De novo design of multitarget ligands with an iterative fragment-growing strategy

Article abstract of DOI:10.1021/ci500021v

The discovery of multitarget drugs has recently attracted much attention. Most of the reported multitarget ligands have been serendipitous discoveries. Although a few methods have been developed for rational multitarget drug discovery, there is a lack of

Full text of DOI:10.1021/ci500021v

Article
pubs.acs.org/jcim
De Novo Design of Multitarget Ligands with an Iterative Fragment-
Growing Strategy
Erchang Shang,† Yaxia Yuan,‡ Xinyi Chen,† Ying Liu,*,†,‡ Jianfeng Pei,*,‡ and Luhua Lai*,†,‡,§
†BNLMS, State Key Laboratory for Structural Chemistry of Unstable and Stable Species, College of Chemistry and Molecular
Engineering, Peking University, Beijing 100871, China
‡Center for Quantitative Biology, AAIS, Peking University, Beijing 100871, China
§Peking-Tsinghua Center for Life Sciences, Peking University, Beijing 100871, China
S
* Supporting Information
ABSTRACT: The discovery of multitarget drugs has recently attracted much attention. Most of the reported multitarget ligands
have been serendipitous discoveries. Although a few methods have been developed for rational multitarget drug discovery, there
is a lack of elegant methods for de novo multitarget drug design and optimization, especially for multiple targets with large
differences in their binding sites. In this paper, we report the first de novo multitarget ligand design method, with an iterative
fragment-growing strategy. Using this method, dual-target inhibitors for COX-2 and LTA4H were designed, with the most potent
one inhibiting PGE2 and LTB4 production in the human whole blood assay with IC50 values of 7.0 and 7.1 μM, respectively. Our
strategy is generally applicable in rational and efficient multitarget drug design, especially for the design of highly integrated
inhibitors for proteins with dissimilar binding pockets.
elegantly and simultaneously design and optimize molecules for
multiple targets.10
INTRODUCTION
■
Single-target drugs are often less effective in controlling
complex diseases with multiple pathogenic factors, such as
diabetes, inflammation, cancer, and CNS disorders.1−3 Multi-
target drugs, which are able to interact with several drug targets
simultaneously, lead to new and more effective medications for
a variety of complex diseases, even with relatively weak
activities.4,5 Consequently, much attention has been given to
multitarget drug design, and methods for multitarget ligand
discovery were developed accordingly.6,7 Among them, frame-
work combination and screening are two widely used
approaches to multitarget lead generation. However, framework
combination usually produces ligands with large molecular
weights and low ligand efficiencies with possible poor oral
pharmacokinetics. Moreover, the chance of screening success is
generally low, especially for unrelated targets where the
pharmacophores are distinct.8 Therefore, the efficient discovery
of “highly integrated” ligands for unrelated targets remains
challenging, and a general strategy for multitarget rational drug
design for dissimilar targets needs to be developed.
Cyclooxygenase (COX)/prostaglandin E2 synthases (PGES)
and 5-lipoxygenase (5-LOX)/leukotriene A4 hydrolase
(LTA4H) are the two major metabolic pathways that produce
inflammatory mediators from arachidonic acid (AA).11,12
Nonsteroidal anti-inflammatory drugs (NSAIDs) are widely
used in anti-inflammatory therapy. However, commonly used
NSAIDs are reported to associate with gastrointestinal side
effects due to their inhibition of the synthesis of the
prostaglandins PGI2 and PGF2.13 Although selective COX-2
inhibitors (e.g., coxibs) have a lower risk of gastrointestinal
damage, they exert cardiovascular side effects presumably by
altering the balance between PGI2 and thromboxanes.14 Dual-
target inhibitors for COX-2 and LTA4H are considered to be
able to establish a more efficient anti-inflammatory therapy
strategy with reduced side effects.15,16 Until now, only a few
COX-2/LTA4H dual-target inhibitors have been reported.13
A
possible difficulty is due to the large difference in substrate-
binding pockets between the two enzymes.
