Discovery of novel STAT3 small molecule inhibitors via in silico site-directed fragment-based drug design

Article abstract of DOI:10.1021/jm400080c

Constitutive activation of signal transducer and activator of transcription 3 (STAT3) has been validated as an attractive therapeutic target for cancer therapy. To stop both STAT3 activation and dimerization, a viable strategy is to design inhibitors blocking its SH2 domain phosphotyrosine binding site that is responsible for both actions. A new fragment-based drug design (FBDD) strategy, in silico site-directed FBDD, was applied in this study. A designed novel compound, 5,8-dioxo-6-(pyridin-3-ylamino)-5,8-dihydronaphthalene-1-sulfonamide (LY5), was confirmed to bind to STAT3 SH2 by fluorescence polarization assay. In addition, four out of the five chosen compounds have IC₅₀ values lower than 5 μM for the U2OS cancer cells. 8 (LY5) has an IC₅₀ range in 0.5-1.4 μM in various cancer cell lines. 8 also suppresses tumor growth in an in vivo mouse model. This study has demonstrated the utility of this approach and could be used to other drug targets in general.

Full text of DOI:10.1021/jm400080c

Article
pubs.acs.org/jmc
Discovery of Novel STAT3 Small Molecule Inhibitors via in Silico Site-
Directed Fragment-Based Drug Design
Wenying Yu,† Hui Xiao,‡ Jiayuh Lin,‡ and Chenglong Li*,†
†Division of Medicinal Chemistry and Pharmacognosy, College of Pharmacy, The Ohio State University, Columbus, Ohio 43210,
United States
‡Center for Childhood Cancer, The Research Institute at Nationwide Children’s Hospital, Department of Pediatrics, College of
Medicine, The Ohio State University, Columbus, Ohio 43205, United States
S
* Supporting Information
ABSTRACT: Constitutive activation of signal transducer and activator of
transcription 3 (STAT3) has been validated as an attractive therapeutic target
for cancer therapy. To stop both STAT3 activation and dimerization, a viable
strategy is to design inhibitors blocking its SH2 domain phosphotyrosine
binding site that is responsible for both actions. A new fragment-based drug
design (FBDD) strategy, in silico site-directed FBDD, was applied in this
study. A designed novel compound, 5,8-dioxo-6-(pyridin-3-ylamino)-5,8-
dihydronaphthalene-1-sulfonamide (LY5), was confirmed to bind to STAT3
SH2 by fluorescence polarization assay. In addition, four out of the five
chosen compounds have IC50 values lower than 5 μM for the U2OS cancer
cells. 8 (LY5) has an IC50 range in 0.5−1.4 μM in various cancer cell lines. 8
also suppresses tumor growth in an in vivo mouse model. This study has
demonstrated the utility of this approach and could be used to other drug
targets in general.
69 nM, as determined by the fluorescence polarization
method.2 (2) Peptidomimetics were also designed to target
the STAT3 SH2 domain. Peptidomimetics are derived from
phosphopeptides that mimic peptides but do not necessarily
contain phosphate groups. For example, XZH-5 was designed
using a structure-based approach to inhibit the formation of
STAT3 dimers.3,4 (3) Various small molecules have been
reported to inhibit STAT3 dimerization, making them more
druggable candidates. STA-21 discovered by structure-based
virtual screening was one of the first reported small inhibitors.
It inhibits STAT3 dimerization, DNA binding, and STAT3-
dependent luciferase reporter activity in breast cancer cells.5
Another small molecule, Stattic, was discovered by high-
throughput screening and has been shown to selectively inhibit
activation, dimerization, nuclear translocation of STAT3, and to
increase apoptosis in STAT3-dependent cancer cell lines.6
Among all the reported nonpeptidomimetic small inhibitors, 5-
hydroxy-9,10-dioxo-9,10-dihydroanthracene-1-sulfonamide
(LLL12) has the lowest IC50 (0.16−3.09 μM7), inhibiting
STAT3 phosphorylation and the growth of human cancer cells.
However, so far, there has been no STAT3-targeting drug
approved by the FDA. The search for more druggable STAT3
inhibitors with high potency and excellent bioavailability
remains extremely important.
INTRODUCTION
■
Constitutive activation of STAT3 has been found in a wide
variety of cancers, including breast cancer, sarcomas, and other
cancers, promoting it as a very attractive therapeutic target.
Cytokines, hormones, and growth factors binding to the cell
surface receptors can activate the JAK-STAT signaling pathway.
Activated receptors activate JAK kinase(s) and autophosphor-
ylate themselves. Subsequently, the STAT3 monomer is
activated through phosphorylation at its tyrosine705
(pTyr705) by the same kinases through STAT3 SH2 domain
binding to pTyr loop of the activated receptors, leading to
STAT3 homodimer formation through its SH2 domain
dimerization. The dimerized STAT3 then translocates into
the nucleus and binds to DNA, turning on a host of oncogenes.
Altogether, these events lead to cell proliferation, apoptosis
resistance, etc. To block both phosphorylation and dimerization
processes, STAT3 inhibitors should compete with the native
phosphotyrosine (pTyr705) loop by binding to the STAT3
SH2 domain (Figure 1).
Several series of STAT3 dimerization inhibitors have been
discovered via both computational and experimental methods.
(1) Phosphopeptide mimics were initially developed as STAT3
inhibitors to compete with the native phosphopeptide of the
STAT3 protein. For example, PM-73G is a phosphopeptide
mimic STAT3 inhibitor that can completely inhibit STAT3
Tyr705 phosphorylation at 0.5−1 μM level in various cancer
cell lines.1 Another phosphopeptide mimic, pCinn-Leu-cis-3,4-
methanoPro-Gln-NHBn, has the lowest reported IC50 value at
Received: January 17, 2013
© XXXX American Chemical Society
A
dx.doi.org/10.1021/jm400080c | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
Figure 1. Docking model of STAT3 inhibitor 8 binding to the STAT3 SH2 domain (PDB 1BG1), generated by AutoDock4 and viewed by Maestro.
The two key binding pockets are called “site pTyr705” (red circle) and “side pocket” (blue circle). Carbon atoms of 8 are colored purple, and those
of the pTyr peptide are colored green. The style of molecular surface is mesh, colored according to element property.
