Compounds binding to the S2-S3 pockets of thrombin

Article abstract of DOI:10.1021/jm8011849

A set of compounds designed to bind to the S2-S3 pockets of thrombin was prepared. These compounds included examples with no interactions in the S1 pocket. Proline, a common P2 in many thrombin inhibitors, was combined with known P3 residues and P1 substituents of varying size and lipophilicity. Binding constants were determined using surface plasmon resonance (SPR) biosensor technology and were found to be in good agreement with results from an enzyme assay. A dramatic increase in affinity (100-1000 times) was seen for compounds incorporating an amino group capable of forming a hydrogen bond with gly216 in the protein backbone. The ligand efficiency was increased by including substituents that form stronger hydrophobic interactions with the P1 pocket. The binding mode was confirmed by X-ray analysis, which revealed the anticipated binding motif that included hydrogen bonds as well as a tightly bound water molecule. A QSAR model indicated that hydrogen bonding and lipophilicity were important for the prediction of binding constants. The results described here may have implications for how directed compound libraries for shallow protein pockets, like S2 and S3 in serine proteases, can be designed.

Full text of DOI:10.1021/jm8011849

2708
J. Med. Chem. 2009, 52, 2708–2715
Compounds Binding to the S2-S3 Pockets of Thrombin
Mikael Nilsson,*,† Markku Ha¨ma¨la¨inen,‡ Maria Ivarsson,‡ Johan Gottfries,§ Yafeng Xue,§ Sebastian Hansson,§
Roland Isaksson,*,† and Tomas Fex*,§
School of Pure and Applied Natural Sciences, UniVersity of Kalmar, S-391 82 Kalmar, Sweden, GE Healthcare Bio-Sciences AB,
S-754 50 Uppsala, Sweden, R&D AstraZeneca, S-431 83 Mo¨lndal, Sweden
ReceiVed September 19, 2008
A set of compounds designed to bind to the S2-S3 pockets of thrombin was prepared. These compounds
included examples with no interactions in the S1 pocket. Proline, a common P2 in many thrombin inhibitors,
was combined with known P3 residues and P1 substituents of varying size and lipophilicity. Binding constants
were determined using surface plasmon resonance (SPR) biosensor technology and were found to be in
good agreement with results from an enzyme assay. A dramatic increase in affinity (100-1000 times) was
seen for compounds incorporating an amino group capable of forming a hydrogen bond with gly216 in the
protein backbone. The ligand efficiency was increased by including substituents that form stronger hydrophobic
interactions with the P1 pocket. The binding mode was confirmed by X-ray analysis, which revealed the
anticipated binding motif that included hydrogen bonds as well as a tightly bound water molecule. A QSAR
model indicated that hydrogen bonding and lipophilicity were important for the prediction of binding constants.
The results described here may have implications for how directed compound libraries for shallow protein
pockets, like S2 and S3 in serine proteases, can be designed.
Introduction
network. A significant increase in ligand efficiency was seen
when lipophilic P1 substituents penetrated deeper into the P1
pocket.
Proteases represent an important target class, with several
pharmacologically relevant targets, but it has been notoriously
difficult to find useful inhibitors.1-4 The inhibitors are often
peptide-like and try to mimic the amino acids that are
characteristic for the substrate. Serine protease inhibitors can
usually be viewed as consisting of three units, P1, P2, and P3.
P1 binds in the S1 pocket, which is usually quite deep, whereas
P2 and P3 bind in the much more shallow S2 and S3 pockets.
Thrombin is a serine protease that is critically involved in
blood coagulation by converting soluble fibrinogen into insoluble
fibrin. The Asp189 in the S1 pocket of thrombin binds strongly
to basic P1 residues, especially arginine and arginine-mimics
such as benzamidine. Figure 1 shows the X-ray structure of the
thrombin inhibitor melagatran5 in complex with thrombin, also
detailing important hydrogen bonds between the inhibitor and
the protein backbone (PDB IK22).
