Enantiopure five-membered cyclicdiamine derivatives as potent and selective inhibitors of factor Xa. Improving in vitro metabolic stability via core modifications

Article abstract of DOI:10.1016/j.bmcl.2007.07.020

We previously reported a series of enantiopure cis-(1R,2S)-cyclopentyldiamine derivatives as potent and selective inhibitors of Factor Xa (FXa). Herein, we describe our approach to improve the metabolic stability of this series via core modifications. Multiple resulting series of compounds demonstrated similarly high FXa potency and improved metabolic stability in human liver microsomes compared with the cyclopentyldiamide 1. (3R,4S)-Pyrrolidinyldiamide 31 was the best overall compound with human FXa K_i of 0.50 nM, PT EC_2x of 2.1 μM in human plasma, bioavailability of 25% and t_1/2of 2.7 h in dogs. Further biochemical characterization of compound 31 is also presented.

Full text of DOI:10.1016/j.bmcl.2007.07.020

Bioorganic & Medicinal Chemistry Letters 17 (2007) 5041–5048
Enantiopure five-membered cyclicdiamine derivatives as potent
and selective inhibitors of factor Xa. Improving in vitro
metabolic stability via core modifications
Jennifer X. Qiao,* Tammy C. Wang, Gren Z. Wang, Daniel L. Cheney, Kan He,
Alan R. Rendina, Baomin Xin, Joseph M. Luettgen, Robert M. Knabb,
Ruth R. Wexler and Patrick Y. S. Lam
Bristol-Myers Squibb Company, Research and Development, PO Box 5400, Princeton, NJ 08643-5400, USA
Received 3 June 2007; revised 4 July 2007; accepted 6 July 2007
Available online 13 July 2007
Abstract—We previously reported a series of enantiopure cis-(1R,2S)-cyclopentyldiamine derivatives as potent and selective inhib-
itors of Factor Xa (FXa). Herein, we describe our approach to improve the metabolic stability of this series via core modifications.
Multiple resulting series of compounds demonstrated similarly high FXa potency and improved metabolic stability in human liver
microsomes compared with the cyclopentyldiamide 1. (3R,4S)-Pyrrolidinyldiamide 31 was the best overall compound with human
FXa Ki of 0.50 nM, PT EC2x of 2.1 lM in human plasma, bioavailability of 25% and t1/2 of 2.7 h in dogs. Further biochemical char-
acterization of compound 31 is also presented.
ꢀ 2007 Elsevier Ltd. All rights reserved.
Factor Xa (FXa), a serine protease located at the conver-
gent point of the intrinsic and extrinsic pathways, binds
phospholipids, cofactor VIIIa, and calcium ions to form
a prothrombin complex, which is responsible for catalyz-
ing the conversion of prothrombin to thrombin. A prom-
ising strategy to develop novel anticoagulants is to
inhibit thrombin formation via the inhibition of FXa.1–3
Several small-molecule, orally active FXa inhibitors,
including pyrazole-based razaxaban4 and apixaban,5 have
entered clinical development for the treatment and pre-
vention of thrombotic diseases, and apixaban is undergo-
ing evaluation in phase 3 studies in various indications.
P1 and a phenylpyridone P4 group7 remained after
10-min incubation in human liver microsomes (HLM),
and 1 had a microsomal intrinsic clearance rate of
0.063 nmol/min/mg. Cyclopentyldiamine derivatives
bearing several other P1 and P4 groups also displayed
metabolic instability in HLM. Metabolic ID studies on
compound 1 showed that 12% and 8% of 1 were mono-
hydroxylated at the cyclopentyl ring and the phenylpyri-
done P4 moiety, respectively, after 1 h incubation in
HLM in the presence of NADPH.
Several compounds in the cycloalkyldiamine series6 were
tested stable to amide hydrolysis in dog plasma in vitro
for up to 4 h, indicated by the sustained anti-FXa activ-
ity and HPLC peak area. In addition, no metabolites
related to amide cleavage were observed during the met-
abolic stability studies of compound 1.
In a previous communication,6 we disclosed a series of
enantiopure (1R,2S)-cyclopentyldiamine derivatives as
potent and selective FXa inhibitors structurally different
from the pyrazole-based scaffolds. However, low meta-
bolic stability in liver microsomal incubation studies
was an issue common to this series of compounds. For
instance, 79% of compound 1 bearing a chlorothiophene
The interactions of a ligand in the FXa S1 and S4
subsites are essential to the FXa binding affinity of the
ligand. Previous SAR studies6 identified 5-chlorothioph-
ene and 3-chloroindole as optimal P1 groups for FXa po-
tency, and phenylpyridone group as the most potent
neutral P4 group in the cyclopentyldiamine derivatives.
Driven by our continuous interest in structurally diverse
back-up series, we decided to further modify the
Keywords: Factor Xa inhibitors; Enantiopure cyclicdiamine deriva-
tives; SAR; Human liver microsomal stability.
