 mmol) were added thereto and the
mixture was stirred at 90 ꢁC for 30 h. The resulting reaction was
filtered to remove sodium carbonate, extraction was conducted by
addition of water and chloroform to the filtrate and the organic
layer was dried over magnesium sulfate. After evaporation of the
solvent, a mixed solvent of hexane/ethyl acetate (5:1) was added
and the suspension was stirred for 1 h. After that, the crystals
separated out therefrom were filtered and dried in vacuo to pre-
pare compound 55 (4.5 g, 75%). Rf ¼ 0.4 (PE:EA ¼ 1:2); Mp: 179e
181 ꢁC; 1H NMR (300 MHz, CDCl3,
d ppm): 2.43e2.45 (m, 2H),
4.1.1. 6-Benzyl-5,6,7,8-tetrahydropyrido[4,3-d]pyrimidine-
2,4(1H,3H)-dione 52
2.82e2.83 (m, 2H), 3.51 (s, 2H), 3.63 (s, 2H), 3.82 (s, 8H), 6.51e6.54
(m, H), 7.27e7.41 (m, 5H), 8.00 (s, 1H), 8.34e8.36 (s, 2H). MS
(Mwt.: 387.48): m/z 387 (Mþ, 100%), 372 (8%), 253 (80%), 134
(95%).
A commercially available compound 51 (10.0 g, 33.5 mmol) was
dissolved in ethanol (150.0 mL), then urea (10.0 g, 167.0 mmol) and
sodium methoxide (22.7 g, 118.0 mmol) were added thereto and
the mixture was made to react for 24 h under the condition of
heating to reflux. After cooled to 0 ꢁC, crystals separated out
therefrom were filtered. The crystals were suspended in water,
hydrochloric acid (6.0 mol/L) was added thereto and pH was
adjusted to 6.0. Stirring was further conducted at room tempera-
ture for 1 h and the crystals separated out therefrom were filtered
and dried in vacuo to prepare compound 52 (6.5 g, 75%), which was
used without further purification. Rf ¼ 0.1 (EA:MeOH ¼ 5:1); Mp:
4.1.5. 2-(4-(Pyrimidin-2-yl)piperazin-1-yl)-5,6,7,8-
tetrahydropyrido[4,3-d]pyrimidine 56
55 (4.3 g, 11.1 mmol) was dissolved in MeOH (18.0 mL) and
dichloromethane (5 mL), treated with 10% Pd/C (50.0 mg), then
subjected to hydrogenation at atmospheric pressure for 16 h. The
mixture was filtered through Celite and washed with MeOH. The
combined filtrates were concentrated in vacuo. The residue was
triturated with Et2O, filtered, and then dried under vacuum to give
the title compound as an off-white solid (2.64 g, 80%). Mp: 163 ꢁC;
293 ꢁC; 1H NMR (300 MHz, CDCl3,
d ppm): 2.43e2.45 (m, 2H),
2.51e2.52 (m, 2H), 3.01 (s, 2H), 3.53 (s, 2H), 7.23e7.41 (m, 5H),
1H NMR (300 MHz, CDCl3,
d ppm): 2.51e2.52 (m, 2H), 2.70e2.72
10.21 (s, 1H), 11.01 (s, 1H).
(m, 2H), 3.01e3.04 (m, 2H), 3.61e3.82 (m, 8H), 6.64e6.67 (m, 1H),
8.10 (s, 1H), 8.40 (d, J ¼ 4.74, 2H). MS (Mwt.: 297.36): m/z 297 (Mþ,
15%), 163 (73%), 134 (100%).
4.1.2. 6-Benzyl-2,4-dichloro-5,6,7,8-tetrahydropyrido[4,3-d]
pyrimidine 53
52 (8.5 g, 32.9 mmol) was added to phosphoryl chloride
(50.0 mL, 535 mmol) in a 250 mL round-bottom flask equipped
with a stir bar, and the solution was refluxed for 3 h under N2. The
volatiles were removed under reduced pressure, and the concen-
trate was poured over 200 mL of ice. NaOH (3 M) was added to a
final pH of 10, and then the aqueous phase was extracted with
CH2Cl2 (3 ꢂ 100 mL). The combined organics were dried over
MgSO4, filtered, and concentrated to a tan oil. This crude material
was dissolved in dichloromethane, concentrated onto silica gel, and
subjected to column chromatography using a 100 g column, with a
gradient of 0%e50% ethyl acetate in hexanes. The product-
containing fractions were combined and evaporated under
reduced pressure to give 53 as a pale solid (8.4 g, 87%). Rf ¼ 0.5
4.1.6. (2,5-Dimethoxyphenyl)(2-(4-(pyrimidin-2-yl)piperazin-1-
yl)-7,8 dihydropyrido[4,3-d]pyrimidin-6(5H)-yl)methanone 57a
A mixture of 2,5-dimethoxybenzoic acid (151.0 mg, 0.83 mmol),
EDCI (192.0 mg, 1.0 mmol), HOBt (135.0 mg, 1.0 mmol) and 56
(300.0 mg, 1.0 mmol) in CH2Cl2 (10.0 mL) was stirred at room
temperature overnight. The organic layer was washed successively
with 2 M HCl and 2 M NaOH, then dried (MgSO4), filtered,
concentrated and purified by normal phase column chromatog-
raphy (0e100% ethyl acetate in hexanes) to afford 336.0 mg (88%) of
57a as a white solid. Rf ¼ 0.2 (PE:EA ¼ 1:1); Mp: 189e191 ꢁC; 1H
NMR (300 MHz, CDCl3, d ppm): 2.86 (s, 4H), 3.57 (s, 2H), 3.81 (s, 6H),
3.82e3.92 (m, 8H), 6.51e6.54 (m, 1H), 6.80e6.87 (m, 2H), 7.06 (s,
1H), 8.02 (s, 1H), 8.34e8.36 (m, 2H). 13C NMR (75 MHz, CDCl3,
(PE:EA ¼ 8:1); Mp: 273e275 ꢁC; 1H NMR (300 MHz, CDCl3,
d
ppm):
d ppm): 169.11, 165.05, 158.32 (2C), 155.15, 154.95, 149.12, 125.83,
2.23e2.24 (m, 2H), 2.41e2.43 (m, 2H), 2.97 (s, 2H), 3.53 (s, 2H),
7.53e7.61 (m, 5H).
