How a β-D-glucoside side chain enhances binding affinity to thrombin of inhibitors bearing 2-chlorothiophene as P1 moiety: Crystallography, fragment deconstruction study, and evaluation of antithrombotic properties

Article abstract of DOI:10.1021/jm5010754

The β-d-glucose-containing compound 3, bearing 2-chlorothiophene and 1-isopropylpiperidine moieties as binders of the S1 and S4 pockets, respectively, proved to be potent competitive inhibitor of factor Xa (fXa, K_i = 0.090 nM) and thrombin (fIIa, K_i = 100 nM). The potency of 3 increases, over the parent compound 1, against fIIa (110-fold), much more than against fXa (7-fold). Experimental deconstruction of 3 into smaller fragments revealed a binding cooperativity of the P3/P4 and propylene-linked β-d-glucose fragments, stronger in fIIa (15.5 kJ·mol^-1) than in fXa (2.8 kJ·mol^-1). The crystal structure of human fIIa in complex with 3 revealed a binding mode including a strong H-bond network between the glucose O1′, O3′, and O5′ and two critical residues, namely R221a and K224, belonging to the Na⁺-binding site which may allosterically perturb the specificity sites. The potential of 3 as antithrombotic agent was supported by its ability to inhibit thrombin generation and to stimulate fibrinolysis at submicromolar concentration.

Full text of DOI:10.1021/jm5010754

Article
pubs.acs.org/jmc
How a β‑D‑Glucoside Side Chain Enhances Binding Affinity to
Thrombin of Inhibitors Bearing 2‑Chlorothiophene as P1 Moiety:
Crystallography, Fragment Deconstruction Study, and Evaluation of
Antithrombotic Properties
Benny D. Belviso,† Rocco Caliandro,† Modesto de Candia,‡ Giorgia Zaetta,‡ Gianfranco Lopopolo,‡,∥
Francesca Incampo,§ Mario Colucci,§ and Cosimo D. Altomare*,‡
†Institute of Crystallography, Consiglio Nazionale delle Ricerche, Via Amendola 122/o, 70126 Bari, Italy
‡Department of PharmacyDrug Sciences, University of Bari ‘‘Aldo Moro’’, Via E. Orabona 4, 70125 Bari, Italy
§Department of Biomedical Sciences and Human Oncology, University of Bari “Aldo Moro”, Piazza Giulio Cesare 11, 70124 Bari,
Italy
S
* Supporting Information
ABSTRACT: The β-D-glucose-containing compound 3, bearing 2-chlorothio-
phene and 1-isopropylpiperidine moieties as binders of the S1 and S4 pockets,
respectively, proved to be potent competitive inhibitor of factor Xa (fXa, Ki =
0.090 nM) and thrombin (fIIa, Ki = 100 nM). The potency of 3 increases, over
the parent compound 1, against fIIa (110-fold), much more than against fXa (7-
fold). Experimental deconstruction of 3 into smaller fragments revealed a
binding cooperativity of the P3/P4 and propylene-linked β-D-glucose
fragments, stronger in fIIa (15.5 kJ·mol−1) than in fXa (2.8 kJ·mol−1). The
crystal structure of human fIIa in complex with 3 revealed a binding mode
including a strong H-bond network between the glucose O1′, O3′, and O5′ and
two critical residues, namely R221a and K224, belonging to the Na+-binding
site which may allosterically perturb the specificity sites. The potential of 3 as
antithrombotic agent was supported by its ability to inhibit thrombin generation and to stimulate fibrinolysis at submicromolar
concentration.
failure or impaired liver disease. In addition, they have to be
used with caution in patients with renal impairment.5
INTRODUCTION
■
Currently, the management of anticoagulation prophylaxis and
therapy can benefit from a number of new orally active
anticoagulants (NOACs), which have recently become available
as alternatives to warfarin and other vitamin K antagonists
(VKAs).1 The NOACs currently approved are dabigatran,
available as etexilate prodrug (Pradaxa), which acts as inhibitor
of both free and fibrin-bound thrombin (fIIa), and apixaban
(Eliquis) and rivaroxaban (Xarelto), which are factor Xa (fXa)-
selective reversible inhibitors.2,3 Indications of NOACs include
stroke prevention in patients with nonvalvular atrial fibrillation
(AF) and following hip or knee replacement surgery.
Rivaroxaban is also indicated for acute treatment and secondary
deep venous thrombosis (DVT) and pulmonary embolism.4
The NOACs proved able to overcome a number of
shortcomings often associated with the anticoagulant treatment
with VKAs. They have more stable and predictable
pharmacokinetics, no or fewer interactions with food and
drugs, and more importantly, they can be administered in
standard doses without the need for laboratory monitoring for
dose adjustment. However, the NOACs are not indicated for
conditions such as prosthetic heart valve replacement,
hemodynamically significant valve disease, and severe renal
In the past two decades, the rational design of orally
bioavailable small-molecule fIIa and fXa direct inhibitors has
been intensively supported by X-ray of ligand−protein crystal
structures, molecular modeling, and 3D-QSAR approaches,6,7
which has allowed achievement of high inhibitory potency.
However, despite the large number of discovered potent
compounds, only three NOACs are clinically available and a
few others are in clinical trials.
We recently reported several N-(2- or 3-methoxyphenyl)-
piperidine-4-carboxamide derivatives, which selectively inhibit
factor Xa8,9 or thrombin (fIIa)10 depending upon the moieties
they bear as binders of the specificity enzymes’ subsites S1−S4.
In particular, compound 1 (Figure 1), bearing 5-chlorothio-
phen-2-yl and 1-isopropylpiperidin-4-yl moieties as S1 and S4
binders, respectively, proved to be a very potent fXa inhibitor,
with high selectivity over fIIa and other serine proteases and
good anticoagulant in vitro and ex vivo.10,11
Received: July 17, 2014
Published: September 30, 2014
© 2014 American Chemical Society
8563
dx.doi.org/10.1021/jm5010754 | J. Med. Chem. 2014, 57, 8563−8575

Journal of Medicinal Chemistry
Article
Scheme 1. Synthesis of a 5-Chlorothiophen-2yl-Containing
a
P1 Fragment Molecule .
Figure 1. Chemical structures of recently reported potent and selective
factor Xa inhibitors, showing in vitro and ex vivo anticoagulant
activity.12
a
Reagents and conditions: (a) 3-bromopropanol, K2CO3, Me2CO,
reflux, overnight, 65%; (b) SnCl2·2H2O, EtOAc, reflux, 48 h, >95%.
According to our docking calculations and in good
agreement with the X-ray crystal structures of fXa-inhibitor
complexes,12,13 compound 1 interacts with the fXa binding
pocket by directing the 2-chlorothiophene moiety into the S1
subsite, with the chlorine atom pointing toward the center of
the Y228 phenyl ring and the protonated N1-isopropyl residue
(the tertiary ammonium head) at the center of the S4 aromatic
box, wherein it undergoes efficient cation−π interactions and
additional C−H···π interactions by the isopropyl group with
the side chains of F174, Y99, and W215. Interestingly, the
modification of compound 1 by connection of a β-D-glucose
fragment via a propylene bridge at the ortho position of the
isonipecotamido moiety on the central phenyl ring led to the
O-glucoside 3 (and its peracetylated derivative 2), which
proved to be not only significantly more potent than the parent
compound 1 as fXa inhibitor but also a good fIIa inhibitor with
inhibition constant in the nanomolar range, showing
anticoagulant activity in both in vitro and ex vivo assays.12 In
other words, the glucosyl conjugates 2 and 3, while still
retaining fXa-selectivity, displayed improved potency against
fIIa. To possibly understand this behavior, in this work we
report a fragment deconstruction study of compound 3, aimed
at assessing the contributions of its smaller fragments to the
free energy of binding to fIIa and fXa and the crystal structures
of human thrombin in complex with both two glucose-based
compounds 2 and 3, which provided us with useful information
on their binding modes. The antithrombotic potential of these
new inhibitors of fXa and fIIa has been also assessed by
investigating their effects on thrombin generation and
fibrinolysis.
with N-Boc-isonipecotic acid using DCC/HOBt as the
coupling reagents, followed by the removal of the Boc-
protecting group to yield 12. Reductive amination of 12 with
acetone/Na(CN)BH3 and cleavage of acetyl-protecting groups
of 13 with MeONa/MeOH provided compound 14.
RESULTS AND DISCUSSION
■
Fragment-Based Structure−Activity Relationships.
The enzymes’ inhibition data for fragments and derivatives of
the reference molecules 1 and 3, expressed as inhibition
constant (Ki) or percent inhibition at 1 mM concentration, are
listed in Table 1. The Ki values were calculated using the
Cheng−Prusoff equation14 from IC50 values, determined
through standard spectrophotometric assays.
Lineweaver−Burk plots, which were generated for the most
potent inhibitor 3 (Figure 2) and its parent compound 1, using
fixed amounts of the enzymes and varying concentrations of the
substrates in the absence or presence of inhibitor at four
different concentrations, showed a competitive inhibition
mechanism. A replot of their slopes versus the corresponding
inhibitor concentrations provided Ki values (fXa, 0.11 0.06
and 0.87 0.18 nM for 1 and 3, respectively; fIIa, 18 0.45
and 0.096 0.0041 μM for 1 and 3, respectively), which are
close to those assessed by the Cheng−Prusoff equation.
On the basis of the data in Table 1, the β-D-glucose-bearing
compound 3, which is more potent than its peracetylated
derivative 2, inhibits fXa with picomolar potency (Ki = 0.090
nM) and fIIa with nanomolar potency (Ki = 100 nM), whereas
its potency increases, over the parent compound 1, against fIIa
(110-fold) much more than against fXa (7-fold).
Experimental deconstruction of 3 provided us with insights
into fXa and fIIa affinity contributions of the 5-(5-
chlorothiophen-2-yl)isoxazol-3-yl binding fragment (P1), 1-
isopropylpiperidine-4-carboxamide binder (P3/P4), and pro-
pylene-linked β-D-glucose fragment (PG). The data show that
all P3/P4-containing fragment molecules [8a (R = H), 8b (R =
OMe) and 14], regardless of whether they bear the hydrophilic
PG group, do not display any noteworthy (>50%) inhibition
against both blood coagulation factors at 1 mM concentration,
beyond which the determination of IC50 values was prevented
because of solubility limits. In contrast, the P1-directed
fragment decomposition of 3 resulted in compounds 4a, 6,
and 6a, which all showed fIIa/fXa affinity in the micromolar
range and a certain preference toward fXa. The glucose-bearing
molecule 6a (fXa Ki = 6.9 μM) proved to be 3-fold more potent
CHEMISTRY
■
The synthesis of the 5-chlorothiophen-2-yl-bearing compound
6 was accomplished by alkylation of the already reported
compound 412 with 3-bromopropanol, followed by reduction of
the nitro to amino group using SnCl2·2H2O in EtOAc at reflux
(Scheme 1).
The synthesis pathways for the molecules containing the 1-
isopropylpiperidine P4 fragment and related glucose-bearing
derivatives are shown in Scheme 2.
The N-isopropyl derivatives 8a,b were obtained by reductive
amination (acetone/Na(CN)BH3) of the known anilides 7a,b.
Compound 14 was synthesized according to previously
reported procedures.12 The amine derivative 10, prepared
through the reaction between 3-bromopropyl-2,3,4,6-tetra-O-
acetyl-β-D-glucose and 2-nitrophenol to afford 9 and sub-
sequent Pd-catalyzed reduction of the nitro group, was coupled
8564
dx.doi.org/10.1021/jm5010754 | J. Med. Chem. 2014, 57, 8563−8575