*
Corresponding author. Tel.: +1 609 818 5298; fax: +1 609 818
6810; e-mail: jennifer.qiao@bms.com
0960-894X/$ - see front matter ꢀ 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2007.07.020

5042
J. X. Qiao et al. / Bioorg. Med. Chem. Lett. 17 (2007) 5041–5048
Table 1. SAR of enantiopure cis-diaminoindane derivatives
4
M
1
core modification to
improve HLM stability
O
O
O
O
M
NH HN
NH HN
O
O
NH HN
S
S
O
O
P1
1
P1
P1
N
N
P4
Cl
Cl
P4
O
FXa
K
i = 0.43 nM
N
PT EC2x = 1.7 μM
M = five-membered cores
HLM rate = 0.063 nmol/min/mg
Compound
M
P1
FXa
PT EC2x
Figure 1. Overall strategy.
Ki (nM) (lM)
cyclopentyl core of the lead 1 (Fig. 1) to improve in vitro
metabolic stability so as to ultimately improve the phar-
macokinetic properties of this series of compounds.
5
1
3-Cl-Indole-6-yl 0.70
5-Cl-Thiophene-2-yl 0.43
nd
1.7
Figure 2 shows an overlay of the X-ray crystal structures
of compound 16 and the pyrazole-based compound 4
(apixaban)5 in the FXa binding site. Significant differ-
ences in binding are confined to the core region. In the
1-FXa complex, the amide NH forms a hydrogen bond
to the carbonyl oxygen of G216, whereas 4 is hydrogen
bonded to G216 NH via the carbonyl oxygen of the pyr-
azolo pyridinone core. In the 4-FXa complex, the C3
substituent on the pyrazole ring (CONH2 in 4; a variety
of groups such as CF3 are also tolerated) interacts with
the S1b pocket. The overlay shows that the 4 and the 5
positions of the cyclopentyl ring in 1 are close to the C3
substituent of 4, suggesting that modifications on the
periphery of the cyclopentyl ring in 1 could improve
the properties of the molecules (e.g., to improve meta-
bolic stability by blocking the oxidative metabolism on
the cyclopentyl ring) while maintaining or improving
potency. Such modifications include fusing a ring to,
or adding a substituent at the 4 position of, or inserting
a heteroatom at the 4 position of the cyclopentyl ring.
6
7
3-Cl-Indole-6-yl 0.21
5-Cl-Thiophene-2-yl 230
14
nd
1 2
8
9
3-Cl-Indole-6-yl
5-Cl-Thiophene-2-yl 27
0.84
5.5
nd
10
11
3-Cl-Indole-6-yl 7.9
5-Cl-Thiophene-2-yl 10870
nd
nd
2 1
12
13
3-Cl-Indole-6-yl 1.0
5-Cl-Thiophene-2-yl 0.27
nd
3.3
Human purified enzymes were used. Values are averages from multiple
determinations (n P 2). Ki values and PT EC2x (the concentration of
the inhibitor that doubles the prothrombin time from the control in the
prothrombin time assay) values were measured as described in Ref. 9.
Same for all the tables in this publication.
Indeed, fusing the cyclopentyl core with a phenyl ring
led to diaminoindane analogs. Table 1 shows FXa activ-
ities of enantiopure cis diaminoindanes bearing either a
3-chloroindole or a 5-chlorothiophene P1 group. The
two cis 1,2-diaminoindane enantiomers, compound 6
with a cis-up ((1S,2R)-2,3-dihydro-1H-inden-1-yl) con-
figuration and compound 8 with a cis-down ((1R,2S)-
2,3-dihydro-1H-inden-1-yl) configuration bearing
a
3-chloroindole group, were potent with FXa Ki less than
1 nM. On the other hand, the corresponding two cis 1,
2-diaminoindane enantiomers 7 and 9 bearing a 5-chlo-
rothiophene P1 group were much less potent. Com-
pounds 11 and 13 with a 2,1-diaminoindane core and
a chlorothiophene P1 had drastically different FXa po-
tency. Compound 11 with a cis-up ((1R,2S)-2,3-dihy-
dro-1H-inden-2-yl) configuration was weakly active;
while compound 13 with a cis-down configuration
((1S,2R)-2,3-dihydro-1H-inden-2-yl) was the most po-
tent (FXa Ki = 0.27 nM) in the diaminoindane series,
having potency similar to the parent cyclopentyl analog
1. Compound 13, however, was extensively metabolized
in HLM (32% remaining after 10-min incubation) pre-
sumably because of the oxidation of the phenyl ring of
the indane core.