123.52, 121.51, 116.15, 114.10, 111.34, 56.21, 56.23, 46.01 (4C), 44.41,
43.22, 35.64. IR (KBr):3342, 3125, 2954, 1629, 1543, 1399, 1319,
1282, 1246, 1178, 971, 875, 732 cmꢃ1. HRMS (ESI): calcd for
4.1.3. 6-Benzyl-2-chloro-5,6,7,8-tetrahydropyrido[4,3-d]pyrimidine
54
C
24H27N7O3 [M þ H]þ 462.2175, found 462.2151. Purity: 98.83% by
HPLC (MeOH/H2O ¼ 80:20).
To a solution of 53 (8.0 g, 27.3 mmol) in ethanol (150.0 mL) was
added zinc (14.2 g, 218.4 mmol) and ammonium hydroxide
(18.9 mL,136.5 mmol) and the reaction was heated to 78 ꢁC for 15 h.
The reaction was cooled, filtered through Celite, and washed with
ethyl acetate. The reaction was extracted with ethyl acetate
(3 ꢂ 200 mL) and the organic phase was dried over MgSO4. The
organic phase was concentrated and purified by normal phase
column chromatography (0e100% ethyl acetate in hexanes) to
4.1.7. (4-Fluorophenyl)(2-(4-(pyrimidin-2-yl)piperazin-1-yl)-7,8-
dihydropyrido[4,3-d]pyrimidin-6(5H)-yl)methanone 57b
Compound 57b (302.0 mg, 87% yield) was synthesized from 4-
fluorobenzoic acid (116 mg, 0.83 mmol) and 56 (300.0 mg,
1.0 mmol) according to the procedure used to synthesize 57a;
Rf ¼ 0.4 (PE:EA ¼ 1:1); Mp: 196e198 ꢁC; 1H NMR (300 MHz, CDCl3,
d
ppm): 2.87 (s, 2H), 3.71 (s, 2H), 3.92 (s, 8H), 4.61 (s, 2H), 6.53e6.56

H.-P. Sun et al. / European Journal of Medicinal Chemistry 79 (2014) 399e412
409
(m, 1H), 7.11e7.17 (m, 2H), 7.45e7.50 (m, 2H), 7.06 (s, 1H), 8.22 (s,
1H), 8.34e8.35 (m, 2H). 13C NMR (75 MHz, CDCl3,
ppm): 171.03,
4.1.11. (5-Chloro-2,4-dihydroxyphenyl)(2-(4-(pyrimidin-2-yl)
piperazin-1-yl)-7,8-dihydropyrido[4,3-d]pyrimidin-6(5H)-yl)
d
169.21, 164.90, 162.10, 161.98, 158.10 (2C), 155.23, 135.69, 131.26,
131.01, 124.01, 116.27, 115.78, 110.97, 45.98 (4C), 44.11, 43.55, 34.92.
IR (KBr): 3404, 2957, 2862, 1629, 1581, 1466, 1348, 1252, 1047, 922,
863, 793, 719 cmꢃ1. HRMS (ESI): calcd for C22H22FN7O [M þ H]þ
420.1870, found 420.1873. Purity: 98.74% by HPLC (MeOH/
H2O ¼ 80:20).
methanone 57f
Compound 57f with benzyl protection (451.0 mg, 84% yield) was
synthesized from 2,4-bis(benzyloxy)-5-chlorobenzoic acid
(305.0 mg, 0.83 mmol) and 56 (300.0 mg, 1.0 mmol) according to
the procedure used to synthesize 57a; A stirred solution of the
product (400.0 mg, 0.62 mmol) in CH2Cl2 (10.0 mL) at 0 ꢁC was
treated dropwisely with BBr3 in CH2Cl2 (1.86 mL, 1 M) and the
mixture was stirred for 3 h. The reaction was quenched by the
addition of water then extracted with EtOAc. The organic phase was
dried (MgSO4), filtered and the solvent evaporated in vacuo. The
crude material was purified by normal phase column chromatog-
raphy (10% MeOH in EtOAc) to afford 130.0 mg (45%) of 57f as a
white solid. Rf ¼ 0.4 (EA:MeOH ¼ 5:1); Mp: 296 ꢁC; 1H NMR
4.1.8. (4-Chlorophenyl)(2-(4-(pyrimidin-2-yl)piperazin-1-yl)-7,8-
dihydropyrido[4,3-d]pyrimidin-6(5H)-yl)methanone 57c
Compound 57c (307.0 mg, 85% yield) was synthesized from 4-
chlorobenzoic acid (129 mg, 0.83 mmol) and 56 (300.0 mg,
1.0 mmol) according to the procedure used to synthesize 57a;
Rf ¼ 0.4 (PE:EA ¼ 1:1); Mp: 219e222 ꢁC; 1H NMR (300 MHz, CDCl3,
(300 MHz, DMSO,
d ppm): 2.73e2.76 (m, 2H), 3.71 (s, 2H), 3.80 (s,
d
ppm): 2.86 (s, 2H), 3.65 (s, 2H), 3.92 (s, 8H), 4.51e4.71 (m, 2H),
8H), 4.52 (s, 2H), 6.35e6.37 (m, 1H), 6.65e6.68 (m, 1H), 7.11 (d,
J ¼ 8.1, 1H), 8.14 (s, 1H), 8.41 (d, J ¼ 5.1, 1H), 9.63 (s, 1H), 9.94 (s, 1H).
6.53e6.56 (m, 1H), 7.27e7.46 (m, 4H), 8.00e8.25 (m, 1H), 8.34e
8.37 (m, 2H). 13C NMR (75 MHz, CDCl3,
d
ppm): 169.13, 163.15,
13C NMR (75 MHz, DMSO,
d ppm): 171.01, 164.52, 162.23, 162.01,
161.24, 161.02, 156.35 (2C), 153.98, 138.17, 136.95, 130.05 (2C),
127.83 (2C), 123.79, 110.23, 46.01 (4C), 44.34, 43.67, 35.02. IR (KBr):
2998, 2909, 2843, 1629, 1586, 1506, 1422, 1351, 1256, 1178, 1048,
978, 845, 759 cmꢃ1. HRMS (ESI): calcd for C22H22ClN7O [M þ H]þ
436.1574, found 436.1580. Purity: 99.30% by HPLC (MeOH/
H2O ¼ 80:20).