Journal of Medicinal Chemistry
Article
a
Scheme 2. Synthesis of 1-Isopropylpiperidine-Containing P3/P4 Fragment Derivatives .
a
Reagents and conditions: (a) Na(CN)BH3, Me2CO/MeOH, rt, overnight, 50−70%; (b) K2CO3, dry DMF, rt, overnight, 70%; (c) H2 atmosphere,
10% Pd-C, EtOAc, 8 h, 85%; (d) HOBt, DCC, dry DCM, rt, 48 h, 70%; (e) redistilled TFA, DCM, rt, 2 h, 80%; (f) NaOMe, MeOH, rt, overnight,
70%.
Table 1. Factor Xa and Thrombin Inhibition Constants for
Fragments and Derivatives of the Reference Molecules 1 and
3
a
Ki (nM)
compd
fXa
fIIa
b
1
0.60
11000
160
b
2
0.25
b
3
0.090
11600
18300
6900
100
b
4a
31000
29800
133000
6
b
6a
c
c
8a
8b
14
13%
16%
c
c
16%
0%
c
c
14%
28%
a
Inhibition constants, as calculated by the Cheng−Prusoff equation15
from experimentally determined IC50 values, or % inhibition data at
the maximum concentration tested (1 mM) are means of three
independent determinations (SEM < 5% of the mean), each
b
performed in duplicate. Inhibition data for compounds already
reported by us10,12 have been redetermined in this study. Percent
c
inhibition for the corresponding compounds at 1 mM concentration.
than the corresponding glucose-lacking compound 6 as fXa
inhibitor.
On the basis of the experimental inhibition data, taking 4a as
the shortest active molecule encompassing the chlorothiophene
S1 binder and the aminophenoxy linker, the P1-directed
fragment deconstruction allowed us to assess the contributions,
as stepwise relative free energy of binding (ΔΔG, kJ·mol−1), to
fIIa/fXa affinity of P3/P4 and PG moieties. The data were
analyzed using a double fragment replacement thermodynamic
cycle as shown in Figure 3.
Figure 2. Lineweaver−Burk plots of (A) fXa (2 nM) with the
chromogenic substrate S2765, or (B) fIIa (0.45 U·mL−1) with the
chromogenic substrate S2238, in the absence or presence of the
inhibitor 3; replots of the slopes versus [I] to determine Ki values as
the x-axis intercepts are shown in the upper left inserts.
As for fXa, the largest increase in affinity relative to the P1-
bearing compounds 4a and 6a resulted from addition of the
8565
dx.doi.org/10.1021/jm5010754 | J. Med. Chem. 2014, 57, 8563−8575

Journal of Medicinal Chemistry
Article
Figure 3. P1-directed fragment deconstruction cycle. ΔΔG values of binding (kJ·mol−1) are shown on the arrows in red and blue for fIIa and fXa,
respectively. Colors of the fragments’ structures: P1, red; P3/P4, green; β-D-glucosyl propyloxy (PG), blue. Double fragment replacement cycle
reveals a positive cooperative binding between P3/P4 and PG for both fIIa and fXa inhibition. Such a cooperativity effect is much stronger in fIIa
binding (14.5 − 2.6 + 3.6 = 15.5 kJ·mol−1) than in fXa binding (30.1 − 26.1 − 1.2 = 2.8 kJ·mol−1).
isonipecotamido S3/S4 binder (ΔΔG4a→1 = −26.1 kJ·mol−1;
Table 2. Aqueous Solubility, Lipophilicity, and Stability of
Compounds 1−3
ΔΔG6a→3 = −28.9 kJ·mol−1), whereas a small ΔG gain was
observed to depend upon the PG addition (−1.2 and −4.0 kJ·
mol−1 for 4a → 6a and 1 → 3, respectively). In contrast, a
significant increase in thrombin affinity resulted from the
contemporary addition to the minimum active fragment 4a of
P3/P4 and PG (ΔΔG6a→3 = −18.1 kJ·mol−1; ΔΔG1→3 = −11.9
kJ·mol−1) and not from the addition of P3/P4 or PG moieties
alone (−2.6 and +3.6 kJ·mol−1). Indeed, the ΔΔG “super-
additivity” effect of P3/P4 fragment and PG side chain accounts
for a positive cooperativity in both fIIa and fXa, but such a
binders’ synergistic effect is significantly stronger toward fIIa
(15.5 kJ·mol−1) than fXa (2.8 kJ·mol−1). In other words, the
free energy contributions of P3/P4 and PG to the enzyme
binding are almost additive for fXa (small “superadditivity”, less
than 3 kJ mol−1), whereas the noncovalent interactions attained
by these two moieties into the thrombin pockets cooperate to
significantly enhance their mutual binding affinity (large
“superadditivity”, about 15 kJ mol−1).
compd
physicochemical properties
1
2
3
a
aqueous solubility (PBS, pH 7.4) (mM)
0.53
3.49
d
<1 × 10−3
7.33
3.07
d
b
log k′w
PBS (pH 7.4) half-life (h)
5.07
d
c
c
Plasma half-life (h)
d
2.8
d
a
Solubility (mmol L−1) below the detection limit of the quantification
b
HPLC method. Lipophilicity parameter as assessed by the logarithm
of the polycratic capacity factor determined by a reversed-phase HPLC
method (see Experimental Section). Stability evaluated in 0.05 M
c
phosphate buffer (pH 7.4, 0.15 M KCl) and human plasma solutions
d
at 37 °C. More than 90% unhydrolyzed compound at 4 h incubation.
We measured aqueous solubility, lipophilicity by a reversed
phase (RP) HPLC method, and stability both in PBS at
physiological pH and human plasma of compounds 1−3 (Table
2). The O-glucoside derivative 3 exhibited water solubility (7.33
mM) 14-fold higher than that of the parent compound 1,
whereas the peracetylated compound 2 exhibited high lipo-
philicity and low aqueous solubility (<1 μM). All three
compounds were stable enough in PBS at pH 7.4 and, with
the only exception of the acetate ester derivative 2 (t1/2 = 2.8
h), in human plasma at 37 °C up to 4 h incubation (Figure 4).
To understand their binding modes and gain insights into the
cooperative binding of the fragments P3/P4 and PG, the crystal
structures of human thrombin in complex with the glucose-
based compounds 2 and 3 were solved.
●
Figure 4. Hydrolysis profile of compound 2 in ( ) pH 7.4 PBS (0.15
■
M KCl) and ( ) human serum (first-order half-life 2.8 h
Crystal Structures of Human Thrombin in Complex
with β-D-Glucose-Bearing Inhibitors. The crystals of human
8566
dx.doi.org/10.1021/jm5010754 | J. Med. Chem. 2014, 57, 8563−8575