Figure 2. Site-directed FBDD strategy. Step 1: Fragment libraries are categorized from existing inhibitors according to their binding modes. Step 2:
New lead library is merged from randomly selected fragments from site-specific fragment sublibraries. (Any existing inhibitors are removed from the
new lead library.) Leads are selected after evaluating docking modes to the target. Step 3: Final hits are identified after synthesis and testing.
Table 1. Comparison of Conventional FBDD and in Silico Site-Directed FBDD
Fragment-based drug design (FBDD) has emerged as a new
strategy for drug discovery in the past decade. FBDD as a
design-intensive method complements the resource-intensive
conventional drug discovery methods such as high-throughput
screening and combinatorial chemistry.8 Conventional FBDD
has made some improvements on efficiency and cost-
effectiveness in drug design, and several drug candidates
designed by this methodology are currently under clinical trials.
Computational fragment-based drug design is another FBDD
approach. For example, Chen and Shoichet utilized the
B
dx.doi.org/10.1021/jm400080c | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
Figure 3. Docking modes of the selected known STAT3 inhibitors with the STAT3 SH2 domain (PDB 1BG1), generated by AutoDock4. Surface
representation was created with Maestro.
computational fragment-based approach and successfully
discovered potent de novo inhibitors against CTX-M, while
the docking screens of lead-like compound libraries failed to
identify any lead-like inhibitors.9
In this article, we present the concept of in silico site-directed
FBDD as a computational FBDD approach, summarized in
Figure 2.
To test the method, we have applied the approach to the case
of STAT3 inhibitor design. The STAT3 fragment library was
generated from nine known inhibitors with proven affinity and
established pharmacological activities for STAT3. New leads
were designed to target site pTyr705 and the side pocket
(Figure 1). The STAT3 fragment library was divided into two
specific sublibraries for the site pTyr705 and the side pocket
based on the docking poses of the inhibitors to the crystal
structure of STAT3 SH2 domain (PDB 1BG1).
To efficiently evolve the fragments into leads, several
considerations were made when choosing the linker and
performing the merging. A desirable linker should allow
sufficient flexibility for the fragments to maintain their poses
in the binding sites and enhance binding affinity and/or
biophysical features such as water solubility. Most importantly,
the chosen linker should not complicate synthesis. In this case,
a secondary amine was chosen as the linker. The merged
candidates were then screened via computational docking, and
the compounds with the most favorable docking energies and
well-clustered binding poses were selected for synthesis and
experimental testing. This led to a new class of STAT3
inhibitors.
With our in silico approach, fragment libraries are built from
known inhibitors and are divided into binding site specific
sublibraries according to docking poses. Linkers from the
known inhibitors are the first choices to join fragments,
however, new linkers based on the concept of bioisosterism10
may also be applied to join fragments. New linkers designed in
this fashion can often maintain the original binding interactions
or even enhance binding. With merging, the joining of
fragments and linkers is different in our computational
approach compared to conventional FBDD. Conventional
approaches randomly merge fragments together while our
presorted, site-specific fragment sublibraries are recombined to
maximize the possibility of obtaining novel compounds with
potentially better drugguabilty (e.g., potency, selectivity,
ADME/Tox properties, and synthetic ease, etc). Merged
potential candidates may be quickly screened via computational
docking methods to further narrow the number of molecules
that will be selected for synthesis and testing. The comparisons
between conventional and in silico site-directed FBDD
methods are summarized in Table 1.
RESULTS
■
Design. Step 1: Fragment Libraries Were Categorized
from the Known STAT3 Dimerization Inhibitors Based on
Their Binding Modes. To build fragment libraries from the
C
dx.doi.org/10.1021/jm400080c | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
Figure 4. Chemical structures of the selected STAT3 dimerization inhibitors. According to the docking modes in Figure 3, the fragments binding to
the pTyr705 site are colored red and those binding to the side pocket are colored blue.
known STAT3 inhibitors,3,5−7,11−15 the compounds were
docked to the STAT3 SH2 domain (PDB 1BG1). The initial
docking results are shown in Figure 3. On the basis of the
binding modes, the fragments were divided into libraries
specific for each of the two binding sites: site pTyr705 and side
pocket (Figures 4 and 5).
Step 2: New lead Library Was Built by Linking Selected
Fragments from Different Fragment Sublibraries. On the
basis of the docking modes for known STAT3 inhibitors,
naphthalene-5,8-dione-1-sulfonamide (1) exhibited the most
interactions with site pTyr705 by forming two hydrogen bonds
with Arg609 and an additional hydrogen bond with Ser613.
Additionally, as part of 13 (LLL12), one of the most potent
small nonpeptidomimetic inhibitors of STAT3 dimerization, 1
is an attractive starting point for the synthesis of new STAT3
inhibitors. Therefore, 1 was selected as the component
targeting site pTyr705 and was randomly linked to fragments
from the side pocket sublibrary to form new potential
inhibitors.
The linker was designed based on the rationale described
previously, that the new linker could maintain or even enhance
the binding affinity of the two fragments and that the synthetic
strategy would be feasible. On the basis of the docking mode of
13 in Figure 3, the phenol ring of LLL12 is the original linker of
1. Given the difficulty of synthesis, the original linker was
simplified into an isopropyl group, which was subsequently
evolved into an isopropyl amine group based on bioisosterism.
To further reduce the synthetic difficulty, the isopropyl amine
was modified to a dimethyl amine linker as shown in Figure 6.
Figure 6. The evolution of linker design.
Finally, the dimethyl amine group was selected as a suitable
linker after docking results indicated that candidates with this
linker could form hydrogen bonds with the backbone of Ser636
in a similar fashion as the hydroxyl group of 13 (Figure 7).
Step 3: Final Leads for Further Synthesis and Tests Were
Selected by Repositioning the Compounds from Lead
Figure 5. Categorized fragment sublibraries, site pTyr705 (1−7) and
side pocket (a−g) (the fragment phosphotyrosine group was not
included in the fragment libraries due to its peptidomimetic property).
D
dx.doi.org/10.1021/jm400080c | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
Scheme 1. Preparation of Fragment 1
Scheme 2. Synthesis of Designed Inhibitors by
Regioselective Coupling Two Fragments Together (R Group
a
Is Described in Table 2 for Each Product)
Figure 7. The docking modes of 13 (LLL12) and 8 (LY5). Both
linkers, the hydroxyl group in 13 and secondary amine group in 8,
form hydrogen bonds with the Ser636.
