Application of fragment screening and fragment linking to the discovery of novel thrombin inhibitors

Article abstract of DOI:10.1021/jm050850v

The screening of fragments is an alternative approach to high-throughput screening for the identification of leads for therapeutic targets. Fragment hits have been discovered using X-ray crystallographic screening of protein crystals of the serine protease enzyme thrombin. The fragment library was designed to avoid any well-precedented, strongly basic functionality. Screening hits included a novel ligand (3), which binds exclusively to the S2-S4 pocket, in addition to smaller fragments which bind to the S1 pocket. The structure of these protein-ligand complexes are presented. A chemistry strategy to link two such fragments together and to synthesize larger drug-sized compounds resulted in the efficient identification of hybrid inhibitors with nanomolar potency (e.g., 7, IC₅₀ = 3.7 nM). These potent ligands occupy the same area of the active site as previously described peptidic inhibitors, while having very different chemical architecture.

Full text of DOI:10.1021/jm050850v

1346
J. Med. Chem. 2006, 49, 1346-1355
Application of Fragment Screening and Fragment Linking to the Discovery of Novel Thrombin
Inhibitors†
Nigel Howard,§ Chris Abell,§ Wendy Blakemore,‡ Gianni Chessari,‡ Miles Congreve,‡ Steven Howard,*,‡ Harren Jhoti,‡
Christopher W. Murray,‡ Lisa C. A. Seavers,‡ and Rob L. M. van Montfort‡
Astex Therapeutics, 436 Cambridge Science Park, Milton Road, Cambridge CB4 0QA, United Kingdom, and
UniVersity Chemical Laboratory, Lensfield Road, Cambridge, CB2 1EW, United Kingdom
ReceiVed August 26, 2005
The screening of fragments is an alternative approach to high-throughput screening for the identification of
leads for therapeutic targets. Fragment hits have been discovered using X-ray crystallographic screening of
protein crystals of the serine protease enzyme thrombin. The fragment library was designed to avoid any
well-precedented, strongly basic functionality. Screening hits included a novel ligand (3), which binds
exclusively to the S2-S4 pocket, in addition to smaller fragments which bind to the S1 pocket. The structure
of these protein-ligand complexes are presented. A chemistry strategy to link two such fragments together
and to synthesize larger drug-sized compounds resulted in the efficient identification of hybrid inhibitors
with nanomolar potency (e.g., 7, IC50 ) 3.7 nM). These potent ligands occupy the same area of the active
site as previously described peptidic inhibitors, while having very different chemical architecture.
Chart 1. Licensed Drugs that Directly Inhibit Thrombin
Introduction
Thromboembolic disorders are a major cause of mortality in
Western societies. At present, clinical treatment of thrombosis
involves the administration of heparin and its low molecular
weight derivatives or oral anticoagulants of the dicumarol type,
which all indirectly inhibit the trypsin-like serine protease
thrombin. These drugs have limitations with regard to their
efficacy and also their therapeutic index which, being low, leads
to the need for extensive monitoring. Thrombin has been under
intense investigation for over a decade with the aim of
identifying potent inhibitors that might be useful as anticoagu-
lation agents.1-3 An orally active thrombin inhibitor should be
useful for treatment of pathological states characterized by
thrombosis, such as deep vein thrombosis and stroke.4 Thrombin
acts in the blood coagulation cascade by catalyzing the
conversion of fibrinogen to fibrin and also converting factor
XIII to factor XIIIa, which then cross-links the fibrin clot.
Thrombin also activates upstream zymogens factors V, VIII,
and XI, which then further accelerate the clotting cascade by
thrombin synthesis.5
An injectable form of a direct thrombin inhibitor, argatroban,
(Chart 1) has been approved by the FDA, but only for the
relatively rare condition of heparin-induced thrombocytopenia.6
Very recently, the first orally active thrombin inhibitor, ximela-
gatran (Exanta), was licensed for the treatment of venous
thromboembolic events in Europe. Ximelagatran is a double-
prodrug molecule (Chart 1) and is converted in vivo to the active
parent (melagatran) that contains a highly basic amidine
functionality which is key to inhibitor potency through salt
bridge formation with aspartate 189 in the S1 specificity
pocket.7,8 This amidine group has been incorporated in many
classes of thrombin inhibitors, but while such compounds
demonstrate good in vivo efficacy, their use is generally limited
by poor oral bioavailability. Although a prodrug approach offers
one solution to this problem, the challenge over recent years
has been to develop second-generation compounds that lack a
highly basic P1 motif and therefore have superior druglike
properties.9
Over the last 20 years there has been considerable interest in
new approaches to drug discovery that offer improvements in
the process of identifying new therapeutic agents. Large
investments in platform technologies such as high-throughput
screening and combinatorial chemistry have been made across
the pharmaceutical industry. One area that has received sig-
nificant attention is the drug-likeness of hits and leads, because
the quality of lead compounds is thought to have a major impact
on the attrition rates in drug development.10-12 More recently,
interest has grown in a different approach to lead generation
that involves screening libraries of molecules that are much
smaller and functionally simpler than drugs themselves, often
referred to as “fragment-based” drug discovery.13-21 Fragments
are low molecular weight compounds (typically 100-250 Da)
and as a consequence will usually exhibit low binding affinities
(100 µM-mM). A fragment-based approach is argued to have
advantages over conventional screening methods for several
reasons.12,18 First, the number of compounds that need to be
screened to find useful hits is in the order of only a few
hundreds, because the lower complexity compounds (fragments)
have a higher probability of matching a target protein binding
† Coordinates of the thrombin complexes with compounds 1-5 and
compound 7 have been deposited in the Protein Data Bank under accession
codes 2c90, 2c8z, 2c8y, 2c93, 2c8x, and 2c8w. The corresponding structure
factor files have been deposited under accession codes r2c90sf, r2c8zsf,
r2c8ysf, r2c93sf, r2c8xsf, and r2c8wsf.
* To whom correspondence should be addressed. Phone: +44 (0)1223
226209. Fax: +44 (0)1223 226201. E-mail: S.Howard@astex-
therapeutics.com.
‡ Astex Therapeutics.
§ University Chemical Laboratory.
10.1021/jm050850v CCC: $33.50 © 2006 American Chemical Society
Published on Web 01/20/2006

DiscoVery of NoVel Thrombin Inhibitors
Journal of Medicinal Chemistry, 2006, Vol. 49, No. 4 1347
site. Second, a higher proportion of the atoms in a low molecular
weight fragment hit are directly involved in the desired protein-
ligand binding interaction, so fragments can be seen as more
efficient binders when compared with conventional (HTS)
screening hits. Last, when guided by protein-ligand crystal
structures, fewer molecules will need to be synthesized in order
to optimize the hits to potent leads, and the chemical synthesis
approach itself will often be simple and straightforward due to
the low level of complexity of the initial fragment hits.
Recently, Congreve et al. analyzed a diverse set of fragment
hits that had been identified against a range of targets.22 This
study indicated that such hits seemed to obey, on average, a
“rule of three” in which molecular weight is <300, the number
of hydrogen bond donors and acceptors is, in each case, e3,
and clogP e 3. The authors suggest that this “rule of three”
provides a useful guideline for physicochemical properties when
designing fragment-based screening sets.
Due to the low binding affinity of fragments they are usually
difficult to detect using bioassay-based screening methods. A
variety of alternative biophysical methods have therefore been
used to detect the binding of such fragments, and the area has
recently been reviewed.18-20,23 Recently, Nienaber et al. identi-
fied novel inhibitors for urokinase using protein X-ray crystal-
lographic screening.21 In our laboratories we have developed
methodologies for screening fragments, that we call “Pyramid”,
which include the high-throughput X-ray protein crystallographic
screening of fragment libraries.24,25 In our previous reports the
subsequent optimization of hits identified from fragment screen-
ing has been based on a “fragment-growing” strategy. This
involves using structure-based drug design to form additional
interactions by growing out from the starting fragment. In this
current work, we describe an example of an alternative,
“fragment-linking”, approach. This general approach has been
described by others and is associated with its own particular
set of difficulties.13,14,26
using virtual screening as described below. Third, thrombin
crystals were soaked in compound mixtures, followed by
elucidation of the protein-ligand structures. Fragment hits were
identified in Fo-Fc electron density maps calculated after initial
automatic refinement against an unliganded thrombin structure
and fitted and scored using AutoSolve.17 Protein-ligand
structures of bound fragments were then subjected to further
refinement steps. The output from the Pyramid process was a
set of experimentally determined binding modes for thrombin-
bound fragments which were then manually examined using
AstexViewer.28 Ligands chosen for hit-to-lead chemistry were
selected based on a number of criteria which are described
below.
