Structure-based approach to improve a small-molecule inhibitor by the use of a competitive peptide ligand

Article abstract of DOI:10.1002/anie.201310749

Structural information about the target-compound complex is invaluable in the early stage of drug discovery. In particular, it is important to know into which part of the initial compound additional interaction sites could be introduced to improve its affinity. Herein, we demonstrate that the affinity of a small-molecule inhibitor for its target protein could be successfully improved by the constructive introduction of the interaction mode of a competitive peptide. The strategy involved the discrimination of overlapping and non-overlapping peptide-compound pharmacophores by the use of a ligand-based NMR spectroscopic approach, INPHARMA. The obtained results enabled the design of a new compound with improved affinity for the platelet receptor glycoprotein VI (GPVI). The approach proposed herein efficiently combines the advantages of compounds and peptides for the development of higher-affinity druglike ligands. The best of both worlds: In a structure-based strategy to improve the affinity of a small-molecule inhibitor for its target protein, the interaction mode of a competitive peptide was constructively introduced. Thus, the discrimination of overlapping and non-overlapping peptide-compound pharmacophores by INPHARMA NMR spectroscopy enabled the design of a new compound (see structure) with improved affinity for the platelet receptor glycoprotein VI.

Full text of DOI:10.1002/anie.201310749

Angewandte
Chemie
DOI: 10.1002/anie.201310749
Structure-Based Drug Discovery
Structure-Based Approach To Improve a Small-Molecule Inhibitor by
the Use of a Competitive Peptide Ligand**
Katsuki Ono, Koh Takeuchi, Hiroshi Ueda, Yasuhiro Morita, Ryuji Tanimura, Ichio Shimada,*
and Hideo Takahashi*
Abstract: Structural information about the target–compound
complex is invaluable in the early stage of drug discovery. In
particular, it is important to know into which part of the initial
compound additional interaction sites could be introduced to
improve its affinity. Herein, we demonstrate that the affinity of
a small-molecule inhibitor for its target protein could be
successfully improved by the constructive introduction of the
interaction mode of a competitive peptide. The strategy
involved the discrimination of overlapping and non-overlap-
ping peptide–compound pharmacophores by the use of
a ligand-based NMR spectroscopic approach, INPHARMA.
The obtained results enabled the design of a new compound
with improved affinity for the platelet receptor glycoprotein VI
(GPVI). The approach proposed herein efficiently combines
the advantages of compounds and peptides for the develop-
ment of higher-affinity druglike ligands.
tein-based NMR spectroscopic approaches, exemplified by
“SAR by NMR”, provide information on compound-binding
sites in the target protein. This information enables the
modeling of target–compound complex structures in combi-
nation with computational approaches. However, protein-
based NMR spectroscopic approaches require 15N and/or 13
C
isotopic labeling, and more costly deuterium labeling should
be considered when the molecular weight of the target protein
is large (> 30 K).[4] On the other hand, ligand-based NMR
spectroscopic approaches, such as saturation transfer differ-
ence (STD) and WaterLOGSY experiments,[5] have no
restrictions on the size of the target protein, and there is no
requirement for isotopic labeling. However, it is often
difficult to obtain the relative position and orientation of
fragments with a ligand-based NMR spectroscopic approach.
The only exception is the INPHARMA method (interligand
NOEs for pharmacophore mapping), which allows the
identification of a common pharmacophore between two
competing compounds.[6] Although INPHARMA may infer
common epitopes, the information from overlapping small-
molecular ligands may often be insufficient to link or grow
compounds to improve their affinity. Instead, information on
additional non-overlapping interaction sites and/or the rela-
tive position of compounds that bind distinct sites in the target
protein is more important for improving the affinity of
a compound.