Library to STAT3 SH2 domain. Docking the new lead library to
the STAT3 SH2 domain, compounds were ranked based on the
docking scores and clustering, and compounds that could not
reposition to the binding pockets were removed from
consideration. The compounds that were ultimately selected
for synthesis and testing are shown in Table 2.
Chemistry. Synthesis of Fragment 1. 1 was synthesized as
previously described7 with some modifications; procedures are
shown in Scheme 1. Naphthalenesulfonylchloride reacted with
ammonium hydroxide at room temperature and precipitated
white crystal naphthalenesulfonamide with high purity ready for
next step synthesis and 90.2% yield. Naphthalenesulfonamide
was oxidized by chromium trioxide in an acidic environment.
This reaction is highly time sensitive because the longer
reaction times result in more byproducts. To achieve the
optimal reaction time, the solvent can be preheated to reflux,
and then chromium trioxide added into the reaction system. 1
was not very stable during the column separation, so flash
column was used for purification.
a
Regents and conditions: (a) Cu(OAc)2·H2O, AcOH; (b) Cu-
(OAc)2·H2O, AcOH and H2O (1:10 v/v).
and terminated when 1 was consumed completely. This
reaction is a novel regioselective method which can selectively
react to position 6 of fragment 1. No regioisomers at position 7
were detected in the byproducts. The atom numbering of 1 is
shown in Scheme 2. The purification of STAT3 inhibitors is
also very interesting. As the bands of novel STAT3 inhibitors
were red or brown, the elution of products was quickly
monitored based on the visible bands. As reported in other
naphthoquinone coupling reactions, solvent has a large effect
Synthesis of Novel STAT3 Inhibitors. Fragment 1 and
aromatic amines were catalyzed by Cu(OAc)2·H2O in an acidic
environment (Scheme 2). Reaction system was heated to reflux
Table 2. Chemical Structures, Docking Scores, and IC50 of the Designed STAT3 Inhibitors
a
b
Docking scores were calculated by AutoDock4. IC50 was tested on sarcoma cell line U2OS.
E
dx.doi.org/10.1021/jm400080c | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
on the yield. Glacial acetic acid was the best choice for our
reaction. If water is added to the reaction, the yield will
decrease. However, if glacial acetic acid is the only solvent, the
major product will be cyclized compounds because the
sulfonamide group can attack the nearby 1-ketone and deplete
a water molecule to form the cyclized products such as 10, 11,
and 12. To keep the free sulfonamide group to maintain the
best docking pose, we had to use a mixed solvent system by
adding water to glacial acetic acid in a volume ratio 1:10. This
decreased the yield by approximate 20%.
Biological and Biochemical Assays. Cell Viability Assay.
Initially, IC50 values for all of the compounds were tested with
U2OS cell line. Then two most potent compounds, 8 (LY5)
and 11, were further tested together with 13 in two other RD2
and RH30 cell lines. As 13 had the lowest IC50 among the
known small inhibitors, it was selected as a control compound.
The results showed that 8 had lower IC50 than 13 in cell lines
RD2 and U2OS and similar IC50 to 13 in cell line RH30. 11
had a comparable IC50 as 13 in cell lines RD2 and U2OS
(Table 3).
Figure 9. Compound 8 inhibits constitutive STAT3 phosphorylation.
Ewing’s sarcoma cell line EW8 was treated with 8 (0.5−1 μM) for 8 h,
then whole-cell extracts were prepared and phospho-STAT3 was
detected by Western blot assay, revealing a decrease in STAT3
phosphorylation.
of the kinase ERK1/2. The inhibition of STAT3 phosphor-
ylation by 8 seems to be consistent with the induction of
apoptosis, as evidenced by the presence of cleaved caspase 3
(Figure 9).
Table 3. IC50 Values for Human Sarcoma Cells (RD2, RH30,
and U2OS) Inhibition by STAT3 Inhibitors, 13 (LLL12), 8,
and 11
Compound 8 Inhibits STAT3 Phosphorylation Induced by
IL-6. IL-6 can induce the activation of STAT3.16 The MCF-7
breast cancer cell line was used to determine the ability of 8 to
inhibit IL-6 induced STAT3 phosphorylation because MCF-7
cells do not express persistently phosphorylated STAT3. In
control experiments, IL-6 elevated STAT3 phosphorylation in
MCF-7 cells, while 8 and 11 blocked the stimulation, with
phosphorylated STAT3 abrogated at low concentrations
(Figure 10). The DAOY medulloblastoma cancer cell line
was used to further determine the ability of 8 to inhibit IL-6
induced STAT3 phosphorylation. In control experiments, IL-6
elevated STAT3 phosphorylation in DAOY cells, while 8
blocked the stimulation, with phosphorylated STAT3 abro-
gated at low concentrations (Figure 11).
Compound 8 Does Not Inhibit STAT1 Phosphorylation
Induced by IFN-γ. IFN-γ can induce the activation of STAT117
that acts as a tumor suppressor in certain occasions. The
selectivity of 8 for STAT3 inhibition was evident when
compared with that for STAT1. IFN-γ treatment elevated
STAT1 phosphorylation in DAOY cells, whereas 8 pretreat-
ment had no effect on the extent of STAT1 phosphorylation
(Figure 12).
IC50 values (μM) vs cell lines
compd
U2OS
RH30
RD2
13
8
1.00
0.52
1.39
0.47
0.55
1.14
2.85
1.39
2.56
11
Compound 8 Inhibits STAT3 Phosphorylation and Induces
Apoptosis in Human Sarcoma Cancer Cells. To examine the
inhibition of STAT3 phosphorylation, Western blot assays were
performed to detect the abundance of phosphorylated STAT3
(P-STAT3) after RH30 cells were treated with 11 (0.5−2.5
μM) or 8 (0.25−1 μM) for 16 h. STAT3 phosphorylation was
reduced in a dose-dependent manner, with 8 and 11 almost
completely inhibiting Tyr705 phosphorylation at 0.5 and 1 μM,
respectively (Figure 8). Since 8 was shown to be a more potent
inhibitor of STAT3 phosphorylation than 11 in the RH30 cell
line, it was further tested in EW8 sarcoma cells. Western blots
of EW8 cells treated with 8 (0.5−1.0 μM) for 8 h again reveals
a dose-dependent decrease in formation of P-STAT3 (Figure
9). The expression of total STAT3 was not changed in both
RH30 and EW8 cell lines, indicating that the decrease of P-
STAT3 was not due to a constitutional decrease of total
STAT3 expression. 8 was not found to inhibit phosphorylation
Compound 8 Suppresses Tumor Growth of Breast Cancer
in a Mouse Tumor Model. We performed in vivo studies to
Figure 8. 11 and 8 inhibit constitutive STAT3 phosphorylation. Human rhabdomyosarcoma cell line RH30 was treated with 11 (0.5−2.5 μM) and 8
(0.25−1 μM) for 16 h, then whole-cell extracts were prepared and phospho-STAT3 was detected by Western blot assay, revealing a decrease in
STAT3 phosphorylation.