Construction of a Focused Fragment Library for Throm-
bin. Virtual screening was used to build a focused set of
fragments to screen against thrombin. Filtering on the basis of
1D and 2D properties was employed to create databases of
compounds from the Astex Therapeutics library of available
substances (ATLAS).29 1D and 2D properties included heavy
atom count (HAC), number of hydrogen bond donors (NDON),
and clogP. Compounds in these databases were subsequently
docked against different protein conformations of thrombin using
a proprietary version of the program GOLD.30 This method has
been previously reported by Hartshorn et al.24 Pharmacophores
were designed to reward the placement of hydrophobic groups
in the S1 pocket and subsequently applied during selected
dockings. Goldscore31 and Chemscore32 scoring functions were
used to score and to rank the different fragments.
This approach generated a library of 80 fragments, which
could be categorized as follows:
(1) Charged S1 binders (HAC e 13, NDON e 4, clogP e
3): fragments carrying a weakly basic group able to interact
with Asp189 (e.g., aminopyridine and aminoquinoline frag-
ments). Strong bases and obvious arginine mimetics were
avoided.
We have already reported our preliminary findings of
fragment-based crystallographic screening against thrombin in
a previous publication.24 We now describe the fragment hits
identified from Pyramid screening in more detail and how the
subsequent structure-based fragment-linking strategy led to the
identification of novel, potent, thrombin inhibitors. In the screen
we soaked libraries of fragments into thrombin crystals with
the aim of identifying new leads for this important drug target.
In particular, our rationale was to exclude fragments that
contained a strongly basic motif, such as an amidine or
guanidine, and instead to focus on identifying hits with better
druglike properties. We also wished to establish our approach
as a useful method for identifying potent ligands for a
representative protease target. As a class of enzymes, proteases
have often been challenging as drug discovery targets since leads
against these targets are often large and peptidic in nature,
making the development of druglike inhibitors difficult.27
(2) Neutral S1 binders (HAC e 13, NDON e 1, clogP e
3.5): uncharged fragments able to bind deeply in the S1 pocket
through hydrophobic interaction (e.g., substituted chlorophenyl
and indole fragments). Although there were no neutral fragments
reported that bind to thrombin, drug-sized inhibitors with neutral
S1 binding groups have been described, suggesting that such
fragments might be identified.33-35
(3) Neutral S1 binders with a solubilizing group (HAC e
18, NDON e 3, clogP e 2.5): in order to test some uncharged
scaffolds it was necessary to introduce a solubilizing group to
make screening at high concentrations possible.
(4) S1-S1â and larger S1-S2 binders (HAC g 16, NDON
< 5, clogP < 5): a set of slightly larger compounds selected
with the aim of simultaneously targeting the S1 pocket and either
the S1â pocket or the S2 pocket of the active site.
Thrombin Crystal Structure Determinations. Crystal-
lographic screening of the focused thrombin set was carried out
on crystals of R-thrombin complexed with the peptidic inhibitor
hirugen which renders the protein inactive by binding to the
thrombin anion-binding exosite. However, hirugen binding does
not obstruct the thrombin active site, thus providing a crystal
system suitable for fragment soaking.36 To maximize throughput
and minimize subsequent deconvolution experiments, Pyramid
screening of the targeted library was carried out using cocktails
(mixtures) of two fragments. Even after overnight soaking in
high concentrations of fragments, the thrombin-hirugen crystals
generally diffracted to about 2.0 Å resolution, which allowed
high-throughput X-ray data collection on in-house equipment
using an automatic sample-changing robot ACTOR.37
Results and Discussion
Overview of Pyramid Screening. The Pyramid screening
approach and its application to a number or targets, including
thrombin, has been described previously.24 Further details on
specific steps in the application to thrombin are described in
other sections. In brief, thrombin screening consisted of the
following steps: First, thrombin crystals suitable for soaking
experiments were generated, and the robustness of soaking
techniques against a variety of fragments and fragment mixtures
was demonstrated. Second, a library of 80 fragments was defined

1348 Journal of Medicinal Chemistry, 2006, Vol. 49, No. 4
Howard et al.
Overview of Screening Results. Although fragments often
have low affinity, they usually exhibit a high ligand efficiency,
i.e., a high value for the ratio of free energy of binding to the
number of heavy atoms.38,39 This concept of ligand efficiency
was one of the factors used to help prioritize the fragments
selected for hits-to-leads chemistry. Overall, hits from the
fragment screening were selected based upon the following
criteria: moderate to good ligand efficiency, good druglike
properties (such as neutral or weakly basic, nonpeptidic
compounds), and potential for rapid optimization to potent leads
based on a consideration of their protein-ligand binding modes.
The Pyramid screen identified several fragments bound to
the S1 pocket of the enzyme. In each case, the crystal structure
of the thrombin-bound fragment was solved and demonstrated
a clearly defined binding mode by fitting to electron density in
the active site. The hits included weakly basic fragments, such
as aminopyridines and aminoquinolines, and also neutral frag-
ments, such as chlorophenyl and indole-based compounds. For
the purpose of this work, two chlorophenyl fragments, 1 (Figure
1a) and 2 (Figure 2a) were chosen for chemistry optimization.
Crystal structures of thrombin containing uncharged chloro-
phenyl P1 motifs have (as noted earlier) been previously reported
in the literature, but only in the context of larger, more potent,
inhibitors.33-35 The detection of small uncharged fragments, such
as compound 1, binding to the S1 pocket, without the require-
ment for additional hydrogen bonding in adjacent subsites, is a
key finding. This provides validation of high-throughput X-ray
crystallography’s ability to detect fragments driven primarily
by hydrophobic binding to active-site pockets.
Figure 1a shows the crystal structure of the chlorophenyl-
tetrazole fragment 1. The chlorophenyl group binds deeply in
the S1 pocket with the chlorine atom located above the aromatic
ring of the Tyr228 side chain. The tetrazole ring makes close
contact (3.6 Å) with the sulfur atom of Cys220 in the S1â
pocket, possibly forming a donor-atom-π interaction.35 These
type of interactions have been identified in other systems such
as flavoenzymes.40,41 The higher in vitro potency of compound
1 (IC50 ) 330 µM) made this compound the most potent, and
also ligand efficient, of the uncharged S1 binders identified from
the fragment screen.
Figure 1. Fragment hits from thrombin Pyramid screening. (a) Crystal
structure showing the binding of 1 in the thrombin S1 pocket at 2.3 Å
resolution. 1 is shown in blue with the initial Fo-Fc electron density
identifying the presence of the ligand superimposed in light blue and
contoured at 3σ. Key protein residues are shown with carbon atoms in
green, nitrogen in blue, oxygen in red, and sulfur in yellow. (b) Crystal
structure showing the binding of 3 in the thrombin S2-S4 pockets at
2.2 Å resolution. Ligand and protein residues are shown using the same
color scheme as in panel a, and the electron density is contoured at
3σ. Chemical structures of the fragment hits are shown on the left-
hand side of the panels. All figures of protein-ligand structures were
made with AstexViewer. IC50 values are given as the average of two
or more determinations.
Figure 2a shows the binding mode of compound 2. The
chlorophenyl group binds in the S1 pocket in an analogous
fashion to 1 (Figure 1a). The amino group forms hydrogen bonds
with the hydroxyl group of the catalytic Ser195, with the
backbone carbonyl of Ser214, and also interacts via water
mediation with Gly216. Although compound 2 was insuf-
ficiently soluble under the assay conditions for an IC50 to be
measured accurately (IC50 > 1 mM), its very simple structure
makes it a synthetically attractive starting point for hit-to-lead
chemistry.
Moreover, the hydroxyl group and the sulfonamide NH formed
hydrogen bonds with Gly216. Similar hydrogen-bonding inter-
actions to Gly216 are commonly observed with many peptido-
mimetics thrombin inhibitors, including melagatran and arga-
troban. The folded conformation adopted by 3 was analyzed
using molecular dynamics calculations in explicit water at room
temperature.42 It was observed that after 55 ps of dynamics
simulation, the structure of compound 3 changed from a high-
energy open conformation (rmsd between the X-ray structure
and the open conformation was 3.4 Å) to a low-energy folded
conformation, similar to the one observed in the crystal structure
(rmsd equal to 1.3 Å). These results suggested that S2-S4
binder 3 may adopt a prearranged, folded, conformation in
solution which is similar to the observed bioactive form. This
phenomenon, often termed “hydrophobic collapse”, has been
described for a number of important ligands, including thrombin
inhibitors.43,44 Due to the novel binding mode and the nonpep-
tidic nature of this molecule, compound 3 offered an attractive
start-point for chemistry optimization.