T
he design of high-affinity compounds from initially
screened “hits” is essential in the early stage of drug
discovery. At this stage, the original compounds are modified
by chemical approaches to pick up additional interactions in
the binding site. Structural information on the target–
compound complex provided by NMR spectroscopy and/or
X-ray crystallography is critical to guiding strategic improve-
ment of the small-molecule inhibitor.[1]
NMR spectroscopy has been widely used for hit improve-
ment, as key early developments in this field were made on
the basis of the technique “SAR by NMR” (the use of NMR
spectroscopy to determine structure–activity relationships) in
the late 1990s.[2] NMR spectroscopic techniques hold a firm
position, even after recent expansions in strategies based on
crystal-structure determination,[3] especially when the crys-
tallization of target–compound complexes is difficult. Pro-
To overcome this weakness in the ligand-based NMR
spectroscopic approach, we propose a new strategy, in which
information on target protein–peptide interactions is used
constructively. A protein–peptide interface is often larger
than low-molecular-weight compounds and consists of multi-
ple key interactions. Furthermore, peptide ligands that
directly bind to a biologically relevant site in the target
[*] Dr. K. Ono, Dr. K. Takeuchi, Prof. I. Shimada, Prof. H. Takahashi
Biomedicinal Information Research Center (BIRC), National
Institute of Advanced Industrial Science and Technology (AIST)
2-3-26 Aomi, Koto-ku, Tokyo 135-0064 (Japan)
Prof. H. Takahashi
Graduate School of Medical Life Science, Yokohama City University
1-7-29 Suehirocho, Tsurumi-ku, Yokohama, Kanagawa 230-0045
(Japan)
E-mail: hid@tsurumi.yokohama-cu.ac.jp
[**] This project was supported by a grant from the Japan New Energy
and Industrial Technology Development Organization (NEDO) and
the Ministry of Economy, Trade, and Industry (METI; to I.S.).
Funding was also provided by Grants-in-Aid for Scientific Research
(KAKENHI; grant numbers 24370048 to H.T. and K.T., 24102524 to
H.T., and 25121743 to K.T.) from the Ministry of Education, Culture,
Sports, Science and Technology (MEXT) and the Japan Society for
the Promotion of Science (JSPS). We thank M. Nagasu for technical
assistance during sample preparation.
Dr. K. Ono
Japan Biological Informatics Consortium (JBIC)
2-3-26 Aomi, Koto-ku, Tokyo 135-0064 (Japan)
H. Ueda, Y. Morita, Dr. R. Tanimura
Toray Branch Office of JBIC
6-10-1 Tebiro, Kamakura-shi, Kanagawa 248-8555 (Japan)
Prof. I. Shimada
Graduate School of Pharmaceutical Sciences
The University of Tokyo
7-3-1 Hongo, Bunkyo-ku, Tokyo 133-0033 (Japan)
E-mail: shimada@iw-nmr.f.u-tokyo.ac.jp
Supporting information for this article is available on the WWW
under http://dx.doi.org/10.1002/anie.201310749.
Angew. Chem. Int. Ed. 2014, 53, 2597 –2601
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
2597

.
Angewandte
Communications
protein can be obtained relatively easily by phage display[7] or
other approaches for in vitro evolution.[8] Although peptides
have limited utility as therapeutic agents, their efficacy,
selectivity, and specificity are often better than those of
small fragments.[9] Thus, the introduction of information from
target-protein–peptide interactions in the design of high-
affinity ligands would be an effective way to compensate for
the problem originating from the small interaction surface of
compounds.
Herein, we demonstrate that the affinity of a small-
molecule inhibitor for its target protein could be successfully
improved by the generation of hybrid molecules in which non-
overlapping moieties from a competitive peptide were
introduced. In this strategy, overlapping and non-overlapping
peptide–compound pharmacophores were discriminated by
using INPHARMA. This novel approach combines the
advantages of peptides and small molecules for the develop-
ment of higher-affinity ligands by the ligand-based NMR
spectroscopic approach.
Figure 1. Structural representation of pep-10L peptide and losartan.
A) The GPVI-bound structure of pep-10L, as determined by structural
calculation by the use of transferred NOE (TrNOE) information.[13]
Only residues 4–10, which represent the minimal binding sequence for
binding to GPVI, are shown. Residue numbers and three-letter codes
are indicated. The residues important for GPVI binding are indicated
by black boxes. B) The GPVI-bound structure of losartan.[15] All carbon
atoms in losartan are numbered. The carbon atoms in the phenyl-
tetrazole (PTZ) moiety, which is important for GPVI binding, are
circled.