Journal of Medicinal Chemistry
Article
thrombin (hfIIa) in the presence of the inhibitors 2 and 3 were
obtained and structurally solved by X-ray diffraction. Protein/
inhibitor adducts were crystallized starting from hfIIa crystals
through the soaking technique; therefore, they preserved the
monoclinic unit cell, the space group C2, the molecular
arrangement, and the hirugen molecule (used to prevent the
autolysis of the protein during the crystallization experiment) in
the exosite I of the progenitor hfIIa crystal.15 The unit cell and
refinement parameters for crystal structures of hfIIa in complex
with 2 (PDB 4NZE) and 3 (PDB 4N3L) are listed in Table 3.
between the two crystal structures (RMSD equals 0.18 Å).
Residues belonging to the active site have RMSD values of 0.9
Å at most, suggesting that the hfIIa active site does not undergo
significant changes because of the presence of the inhibitors 2
and 3. However, it should be noted that the last result could be
biased by the technique used to produce protein−ligand
crystals; indeed, protein movements triggered by ligand binding
could be hindered by crystal packing in the case of soaking
experiments.
Inhibitor Binding Modes to hfIIa. Figure 5 shows the
inhibitors 2 (cyan-colored C atoms) and 3 (green-colored C
Table 3. Unit Cell and Refinement Parameters for 4NZE (2)
a
and 4N3L (3) Structures
4NZE (2)
4N3L (3)
wavelength (Å)
0.978
1.5418
space group
C2
C2
cell dimensions (Å)
a
67.460
71.640
71.800
1.98−37.86
22000
97.8 (97.0)
9.2
66.990
71.660
71.790
1.94−13.65
24649
99.0 (94.0)
6.8
b
c
resolution range (Å)
unique reflections
completeness (%)
I/σ(I)
average B-factor (Å2) (from Wilson plot)
23.2
22.8
average B-factor of inhibitor molecule
65.3
31.8
atoms (Å2)
R (%)
19.9 (24.3)
24.8 (29.2)
2536
21.9 (34.0)
Rfree (%)
24.6 (36.9)
protein atoms
2580
47
inhibitor molecule atoms
DMSO molecules
water molecules
RMSD bond lengths (Å)
RMSD bond angles (deg)
59
Figure 5. Structures of the inhibitors 2 (C atoms in cyan) and 3 (C
atoms in green) in the human α-thrombin binding site after the
superimposition of 4NZE and 4N3L X-ray crystal structures. The
inhibitor structures are shown in stick representation with O in red
and N in blue color (sulfur and chloride atoms are hidden in the S1
pocket). The molecular surface of thrombin in the 4N3L structure is
shown in gray. The thrombin binding sites (S1, S2, and S4) for small-
molecule direct inhibitors and sodium binding site (Na+ as violet
sphere from thrombin structure, PDB 2ZFE) are highlighted.
6
6
118
79
0.007
1.419
0.007
1.078
a
Data in parentheses refer to the higher resolution shell.
In both crystal structures, the inhibitor molecules are located
in the active site of the protein and mainly interact with
residues belonging to S1 and S3/S4 sites. The thrombin
conformation is not affected by the presence of the inhibitor
molecules, as suggested by the low values of root-mean-square
deviation (RMSD) of all the Cα atoms, calculated between our
structures and the one used as the search model for the
structure determination process (PDB 1HGT16), which does
not contain active-site inhibitor molecules (RMSD equals 0.38
and 0.54 Å for 4N3L and 4NZE, respectively). Atomic
displacements in specific sites are instead observed: (i) the
autolysis site of the protein from E146 to K149e (RMSD equals
0.55 and 1.76 Å for 4N3L and 4NZE, respectively) and (ii) the
S2 loop from Y60a to W60d (RMSD equals 1.17 and 0.79 Å for
4N3L and 4NZE, respectively). They can be explained taking
into account (i) the mobility for the new terminations
produced by the autolysis of the protein and (ii) the small
variation induced by the presence of a molecule in the
thrombin active site on the rigid hydrophobic S2 loop. In
addition, the residue E186d shows large RMSD value (1.60 and
1.84 Å for 4N3L and 4NZE, respectively), but notwithstanding
the closeness of this residue to the active site, the atomic
displacement does not seem to arise from the inhibitor−protein
coupling. Despite the presence of two different inhibitor
molecules, there were no apparent conformational differences
atoms) in the binding site of hfIIa after the superimposition of
the protein crystal structures. The chlorothiophene moieties of
the two inhibitors are completely overlapped in the S1 pocket,
whereas just small variations occur in the 1-isopropylpiperidine
fragments located in S4.
In both cases, the chlorine atom is involved in nonbonded
interactions with the phenyl ring of Y228 located on the back
wall of S1 (Figure 6). The interactions have a face-on geometry
with the Cl atoms distant 3.9 Å from the centroid of the Tyr
aromatic ring, in accordance with the results expected for a
Cl−π interaction.16 The amide nitrogen of 2 and 3 forms an
extended H-bond network with G216 and a water molecule,
which contributes to strengthen the binding of ligand to the
protein. In the case of 3 (Figure 6a), the G219-NH forms a H-
bond with the ligand ether oxygen connecting the central
phenyl linker with the propylene-linked glucose (N−O distance
= 3.4 Å).
The main difference between the binding modes of 3 (Figure
6a) and 2 (Figure 6b) resides in the orientation of the
propylene-glucose side chains (PG) into the binding site of
hfIIa. Figure 5 shows that the PG fragment in 3 lies on the lower
part of the access channel to the active site, whereas it tilts and
shifts toward S4 and S2 (loop 60) when β-D-glucose is
8567
dx.doi.org/10.1021/jm5010754 | J. Med. Chem. 2014, 57, 8563−8575

Journal of Medicinal Chemistry
Article
as S129b-Oγ and w453 (Figure 6a), interacting with them to
stabilize the enzyme−inhibitor complex. As a matter of fact, the
absence of H-bonds involving the peracetylated glucose moiety
of 2, along with other factors such as steric repulsion, may
prevent interactions with the access channel to the active site by
the PG fragment, which instead is oriented toward the solvent
outside the protein close to S2 and S4 (Figure 6b). The
stabilization effect of these H-bonds in 2 and 3 can be inferred
by examining the refined values of the crystallographic B-factors
of their atoms (Figure 7). Atoms having higher B-factors (high
vibrational motions) are colored in red, whereas atoms having
lower B-factors (low vibrational motions) are colored in blue.
Figure 7. Crystallographic B-factor for inhibitors 2 (right) and 3 (left).
Atoms are colored depending on their B-factors: from blue for atoms
having lower B-factor values to red for atoms having higher B-factor
values.
The largest differences can be observed in the glucose
portions (prevailing red in 2, prevailing blue in 3), but
significant differences also occur in the remaining part of the
two ligand structures; the structure of 3 is prevalently colored
in blue, suggesting that the whole molecule is strongly anchored
to the protein, whereas the central portion of 2 (encompassing
the phenyl linker and the isonipecotamide P3/P4 moiety) is
less stable than in compound 3, with effects on the stability of
its 5-(5-chlorothiophen-2-yl)isoxazol-3-yl group as the S1
binder. This result suggests that the stabilization effect due to
formation of H-bonds with K224 and R221a is not limited to
the glucose moiety but affects the stability of binding of the
whole ligand structures up to the S1 binding fragment. The
effect of stabilization also affects the B-factor of the protein
residues close to the two ligands, as shown in Table 4, which
summarizes the average B-factors of the amino acid residues
and molecules (DMSO, water) closer than 4 Å to the ligands.
For each residue, the B-factor in 4N3L (3) is lower than that
in 4NZE (2), highlighting that the binding interactions in
4N3L are more stable than in 4NZE. The B-factors relating to
DMSO (DMS305 and 256 for 4N3L and 4NZE, respectively)
and water molecules (HOH474 and 375 for 4N3L and 4NZE,
respectively), which participate to form an H-bond network
with the two ligands near the entrance to the S1 pocket (Figure
6), are those mostly affected by the acetylation of the glucose
OH groups in the ligand structures, whereas G216 is the
residue in close proximity to S3 that shows the larger difference
in the B-factor between the 4N3L and 4NZE structures.
Significant differences (ca. 10%) also emerge for the residues
R221a and W60d belonging to Na+ binding loop and the
specificity site S2, respectively.
Figure 6. Binding modes of inhibitors 3 (a) and 2 (b) to human α-
thrombin as revealed by X-ray crystallography. Compounds 2 and 3
are shown as stick representations with O in red, N in blue, S in
yellow, and Cl in dark green color; C atoms are in green and cyan
colors for 3 and 2, respectively. Protein residues interacting with
ligands are in stick representation with O in red, N in blue, and C in
gray. H-bonds with H-donor/H-acceptor distance <3.5 Å are shown as
black dashed lines; weaker H-bonds (3.5 < H-donor/H-acceptor
distance <4 Å) are in magenta dashed lines.
peracetylated as in 2. In the crystal structure of hfIIa complexed
with 3, the orientation of the PG fragment is driven by an H-
bond network formed between the basic side chains of R221a
and K224, which are two important residues involved in the
Na+ binding site (SBS), and O1′, O3′, and O5′ of the glucose
moiety (Figure 6a). In detail, the guanidyl group of R221a
makes H-bonds with the glucose O1′, O3′, and O5′, whereas
the K224 amino group is H-bonded with O3′. Anchored in the
lower part of the access channel to the fIIa active site, the
glucose moiety of 3 approaches the S129b side chain of a
symmetry-related fIIa molecule, allowing H-bond formation
between S129b-Oγ, glucose O6′, and a water molecule (HOH-
453, Figure 6a). Although these H-bonds may be due to a
molecular rearrangement of the protein in the crystal of the
adduct with compound 3, they provide evidence that only
“free” OH groups in the glucose moiety of 3, likely due to a
higher density of negative charge compared to the OAc groups
in 2, could more easily retrieve H-donor/acceptor groups, such
8568
dx.doi.org/10.1021/jm5010754 | J. Med. Chem. 2014, 57, 8563−8575