The Pyramid screening approach not only identified S1
binders but also identified a larger compound, 3, which
unexpectedly bound to the S2-S4 region, while leaving the S1
pocket vacant (Figure 1b). To our knowledge the crystal
structure of compound 3 bound to thrombin provides the first
reported structural characterization of a small molecule binding
exclusively in the enzyme’s S2 and S4 pockets. Compound 3
had been included in the screening set following the initial
GOLD-based virtual screening which had placed the benzyl
motif of this compound in the S1 pocket. The structure of
compound 3 complexed with thrombin showed that the com-
pound bound in a folded conformation where the naphthyl and
the benzyl groups formed an intramolecular aromatic interaction
while binding to the S4 and the S2 pockets, respectively.

DiscoVery of NoVel Thrombin Inhibitors
Journal of Medicinal Chemistry, 2006, Vol. 49, No. 4 1349
benzenesulfonamide analogue 4 was found to be the most potent
compound (IC50 ) 12 µM, Figure 2b). This substituent had
previously been identified for another class of thrombin inhibi-
tors.45 The crystal structure of compound 4 bound to thrombin
was solved and revealed a similar binding mode to S2-S4
binder 3 with the methoxy-trimethylbenzene group showing
exceptional shape complementarity with the S4 pocket. In
addition, compound 4 demonstrated good selectivity over other
serine proteases such as trypsin and FXa, being essentially
inactive at 1 mM.
Having identified different ligands which bind in adjacent
pockets, a “fragment-linking” approach seemed to present an
attractive strategy, particularly since the global protein structure
did not change upon ligand binding. It has been well established
that there should be a “superadditivity” of potency when two
fragments are linked in an optimal fashion.46,47 Achieving this
in practice is often very difficult, but there are a number of
cases reported in which two fragments have been successfully
linked together to give highly potent ligands. For example,
NMR-directed fragment linking has been used to identify potent
ligands for a number of targets, including the FK506-binding
protein, stromelysin, and protein tyrosine phosphatase 1B
(PTP1B).13,14,26
To define the best fragment-linking strategy, thrombin
structures which contained a fragment in the S1 pocket were
overlaid with the structure of the optimized S2-S4 binder 4 in
the same frame of reference. The overlay allowed direct
comparison between the different binding modes and the
possible attachment points between fragments. Compound 4
contains a hydroxyl group and a primary amine which are
directed toward the S1 pocket and therefore provided attractive
points through which S2-S4 and S1 fragments could be linked.
Guided by structure-based design and consideration of synthetic
tractability, the S2-S4 binder 4 was linked through the basic
nitrogen to the meta position of the chlorophenyl ring of the
S1 fragment 2 via an aminomethyl linker. The hybrid compound
5 (IC50 ) 220 nM) demonstrated a 50-fold improvement in
affinity over compound 4 and was at least 1000-fold selective
over trypsin (5: IC50 trypsin > 100 µM). The fragment-linking
strategy is summarized in Figure 2a-c which illustrates the
crystal structures of compounds 2, 4, and 5 respectively.
The crystal structure of compound 5 bound to thrombin shows
the benzyl group and the phenyl sulfonamide occupying the S2
and S4 pockets, respectively, as observed in compound 4, with
the sulfonamide NH hydrogen-bonded to the carbonyl of
Gly216. The secondary amine points toward Gly216 in order
to allow the 3-chlorophenyl to bind to the S1 pocket, as seen
previously for S1 fragment 2. Finally, the hydroxyl group is
directed away from Gly216, and the hydrogen bond to the
protein, as observed in compound 4, is absent.
Figure 2. Fragment linking of thrombin S1 and S2-S4 binders. (a)
Binding of a chlorophenyl fragment in the thrombin S1 pocket revealed
at 2.2 Å resolution. (b) Crystal structure of the S2-S4 binder 4 at 2.2
Å resolution. (c) Crystal structure of 5, designed by linking of 2 and 4.
In all panels the color scheme and electron density contour levels are
the same as in Figure 1. IC50 values are given as the average of two or
more determinations.
Although a 50-fold increase in potency of the hybrid
compound 5 over the S2-S4 binder 4 was achieved, the
improvement was not of the magnitude expected for optimally
linked fragments.46 The conformational changes observed for
compound 5 with the concomitant loss of one hydrogen bond
may go some way to explain this. To investigate further
optimization, we turned our attention to fragment linking with
the more potent S1 fragment 1 (IC50 ) 330 µM). Two hybrid
compounds were designed and synthesized, one analogue 6 with
the fully substituted P4 motif (methoxy-trimethylbenzene), and
a second analogue 7 with a less substituted P4 motif (p-
methoxybenzene) aimed at minimizing molecular weight, lipo-
philicity, and molecular complexity (Chart 2). Compound 6 (IC50
) 1.4 nM) was the most potent analogue identified, demonstrat-
Hit Optimization. The results from fragment screening
against thrombin demonstrate the ability of this approach to
probe different pockets within the active site. Furthermore, the
high-quality structural information generated by the Pyramid
screen can then be used to facilitate a rapid hit-to-lead
optimization phase. We first looked at optimizing the novel S2-
S4 binder, 3. Analysis of the crystal structure of compound 3
suggested the naphthyl motif was suboptimal with regard to its
steric complementarity with the S4 pocket. A small number of
analogues aimed at modifying the sulfonamide motif were
designed and synthesized. Of these, the methoxy-trimethyl-

1350 Journal of Medicinal Chemistry, 2006, Vol. 49, No. 4
Howard et al.
Chart 2a
both compounds present aromatic motifs which bind to the S1
pocket, although these groups are chemically quite different.
In the case of compound 7, P1 is a nonbasic chlorophenyl group,
whereas melagatran presents a basic benzamidine motif. Al-
though they occupy the same active-site region, the two
compounds are based on very different scaffolds. Melagatran
is a pseudo-peptide, whereas compound 7 is nonpeptidic.
In this example, fragment linking has produced the potent
inhibitor 6 (IC50 ) 1.4 nM) which can be attributed to the
binding affinity of the initial fragments being combined in an
additive manner. However, the “superadditivity” which could
theoretically be expected from optimal linked fragments was
not observed. This lack of optimal ligand efficiency can be
ascribed to the fact that the interactions of the starting hits were
not completely retained in the fragment-linked compounds. This
result fits with the assertion that linking two fragments in an
optimal fashion is often very difficult in practice and, further-
more, that large molecular weight inhibitors are often a
consequence of such an approach.
An alternative strategy to fragment linking is to start from a
fragment with high ligand efficiency and to use structure-based
drug design to form additional interactions by growing out from
the fragment. It has been demonstrated elsewhere that the
fragment-growing approach can maintain ligand efficiency and
thus identify attractive lead molecules.39 Although both ap-
proaches can deliver potent inhibitors, a fragment-growing
approach is likely to deliver compounds which are better starting
points for lead optimization.
a Average of two or more determinations.
Conclusions
The fragment-screening and linking strategy has facilitated
the rapid identification of novel nanomolar inhibitors of
thrombin which are nonpeptidic and lack a highly basic
guanidine/amidine mimetic. Only 80 targeted fragments were
screened, and less than 40 compounds were synthesized.
Fragment hits with novel binding modes, e.g., compound 3, were
discovered which demonstrates the potential of fragment-based
screening to probe beyond just the key S1 specificity pocket in
thrombin.
It is recognized that compound 7 is not ideal with respect to
Lipinski’s molecular weight guidelines for an oral drug.
Nonetheless, this compound presents a useful starting point for
further optimization strategies. Optimization of this series and
a comparison with potent lead compounds identified from a
fragment-growing approach will be the topic of a future
publication.
Figure 3. (a) Binding of 7. Crystal structure showing the binding of
7 at 2.0 Å resolution. In all panels the color scheme and electron density
contour levels are the same as in Figure 1. (b) Binding mode of 7
overlaid with melagatran.
ing a 200-fold improvement in affinity over compound 5, and
the simplified analogue 7 (IC50 ) 3.7 nM) was also only
marginally less potent. The crystal structure of the complex
between thrombin and analogue 7 (Figure 3a) showed that the
relative binding modes of the P1, P2, and P4 functionalities
were analogous to those observed in both the initial fragments
and in compound 5. However, some notable differences were
again observed, particularly regarding the conformation of the
ethanolamine linker. Rotation around the C1-C2 and C2-C3
bonds of the ethanolamine chain in compound 7 causes both
the hydroxyl group and the sulfonamide NH to point away from
the Gly216 backbone residue, thereby preventing hydrogen
bonding to Gly216. The binding of compound 7 to thrombin
appears to be driven purely by hydrophobic interactions with
no obvious contribution from hydrogen bonding.