In the present study, we have prepared compounds having a
proline P2 scaffold and P3 residues from known thrombin
inhibitors in combination with P1 substituents of varying size
and lipophilicity. This enabled the importance of P1 for binding
to the enzyme to be investigated and to assess whether
compounds without P1 could be detected as binders. Also, the
importance of P3 substituents with capacity to form hydrogen
bonds to the protein backbone was studied. Using SPR biosensor
measurement of binding affinities, we found that an amino group
in P3 had a remarkably strong effect on potency and an X-ray
structure confirmed its paticipation in a hydrogen bonding
Results and Discussion
The compounds making up the small directed library are
shown in Table 1, along with their corresponding affinity and
IC50 values. Proline is a common P2 residue in many thrombin
inhibitors, and thus the potential to form two important hydrogen
bonds (similar as depicted for melagatran in Figure 1) was
retained. Lipophilic D-amino acids were incorporated in the P3
position. In addition, building blocks without the amino group
were used, and the amino group was also replaced with a
hydroxyl group. Binding constants of up to 1 mM could be
determined using SPR biosensor technology. Thrombin inhibi-
tion was also tested in a traditional enzyme assay using a
maximum compound concentration of 167 µM. For some
compounds, extrapolation of inhibition curves allowed deter-
mination of approximate IC50 values even above this concentra-
tion. For those compounds which yielded IC50 values in the
enzyme assay, it can be seen that there is good correlation
between the two methods (Figure 2).
An overview of some interesting structure-activity relation-
ships is shown in Figure 3.
When R1 is a methyl group, only compound 25 (KD 38 µM,
Table 1)) with a very lipophilic P3 had high enough affinity to
be readily detected. Compounds 21 (KD 550 µM) and 29
(KD 790 µM) are much less potent and the affinities are barely
detectable. Looking at compounds 14 (KD 99 µM), 22
(KD 45 µM), 26 (KD 3.1 µM), and 30 (KD 27 µM), it can be
seen that replacing methyl with n-propyl increases potency by
more than 10 times. This correlates with increased lipophilicity
of the substituent and shows that hydrophobic interactions in
the S1 region of the enzyme are favorable.
The methoxyethyl substituent in compounds 15 (KD 750 µM),
23 (KD 140 µM), 27 (KD 6.8 µM), and 31 (KD 110 µM) is more
polar and results in compounds with intermediate potency. The
thienylmethyl substituent, in compounds 4 (KD 700 µM), 8
* To whom correspondence should be addressed. For T.F.: . phone, +46
31 7065636; fax, +46 31 7763818; E-mail, Tomas.Fex@astrazeneca.com.
For R.I.: phone, +46480446751; fax, +46480447305; E-mail,
Roland.Isaksson@hik.se. For M.N.: phone, +46480446480; fax, +46480447305;
E-mail, Mikael.Nilsson@hik.se. Present address for M.N.: Cambrex Karlskoga
AB, SE-69185 Karlskoga Sweden; phone, +46586783116; fax, +46586783129;
E-mail, mikael.nilsson@cambrex.com.
†
School of Pure and Applied Natural Sciences, University of Kalmar.
GE Healthcare Bio-Sciences AB.
R&D AstraZeneca.
‡
§
10.1021/jm8011849 CCC: $40.75
2009 American Chemical Society
Published on Web 04/16/2009

Compounds Binding to the S2-S3 Pockets of Thrombin
Journal of Medicinal Chemistry, 2009, Vol. 52, No. 9 2709
Figure 1. Melagatran in the active site of thrombin with the H-bonds to the enzyme backbone highlighted (PDB 1K22).
(KD 34 µM), 12 (KD 1000 µM), 16 (KD 23 µM), 20 (KD 250
µM), 24 (KD 2.2 µM), 28 (KD 0.27 µM), and 32 (KD 1.1 µM)
clearly produces the most potent compounds, with activity
observed for all different R2 substituents, underlining that large
contributions of binding energy can be obtained through
interactions in the S1 pocket. Compound 33 (KD 49 µM), with
an isobutyl substituent is less active than the corresponding
n-propyl derivative 30 (KD 27 µM), indicating that the extra
methyl group may result in unfavorable steric interactions.
Compound 34 (KD 9 µM) is the related cyclohexylmethyl
derivative, but here the slight increase in potency does not
parallel the substantial increase in lipophilicity as compared to
the propyl substituent. This is rather surprising because previous
investigations have shown that the cyclohexylamine analogue
is a potent S1 binder.5
The results for different R2 substituents are striking. Those
having an amino group are by far the most potent, with
100-1000 times higher affinities than their nonamino equiva-
lents. This is seen when comparing compounds 24 (KD 2.2 µM),
28 (KD 0.27 µM), and 32 (KD 1.1 µM) with 4 (KD 700 µM), 8
(KD 34 µM), and 12 (KD 1000 µM). The amino group can form
a hydrogen bond to gly216 in the enzyme (see Figure 1), which
contributes very significantly to the total binding energy. A
hydroxyl group seems to be much less efficient in this respect,
however, as seen when comparing 36 (no binding) and 31
(KD 110 µM). Even if compound 16 (KD 23 µM) can not be
compared with 28 (KD 0.27 µM) in a straightforward manner,
there is also an indication here that the amino group is preferred
over the hydroxyl.