In the present study, COX-2 and LTA4H were used as the
paradigm. We used our de novo multitarget drug design method
LigBuilder 2 is a de novo drug design program developed in
the authors’ lab with embedded synthesis accessibility analysis
and various scoring schemes.9 We recently upgraded the
program to LigBuilder 3 and developed a new function to
Received: January 11, 2014
Published: March 10, 2014
© 2014 American Chemical Society
1235
dx.doi.org/10.1021/ci500021v | J. Chem. Inf. Model. 2014, 54, 1235−1241

Journal of Chemical Information and Modeling
Article
Figure 1. Flowchart of de novo multitarget ligand design method.
a
Scheme 1. Synthesis of Compounds II-1 and III-1 through III-4
a
Reagent and conditions: (a) EDCI, DMAP, THF, rt; (b) H2, Pd/C, MeOH or AcOEt.
with an iterative fragment-growing strategy to design dual-
target inhibitors for both proteins. Representative compounds
were synthesized, and their inhibition activities to purified
enzymes and human whole blood were tested. Surface plasmon
resonance was used to confirm the direct binding interactions
of the serial optimized compounds with COX-2 and LTA4H.
successful design ligands with high potencies for multiple
targets.
Target Protein and Fragments Preparation. The crystal
structure of COX-2 (PDB code 1PXX) and LTA4H (PDB code
1HS6) were downloaded from the Protein Data Bank.24 Water
molecules and co-crystallized ligands were removed from the
structures using the SYBYL software.25 The structures of all the
small molecule fragments were modeled using the SYBYL
software and optimized with Tripos force field.25
METHODS
■
Overview. The overall strategy is shown in Figure 1. In
order to increase the success rate of de novo multitarget drug
design, an iterative fragment-growing strategy was used in our
design process. This strategy was borrowed from fragment-
based drug discovery (FBDD), which has emerged over recent
years as an effective drug discovery strategy.17−21 The main idea
in FBDD is to start the hit and lead optimization process with
small molecular fragments with molecular weights in the 120−
300 Da range.22 The molecular complexity model proposed by
Hann and co-authors shows that smaller ligands are more likely
to bind to multiple targets than larger ones.23 In the first step,
we designed small fragments with limited molecular weights.
The activities of these fragments to COX-1/2 and LTA4H were
tested with high-concentration biochemical assays, and the
potent fragments with dual inhibition to the both targets were
selected. These potent fragments, named “multitarget seeds”,
were used as starting structures in the following optimization.
In the second step, using LigBuilder 3, the multitarget seeds
grow larger with molecular weights increasing within a limited
range, and we experimentally selected the more potent ones as
seed structures for the next round. This “design−synthesis−
bioassay” loop was repeated for several rounds to ensure
Fragments Docking. AutoDock 4.026 was used to dock the
fragments hits into the binding site of the two enzymes. For
COX-2, the size of docking box is 21.0 Å × 14.0 Å × 17.0 Å,
with the grid spacing of 0.375 Å. For LTA4H, the size of the
docking box is 31.0 Å × 19.0 Å × 23.0 Å, with a grid spacing of
0.375 Å. Gasteiger charges were used for both the protein and
ligand structures. The Lamarckian genetic algorithm (GA) was
used, and the related AutoDock parameters were set as follows:
200 GA runs, each with population size of 200; 25 million
energy evaluations; and a maximum of 270,000 generations per
GA run. The 200 independent GA runs from AutoDock were
processed using the built-in clustering analysis with a 2.0 Å
cutoff.
Multitarget Ligand Design. Liguilder 3 was used to
produce novel compounds based-on the seed structures.
Ligands generated by LigBuilder 3 are expected to bind to
both proteins with different binding conformations. The drug-
like and privileged building blocks used in this work were
inherited from LigBuilder 2. We apply the “Growing Mode” of
Liguilder 3 to manipulate the optimization process. The GA
parameters were set as follows: GA population size of 5000, GA
1236
dx.doi.org/10.1021/ci500021v | J. Chem. Inf. Model. 2014, 54, 1235−1241

Journal of Chemical Information and Modeling
Article
parent ratio of 10%, GA generation number of 15. To enhance
the diversity and quality of the designed compounds, we
designed about 1 million molecules and then apply the
“Recommendation” module of LigBuilder 3 to pick the top
1000 results in each round of optimization. Considering each
GA procedure will produce hundreds of compounds, this
procedure was repeated more than 1000 times, which ensures
that each seed structure in the seed list has been used.