Journal of Medicinal Chemistry
Article
of physiological and synthetic ligands. Indeed, the Na+-bound
fast form is specific toward fibrinogen, the thrombin receptors,
thrombomodulin and antithrombin III, whereas the Na+-free
slow form is specific toward protein C.18 From the structural
point of view, the SBS lies within a cylindrical cavity formed by
three antiparallel β-strands of the B chain, diagonally crossed by
the sequence E188−E192 and shaped by the loop D221−K224,
which connects the last two β-strands. The sequence C220−
G226, encompassing the Na+-binding loop and part of the last
β-strand of the B chain, is almost completely conserved in
thrombin from different species.19 Sodium ion is octahedrally
coordinated by the carbonyl O atoms of R221a and K224 and
four buried water molecules, and the Na+-binding environment
is stabilized by three ion pairs, namely R221a···E146 (a residue
of the neighbor autolysis loop), K224···E217, and a bidentate
ion pair formed by R187 with D221 and D222 flanking R221a.
A disulfide bond between C220 and C191 anchors the Na+-
binding loop to the neighbor sequence 187−192, which
includes two critical residues, namely D189 at the bottom of
S1 (its side chain is 5 Å away from the bound Na+) and E192 in
close proximity to S1.20
Actually, a closer examination of the crystal structures 4NZE
and 4N3L did not reveal any significant difference in their
specificity sites’ folding, irrespective of the small-molecule
ligand structure, and both structures reproduce the fast form of
thrombin, which is overrepresented in PDB. This could be due
to the high Na+ concentration used in our crystallization
experiments. In addition, it should be noted that there are cases
in which putative slow forms of thrombin (e.g., 1SGI) are
conformationally indistinguishable from the fast forms.21
Nonetheless, it is reasonable to assume that multiple and
strong H-bond interactions between the PG fragment of the
most potent inhibitor 3 and R221a and K224, and G219 as
well, all residues belonging to the Na+-binding loop, could
allosterically perturb the thrombin specificity sites, especially
allowing a closer (more efficient) binding of the S1 and S3/S4
binders of the inhibitors. In the case of binding of 3 to fIIa, it is
likely that when the SBS is perturbed through H-bond
interactions of the glucose moiety with key residues (i.e.,
R221a and K224), perturbation at other sites (S1 and S3/S4)
reduces their specificity beyond simple additivity of the single
Table 4. Crystallographic B-Factor of the Thrombin
Residues and Molecules Closer than 4 Å to the Ligands 2
and 3, Listed in Descending Order of the B-Factor
a
Difference between the Two Crystals
4NZE
(2)
4N3L
(3)
molecule/residue
location/specificity site
DMS
48.49
38.51
27.63
38.06
48.95
36.04
31.72
31.72
25.62
28.55
24.99
14.10
31.33
23.57
16.92
27.71
38.95
27.25
24.22
24.88
18.99
23.25
20.17
9.58
DMSO in the active site
water in the active site
close proximity to S3
Na+ binding loop
S2 (loop 60)
HOH
Gly216
Arg221a
Trp60d
Glu192
Glu217
Gly219
Cys191
Cys220
Leu99
close proximity to S1
close proximity to S3
Na+ binding loop
Na+ binding loop
Na+ binding loop
close to S2
Val213
last β-strand close to Na+ binding
loop
Glu97a
Lys224
Ser195
Ala190
Tyr228
Trp215
Asp189
Gly226
Ile174
42.09
25.66
19.25
18.22
12.90
19.82
20.92
15.00
22.84
27.11
38.13
21.99
15.73
15.08
9.94
S4
Na+ binding loop
catalytic residue
S1
S1
16.92
18.05
13.06
21.48
26.46
S3
S1
Na+ binding loop
S4
Tyr60a
S2 (loop 60)
a
Location and/or specificity site of each residue is shown in the last
column.
As mentioned above, R221a and K224, which form a H-bond
network with the glucose moiety of 3 (and not with the
peracetylated glucose of 2), are two critical residues of the Na+
binding loop in thrombin, which is known for allosterically
regulating the enzyme catalytic activity (the residues of the
catalytic triad are about 20 Å away from the bound Na+) and
influences its specificity sites.17 The binding of Na+ triggers the
transition from slow (anticoagulant) to fast (procoagulant)
thrombin forms, affecting thrombin specificity toward a number
Figure 8. Effect of compounds 1−3 and argatroban on thrombin generation as assessed by the calibrated automated thrombography. Citrated
human plasma was spiked with the indicated concentrations of test compounds or vehicle, and clotting was induced by low tissue factor
concentration, phospholipids, and calcium (details in Experimental Section). The following parameters of the thrombin generation curves were
considered: (i) lag-time (left panel), (ii) peak thrombin concentration (middle panel), and (iii) endogenous thrombin potential (ETP, right panel).
Results are expressed as percent of the values recorded in the corresponding control plasma samples supplemented with the vehicle alone. Data are
the means SD of three independent measurements carried out on different plasma samples.
8569
dx.doi.org/10.1021/jm5010754 | J. Med. Chem. 2014, 57, 8563−8575

Journal of Medicinal Chemistry
Article
perturbations, allowing further stabilization of the interactions
with particular residues in these sites (e.g., H-bonds with G216-
CO and G219-NH close to S3). The large and significant
superadditivity of free energy contributions of P3/P4 and PG
fragments into compound 3 to thrombin binding (+15.5 kJ·
mol−1), which revealed a positive cooperative binding of the
inhibitor fragments, is consistent with such a kind of allosteric
regulation of the specificity sites.
Anticoagulant Activity. The anticoagulant activity of our
most potent fXa/fIIa inhibitors 1−3 was investigated in vitro by
measuring their effects on tissue factor-induced thrombin
generation (TG). The assay was performed in human plasma,
using the calibrated automated thrombography (CAT) as
described by Hemker et al.,22 and compared with the direct
thrombin inhibitor (DTI) argatroban. The TG assay is
considered one of the most relevant tests for assessing the
hemostatic function and is one of the most sensitive for
monitoring the anticoagulant activity of drugs targeting fXa. At
variance with the classical clotting test, such as aPTT (activated
partial thromboplastin time) and PT (prothrombin time), the
TG assay, thanks to the use of several parameters, each
reproducing a distinct phase of the clotting cascade, makes it
possible to distinguish between drugs with different mecha-
nisms of action.23
straightforward. For example, the thrombin inhibitors hirudin
and PPACK were unable to stimulate fibrinolysis, even at
concentrations causing a strong anticoagulant effect.26 More-
over, the synthetic pentasaccharide fondaparinux, an indirect
selective fXa inhibitor (its anticoagulant activity is mediated by
the high affinity binding to the anticoagulant factor
antithrombin III), has been shown to stimulate fibrinolysis at
concentrations much higher than achieved in patients.27
The effect of our compounds in an in vitro model of plasma
clot lysis is shown in Figure 9. Results are expressed as lysis
Pooled citrated human plasma was spiked with 1−3 or
argatroban at concentrations of 0.01−10 μM, and the inhibitory
potencies of the test compounds were assessed by three CAT
parameters: (i) the lag-time (LT) that is the time required to
generate the first traces of thrombin, (ii) the thrombin peak
(TP), and (iii) the endogenous thrombin potential (ETP),
which is a measure of the total thrombin formed (i.e., the area
under the TG curve). The concentration-dependent effects of
the test compounds on the CAT parameters are shown in
Figure 8.
Figure 9. Effect of compounds 1−3 and argatroban on plasma
fibrinolysis. Citrated human plasma was spiked with the indicated
concentrations of test compounds or vehicle. The lysis time of tissue
factor-induced clots supplemented with t-PA was evaluated by a
turbidimetric assay (details in Experimental Section). Results are
expressed as percent of the lysis time recorded in the same plasma
samples supplemented with the vehicle alone. Data are the means
SD of three independent measurements carried out on different
plasma samples.
LT was variably prolonged by the three molecules, the β-D-
glucose-bearing compound 3 being the most active and 1 the
least active. At the highest tested concentration (10 μM), 3
prolonged the lag-time more than 15 times over the control,
while 1 and 2 showed about 3-fold and 9-fold LT
prolongations, respectively. Compared with argatroban, 2
showed similar behavior, whereas 3 caused a significantly
greater delay of thrombin appearance. Fairly similar results
were obtained when thrombin peak was considered. Indeed, the
strongest inhibition was observed with 3, which caused a 50%
reduction at about 0.3 μM, as compared to 1.2 and 10 μM for
compounds 2 and 1, respectively. Interestingly, considering the
IC50 values, both the glucosyl derivatives 2 and 3 proved to be
more potent than argatroban (IC50 1.9 μM). Concerning the
ETP parameter, compound 1 did not induce any appreciable
decrease, 2 displayed an intermediate inhibitor activity with
IC50 of 10 μM, and 3 was the most active, with IC50 value
comparable to that of argatroban (3.4 and 2.4 μM,
respectively).
Profibrinolytic Activity. In the past decade, it has become
increasingly clear that the enhancement of fibrinolysis is an
important component of the antithrombotic activity of
anticoagulant drugs.24 Because thrombin has been shown to
inhibit fibrinolysis through different mechanisms,25 the
profibrinolytic activity of anticoagulants is almost entirely due
to their ability to inhibit thrombin formation or activity.
However, not all anticoagulants promote fibrinolysis with the
same efficiency,25 suggesting that the relationship between
inhibition of coagulation and stimulation of fibrinolysis is not
time, that is, the time needed to lyse the fibrin clot by 50% (the
shorter the lysis time the greater the profibrinolytic activity of
the compound). In this model, the maximum effect that can be
achieved is a reduction of lysis time of 50−60%, which is
obtained when the antifibrinolytic effects of thrombin are
completely neutralized. Therefore, the EC50 corresponds to the
concentration that shortens the lysis time by 25−30%. In
reasonable agreement with the ETP results reported above,
compound 3 was the most active, with EC50 of 0.3 μM,
followed by compound 2, with EC50 of 1.8 μM, and compound
1, which was practically inactive. In this model, the EC50 of
argatroban was 0.6 μM.
As a matter of fact, compound 1, which proved a potent and
highly selective fXa inhibitor (Ki = 0.6 nM), with a low
inhibition potency against fIIa (Ki = 11 μM), did not show
significant effects on TG and fibrinolysis up to the
concentration of 10 μM. Its chemical modification, through
connection of a propylene-linked β-D-glucose in suitable
position of the central phenyl ring, led to the O-glucoside
derivative 3, which not only enhanced binding affinity to fXa
(Ki = 0.09 nM) and fIIa (Ki = 100 nM) but also proved able to
stimulate fibrinolysis at the same submicromolar concentrations
producing the anticoagulant effects assessed by the in vitro TG
assay, its efficacy being comparable to that of argatroban, a DTI
currently in clinical use. Compound 2, that is, the peracetyl
precursor of 3, showed intermediate potency between 1 and 3
on enzymes’ inhibition, as well as on anticoagulant and
profibrinolytic activities.
8570
dx.doi.org/10.1021/jm5010754 | J. Med. Chem. 2014, 57, 8563−8575