Experimental Section
Crystallography. Freeze-dried human prothrombin and the
hirugen peptide were purchased from Enzyme Research Labora-
tories, South Bend, IN and Bachem, Bubendorf Switzerland,
respectively. Human R-thrombin was activated and purified using
a modified method described by Ngai and Chang.48 Crystals of the
R-thrombin-hirugen complex were obtained by a procedure based
on the method of Skrzypczak-Jankun et al.36 In brief, human
R-thrombin and the hirugen peptide were complexed in a 1:3 molar
ratio in 20 mM Tris/HCl pH 7.5 and 250 mM NaCl, and the
thrombin-hirugen complex was concentrated to a final concentra-
tion of 10 mg/mL. Initial thrombin-hirugen crystals were grown
by vapor diffusion using drops composed of a 1:1 mixture of protein
and reservoir solution, equilibrated over reservoirs containing 28-
33% PEG4000, 0.1 M phosphate buffer pH 7.3 and 0-0.3 M NaCl.
Diffraction quality crystals could be obtained by microseeding,
whereby drops composed of a 1:1 mixture of protein solution and
thrombin-hirugen seed crystals in 30% PEG4000 and 0.1 M
phosphate buffer pH 7.3 were equilibrated against reservoirs of 23%
The binding mode of compound 7 is compared with that of
melagatran in Figure 3b. An overlay of the two structures reveals
some similarities with respect to the key P1, P2, and P4 motifs
of the two compounds. Both compounds present hydrophobic
groups which occupy the S2 and S4 pockets, although, in the
case of melagatran, the P2 motif is relatively small. Furthermore,

DiscoVery of NoVel Thrombin Inhibitors
Journal of Medicinal Chemistry, 2006, Vol. 49, No. 4 1351
Table 1. Crystallographic Data Collection and Refinement Statistics
compound
1
2
3
4
5
7
Data Collection
X-ray source
resolution (Å)
in house, λ ) 1.54 Å
2.3
2.1
2.2
2.2
2.0
2.0
no. unique reflections
completeness (%)a
ave multiplicity
Rmergea,b
15571
97.3 (96.3)
1.96
17595
95.4(86.3)
1.96
18049
99.7 (99.1)
2.91
17066
97.4(92.7)
1.97
17227
93.6(87.3)
2.07
23682
97.9 (90.0)
3.3
0.136(0.38)
0.081(0.26)
0.133(0.44)
0.061(0.16)
0.044(0.11)
0.060(0.15)
Refinement
Rcryst
Rfree
0.220
0.306
0.011
1.4
39.2
35.3
40.2
0.176
0.246
0.011
1.3
34.7
36.4
43.8
0.185
0.259
0.012
1.4
33.5
32.2
40.0
0.164
0.242
0.011
1.4
21.4
15.3
29.5
0.153
0.229
0.012
1.4
23.9
25.8
33.7
0.164
0.219
0.011
1.3
20.8
34.1
19.6
rmsd bond lengths (Å)
rmsd bond angles (deg)
ave B-factor protein (Å2)
ave B-factor ligand (Å2)
ave B-factor solvent (Å2)
a Numbers in parentheses indicate the highest shell values. b Rmerge ) ∑h∑i|I(h,i) - I (h)|/∑h∑i I (h); I(h,i) is the scaled intensity of the ith observation
of reflection h, and (h) is the mean value. Summation is over all measurements. Rcryst ) ∑hkl,work||Fobs| - k|Fcalc||/∑hkl|Fobs|, where Fobs and Fcalc are the
observed and calculated structure factors, k is a weighting factor, and work denotes the working set of 95% of the reflections used in the refinement. Rfree
) ∑hkl,test||Fobs| - k|Fcalc||/∑hkl|Fobs|, where Fobs and Fcalc are the observed and calculated structure factors, k is a weighting factor, and test denotes the test
set of 5% of the reflections used in cross validation of the refinement. λ refers to wavelength, rmsd to root-mean-square deviations. All ligands have been
refined with occupancies of 1.0.
PEG4000, 0.1 M phosphate buffer pH 7.3 and 0-0.3 M NaCl at
room temperature. Data collection and refinement statistics for
selected crystal structures are presented in Table 1.
ensemble) using the program Discover. The system consisted of 3
solvated with 236 water molecules placed in a cubic box (with
dimensions 20 Å × 20 Å × 20 Å). All atoms of the system were
considered explicitly, and their interaction were computed using
CFF91 force field53 using a dielectric constant equal to 1. First a
500 iteration potential energy minimization was carried out with
harmonic restraint on 3, followed by 3ps molecular dynamics at
298 K. A final dynamics production run of 500 ps (2 fs step) was
carried out at 298 K with no restraint on the ligand. Snapshots from
the trajectories were saved every 0.5 ps.
Chemistry. The compounds 5, 6, and 7 were synthesized from
the amino-alcohol 8 (Scheme 1), which can be prepared in five
steps from L-phenylalanine.54 Sulfonamides 9a and 9b were
prepared by treatment of the amino-alcohol 8 with the corresponding
sulfonyl chloride (Scheme 1). Removal of the Boc-protecting group,
followed by reductive amination with the relevant benzaldehyde,
gave the final products 5, 6, and 7. 5-Chloro-2-tetrazol-1-yl-
benzaldehyde 14 was synthesized from 2-amino-5-chlorobenzoic
acid 11 in three steps (Scheme 2). Tetrazole 12 was prepared from
amine 11 using a modification of the procedure described by Young
et al.35 Subsequent PyBOP-mediateda coupling with N,O-dimeth-
ylhydroxylamine hydrochloride gave the Weinreb amide 13 which,
upon treatment with lithium aluminum hydride, gave the aldehyde
14.
All reactions were carried out under a N2 atmosphere with
commercial grade reagents from Sigma-Aldrich-Fluka, Avocado,
Breckland, or Novabiochem. Solvents were distilled prior to use
or commercially available anhydrous solvents used (Aldrich). The
1H and 13C NMR spectra were recorded on a Bruker DPX400N
spectrometer. Chemical shifts are reported in ppm relative to
residual solvent peak. Flash silica chromatography was performed
using Breckland silica gel 60 (230-400 mesh). Analytical LC-
MS was performed using either method A (Waters 2795 separation
module/Waters micromass ZQ detector with a Waters Atlantis dC18
4.6 mm × 30 mm, 3 µm column and a mobile phase of acetonitrile/
10 mM aq ammonium acetate with 0.1% formic acid) or method
B (Waters 2795 separation module/Waters 2996 PDA detector with
a Phenomenex Synergi 4 µm MAX-RP 80A, 2.0 mm × 50 mm
column and a mobile phase of acetonitrile/water with 0.1% formic
acid). Purification by HPLC was carried out using a Waters
Fractionlynx/Waters ZQ LC/MS platform with a Phenomenex
Synergy MAX-RP, 10 µm, 150 mm × 15 mm column and a mobile
phase of acetonitrile/water with 0.1% formic acid.
All soaking experiments were carried out overnight in a standard
mother liquor consisting of 30% PEG4000 and 0.1 M phosphate
buffer pH 7.3. Data were collected using our in-house equipment
on either a Rigaku-MSC Jupiter CCD or an R-AXIS IV image plate
detector mounted on an RU-H3R rotating anode generator equipped
with Osmic confocal multilayer optics. All data sets were processed
with d*trek49 and scaled using SCALA.50 Structure solution and
initial refinement of the protein-ligand complexes were carried
out with our automated scripts using a modified thrombin-hirugen
structure (PDB id code 1QJ1) as a starting model. Bound ligands
were identified and fitted in the Fo-Fc electron density using
AutoSolve and further refined using our automated scripts followed
by rounds of rebuilding and refinement using Refmac.50
Virtual Screening. In thrombin, the side chain conformation of
Glu192 can vary significantly changing the size and the shape of
the pockets S1 and S1â.51 To take these conformational changes
into account, we decided to virtually screen compounds against
two structures: PDB entry 1ETS and a modified version of 1ETS
where the Glu192 side chain was moved away from the S1â pocket.
For each protein structure two binding sites were constructed: a
large binding site constituted by all protein atoms within 6 Å of an
atom in the 1ETS ligand and a small binding site constituted by all
protein atoms within 6 Å of an atom in the benzamidine substructure
contained in the 1ETS ligand. All the virtual screens were run on
a Linux cluster using the Astex web-based virtual screening
platform29 using methods and settings previously described by
Verdonk et al.52 A number of filters (heavy atoms count, MW,
number of donors, number of acceptors, ClogP, and PSA) combined
with the scoring functions Goldscore and Chemscore were applied
to score and rank the docked libraries. Finally, virtual screens were
visually analyzed, and fragments were selected based on the binding
mode, the chemical tractability, and the score.
The S1 hydrophobic pharmacophore was designed using the
Factor Xa structure 1MQ5, where the ligand contains a chlorophenyl
ring buried deeply in the S1 pocket. The S1 pocket residues of the
Factor Xa structure 1MQ5 were overlaid with the respective S1
pocket residues in the thrombin structure 1ETS. The coordinates
of chlorine atom in 1MQ5 were chosen to define the center of the
pharmacophore. Subsequently, dockings experiments of chloroben-
zene fragments against the thrombin structures 1ETS were used to
optimize the pharmacophore radius to 2.5 Å and the pharmacophore
weight to 10.
a Abbreviations: DMF, N,N-dimethylformamide; EtOAc, ethyl acetate;
PyBOP, (benzotriazol-1-yloxy)tripyrrolidinophosphonium hexafluorophos-
phate.