Chemistry. Compounds II-1 and III-1 through III-4 were
prepared in moderate to high yields (Scheme 1). 5-Nitro-
salicylic acid reacted with different substituted 3-phenylpropyl
alcohols, and 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide
hydrochloride (EDCI) was used as a coupling agent in the
presence of a catalytic amount of 4-dimethylaminopyridine
(DMAP). The synthesis of the first round compounds and
experimental details are supplied in the Supporting Informa-
tion.
In Vitro Enzymatic Assay. The title compounds were
evaluated in enzymatic assay of COX-1/2 and LTA4H. The
enzyme activity of the purified COX-1/2 was measured by a
chromogenic assay based on monitoring the absorption of
oxidized N,N,N′,N′-tetramethyl-p-phenylenediamine (TMPD)
during the reduction of PGG2 to PGH2 at 610 nm.27 The
LTA4H hydrolase activity was measured using an ELISA assay
kit to quantify the amount of LTB4 generated.
Human Whole Blood Assay. II-1 and III-1 were also
evaluated in human whole blood assay. In this study, E. coli
liopolysaccharide (LPS) was used to induce the COX-2/PGES
pathway in human whole blood, and A23187 was used to
induce the 5-LOX/LTA4H pathway (see Surpporting In-
formmation for details).28,29
Surface Plasmon Resonance (SPR) Analysis. SPR
experiments were performed on a Biacore T200 using a CM5
sensor chip. PBS-P+ 10× buffer of GE Healthcare was diluted
to running buffer with 5% DMSO. The immobilization levels
were 10,000−15,000 RU. The references were blocked with
ethanolamine-HCl after activated with EDC and NHS. A small
molecule wizard of T200 was used to test the binding of SH-1
and the following optimized compounds. The binding strengths
of these compounds at 100 μM were tested for comparison.
The same method was used to the measure of kinetic profiles of
III-1 with the two targets (2-fold dilution series, five
concentrations), and a 1:1 interaction model was used to fit
the analysis data.
single-target inhibitors of COX-2 and LTA4H have already
been discovered, we extracted small fragments from these
inhibitors, which can also be done by LigBuilder 3 automati-
cally. In this fragment extraction and selection process,
considering Congreve’s rule-of-three,30 we employed several
rules to select known ligands. (1) The molecular weight of the
screening fragments should be no more than 300 Da. (2) The
number of hydrogen bond acceptors and donors should be less
than or equal to three. (3) The number of rings should be
greater than or equal to one. (4) The structure−activity
relationship (SAR) of known ligands with their targets was
manually considered. Consequently, 21 fragments from the
extracted fragment library were selected and bought from
chemical reagent corporations. Among them, 16 fragments
were derived from known COX-1/2 inhibitors, and the other
five were derived from known LTA4H inhibitors (see Table S-1,
Supporting Information, for these fragment structures). As the
fragments were small and their potencies were expected to be
low, high-concentration biochemical assays, typically 250−1000
μM,31 were used to identify active fragment hits. In our study,
the inhibition activity of these fragments to COX-1/2 and
LTA4H were measured at 1 mM concentrations, and nine dual
functional hits were found. The high hit rate (43%) confirmed
that extracting small fragments from known single-target
inhibitors is a reliable method to build starting seeds for
multitarget ligand design.
Seeds Growing. After discovering multitarget lead frag-
ments as seed structures, LigBuilder 3 was used to optimize
these seeds in a stepwise way. The lead fragments were docked
into each of the two targets, and several possible docking
conformations were selected for further optimization. As it is
difficult to predict the precise binding modes for “small”
fragments in the binding sites of the targets, selecting multiple
binding conformations is essential for follow-up optimization.
To ensure that the optimization direction was reasonable, for
each round, the molecular weight increase was restricted to
within 100 Da in the fragment growing process. The designed
ligands were selected for synthesis and bioassay evaluation, and
the active ligands were subject to the next round of
optimization. Through several rounds optimization, potent
dual target ligands were expected be discovered.
First Round of de Novo Multi-Target Optimization.