157.87 (2C), 156.81, 156.34, 153.25, 131.41, 124.61, 116.94, 113.56,
110.20, 103.01, 45.91, 43.74, 42.50, 34.13. IR (KBr):3432, 2853, 1629,
1600, 1584, 1500, 1436, 1347, 1253, 1187, 1059, 979, 735 cmꢃ1. HRMS
(ESI): calcd for C22H23ClN7O3 [M þ H]þ 468.1473, found 468.1450.
Purity: 98.71% by HPLC (MeOH/H2O ¼ 80:20).
4.1.12. (2-Amino-7,8-dihydropyrido[4,3-d]pyrimidin-6(5H)-yl)(5-
chloro-2,4-dihydroxyphenyl)methanone 57g
4.1.9. (2,4-Dimethoxyphenyl)(2-(4-(pyrimidin-2-yl)piperazin-1-
yl)-7,8-dihydropyrido[4,3-d]pyrimidin-6(5H)-yl)methanone 57d
Compound 57d (333.0 mg, 87% yield) was synthesized from 4-
chlorobenzoic acid (151.0 mg, 0.83 mmol) and 71 (300.0 mg,
1.0 mmol) according to the procedure used to synthesize 57a;
Rf ¼ 0.2 (PE:EA ¼ 1:1); Mp: 226e228 ꢁC; 1H NMR (300 MHz, CDCl3,
N-tert-butoxycarbonyl-4-piperidone (58, 5.8 g, 31.4 mmol) was
dissolved in N,N-dimethylformamide dimethyl acetal (45.0 mL),
and the solution was heated under reflux for 1.5 h and concen-
trated. The residue was triturated with hexane, filtered, and
washed with hexane to give 59 as a yellow powder (5.1 g, 63.8%):
mp 135e136 ꢁC; To a solution of 59 (5.0 g, 20.8 mmol) in EtOH
(200.0 mL) were added guanidine carbonate (15.0 g, 84.0 mmol)
and sodium acetate (13.7 g, 167.0 mmol), and the solution was
heated under reflux for 48 h. The reaction mixture was filtered, and
the insoluble material was extracted with CHCl3 and washed with
water. The organic layer was dried over anhydrous MgSO4 and
evaporated. The resultant solid was triturated with 2-propanol,
filtered, and washed with 2-propanol and Et2O to give a colorless
powder. It was dissolved in TFA (50.0 mL) at 0 ꢁC, and the solution
was stirred at room temperature for 1 h and concentrated. The
residue was dissolved in 2-propanol and treated with concentrated
HCl (4.0 mL). The precipitated solid was filtered and washed with 2-
propanol and Et2O to give 60a (4.2 g, 81.6%) as a colorless powder:
Mp 258e260 ꢁC; Compound 57g was obtained from 60a in the
same way as 57f. A white powder: yield 0.73 g (43%); Rf ¼ 0.3
(EA:MeOH ¼ 5:1); Mp: 281e282 ꢁC; 1H NMR (300 MHz, DMSO,
d
ppm): 2.72e2.88 (m, 2H), 3.51e3.56 (m, 2H), 3.80 (s, 6H), 3.82e
3.89 (m, 8H), 4.51e4.78 (m, 2H), 6.45e6.52 (m, 3H), 7.18e7.26 (m,
1H), 8.12 (s, 1H), 8.29e8.34 (m, 2H). 13C NMR (75 MHz, CDCl3,
d
ppm): 170.13, 165.33, 164.86, 162.02, 161.89, 161.17, 158.12 (2C),
154.01, 127.93, 125.98, 121.01, 109.83, 105.05, 98.54, 56.03, 57.98,
45.88 (4C), 43.65, 42.12, 33.98. IR (KBr):3430, 2906, 1630, 1586,
1507, 1419, 1258, 1084, 978, 839, 796, 757 cmꢃ1. HRMS (ESI): calcd
for C24H27N7O3 [M þ H]þ 462.2175, found 462.2179. Purity: 99.23%
by HPLC (MeOH/H2O ¼ 80:20).
4.1.10. (2,4-Dihydroxyphenyl)(2-(4-(pyrimidin-2-yl)piperazin-1-
yl)-7,8-dihydropyrido[4,3-d]pyrimidin-6(5H)-yl)methanone 57e
Compound 57e with benzyl protection(425.0 mg, 83% yield)
was synthesized from 2,4-bis(benzyloxy)benzoic acid [add 3]
(277.0 mg, 0.83 mmol) and 56 (300.0 mg, 1.0 mmol) according to
the procedure used to synthesize 57a; the product (400.0 mg,
0.65 mmol) was dissolved in MeOH (8.0 mL) and CH2Cl2 (3.0 mL),
treated with 10% Pd/C (30.0 mg) then subjected to hydrogenation
at atmospheric pressure for 2 h. The mixture was filtered through
Celite and washed with MeOH. The combined filtrates were
concentrated in vacuo. The residue was triturated with Et2O,
filtered and then dried under vacuum to give the title compound as
an off-white solid (225.0 mg, 80%); Rf ¼ 0.35 (EA:MeOH ¼ 5:1);
d
ppm): 2.83e2.87 (m, 2H), 3.75e3.77 (m, 2H), 4.58e4.60 (m, 2H),
6.37 (s, 1H), 6.81e6.83 (m, 2H), 7.34 (s, 1H), 8.49 (s, 1H), 9.69e9.71
(d, 2H). 13C NMR (75 MHz, DMSO,
ppm): 171.42, 163.17, 161.02,
d
157.16, 156.32, 153.66, 133.72, 118.01, 117.54, 113.26, 102.18, 43.44,
42.54, 33.15. IR (KBr): 3152, 2360,1609,1565,1477,1258,1198,1049,
948, 839, 769 cmꢃ1. HRMS (ESI): calcd for C14H13ClN4O3 [M þ H]þ
321.0676, found 321.0685. Purity: 98.01% by HPLC (MeOH/
H2O ¼ 80:20).