Docking studies10 of the 1,2-diaminoindane (com-
pounds 6 and 8) and the 2,1-diaminoindane (compounds
11 and 13) analogs were conducted to gain structural in-
sights into the observed SAR. Figure 3 shows the over-
lay of the top-scoring binding models of compounds 6
Figure 2. Overlay of X-ray structures of FXa bound cis-(1R,2S)-
cyclopentyldiamide1and pyrazole bicyclic-based apixaban (4). Created
using PyMol.8

J. X. Qiao et al. / Bioorg. Med. Chem. Lett. 17 (2007) 5041–5048
5043
and 8, and the X-ray structure of 1 in the active site of
FXa. The corresponding P1 and P4 regions of 6 and 8
overlap very well. However, the two indane cores occu-
py different subsites of the FXa enzyme: the indane ring
in 6 with a cis-up configuration sits above the 42–58
disulfide bridge, near the S10 pocket; while the indane
ring in 8 with a cis-down configuration locates in the
vicinity of the 191–220 disulfide in the S1b region.
Compared with compound 1, the sterically demanding
1,2-diaminoindane cores of 6 and 8 tend to pull the mol-
ecules out farther away from the active site. Thus, larger
P1 groups, 3-chloroindole in the case of 6 and 8, are
needed to form necessary interactions with the S1 pock-
et to maintain FXa potency. This also explains the de-
creased FXa potency observed in 7 and 9 bearing a
smaller 5-chlorothiophene P1 group. Interactions in
the FXa S1b and the entrance of the S10 region may also
affect FXa potency of diaminoindane compounds, as
suggested by the potency difference between 7 and 9.
Figure 4 shows the proposed binding model of the 2,1-
diaminoindane 13, wherein the inversion of the cyclo-
pentyl ring results in a reversal of the axial/equatorial
orientations of the P1 and P4 groups with respect to that
observed in the crystal structure of 1-FXa. This model
suggests that the core phenyl is projected into solvent
well away from the protein surface, allowing the 5-chlo-
rothiophene to bind deeply in the S1 pocket. Not
surprisingly, a compelling binding model in FXa of
the weakly active 2,1-diaminoindane 11 with optimal
interactions could not be generated.11
Figure 4. Binding model of the 2,1-diaminoindane 13 in the active site
of FXa. Flipping of the cyclopentyl ring results in the reversal of the
axial/equatorial orientations of the P1 and P4 groups with respect to
that observed in the crystal structure of 1 in FXa. Created using
PyMol.8
tent analog in the 4-substituted cyclopentyl series stud-
ied, had an improved metabolic stability (100%
remaining after 10 min) compared with
remaining).
1 (79%
Using the 3-chloroindole and the 5-chlorothiophene P1
groups, we investigated the FXa inhibitory activities of
five-membered ring analogs (Table 3). The tetrahydro-
furanyldiamine derivative 23 and the substituted pyr-
rolidinyldiamine derivatives 29 and 31 showed good
FXa potency and anticoagulant activity. Compounds
with a 5-chlorothiophene P1 group were consistently
more potent than those with a 3-chloroindole group
across the five-membered cores studied (4-substituted
cyclopentyl, tetrahydrofuranyl, pyrrolidinyl core series).
This P1 preference was also observed previously be-
tween the cyclopentyldiamine and cyclohexyldiamine
Adding a COOMe group on the 4 position of the cyclo-
pentyl core in 1 generated compound 14 with maintained
FXa potency and anticoagulant activity (Table 2).
Reduction of the methyl ester 14 to methyl alcohol 15
led to the same affinity as the unsubstituted cyclopen-
tane 1. The acid 16 and the amides 17–20 had similar
FXa potency and were about three to four times less po-
tent than the methyl ester 14. They were also less potent
in anticoagulant activity. Compound 15, the most po-
Table 2. SAR of 4-substituted cyclopentyldiamine derivatives
R
O
O
N
O
H
NH
N
S
Cl
Compound
R
FXa
Ki (nM)
PT EC2x
(lM)
1
14
15
16
17
18
19
20
H
0.43
0.53
0.56
1.9
1.7
1.8
1.5
4.7
6.5
7.1
4.9
4.7
COOMe
CH2OH
COOH
CONMe2
2.1
1.6
Figure 3. Overlay of binding models of two 1,2-diaminoindane
enantiomers 6 (white) and 8 (pink) with the X-ray structure of
cyclopentyldiamide 1 (blue) in the active site of FXa. Created using
PyMol.8
CONH-Cyclopropyl
CO-N-Morpholinyl
CONHCH2CH2OMe
1.6
2.9

5044
J. X. Qiao et al. / Bioorg. Med. Chem. Lett. 17 (2007) 5041–5048
Table 3. SAR of enantiopure 4-substituted five-membered cyclicdiamine derivatives
X
O
O
NH HN
P1
O
N
Compound
X
P1
FXa Ki (nM)
PT EC2x (lM)
1
5
CH2
CH2
5-Cl-Thiophene-2-yl
3-Cl-Indole-6-yl
5-Cl-Thiophene
0.43
1.0
1.7
5.5
1.8
nd
21
22
23
24
25
26
27
28
29
30
31
32
CH-COOMe
CH-COOMe
O
0.39
3.0
3-Cl-Indole-6-yl
5-Cl-Thiophene-2-yl
3-Cl-Indole-6-yl
2.6
2.6
3.2
7.1
nd
O
NH
5-Cl-Thiophene-2-yl
3-Cl-Indole-6-yl
9.5
NH
18
nd
nd
N-Fmoc
N-Fmoc
N-COMe
N-COMe
N-COOEt
N-COOEt
5-Cl-Thiophene-2-yl
3-Cl-Indole-6-yl
4.5
163
0.58
3.9
nd
5-Cl-Thiophene-2-yl
3-Cl-Indole-6-yl
0.90
nd
5-Cl-Thiophene-2-yl
3-Cl-Indole-6-yl
0.50
2.8
2.1
nd
Comparison of two P1 groups: 5-chlorothiophene and 3-chloroindole.