AutoDock 4.0 was used to dock the nine lead fragments into
COX-2 and LTA4H (PDB: 1pxx/COX-2 and 1hs6/LTA4H).
The docking conformations with the lowest energies and the
representative conformations of the five largest clusters were
selected as seeds for LigBuilder 3 growing. Duplicates caused by
the five largest clusters containing the lowest energy
conformations were removed, and 50 conformations were
selected for each receptor in this step. Using the 50
conformations of the nine lead fragments as seeds for both
receptors, LigBuilder 3 produced 1.1 million compounds and
exported 1000 top-ranked compounds. Analysis of these
molecules showed that the derivatives of three “single ring”
lead fragments, SH-1, SH-7, and SH-9 (Figure 2) were in the
1000 top-ranked designed molecules, while the derivatives of
the other fragments were ranked beyond the top 1000 designed
molecules. Three criterions were used to choose compounds
for the experimental testing: (1) structures with “minimum
modification”’, (2) “common frameworks” from a series of
analogues, and (3) structures that are easily prepared.
Consequently, six representative exported molecules of SH-1,
RESULTS
■
Dual-Target Seeds Identification. LigBuilder 3 is a de
novo multitarget design method. It can start with lead structures
Figure 2. Structures of SH-1, SH-7, and SH-9.
for multitargets as seeds and perform the optimization process
until larger molecules with high potencies for each target are
designed. If there are no lead structures available, LigBuilder 3
can generate lead structures automatically. However, small
fragments from known ligands of single targets are more
reliable to become lead structures than random ones. As many
1237
dx.doi.org/10.1021/ci500021v | J. Chem. Inf. Model. 2014, 54, 1235−1241

Journal of Chemical Information and Modeling
Article
Figure 3. Ligands selected from first-round optimization (fragments in blue are parent frameworks).
Table 1. COX-1/2 and LTA4H Inhibition Rates of First-
Round Optimized Ligands at 100 μM
inhibition %
COX-1
inhibition %
COX-2
inhibition %
LTA4H
compound
a
SH-1
13
5
12
1
48 21
17.5 1.0
43.2 4.4
32.0 2.5
35.0 1.5
n. i.
I-1
24.0 2.0
92.4 1.0
14.6 1.6
98.3 1.2
n. i.
b
I-1
c
I-2
I-3
I-4
I-5
I-6
n. i.
n. i.
n. i.
n. i.
n. i.
n. i.
n. i.
70.3 0.8
27.3 1.5
15.2 2.2
n. i.
a
b
c
SH-1was measured at 1 mM. I-1 was measured at 500 μM. n.i. = no
inhibition.
Figure 5. SPR binding curves. Binding levels of optimized compounds
with COX-2 (a) and LTA4H (b) at 100 μM. Interaction profiles of III-
1 with COX-2 (c) and LTA4H (d).
The in vitro enzymatic assays of the two targets (see
Supporting Information for details) were used to evaluate the
first-round optimized ligands, and the compound I-1, which
was evolved from SH-1, shows dual target inhibition activity
against COX-2 and LTA4H (Table 1). Therefore, using I-1 as
the seed, second-round optimization with LigBuilder 3 was
performed.
Figure 4. Designed poses of II-1 with (a) LTA4H and (b) COX-2.
The effective binding modes of I-1 with the two targets are
not evolved from the lowest energy conformations of the SH-1
docking results. Consequently, the use of multiple different
SH-7, and SH-9 derivatives were selected for synthesis and
bioassay (Figure 3).
Table 2. Bio-Evaluation of the Title Compounds
in vitro IC50 (μM)
HWB IC50 (μM)
compound
Z
COX-2
LTA4H
PGE2
LTB4
a
ref-1
−
−
−
−
0.025 0.002
−
−
0.63 0.17
−
6.9 0.1
12.6 0.5
7.1 0.7
n.t.
b
Flur
0.018 0.005
7.7 0.3
56.0 8.0
16.8 3.4
>50
4.3 1.2
8.4 0.7
18.0 1.8
7.0 2.5
c
5a
0.68 0.27
16.6 0.7
7.2 1.3
1.7 0.2
3.0 0.1
>50
II-1
III-1
III-2
III-3
III-4
H
4-Cl
4-F
d
n.t.