increases potency by about 20 times. One of the sulfonyl
oxygens can form a hydrogen bond to gly218, while the phenyl
group can reach toward the so-called S1ꢀ binding site above
the S1 pocket.6-8
As can be seen in Figure 4, those compounds where R2
contains an amino functionality exhibit higher ligand efficiency
than those without. Ligand efficiency (LE) is defined as -∆G/
number of non-hydrogen atoms in the molecule. LE are obtained
from SPR biosensor measurements. This suggests that for
optimizing a hit (i.e., confirmed binder), finding additional
hydrogen bonding possibilities may be very productive in terms
of increasing potency while minimizing the molecular weight.
Another interesting observation is that the ligand efficiency
increases almost linearly with an increase in size of the P1
residue. This finding is somewhat unusual compared to what is
often seen during optimizations, where constant or decreasing
ligand efficiency is more common.9 The S2-S3 pockets are
relatively shallow, and compounds binding in them will have
relatively weak affinity compared to compounds binding in the
S1 pocket. Thus, ligand efficiency will increase when com-
pounds can also utilize the S1 pocket for binding.
For a more detailed analysis, KD values were used for
QSARa calculations. The library compounds were characterized
by chemical descriptors relating to lipophilicity (ClogP, ClogD,
and molecular refractivity MR), size and shape (MW and
Volume), flexibility and rigidity (rotatable bonds and ring count),
polar features (H-bond donors and acceptors and polar surface
a
Abbreviations: SPR, surface plasmon resonance; HTS, high throughput
screening; DMPK, distribution metabolism and pharmacokinetics; QSAR,
quantatative structure activity relationships; OPLS, orthogonal projections
to latent structures.
Compound 35 (KD 35 µM) is a phenylmethylsulfonamide
derivative of 29 (KD 790 µM), and this added substituent

2710 Journal of Medicinal Chemistry, 2009, Vol. 52, No. 9
Table 1. Synthesized Compounds Including KD and IC50 Dataa
Nilsson et al.
a NB ) BIACORE, KD > 1 mM, (NA) ) enzyme assay, IC50 > 400 µM.
(k ) 15) were correlated to the KD values. Compounds
generating IC50 and KD values were included, n ) 24. A QSAR
model was obtained, which by OPLS10 provided a 3 OPLS
component model (one predictive and two orthogonal PLS
components) with 78% explained Y variance (R2 ) 0.78 and
Q2 ) 0.73). The prediction power of the QSAR model, as
estimated by cross-validation, provided an RMSEE of 0.49 pKD
units, and the observed versus predicted values are plotted in
Figure 5. Interpretation of model coefficients indicated that
H-bonding and lipophilicity were of importance for prediction
(Figure 6). First the presence of H-bond donor(s), but not
acceptors, appeared essential, although the pKa range included
(7.2-8.4 by ACD laboratories) revealed only minor difference
between the affinities. Second, lipophilicity appeared important
for the QSAR model, indicated by significant ClogP and MR
QSAR loadings (i.e., OPLS p1p, significance tested by jack-
knifing). The LogD descriptors were close to nonsignificance,
which in combination with the H-bond donor information
supports the conclusion that the presence of a H-bond donor,
rather than its basicity, generated the binding energy. The
contribution to affinity of the R1 substituents fits well with the
Figure 2. The correlation between IC50 and KD.
area), and pKa (indicator of base and, if available, the actual
pKa was calculated). The descriptors that were generated

Compounds Binding to the S2-S3 Pockets of Thrombin
Journal of Medicinal Chemistry, 2009, Vol. 52, No. 9 2711
Figure 3. SAR for selected compounds showing the effect on affinities (KD) of variation in R1 (methyl, n-propyl, methoxyethyl, thienylmethyl)
for three different R3 substituents.
Figure 4. Ligand efficiency (LE) as a function of pKD for selected series. LE ) (-∆G/N); N is the number of non-hydrogen atoms, and
∆G ) - RT ln(1/KD).
QSAR model comprising positive and significant coefficients
for ClogP and MR.
obtained was in good agreement with the anticipated binding
mode. There are three strong hydrogen bonds (all around 2.9
Å) formed between the inhibitor and thrombin. These hydrogen
bonds contribute significantly to the observed binding affinity.