Mp: 290e291 ꢁC; 1H NMR (300 MHz, DMSO,
d ppm): 2.71e2.75 (m,
2H), 3.67 (s, 2H), 3.79 (s, 8H), 4.50 (s, 2H), 6.25e6.32 (m, 2H), 6.64e
6.67 (m, 1H), 6.98 (d, J ¼ 8.1, 1H), 8.24 (s, 1H), 8.38 (d, J ¼ 4.5, 1H),
4.1.13. (2-Methyl-7,8-dihydropyrido[4,3-d]pyrimidin-6(5H)-yl)(5-
chloro-2,4-dihydroxyphenyl)methanone 57h
9.59 (s, 1H), 9.74 (s, 1H). 13C NMR (75 MHz, DMSO,
d ppm): 169.18,
164.72, 161.40, 161.19, 160.29, 157.18 (2C), 155.29, 154.24, 128.80,
117.76, 114.32, 109.54, 104.51, 98.09, 43.08 (4C), 38.72, 32.05, 31.16.
IR (KBr):3431, 3156, 2993, 2866, 1626, 1599, 1486, 1436, 1347, 1251,
1166, 1032, 975, 835, 789 cmꢃ1. HRMS (ESI): calcd for C22H23N7O3
[M þ H]þ 434.1862, found 434.1872. Purity: 98.46% by HPLC
(MeOH/H2O ¼ 80:20).
Compound 57h was obtained from 60b in the same way as 57g.
A white powder: yield 0.84 g (51%); Rf ¼ 0.3 (EA:MeOH ¼ 10:1); Mp:
284e285 ꢁC; 1H NMR (300 MHz, DMSO,
d ppm): 2.53 (s, 3H), 2.81e
2.85 (m, 2H), 3.66e3.69 (m, 2H), 4.61 (s, 2H), 6.35 (s, 1H), 6.87 (s,
1H), 8.50 (s, 1H), 9.49 (s, 1H), 9.52 (s, 1H). 13C NMR (75 MHz, DMSO,
d
ppm): 169.22, 161.20, 160.32, 157.21, 155.02 (2C), 133.44, 124.99,

410
H.-P. Sun et al. / European Journal of Medicinal Chemistry 79 (2014) 399e412
117.98, 113.82, 104.51, 43.10, 41.16, 31.50, 27.99. IR (KBr): 3127, 2923,
minimization of the hit compounds (convergence criterion of
0.0005 kcal/mol/M, energy minimizations in 20,000 steps) in
CHARMm force field. The molecular docking was carried out using
GOLD 5.1. GOLD uses a powerful genetic algorithm (GA) method for
conformation search and docking. As one of the most successful
and widely used docking programs, GOLD was tested on a data set
of over 300 complexes extracted from the Protein Data Bank (PDB)
and succeeded in more than 70% cases in reproducing the bound
conformation of the ligand obtained from experiment. In the pre-
sent study, the Hsp90-ligand complex (PDB id: 2XJX) was down-
load from PDB for docking studies. Residues around the original
2361, 1600, 1438, 1404, 1262, 1198, 1122, 1052, 908, 776, 737 cmꢃ1
.
HRMS (ESI): calcd for C15H14ClN3O3 [M þ H]þ 320.0724, found
320.0732. Purity: 97.36% by HPLC (MeOH/H2O ¼ 80:20).
4.2. Computational method
4.2.1. Preparation of the active set, decoy set and the Topscience
database
The Topscience database was downloaded from the official
website (www.tsbiochem.com). Multiple conformations of the
compounds in the active set, decoy set and the Topscience database
were generated by using OMEGA (Open Eye Scientific Software)
with the following parameters: number of allowed conformations
(nconfs) ¼ 400, Ewindow ¼ 10 kcal/mol and root-mean-square
ꢀ
ligand (radius 7.5 A) that completely covered the ATP binding site
were defined as the binding site. Docking studies were performed
using the standard default settings with 100 GA runs of molecules.
For each GA run, a maximum of 125,000 operations were per-
formed. Considering ligand flexibility, special care has been taken
by including options such as flipping of ring corners, pyramidal
nitrogens, amides, secondary and tertiary amines, rotation of
carboxylate groups, as well as torsion angle distribution and post-
process rotatable bonds as default. The annealing parameters
were used as default. Hydrophobic fitting points were calculated to
facilitate the correct starting orientation of the compound for
docking by placing the hydrophobic atoms appropriately in the
ꢀ
distance (RMS) ¼ 0.5 A. The Merck Molecular Force Field 94
(MMFF94) was used during the conformation generation. Ewindow
is the value used to discard high-energy conformations. The
maximum allowed conformations per compound was set to 400 to
ensure complete conformational coverage.
4.2.2. Creation of the ROCS model and the validation of the model
ROCS model was generated on the basis of the bound confor-
mation of AT-13387, which was directly separated from the coc-
rystal structure downloaded from protein data bank (PDB, ID:
2XJX). The algorithm is based on the idea that the molecular shape
of compounds is similar if the molecules overlay well and any
volume mismatch is resulting from shape dissimilarity. ROCS
maximize the rigid body overlap of the molecular Gaussian func-
tions and therefore the shared volume between a query molecule
and a conformation of a database molecule. For the superimposi-
tion of molecules, a smooth Gaussian function is used to represent
the molecular volume. Subsequently, the overlay of the molecules
is corrected by matching of chemical functionalities described by
“Color Force Field”, which can be used to measure chemical
complementarity, and to refine shape based superpositions based
on chemical similarity. The CFF is composed of SMARTS rules that
determine chemical centers plus rules to determine how such
centers interact. The color force fields in ROCS define six similar
TYPE color force-fields. The types are hydrogen-bond donors,
hydrogen-bond acceptors, hydrophobes, anions, cations, and rings.