Molecular Dynamics. The molecular dynamics simulation on
3 was performed with cubic periodic boundary condition (NVT

1352 Journal of Medicinal Chemistry, 2006, Vol. 49, No. 4
Scheme 1. Synthesis of Inhibitors 5, 6, and 7a
Howard et al.
a Reagents: (a) Mtr-Cl or 4-MeOC6H6SO2Cl, iPr2EtN, DMF; (b) TFA, CH2Cl2; (c) 3-chlorobenzaldehyde, NaCNBH3, MeOH, AcOH; (d) 14, NaCNBH3,
MeOH, AcOH.
Scheme 2. Synthesis of Aldehyde 14a
i
a Reagents: (a) CH(OMe)3, NaN3, AcOH, 88%; (b) MeNHOMe‚HCl, PyBOP, Pr2EtN, CH2Cl2, 83%; (c) LiAlH4, THF, -78 °C, 94%.
(1S,2R)-[1-Benzyl-2-hydroxy-3-(4-methoxy-benzenesulfon-
ylamino)-propyl]-carbamic Acid tert-Butyl Ester (9a). A solution
of 4-methoxy-benzenesulfonyl chloride (427 mg, 2.07 mmol) in
DMF (5 mL) was added dropwise to a solution of (1S,2R)-(3-amino-
1-benzyl-2-hydroxy-propyl)-carbamic acid tert-butyl ester 8 (527
mg, 1.88 mmol) and N,N-diisopropylethylamine (655 µL, 3.76
mmol) in DMF (15 mL) over the period of 10 min. The reaction
was stirred for 18 h. The mixture was concentrated in vacuo, the
residue was dissolved in EtOAc (25 mL) and washed with water
(25 mL), 10% aq KHSO4 (25 mL), saturated aq NaHCO3 (25 mL),
and saturated brine (25 mL). The organic phase was dried (Na2-
SO4), filtered, and concentrated in vacuo. The residue was purified
by silica chromatography with hexane/EtOAc (1.25:1) to give the
title compound 9a, a white solid, 531 mg, 63%. 1H NMR (CDCl3,
400 MHz): δ 7.78 (m, 2H), 7.16-7.34 (m, 5H), 6.94 (m, 2H),
5.74 (br, 1H), 4.42 (d, J ) 8.2 Hz, 1H), 3.84 (s, 3H), 3.77 (dt, J )
4.6, 8.2 Hz, 1H), 3.45 (br, 1H), 3.15 (ddd, J ) 3.1, 9.1, 12.7 Hz,
1H), 3.01 (dd, J ) 4.6, 14.2 Hz, 1H), 2.85 (m, 2H), 1.29 (s, 9H)
ppm. 13C NMR (CDCl3, 100 MHz): δ 163.3, 157.2, 137.4, 131.7,
129.8, 129.6, 129.2, 127.1, 114.7, 80.7, 71.8, 56.0, 53.8, 45.8, 36.7,
28.5 ppm. LC-MS (method A) m/z (ES+) 351 [M - Boc + H], Rt
) 4.12 min.
(1S,2R)-[1-Benzyl-2-hydroxy-3-(4-methoxy-2,3,6-trimethyl-
benzenesulfonylamino)-propyl]-carbamic Acid tert-Butyl Ester
(9b). Prepared in an analogous fashion to 9a using 8 (89 mg, 303
µmol) and 4-methoxy-2,3,6-trimethyl-benzenesulfonyl chloride to
give the title compound 9b, a white solid, 95 mg, 61%.1H NMR
(CDCl3, 400 MHz): δ 7.18-7.34 (m, 5H), 6.58 (s, 1H), 5.96 (dd,
J ) 3.9, 7.8 Hz, 1H), 4.44 (d, J ) 8.4 Hz, 1H), 3.85 (s, 3H), 3.82
(m, 1H), 3.42 (br, 2H), 3.05 (m, 2H), 2.87 (m, 2H), 2.68 (s, 3H),
2.60 (s, 3H), 2.14 (s, 3H), 1.33 (s, 9H) ppm. 13C NMR (CDCl3,
100 MHz): δ 158.1, 155.5, 137.4, 135.9, 133.0, 128.2, 127.8, 127.4,
125.5, 124.1, 110.9, 79.0, 70.3, 54.3, 52.0, 43.3, 35.1, 27.0, 23.0,
16.5, 10.8 ppm. LC-MS (method B) m/z (ES+) 493 [M + H],
(ES-) 491 [M - H], Rt ) 3.45 min.
(2R,3S)-N-(3-Amino-2-hydroxy-4-phenyl-butyl)-4-methoxy-
benzenesulfonamide (10). (1S,2R)-[Benzyl-2-hydroxy-3-(4-meth-
oxy-benzenesulfonylamino)-propyl]-carbamic acid tert-butyl ester
9a (347 mg, 770 µmol) was dissolved in (1:1) trifluoroacetic acid/
CH2Cl2 (10 mL), and stirred for 2 h, then concentrated in vacuo.
The residue was partitioned between EtOAc (20 mL) and (1:1)
saturated aq NaCO3/water (20 mL). The organic phase was dried
(Na2SO4), filtered, and concentrated in vacuo. The residue was
purified by silica chromatography with methanol/CH2Cl2 (1:9) to
give the title compound 10, a white foam solid, 187 mg, 69%.1H
NMR (CD3OD, 400 MHz): δ 7.80 (m, 2H), 7.29 (m, 2H), 7.22
(m, 3H), 7.06 (m, 2H), 3.86 (s, 3H), 3.60 (td, J ) 4.7, 6.8 Hz, 1H),
3.15 (td, J ) 4.7, 9.5 Hz, 1H), 3.06 (dd, J ) 4.7, 13.3 Hz, 1H),
2.98 (dd, J ) 4.7, 13.8 Hz, 1H), 2.92 (dd, J ) 6.8, 13.3 Hz, 1H),
2.52 (dd, J ) 9.5, 13.8 Hz, 1H) ppm. 13C NMR (CD3OD, 100
MHz): δ 164.4, 139.7, 133.0, 130.4, 130.2, 129.7, 127.6, 115.4,
73.3, 56.3, 56.2, 46.5, 38.4 ppm. LC-MS (method A) m/z (ES+)
351 [M + H], (ES-) 349 [M - H], Rt ) 2.96 min.
(2R,3S)-N-(3-Amino-2-hydroxy-4-phenyl-butyl)-4-methoxy-
2,3,6-trimethyl-benzenesulfonamide (4). Prepared in an analogous
fashion to 10 but using 9b (840 mg, 1.71 mmol) as starting material
to give the title compound 4, an off-white solid, 630 mg, 94%. 1H
NMR (CDCl3, 400 MHz): δ 7.25 (m, 2H), 7.15 (m, 3H), 6.80 (s,
1H), 4.82 (br, 1H), 3.83 (s, 3H), 3.29 (dd, J ) 5.4, 10.7 Hz, 1H),
2.96 (dd, J ) 4.7, 12.8 Hz, 1H), 2.76 (m, 3H), 2.60 (s, 3H), 2.50
(s, 3H), 2.30 (m, 1H), 2.09 (s, 3H) ppm. 13C NMR (CDCl3, 100
MHz): δ 158.6, 139.9, 138.3, 138.0, 129.6, 129.4, 128.2, 125.9,
124.2, 112.4, 72.4, 55.7 (×2), 45.4, 39.0, 23.9, 17.7, 12.0 ppm.
LC-MS (method A) m/z (ES+) 393 [M + H], (ES-) 391 [M -
H], Rt ) 3.06 min.

