Orally bioavailable amine-linked macrocyclic inhibitors of factor XIa

Article abstract of DOI:10.1016/j.bmcl.2020.126949

The discovery of orally bioavailable FXIa inhibitors has been a challenge. Herein, we describe our efforts to address this challenge by optimization of our imidazole-based macrocyclic series. Our optimization strategy focused on modifications to the P2 prime, macrocyclic amide linker, and the imidazole scaffold. Replacing the amide of the macrocyclic linker with amide isosteres led to the discovery of substituted amine linkers which not only maintained FXIa binding affinity but also improved oral exposure in rats. Combining the optimized macrocyclic amine linker with a pyridine scaffold afforded compounds 23 and 24 that were orally bioavailable, single-digit nanomolar FXIa inhibitors with excellent selectivity against relevant blood coagulation enzymes.

Full text of DOI:10.1016/j.bmcl.2020.126949

Bioorganic&MedicinalChemistryLettersxxx(xxxx)xxxx
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry Letters
journal homepage: www.elsevier.com/locate/bmcl
Orally bioavailable amine-linked macrocyclic inhibitors of factor XIa
⁎
⁎
Tianan Fanga, , James R. Cortea, , Paul J. Gilligana, Yoon Jeona, Honey Osunaa, Karen A. Rossia,
Joseph E. Myers Jr.b, Steven Sheriffb, Zhen Louc, Joanna J. Zhengc, Timothy W. Harperc,
Jeffrey M. Bozarthc, Yiming Wuc, Joseph M. Luettgenc, Dietmar A. Seiffertc, Ruth R. Wexlera,
Patrick Y.S. Lama
a Bristol-Myers Squibb Company, Research and Development, 350 Carter Road, Hopewell, NJ 08540, United States
b Bristol-Myers Squibb Company, Research and Development, US Rt. 206 & Province Line Road, Princeton, NJ 08540, United States
c Bristol-Myers Squibb Company, Research and Development, 311 Pennington-Rocky Hill Road, Pennington, NJ 08543-2130, United States
A R T I C L E I N F O
A B S T R A C T
Keywords:
The discovery of orally bioavailable FXIa inhibitors has been a challenge. Herein, we describe our efforts to
address this challenge by optimization of our imidazole-based macrocyclic series. Our optimization strategy
focused on modifications to the P2 prime, macrocyclic amide linker, and the imidazole scaffold. Replacing the
amide of the macrocyclic linker with amide isosteres led to the discovery of substituted amine linkers which not
only maintained FXIa binding affinity but also improved oral exposure in rats. Combining the optimized mac-
rocyclic amine linker with a pyridine scaffold afforded compounds 23 and 24 that were orally bioavailable,
single-digit nanomolar FXIa inhibitors with excellent selectivity against relevant blood coagulation enzymes.
Factor Xia
FXIa
Activated partial thromboplastin time
aPTT
Thrombosis
Cardiovascular diseases (CVDs), the leading cause of death world-
wide, accounted for about 32% of the 55.9 million global deaths in
2017.1 The underlying pathophysiology of CVDs often involves the
formation of a thrombus that rapidly slows or stops blood flow poten-
tially leading to heart attack or stroke. The direct thrombin inhibitor
dabigatran and the factor Xa inhibitors (rivaroxaban, apixaban, edox-
aban, and betrixaban) are effective for the treatment or prevention of
deep vein thrombosis (DVT), pulmonary embolism (PE) or stroke
arising from nonvalvular atrial fibrillation (NVAF).2,3 However, data
from clinical trials and post-marketing data indicates that bleeding re-
mains a safety concern, especially in patients with increased bleeding
risk.4 Thus, a need exists for an oral antithrombotic with a wider
therapeutic index with respect to the bleeding safety profile. Factor XIa
(FXIa), the activated form of the zymogen FXI, plays a major role in the
amplification pathway of the coagulation cascade but not in hemos-
tasis.5 Inhibition of FXIa has the potential of providing an effective
antithrombotic agent with limited effects on hemostasis. Preclinical
studies have shown that peptide and small molecule FXIa inhibitors
prevent thrombosis with less bleeding liability.6,7
(Table 1).12 These macrocycles with an amide-linker displayed high
FXIa binding affinity (FXIa Ki ≤1 nM)13 and are potent in the activated
partial
thromboplastin
(aPTT)
clotting
assay
(aPTT
EC1.5x ≤ 0.30 μM);14 however, these compounds were not orally
bioavailable (%Frat < 1). We hypothesized that high polar surface area
(PSA = 169 Å2 for 1 and 2) may contribute to poor permeability [Caco-
2 AB/BA (nm/s): < 15/67 for 1 and < 15/58 for 2] and lack of oral
bioavailability.15 Our initial optimization strategy for improving oral
bioavailability by replacing the chlorophenyltetrazole P1 with less
polar groups to reduce PSA led to the discovery of novel P1 groups that
proved to be successful for improving oral bioavailability.16 Described
herein is an alternative strategy in which the chlorophenyltetrazole P1
was kept fixed and the P2 prime, macrocyclic amide linker, and scaffold
regions of the macrocycle were modified with the goal of improving the
oral bioavailability. In this communication, we describe SAR in these
three regions culminating in the discovery of a macrocyclic amine
linker which maintained FXIa binding affinity and improved oral
bioavailability for both the imidazole- and pyridine-based macrocyclic
series.