derivatives.6 Furthermore, the more sterically demand-
ing the X substituent, the larger the potency difference
is, for example, 25 versus 26 (X = NH), twofold; while
27 versus 28 (X = N-Fmoc), 36-fold.
Table 4. SAR of R groups of the enantiopure cis-(3R,4S)-pyrrolidinyl
diamine derivatives
R
N
O
S
O
NH HN
Using the preferred 5-chlorothiophene P1 group, R sub-
stituents on the nitrogen atom of the pyrrolidinyldia-
O
mide core were further investigated. Table
4
N
Cl
demonstrates that a variety of substituents were toler-
ated. Neutralization of the pyrrolidine ring in 25 to form
amides, carbamates, sulfonamides, and ureas resulted in
improved FXa inhibitory activity (29, 31, and 33–46).
Among the compounds with amide substituents, the
methyl amide analog 29 (FXa Ki = 0.58 nM, PT
EC2x = 0.98 lM) was three to four times more potent
than those with larger hydrophobic groups such as
COCH2Me, COCHMe2, CO-t-Bu, CO-cyclopropyl,
CO-phenyl. Polar substitution in the amide, such as CO-
CH2OMe, was tolerated, and compound 38 was equipo-
tent to the methyl amide 29 in FXa inhibitory activity,
but slightly less active in the anticoagulant activity.
The methyl and ethyl carbamates 39 and 31 as well as
the methyl and ethyl sulfonamides 41 and 42 were also
potent FXa inhibitors (FXa Ki < 1 nM, PT
EC2x < 3 lM). The dimethyl urea 46 was the most
potent urea analog, having a FXa Ki of 0.56 nM and a
PT EC2x of 1.8 lM. Similar to the amides, sulfonamides
and ureas with larger alkyl groups, such as 43 bearing a
isopropylsulfonamide and 44 with a pyrrolidinyl urea,
were less potent than those with a smaller alkyl group,
such as 41, 42, and 46.
Compound
R
H
FXa
Ki (nM) (lM)
PT EC2x Microsomal
rate
(nmol/
min/mg)a
25
28
29
33
34
35
36
37
38
39
31
40
9.5
4.5
0.58
2.0
1.3
1.7
nd
nd
0.00
nd
Fmoc
COMe
0.94
3.2
2.9
4.5
4.0
9.2
3.6
3.0
2.1
2.8
0.00
nd
COEt
COCHMe2
CO-t-Bu
0.00
nd
CO-Cyclopropyl 2.1
nd
nd
CO-Phenyl
COCH2OMe
COOMe
COOEt
2.2
0.36
0.74
0.50
0.93
nd
0.00
0.00
nd
COOCH2
CH2OMe
SO2Me
41
42
43
44
0.44
0.64
1.7
1.6
3.1
3.1
6.0
0.00
0.00
nd
SO2Et
SO2i-Pr
CO-N-
1.7
nd
Pyrrolidinyl
CONHMe
CONMe2
45
46
1.1
0.56
2.4
1.8
0.00
0.00
Figure 5 depicts the binding model for pyrrolidinyldi-
amine derivative 31, in which a ring flip similar to that
in 13 (Fig. 4) is proposed, extending the carbamate into
a Based on % remaining after 10-min incubation in human liver
microsomes.

J. X. Qiao et al. / Bioorg. Med. Chem. Lett. 17 (2007) 5041–5048
5045
all the compounds tested were very weak against all
P450 isozymes (except for the basic pyrrolidinyl 25
showing low lM activity for CYP3A4), non-cytotoxic
in the cytotoxicity assay with IC50 > 100 lM, and weak
against hERG channel (most compounds had hERG
flux IC50 > 80 lM).12 Importantly, all the compounds
tested in HLM incubation showed improved metabolic
stability (100% remaining after 10 min) compared with
the cyclopentyl analog 1.