Our approach was applied to the interaction between the
platelet receptor glycoprotein VI (GPVI) and collagen. GPVI
is primarily responsible for initial collagen-dependent platelet
adhesion and aggregation; deficiency of GPVI causes a loss of
collagen-stimulated platelet activation.[10] Interestingly, the
absence of GPVI does not significantly prolong bleeding time,
and a recent phase I clinical study demonstrated that GPVI–
Fc dimers (Fc is the fragment crystallizable region of an
antibody) inhibited collagen-induced platelet aggregation
without affecting general hemostasis.[11] Therefore, unlike
other antiplatelet agents, which are often associated with
adverse bleeding events, an inhibitor of the GPVI–collagen
interaction would be an interesting target for a safer anti-
aggregation drug. In fact, it has been shown that the
nonpeptide angiotensin II type 1 (AT1) receptor antagonist
losartan and its metabolite also inhibit collagen-dependent
platelet activation by GPVI at the clinical dose.[12]
Recently, we used an approach based on a phage-dis-
played peptide library to create a 12 amino acid peptide, pep-
10L (H2N-YSDTDWLYFSTS-COOH), which inhibits the
GPVI–collagen interaction.[13] The structure of pep-10L in the
GPVI-bound state was determined by structural calculation
by using transferred NOE (TrNOE) information.[13] The
structure revealed that the central part of pep-10L (Asp5-
Phe9) adopts a helical conformation (Figure 1A), in which
the side chains of Trp6, Leu7, and Phe9 form a hydrophobic
cluster upon binding to GPVI.[13] By the use of STD experi-
ments and site-directed mutagenesis, the hydrophobic cluster
was identified as the key interaction site of pep-10L.[13] The
chemical-shift perturbation (CSP) of GPVI resonances upon
binding to pep-10L indicated that the peptide binds to the
proposed collagen-binding site[14] with a KD value of 5.7 ꢀ
10À5 m (see Figure S1 in the Supporting Information).
chemical structure for the interaction with GPVI (Fig-
ure 1B).[15]
To determine whether there was direct competition
between losartan and pep-10L on the GPVI surface, we
performed NMR spectroscopic competition experiments. We
1
recorded H–15N correlated spectra of 15N-labeled pep-10L
alone (see Figure S2A), in the presence of unlabeled GPVI–
Fc (see Figure S2B), and after adding losartan to the mixture
(see Figure S2C). When the pep-10L/GPVI-Fc ratio was set to
1:0.25, the signal intensity of pep-10L significantly decreased
in comparison to the free state (see Figure S2B). This
decrease in intensity reflects accelerated relaxation owing to
interaction with a larger molecule, GPVI–Fc. Upon the
subsequent addition of losartan (25-fold concentration rela-
tive to pep-10L) to the mixture of pep-10L and GPVI–Fc, the
intensity of the pep-10L resonances were restored to up to
80% of their original intensity (see Figure S2C). These results
indicated that pep-10L and losartan compete for the same
binding site in the GPVI molecule.
With the competing low-molecular-weight compound
losartan and the peptide pep-10L, we started to investigate
which parts of the compound overlap with pep-10L by using
1
the INPHARMA method (Figure 2). Two-dimensional H–
1H NOESY experiments were performed with various
mixing times (60, 100, 200, and 300 ms) with losartan
(1.4 mm), pep-10L (0.9 mm), and GPVI–Fc (25 mm) in D2O
buffer. At a mixing time of 200 ms, interligand NOE peaks
between the PTZ moiety in losartan and pep-10L were clearly
observed in the NOESY spectrum (Figure 2C). Interligand
NOE peaks were not detected in the absence of GPVI–Fc or
with the Fc portion alone, thus excluding the possibility of
a direct interaction between losartan and pep-10L or of
indirect magnetization transfer via the Fc fragment (data not
shown). In view of the competition between pep-10L and
losartan for the GPVI–Fc binding site, these NOE peaks
could be identified as protein (GPVI)-mediated
INPHARMA peaks.