F
dx.doi.org/10.1021/jm400080c | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
Figure 10. 11 and 8 inhibit IL-6 induced STAT3 phosphorylation in MCF-7 breast cancer cells. The MCF-7 cells were serum-starved overnight,
then left untreated or treated with 11 (0.5−2.5 μM) or 8 (0.25−1 μM) for 5 h, followed by stimulation by IL-6 (25 ng/mL). The cells were
harvested at 30 min and analyzed by Western blot assays.
which is comparable to the AutoDock4 computed ΔG value of
−7.3 kcal/mol.
DISCUSSION
■
Even though there have been numerous attempts on drug
design using computational fragment-based approach and
molecular hybridization, in this work, we introduce a new
approach that combines target-structure-based in silico docking,
site-directed fragment hybridization, and bioisosterism to
recombine fragments of known inhibitors to form new lead
molecules. The successful application of this method is detailed
in the design of a novel STAT3 inhibitor scaffold, leading to
compounds with low IC50 values and strong binding affinity to
STAT3 SH2 domain. This study has several significant findings:
(1) The use of fragment sublibraries based on the docking
modes of known inhibitors was an efficient method to design
potential new inhibitors. This method is potentially more
reliable than virtual screening of fragment libraries and less
costly than conventional methods like high-throughput screen-
ing. It is also distinct from the molecular hybridization,18 as in
our approach, fragments are linked together based on structural
information of a specific target and the linked new compound is
designed to bind to the same target. In addition, the final
docking step of the merged fragments acts as an extra filter to
ensure that key interactions are maintained or even optimized
prior to synthesis. (2) The secondary amine linker used in the
design is a bioisostere of the hydroxyl group in 13 that can
maintain both the binding poses of the fragments and the
binding affinity. (3) The synthesis of novel STAT3 inhibitors
presents a new highly regioselective coupling method. The
major product is connected at position 6 of the napthoquinone.
Compared to 13, the novel series of STAT3 inhibitors are
superior, with lower synthetic cost and larger space for further
modification. 13 has some disadvantages such as the need for
the expensive synthetic material 3-hydroxy-2-pyrone, time-
consuming purification to separate the regioisomers, and
difficulty in modifying the three-ring system. The novel
STAT3 inhibitors can overcome these shortcomings. The
synthetic materials for novel STAT3 inhibitors are about 10
times less costly, the reaction has much higher selectivity to
react to 6-naphthalene-5,8-dione-1-sulfonamide than to 7-
naphthalene-5,8-dione-1-sulfonamide, and it is very easy to
modify the novel STAT3 inhibitors to fit the side pocket by
alternating the amine reagents. (4) The fluorescence polar-
ization (FP) assay proved that 8 binds to STAT3 SH2 domain
with a Ki value of 2.5 μM. (5) For in vitro study, the IC50 of 8 is
superior to that of 13 in two different cancer cell lines, U2OS
and RD2, while 13 has the lowest IC50 among all the reported
Figure 11. 8 inhibits IL-6 induced STAT3 phosphorylation in DAOY
cancer cells. The DAOY cells were serum-starved overnight, then left
untreated or treated with 8 (1−5 μM) for 2 h, followed by stimulation
by IL-6 (50 ng/mL). The cells were harvested at 30 min and analyzed
by Western blot assays.
Figure 12. IFN-γ induced STAT1 phosphorylation in DAOY cancer
cells. The DAOY cells were serum-starved overnight, then left
untreated or treated with 8 (1−5 μM) for 2 h, followed by stimulation
by IFN-γ (50 ng/mL). The cells were harvested at 30 min and
analyzed by Western blot assays. The results show that 8 did not
inhibit STAT1 phosphorylation.
confirm the tumor suppression efficacy of 8 and its potential
usage for breast cancer treatment. MDA-MB-231 cells were
injected subcutaneously into nude mice for the induction of
xenografts. After a 21 days of treatment, the morphological data
showed that 8 suppressed xenograft growth and significantly (P
< 0.001) shrank the tumor sizes (Figure 13).
Fluorescence Polarization (FP) Assay. FP assay was used to
further prove the binding of 8 to the STAT3 SH2 domain.
Screening was performed at physiologically relevant temper-
ature 37 °C. Before each experiment, 8 was prepared freshly
from the stock solution. Compound 8 increased its background
fluorescence reading at 595 nm at high concentrations, so the
control solutions for each concentration were prepared by
mixing buffer, 8, and fluorescence-labeled peptides together.
Inhibition curves were fitted in Sigmaplot11.0 (Figure 14), and
Ki value is 2.5 μM using a mathematical model developed for
FP assay. So the experimental ΔG value of 8 is −7.9 kcal/mol,
G
dx.doi.org/10.1021/jm400080c | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
Figure 13. Compound 8 suppresses tumor growth of MDA-MB-231 breast cancer cells in mouse tumor model in vivo. Tumor growth was
determined by measuring length (L) and width (W) of the tumor every other day with a caliper. The tumor volume was calculated according to the
formula: tumor volume = 0.5236 × L × W2. The treatment lasted for 21 days. The results showed that 8 significantly suppresses the tumor growth
(P < 0.001).
shortcoming is that ligand efficiency (LE) is unlikely to increase
significantly during the fragment recombination unless the
binding modes of fragments can be greatly improved without
increasing their sizes. One might not expect deconstruction of a
potent lead to always derive high quality fragments if the
synergy between two nonadjacent binding interactions is lost
during the fragmentation. How reliably fragments can be
recombined to maintain synergy will need to be further
examined. In this work, potency does not improve a lot but the
new compound is better in several other aspects as described
above. In this sense, this approach is a useful option for certain
design purposes, be it on synthetic easiness, potency or
selectivity improvement, necessary drug property (ADMT/
Tox) alteration to move the compound forward in the drug
design pipeline, etc.