Table 5 illustrates the selectivity profile of example com-
pounds bearing the 4-substituted cyclopentyldiamine,
the tetrahydrofuranyldiamine, and the N-substituted
pyrrolidinyldiamine cores. In general, these compounds
were highly selective against other serine proteases.
Table 6 illustrates the pharmacokinetic properties of
selected compounds in dogs. All showed low to moder-
ate Vdss. Compared with the corresponding cyclopentyl
analog 1, the tetrahydrofuran 23 and the pyrrolidinyl
analogs 29, 31, and 42 had increased HLM stability,
but higher in vivo clearance. This may be due to a
non-metabolic clearance mechanism, such as renal or
biliary clearance in vivo, and/or species differences in
metabolism. Changing the cyclopentyl core in 1 to the
tetrahydrofuran core in 23 improved bioavailability.
Figure 5. Binding model of the pyrrolidinyldiamide 31 in the active site
of FXa showing the carbamate functionality extended into solvent.
Created using PyMol.8
solvent,11 supporting the similar FXa binding affinity
observed in compounds 31 and 1.
The in vitro liability profile of the pyrrolidinyldiamine
compounds (Table 4) was generally good. For instance,
Table 5. Selectivity profile of enantiopure 4-substituted cyclopentyldiamine, tetrahydrofuranyldiamine, and 3,4-pyrrolidinyldiamine derivatives
Compound FXa Ki FXIa Ki FVIIa Chymotrypsin Plasma Trypsin Thrombin aPC Plasmin tPA Urokinase
Ki (nM) Ki (nM) kallikrein Ki (nM) Ki (nM)
(nM)
(nM)
Ki (nM) Ki (nM) Ki (nM) Ki (nM)
Ki (nM)
15
16
23
29
31
42
0.56
1.9
nd
>11,000 >20,000
>10,800
>6000
nd
>5000
>5000
>5000
>5000
>5000
>5000
nd
>21,000 >22,000 >21,000 >14,000
>21,000 >22,000 >21,000 >14,000
>21,000 >22,000 >21,000 >14,000
>21,000 >22,000 >21,000 >14,000
>21,000 >22,000 >21,000 >14,000
>21,000 >22,000 >21,000 >14,000
>11,000 >11,000 >20,000
>11,000 >11,000 >20,000
10,170 >11,000 >20,000
7380 >11,000 >20,000
6680 >11,000 >20,000
>12,000
>12,000
5440
2.6
0.58
0.50
0.64
>6000
>6000
>10,800
4020
3950
Table 6. In vitro and dog PK profiles of representative five-membered cyclicdiamines containing phenylpyridone P4 residue
R
X
O
O
NH HN
S
O
Cl
N
Compound
X
R
FXa
Ki (nM)
PT EC2x
(lM)
Microsomal
rate
(nmol/min/mg)a
Caco-2 Pc
(nm/s)
Cl
(L/Kg/h)b
Vdss
(L/Kg)b
iv t1/2
(h)
po t1/2
(h)c
F%c
1
23
29
31
42
CH2
O
—
—
0.43
2.6
1.7
3.2
0.90
2.1
3.1
0.063
0.031
0.000
0.000
0.000
92
—
0.7
1.6
1.8
2.6
1.5
0.8
1
0.7
1.2
0.8
1.4
0.7
0.7
1.2
0.8
2.7
1.3
60
97
5
N
COMe
COOEt
SO2Et
0.58
0.50
0.64
<15
25
2.4
2.4
1.1
N
25
16
N
<15
a Based on % remaining after 10-min incubation in human liver microsomes.
b iv dose: 0.5 mg/kg.
c po dose: 0.2 mg/kg.

5046
J. X. Qiao et al. / Bioorg. Med. Chem. Lett. 17 (2007) 5041–5048
Table 7. Detailed kinetic parameters for 31a,b
Compound 31 was selected for detailed mechanistic
studies and the key kinetic parameters are summarized
in Table 7. As predicted for a reversible inhibitor bind-
ing in the active site, 31 exhibits competitive inhibition
versus a tripeptide chromogenic substrate, but exhibits
mixed-type inhibition (with ca. fourfold lower affinity
for the ES complex) versus prothrombin, the physiolog-
ical substrate, which interacts with FXa primarily at
exosites.13,14 An advantageous consequence of this inhi-
bition mechanism is that 31 is a potent inhibitor at both
sub-saturating and saturating levels of prothrombin.