Journal of Medicinal Chemistry
Article
CONCLUSIONS
EXPERIMENTAL SECTION
■
■
General Methods. Several compounds were synthesized according
to known procedures with slight modifications; their melting points
and spectral data were in full agreement with those reported in
literature, and no effort was made at this stage to optimize the yields.
Unless otherwise stated, starting materials, and all chemicals and
solvents as well, were purchased from Sigma-Aldrich. Chromato-
graphic separations were performed on silica gel 60 for column
chromatography (Merck 70−230 mesh, or alternatively 15−40 mesh
for flash chromatography). Melting points were determined by using
the capillary method on a Stuart Scientific SMP3 electrothermal
apparatus and are uncorrected. IR spectra were recorded using KBr
disks on a PerkinElmer Spectrum One FT-IR spectrophotometer
(PerkinElmer Ltd., Buckinghamshire, UK), and the most significant
The propylene-linked β-D-glucosyl derivative 3 has been
discovered which inhibits fXa and fIIa with different potencies
(Ki values equal to 0.090 and 100 nM, respectively). Compared
to the parent compound 1, compound 3, while remaining a
fXa-selective inhibitor, increases its binding affinity to fIIa (110-
fold) much more than to fXa (7-fold). Thanks to its unique
features, compound 3 is different from the direct fXa inhibitors
currently in use in the clinic, i.e., rivaroxaban and apixaban,
which display a greater than 10000-fold selectivity for fXa over
other human coagulation proteases, including thrombin.28,29
In this study, we virtually decomposed 3 into smaller
fragments, synthesized them, and determined their inhibition
effects on both fIIa and fXa in order to explore ΔG
“superadditivity” of the linked fragments as proof of possible
cooperative effects in enzyme−ligand interactions. Our
experimental fragment deconstruction study highlighted a
long-range binding cooperative effect between the 1-isopro-
pylpiperidine-4-carboxamide S3/S4 binder and the C3-alkyl-
linked β-D-glucose fragment, which resulted in stronger fIIa
(15.5 kJ·mol−1) than fXa (2.8 kJ·mol−1), thereby suggesting
that the interactions attained by the two fragments can mutually
increase in their strength.
The crystal structure of the hfIIa−3 complex provided a
picture of the binding mode of 3, which includes, besides the
expected interactions of the 2-chlorothiophene and 1-
isopropylpiperidine moieties with the S1 and S4 pockets,
respectively, an H-bond network between the glucose O1′, O3′,
and O5′ and the side chains of R221a and K224, which are two
key residues belonging to the thrombin SBS. Similar H-bond
interactions are not formed when β-D-glucose is peracetylated
as in 2, and the sugar-bearing side chain in fact tilts and shifts
toward S4 and S2 (loop 60). It has been established that SBS,
depending upon the Na+ concentration, may allosterically
regulate the thrombin catalytic activity, also affecting the
specificity sites.18 It is likely that when SBS is perturbed, due to
H-bond interactions between the glucose moiety in 3 and
R221a and K224 side chains, it could affect the S1−S3
environments, reducing their specificity for the ligand moieties
and/or allowing further stabilization of the interactions with
particular residues in these sites (e.g., H-bonds with G216-C
O and G219-NH close to S3). Reasonably, the large
“superadditivity” (+15.5 kJ·mol−1) of the ΔG contributions to
thrombin binding by P3/P4 and PG fragments suitably placed
in compound 3 could reflect the SBS-dependent allosteric
regulation of the specificity sites.
1
absorption bands are listed. H NMR spectra were recorded at 300
MHz on a Varian Mercury 300 instrument. Chemical shifts are
expressed in δ and the coupling constants J in hertz (hz). The
following abbreviations are used: s, singlet; d, doublet; dd, doublet−
doublet; t, triplet; q, quartet; qt, quintet; m, multiplet. Signals due to
NH and OH protons were located by deuterium exchange with D2O.
Mass spectra of final products and intermediates were recorded on an
Agilent GC/MS 6890−5973 using electrospray ionization (ESI).
For newly synthesized tested compounds, HRMS data, as a further
criterion of compound identity, and combustion analysis, as a criterion
of purity (≥95%), have been supplied. HRMS were recorded on
Agilent 6530 Accurate-Mass Q-TOF LC/MS system (laboratory
service of the Consorzio Interuniversitario di Ricerca in Chimica dei
Metalli nei Sistemi Biologici, CIRCMSB, Bari, Italy). Elemental
analyses (C, H, N) were performed on an Euro EA3000 analyzer
(Eurovector, Milan, Italy), and the results agreed to within 0.40% of
the theoretical values. The syntheses of the β-D-glucose-containing
compounds 9−14 are described below, whereas experimental details
on synthesis, along with analytical and spectral data, of non-
glycosylated fragments and intermediates of the reference molecules
1 and 3 (i.e., compounds 5, 6, 7a, 7b, 8a, and 8b) are reported in
Supporting Information.
3-(2-Nitrophenoxy)propyl 2,3,4,6-tetra-O-acetyl-β-D-glucopyra-
noside (9). K2CO3 (620 mg, 4.50 mmol) was added to a solution of
2-nitrophenol (300 mg, 2.14 mmol) in dry DMF (10 mL), and the
mixture was stirred at room temperature for 30 min. Then, 3-
bromopropyl 2,3,4,6-tetra-O-acetyl-β-D-glucopyranoside (1.0 g, 2.14
mmol) dissolved in dry DMF (5 mL) was added, and the reaction
mixture was stirred overnight at room temperature, diluted with water
(250 mL), and extracted with EtOAc (3 × 50 mL). The combined
organic layers were sequentially washed with 2 N NaOH (3 × 30 mL)
and brine (50 mL), dried (Na2SO4), filtered, and concentrated under
reduced pressure to afford an oil, which was purified by flash
chromatography (mobile phase n-hexane/EtOAc, 1:1 v/v) to provide
9 as white solid (650 mg, 70% yield), mp 129−131 °C. IR (KBr):
2966, 2917, 1755, 1659, 1521, 1367, 1231, 1166, 1038, 858, 742 cm−1.
1H NMR (300 MHz, CDCl3) δ 7.82 (d, J = 9.0 Hz, 1H), 7.52 (t, J =
On the basis of the TG assay, which is the most
pathophysiologically relevant clotting test currently used, we
showed that compound 3 has the strongest anticoagulant
activity. It prolongs the lag-time and reduces both thrombin
peak and ETP with an efficacy comparable to that of the
intravenous DTI argatroban. Another major feature of 3 is its
ability to markedly stimulate fibrinolysis at the same
concentrations producing the anticoagulant effect. In contrast,
the parent compound 1, which showed about 20000-fold
selectivity for fXa over thrombin, does not significantly affect
both TG and fibrinolysis up to the concentration of 10 μM.
While the SARs and crystallographic information presented
here will be valuable for the design of new efficacious fIIa and
fXa inhibitors, compound 3 is worthy of further investigation
aimed at better characterizing its potential as antithrombotic
agent.
7.5 Hz, 1H), 7.07 (d, J = 8.0 Hz, 1H), 5.21 (t, J = 9.0 Hz, 1H), 5.06 (t,
J = 9.0 Hz, 1H), 4.97 (t, J = 9.0 Hz, 1H), 4.53 (d, J = 8.0 Hz, 1H), 4.23
(d, J = 6.0 Hz, 1H), 4.16 (t, J = 6.0 Hz, 2H), 4.09 (m, 2H), 3.83 (m,
2H), 2.09 (m, 2H), 2.07 (s, 3H), 2.05 (s, 3H), 2.01 (s, 3H), 1.98 (s,
3H). ESIMS m/z 550 (M + Na)+.
3-(2-Aminophenoxy)propyl 2,3,4,6-tetra-O-acetyl-β-D-glucopyra-
noside (10). A suspension of 9 (360 mg, 0.68 mmol) and 10% Pd-C
(50 mg) in EtOAc (50 mL) was stirred in H2 atmosphere for 2 h and
then filtered on a Celite Pad. After the solvent removal, compound 10
was obtained as brown oil (290 mg, 85% yield). IR (KBr): 3441, 2960,
2893, 1755, 1622, 1507, 1368, 1225, 1166, 1041, 908, 750, 613 cm−1.
1H NMR (300 MHz, CDCl3) δ 7.82 (d, J = 9.0 Hz, 1H), 7.52 (t, J =
7.5 Hz, 1H), 6.76 (m, 4H), 5.19 (t, J = 9.0 Hz, 1H), 5.08 (t, J = 9.0 Hz,
1H), 5.00 (t, J = 9.0 Hz, 1H), 4.52 (d, J = 8.0 Hz, 1H), 4.26 (d, J = 6.0
Hz, 1H), 4.13 (m, 2H), 4.08 (m, 2H), 3.85 (m, 2H), 2.09 (m, 2H),
2.08 (s, 3H), 2.02 (s, 3H), 2.00 (s, 3H), 1.99 (s, 3H). ESIMS m/z 520
(M + Na)+.
8571
dx.doi.org/10.1021/jm5010754 | J. Med. Chem. 2014, 57, 8563−8575