Interestingly, the site seems to overlap with the binding
site of losartan, which we proposed recently on the basis of
NMR spectroscopy.[15] By using NMR spectroscopic methods
in combination with in silico tools, we found that losartan
specifically interacts with extracellular immunoglobulin (Ig)-
like domain 1 of GPVI with a KD value of 1.7 ꢀ 10À4 m, and
that the phenyltetrazole (PTZ) moiety in losartan is the key
2598
www.angewandte.org
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2014, 53, 2597 –2601

Angewandte
Chemie
suggested by molecular-dynamics simulations of the GPVI–
losartan complex.[15] The AT1-receptor antagonists valsartan
and olmesartan showed affinity comparable to that of
losartan (see Figure S3); thus, the substitution of the imida-
zole group has only a marginal effect on the affinity of AT1-
receptor antagonists. On the other hand, telmisartan, which
does not have a PTZ moiety, did not bind to GPVI (see
Figure S3). In line with this observation, 2-biphenylcarboxylic
acid has fourfold lower affinity as compared to 5-biphenyl-2-
yl-2H-tetrazole. Thus, the lack of a tetrazole ring in telmi-
sartan at least partially explains its lack of affinity for GPVI.
Interestingly, although the side chain of Phe9 is important for
the interaction of GPVI and pep-10L, there were no
INPHARMA peaks between the aromatic ring of Phe9 in
pep-10L and losartan. Thus, pep-10L has an additional
interaction site (the phenyl portion of Phe9) that does not
overlap with losartan, and the compound-binding affinity of
GPVI might be improved by introducing a similar interaction
to that of Phe9 of pep-10L.
Figure 2. Expanded region of the 1H–1H NOESY spectra recorded for
mixtures of A) losartan (1.4 mm) and GPVI–Fc (25 mm), B) pep-10L
(0.9 mm) and 25 mm GPVI–Fc, and C) losartan (1.4 mm), pep-10L
(0.9 mm), and GPVI-Fc (25 mm). The chemical shifts of losartan and
pep-10L are indicated by orthogonally crossed solid and broken lines,
respectively. Peaks in (A) and (B) indicate intramolecular TrNOE
peaks, whereas signals in black boxes in (C) are the INPHARMA peaks
mediated by the GPVI–Fc hydrogen atoms. NMR spectroscopic data
were collected on an Avance 800 MHz spectrometer equipped with
a cryogenically cooled probe head. The mixing time (tm) was set to
200 ms. All spectra were recorded in 20 mm sodium phosphate buffer
(NaPi; pH 6.5) in D2O. The procedures used for the expression and
purification of GPVI–Fc were as previously described.[13]
As discussed above, the phenyl ring in the PTZ moiety can
be positioned between the side chains of Trp6 and Leu7
(Figure 3). The orientation of the PTZ moiety was estimated
from INPHARMA peaks between H7/H8 in losartan and
Twenty-four INPHARMA peaks were observed in the
experiment (Table 1). The strongest sets of INPHARMA
peaks were observed between the H14/H16 hydrogen atoms
in the PTZ moiety and the side-chain hydrogen atoms of Trp6
Table 1: Observed INPHARMA peaks between the hydrogen atoms of
losartan and pep-10L.
Losartan[a]
Pep-10L
Figure 3. Superimposition of pep-10L (green) and the PTZ moiety in
losartan (yellow) on the basis of INPHARMA information. The main
chain of pep-10L (Thr4-Ser10) is rendered as a line model, whereas the
side chains of key residues in the GPVI interaction, Trp6, Leu7, and
Phe9, are shown as a stick model. Carbon atoms in pep-10L, the
carbon atoms of the PTZ moiety, nitrogen atoms, and oxygen atoms
are colored green, yellow, blue, and red, respectively. The red broken
line indicates the linker region between two fragments.
H7/H11 or H8/H10
T4 (Ha), W6 (Hb), L7 (Hb, Hd), Y8 (Hb),
F9 (Ha, Hb), S10 (Hb)
W6 (Hb), L7 (Hb, Hd), S10 (Hb)
T4 (Ha), D5 (Ha), W6 (Hb, Hh2, Hz2, Hz3),
L7 (Hb, Hd), S10 (Hb)
H13
H14/H16
H15
T4 (Ha), D5 (Ha), W6 (Hz2)
[a] The numbering of losartan hydrogen atoms corresponds to the
carbon-atom numbering in Figure 1B.
residues forming a hydrophobic cluster in pep-10L as well as
the other INPHARMA peaks from H13 and H15 of the PTZ
ring (Table 1). The estimated distance between the PTZ
moiety in losartan and the phenyl ring of Phe9 in pep-10L was
expected to be approximately 9 ꢁ, with Phe9 oriented in the
direction corresponding to the meta position of the phenyl
ring of the PTZ moiety.