CONCLUSIONS
■
Figure 14. Compound 8 binds to STAT3 SH2 domain. Inhibition of
Recombination of two fragments from existing STAT3
activation and dimerization inhibitors through in silico site-
directed FBDD is an efficient and feasible method to design
novel potent STAT3 inhibitors. Linker selection was explored
so that the synthetic strategy was feasible, the designed binding
modes of the fragments were preserved, and the new linker
itself enhanced the binding affinity of the fragments. Four out
of five synthesized compounds have IC50 lower than 5 μM for
cancer cell line U2OS. The lead compound 8 has an IC50
between 0.5 and 1.4 μM in various cancer cell lines. The
fluorescence polarization (FP) assay validated the binding of 8
to STAT3 SH2 domain. The experimental ΔG value of 8 is
−7.9 kcal/mol, which is comparable to the AutoDock4
computed ΔG value of −7.3 kcal/mol. In addition, 8 has
been demonstrated to significantly decrease STAT3 phosphor-
ylation/activation, shows a clear inhibitory selectivity of P-
STAT3 over P-STAT1, and significantly suppresses tumor
growth in vivo. This in silico site-directed FBDD strategy could
be used for small molecule design to other drug targets.
binding of fluorescein-labeled phosphopeptides to the SH2 domains of
STAT3 by 8 at 37 °C was assayed by fluorescence polarization. Error
bars represent standard deviation.
nonpeptidomimetic small molecule inhibitors targeting SH2
domain previously. (6) For DAOY cell line, 8 reveals a clear
selectivity between IL-6 induced P-STAT3 and IFN-γ induced
P-STAT1, as shown in Figure 11 and 12. 7) For in vivo study, 8
also significantly suppresses tumor growth of breast cancer cells
in a mouse tumor model.
This method has the potential to be efficient in new lead
generation, but it requires a few known inhibitors for certain
targets. For targets with many known inhibitors, such as
kinases, in silico site-directed FBDD would likely be useful and
produce many new leads. However, this drug design method
could also be used for a target without known inhibitors, if its
homologous targets exist with sufficient known inhibitors.
Fragment libraries built from these inhibitors still could be used
to search for new inhibitors for the target protein. An additional
H
dx.doi.org/10.1021/jm400080c | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
eluting with CH2Cl2/EtOAc (8:1 v/v), followed by CH2Cl2/EtOAc
(2:1 v/v), then further purified by recrystallization. Bright-red crystals
(89 mg, yield: 27%); mp (>260 °C). 1H NMR (300 MHz, DMSO) δ
9.52 (s, 1H), 8.63 (s, 1H), 8.44 (d, J = 7.9 Hz, 2H), 8.35 (d, J = 7.5
Hz, 1H), 7.96 (t, J = 7.7 Hz, 1H), 7.84 (d, J = 7.7 Hz, 1H), 7.48 (s,
1H), 7.35 (s, 2H), 6.13 (s, 1H). HRMS (ESI) of C15H11N3O4SNa [M
+ Na]+ calcd, 352.0362; found, 352.0358.
5,8-Dioxo-6-(phenylamino)-5,8-dihydronaphthalene-1-sulfona-
mide (9). The reactant amine is aniline (0.11 mL, 1.2 mmol). Then the
product was purified by silica gel column chromatography eluting with
CH2Cl2, followed by CH2Cl21/EtOAc (8:1 v/v). Red crystals (85 mg,
yield: 26%); mp (>260 °C). H NMR (300 MHz, DMSO) δ 9.46 (s,
1H), 8.48−8.31 (m, 2H), 7.95 (t, J = 6.9 Hz, 1H), 7.61−7.21 (m, 7H),
6.14 (s, 1H). HRMS (ESI) of C16H12N2O4SNa [M + Na]+ calcd,
351.0415; found, 351.0423.
EXPERIMENTAL SECTION
■
FBDD. General procedure of docking. Computational docking
program AutoDock4 was used to dock all the existing inhibitors and
our designed small molecules to predict their binding modes and
approximate binding free energies to the STAT3 SH2 dimerization
site.19 Compounds were docked using the Lamarckian Genetic
Algorithm. The docking procedure involved the preparation of the
ligand and macromolecule using the Schrodinger software.20,21
AutoDockTools was used to assign Gasteiger charges to the ligands.
AutoGrid maps were then precomputed for all atom types in the
ligand set. After 10 million energy evaluations were completed, all the
resulting conformations of the ligands in the binding pocket of the
macromolecule were clustered into groups according to their
conformations with a root-mean-square deviation threshold of 1.5 Å.
The most significant low energy clusters were identified and binding
energies were evaluated.
Chemistry. Chemicals and Reagents. The solvents and reagents
used in the present study were purchased from commercial suppliers
and were used as received. Thin layer chromatography was performed
with fluorescent silica coated aluminum sheets. Silica gel was
purchased from Sigma-Aldrich Chemical Co. (Milwaukee,WI). The
purities of all tested compounds are higher than 95% by HPLC, which
were performed by analytical HPLC. Analytical HPLC was carried out
with a Gemini 5 μ C18 110A column (250 mm × 4.6 mm) supplied by
Phenomenex Inc. CA, USA. Two different gradients [water (0.1%
TFA)/acetonitrile (0.1% TFA) and water (0.1% TFA)/methanol
(0.1% TFA)] were used at 1 mL/min flow rate (method: 100:0 to
0:100 over 20 min followed by 0:100 to 100:0 over 10 min). Melting
points (mp) were determined on a Thomas−Hoover capillary melting
point apparatus. Proton nuclear magnetic resonance spectra were
obtained with a Bruker Avance 300 (300 MHz) spectrophotometer
(Billerica, MA).