In addition, the propyl tail of the inhibitor fits into the entrance
of S1 and the diphenyl group fills into the S3 pocket. Together
with the pyrrolidine group, which occupies the hydrophobic
space formed by Tyr83/Trp86/Leu132 in S2, the shape of the
inhibitor matches/fills the binding pockets of the protein very
well. Furthermore, there are a number of water molecules
surrounding the inhibitor, some of which bridge the interaction
between the inhibitor and the protein. In particular, Wat154
seems to stabilize the inhibitor conformation by having hydrogen
bonds to both the amino group and the amide carbonyl group.
This may contribute to the overall high affinity of compounds
with a terminal amino group.
The quite dramatic effect observed for the amino group may
have some important implications. Molecules that can form three
interactions have three points of attachment to the enzyme,
which may result in more rigid and stable complexes. In light
of the fact that the hydrogen bond is formed between the amino
group and a carbonyl group on the surface of the enzyme, the
increased binding is even more intriguing. Thus, when a binder
has been identified, one option for improving potency should
be to look for potential additional hydrogen bonding interactions.
If found, these can provide dramatic increases in potency. This
is somewhat contradictory to the general notion that increased
potency during lead optimization is achieved by increasing bulk
and lipophilicity.
Compound 26, which showed quite good potency and a P1
that could occupy only part of the S1 pocket, was selected for
crystallization with thrombin (Figures 7-8). The X-ray structure
The results from this study show that synthesis of P2-P3
libraries may be a useful approach to find novel serine protease

2712 Journal of Medicinal Chemistry, 2009, Vol. 52, No. 9
Nilsson et al.
dimethyl sulfoxide (Riedel-de Hae¨n, Seelze, Germany) was used
as running buffer.
The proteins where amine coupled (EDC/NHS, amine coupling
kit, GE Healthcare Bio-Sciences AB, Uppsala, Sweden) to the
sensor chip using a 10 mM NaAc pH 5.0 as a coupling buffer (GE
Healthcare Bio-Sciences AB, Uppsala, Sweden). Typical im-
mobilization levels were 5000-6000 RU (resonance units) for wt-
thrombin (human R-thrombin, Haematologic Technologies Inc., VT)
and blocked thrombin (human R-thrombin-DFP, Haematologic
Technologies Inc., VT). Routinely, the following concentrations
(µg/mL) and injection times (min) were used for immobilizations:
Thrombin, 20 µg/mL, 5 min, and Thrombin-DFP, 20 µg/mL, 7 min.
The flow rate was 30 µL/min. Test substances and positive controls
were injected in a cycle with a contact time of 30 s and dissociation
time of 30 s, followed by an extra wash of the flow system with
50% DMSO. The effect of bulk refractive index variations due to
potential slight variations in DMSO content of the samples was
compensated using a standard calibration routine implemented in
the A100 Evaluation software. The calibration was performed in
the 4.5-5.8% DMSO range.
The first affinity screen was run using a concentration range of
3-200 µM. Affinity calculations for the medium affinity com-
pounds (KD 6-100 µM) were based on this concentration range.
High affinity compounds (0.3-4 µM) were analyzed using a
concentration range of 1 nM to 10 µM. Nonbinders and weak
binders were run using a concentration range of 30-500 µM to
7-1000 µM. Affinity calculations for the weak binders were based
on these higher concentration ranges. The compounds were injected
in a series of increasing concentrations. The response used for
affinity calculations was the differential steady-state binding signal
between wt-thrombin and the blocked thrombin, i.e., the wt-
thrombin signal minus that of the blocked thrombin. For equilibrium
binding responses of compounds binding to thrombin (wt minus
blocked), the affinity (KD) was calculated by fitting to the equation
of a single-site binding model (R ) (conc × Rmax)/(conc + KD). A
minimum of three replicate measurements were made for each
compound. The average relative standard deviation of the affinity
estimates was 20.7%.
Figure 5. Observed pKD values for compounds (numbering according
to Table 1) binding to thrombin assessed by SPR plotted against the
predicted pKD. The predictions were generated by cross validation
(Q2 ) 0.71), with an RMSEE ) 0.49, using an OPLS model explaining
78% of pKD according to R2.
Enzyme Assay. The thrombin inhibitor potency was measured
with a chromogenic substrate method in a robotic microplate
processor, using 96-well, half-volume microtiter plates.11 The linear
absorbance increase values were determined by measurements at
405 nm (37 °C) during 40 min with Melagatran as control substance.
The IC50 values were calculated by fitting the data to a three-
parameter equation by Microsoft XLfit. For some compounds with
IC50 > 167 µM, an approximate IC50 value was obtained by
extrapolation of inhibition curves. For melagatran, the IC50 ) 0.008
µM.