Substrate-dependent inhibition mechanisms have also
been reported for other FXa inhibitors.13,15 At both
25 ꢁC and 37 ꢁC the second-order rate constant for asso-
ciation of 31 with FXa, determined by stopped-flow
spectrofluorimetry, is rapid and approaches the diffusion
controlled limit. Rapid onset of inhibition of blood
coagulation is preferable and has been observed for
optimized thrombin16 and FXa inhibitors.13,15b,c,17
Binding of 31 to FXa proceeds by a simple one-step
mechanism or by a two-step mechanism with a very
weak initial complex (i.e., initial Ki ꢀ 5000 nM), since
Human Factor Xa
parameter
31 Parameter
value
25 ꢁC Ki (tripeptide substrate)
37 ꢁC Ki (tripeptide substrate)
25 ꢁC Ki (prothrombin)
37 ꢁC Ki (saturating prothrombin)
25 ꢁC Association rate constant
25 ꢁC Dissociation rate constant
37 ꢁC Association rate constant
37 ꢁC Dissociation rate constant
0.73 nM
2.6 nM
7.5 nM
30 nM
3.2 · 107 Mꢁ1 sꢁ1
2.3 · 10ꢁ2 sꢁ1
2.3 · 107 Mꢁ1 sꢁ1
6.1 · 10ꢁ2 sꢁ1
a Ki’s measured with purified human enzymes and averaged from
multiple determinations (n P 2).
b Prothrombinase inhibition, association and dissociation rate con-
stants were obtained as described in Ref. 13.
The ethyl carbamate 31 was the only pyrrolidinyldi-
amine analog having detectable cell permeability
(Caco-2 Pc 25 nm/s). As anticipated based on permeabil-
ity, 31 showed the best oral bioavailability of the pyr-
rolidinyldiamine compounds (F% = 25%) and a slightly
prolonged t1/2 of 1.4 h (iv) and 2.7 h (po) in dogs.
f
a,b,c
d,e
BocHN
O
NH2
BocHN
O
OH
BocHN
N3
H2N
OH
N
O
h,i
O
H
g
NH
N
N
H
O
BocHN
8
N
HN
Cl
Scheme 1. Reagents and conditions: (a) (Boc)2O, Et3N, THF, 99%; (b) p-NO2C6H4COOH, DEAD, PPh3, 57%; (c) NaOMe, CH3OH, 59%; (d)
MsCl, Et3N, CH2Cl2, 0 ꢁC, 95%; (e) NaN3, DMSO, 100 ꢁC, overnight, 47%; (f) H2, Pd–C (5%), EtOH, 94%; (g) 4-(2-oxopyridin-1(2H)-yl)benzoic
acid, BOP, NMM, DMF, 1 h, 96%; (h) TFA, CH2Cl2, rt, 1 h, 61%; (i) 3-Chloro-1H-indole-6-carboxylic acid, BOP, NMM, DMF, 37%.
Fmoc
N
Fmoc
Fmoc
N
Fmoc
N
N
O
H
a,b,c,d
g,h
f
e
N
N
H
O
HO
NHBoc
HO
N3
N3
NHBoc
BocHN
H
H
N
49
48
N
N
50
Cr
Cl
O
O
(S
,
S)-47
O
O
Fmoc
HN
N
O
N
O
O
N
H
O
l
N
O
k
i,j
N
H
O
O
O
H
O
NH
NH
NH
N
N
N
51
31
S
S
S
Cl
Cl
Cl
Scheme 2. Reagents and conditions: (a) Fmoc–Cl, DIEA, CH2Cl2, 95%; (b) m-CPBA, NaHCO3, CH2Cl2, 60%; (c) TMSN3, (S,S)-47 (5 mol%), Et2O,
89%; (d) 10-CSA (cat.), MeOH, 85%; (e) H2, 10% Pd–C, (Boc)2O, EtOAc, 91%; (f) NaN3, H2SO4, H2O/toluene, 0 ꢁC, then PPh3, DEAD, HN3, THF,
ꢁ78 ꢁC to rt; (g) H2, 10% Pd–C, EtOH; (h) 4-(2-oxopyridin-1(2H)-yl)benzoic acid, BOP, NMM, THF, 44% for three steps.; (i) 30% TFA, CH2Cl2; (j)
5-Cl-thiophene-2-carboxylic acid, BOP, NMM, THF, 93% for two steps; (k) 10% piperidine, THF, 90%; (l) ClCOOEt, Et3N, THF, 80%.

J. X. Qiao et al. / Bioorg. Med. Chem. Lett. 17 (2007) 5041–5048
5047
plots of the observed rate constants versus inhibitor
were linear up to 5000 nM. Dissociation rate constants
could thus be calculated from the relationship,
Ki = kdissoc/kassoc, and account for the higher Ki’s at
37 ꢁC. These results are similar to those obtained with
both razaxaban and apixaban.13
A. Barbera, Tracy A. Bozarth, and Karen S. Hartl for
in vitro assays.
References and notes
1. Presented in part at the 232nd ACS National Meeting San
Francisco, CA, September 10–14, 2006, MEDI-393.
2. For review papers of FXa inhibitors, see: (a) Quan, M.;
Smallheer, J. Curr. Opin. Drug Disc. Dev. 2004, 7, 460; (b)
Walenga, J. M.; Jeske, W. P.; Hoppensteadt, D.; Fareed,
J. Curr. Opin. Invest. Drugs 2003, 4, 272; (c) Gould, W. R.;
Leadley, R. J. Curr. Pharm. Design 2003, 9, 2337; (d)
Quan, M. L.; Wexler, R. R. Curr. Top. Med. Chem.