Naphthalene-5,8-dione-1-sulfonamide (1). Naphthalenesulfonyl-
chloride (16.8 g, 74.3 mmol) was dissolved in acetone (100 mL) and
was stirred at 0 °C for 5 min. Ammonium hydroxide (100 mL) was
dropped into the above mixture and stirred at room temperature for 3
h. Precipitated white crystals were filtered then acetone was removed
at reduced pressure. The residue was washed in ethylacetate (3 × 10
mL), producing a white solid powder which was used without further
purification, yielding naphthalenesulfonamide (15.9 g, 90.2%); mp
(147−149 °C). Naphthalenesulfonamide (500 mg, 2.41 mmol) was
dissolved in slowly warming glacial acetic acid (5.0 mL). The mixture
was heated to 90 °C and chromium trioxide (1.08 g, 10.85 mmol),
which was dissolved in a mixture of water and glacial acetic acid (1:1
v/v, 3 mL), was added to the mixture solution. The above solution was
stirred under reflux for 18 min. The solution was cooled to 0 °C, and
water (50 mL) was added and stirred overnight at room temperature.
The precipitated yellow powder was filtered, and the remaining
solution was extracted with ether (3 × 100 mL). The organic layer was
collected, dried, and removed at reduced pressure. The yellow powder
was combined and purified with silica column chromatography ethyl
acetate−hexane (2:3 v/v) to yield 1 (110 mg, 19.3%); mp (186−188
°C).
5H-Naphth[1,8-cd]isothiazol-5-one,1,1-dioxide,6-(phenylamino)
(10). The reactant amine is aniline (0.11 mL, 1.2 mmol). The product
was purified by silica gel column chromatography eluting with CH2Cl2,
then further purified by recrystallization. Red crystals (143 mg, yield:
1
46%); mp (>260 °C). H NMR (300 MHz, DMSO) δ 9.78 (s, 1H),
8.39 (d, J = 7.1 Hz, 1H), 8.04 (q, J = 7.3 Hz, 2H), 7.55−7.33 (m, 4H),
7.25 (t, J = 6.9 Hz, 1H), 6.14 (s, 1H). HRMS (ESI) of
C16H10N2O3SNa [M + Na]+ calcd, 333.0304; found, 333.0329.
5H-Naphth[1,8-cd]isothiazol-5-one,1,1−dioxide,6-(1′-chloro-3′-
nitro-2′-phenylamino) (11). The reactant amine is 2-chloro-4-
nitroaniline (206 mg, 1.2 mmol). The product was purified by silica
gel column chromatography eluting with CH2Cl2, followed by a
CH2Cl2/EtOAc (8:1 v/v), then further purified by recrystallization.
Orange crystals (189 mg, yield: 48.6%); mp (>260 °C). 1H NMR (300
MHz, DMSO) δ 10.19 (s, 1H), 8.51 (d, J = 2.0 Hz, 1H), 8.45−7.93
(m, 4H), 7.83 (d, J = 8.8 Hz, 1H), 6.15 (s, 1H). HRMS (ESI) of
C16H8ClN3O5SNa [M + Na]+ calcd, 411.9765; found, 411.9764.
5H-Naphth[1,8-cd]isothiazol-5-one,1,1−dioxide,6-(naphthylami-
no) (12). The reactant amine is 2-naphthylamine (172 mg, 1.2 mmol).
The product was purified by silica gel column chromatography eluting
with hexane/CH2Cl2 (1:4 v/v), followed by CH2Cl2. Red crystals (170
mg, yield: 47.3%), then further purified by recrystallization; mp (>260
°C). 1H NMR (300 MHz, DMSO) δ 10.00 (s, 1H), 8.44 (dd, J = 7.2,
1.3 Hz, 1H), 8.31−7.86 (m, 7H), 7.78−7.30 (m, 4H), 6.33 (s, 1H).
HRMS (ESI) of C20H14N2O4SNa [M + Na]+ calcd, 383.0466; found,
383.0453.
Biological Assays. Cell Lines. Human breast cancer cell line
(MCF-7), human sarcoma cell lines (RD2, RH30 and U2OS), and
human medulloblastoma cell line (DAOY) were maintained in
Dulbecco’s Modified Eagle Medium (DMEM) supplemented with
10% penicillin/streptomycin FBS and stored in a humidified 37 °C
incubator with 5% CO2.
STAT3 Inhibitors. 8, 11, and 13 were dissolved in sterile DMSO to
make 20 mM stock solutions. Aliquots of the stock solutions were
stored at −20 °C.
Cell Viability Assay. Human sarcoma cell lines (RD2, RH30,
U2OS) were seeded in 96-well plates at a density of 3000 cells per
well. The cells were incubated at 37 °C for a period of 24 h. Different
concentrations of 8 (0.1−10 μM), 11 (0.1−20 μM), and 13 (0.1−10
μM) were added in triplicate to the plates in the presence of 10% FBS.
3-(4, 5-Dimethylthiazolyl)-2, 5-diphenyltetrazoliumbromide (MTT)
was added to evaluate cell viability. The absorbance was read at 595
nm. IC50 values were determined using SigmaPlot 9.0 Software (Systat
Software, Inc., San Jose, CA).
Western Blot Analysis. RH30 cells were treated with 8 (0.25−1
μM) or 11 (0.5−2.5 μM) or DMSO at 60−80% confluence in the
presence of 10% FBS for 24 h. The sarcoma cell line EW8 was treated
with 8 (0.5−2.5 μM) or DMSO at 60−80% confluence in the presence
of 10% FBS for 24 h. MCF-7 cells and DAOY cells were treated with
DMSO, 25−50 ng/mL of IL-6, 8 (0.25−1 μM), or 11 (0.5−2.5 μM),
protein expressions of P-STAT3 (Tyr705) and STAT3 were tested.
DAOY cells were treated with DMSO, 50 ng/mL of IFN-γ, or 8 (1−5
μM), protein expressions of P-STAT3 (Tyr705) and STAT3 were
tested.
General Experimental Procedure for Scheme 2. Fragment 1 (237
mg,1 mmol), amine (1.2 mmol) and Cu(OAc)2·H2O (20 mg, 0.1
mmol) were solubilized by gently warming in AcOH (5 mL), refluxing
for about 3 h. As detected by TLC, after 1 had completely reacted, all
the volatiles were removed under reduced pressure. The resulting
crude product was dissolved in a minimal volume of CH2Cl2 and
applied to a column of silica gel. The column was eluted with CH2Cl2
or other solvents as indicated in the description of each product. The
solvent was removed under reduced pressure to give products (10−
12). Alternatively, 1 (237 mg, 1 mmol), amine (1.2 mmol), and
Cu(OAc)2·H2O (20 mg, 0.1 mmol) were solubilized by gently
warming in AcOH and H2O (1:10 v/v, 5.5 mL), refluxing for about 3
h. Following the same procedures, 8−12 were synthesized.