4-OMe
3-F-4-Cl
>50
n.t.
n.t.
n.t.
76.5 0.1
n.t.
a
b
c
d
Positive control of LTA4H. Positive control of COX-2. The best dual-target inhibitor of COX-2 and LTA4H reported.15 n.t.= no test.
1238
dx.doi.org/10.1021/ci500021v | J. Chem. Inf. Model. 2014, 54, 1235−1241

Journal of Chemical Information and Modeling
Article
Figure 6. Optimization path of SH-1. (a) Designed poses of LTA4H with SH-1, I-1, II-1, and III-1. (b) Structure of SH-1, I-1, II-1, and III-1. (c)
Designed poses of COX-2 with SH-1, I-1, II-1, and III-1 (generated by LigBuilder 3).
docking conformations of each fragment hit as a starting point
for LigBuilder 3 is vital for the success of de novo multitarget
ligand design. LigBuilder 3 was developed to have a strong
ability to simultaneously design and optimize multiple ligands
with multiple binding poses.
allowing II-1 to adopt different conformations in the two
binding sites. Thus, four derivatives of II-1 with different
hydrophobic phenyls were designed and synthesized. Table 2
shows the structures and in vitro activities of these derivatives.
The 4-chlorophenyl derivative III-1 possesses improved activity
to both targets (IC50 values: 16.8 μM to COX-2 and 7.2 μM to
LTA4H). Although the 4-fluorophenyl derivative III-2 and the
4-methoxyphenyl derivative III-3 showed improved inhibition
activities for LTA4H, their inhibition activities for COX-2
disappeared. On the other hand, the 4-chloro-3-fluorophenyl
derivative III-4 shows comparable COX-2 inhibition activity,
but loses LTA4H inhibition activity. Consequently, III-1, as a
more potent COX-2 and LTA4H dual-target inhibitor, was
found.
Prontein−Ligand Binding Assay. In order to confirm the
interactions of the serial optimized compounds with both
COX-2 and LTA4H, the binding strengths of these compounds
were tested using SPR at 100 μM concentration, and the
dissociation constants of the best compound III-1 with the
both targets were measured too. As shown in Figure 5a and b,
the binding strengths increased with each round of
optimization. The dissociation constants of III-1 were 28 μM
for COX-2 and 34 μM for LTA4H (Figure 5c and d).
Second-Round of de Novo Multi-Target Optimization.
All binding modes of I-1 with the both targets in the first round
optimization were used as seeds in this round. In this round,
the ligands became larger, so the log P value was restricted from
−1.4 to 6.6, which is a slight extension of the qualifying range
(−0.4 to 5.6) suggested by Ghose32 for further derivatization.
For the same reason, the number of hydrogen bond acceptors
and donors was set to less than five. For the genetic algorithm
in LigBuilder 3, the number of genetic algorithm runs was set to
15, and population size was set to 3000. One million molecules
were produced with LigBuilder 3 and ranked by size-
independent ligand efficiency as proposed by Nissink,33 and
the 1000 top-ranked compounds were then exported. Analysis
of the designed molecules shows that a common framework 3-
phenylpropyl-5-amino-2-hydroxybenzoate (II-1) can represent
all the key interactions of these molecules to both targets
(Figure4); hence, II-1 was selected for experimental
verification. We synthesized II-1 and measured its inhibition
activities toward the two enzymes. The IC50 values of II-1 to
the two targets were measured in the in vitro enzymatic and the
human whole blood (HWB) assays (Table 2). Two known
inhibitors, 1-(2-(4-phenoxyphenoxy)ethyl)pyrrolidine (ref-
1),34 a reported LTA4H inhibitor, and fluorobiprofen
(Flur),35 a reported COX-2 inhibitor, were used as positive
controls (see Supporting Information for the structures of ref-1
and Flur).