Journal of Medicinal Chemistry
Article
N-(2-{3-[(2,3,4,6-Tetra-O-acetyl-β-D-glucopyranosyl)oxy]-
propoxy}phenyl)piperidine-4-carboxamide (12). HOBt (165 mg,
1.21 mmol) and DCC (250 mg, 1.21 mmol) were added to a solution
of 1-(t-butoxycarbonyl)piperidine-4-carboxylic acid (280 mg, 1.21
mmol) in dry DCM (15 mL). Compound 10 (600 mg, 1.21 mmol)
was added after 30 min of stirring, and the reaction mixture was further
stirred at room temperature for 48 h. DCU was removed by filtration,
and the filtrate was concentrated under reduced pressure. The oily
residue was dissolved in EtOAc (50 mL) and sequentially washed
twice with saturated 10% Na2CO3, 2 N HCl, and brine. The organic
phase was dried (Na2SO4), filtered, and concentrated under reduced
pressure to provide the N-Boc-protected compound 11 as an oil (600
mg, 70% yield). After removal of the Boc protecting group with
redistilled TFA (21 mL) in CHCl3 (10 mL), the TFA salt 12 was
obtained as brown oil. The oil residue was suspended in water (30
mL), and 2 N NaOH was added to alkaline pH. The aqueous
suspension was then extracted with EtOAC (3 × 30 mL), and the
combined organic layers were dried (Na2SO4) and concentrated under
reduced pressure to afford 12 free base as pale-brown oil (420 mg,
80% yield), which was used in the next reaction without further
purification. IR (KBr) 3430, 1750, 1625, 1220, 1168, 1041, 750, 613
column (Phenomenex Italy, Castel Maggiore, BO, Italy) as the
nonpolar stationary phase, and retention data were measured at regular
increments of the volume fraction of MeOH in pH 4.5 phosphate
buffer (0.05 M). All the RP-HPLC measurements were carried out at
temperature of 25
0.2 °C, flow-rate of 1.0 mL min−1 and UV
detection at 288 nm wavelength, using a Waters HPLC 1525
multisolvent delivery system equipped with a Waters 2487 variable
wavelength UV detector (Waters Assoc., Milford, MA, USA). Capacity
factors (k′) of each compound at different mobile phase compositions
(0.05 increments of the MeOH volume fraction, φ, ranging between
0.8 and 0.4) were calculated as k′ = (tR − t0)/t0, where tR is the
retention time of the solute and t0 is the column dead time, measured
as the solvent front. The log k′ values increased linearly (r2 > 0.95)
with decreasing ϕ. Linear regression analysis was performed on at least
five data points (the lowest MeOH volume factions) for each
compound and the linear relationship extrapolated to 100% aqueous
mobile phase to yield log k′w value.
Aqueous Solubility. The aqueous solubility of compounds 1−3
was determined at pH 7.4 in a 0.05 M phosphate buffer (containing
0.15 M KCl as ionic strength regulator) in deionized water. An excess
of each compound (4−10 mg) was added to 1 mL of buffered
solution, and the suspension was shaken 1 h at 25 °C. The suspension
was then maintained for 15 min at the same temperature and without
shake, to ensure the solubility equilibrium. Subsequently, the
supernatant was filtered (Millipore 0.45 μm), and the solution was
analyzed by HPLC. Analyses were performed on a Phenomenex
Gemini C18 column (5 μm, 150 mm × 4.6 mm i.d.) using MeOH and
0.05 M phosphate buffer solution (pH 4.5) mixed in different fraction
composition depending upon the retention of the analyte as the
mobile phase (flow rate 1 mL·min−1; detection at 288 nm).
Calibration curves were obtained by measuring peak areas for each
compound at known concentrations.
Stability in Water and Human Plasma. Stability of compounds
1−3 was determined either in phosphate buffer solution (pH 7.4) and
human plasma at 37 °C. For each compound, the stock solution in
ACN (10 mM) was diluted to a final incubation concentration of 250
μM. The incubation at 37 °C was stopped at 1, 2, and 4 h and the
samples treated and analyzed by HPLC using the above analytical
conditions.
1
cm−1. H NMR (300 MHz, DMSO-d6) δ 10.05 (s, 1H), 7.85 (d, J =
9.0 Hz, 1H), 7.50 (t, J = 7.5 Hz, 1H), 6.77 (m, 2H), 5.21 (t, J = 9.0 Hz,
1H), 5.12 (t, J = 9.0 Hz, 1H), 5.05 (t, J = 9.0 Hz, 1H), 4.55 (d, J = 8.0
Hz, 1H), 4.26 (d, J = 6.0 Hz, 1H), 4.15−4.10 (m, 2H), 4.08−4.00 (m,
2H), 3.90−3.85 (m, 2H), 3.50−3.30 (m, 2H), 3.25−3.15 (m, 2H),
2.65−2.50 (m, 1H), 2.15−2.08 (m, 2H), 2.08 (s, 3H), 2.02 (s, 3H),
2.00 (s, 3H), 1.99 (s, 3H), 1.90−1.80 (m, 2H), 1.75−1.60 (m, 2H).
ESIMS m/z 631 (M + Na)+.
1-Isopropyl-N-(2-{3-[(2,3,4,6-tetra-O-acetyl-β-D-glucopyranosyl)-
oxy]propoxy}phenyl)piperidine-4-carboxamide (13). Na(CN)BH3
(100 mg, 1.77 mmol) and 1.25 M HCl solution in MeOH (1 mL)
were added to a 0 °C solution of 12 (400 mg, 0.82 mmol) in Me2CO
(5 mL) and MeOH (15 mL), and the mixture was stirred 1 h at 0 °C
and overnight at room temperature. After the solvent removal, the
residue was dissolved in EtOAc (50 mL), and the organic phase was
washed twice with brine, dried (Na2SO4), filtered, and concentrated.
The residue was purified by flash chromatography (mobile phase
DCM/MeOH, 95:5 v/v) to afford 13 as brown oil (215 mg, 50%
1
yield). IR (KBr) 3320, 1748, 1625, 1225, 1040, 750, 613 cm−1. H
For the stability study in the aqueous medium, the stock solution
(0.25 mL) was added to 9.75 mL of 0.05 M phosphate buffer solution
at pH 7.40 (0.15 M KCl as ionic strength regulator). At t0 and each
other fixed time, a solution sample (0.5 mL) was withdrawn, filtered
(Millipore 0.45 μm), and analyzed by HPLC. For the stability
measurements in human plasma, the stock solution (0.037 mL) was
added to 1.463 mL of preheated (37 °C) human serum (lyophilized
and reconstituted with 4 mL of deionized water). The incubation at 37
°C was stopped at each fixed time by adding 0.4 mL of ice-cold ACN
to 0.2 mL of serum solution. The sample was vortexed for 1 min and
then centrifuged for 10 min at 3500 rcf before HPLC injection of the
supernatant. The percentage of the remaining test compound was
measured by monitoring the peak area of the chromatogram.
Crystallization of Human Thrombin with Inhibitors 2 and 3.
Human α-thrombin supplied in 50% (v/v) glycerol/H2O solution and
hirugen peptide (exosite inhibitor having sequence Ac-
NMR (300 MHz, DMSO-d6) δ 10.05 (s, 1H), 7.85 (d, J = 9.0 Hz,
1H), 7.50 (t, J = 7.5 Hz, 1H), 6.77 (m, 2H), 5.21 (t, J = 9.0 Hz, 1H),
5.12 (t, J = 9.0 Hz, 1H), 5.05 (t, J = 9.0 Hz, 1H), 4.55 (d, J = 8.0 Hz,
1H), 4.26 (d, J = 6.0 Hz, 1H), 4.15−4.10 (m, 2H), 4.08−4.00 (m,
2H), 3.90−3.85 (m, 2H), 3.50−3.30 (m, 2H), 3.25−3.15 (m, 2H),
3.15−3.05, (m 1H), 2.65−2.50 (m, 1H), 2.15−2.08 (m, 2H), 2.08 (s,
3H), 2.02 (s, 3H), 2.00 (s, 3H), 1.99 (s, 3H), 1.90−1.80 (m, 2H),
1.75−1.60 (m, 2H), 1.26 (d, J = 6.5 Hz, 6H). ESIMS m/z 673 (M +
Na)+.
N-{2-[3-(β-D-Glucopyranosyloxy)propoxy]phenyl}-1-isopropylpi-
peridine-4-carboxamide (14). MeONa (100 mg, 1.85 mmol) was
added to a 0 °C solution of 13 (200 mg, 0.31 mmol) in MeOH (10
mL), and the mixture was stirred overnight at room temperature. After
concentration under reduced pressure, the solid residue was purified
by silica gel flash chromatography (mobile phase DCM/MeOH, 95:5
v/v), providing 14 as white solid (110 mg, 70% yield), mp 134−136
°C. IR (KBr) 3325, 1650, 1260, 1040, 750 cm−1. 1H NMR (300 MHz,
DMSO-d6) δ 9.95 (s, 1H), 7.82 (d, J = 9.0 Hz, 1H), 7.51 (t, J = 7.5 Hz,
1H), 6.80−6.75 (m, 2H), 5.19 (t, J = 9.0 Hz, 1H), 5.10 (t, J = 9.0 Hz,
1H), 5.05 (t, J = 9.0 Hz, 1H), 4.90 (s br, 4H), 4.55 (d, J = 8.0 Hz, 1H),
4.25 (d, J = 6.0 Hz, 1H), 4.15−4.10 (m, 2H), 4.10−4.00 (m, 2H),
3.90−3.85 (m, 2H), 3.50−3.30 (m, 2H), 3.25−3.15 (m, 2H), 3.15−
3.05, (m 1H), 2.65−2.50 (m, 1H), 2.15−2.08 (m, 2H), 1.85−1.70 (m,
2H), 1.65−1.50 (m, 2H), 1.10 (d, J = 6.5 Hz, 6H). ESIMS m/z 505
(M + Na)+. HRMS (ESI) calcd for C24H38N2O8 ([M + H]+)
483.2701; found 483.2597. Anal. Calcd for C24H38N2O8·2.5H2O: C,
54.60; H, 8.20; N, 5.31. Found: C, 54.81; H, 7.76; N, 5.40.
−
NGDFEEIPEEY(SO3 )L) used for crystallization experiments were
purchased from Haematologic Technologies Inc. (Essex Junction, VT,
USA) and GL Biochem (Shanghai), respectively. To avoid the fast
autoproteolysis of thrombin during crystallization, a 0.25 mM solution
containing protein was preliminarily incubated in the presence of 2.5
mM hirugen peptide for 4 h at 20 °C. The resulting thrombin−
hirugen adduct solution was centrifuged (10.000 rpm at 4 °C) by using
Amicon Ultra centrifuge filter having a cut off of 10 kDa (Millipore
Corporation) to exchange the starting buffer containing glycerol with a
solution containing 20 mM HEPES pH 7.1, 750 mM NaCl, and 0.1
mM hirugen. At the end of buffer exchange, the protein−hirugen
adduct was concentrated up to 13 mg·mL−1. Large lanceolate crystals
(100−200 μm) of thrombin−hirugen complex were grown in 1 week
as described by Skrzypczak-Jankun et al.:30 a drop having volume of 2
μL, formed by 1 μL of solution containing 12−13 mg·mL−1 of
Lipophilicity Measurements. The lipophilicity of compounds 1−
3 was determined by a reversed-phase (RP) HPLC method, using a
Phenomenex Synergi C18 (4.6 mm × 150 mm i.d., 5 μm particles)
8572
dx.doi.org/10.1021/jm5010754 | J. Med. Chem. 2014, 57, 8563−8575