Several small molecule irreversible,6,8 reversible,7,9,10 and allos-
teric11 FXIa inhibitors have been reported in the literature. Bristol-
Myers Squibb has reported on the discovery of a novel series of 12- and
13-membered imidazole-based macrocyclic FXIa inhibitors 1 and 2
First, we explored the P2 prime region of the macrocycle by re-
moving the carbamate moiety to reduce PSA (Table 2). Based on X-ray
crystallography, the carbamate moiety is engaged in several H-bonding
interactions with FXIa (the NH of the carbamate forms an H-bond to
⁎ Corresponding authors.
E-mail addresses: tianan.fang@bms.com (T. Fang), james.corte@bms.com (J.R. Corte).
https://doi.org/10.1016/j.bmcl.2020.126949
Received 15 November 2019; Received in revised form 23 December 2019; Accepted 1 January 2020
0960-894X/©2020ElsevierLtd.Allrightsreserved.
Pleasecitethisarticleas:TiananFang,etal.,Bioorganic&MedicinalChemistryLetters,https://doi.org/10.1016/j.bmcl.2020.126949

T. Fang, et al.
Bioorganic&MedicinalChemistryLettersxxx(xxxx)xxxx
Table 1
however a 9-fold loss in aPTT clotting activity (EC1.5x = 2.6 μM) was
seen. The trifluoroethylamine proved to be a good replacement for the
amide moiety in the 12-membered macrocycle but not the 13-mem-
bered macrocycle 14, where a 330-fold loss in FXIa affinity (FXIa
Ki = 53 nM) was observed. Since 12-membered macrocycles with an
amine linker were more potent, additional substitutions were explored.
Diastereomer (R)-CHF2 11 was equipotent to (R)-CF3 10. When the
fluoro alkyl groups were replaced with either CH3 as in 12 or hydrogen
as in 13, a loss in both FXIa affinity (> 6-fold) and metabolic stability
was observed. The difference in FXIa binding affinity can be rationa-
lized based on the electron withdrawing ability of the R-substituent.
Specifically, the stronger electron withdrawing ability of the fluoro
alkyl substituents in 10 and 11 makes the adjacent NH a better H-bond
donor for Leu41 compared to either the methyl or hydrogen substituent
in 12 and 13.
12- and 13-membered macrocyclic FXIa inhibitors (the macrocyclic amide
linker is highlighted in blue).
Compd #
n
Saturation/unsaturation
FXIa Ki
(nM)13
aPTT EC1.5x
(μM)14
(ring size)
1
2
0 (12)
1 (13)
saturated
E-alkene
1.0
0.30
0.27
An X-ray crystal structure of 10 bound to human FXIa active site
(2.2 Å resolution, Fig. 1) was obtained.20 The compound occupies the
S1, S1 prime, and the S2 prime binding pockets. The absolute stereo-
chemistry of the trifluoroethylamine moiety was determined to be the
R-configuration. An overlay of trifluoroethylamine linker 10 and amide
linker 1 is shown in Fig. 2 and highlights the interaction of the tri-
fluoroethylamine moiety. As postulated in the design, the tri-
fluoroethylamine moiety maintains the key H-bond (3.0 Å) with Leu41
and mimics the amide interaction. Interestingly, the X-ray crystal
structure also revealed that the CF3 group, known to be a weak H-bond
acceptor,21 is within hydrogen bonding distance (3.0 Å) to Arg39. This
weak H-bonding interaction may also be contributing to the FXIa
binding affinity of the fluoro alkyl analogs 10 and 11, but not 12 or 13.