DISCUSSION
■
The optimization from SH-1 to III-1 is showed in Figure 6. In
the binding mode of SH-1 with the two targets, the hydroxyl
and carboxyl groups interact with the two targets as hydrogen
bond acceptors, and the H of the carboxyl group is free. Hence,
the displacement of H with an alkyl group retains the activity of
the parent structure. Moreover, the amino group introduced
during first-round optimization makes hydrogen bonds with
Tyr365 of COX-2 and Arg563 of LTA4H (Figure 4), resulting
in an activity increase of I-1 toward both enzymes (Table 1). In
the second round of optimization, the newly added phenyl
group leads to a notable inhibition increase in II-1 to both the
Further Improve the Potency of II-1. As revealed in
Figure 4, both the hydroxyl group and amino group are
engaged in the key hydrogen bonds, which anchor the
molecules in the binding sites. The three-carbon flexible linker
between the polar moiety and hydrophobic moiety is
appropriate for the interaction of II-1 with both the targets,
1239
dx.doi.org/10.1021/ci500021v | J. Chem. Inf. Model. 2014, 54, 1235−1241

Journal of Chemical Information and Modeling
Article
targets due to the hydrophobic interaction between the phenyl
group and the two binding sites (Figure 4). Hence, using a de
novo multitarget design method, a high quality dual-target
inhibitor hit II-1 was designed efficiently, and a more potent
III-1 was found after a simple modification of II-1. III-1
inhibits PGE2 and LTB4 production in the human whole blood
assay with IC50 values of 7.0 and 7.1 μM, respectively. The SPR
binding assay confirms that the inhibition activities of these
ligands are related to their direct binding to the targets.
The COX-2 and COX-1 selectivity indicator (SI = IC50
(COX-1)/IC50 (COX-2)) is one of the main evaluation
indicators for the safety of anti-inflammatory drugs. In previous
work, we built a mathematical model for the inflammation-
related arachidonic acid metabolic network and predicted that
the use of COX-2 and LTA4H dual-target inhibitors with COX-
2 selectivity (SI approximately 7) is an optimum intervention
solution. We tested the inhibition activity of III-1 to COX-1
(179 21 μM) and found that its selectivity for COX-2 is 11.1,
comparable to the ideal value. With good potency in both in
vitro and HWB assays as well as COX-2 selectivity, III-1 is a
promising lead compound for further development of novel
anti-inflammatory drug.
ABBREVIATIONS
■
NSAIDs, nonsteriodal anti-inflammatory drugs; FBDD, frag-
ment-based drug design; COX, cyclooxygenase; 5-LOX, 5-
lipoxygenase; PGES, prostaglandin E synthase; LTA4H,
leukotriene A4 hydrolase; PGE2, prostaglandin E2; LTB4,
leukotriene B4
REFERENCES
■
(1) Brown, D.; Superti-Furga, G. Rediscovering the sweet spot in
drug discovery. Drug Discovery Today 2003, 8, 1067−1077.
(2) Kamb, A.; Wee, S.; Lengauer, C. Why is cancer drug discovery so
difficult? Nat. Rev. Drug Discovery 2007, 6, 115−120.
(3) Cavalli, A.; Bolognesi, M. L.; Minarini, A.; Rosini, M.; Tumiatti,
V.; Recanatini, M.; Melchiorre, C. Multi-target-directed ligands to
combat neurodegenerative diseases. J. Med. Chem. 2008, 51, 347−372.
́ ́ ́
(4) Korcsmaros, T.; Szalay, M. S.; Bode, C.; Kovacsvacs, I. A.;
̈
Csermely, P. How to design multi-target drugs: target search options
in cellular networks. Expert Opin. Drug Discovery 2007, 2, 1−10.
́
(5) Zimmermann, G. R.; Lehar, J.; Keith, C. T. Multi-target
therapeutics: When the whole is greater than the sum of the parts.
Drug Discovery Today 2007, 12, 34−42.
(6) Morphy, R.; Kay, C.; Rankovic, Z. From magic bullets to
designed multiple ligands. Drug Discovery Today 2004, 9, 641−651.
(7) Morphy, R.; Rankovic, Z. Designed multiple ligands. An
emerging drug discovery paradigm. J. Med. Chem. 2005, 48, 6523−
6543.
Compared to framework combination methods using
molecular linking, substitution, or fusion strategies, LigBuilder
3 builds highly efficient multifunctional ligands. Ligands are
designed to interact with different targets with distinct
conformations, and atoms of the ligands are designed to
maximize their interaction with all the targets.