5,8-Dioxo-6-(pyridin-3-ylamino)-5,8-dihydronaphthalene-1-sul-
fonamide (8). The reactant amine is 3-amine-pyridine (113 mg, 1.2
mmol). The product was purified by silica gel column chromatography
I
dx.doi.org/10.1021/jm400080c | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
Western blot procedure: The cells were harvested and lysed in cold
radioimmunoprecipitation assay (RIPA) lysis buffer containing
proteasome inhibitor cocktail and phosphatase inhibitor cocktail, and
were subjected to SDS-PAGE. Then they were transferred to PVDF
membrane. Membranes were probed with primary antibodies
(1:1,000) and horseradish peroxidase (HRP)-conjugated secondary
antibody (1:10000) (Cell Signaling Technology, Beverly, MA).
Membranes were analyzed using Enhanced Chemiluminescence Plus
reagents and scanned with the Storm scanner (Amersham Pharmacia
Biotech, Inc., Piscataway, NJ).
IL-6 Induction of STAT3 Phosphorylation. MCF-7 breast cancer
cells were seeded in 10 cm plates and allowed to adhere overnight.
The following day, the cells were serum-starved. The cells were then
left untreated or were treated with 8 (0.25−1 μM) or 11 (0.5−2.5
μM) or DMSO. After 5 h, the untreated and 8 or 11 treated cells were
stimulated by IL-6 (25 ng/mL). The cells were harvested after 30 min
and analyzed by Western blot. DAOY medulloblastoma cancer cells
were seeded in 10 cm plates and allowed to adhere overnight. The
following day, the cells were serum-starved. The cells were then left
untreated or were treated with 8 (1−5 μM) or DMSO. After 2 h, the
untreated and 8 treated cells were stimulated by IL-6 (25 ng/mL). The
cells were harvested after 30 min and analyzed by Western blot.
IFN-γ Induction of STAT1 Phosphorylation. DAOY medulloblas-
toma cancer cells were seeded in 10 cm plates and allowed to adhere
overnight. The following day, the cells were serum-starved. The cells
were then left untreated or were treated with 8 (1−5 μM) or DMSO.
After 2 h, the untreated and 8 treated cells were stimulated by IFN-γ
(50 ng/mL). The cells were harvested after 30 min and analyzed by
Western blot.
Mouse Xenograft Tumor Model. MDA-MB-231 human breast
cancer cells (5 × 106) were injected subcutaneously into the flank area
of 6-week-old female athymic nude mice which were purchased from
Harlan (Indianapolis, IN, USA). After tumor development, mice were
divided into two treatment groups consisting of eight tumors per
group: DMSO vehicle control, IP injection of 8 (5 mg/kg). The
inhibitors were formulated with Cremaphor, DMSO, and 5% dextrose
to enhance delivery and limit toxicity encountered with DMSO alone
as the mixing base. Tumor growth was determined by measuring
length (L) and width (W) of the tumor every other day with a caliper.
The tumor volume was calculated according to the formula: Tumor
volume = 0.5236 × L × W2. The treatment lasted for 21 days.
Fluorescence Polarization (FP) Assay. FP assay was used to analyze
the ability of 8 to inhibit phosphopeptide binding to the STAT3 SH2
domain. STAT3 protein and peptides were provided by Dr. Pui-Kai
Li’s lab. Screening was performed at approximately 37 °C. STAT3
protein was used at 300 nM. The final concentration of buffer
components used for all FP assays was 10 mM HEPES (pH 7.5), 1
mM EDTA, 0.1% Nonidet P-40, 50 mM NaCl, and 20% DMSO.
Dithiothreitol was not added. 8 was diluted to a series of
concentrations (1−200 μM) from a 20 mM stock in DMSO. Proteins
were incubated with 8 in Eppendorf tubes at room temperature for 60
min prior addition of the 5-carboxyfluorescein labeled peptides. The
final concentration of labeled peptide is 10 nM. The mixtures were
transferred to a 384 Well Costar Black plate with each concentration
repeated in three wells. The mixtures were equilibrated in the
incubator at 37 °C for at least 30 min. The FP readings were
measured, and the experiments were repeated three times. Inhibition
curves with standard deviation were fitted, and Ki value was
determined accordingly.
AUTHOR INFORMATION
■
Corresponding Author
*Phone: 614-247-8786. Fax: 614-292-2435. E-mail: li.728@osu.
edu.
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
We are grateful to Dr. Pui-Kai Li and his graduate student,
Jonathan P. Etter, for providing the purified STAT3 protein
and their assistance in FP assays. We are grateful to Dr. Werner
Tjarks for analytic HPLC tests. We thank Ryan E. Pavlovicz
and Dr. Xiaohua Zhu for paper editing. We thank the Ohio
Supercomputer Center for computing resources. The study is
partially funded by the Alex’s Lemonade Stand Foundation and
NIH/NCI 5R21CA133652.
ABBREVIATIONS USED
■
HPLC, high performance liquid chromatography; DMSO,
dimethyl sulfoxide; HEPES, (4-(2-hydroxyethyl)-1-piperazinee-
thanesulfonic acid); EDTA, ethylenediaminetetraacetic acid;
TLC, thin layer chromatography; IL-6, interleukin 6; IFN-γ,
interferon γ
REFERENCES
■
(1) Mandal, P. K.; Gao, F.; Lu, Z.; Ren, Z.; Ramesh, R.; Birtwistle, J.
S.; Kaluarachchi, K. K.; Chen, X.; Bast, R. C.; Liao, W. S.; McMurray, J.
S. Potent and Selective Phosphopeptide Mimetic Prodrugs Targeted
to the Src Homology 2 (SH2) Domain of Signal Transducer and
Activator of Transcription 3. J. Med. Chem. 2011, 54, 3549−3563.
(2) Mandal, P. K.; Ren, Z. Y.; Chen, X. M.; Xiong, C. Y.; McMurray,
J. S. Structure−Affinity Relationships of Glutamine Mimics Incorpo-
rated into Phosphopeptides Targeted to the SH2 Domain of Signal
Transducer and Activator of Transcription 3. J. Med. Chem. 2009, 52,
6126−6141.