(Hilversum, Netherlands) 2001, 1, 137.
3. (a) Lassen, M. R.; Davidson, B. L.; Gallus, A.; Pineo, G.;
Ansell, J.; Deitchman, D. Blood 2003, 102, 11; (b) Straub,
A.; Pohlmann, J.; Lampe, T.; Pernerstorfer, J.; Schlem-
mer, K.-H.; Reinemer, P.; Perzborn, E.; Roehrig, S. J.
Med. Chem. 2005, 48, 5900; (c) Eriksson, B. L.; Borris, L.;
Dahl, O. E.; Haas, S.; Huisman, M. V.; Kakkar, A. K. J.
Thromb. Haemost. 2006, 4, 121; (d) Hampton, T. JAMA
2006, 295, 743; (e) Wong, P. C.; Crain, E. J.; Watson, C.
A.; Zaspel, A. M.; Wright, M. R.; Lam, P. Y. S.; Pinto, D.
J.; Wexler, R. R.; Knabb, R. M. J. Pharmacol. Exp. Ther.
2002, 303, 993.
Scheme 1 illustrates the synthesis of enantiopure diamin-
oindane derivatives using 8 as an example following the
similar sequence as that for compounds 1.6
The synthetic route for enantiopure cis-3,4-diamino-
pyrrolidine core is outlined in Scheme 2 using the prep-
aration of compound 31 as an example. The 12-step
synthesis involved two key reactions. One is the synthe-
sis of the Fmoc protected hydroxylazide pyrrolidine 48
via stereospecific ring opening of the meso epoxide with
TMS azide catalyzed by Jacobsen’s chiral (salen)chro-
mium(III) catalyst 47.18 The Fmoc group was chosen
as the amino protecting group because it allows high
ee and easy monitoring of the reaction progress. The
other key transformation is from the trans Boc-pro-
tected amino alcohol 49 to the cis Boc-protected azide
50. Normal azide formation via sodium azide displace-
ment of the corresponding mesylate in either DMF or
DMSO at either room temperature or elevated temper-
ature to 80 ꢁC generated undesirable products, presum-
ably due to the initial Fmoc decomposition followed by
other reactions. However, treatment of alcohol 49 with a
freshly prepared HN3 toluene solution under Mitsunobu
condition led to the desired cis Boc-protected amino
azide 50. Reduction of the crude azide 50, and then
coupling the resulting amine with 4-(2-oxopyridin-
1(2H)-yl)benzoic acid, followed by deprotection and a
subsequent amide formation with 3-chloro-1H-indole-
6-carboxylic acid afforded the Fmoc protected diamino-
pyrrolidine 51. Deprotection of Fmoc group followed by
carbamate formation provided compound 31. Using
similar strategies, the tetrahydrofuranyldiamides 23
and 24 were also prepared with an ee >97% measured
by chiral analytical HPLC.
4. Quan, M. L.; Lam, P. Y. S.; Han, Q.; Pinto, D. J.; He, M.;
Li, R.; Ellis, C. D.; Clark, C. G.; Teleha, C. A.; Sun, J. H.;
Alexander, R. S.; Bai, S. A.; Luettgen, J. M.; Knabb, R.
M.; Wong, P. C.; Wexler, R. R. J. Med. Chem. 2005, 48,
1729.
5. Pinto, D. J. P.; Orwat, M. J.; Koch, S.; Rossi, K. A.;
Alexander, R. S.; Smallwood, A.; Wong, P. C.; Rendina,
R. A.; Luettgen, J. M.; Knabb, R. M.; He, K.; Xin, B.;
Wexler R.R.; Lam, P.Y.S. J. Med. Chem. in press.
6. Qiao, J. X.; Chang, C.-H.; Cheney, D. L.; Morin, P. E.;
King, S. R.; Wang, G. Z.; Wang, T. C.; Rendina, A. R.;
Luettgen, J. M.; Knabb, R. M.; Wexler, R. R.; Lam, P. Y. S.
Bioorg.Med.Chem.Lett.2007.doi:10.1016/j.bmcl.2007.06.029.
7. Pinto, D.; Quan, M.; Orwat, M.; Li, Y.; Han, W.; Qiao, J.;
Lam, P.; Koch, S. PCT Int. Appl. WO 03026652 A1, 2003.
8. PyMol. DeLano, W. L. The PyMOL Molecular Graphics
System (2002) on World Wide Web http://www.
pymol.org.
9. (a) Knabb, R. M.; Kettner, C. A.; Timmermans, P. B. M.
W. M.; Reilly, T. M. Thromb. Haemost. 1992, 67, 56; (b)
Kettner, C. A.; Mersinger, L. J.; Knabb, R. M. J. Biol.