In order to explore this potential interaction between the CF3 group
and Arg39, the CF3 moiety was replaced with other potential H-bond
acceptors (Table 4). We focused on 12-membered macrocycles since the
13-membered macrocycle with the trifluoroethylamine linker was
much less potent. Replacing the CF3 with methyl ester 15 resulted in a
2-fold increase in FXIa affinity (FXIa Ki = 0.46 nM) and a 5-fold im-
0.16
His40 and the carbonyl of the carbamate forms an H-bond to Ile151 via
a water molecule) and it was not clear how the loss of these interactions
would affect FXIa affinity.12 Removing the carbamate moiety in 12-
membered macrocycle 3 lowered PSA (130 Å2) but also led to a sig-
nificant loss in FXIa affinity (48-fold; FXIa Ki = 48 nM), aPTT potency
(> 100 fold; EC1.5x > 40 μM) and metabolic stability [15-fold; human
liver microsome (HLM) t1/2 = 6 min]. A similar loss in FXIa binding
affinity (32-fold) and metabolic stability was observed when the car-
bamate was removed in the 13-membered macrocycle 4; however, a
single-digit nanomolar FXIa inhibitor (FXIa Ki = 5.1 nM) with good
aPTT potency (EC1.5x = 2.9 μM) was realized. Encouraged by this result
in the 13-membered macrocycle, additional substituents were explored
at the 4- and 5-positions of the P2 prime phenyl group with the goal of
improving metabolic stability. Whereas introduction of a fluoro at C5
(5) resulted in a slight loss in FXIa affinity, C4-F (6) was equipotent
with 4. Unfortunately, neither 5 nor 6 improved HLM stability.
Compound 8, with a carboxylic acid group at C4, was the only analog
that maintained FXIa affinity and improved metabolic stability
(HLM t1/2 = 94 min), but unfortunately PSA (168 Å2) was high. Since
efforts to reduce PSA in the P2 prime region led to a loss in both FXIa
affinity and metabolic stability, we turned our attention to modifying
the amide moiety in the macrocyclic linker.
provement in aPTT clotting potency (EC1.5x
= 0.55 μM). The
ethyl ester 16 was similar in potency to methyl ester 15. These ester
analogs had high PSA values (178 Å2) and poor metabolic stability
(HLM t1/2 ≤10 min). Acid 17 was similar in potency to the corre-
sponding ester analogs and exhibited improved metabolic stability
(HLM t1/2 = 70 min). Amide 18 and hydroxymethyl 19 analogs dis-
played good FXIa affinity and aPTT potency but also suffered from poor
metabolic stability. Homologation of methyl ester 20 led to a 4- to
5-fold loss in FXIa affinity. The NH moiety in the macrocyclic linker
proved to be the major contributor to FXIa affinity, as removal of the
NH led to a 50-fold loss in FXIa binding affinity for the all-carbon linker
21.22 In general, the imidazole-based macrocycles containing the sub-
stituted amine linkers in Table 4 had high PSA values (≥172 Å2)
and, aside from acid 17, suffered from poor metabolic stability
(HLM t1/2 ≤ 14 min).
During our initial exploration of the imidazole-based macrocyclic
series, the macrocycles with an amide linker were identified as giving
the most potent FXIa inhibitors.12 It was demonstrated by X-ray crys-
tallography that the NH of the amide moiety forms a key H-bond with
Leu41 (2.8 Å) which contributed significantly to FXIa binding affinity
but proved to be deleterious for oral bioavailability.12 The literature
suggests that amide isosteres may be a means of modulating polarity
and oral bioavailability.17 We were particularly interested in replacing
the amide moiety with a suitable isostere that had the potential to
maintain the key H-bonding interaction with Leu41 such as a tri-
fluoroethylamine [–CH(CF3)NH–] group. The trifluoroethylamine unit
has been adopted as an effective amide isostere in drug design since it is
a non-basic amine with excellent H-bond donating ability.18,19 How-
ever, it was not clear whether the trifluoroethylamine moiety, once it
was incorporated into the linker, would be able to adopt the desired
conformation to maintain the key H-bond with Leu41. Moreover, an
unsubstituted macrocycle linked by an amine was prepared earlier in a
13-membered macrocycle and it showed a significant loss in FXIa af-
finity.12
X-ray crystal structures of methyl ester 15 (2.12 Å resolution, Fig. 3)
and ethyl ester 16 (2.07 Å resolution, Fig. 4A) bound to human FXIa
active site were obtained.23 The carbonyl of the methyl ester in 15
formed an H-bond to Arg39 (2.8 Å). The ether oxygen forms an H-bond
to the Cys58 backbone carbonyl via two water molecules. Interestingly,
the ethyl ester in 16 does not form an H-bond to Arg39. Rather, the
ethyl ester is flipped so that the ethyl portion forms a van der Waals
contact with a small pocket created by residues 39–41 and the carbonyl
makes an H-bond to Cys58 carbonyl through two water molecules
(Fig. 4B).