(8) Morphy, J. R. In Designing Multi-Target Drugs; Morphy, J. R.,
Harris, C. J., Eds.; RSC Publishing: Cambridge, U.K., 2012; Chapter 8,
pp 111−129.
(9) Yuan, Y.; Pei, J.; Lai, L. LigBuilder 2: a practical de novo drug
design approach. J. Chem. Inf. Model. 2011, 51, 1083−1091.
(10) The program LigBuilder 3 will be published elsewhere.
(11) Simmons, D. L.; Botting, R. M.; Hla, T. Cyclooxygenase
isozymes: The biology of prostaglandin synthesis and inhibition.
Pharmacol. Rev. 2004, 56, 387−487.
CONCLUSIONS
■
A de novo multitarget drug design method, with a special
iterative strategy to improve the design success rate, was
developed. This strategy is especially for the design of highly
integrated inhibitors for proteins with dissimilar binding
pockets. Using this method, the best in class dual target
inhibitor for COX-2 and LTA4H was discovered. The
successful application of our strategy shows that is a powerful
tool for multitarget ligand design and optimization. This
method can be generally used to design and optimize
multifunction ligands binding to two or more targets.
(12) Haeggstrom, J. Z. Leukotriene A4 hydrolase/aminopeptidase,
̈
the gatekeeper of chemotactic leukotriene B4 biosynthesis. J. Biol.
Chem. 2004, 279, 50639−50642.
(13) Rainsford, K. D. Anti-Inflammatory Drugs in the 21st Century.
In Subcellular Biochemistry; Harris, R.E., Ed.; Springer: New York,
2007; pp 3−27, 31−141, 145−190, 193−225, 229−279, 283−318.
(14) McGettigan, P.; Henry, D. Cardiovascular risk and inhibition of
cyclooxygenase - A systematic review of the observational studies of
selective and nonselective inhibitors of cyclooxygenase. JAMA, J. Am.
Med. Assoc. 2006, 296, 1633−1644.
ASSOCIATED CONTENT
* Supporting Information
Fragments library, synthesis, bioassay, and scanned spectrum of
the 1H NMR and 13C NMR data of representative compounds.
This material is available free of charge via the Internet at
http://pubs.acs.org.
■
S
(15) Chen, Z.; Wu, Y.; Liu, Y.; Yang, S.; Chen, Y.; Lai, L. Discovery
of dual target inhibitors against cyclooxygenases and leukotriene A4
hydrolyase. J. Med. Chem. 2011, 54, 3650−3660.
(16) Yang, K.; Bai, H.; Ouyang, Q.; Lai, L.; Tang, C. Finding multiple
target optimal intervention in disease-related molecular network. Mol.
Syst. Biol. 2008, 4, 228.
(17) Hajduk, P. J.; Greer, J. A decade of fragment-based drug design:
Strategic advances and lessons learned. Nat. Rev. Drug Discovery 2007,
6, 211−219.
(18) Congreve, M.; Chessari, G.; Tisi, D.; Woodhead, A. J. Recent
developments in fragment-based drug discovery. J. Med. Chem. 2008,
51, 3661−3680.
(19) Chessari, G.; Woodhead, A. J. From fragment to clinical
candidateA historical perspective. Drug Discovery Today 2009, 14,
668−675.
(20) Morphy, R.; Rankovic, Z. Fragments, network biology and
designing multiple ligands. Drug Discovery Today 2007, 12, 156−160.
(21) Hopkins, A. L.; Mason, J. S.; Overington, J. P. Can we rationally
design promiscuous drugs? Curr. Opin. Struct. Biol. 2006, 16, 127−136.
(22) Schulz, M. N.; Hubbard, R. E. Recent progress in fragment-
based lead discovery. Curr. Opin. Pharmacol. 2009, 9, 615−621.
AUTHOR INFORMATION
Corresponding Authors
*Fax: (+86)10-62751725. E-mail: liuying@pku.edu.cn Y.L.).
*Fax: (+86)10-62751725. E-mail: jfpei@pku.edu.cn (J.P.).
*Fax: (+86)10-62751725. E-mail: lhlai@pku.edu.cn (L.L.).