Finally, ROCS combines the volume matching and features overlay
to describe the shape similarity. The angles of the key bonds in AT-
13387 were calculated in Discovery Studio.
ꢀ
corresponding areas of the active site. Cutoff values of 3.0 A for
ꢀ
hydrogen bonds and 4.0 A for van der Waals interactions were set.
The docking was terminated when top ten solutions attained root-
ꢀ
mean-square deviation (RMSD) values within 1.5 A, The imple-
mented scoring function Goldscore and Chemscore were used to
consider the intra- and intermolecular hydrogen bonding interac-
tion energy, van der Waals energy and ligand torsion energy.
4.3. Biology
4.3.1. Malachite green assay
The assay procedures were based on the literature [50]. The test
compounds were diluted from mother plates (10 mM in 100% (v/v)
DMSO) into daughter plates (200
test compound solution was added to each well (equivalent to a
final concentration of 40 M) of the 96-well assay plate. The first
mM in 2.0% (v/v) DMSO); 5 mL of
m
and last rows of the 96-well plate contained the appropriate con-
centration of DMSO were used as blank control. ATP was dissolved
in the assay buffer to give a stock concentration of 2.5 mM and
stored at room temperature. A 10 mL aliquot of ATP solution was
added to each well to give a final assay concentration of 1 mM.
Before the usage, Hsp90 protein was thawed on ice and suspended
in chilled assay buffer to a stock concentration of 0.45 mg/mL, and
the solution was kept on ice. The incubation was started by adding
4.2.3. Virtual screening of the topscience database
Virtual screenings were then performed based on the query
model of AT-13387 using ROCS. The parameters for the ROCS run
were set as follow: rank by ¼ combo and shape tanimoto,
besthits ¼ 1. During the screening, ROCS compares database com-
pounds and AT-13387 by aligning the compounds and calculating
the similarities including their volumes and chemical features. The
similarity is represented by a combo score, ranging from 0 to 2. If
the combo score is close to 2, then the molecules have an excellent
shape and chemical-feature match, while values close to 0 imply
poor shape and chemical-feature similarities. Finally, 50 com-
pounds were retained and purchased from Topscience database
with purity > 95% (Liquid chromatographyemass spectrometry,
LCeMS).
10
mL of the stock Hsp90 to each well (except for the background
wells which received 10
mL of assay buffer), giving a final assay
volume of 25 mL. The plates were shaken for approximately 2 min
and incubated for 3 h at 37 ꢁC.
To stop the ATPase reactions, 10
mL of the Malachite green re-
agent A (Sciencell, 8118) was added to each well. Following the
addition of 10 L of Malachite green reagent B (Sciencell, 8118) to
m
each well, the plate was incubated at room temperature for about
15 min, and the absorbance at 620 nm was measured by Varioskan
Multimode Microplate Spectrophotometer (Thermo, Waltham, MA,
USA).
4.3.2. The preparation of Hsp90 and measurement of ATP hydrolysis
inhibition
4.2.4. Molecular docking
The hit compounds from ROCS were saved as SD file and then
imported to Discovery Studio (DS) 3.0. They were firstly ionized and
prepared conformations by ‘Prepare Ligands’ protocol in DS at pH
7.0. Powell conjugate gradient algorithm was then applied for the
The region encoding full-length Hsp90 was subcloned into
pET28a. Escherichia coli cells with protein expression were induced
with 0.5 mM IPTG. Cells were incubated at 16 ꢁC for 20 h and then
were harvested and disrupted by sonication. After centrifugation of

H.-P. Sun et al. / European Journal of Medicinal Chemistry 79 (2014) 399e412
411
the soluble lysate, a Ni2þ-nitrilo-triacetic acid (NTA) agarose col-
umn in a buffer (50 mM Tris-Cl, 300 mM NaCl, 10 mM Imidazole,
10% [v/v] Glycerol, 10 mM PMSF, 10 mM DTT) was used for the
separation of Hsp90, which was eluted with a linear gradient of
20e1000 mM imidazole. Hsp90 was identified by SDS-PAGE, and
the high concentrated fraction was dialyzed against ATPase buffer
(20 mM Tris-Cl, pH 7.5; 6 mM MgCl2; 20 mM KCl). It was then
aliquoted, frozen in liquid nitrogen and stored at ꢃ80 ꢁC.
Luc and pRL-SV40 (a plasmid encoding Renilla luciferase) by using
lipofectamine and after 6 h the cells were treated with the indi-
cated concentrations of the various compounds for 24 h. Luciferase
activity was assessed by dual-luciferase reporter assay system
(Promega, E1910) using a luminometer (Thermo Scientific Lumi-
noskan Ascent). The level of HSP70 activation for a sample was
calculated as the ratio between firefly and Renilla luciferase activity
for the same sample. Typically, each sample was run in triplicate.
Compounds giving firefly/Renilla luciferase ratios >1-fold higher
than the corresponding ratios for the control samples (DMSO) were
selected for retesting in a doseeresponse format, using the same
assay. Results shown are from triplicate wells obtained in at least
three separate experiments. Preliminary experiments validated
that the reporter was strongly activated as expected by AT-13387
(w5-fold).
The Hsp90 ATP hydrolysis inhibitory effects of the compounds
were evaluated by using the Discover RX ADP HunterÔ Plus Assay
kit (Discoverx, Fremont, CA) as reported previously.53 Briefly, The
test compounds were diluted to 200 mM in 2.0% (v/v) DMSO; 5 mL
of test compound solution with different concentrations was added
to each well of the 96-well assay plate. The first and last rows of the
96-well plate contained the appropriate concentration of DMSO
were used as blank control. ATP was dissolved in the assay buffer to
give a stock concentration of 2.5 mM and stored at room temper-
4.3.6. Western-blot analysis
ature. 10 mL aliquot of prepared ATP solution was added to each well
SK-Br-3 cells were pretreated with various concentrations of the
test compounds. After stimulation, cells were collected; lysed in
lysis buffer [50 mM Tris-Cl, pH 7.6, 150 Mm NaCl, 1 mM EDTA, 1%
(m/v) Nonidet P-40 (NP-40), 0.2 mM Phenylmethanesulfonyl fluo-
ride (PMSF), 0.1 mM NaF and 1.0 mM dithiothreitol (DTT)], and the
supernatant was obtained after centrifugation at 13,000ꢂ g for
10 min at 4 ꢁC. The concentration of protein in the supernatants was
measured by the bicinchoninic acid (BCA) assay. Then equal
to give a final assay concentration of 1 mM. Hsp90 protein was
thawed on ice before usage and then suspended in chilled assay
buffer to a stock concentration of 0.45 mg/mL. The incubation was
started by adding 10
the background wells which received 10
to a final assay volume of 25
L and then incubated for 3 h at 37 ꢁC.