(3) Liu, A.; Liu, Y.; Xu, Z.; Yu, W.; Wang, H.; Li, C.; Lin, J. Novel
small molecule, XZH-5, inhibits constitutive and interleukin-6-induced
STAT3 phosphorylation in human rhabdomyosarcoma cells. Cancer
Sci. 2011, 102, 1381−1387.
(4) Liu, Y.; Liu, A.; Xu, Z.; Yu, W.; Wang, H.; Li, C.; Lin, J. XZH-5
inhibits STAT3 phosphorylation and causes apoptosis in human
hepatocellular carcinoma cells. Apoptosis 2011, 16, 502−510.
(5) Song, H.; Wang, R.; Wang, S.; Lin, J.; Stark, G. R. A Low-
Molecular-Weight Compound Discovered through Virtual Database
Screening Inhibits Stat3 Function in Breast Cancer Cells. Proc. Natl.
Acad. Sci. U. S. A. 2005, 102, 4700−4705.
(6) Schust, J.; Sperl, B.; Hollis, A.; Mayer, T. U.; Berg, T. Stattic: a
small-molecule inhibitor of STAT3 activation and dimerization. Chem.
Biol. 2006, 13, 1235−1242.
(7) Lin, L.; Hutzen, B.; Li, P.-K.; Ball, S.; Zuo, M.; DeAngelis, S.;
Foust, E.; Sobo, M.; Friedman, L.; Bhasin, D.; Cen, L.; Li, C.; Lin, J. A
novel small molecule, LLL12, inhibits STAT3 phosphorylation and
activities and exhibits potent growth suppressive activity in human
cancer cells. Neoplasia 2010, 12, 39−50.
(8) Hajduk, P. J.; Greer, J. A decade of fragment-based drug design:
strategic advances and lessons learned. Nature Rev. Drug Discovery
2007, 6, 211−219.
ASSOCIATED CONTENT
(9) Chen, Y.; Shoichet, B. K. Molecular docking and ligand specificity
in fragment-based inhibitor discovery. Nature Chem. Biol. 2009, 5,
358−364.
■
S
* Supporting Information
Data pertaining to the 1H and 13C NMR and HPLC
chromatograms of key compounds and compound 8 inhibitions
on OSM stimulated MCF-7 cell STAT1/3/5 phosplorylations
results. This material is available free of charge via the Internet
at http://pubs.acs.org.
(10) Lima, L. M.; Barreiro, E. J. Bioisosterism: a useful strategy for
molecular modification and drug design. Curr. Med. Chem. 2005, 12,
23−49.
(11) Bhasin, D.; Cisek, K.; Pandharkar, T.; Regan, N.; Li, C. L.;
Pandit, B.; Lin, J.; Li, P. K. Design, synthesis, and studies of small
J
dx.doi.org/10.1021/jm400080c | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
molecule STAT3 inhibitors. Bioorg. Med. Chem. Lett. 2008, 18, 391−
395.
(12) Matsuno, K.; Masuda, Y.; Uehara, Y.; Sato, H.; Muroya, A.;
Takahashi, O.; Yokotagawa, T.; Furuya, T.; Okawara, T.; Otsuka, M.;
Ogo, N.; Ashizawa, T.; Oshita, C.; Tai, S.; Ishii, H.; Akiyama, Y.; Asai,
A. Identification of a New Series of STAT3 Inhibitors by Virtual
Screening. ACS Med. Chem. Lett. 2010, 1, 371−375.
(13) Turkson, J.; Kim, J. S.; Zhang, S. M.; Yuan, J.; Huang, M.;
Glenn, M.; Haura, E.; Sebti, S.; Hamilton, A. D.; Jove, R. Novel
peptidomimetic inhibitors of signal transducer and activator of
transcription 3 dimerization and biological activity. Mol. Cancer Ther.
2004, 3, 261−269.
(14) Ren, X. M.; Duan, L.; He, Q. A.; Zhang, Z.; Zhou, Y.; Wu, D.
H.; Pan, J. X.; Pei, D. Q.; Ding, K. Identification of Niclosamide as a
New Small-Molecule Inhibitor of the STAT3 Signaling Pathway. ACS
Med. Chem. Lett. 1, 454−459.
(15) Siddiquee, K. A. Z.; Gunning, P. T.; Glenn, M.; Katt, W. P.;
Zhang, S.; Schroeck, C.; Sebti, S. M.; Jove, R.; Hamilton, A. D.;
Turkson, J. An oxazole-based small-molecule Stat3 inhibitor modulates
Stat3 stability and processing and induces antitumor cell effects. ACS
Chem. Biol. 2007, 2, 787−798.
(16) Faruqi, T. R.; Gomez, D.; Bustelo, X. R.; Bar-Sagi, D.; Reich, N.
C. Rac1 mediates STAT3 activation by autocrine IL-6. Proc. Natl.
Acad. Sci. U. S. A. 2001, 98, 9014−9019.
(17) Yu, H.; Pardoll, D.; Jove, R. STATs in cancer inflammation and
immunity: a leading role for STAT3. Nature Rev. Cancer 2009, 9, 798−
809.
(18) Viegas-Junior, C.; Danuello, A.; Bolzani, V. S.; Barreiro, E. J.;
Fraga, C. A. M. Molecular hybridization: a useful tool in the design of
new drug prototypes. Curr. Med. Chem. 2007, 14, 1829−1852.
(19) Huey, R.; Morris, G. M.; Olson, A. J.; Goodsell, D. S. A
semiempirical free energy force field with charge-based desolvation. J.
Comput. Chem. 2007, 28, 1145−1152.
(20) 2012: LigPrep Suite 2012, version 2.5; Schrodinger, LLC, New
̈
York, 2012.
(21) Schrodinger Suite 2011 Protein Preparation Wizard; Schrodinger,
̈
̈
LLC, New York, 2012; Epik , version 2.3; Schrodinger, LLC: New
̈
York, 2012; Impact, version 5.8; Schrodinger, LLC: New York, 2012;
Prime, version 3.1; Schrodinger, LLC: New York, 2012.
̈
̈
K
dx.doi.org/10.1021/jm400080c | J. Med. Chem. XXXX, XXX, XXX−XXX