Figure 6. QSAR coefficients (i.e., P1p loadings, confidence intervals
by jack-knifing), using OPLS for regression with compounds binding
to thrombin (pKD) from SPR assessments as Y (R2 ) 0.78; Q2 ) 0.73;
RMSEE ) 0.49).
Experimental Procedure Crystallization and X-Ray Struc-
ture Determination. Crystallization and soaking of compound:
human R-thrombin (Factor IIa) was purchased from Enzyme
Research Laboratories. Preformed thrombin-hirudin complex
(stored at 4 °C) was used for crystallization. Apo-crystals were
grown using hanging-drop method with microseeding. The drop
was made by mixing 1.5 µL of the thrombin-hirudin complex and
1.5 µL of reservior solution containing: 0.05 M sodium phosphate
buffer pH 7.3, 28% PEG8000. The soaking was carried out by
adding powdered compound directly to the drop and leaving it at
room temperature overnight.
Data Collection and Structure Determination. The diffraction
data were collected from frozen crystals on a MarCCD detector
mounted on a micro focus rotating anode generator FR-E Super-
Bright from Rigaku. The crystal-to-detector distance was set to be
90 mm, and 360 images were collected with oscillation of 1°. The
data were processed with MOSFLM12 and programs in the CCP4
suite.13 Refinement and model rebuilding was carried out using
programs REFMAC514 and COOT.15 The structure has been
deposited with PDB, ID code 3DA9.
inhibitors. The importance of hydrogen bonding interactions is
underlined, and compounds should be capable of accepting and/
or donating hydrogen bonds. Once a binder has been identified,
a further increase in potency can be achieved by a combination
of adding new hydrogen bonding groups and changes that
increase hydrophobicity. Hydrogen bonds to the protein surface
can greatly enhance binding.
Protein-protein interactions are often made up of interactions
between protein surfaces with relatively shallow pockets,
perhaps resembling the shallow S2 and S3 pockets in proteases.
Thus, the results presented here may have implications for how
we try to find molecules that can interact with proteases and
hence to develop effective drugs against these very challenging
targets.
Experimental Section
SPR Methodology. All experiments where performed with a
Biacore A100 instrument using Series S sensor chip CM5 (GE
Healthcare Bio-Sciences AB, Uppsala, Sweden), 10 mM phosphate
buffer with 137 mM NaCl, 2.7 mM KCl, 0.05% surfactant P20
(GE Healthcare Bio-Sciences AB, Uppsala, Sweden), and 5%
General Procedures. All chemicals were used as received, and
solvents for dry conditions were kept over molecular sieves. All
1H NMR spectra were recorded using a Varian Unity 500 at 500

Compounds Binding to the S2-S3 Pockets of Thrombin
Journal of Medicinal Chemistry, 2009, Vol. 52, No. 9 2713
Figure 7. X-ray structure of compound 26 in the active site of thrombin with the H bonds to the enzyme backbone and wat154 highlighted (PDB,
ID code 3DA9).
MHz. Flash column chromatography was carried out using Silica
Gel 60 (0.063-0.200 mm). Semipreparative HPLC was performed
using a Bishoff system, consisting of 2250 HPLC compact pumps
and Lambda 1010 detector, equipped with a Sperisorb ODS2 (C18)
10 µ 250 mm × 10 mm or a Kromasil C18 10 µ 250 mm × 10
mm. Purification was performed either using different gradients of
acetonitrile/H2O, or isocratically. Normal phase semipreparative
HPLC was performed on a Chromtech CT-sil 5 µ 250 mm × 10
mm isocratically using different proportions of heptane and
isopropyl alcohol. Purity was checked on a HP 1090 using a
Kromasil C18 5 µ 250 mm × 4.6 mm with a gradient ranging from
10 to 70% acetonitrile in 30 min, thereafter isocratically at 70%
for an additional 15 min. Purity always exceeded 95% when
monitoring the eluent with UV at 214 nm or 254 nm.
High resolution mass spectra was obtained on a Waters LCT
Premier using positive electrospray for ion generation. Leucine
Enkephaline C28H37N5O7 (m/z 556.2771) was used as reference.
Procedures and Data for 1-36. General procedures for A-D:
1 g of Boc-L-proline hydroxysuccinimide ester was dissolved in
30 mL dry THF. Three eqviv of the amine was added, and the
mixture was stirred during 3 days at room temperature. The
insoluble part was then filtered off, washed with THF, and
the filtrate combined with the “wash solution” was concentrated in
vacuo and subjected to flash chromatography using EtOAc as eluent.