On the basis of this observation, we then constructed
a new compound that linked the important binding fragment
in losartan, the PTZ ring, and the phenyl ring of pep-10L
(Figure 3). First, we reconsidered the core structure to be used
in the hybrid compounds. Whereas 5-biphenyl-2-yl-2H-tetra-
zole retained an affinity (Kd = 3.6 Æ 0.5 ꢀ 10À4 m) comparable
to that of losartan, the methylated analogue of PTZ, 5-(2-
methylphenyl)-2H-tetrazole, showed only limited affinity
(Hh2 and Hz3) and Leu7 (Hb and Hd). These results indicate
that the PTZ moiety in losartan is positioned at a similar site
to that occupied by the side chains of Trp6 and Leu7 in pep-
10L. The phenyl group (H7/H11 and H8/H10) in losartan
showed INPHARMA peaks with Trp6 (Hb), Leu7 (Hb and
Hd), Tyr8 (Hb), Phe9 (Ha and Hb), and Ser10 (Hb) of pep-
10L at a mixing time of 200 ms. Thus, the phenyl group in
losartan is located in the position corresponding to the center
of the hydrophobic cluster formed by Trp6, Leu7, and Phe9 on
the surface of GPVI. No INPHARMA peaks from the
imidazole group and the alkyl chain of losartan were
observed, even at a longer mixing time of 300 ms. The
absence of such peaks may indicate that the imidazole group
and alkyl chain do not stably bind to the GPVI surface, as
Angew. Chem. Int. Ed. 2014, 53, 2597 –2601
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org 2599

.
Angewandte
Communications
toward GPVI (Kd = 1.1 Æ 0.1 mm). Thus, biphenyl tetrazole
was selected as the mother scaffold. We also tried attaching
the second phenyl moiety to the meta position of the phenyl
ring of PTZ instead of the ortho position, since the peptide
pharmacophore was positioned in the direction correspond-
ing to the meta position. Interestingly, the resulting 5-
the chemical-shift time scale. From the concentration depend-
ence of the chemical-shift values, the KD value of com-
pound A for GPVI was estimated to be 5.2 Æ 0.4 ꢀ 10À5 m,
which is comparable to the binding affinity of pep-10L (KD =
5.7 ꢀ 10À5 m) and stronger than that of both losartan and the
core fragment, 5-biphenyl-3-yl-2H-tetrazole. The KD value of
compound C was much weaker at 8.5 Æ 0.5 ꢀ 10À4 m.
biphenyl-3-yl-2H-tetrazole showed stronger affinity (Kd
=
8.3 Æ 0.4 ꢀ 10À5 m) than that of 5-biphenyl-2-yl-2H-tetrazole
(see Figure S3). Thus, we used 5-biphenyl-3-yl-2H-tetrazole
as the core fragment for affinity improvement. Then, consid-
ering the spatial accuracy of NOE information and the
flexibility of the Phe9 side chain in the pep-10L structure, we
designed a linker to connect the phenyl ring corresponding to
Phe9 in the peptide in such a way as to allow a certain
ambiguity in length. Furthermore, the following desired
characteristics were taken into account: 1) a simple chemical
structure; 2) rigidity to maintain the spatial arrangement; and
3) ease of chemical synthesis. Compounds A–C with different
linker lengths were synthesized (Scheme 1; see the Support-
ing Information for details), whereby the linker length of
compound B is most consistent with our estimate.
To determine the binding mode of the best compound,
compound A, the INPHARMA experiments were performed
with a mixture of compound A (0.5 mm), pep-10L (0.9 mm),
and GPVI–Fc (25 mm) in D2O at 288 K. The mixing time was
set to 60 and 300 ms. A total of 25 INPHARMA peaks were
observed in the experiment (Table 2). As expected, the PTZ
Table 2: Observed INPHARMA peaks between the hydrogen atoms of
compound A and pep-10L.