DiscoVery of NoVel Thrombin Inhibitors
Journal of Medicinal Chemistry, 2006, Vol. 49, No. 4 1353
(2R,3S)-N-[3-(3-Chloro-benzylamino)-2-hydroxy-4-phenyl-
butyl)-4-methoxy-2,3,6-trimethyl-benzenesulfonamide (5). (2R,3S)-
N-(3-Amino-2-hydroxy-4-phenyl-butyl)-4-methoxy-2,3,6-trimethyl-
benzenesulfonamide 4 (158 mg, 402 µmol) and 3-chlorobenzalde-
hyde (46 µL, 406 µmol) were dissolved in methanol (10 mL) and
stirred for 30 min. Acetic acid (75 µL, 1.31 mmol) and sodium
cyanoborohydride (51 mg, 812 µmol) were added, and the reaction
mixture was stirred for 18 h. The mixture was concentrated in
vacuo, and the residue was partitioned between EtOAc (25 mL)
and 10% aq K2CO3 (25 mL). The organic phase was dried (Na2-
SO4), filtered, and concentrated in vacuo. The residue was purified
by silica chromatography with hexane/EtOAc (1:1) to give the title
compound 5, a white foam solid, 141 mg, 68%.1H NMR (CDCl3,
400 MHz): δ 7.11-7.29 (m, 5H), 7.06 (m, 2H), 6.98 (t, J ) 1.6
Hz, 1H), 6.89 (td, J ) 1.6, 6.8 Hz, 1H), 6.58 (s, 1H), 3.83 (s, 3H),
3.66 (td, J ) 4.4, 7.7 Hz, 1H), 3.58 (d, J ) 13.5 Hz, 1H), 3.52 (d,
J ) 13.5 Hz, 1H), 3.11 (dd, J ) 4.4, 12.9 Hz, 1H), 2.97 (dd, J )
7.7, 12.9 Hz, 1H), 2.86 (td, J ) 4.4, 9.3 Hz, 1H), 2.73 (dd, J )
4.4, 13.9 Hz, 1H), 2.67 (s, 3H), 2.62 (m, 1H), 2.58 (s, 3H), 2.14 (s.
3H) ppm. 13C NMR (CDCl3, 100 MHz): δ 159.8, 142.1, 139.2,
138.2, 134.6, 130.1, 129.5, 129.1, 129.0, 128.5, 127.7, 127.1, 126.6,
125.7, 112.5, 69.9, 60.8, 55.9, 51.4, 45.5, 35.9, 24.7, 18.3, 12.4
ppm. LC-MS (method A) m/z (ES+) 517, 519 [M + H], (ES-)
515, 517 [M - H], Rt ) 3.53 min.
(2R,3S)-N-[3-(5-Chloro-2-tetrazol-1-yl-benzylamino)-2-
hydroxy-4-phenyl-butyl)-4-methoxy-2,3,6-trimethyl-benzene-
sulfonamide (6). Prepared in an analogous fashion to 5 but using
4 (451 mg, 1.15 mmol) and 5-chloro-2-tetrazol-1-yl-benzaldehyde
14 (240 mg, 1.15 mmol) in methanol (35 mL) in the presence of
activated 3 Å molecular sieves. The residue was applied to a silica
chromatography column with methanol/CH2Cl2 (3:97) to give the
crude product (355 mg). A 100 mg sample was further purified by
preparative HPLC to give the title compound 6, a white solid, 73
mg, 40%. 1H NMR (CD3OD, 400 MHz): δ 9.26 (s, 1H), 7.51 (m,
2H), 7.44 (m, 1H), 7.10-7.22 (m, 3H), 7.05 (m, 2H), 6.69 (s, 1H),
3.81 (s, 3H), 3.73 (m, 1H), 3.54 (d, J ) 13.9 Hz, 1H), 3.44 (d, J
) 13.9 Hz, 1H), 2.97 (m, 2H), 2.82 (m, 2H), 2.60 (s, 3H), 2.57
(m, 1H), 2.52 (s, 3H), 2.09 (s. 3H) ppm. 13C NMR (CD3OD, 100
MHz): δ 159.7, 144.5, 139.0, 138.8, 138.6, 136.5, 135.4, 131.9,
131.5, 129.2, 129.0, 128.6, 127.4, 126.6, 125.2, 112.2, 69.7, 61.3,
55.0, 46.5, 44.8, 34.7, 23.5, 17.2, 11.1 ppm. LC-MS (method A)
m/z (ES+) 585, 587 [M + H], (ES-) 583, 585 [M - H], Rt ) 4.51
min.
(2R,3S)-N-{3-[5-Chloro-2-tetrazol-1-yl-benzylamino]-2-hydroxy-
4-phenyl-butyl}-4-methoxy-benzenesulfonamide (7). Prepared in
an analogous fashion to 5 but using (2R,3S)-N-(3-amino-2-hydroxy-
4-phenyl-butyl)-4-methoxy-benzenesulfonamide 10 (264 mg, 753
µmol) and 5-chloro-2-tetrazol-1-yl-benzaldehyde 14 (157 mg, 753
µmol). The residue was applied to a silica chromatography column
with methanol/CH2Cl2 (3:97) to give the crude product 7 (294 mg).
A 100 mg sample was further purified by preparative HPLC to
give a white solid, 50 mg, 36%. 1H NMR (CD3OD, 400 MHz): δ
9.22 (s, 1H), 7.77 (m, 2H), 7.47 (m, 2H), 7.40 (m, 2H), 7.20 (m,
3H), 7.15 (m, 1H), 7.08 (m, 2H), 7.03 (m, 2H), 3.85 (s, 3H), 3.62
(td, J ) 4.8, 7.4 Hz, 1H), 3.44 (d, J ) 13.9 Hz, 1H), 3.37 d, J )
13.9 Hz, 1H), 3.03 (dd, J ) 4.8, 13.2 Hz, 1H), 2.80 (m, 3H), 2.50
(dd, J ) 8.6, 13.3 Hz, 1H) ppm. 13C NMR (CD3OD, 100 MHz):
δ 164.4, 145.7, 140.5, 138.8, 137.5, 133.0, 132.8, 130.4, 130.2,
129.6, 129.5, 128.7, 127.4, 115.4, 71.8, 62.1, 56.2, 47.8, 46.8, 36.9
ppm. LC-MS (method B) m/z (ES+) 543, 545 [M + H], (ES-)
541, 543 [M - H], Rt ) 2.33 min.
5-Chloro-2-tetrazol-1-yl-benzoic Acid (12). 2-Amino-5-chlo-
robenzoic acid 11 (10.90 g, 63.53 mmol) and sodium azide (12.38
g, 190.43 mmol) were suspended in trimethyl orthoformate (21.0
mL, 191.95 mmol) and cooled to 0 °C. Glacial acetic acid (220
mL) was added, and the mixture was stirred at 0 °C for 3 h then at
ambient temperature for 16 h. The slurry was concentrated in vacuo,
and the residue was partitioned between EtOAc (750 mL) and 3 N
HCl (500 mL). The organic phase was dried (Na2SO4), filtered,
and concentrated in vacuo to give the title compound 12, a light
yellow solid, 12.57 g, 88%.1H NMR (DMSO-d6, 400 MHz): δ
13.75 (s, 1H), 9.81 (s, 1H), 8.05 (d, J ) 2.5 Hz, 1H), 7.92 (dd, J
) 2.5, 8.5 Hz, 1H), 7.77 (d, J ) 8.5 Hz, 1H) ppm. 13C NMR
(DMSO-d6, 100 MHz): δ 164.6, 145.2, 135.9, 133.2, 131.6, 131.0,
130.1, 129.9 ppm. LC-MS (method A) m/z (ES+) 225, 227 [M +
H], Rt ) 3.70 min.
5-Chloro-N-methoxy-N-methyl-2-tetrazol-1-yl-benzamide (13).
5-Chloro-2-tetrazol-1-yl-benzoic acid 12 (1.80 g, 8.01 mmol), N,O-
dimethylhydroxylamine hydrochloride (781 mg, 8.01 mmol), and
PyBOP (4.17 g, 8.01 mmol) were suspended in CH2Cl2 (150 mL).
N,N-Diisopropylethylamine (4.2 mL, 24.03 mmol) was added, and
the resultant solution was stirred for 36 h. The solution was
concentrated in vacuo to ∼15-20 mL, diluted with EtOAc (200
mL), and washed with water (200 mL), 10% aq KHSO4 (200 mL),
saturated aq NaHCO3 (200 mL), and saturated brine (200 mL). The
organic phase was dried (Na2SO4), filtered, and concentrated in
vacuo. The residue was purified by silica chromatography (EtOAc)
to give the title compound 13, a white solid, 1.78 g, 83%. 1H NMR
(CD3OD, 400 MHz): δ 9.53 (s, 1H), 7.76 (d, J ) 1.9 Hz, 1H),
7.73 (m, 2H), 3.45 (s, 3H), 3.18 (s, 3H) ppm. 13C NMR (CD3OD,
100 MHz): δ 145.1, 133.9, 133.0, 132.7, 131.4, 130.6, 127.6, 62.6,
33.1 ppm. LC-MS (method A) m/z (ES+) 268, 270, Rt ) 3.15
min.
5-Chloro-2-tetrazol-1-yl-benzaldehyde (14). A solution of
5-chloro-N-methoxy-N-methyl-2-tetrazol-1-yl-benzamide 13 (1.72
g, 6.34 mmol) in THF (20 mL) was added dropwise to a solution
of lithium aluminum hydride (1 M in THF) (12.9 mL, 12.9 mmol)
at -78 °C over the period of 30 min. The mixture was stirred for
1 h, then cold water (5 mL) added dropwise to quench the reaction.