Journal of Medicinal Chemistry
Article
protein−hirugen adduct and 1 μL of reservoir solution, containing
28−30% PEG4000, 100 mM HEPES, pH 7.0, and 0.04% NaN3, was
equilibrated against 1 mL of reservoir solution. Crystallization
experiments were performed in vapor diffusion, hanging drop setup,
at 20 °C. Thrombin−hirugen crystals were stabilized for 5 min in a
solution containing 30.9% PEG4000, 100 mM HEPES, pH 7.0, and
0.04% NaN3 and soaked for 3 days in a solution with the same
composition plus 43 mM compound 3 or 100% satd 2. In both cases,
the soaking solution contained 20% (v/v) DMSO to facilitate the
solubilization of the inhibitor molecule.
5 pmol/L recombinant tissue factor (TF) and procoagulant
phospholipids (4 μmol/L, MP Reagent). Analysis was performed in
a final volume of 120 μL, and started upon addition of 20 μL of a
mixture containing 100 mmol/L CaCl2 and 2.5 mmol/L ZGGR-AMC
(Fluca Reagent, Diagnostica Stago, Asnieres, France). Measurements
were taken every 20 s in a Fluoroskan Ascent fluorometer (Thermo
Fisher Scientific, Waltham, MA, USA) and data were analyzed using
the Thrombinoscope software (Diagnostica Stago). The parameters
calculated by the software were the lag-time (LT), the thrombin peak
(TP), and endogenous thrombin potential (ETP). All samples and
calibrators were run at least in duplicate.
Data Collection, Structure Solution, and Refinement.
Diffraction data for thrombin−3 adduct were collected by using a
home diffractometer equipped by a Rigaku R-axis-IIc imaging plate
system, with Cu Kα radiation (IBB-CNR, Napoli, IT), whereas
diffraction data for thrombin−2 adduct were collected at the beamline
I0−04 of the Diamond Light Source (Oxford, UK) by using a
radiation wavelength of 0.978 Å. The oscillation angle was set to 1° for
each data collection, while the sample−detector distance was chosen
according to the diffraction data resolution. Data reduction was
performed by using the XDS program,31 while POINTLESS and
SCALA32 were used to define the space group and to scale the
intensity of the reflections, respectively. The molecular replacement
(MR) program REMO,33 included into ILMILIONE,34 was employed
for phasing. The Protein Data Bank (PDB) file 1HGT,31 having 100%
of homology sequence with the protein used in the crystallization
experiments, was used as search model. For both crystals, the MR
solution was improved by using Phenix.Autobuild wizard (PHENIX
package)35 in the “rebuild-in-place” mode. The COOT software36 was
used to inspect the electron density map and to fit the inhibitor
molecules 2 and 3, whose addition resulted in a decrease of the R and
Rfree values. The protein−ligand adduct structures were refined by
using Phenix.refine37 (PHENIX package) and validated by MolPro-
bity.38 Their atomic coordinates and structure factors have been
deposited in the RCSB PDB under the accession codes 4NZE and
4N3L for the thrombin−2 and −3 adducts, respectively. Information
on unit cell and refinement parameters for each structure are
summarized in Table 3.
Profibrinolytic Activity. The effect of compounds 1−3 on plasma
fibrinolysis was investigated by a turbidimetric assay, which evaluates
the fibrinolysis time of TF-induced plasma clots exposed to exogenous
tissue plasminogen activator (t-PA). The assay was performed as
reported39,40 with slight modifications. First, 100 μL of plasma, 10 μL
of phospholipids (4 μmol/L, MP Reagent, Diagnostica Stago), 5 μL of
t-PA (40 ng/mL), 20 μL of Tris-NaCl buffer (20 mmol/L Tris-HCl,
150 mmol/L NaCl, pH 7.4), and 5 μL of TF (5 pmol/L) were loaded
in microplate wells, after which the clotting reaction was started with
100 μL of CaCl2 (8.3 mmol/L). The plate was then incubated at 37 °C
in a microplate reader (Multiskan FC; Thermo Fisher Scientific), and
the changes in optical density measured at 405 nm every minute for 3
h. Clotting time was defined as the time to reach the midpoint of clear-
to-maximum turbid transition, whereas clot lysis time was defined as
the time from clot formation to the midpoint of the maximum turbid-
to-clear transition.
ASSOCIATED CONTENT
■
S
* Supporting Information
Experimental details on synthesis, along with spectral data (IR,
NMR, MS) and elemental analyses (C, H, N), of non-
glycosylated compounds 5, 6, 7a, 7b, 8a, and 8b. This material
is available free of charge via the Internet at http://pubs.acs.org.
Inhibition Assays of Factor Xa and Thrombin. The test
compounds were assayed in vitro for their inhibitory activity toward
fXa and fIIa, determining the hydrolysis rates of the synthetic
chromogenic substrates monitored at 405 nm. Enzymes and substrates
were used as follows, in final concentrations: 2 nM human fXa and
0.04 mM S-2765 (Z-D-Arg-Gly-Arg-p-NA) from Chromogenix AB-
Instrumentation Laboratories (Milan, Italy), 0.41 U·mL−1 bovine
thrombin from Sigma-Aldrich (Milan, Italy), and 0.05 mM S-2238 (D-
Phe-Pip-Arg-p-NA) from Chromogenix ABInstrumentation Laborato-
ries (Milan, Italy). Enzyme solutions were incubated with DMSO
solutions of the test inhibitor (DMSO did not exceed 1%) in various
concentrations (0.01−100 nM or 0.01−100 μM), before the respective
chromogenic substrates added to start the enzyme kinetics. The kinetic
studies were performed at pH 8. Reactions were initiated with adding
500 μL of substrate solutions and increase in absorbance monitored
for 5 min. The initial velocities were determined and the concentration
of the inhibitor required to diminish by 50% the control velocity
(IC50), calculated by nonlinear (sigmoid) regression. Three
independent IC50 values were determined for calculating inhibition
constants Ki using the Cheng−Prusoff equation.15 To determine the
type of inhibition for the most potent inhibitor 3 and the parent
compound 1, the Lineweaver−Burk eq (1/v vs 1/[S]) was fitted for
varying concentrations of substrates (25−200 μM and 6−400 μM for
fXa and fIIa, respectively) in the absence or presence of inhibitor at
four different concentrations, using fixed amounts of enzymes (2 nM
dor fXa and 0.41 U·mL−1 for fIIa). Replotting the slopes of the above
plots against the inhibitor concentration yielded the Ki value as the x-
axis intercept (Figure 2).
AUTHOR INFORMATION
■
Corresponding Author
*Phone: +39-080-5442781. Fax: +39-080-5442230. E-mail:
cosimodamiano.altomare@uniba.it.
Present Address
∥For G.L.: Drug Design & Discovery, Aptuit, Via A. Fleming,
37135 Verona, Italy.
Author Contributions
M.d.C., G.Z., G.L., and C.D.A. contributed to chemistry and
performed enzymes’ inhibition measurements. B.D.D. and R.C.
carried out crystallographic experiments and data analysis. F.I.
and M.C. carried out evaluation of the antithrombotic
properties and interpretation of the data. C.D.A. designed the
research project and interpreted data. All the authors
contributed to write the manuscript.
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
We thank DIAMOND LIGHT LTD synchrotron (Oxford,
UK) for beam time, Mr. Giosue Sorrentino and Mr. Maurizio
̀
Amendola for their support during X-ray measurements at the
IBB (CNR, Napoli), and Dr. Maria Brunella Aresta for their
support in the crystallization experiments. The support of
CCNC consortium (Collezione Nazionale di Composti
Chimici e Centro Screening) is gratefully acknowledged.
Anticoagulant Activity. The effect of compounds 1−3 on
thrombin generation (TG) in human plasma was evaluated by the
calibrated automated thrombography (CAT) as described by Hemker
et al.23 Briefly, 80 μL of human citrated plasma were dispensed into
round-bottomed 96-well plates and incubated for 10 min at 37 °C with
8573
dx.doi.org/10.1021/jm5010754 | J. Med. Chem. 2014, 57, 8563−8575