mL of the stock Hsp90 to each well (except for
mL of assay buffer), leading
m
The Discover RX ADP HunterÔ Plus Assay kit was used for the
detection of the ATP hydrolysis following the manufacturer’s in-
structions. The concentration of ADP was measured using Vari-
oskan spectrophotometer (Thermo, 540 nm excitation and 620 nm
emission). Fluorescence intensity values measured for Hsp90
without any testing compound was assumed as 100% of enzyme
activity. The background reaction rate was measured in a reaction
lacking enzyme or substrate and subtracted from the experimental
rates.
amounts of protein (50 mg) were separated by 8% or 10% sodium
dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and
transferred onto the PVDF membranes (Millipore, Billerica, MA).
The blots were incubated with specific antibodies against the
indicated primary antibodies overnight at 4 ꢁC followed by
IRDyeTM800-conjugated secondary antibody for 1 h at 37 ꢁC.
Detection was performed by the Odyssey Infrared Imaging System
(LI-COR; Lincoln, NE). All blots were stripped and incubated with
polyclonal anti-
proteins.
b-actin antibody to ascertain equal loading of
4.3.3. The fluorescence polarization (FP) competitive binding assay
Assays contained Hsp90/GA-FITC assay mix (100 mM Tris.Cl at
pH 7.4, 20 mM KCl, 6 mM MgCl2, 5 mg/ml BSA, 10 nM Hsp90, 10 nM
GA-FITC) and 2% DMSO as a vehicle for the compounds. Compounds
were added to 384 well plates (Corning #3575) as 10-fold stocks in
Acknowledgments
This work was supported by the grants from the National Nat-
ural Science Foundation of China (No. 81230078 (key program),
81202463 (youth foundation) and 91129732 (general program)),
the grant from the National High Technology Research and Devel-
opment Program of China (863 Program, No. 2012AA020301)
and the grants from National Major Science and Technology Project
of China (Innovation and Development of New Drugs, No.
2010ZX09401-401 and 2009ZX09501-003). The authors declare no
other conflicts of interest.
5
ml of 20% DMSO/80% assay mix, then 45 mL of assay mix was
added. After 30 min equilibration at room temperature in the dark,
fluorescence polarization was measured using a Synergy plate
reader (Molecular Devices SpectraMax Paradigm; Ex 485 nm, Em
535 nm). For this assay Z0 values were typically >0.8. IC50 were
calculated using Graphpad Prims 5.
4.3.4. Anti-proliferation activity
Cell viabilities were determined by a colorimetric assay using 3-
(4,5-dimethylthiaz-ol-2-yl)-2,5-
diphenyltetrazoliumbromide
Appendix A. Supplementary data
(MTT, Sigma, St. Louis, MO) as described previously [51]. Cells that
were given only culture media were used as control. After incu-
bation for 48 h, absorbance (A) was measured at 570 nm. Survival
ratio (%) was calculated as follow:
Supplementary data related to this article can be found at http://
dx.doi.org/10.1016/j.ejmech.2014.03.061.
Survival ratio ð%Þ ¼ ðAtreatment=AcontrolÞ ꢂ 100%:
References
IC50 was taken as the concentration that caused 50% inhibition
of cell viabilities and calculated by the Graphpad 5.0 software.
[1] C. Jolly, R.I. Morimoto, Journal of the National Cancer Institute 92 (2000)
1564e1572.
[2] J.D. West, Y. Wang, K.A. Morano, Chemical Research in Toxicology 25 (2012)
2036e2053.
[3] Y.M. Xu, D.Y. Huang, J.F. Chiu, A.T. Lau, Journal of Proteome Research 11 (2012)
2625e2634.
[4] Y.L. Janin, Drug Discovery Today 15 (2010) 342e353.
[5] A. Maloney, P. Workman, Expert Opinion on Biological Therapy 2 (2002) 3e24.
[6] L.H. Pearl, C. Prodromou, Annual Review of Biochemistry 75 (2006) 271e294.
[7] J. Trepel, M. Mollapour, G. Giaccone, L. Neckers, Nature Reviews Cancer 10
(2010) 537e549.
4.3.5. Hsp70 luciferase reporter assay of the hit compounds
HSPA1A promoter-driven luciferase plasmid was granted by
Professor Jianhua Liu [52] and the methods have been described
previously with little modification [53,54]. SK-BR-3 cells were
grown in 24 well plates in 1640 containing sodium pyruvate
0.11 mg/mL, 10% FCS. The cells were co-transfected with pHSPA1A-

412
H.-P. Sun et al. / European Journal of Medicinal Chemistry 79 (2014) 399e412
[8] H. Hao, Y. Naomoto, X. Bao, N. Watanabe, K. Sakurama, K. Noma, T. Motoki,
[32] S.R. Kasibhatla, K. Hong, M.A. Biamonte, D.J. Busch, P.L. Karjian,
J.L. Sensintaffar, A. Kamal, R.E. Lough, J. Brekken, K. Lundgren, R. Grecko,
G.A. Timony, Y. Ran, R. Mansfield, L.C. Fritz, E. Ulm, F.J. Burrows, M.F. Boehm,
Journal of Medicinal Chemistry 50 (2007) 2767e2778.