Compound A[a]
Pep-10L
H2/H6 or H3/H5
W6 (Ha), L7 (Ha, Hd), Y8 (Ha, Hb), F9 (Ha),
S10 (Ha, Hb)
H10/H14
H11/H13 or H16
H17
W6 (Ha, Hz3), L7 (Hb, Hd), S10 (Ha, Hb)
L7 (Hd), S10 (Ha, Hb)
S8 (Ha)
H18
L7 (Hd), S10 (Ha)
H20
T4 (Hg2), W6 (He3), L7 (Hd), S10 (Ha, Hb)
[a] The numbering of the hydrogen atoms in compound A corresponds
to the carbon-atom numbering in Scheme 1.
moiety in compound A provided INPHARMA peaks with
the side chain of Trp6 and Leu7, which indicates that the
binding mode found for the PTZ moiety of losartan and
GPVI is preserved in compound A (see Figure S4). Although
the INPHARMA peaks between phenyl ring I in com-
pound A and the phenyl ring of Phe9 in pep-10L cannot be
discriminated from the strong intraligand NOE peaks owing
to signal overlap, the INPHARMA peak between phenyl
ring I in compound A and the Phe9 Ha atom was clearly
observed (see Figure S3). This result suggests that phenyl
ring I in compound A is positioned near the side chain of
Phe9 in pep-10L.
A ligand-based NMR spectroscopic approach has the
advantage of nonrestricted molecular size and no require-
ment for isotopic labeling of target proteins. However,
determination of the relative position of distinct compounds
is often difficult by such an approach. Herein, we have
proposed a novel ligand-based strategy for improving the
affinity of small-molecule compounds in which information
gained from a competitive peptide ligand is used to overcome
this problem. The strategy takes advantage of the larger
interaction surface of a peptide ligand and enables the
introduction of additional interaction sites that do not overlap
with a small compound. The INPHARMA information
enables the discrimination of overlapping sites from non-
overlapping sites and would most effectively be used to design
a new compound by a linking or growing approach, as shown
herein. In theory, this approach would be most appropriate
for systems with KD values higher than 10 mm, and the relative
fragment/peptide KD value should be between 0.1 and 10.[6b] It
Scheme 1. Chemical structure of the synthesized compounds. The
carbon atoms of aliphatic and aromatic groups are labeled with
numbers for compound A.
We performed CSP experiments to investigate the binding
affinity of compounds A, B, and C for GPVI. Two-dimen-
sional 1H–15N TROSY-HSQC spectra of the Ig-like domains 1
and 2 in GPVI were recorded with increasing concentrations
of each compound. The maximum concentration was set to
0.8 mm. Except for compound B, which did not show chem-
ical-shift perturbation, each cross-peak in the TROSY-HSQC
spectra of the GPVI/compound mixtures appeared as a single
average resonance that reflected the free/bound ratio and
thus the fast exchange between the free and bound states on
2600 www.angewandte.org
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2014, 53, 2597 –2601

Angewandte
Chemie
[3] H. Jhoti, A. Cleasby, M. Verdonk, G. Williams, Curr. Opin.
would be applicable to a system with a KD value of 1 mm;
however, the sensitivity would be lower. A lower-affinity
fragment would be less problematic; thus, the strategy may be
particularly be useful for the improvement of low-molecular-
weight hit compounds from fragment-based screening into
medium-sized compounds with higher activity. By this
strategy it is also possible to define the active moiety in
compounds to replace nonfunctional groups with new func-
tional groups that may be introduced in a different part of the
original compound. The strategy preserves the advantages of
a ligand-based drug-discovery strategy by NMR spectroscopy
while overcoming the lack of information for the improve-
ment of the affinity of compounds.
Chem. Biol. 2007, 11, 485 – 493.
[4] J. L. Markley, I. Putter, O. Jardetzky, Science 1968, 161, 1249 –
1251; D. A. Torchia, S. W. Sparks, A. Bax, J. Am. Chem. Soc.
1988, 110, 2320 – 2321.
[5] M. Mayer, B. Meyer, J. Am. Chem. Soc. 2001, 123, 6108 – 6117; C.