The resultant mixture was diluted with EtOAc (250 mL) and washed
with 1 N HCl (240 mL) and saturated brine (240 mL). The organic
phase was dried (Na2SO4), filtered, and concentrated in vacuo to
give the title compound 14, a yellow solid, 1.26 g, 94%. 1H NMR
(CDCl3, 400 MHz): δ 9.85 (s, 1H), 8.98 (s, 1H), 8.07 (d, J ) 2.4
Hz, 1H), 7.79 (dd, J ) 2.4, 8.5 Hz, 1H), 7.52 (d, J ) 8.5 Hz, 1H)
ppm. 13C NMR (CDCl3, 100 MHz): δ 186.9, 144.0, 138.4, 135.3,
132.4, 132.0, 131.6, 128.3 ppm. LC-MS (method A) m/z (ES+)
209 [M + H], Rt ) 2.96 min.
Thrombin Assay. Compounds were incubated with activated
thrombin and 14 µM fluorogenic substrate, BOC-Val-Pro-Arg-MCA
(Bachem), in 50 mM Tris pH 8, 5 mM CaCl2, 100 mM NaCl, 5%
DMSO in 96-well plates. The cleavage of the substrate was followed
by monitoring the change in fluorescence at 460 nm (excitation at
365 nm) for 25 min at room temperature on a Packard Fusion plate
reader. Initial reaction rates were measured, and IC50s were
calculated from replicate curves using GraphPad Prizm software,
and standard errors and P values were within statistically acceptable
limits. Assay standards were provided by benzamidine and pro-
flavin, both of which are well described in the thrombin litera-
ture.55,56 Benzamidine and proflavin were found to have IC50 values
of 440 and 12 µM, respectively.
Acknowledgment. The authors gratefully acknowledge the
contributions from Susanne Saalau-Bethell, Anne Cleasby, and
Cesare Granata with respect to protein purification, protein
crystallographic work, and HPLC purification, respectively. This
work was partly supported by a grant from Trinity College,
Cambridge, awarded to N. Howard.
Supporting Information Available: Table of results from
purity analyses by HPLC for key target compounds. This material
is available free of charge via the Internet at http://pubs.acs.org.
References
(1) Bode, W. The structure of thrombin, a chameleon-like proteinase. J.
Thromb. Haemostasis [Online] 2005, Epub ahead of print, 09/05/
2005.
(2) Stubbs, M. T.; Bode, W. The clot thickens: clues provided by
thrombin structure. Trends Biochem. Sci. 1995, 20, 23-28.
(3) Riester, D.; Wirsching, F.; Salinas, G.; Keller, M.; Gebinoga, M.;
Kamphausen, S.; Merkwirth, C.; Goetz, R.; Wiesenfeldt, M.; Stur-
zebecher, J.; Bode, W.; Friedrich, R.; Thurk, M.; Schwienhorst, A.
Thrombin inhibitors identified by computer-assisted multiparameter
design. Proc. Natl. Acad. Sci. U.S.A. 2005, 102, 8597-8602.

1354 Journal of Medicinal Chemistry, 2006, Vol. 49, No. 4
Howard et al.
(4) Steinmetzer, T.; Sturzebecher, J. Progress in the development of
synthetic thrombin inhibitors as new orally active anticoagulants.
Curr. Med. Chem. 2004, 11, 2297-2321.
(28) Hartshorn, M. J. AstexViewer: a visualisation aid for structure-based
drug design. J. Comput.-Aided Mol. Des. 2002, 16, 871-881.
(29) Watson, P.; Verdonk, M. L.; Hartshorn, M. J. A web-based platform
for virtual screening. J. Mol. Graphics Modell. 2003, 22, 71-82.
(30) Verdonk, M. L.; Cole, J. C.; Hartshorn, M.; Murray, C. W.; Taylor,
R. D. Improved protein-ligand docking using GOLD. Proteins 2003,
52, 609-623.
(31) Jones, G.; Willett, P.; Glen, R. C. Molecular recognition of receptor
sites using a genetic algorithm with a description of desolvation. J.
Mol. Biol. 1995, 245, 43-53.
(5) Stassen, J. M.; Arnout, J.; Deckmyn, H. The hemostatic system. Curr.
Med. Chem. 2004, 11, 2245-2260.
(6) Matsuo, T.; Koide, M.; Kario, K. Development of argatroban, a direct
thrombin inhibitor, and its clinical application. Semin. Thromb.
Hemostasis 1997, 23, 517-522.
(7) Sorbera, L. A.; Bayes, M. C. J.; Silvestre, J. Melagatran and
Ximelagatran. Drugs Future 2001, 26, 1155-1170.
(8) Dullweber, F.; Stubbs, M. T.; Musil, D.; Sturzebecher, J.; Klebe, G.
Factorising ligand affinity: a combined thermodynamic and crystal-
lographic study of trypsin and thrombin inhibition. J. Mol. Biol. 2001,
313, 593-614.
(9) Sturzebecher, J.; Hauptmann, J.; Steinmetzer, T. Thrombin. In
Proteinase and Peptidase Inhibition; Smith, H. J., Simons, C., Eds.;
Taylor and Francis: London, 2002; pp 178-201.
(10) Oprea, T. I.; Davis, A. M.; Teague, S. J.; Leeson, P. D. Is there a
difference between leads and drugs? A historical perspective. J. Chem.
Inf. Comput. Sci. 2001, 41, 1308-1315.
(11) Teague, S. J.; Davis, A. M.; Leeson, P. D.; Oprea, T. The design of
leadlike combinatorial libraries. Angew. Chem., Int. Ed. 1999, 38,
3743-3748.
(12) 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.
(32) Baxter, C. A.; Murray, C. W.; Clark, D. E.; Westhead, D. R.; Eldridge,
M. D. Flexible docking using Tabu search and an empirical estimate
of binding affinity. Proteins 1998, 33, 367-382.
(33) Sanderson, P. E.; Cutrona, K. J.; Dyer, D. L.; Krueger, J. A.; Kuo,
L. C.; Lewis, S. D.; Lucas, B. J.; Yan, Y. Small, low nanomolar,
noncovalent thrombin inhibitors lacking a group to fill the “distal
binding pocket”. Bioorg. Med. Chem. Lett. 2003, 13, 161-164.
(34) Tucker, T. J.; Brady, S. F.; Lumma, W. C.; Lewis, S. D.; Gardell, S.
J.; Naylor-Olsen, A. M.; Yan, Y.; Sisko, J. T.; Stauffer, K. J.; Lucas,
B. J.; Lynch, J. J.; Cook, J. J.; Stranieri, M. T.; Holahan, M. A.;
Lyle, E. A.; Baskin, E. P.; Chen, I. W.; Dancheck, K. B.; Krueger,
J. A.; Cooper, C. M.; Vacca, J. P. Design and synthesis of a series
of potent and orally bioavailable noncovalent thrombin inhibitors that
utilize nonbasic groups in the P1 position. J. Med. Chem. 1998, 41,
3210-3219.
(35) Young, M. B.; Barrow, J. C.; Glass, K. L.; Lundell, G. F.; Newton,
C. L.; Pellicore, J. M.; Rittle, K. E.; Selnick, H. G.; Stauffer, K. J.;
Vacca, J. P.; Williams, P. D.; Bohn, D.; Clayton, F. C.; Cook, J. J.;
Krueger, J. A.; Kuo, L. C.; Lewis, S. D.; Lucas, B. J.; McMasters,
D. R.; Miller-Stein, C.; Pietrak, B. L.; Wallace, A. A.; White, R. B.;
Wong, B.; Yan, Y.; Nantermet, P. G. Discovery and evaluation of
potent P1 aryl heterocycle-based thrombin inhibitors. J. Med. Chem.
2004, 47, 2995-3008.
(36) Skrzypczak-Jankun, E.; Carperos, V. E.; Ravichandran, K. G.;
Tulinsky, A.; Westbrook, M.; Maraganore, J. M. Structure of the
hirugen and hirulog 1 complexes of R-thrombin. J. Mol. Biol. 1991,
221, 1379-1393.
(37) Sharff, A. J. High throughput crystallography on an in-house source,
using ACTOR. Rigaku J. 2004, 20, 10.
(38) Hopkins, A. L.; Groom, C. R.; Alex, A. Ligand efficiency: a useful
metric for lead selection. Drug DiscoVery Today 2004, 9, 430-431.
(39) Carr, R. A.; Congreve, M.; Murray, C. W.; Rees, D. C. Fragment-
based lead discovery: leads by design. Drug DiscoVery Today 2005,
10, 987-992.
(40) Breinlinger, E. C.; Keenan, C. J.; Rotello, V. M. Modulation of flavin
recognition and redox properties through donor atom-π interactions.
J. Am. Chem. Soc. 1998, 120, 8606-8609.
(41) Rotello, V. M. The donor atom-π interaction of sulfur with flavin.
A density functional investigation. Heteroat. Chem. 1998, 9, 605-
606.