The product obtained was subsequently deprotected with freshly
prepared 10 mL of AcCl/EtOH (1:2) for 1 h, evaporated into
dryness, and left in a desiccator overnight. To afford the free base,
the HCl salt was partitioned between 30 mL of 5 M NaOH and 3
× 30 mL of EtOAc or diethylether and dried over Na2SO4.
Pyrrolidine-2-carboxylic Acid Methylamide (A). Yield: 72%;
1H NMR: 7.60 (s, 1H), 3.75 (dd, 1H, J ) 7.5 Hz, J ) 13.0 Hz),
3.02 (dd, 1H, J ) 6.7 Hz, J ) 15.5 Hz), 2.91 (m, 1H), 2.82 (d, 3H,
J ) 4.8 Hz), 2.15 (dt, 1H, J ) 7.4 Hz, J ) 15.3 Hz), 1.93 (td, 1H,
J ) 5.5 Hz, J ) 12.4 Hz), 1.80 (br s, 1H), 1.72 (tt, 1H, J ) 6.2
Hz, J ) 12.3 Hz).
in 5 mL ACN, were added. The reaction was allowed to proceed
overnight. Work up was performed by partition between the ACN
solution and 10 mL of saturated NaCl. The water solution was then
extracted 3 × 10 mL EtOAc, the combined organic extracts was
washed with 30 mL of 1 M HCl and 30 mL of saturated NaHCO3
and was finally dried over Na2SO4. Purification performed on
semipreparative HPLC using 60 H2O/40 ACN on HPLC afforded
1
1 as clear oil in 78% yield. H NMR: 7.25 (m, 2H), 7.16 (d, 2H,
J ) 8.3 Hz), 6.93 (s, 1H), 4.55 (d, 1H, J ) 8.0 Hz), 3.45 (m, 1H),
3.31 (m, 1H), 2.96 (m, 2H), 2.76 (d, 3H, J ) 4.8 Hz), 2.60 (m,
2H), 2.44 (m, 1H), 2.09 (m, 1H), 1.95 (m, 1H), 1.76 (m, 1H).
HRMS found 295.1220; calcd 295.1213.
1-(3,3-Diphenyl-propionyl)-pyrrolidine-2-carboxylic Acid Me-
thylamide (5). As a general example: 72 mg of 3,3-diphenyl-
propionic acid (1 equiv) were dissolved in 20 mL of DMF and
121 mg of HBTU (1 equiv) were added. After 3 min of activation,
115 µL (2 equiv) of DIPEA and 41 mg of A, dissolved in 5 mL of
DMF, were added. The reaction was allowed to proceed overnight.
The mixture was subsequently concentrated in vacuo, and purifica-
tion was performed on semipreparative HPLC using 65 H2O/35
1
ACN on HPLC, affording white crystals in 70% yield. H NMR:
7.25 (m, 10H), 6.55 (s, 1H), 4.68 (t, 1H, J ) 7.8 Hz), 4.53 (d, 1H,
J ) 7.8 Hz), 3.44 (dt, 1H, J ) 2.5 Hz, J ) 9.4 Hz), 3.32 (dt, 1H,
J ) 7.3 Hz, J ) 9.5 Hz), 3.06 (m, 2H), 2.60 (d, 3H, J ) 4.8 Hz),
2.39 (m, 1H), 1.97 (m, 1H), 1.86 (m, 1H), 1.70 (tt, 1H, J ) 7.3
Hz, J ) 11.9 Hz). HRMS found 337.1932; calcd 337.1916.
1-[2-Amino-3-(4-chloro-phenyl)-propionyl]-pyrrolidine-2-car-
boxylic Acid Methylamide (21). First, 86 mg of D-2-Boc-3-(4-
chloro-phenyl)-propionic acid were dissolved in 5 mL of ACN.
Then 80 mg of HBTU were used to activate the acid for 3 min.
Thereafter, 100 µL of DIPEA and 27 mg of A dissolved in 5 mL
of ACN were added. The reaction was allowed to proceed overnight.