■
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
This research was supported in part by the Ministry of Science
and Technology of China (grant numbers: 2012AA020308,
2012AA020301, 2009CB918503) and the National Natural
Science Foundation of China (grant numbers: 81273436,
11021463, 20873003).
1240
dx.doi.org/10.1021/ci500021v | J. Chem. Inf. Model. 2014, 54, 1235−1241

Journal of Chemical Information and Modeling
Article
(23) Hann, M. M.; Leach, A. R.; Harper, G. Molecular complexity
and its impact on the probability of finding leads for drug discovery. J.
Chem. Inf. Comput. Sci. 2001, 41, 856−864.
(24) Berman, H. M.; Westbrook, J.; Feng, Z.; Gilliland, G.; Bhat, T.
N.; Weissig, H.; Shindyalov, I. N.; Bourne, P. E. The Protein Data
Bank. Nucleic Acids Res. 2000, 28, 235−242.
(25) Tripos, Inc., St. Louis, MO.
(26) Autodock, version 4; The Scripps Research Institute: La Jolla,
CA, 2007.
(27) Ouellet, M.; Falgueyret, J. P.; Percival, M. D. Detergents
profoundly affect inhibitor potencies against both cyclo-oxygenase
isoforms. Biochem. J. 2004, 377, 675−684.
(28) Frohberg, P.; Drutkowski, G.; Wobst, I. Monitoring eicosanoid
biosynthesis via lipoxygenase and cyclooxygenase pathways in human
whole blood by single HPLC run. J. Pharma. Biomed. Anal. 2006, 41,
1317−1324.
(29) de Leval, X.; Delarge, J.; Neven, P.; Michaux, C.; Masereel, B.;
Pirotte, B.; David, J. L.; Henrotin, Y.; Dogne, J. M. Evaluation of
classical NSAIDs and COX-2 selective inhibitors on purified ovine
enzymes and human whole blood. Prostaglandins, Leukotrienes Essent.
Fatty Acids 2001, 64, 211−216.
(30) Congreve, M.; Carr, R.; Murray, C.; Jhoti, H. A “Rule of Three”
for fragment-based lead discovery. Drug Discovery Today 2003, 8, 876−
877.
(31) Barker, J.; Courtney, S.; Hesterkamp, T.; Ullmann, D.;
Whittaker, M. Fragment screening by biochemical assay. Expert
Opin. Drug Discovery 2006, 1, 225−236.
(32) Ghose, A. K.; Viswanadhan, V. N.; Wendoloski, J. J. A
knowledge-based approach in designing combinatorial or medicinal
chemistry libraries for drug discovery. 1. A qualitative and quantitative
characterization of known drug databases. J. Comb. Chem. 1999, 1, 55−
68.
(33) Willem, J.; Nissink, M. Simple size-independent measure of
ligand efficiency. J. Chem. Inf. Model. 2009, 49, 1617−1622.
(34) Penning, T. D.; Chandrakumar, N. S.; Chen, B. B.; Chen, H. Y.;
Desai, B. N.; Djuric, S. W.; Docter, S. H.; Gasiecki, A. F.; Haack, R. A.;
Miyashiro, J. M.; Russell, M. A.; Yu, S. S.; Corley, D. G.; Durley, R. C.;
Kilpatrick, B. F.; Parnas, B. L.; Askonas, L. J.; Gierse, J. K.; Harding, E.
I.; Highkin, M. K.; Kachur, J. F.; Kim, S. H.; Krivi, G. G.; Villani-Price,
D.; Pyla, E. Y.; Smith, W. G.; Ghoreishi-Haack, N. S. Structure-activity
relationship studies on 1-[2-(4-phenylphenoxy)ethyl]pyrrolidine (SC-
22716), a potent inhibitor of leukotriene A(4) (LTA(4)) hydrolase. J.
Med. Chem. 2000, 43, 721−735.
(35) Ouellet, M.; Falgueyret, J. P.; Percival, M. D. Detergents
profoundly affect inhibitor potencies against both cyclo-oxygenase
isoforms. Biochem. J. 2004, 377, 675−684.
1241
dx.doi.org/10.1021/ci500021v | J. Chem. Inf. Model. 2014, 54, 1235−1241