(42) Rossky, P. J.; Karplus, M. Solvation. A molecular dynamics study
of a dipeptide in water. J. Am. Chem. Soc. 1979, 101, 1913-1937.
(43) Lim, M. S.; Johnston, E. R.; Kettner, C. A. The solution conformation
of (D)Phe-Pro-containing peptides: implications on the activity of
Ac-(D)Phe-Pro-boroArg-OH, a potent thrombin inhibitor. J. Med.
Chem. 1993, 36, 1831-1838.
(44) Obst, U.; Banner, D. W.; Weber, L.; Diederich, F. Molecular
recognition at the thrombin active site: structure-based design and
synthesis of potent and selective thrombin inhibitors and the X-ray
crystal structures of two thrombin inhibitor complexes. Chem. Biol.
1997, 4, 287-295.
(45) Reers, M.; Koschinsky, R.; Dickneite, G.; Hoffmann, D.; Czech, J.;
Stuber, W. Synthesis and characterisation of novel thrombin inhibitors
based on 4-amidinophenylalanine. J. Enzyme Inhib. 1995, 9, 61-
72.
(46) Murray, C. W.; Verdonk, M. L. The consequences of translational
and rotational entropy lost by small molecules on binding to proteins.
J. Comput.-Aided Mol. Des. 2002, 16, 741-753.
(47) Page, M. I.; Jencks, W. P. Entropic contributions to rate accelerations
in enzymic and intramolecular reactions and the chelate effect. Proc.
Natl. Acad. Sci. U.S.A. 1971, 68, 1678-1683.
(48) Ngai, P. K.; Chang, J. Y. A novel one-step purification of human
R-thrombin after direct activation of crude prothrombin enriched from
plasma. Biochem. J. 1991, 280 (Part 3), 805-808.
(49) Pflugrath, J. W. The finer things in X-ray diffraction data collection.
Acta Crystallogr., Sect. D 1999, 55 (Part 10), 1718-1725.
(50) Collaborative Computational Project, N. 4. The CCP4 suite: programs
for protein crystallography. Acta Crystallogr., Sect. D 1994, 50, 760-
763.
(13) Hajduk, P. J.; Sheppard, G.; Nettesheim, D. G.; Olejniczak, E. T.;
Shuker, S. B.; Meadows, R. P.; Steinman, D. H.; Carrera, G. M., Jr.;
Marcotte, P. A.; Severin, J.; Walter, K.; Smith, H.; Gubbins, E.;
Simmer, R.; Holzman, T. F.; Morgan, D. W.; Davidsen, S. K.;
Summers, J. B.; Fesik, S. W. Discovery of potent nonpeptide
inhibitors of stromelysin using SAR by NMR. J. Am. Chem. Soc.
1997, 119, 5818-5827.
(14) Shuker, S. B.; Hajduk, P. J.; Meadows, R. P.; Fesik, S. W.
Discovering high-affinity ligands for proteins: SAR by NMR. Science
1996, 274, 1531-1534.
(15) Carr, R.; Jhoti, H. Structure-based screening of low-affinity com-
pounds. Drug DiscoVery Today 2002, 7, 522-527.
(16) Lesuisse, D.; Lange, G.; Deprez, P.; Benard, D.; Schoot, B.; Delettre,
G.; Marquette, J. P.; Broto, P.; Jean-Baptiste, V.; Bichet, P.; Sarubbi,
E.; Mandine, E. SAR and X-ray. A new approach combining
fragment-based screening and rational drug design: application to
the discovery of nanomolar inhibitors of Src SH2. J. Med. Chem.
2002, 45, 2379-2387.
(17) Blundell, T. L.; Abell, C.; Cleasby, A.; Hartshorn, M. J.; Tickle, I.
J.; Parasini, E.; Jhoti, H. High-throughput X-ray crystallography for
drug discovery. In Drug Design: Special Publication; Flower, D.
R., Ed.; Royal Society of Chemistry: Cambridge, U.K., 2002; pp
53-59.
(18) Rees, D. C.; Congreve, M.; Murray, C. W.; Carr, R. Fragment-based
lead discovery. Nat. ReV. Drug DiscoVery 2004, 3, 660-672.
(19) Erlanson, D. A.; McDowell, R. S.; O’Brien, T. Fragment-based drug
discovery. J. Med. Chem. 2004, 47, 3463-3482.
(20) Zartler, E. R.; Shapiro, M. R. Fragnomics: fragment-based drug
discovery. Curr. Opin. Chem. Biol. 2005, 9, 366-370.
(21) Nienaber, V. L.; Richardson, P. L.; Klighofer, V.; Bouska, J. J.;
Giranda, V. L.; Greer, J. Discovering novel ligands for macro-
molecules using X-ray crystallographic screening. Nat. Biotechnol.
2000, 18, 1105-1108.
(22) Congreve, M.; Carr, R.; Murray, C.; Jhoti, H. A “rule of three” for
fragment-based lead discovery? Drug DiscoVery Today 2003, 8, 876-
877.
(23) Muchmore, S. W.; Olson, J.; Jones, R.; Pan, J.; Blum, M.; Greer, J.;
Merrick, S. M.; Magdalinos, P.; Nienaber, V. L. Automated crystal
mounting and data collection for protein crystallography. Struct.
Folding Des. 2000, 8, R243-R246.
(24) Hartshorn, M. J.; Murray, C. W.; Cleasby, A.; Frederickson, M.;
Tickle, I. J.; Jhoti, H. Fragment-based lead discovery using X-ray
crystallography. J. Med. Chem. 2005, 48, 403-413.
(25) Gill, A. L.; Frederickson, M.; Cleasby, A.; Woodhead, S. J.; Carr,
M. G.; Woodhead, A. J.; Walker, M. T.; Congreve, M. S.; Devine,
L. A.; Tisi, D.; O’Reilly, M.; Seavers, L. C. A.; Davis, D. J.; Curry,
J.; Anthony, R.; Padova, A.; Murray, C. W.; Carr, R. A. E.; Jhoti,
H. Identification of Novel p38R MAP Kinase Inhibitors using
fragment-based lead generation. J. Med. Chem. 2004, 48, 414-426.
(26) Szczepankiewicz, B. G.; Liu, G.; Hajduk, P. J.; Abad-Zapatero, C.;
Pei, Z.; Xin, Z.; Lubben, T. H.; Trevillyan, J. M.; Stashko, M. A.;
Ballaron, S. J.; et al. Discovery of a potent, selective protein tyrosine
phosphatase 1B inhibitor using a linked-fragment strategy. J. Am.
Chem. Soc. 2003, 125, 4087-4096.
(27) Leung, D.; Abbenante, G.; Fairlie, D. P. Protease inhibitors: current
status and future prospects. J. Med. Chem. 2000, 43, 305-341.

DiscoVery of NoVel Thrombin Inhibitors
Journal of Medicinal Chemistry, 2006, Vol. 49, No. 4 1355
(51) Engh, R. A.; Brandstetter, H.; Sucher, G.; Eichinger, A.; Baumann,
U.; Bode, W.; Huber, R.; Poll, T.; Rainer, R.; von der Saal, W.
Enzyme flexibility, solvent and “weak” interactions characterize
thrombin-ligand interactions: implications for drug design. Structure
1996, 4, 1353-1362.
(52) Verdonk, M. L.; Hartshorn, M.; Mooij, W. T. M.; Murray, C. W.;
Taylor, R. D.; Watson, P. Virtual screening using protein-ligand
docking: avoiding artificial enrichment. J. Chem. Inf. Comput. Sci.
2004, 44(3), 793-806.
(53) Maple, J. R.; Hwang, M.-J.; Jalkanen, K. J.; Stockfish, T. P. H. A.
T. Derivation of class II force fields: V. Quantum force field for
amides, peptides and related compounds. J. Comput. Chem. 1998,
19, 430-458.
(54) Barrish, J. C.; Gordon, E.; Alam, M.; Lin, P. F.; Bisacchi, G. S.;
Chen, P.; Cheng, P. T.; Fritz, A. W.; Greytok, J. A.; Hermsmeier,
M. A. Aminodiol HIV protease inhibitors. 1. Design, synthesis, and
preliminary SAR. J. Med. Chem. 1994, 37, 1758-1768.
(55) Ascenzi, P.; Fruttero, R.; Amiconi, G.; Pugluse, L.; Bolognesi, M.;
Massimo, C.; Onesti, S.; Guarneri, M.; Menegatti, E. Inhibition of
R-, and â- and γ- thrombin by mono-, bis-, tris- and tetra-benzamidine
structures: thermodynamic study. J. Enzyme Inhib. 1992, 6, 131-
139. 1992.
(56) Magnusson, K. K. A. The binding of proflavin to thrombin. Arch.
Biochem. Biophys. 1974, 160, 175-184.
JM050850V