Work up was performed by partition between the ACN solution
and 10 mL of saturated NaCl. The water solution was then extracted
with 3 × 10 mL EtOAc, the combined organic extracts were
washed with 30 mL of 1 M HCl and 30 mL of saturated NaHCO3,
and were finally dried over Na2SO4. Purification was performed
using gradient elution, starting with 70% H2O to 33% H2O in 34
min on semipreparative HPLC. The product obtained was subse-
quently deprotected with freshly prepared 2 mL of AcCl/EtOH (1:
1-[3-(4-Chloro-phenyl)-propionyl]-pyrrolidine-2-carboxylic Acid
Methylamide (1). As a general example: 46 mg of 3-(4-Chloro-
phenyl)-propionic acid (1 equiv) were dissolved in 5 mL of ACN
and 95 mg of HBTU (1 equiv) were added. After 3 min of
activation, 100 µL (2 equiv) of DIPEA and 32 mg of A, dissolved

2714 Journal of Medicinal Chemistry, 2009, Vol. 52, No. 9
Nilsson et al.
Figure 8. Stereo pictures of (a) melagatran (pdb:1Kk22) and (b) compound 26 (pdb: 3DA9) binding to active site of thrombin.
2) for 1 h, evaporated into dryness, and left in a desiccator overnight.
To afford the free base, the HCl salt was partitioned between 5
mL 5 M NaOH and 3 × 10 mL EtOAc or diethylether and dried
over Na2SO4. Yield 60% as transparent semi solid. 1H NMR: 7.29
(d, 2H) 7.14 (d, 2H, J ) 8.3 Hz) 6.95 (s, 1H) 4.46 (dd, 1H, J )
1.5 Hz, J ) 8.2 Hz) 3.75 (t, 1H, J ) 7.2 Hz) 3.51 (dt, 1H, J ) 2.8
Hz, J ) 9.0 Hz) 2.92 (m, 2H) 2.79 (m, 4H) 2.36 (m, 1H) 2.00 (m,
1H) 1.73 (m, 1H) 1.60 (s, 1H). HRMS found 310.1325; calcd
310.1322.
1-(2-Cyclohexyl-2-phenylmethanesulfonylamino-acetyl)-pyr-
rolidine-2-carboxylic Acid Methylamide (35). First, 34 mg of 29
were dissolved in 5 mL of DCM and cooled on ice. Thereafter, 3
equiv of triethylamine (56 µL) and 3 equiv of phenylmethanesulfo-
nylchloride (77 mg) were added and the mixture was allowed to
stir overnight. The mixture was concentrated in vacuo and purified
using 60 H2O/40 ACN on HPLC, yielding 69% as transparent
semisolid. 1H NMR: 7.45 (m, 2H), 7.39 (m, 2H), 6.81 (s, 1H), 4.90
(d, 1H, J ) 9.7 Hz), 4.52 (dd, 1H, J ) 1.9 Hz, J ) 7.8 Hz), 4.29
(q, 2H, J ) 13.9 Hz), 3.60 (dd, 1H, J ) 7.9 Hz, J ) 9.1 Hz), 3.49
(m, 1H), 3.31 (td, 1H, J ) 7.9 Hz, J ) 15.9 Hz), 2.78 (d, 3H, J )
4.8 Hz), 2.34 (m, 1H), 1.98 (m, 3H), 1.87 (d, 1H, J ) 12.9 Hz),
1.72 (m, 3H), 1.52 (m, 2H), 1.13 (m, 4H), 0.90 (dq, 1H, J ) 3.3
Hz, J ) 12.5 Hz). HRMS found 422.2117; calculated 422.211.
1-(2-Amino-2-cyclohexyl-acetyl)-pyrrolidine-2-carboxylic acid
isobutyl-amide (33). Pyrrolidine-2-carboxylic acid isopropylamide
was synthesized according to A using Boc-L-proline hydroxysuc-
cinimde and isobutylamine. Synthesis and work up of the final
product was performed according to 21 using the obtained pyrro-
lidine-2-carboxylic acid isopropylamide and Boc-D-cyclohexylg-
lycine purified using 45 H2O/55 ACN on HPLC, yielding 57% as
transparent semisolid. 1H-NMR: 7.20 (s, 1H), 4.59 (d, 1H, J ) 7.5
Hz), 3.64 (dt, 1H, J ) 2.7 Hz, J ) 9.4 Hz), 3.49 (dt, 1H, J ) 7.5
Hz, J ) 9.5 Hz), 3.28 (d, 1H, J ) 7.0 Hz), 3.05 (m, 2H), 2.46 (m,
1H), 2.12 (m, 1H), 1.96 (m, 2H), 1.78 (m, 5H), 1.52 (m, 2H), 1.26
(m, 2H), 1.10 (m, 2H), 0.99 (dq, 1H, J ) 2.9 Hz, J ) 12.3 Hz),
0.89 (d, 6H, J ) 6.7 Hz). HRMS found 310.2500; calcd: 310.2495.