[33] P.A. Brough, W. Aherne, X. Barril, J. Borgognoni, K. Boxall, J.E. Cansfield,
K.M. Cheung, I. Collins, N.G. Davies, M.J. Drysdale, B. Dymock, S.A. Eccles,
H. Finch, A. Fink, A. Hayes, R. Howes, R.E. Hubbard, K. James, A.M. Jordan,
A. Lockie, V. Martins, A. Massey, T.P. Matthews, E. McDonald, C.J. Northfield,
L.H. Pearl, C. Prodromou, S. Ray, F.I. Raynaud, S.D. Roughley, S.Y. Sharp,
A. Surgenor, D.L. Walmsley, P. Webb, M. Wood, P. Workman, L. Wright,
Journal of Medicinal Chemistry 51 (2008) 196e218.
[34] A.J. Woodhead, H. Angove, M.G. Carr, G. Chessari, M. Congreve, J.E. Coyle,
J. Cosme, B. Graham, P.J. Day, R. Downham, L. Fazal, R. Feltell, E. Figueroa,
M. Frederickson, J. Lewis, R. McMenamin, C.W. Murray, M.A. O’Brien, L. Parra,
S. Patel, T. Phillips, D.C. Rees, S. Rich, D.M. Smith, G. Trewartha, M. Vinkovic,
B. Williams, A.J. Woolford, Journal of Medicinal Chemistry 53 (2010) 5956e
5969.
[35] W. Ying, Z. Du, L. Sun, K.P. Foley, D.A. Proia, R.K. Blackman, D. Zhou, T. Inoue,
N. Tatsuta, J. Sang, S. Ye, J. Acquaviva, L.S. Ogawa, Y. Wada, J. Barsoum, K. Koya,
Molecular Cancer Therapeutics 11 (2012) 475e484.
[36] T. Nakashima, T. Ishii, H. Tagaya, T. Seike, H. Nakagawa, Y. Kanda, S. Akinaga,
S. Soga, Y. Shiotsu, Clinical Cancer Research 16 (2010) 2792e2802.
[37] Combination of ligand- and structure-based methods in virtual screening,
Drug Discovery Today: Technologies 10 (3) (2013 Sep) e395e401, http://
dx.doi.org/10.1016/j.ddtec.2013.02.002.
[38] T. Scior, A. Bender, G. Tresadern, J.L. Medina-Franco, K. Martínez-Mayorga,
T. Langer, K. Cuanalo-Contreras, D.K. Agrafiotis, Recognizing pitfalls in virtual
screening: a critical review, Journal of Chemical Information and Modeling 52
(4) (2012 Apr 23) 867e881, http://dx.doi.org/10.1021/ci200528d. Epub 2012
Apr 6.
Y. Tomono, T. Fukazawa, Y. Shirakawa, T. Yamatsuji, J. Matsuoka, M. Takaoka,
Oncology Reports 23 (2010) 1483e1492.
[9] A. Kamal, M.F. Boehm, F.J. Burrows, Trends in Molecular Medicine 10 (2004)
283e290.
[10] S. Messaoudi, J.F. Peyrat, J.D. Brion, M. Alami, Anti-Cancer Agents in Medicinal
Chemistry 8 (2008) 761e782.
[11] W.B. Pratt, Y. Morishima, H.M. Peng, Y. Osawa, Experimental Biology and
Medicine (Maywood) 235 (2010) 278e289.
[12] J.S. Isaacs, W. Xu, L. Neckers, Cancer Cell 3 (2003) 213e217.
[13] D.J. Stravopodis, L.H. Margaritis, G.E. Voutsinas, Current Medicinal Chemistry
14 (2007) 3122e3138.
[14] L. Whitesell, S.L. Lindquist, Nature Reviews Cancer 5 (2005) 761e772.
[15] D. Picard, Cellular and Molecular Life Sciences 59 (2002) 1640e1648.
[16] A. Kamal, L. Thao, J. Sensintaffar, L. Zhang, M.F. Boehm, L.C. Fritz, F.J. Burrows,
Nature 425 (2003) 407e410.
[17] K. Moulick, J.H. Ahn, H. Zong, A. Rodina, L. Cerchietti, E.M. Gomes DaGama,
E. Caldas-Lopes, K. Beebe, F. Perna, K. Hatzi, L.P. Vu, X. Zhao, D. Zatorska,
T. Taldone, P. Smith-Jones, M. Alpaugh, S.S. Gross, N. Pillarsetty, T. Ku,
J.S. Lewis, S.M. Larson, R. Levine, H. Erdjument-Bromage, M.L. Guzman,
S.D. Nimer, A. Melnick, L. Neckers, G. Chiosis, Nature Chemical Biology 7
(2011) 818e826.
[18] J. Shi, R. Van de Water, K. Hong, R.B. Lamer, K.W. Weichert, C.M. Sandoval,
S.R. Kasibhatla, M.F. Boehm, J. Chao, K. Lundgren, N. Timple, R. Lough,
G. Ibanez, C. Boykin, F.J. Burrows, M.R. Kehry, T.J. Yun, E.K. Harning,
C. Ambrose, J. Thompson, S.A. Bixler, A. Dunah, P. Snodgrass-Belt, J. Arndt,
I.J. Enyedy, P. Li, V.S. Hong, A. McKenzie, M.A. Biamonte, Journal of Medicinal
Chemistry 55 (2012) 7786e7795.
[19] L. Neckers, Current Medicinal Chemistry 10 (2003) 733e739.
[20] P. Workman, Cancer Letters 206 (2004) 149e157.
[21] L. Whitesell, S.D. Shifrin, G. Schwab, L.M. Neckers, Cancer Research 52 (1992)
1721e1728.
[22] R.C. Schnur, M.L. Corman, R.J. Gallaschun, B.A. Cooper, M.F. Dee, J.L. Doty,
M.L. Muzzi, C.I. DiOrio, E.G. Barbacci, P.E. Miller, et al., Journal of Medicinal
Chemistry 38 (1995) 3813e3820.
[39] A. Lavecchia, C. Di Giovanni, Virtual screening strategies in drug discovery: a
critical review, Current Medicinal Chemistry 20 (23) (2013) 2839e2860.
[40] P. Purushottamachar, A. Khandelwal, P. Chopra, N. Maheshwari, L.K. Gediya,
T.S. Vasaitis, R.D. Bruno, O.O. Clement, V.C. Njar, Bioorganic & Medicinal
Chemistry 15 (2007) 3413e3421.