Dalvit, P. Pevarello, M. Tatꢂ, M. Veronesi, A. Vulpetti, M.
Sundstrçm, J. Biomol. NMR 2000, 18, 65 – 68.
[6] a) V. Sꢃnchez-Pedregal, M. Reese, J. Meiler, M. Blommers, C.
Griesinger, T. Carlomagno, Angew. Chem. 2005, 117, 4244 –
4247; Angew. Chem. Int. Ed. 2005, 44, 4172 – 4175; b) J. Orts,
C. Griesinger, T. Carlomagno, J. Magn. Reson. 2009, 200, 64 – 73.
[7] S. S. Sidhu, Curr. Opin. Biotechnol. 2000, 11, 610 – 616; Y.
Mizukoshi, H. Takahashi, I. Shimada, J. Biomol. NMR 2006, 34,
23 – 30.
[8] S. J. Wrenn, P. B. Harbury, Annu. Rev. Biochem. 2007, 76, 331 –
349; L. Mannocci, M. Leimbacher, M. Wichert, J. Scheuermann,
D. Neri, Chem. Commun. 2011, 47, 12747 – 12753.
[9] P. Vlieghe, V. Lisowski, J. Martinez, M. Khrestchatisky, Drug
Discovery Today 2010, 15, 40 – 56; G. Hummel, U. Reineke, U.
Reimer, Mol. Biosyst. 2006, 2, 499 – 508; B. Bray, Nat. Rev. Drug
Discovery 2003, 2, 587 – 593.
[10] M. Moroi, S. M. Jung, M. Okuma, K. Shinmyozu, J. Clin. Invest.
1989, 84, 1440 – 1445; B. Nieswandt, C. Brakebusch, W. Berg-
meier, V. Schulte, D. Bouvard, R. Mokhtari-Nejad, T. Lindhout,
J. Heemskerk, H. Zirngibl, R. Fꢄssler, EMBO J. 2001, 20, 2120 –
2130.
[11] M. Ungerer, K. Rosport, A. Bꢅltmann, R. Piechatzek, K.
Uhland, P. Schlieper, M. Gawaz, G. Mꢅnch, Circulation 2011,
123, 1891 – 1899.
[12] C. Grothusen, S. Umbreen, I. Konrad, K. Stellos, C. Schulz, B.
Schmidt, E. Kremmer, O. Teebken, S. Massberg, M. Luchtefeld,
B. Schieffer, M. Gawaz, Arterioscler. Thromb. Vasc. Biol. 2007,
27, 1184 – 1190.
[13] K. Kato-Takagaki, Y. Mizukoshi, Y. Yoshizawa, D. Akazawa, Y.
Torii, K. Ono, R. Tanimura, I. Shimada, H. Takahashi, J. Biol.
Chem. 2009, 284, 10720 – 10727.
[14] M. Moroi, S. M. Jung, Thromb. Res. 2004, 114, 221 – 233.
[15] K. Ono, H. Ueda, Y. Yoshizawa, D. Akazawa, R. Tanimura, I.
Shimada, H. Takahashi, J. Med. Chem. 2010, 53, 2087 – 2093.
In most cases, immature compounds in the drug-discovery
stage have affinities for nontarget proteins. This off-target
effect often causes side effects and is one of the major reasons
that bioactive compounds fail to pass clinical trials. Such
failures could be avoided by identifying the fragment
necessary for target binding and excluding unnecessary
moieties as much as possible during compound improvement,
thus improving specificity. The method proposed herein could
be used to efficiently develop such compounds by evaluating
the feasibility of linking or growing strategies in a structure-
guided manner.
Received: December 11, 2013
Published online: January 30, 2014
Keywords: drug discovery · glycoproteins · NMR spectroscopy ·
.
peptides · pharmacophores
[1] P. J. Gane, P. M. Dean, Curr. Opin. Struct. Biol. 2000, 10, 401 –
404.
[2] S. B. Shuker, P. J. Hajduk, R. P. Meadows, S. W. Fesik, Science
1996, 274, 1531 – 1534.
Angew. Chem. Int. Ed. 2014, 53, 2597 –2601
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org 2601