Journal of Medicinal Chemistry
Article
(15) Cheng, Y.; Prusoff, W. H. Relation between the inhibition
constant (Ki) and the concentration of inhibitor which causes fifty per
cent inhibition (I50) of an enzymic reaction. Biochem. Pharmacol. 1973,
22, 3099−3108.
(16) 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 α-thrombin. J. Mol. Biol. 1991,
221, 1379−1393.
(17) Imai, Y. N.; Inoue, Y.; Nakanishi, I.; Kitaura, K. Cl−pi
interactions in protein−ligand complexes. Protein Sci. 2008, 17, 1129−
1137.
(18) Di Cera, E.; Dang, Q. D.; Ayala, Y. M. Molecular mechanisms of
thrombin function. Cell. Mol. Life Sci. 1997, 53, 701−730.
(19) Dang, Q. D.; Vindigni, A.; Di Cera, E. An allosteric switch
controls the procoagulant and anticoagulant activities of thrombin.
Proc. Natl. Acad. Sci. U. S. A. 1995, 92, 5977−5981.
ABBREVIATIONS USED
■
aPTT, activated partial thromboplastin time; ACN, acetonitrile;
AF, atrial fibrillation; CAT, calibrated automated thrombog-
raphy; DCU, dicyclohexylurea; DTI, direct thrombin inhibitor;
DVT, deep venous thrombosis; ETP, endogenous thrombin
potential; fIIa, thrombin; fXa, activated factor X; hfIIa, human
thrombin; HEPES, 2-[4-(2-hydroxyethyl)piperazin-1-yl)-
ethanesulfonic acid; HOBt, 1-benzotriazole; INR, international
normalized ratio; LT, lag time; NOACs, new orally active
anticoagulants; PT, prothrombin time; rcf, relative centrifugal
force; RP, reversed phase; TF, tissue factor; TG, thrombin
generation; t-PA, tissue plasminogen activator; TP, thrombin
peak; VKAs, vitamin K antagonists
(20) Banfield, D. K.; MacGillivray, R. T. A. Partial characterization of
vertebrate prothrombin cDNAs: amplification and sequence analysis of
the B chain of thrombin from nine different species. Proc. Natl. Acad.
Sci. U. S. A. 1992, 89, 2779−2783.
REFERENCES
■
(1) Reiffel, J. A. Novel oral anticoagulants. Am. J. Med. 2014, 127,
e16−17.
(2) Haas, S. New oral Xa and IIa inhibitors: updates on clinical trial
results. J. Thromb. Thrombolysis 2008, 25, 52−60.
(21) Zhang, E.; Tulinsky, A. The molecular environment of the Na+
binding site of thrombin. Biophys. Chem. 1997, 63, 185−200.
(22) Huntington, J. A. How Na+ activates thrombina review of the
functional and structural data. Biol. Chem. 2008, 389, 1025−1035.
(3) Alban, S. Pharmacological strategies for inhibition of thrombin
activity. Curr. Pharm. Des. 2008, 14, 1152−1175.
(4) Deitelzweig, S. Practical considerations in the use of novel oral
anticoagulants for stroke prevention in nonvalvular atrial fibrillation.
Cardiovasc. Ther. 2014, 32, 74−81.
́
(23) Hemker, H. C.; Al Dieri, R.; De Smedt, E.; Beguin, S. Thrombin
generation, a function test of the haemostatic−thrombotic system.
Thromb. Haemostasis 2006, 96, 553−561.
(24) Samama, M. M.; Le Flem, L.; Guinet, C.; Gerotziafas, G.;
Depasse, F. Three different patterns of calibrated automated
thrombogram obtained with six different anticoagulants. J. Thromb.
Haemostasis 2007, 5, 2554−2556.
(25) Colucci, M.; Semeraro, N. Thrombin activatable fibrinolysis
inhibitor: at the nexus of fibrinolysis and inflammation. Thromb. Res.
2012, 129, 314−9.
(26) Bouma, B. N.; Mosnier, L. O. Thrombin activatable fibrinolysis
inhibitor (TAFI)how does thrombin regulate fibrinolysis? Ann. Med.
2006, 38, 378−388.
(27) Lisman, T.; Adelmeijer, J.; Nieuwenhuis, H. K.; de Groot, P. G.
Enhancement of fibrinolytic potential in vitro by anticoagulant drugs
targeting activated factor X, but not by those inhibiting thrombin or
tissue factor. Blood Coagulation Fibrinolysis 2003, 14, 557−562.
(28) Ammollo, C. T.; Semeraro, F.; Semeraro, N.; Colucci, M. The
contribution of anti-Xa and anti-IIa activities to the profibrinolytic
activity of low-molecular-weight heparins. Thromb Haemostasis 2009,
101, 782−785.
(5) Wang, Y.; Bajorek, B. New oral anticoagulants in practice:
pharmacological and practical considerations. Am. J. Cardiovasc. Drugs
2014, 14, 175−189.
(6) Hauel, N. H.; Nar, H.; Priepke, H.; Ries, U.; Stassen, J.-M.;
Wienen, W. Structure-based design of novel potent nonpeptide
thrombin inhibitors. J. Med. Chem. 2002, 45, 1757−1766.
(7) Maignan, S.; Guilloteau, J. P.; Pouzieux, S.; Choi-Sledeski, Y. M.;
Becker, M. R.; Klein, S. I.; Ewing, W. R.; Pauls, H. W.; Spada, A. P.;
Mikol, V. Crystal structures of human factor Xa complexed with
potent inhibitors. J. Med. Chem. 2000, 43, 3226−3232.
(8) de Candia, M.; Lopopolo, G.; Altomare, C. Novel factor Xa
inhibitors: a patent review. Expert Opin. Ther. Pat. 2009, 19, 1535−
1580.
(9) de Candia, M.; Liantonio, F.; Carotti, A.; De Cristofaro, R.;
Altomare, C. Fluorinated benzyloxyphenyl piperidine-4-carboxamides
with dual function against thrombosis: inhibitors of factor Xa and
platelet aggregation. J. Med. Chem. 2009, 52, 1018−1028.
(10) Lopopolo, G.; Fiorella, F.; de Candia, M.; Nicolotti, O.; Martel,
S.; Carrupt, P.-A.; Altomare, C. Biarylmethoxy isonipecotanilides as
potent and selective inhibitors of blood coagulation factor Xa. Eur. J.
Pharm. Sci. 2011, 42, 180−191.
(11) de Candia, M.; Fiorella, F.; Lopopolo, G.; Carotti, A.; Romano,
M. R.; Lograno, M. D.; Martel, S.; Carrupt, P. A.; Belviso, B. D.;
Caliandro, R.; Altomare, C. Synthesis and biological evaluation of
direct thrombin inhibitors bearing 4-(piperidin-1-yl)pyridine at the P1
position with potent anticoagulant activity. J. Med. Chem. 2013, 56,
8696−8711.
(29) Turpie, A. G. G. Oral, direct factor Xa inhibitors in development
for the prevention and treatment of thromboembolic diseases.
Arterioscler. Thromb. Vasc. Biol. 2007, 27, 1238−1247.
(30) Wong, P. C.; Pinto, D. J. P.; Zhang, D. Preclinical discovery of
apixaban, a direct and orally bioavailable factor Xa inhibitor. J. Thromb.
Thrombolysis 2011, 31, 478−492.
(31) 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 a-
1379−1393.
(32) Kabsch, W. XDS. Acta Crystallogr., Sect. D: Biol. Crystallogr.
2010, D66, 125−132.
́
thrombin. J. Mol. Biol. 1991, 221,
(12) Lopopolo, G.; de Candia, M.; Panza, L.; Romano, M. R.;
Lograno, M. D.; Campagna, F.; Altomare, C. β-D-Glucosyl conjugates
of highly potent inhibitors of blood coagulation factor Xa bearing 2-
chorothiophene as a P1 motif. ChemMedChem 2012, 7, 1669−1677.
(33) Evans, P. R. Scaling and assessment of data quality. Acta
Crystallogr., Sect. D: Biol. Crystallogr. 2006, D62, 72−82.
(34) Caliandro, R.; Carrozzini, B.; Cascarano, G. L.; De Caro, L.;
Giacovazzo, C.; Mazzone, A.; Siliqi, D. Molecular replacement: the
approach of the program REMO. J. Appl. Crystallogr. 2006, 39, 185−
193.
́
(13) Nazare, M.; Will, D. W.; Matter, H.; Schreuder, H.; Ritter, K.;
Urmann, M.; Essrich, M.; Bauer, A.; Wagner, M.; Czech, J.; Lorenz,
M.; Laux, V.; Wehner, V. Probing the subpockets of factor Xa reveals
two binding modes for inhibitors based on a 2-carboxyindole scaffold:
a study combining structure−activity relationship and X-ray
crystallography. J. Med. Chem. 2005, 48, 4511−4525.
(35) Burla, M. C.; Caliandro, R.; Camalli, M.; Carrozzini, B.;
Cascarano, G. L.; De Caro, L.; Giacovazzo, C.; Polidori, G.; Siliqi, D.;
Spagna, R. IL MILIONE: a suite of computer programs for crystal
structure solution of proteins. J. Appl. Crystallogr. 2007, 40, 609−613.
(14) Salonen, L. M.; Bucher, C.; Banner, D. W.; Haap, W.; Mary, J.-
L.; Benz, J.; Kuster, O.; Seiler, P.; Schweizer, W. B.; Diederich, F.
Cation-pi interactions at the active site of factor Xa: dramatic
enhancement upon stepwise N-alkylation of ammonium ions. Angew.
Chem., Int. Ed. 2009, 48, 811−814.
(36) Adams, P. D.; Afonine, P. V.; Bunkoczi, G.; Chen, V. B.; Davis,
́
I. W.; Echols, N.; Headd, J. J.; Hung, L.-W.; Kapral, R. W.; Grosse-
8574
dx.doi.org/10.1021/jm5010754 | J. Med. Chem. 2014, 57, 8563−8575

Journal of Medicinal Chemistry
Article
Kunstleve, A. J.; McCoy, N. W.; Moriarty, R.; Oeffner, R. J.; Read, D.
C.; Richardson, J. S.; Richardson, G. J.; Terwilliger, T. C.; Zwart, P. H.
PHENIX: a comprehensive Python-based system for macromolecular
structure solution. Acta Crystallogr., Sect. D: Biol. Crystallogr. 2010,
D66, 213−221.
(37) Emsley, P.; Cowtan, K. COOT: model-building tools for
molecular graphics. Acta Crystallogr., Sect. D: Biol. Crystallogr. 2004,
D60, 2126−2132.
(38) Afonine, P. V.; Grosse-Kunstleve, R. W.; Echols, N.; Headd, J. J.;
Moriarty, N. W.; Mustyakimov, M.; Terwilliger, T. C.; Urzhumtsev, A.;
Zwart, P. H.; Adams, P. D. Towards automated crystallographic
structure refinement with phenix.refine. Acta Crystallogr., Sect. D: Biol.
Crystallogr. 2012, D68, 352−367.
(39) Chen, V. B.; Arendall, W. B., III; Headd, J. J.; Keedy, D. A.;
Immormino, R. M.; Kapral, G. J.; Murray, L. W.; Richardson, J. S.;
Richardson, D. C. MolProbity: all-atom structure validation for
macromolecular crystallography. Acta Crystallogr., Sect. D: Biol.
Crystallogr. 2010, D66, 12−21.
(40) Colucci, M.; Binetti, B. M.; Tripodi, A.; Chantarangkul, V.;
Semeraro, N. Hyperprothrombinemia associated with prothrombin
G20210A mutation inhibits plasma fibrinolysis through a TAFI-
mediated mechanism. Blood 2004, 103, 2157−2161.
8575
dx.doi.org/10.1021/jm5010754 | J. Med. Chem. 2014, 57, 8563−8575