Acknowledgment. We thank Maria Ericsson at AstraZeneca
R&D Mo¨lndal, Sweden, for providing us with results from the
enzyme assay, Gary Franklin regarding improvement of the
language, and finally, the University of Kalmar for funding.
Supporting Information Available: Synthesis and HRMS and
H NMR identification of all other compounds. Chromatographic

Compounds Binding to the S2-S3 Pockets of Thrombin
Journal of Medicinal Chemistry, 2009, Vol. 52, No. 9 2715
Chang, C. Y.; Chong, S.; Wang-Iverson, D. B.; Roberts, D. G. M.;
Seiler, S. M.; Schumacher, W. A.; Ogletree, M. L. Molecular Design
and Structure-Activity Relationships Leading to the Potent, Selective,
and Orally Active Thrombin Active Site Inhibitor BMS-189664. Biorg.
Med. Chem. Lett. 2002, 12, 45–49.
analyses of compound purity. Additional information on X-ray
structure determination. This material is available free of charge
via the Internet at http://pubs.acs.org.
References
(9) Hajduk, P. J.; Greer, J. A decade of fragment-based drug design:
strategic advances and lessons learned. Nat. ReV. Drug DiscoVery 2007,
6, 211–219.
(1) Leung, D.; Abbenante, G.; Fairlie, D. P. Protease inhibitors: current
status and future prospects. J. Med. Chem. 2000, 43, 305–341.
(2) Ilies, M. A.; Scozzafava, A.; Supuran, C. T. Therapeutic applications
of serine protease inhibitors. Expert Opin. Ther. Pat. 2002, 12, 1181–
1214.
(3) Maryanoff, B. E. Inhibitors of serine proteases as potential therapeutic
agents: the road from thrombin to tryptase to cathepsin G. J. Med.
Chem. 2004, 47, 1–19.
(4) Abbenante, G.; Fairlie, D. P. Protease inhibitors in the clinic. Med.
Chem. 2005, 1, 71–104.
(5) Sorbera, L. A.; Bayes, M. C. J.; Silvestre, J. Melagatran and
Ximelagatran. Drugs Future 2001, 26, 1155–1170.
(6) Lumma, W. C., Jr.; Witherup, K. M.; Tucker, T. J.; Brady, S. F.; Sisko,
J. T.; Naylor- Olsen, A. M.; Lewis, S. D.; Lucas, B. J.; Vacca, J. P.
Design of novel, potent, noncovalent inhibitors of thrombin with
nonbasic P-1 substructures: Rapid structure-activity studies by solid
phase synthesis. J. Med. Chem. 1998, 41, 1011–1013.
(7) Huber, W.; Mueller, F. Biomolecular interaction analysis in drug
discovery using surface plasmon resonance technology. Curr. Pharm.
Des. 2006, 23, 3999–4021.
(10) Trygg, J.; Wold, S. Orthogonal projections to latent structures (OPLS).
J. Chemom. 2002, 16, 283–293.
(11) Nilsson, J. W.; Kvarnstrom, I. D.; Musil, D.; Nilsson, I.; Samulesson,
B. Synthesis and SAR of thrombin inhibitors incorporating a novel
4-amino-morpholinone scaffold: analysis of X-ray crystal structure of
enzyme-inhibitor complex. J. Med. Chem. 2003, 46, 3985.
(12) Leslie, A. G. Integration of macromolecular diffraction data. Acta
Crystallogr., Sect. D: Biol. Crystallogr. 1999, 55, 1696–1702.
(13) Collaborative Computational Project no. 4. The CCP4 Suite: Programs
for Protein Crystallography. Acta Crystallogr., Sect. D: Biol. Crys-
tallogr. 1994, 50, 760-763.
(14) Murshudov, G. N.; Vagin, A. A.; Lebedev, A.; Wilson, K. S.; Dodson,
E. J. Efficient anisotropic refinement of macromolecular structures
using FFT. Acta Crystallogr., Sect. D: Biol. Crystallogr. 1999, 55,
247–255.
(15) Emsley, P.; Cowtan, K. Coot: model-building tools for molecular
graphics. Acta Crystallogr., Sect. D: Biol. Crystallogr. 2004, 60,
2126–2132.
(8) Jagabandhu, D.; Kimball, S. D.; Hall, S. E.; Han, W.-C.; Iwanowics,
E.; Lin, J.; Moquin, R. V.; Reid, J. A.; Sack, J. S.; Malley, M. F.;
JM8011849