[23] E.R. Glaze, A.C. Smith, D.W. Johnson, D.L. McCormick, A.B. Brown, B.S. Levin,
R. Krishnaraj, A. Lyubimov, M.J. Egorin, J.E. Tomaszewski, Cancer Research 44
(2003), 162e162.
[24] J. Ge, E. Normant, J.R. Porter, J.A. Ali, M.S. Dembski, Y. Gao, A.T. Georges,
L. Grenier, R.H. Pak, J. Patterson, J.R. Sydor, T.T. Tibbitts, J.K. Tong, J. Adams,
V.J. Palombella, Journal of Medicinal Chemistry 49 (2006) 4606e4615.
[25] L.V. Sequist, S. Gettinger, N.N. Senzer, R.G. Martins, P.A. Janne, R. Lilenbaum,
J.E. Gray, A.J. Iafrate, R. Katayama, N. Hafeez, J. Sweeney, J.R. Walker, C. Fritz,
R.W. Ross, D. Grayzel, J.A. Engelman, D.R. Borger, G. Paez, R. Natale, Journal of
Clinical Oncology 28 (2010) 4953e4960.
[26] M.A. Biamonte, R. Van de Water, J.W. Arndt, R.H. Scannevin, D. Perret,
W.C. Lee, Journal of Medicinal Chemistry 53 (2010) 3e17.
[27] J.M. Holzbeierlein, A. Windsperger, G. Vielhauer, Current Oncology Reports 12
(2010) 95e101.
[28] H. He, D. Zatorska, J. Kim, J. Aguirre, L. Llauger, Y. She, N. Wu, R.M. Immormino,
D.T. Gewirth, G. Chiosis, Journal of Medicinal Chemistry 49 (2006) 381e390.
[29] (a) R.M. Immormino, Y. Kang, G. Chiosis, D.T. Gewirth, Journal of Medicinal
Chemistry 49 (2006) 4953e4960;
[41] E. Sala, L. Guasch, J. Iwaszkiewicz, M. Mulero, M.J. Salvado, C. Blade,
M. Ceballos, C. Valls, V. Zoete, A. Grosdidier, S. Garcia-Vallve, O. Michielin,
G. Pujadas, European Journal of Medicinal Chemistry 46 (2011) 6098e6103.
[42] ROCS, 3.1.2; OpenEye Scientific Software: Santa Fe, NM, 2012.
[43] J.A. Grant, M.A. Gallardo, B.T. Pickup, Journal of Computational Chemistry 17
(1996) 1653e1666.
[44] D. Schuster, M. Spetea, M. Music, S. Rief, M. Fink, J. Kirchmair, J. Schutz,
G. Wolber, T. Langer, H. Stuppner, H. Schmidhammer, J.M. Rollinger, Bio-
organic & Medicinal Chemistry 18 (2010) 5071e5080.
[45] P.C.D. Hawkins, A.G. Skillman, A. Nicholls, Journal of Medicinal Chemistry 50
(2006) 74e82.
[46] C.W. Murray, M.G. Carr, O. Callaghan, G. Chessari, M. Congreve, S. Cowan,
J.E. Coyle, R. Downham, E. Figueroa, M. Frederickson, B. Graham,
R. McMenamin, M.A. O’Brien, S. Patel, T.R. Phillips, G. Williams, A.J. Woodhead,
A.J.A. Woolford, Journal of Medicinal Chemistry 53 (2010) 5942e5955.
[47] E.B. Cogan, G.B. Birrell, O.H. Griffith, Analytical Biochemistry 271 (1999) 29e35.
[48] J. Rumsfeld, K. Ziegelbauer, F. Spaltmann, Protein Expression and Purification
18 (2000) 303e309.
(b) E. Caldas-Lopes, L. Cerchietti, J.H. Ahn, C.C. Clement, A.I. Robles, A. Rodina,
K. Moulick, T. Taldone, A. Gozman, Y. Guo, N. Wu, E. de Stanchina, J. White,
S.S. Gross, Y. Ma, L. Varticovski, A. Melnick, G. Chiosis, Proceedings of the
National Academy of Sciences of the United States of America 106 (2009)
8368e8373.
[49] S.M. Roe, C. Prodromou, R. O’Brien, J.E. Ladbury, P.W. Piper, L.H. Pearl, Journal
of Medicinal Chemistry 42 (1999) 260e266.
[50] M.G. Rowlands, Y.M. Newbatt, C. Prodromou, L.H. Pearl, P. Workman,
W. Aherne, Analytical Biochemistry 327 (2004) 176e183.
[51] J. Jia, X. Xu, F. Liu, X. Guo, M. Zhang, M. Lu, L. Xu, J. Wei, J. Zhu, S. Zhang,
S. Zhang, H. Sun, Q. You, PLoS One 8 (2013) e59315.
[52] Y. Zheng, V. Le, Z. Cheng, S. Xie, H. Li, J. Tian, J. Liu, Cell Stress & Chaperones 18
(2013) 203e213.
[30] R. Bao, C.J. Lai, H. Qu, D. Wang, L. Yin, B. Zifcak, R. Atoyan, J. Wang, M. Samson,
J. Forrester, S. DellaRocca, G.X. Xu, X. Tao, H.X. Zhai, X. Cai, C. Qian, Clinical
Cancer Research 15 (2009) 4046e4057.
[31] D. Papac, J.S. Patton, L. Reeves, L. DeMie, C. Bradford, B. Hachey,
B. Clemetson, C. Christensen, S.H. Kim, R. Tangallapally, D. Parker, R. Trovato,
I.C. Kim, D. Wettstein, V.R. Baichwal, A. Bajji, Cancer Research 71 (2011)
3237e3245.
[53] L. Xin, X. Li, H. Deng, D. Kuang, X. Dai, S. Huang, F. Wang, M. He, R.W. Currie,
T. Wu, Cell Stress & Chaperones 17 (2012) 567e576.
[54] B.R. Kusuma, L. Zhang, T. Sundstrom, L.B. Peterson, R.T. Dobrowsky, B.S. Blagg,
Journal of Medicinal Chemistry 55 (2012) 5797e5812.