Chem. 1990, 265, 18289.
10. Structures 6 and 8 were docked into a representative, in-
house crystal structure of FXa using GLIDE XP with
refinement using the OPLS-AA force-field and SGB
continuum model in the IMPACT module in Maestro.
Maestro, version 6.0, Schro¨dinger, LLC, New York, NY,
2005.
11. Structures of compounds 11, 13, and 31 were docked with
GLIDE XP version 4.0 into a conformationally represen-
tative ensemble of crystal structures of the FXa binding
site. Top-scoring poses were further refined using the MM-
GBSA routine in Maestro in combination with manual
modeling. Maestro, version 7.5, Schro¨dinger, LLC, New
York, NY, 2005. For protein ensemble docking see:
Cheney, D. L.; Mueller, L. Abstracts of Papers, 229th
ACS National Meeting, San Diego, CA, United States,
March 13–17, 2005, COMP-307.
In summary, to improve the metabolic stability of the
cyclopentyldiamine derivative 1 while maintaining the
sub-nanomolar FXa potency, we synthesized several
enantiopure five-membered cyclicdiamine series: the
4-substituted cyclopentyldiamine, the tetrahydrofura-
nyl-diamine, and the pyrrolidinyldiamine derivatives.
Compared with 1, those compounds having a similarly
high FXa potency had an improved metabolic stability
in HLM (90–100% remaining after 10 min). The tetrahy-
drofuranyldiamide 23 had an improved bioavailability
in dogs. The ethyl carbamate 31 in the (3R,4S)-pyr-
rolidinyldiamine series, having excellent potency and
selectivity, and an HLM stability better than 1, was
the best overall compound with a bioavailability of
25% and a t1/2 of 2.7 h in dogs.
Acknowledgments
12. For general descriptions regarding assays for Cyp P450,
cytotoxicity, and hERG, please see: (a) Crespi, C. L.;
Miller, V. P.; Penman, B. W. Anal. Biochem. 1997, 248,
188; (b) Graham, F. L.; Smiley, J.; Russell, W. C.; Nairn,
The authors thank Mary F. Grubb for metabolic ID
study of compound 1, and Jeffrey M. Bozarth, Frank

5048
J. X. Qiao et al. / Bioorg. Med. Chem. Lett. 17 (2007) 5041–5048
R., J. Gen. Virol. 1977, 36, 59; (c) Sanguinetti, M. C.;
He, K.; Wexler, R. R.; Lam, P. Y. S. Bioorg. Med.
Chem. Lett. 2006, 16, 5584; (c) Pinto, D. J. P.; Orwat,
M. J.; Quan, M. L.; Han, Q.; Galemmo, R. A.;
Amparo, E.; Wells, B.; Ellis, C.; He, M. Y.; Alexander,
R. S.; Rossi, K. A.; Smallwood, A.; Wong, P. C.;
Luettgen, J. M.; Rendina, A. R.; Knabb, R. M.;
Mersinger, L.; Kettner, C.; Bai, S.; He, K.; Wexler, R.
R.; Lam, P. Y. S. Bioorg. Med. Chem. Lett. 2006, 16,
4141.
Jurkiewicz, N. K. J. Gen. Physiol. 1990, 96, 195; (d)
Sanguinetti, M. C.; Curran, M. E.; Spector, P. S.; Keating,
M. T. Proc. Natl. Acad. Sci. 1996, 93, 2208; (e) Trudeau,
M. C.; Warmke, J. W.; Ganetzky, B.; Robertson, G. A.
Science 1995, 269, 92; (f) Warmke, J. W.; Ganetzky, B.
Proc. Natl. Acad. Sci. 1994, 91, 3438.
13. Experimental details for the mechanism studies will be
described in a manuscript in preparation on razaxaban
and apixaban.
14. Krishnaswamy, S. J. Thromb. Haemost. 2005, 3, 54.
15. (a) Rezaie, A. R. Thromb. Haemost. 2003, 89, 112; (b)
Pinto, D. J. P.; Galemmo, R. A.; Quan, M. L.; Orwat,
M. J.; Clark, C.; Li, R.; Wells, B.; Woerner, F.;
Alexander, R. S.; Rossi, K. A.; Smallwood, A.; Wong,
P. C.; Luettgen, J. M.; Rendina, A. R.; Knabb, R. M.;
16. Elg, M.; Gustafsson, D.; Deinum, J. Thromb. Haemost.
1997, 78, 1286.
17. Gould, W. R.; Cladera, E.; Harris, M. S.; Zhang, E.;
Narasimhan, L.; Thorn, J. M.; Leadley, R. J., Jr.
Biochemistry 2005, 44, 9280.
18. Schaus, S. E.; Larrow, J. F.; Jacobsen, E. N. J. Org. Chem.
1997, 62, 4197.