Previous work from our lab on the macrocyclic series showed that
the imidazole scaffold can be replaced with a pyridine ring system and
maintain FXIa binding affinity.16c As a result, the optimized amine
linkers were combined with a pyridine scaffold (Table 5). The tri-
fluoroethylamine linker in pyridine macrocycle 22 led to a 14-fold loss
in FXIa affinity (FXIa Ki = 16 nM) and a 5-fold loss in aPTT activity
(EC1.5x = 13 μM) when compared to imidazole-based analog 10.
Due to the synthetic route, two diastereomers were produced by
incorporating the trifluoroethylamine group into the macrocyclic
backbone. In the 12-membered macrocycle, both diastereomers (S)-CF3
9 and (R)-CF3 10 (Table 3) exhibited a desired, lower PSA value (152 Å2
vs 169 Å2 for amide 1). Diastereomer (S)-CF3 9 exhibited a significant
loss (370-fold) in FXIa affinity, while diastereomer (R)-CF3 10 showed
similar FXIa binding affinity (FXIa Ki = 1.1 nM) compared to amide 1,
2

T. Fang, et al.
Bioorganic&MedicinalChemistryLettersxxx(xxxx)xxxx
Table 2
P2 prime SAR in the 12- and 13-membered imidazole macrocycles.
Compd #
n
Saturation/unsaturation
R
FXIa Ki
(nM)13
aPTT EC1.5x
(μM)14
PSAa
(Å2)
HLMb
(ring size)
t1/2 (min)
1
3
2
4
5
6
7
8
0 (12)
0 (12)
1 (13)
1 (13)
1 (13)
1 (13)
1 (13)
1 (13)
saturated
saturated
E-alkene
E-alkene
E-alkene
E-alkene
E-alkene
E-alkene
5-NHCO2Me
1.0
48
0.30
> 40
0.27
2.9
169
130
169
130
130
130
157
168
90
6
H
5-NHCO2Me
H
0.16
5.1
12
41
3
5-F
16
3
4-F
6
2.5
2
4-CO2Me
4-CO2H
245
3.6
NTc
1.7
6
94
a
b
c
PSA = polar surface area.
HLM = human liver microsome stability.
NT = not tested.
Replacing the CF3 moiety with either CO2Me 23 or CO2Et 24 improved
FXIa binding affinity, resulting in single-digit nanomolar inhibitors
(FXIa Ki = 3.4 nM and 2.5 nM, respectively) with acceptable aPTT
clotting activity. While less potent than the corresponding imidazole-
based macrocycles, these pyridine-based macrocycles containing the
amine-linker had lower PSA values (162 Å2) and improved metabolic
stabilities (HLM t1/2 = 111 min. for 23 and 30 min. for 24) compared
to their imidazole analogs. Acid 25 was equipotent to the corresponding
ester analogs, but further increased the PSA value (173 Å2).
An X-ray crystal structure of ethyl ester 2424 bound to human FXIa
active site (1.89 Å resolution, Supplementary Fig. S1) was obtained and
binds like imidazole macrocycle 16, where the ethyl portion of the ester
is interacting with the small pocket and not engaging Arg39 via an H-
bond.
macrocycle 1 exhibited a low clearance, short half-life, low volume of
distribution, and was not orally bioavailable. Removing the carbamate
from the P2 prime region in 3 resulted in a significant increase in
clearance and no oral exposure, in spite of a lower PSA (130 Å2); the
efflux ratio was high (Caco-2 AB/BA (nm/s) < 15/205, ER > 13).26
Replacing the amide moiety in 1 with the trifluoroethylamine group in
10 not only maintained low clearance but improved oral exposure (10,
AUC = 307 nM*h versus 1, AUC = 24 nM*h) leading to an oral
bioavailability of 9%. Methyl ester, 15, was found to be a very low
clearance compound and was orally bioavailable (%F = 8) despite high
PSA (178 Å2) and poor permeability (Caco-2 AB/BA (nm/s) < 15/
103). While ethyl ester 16 displayed a similar PK profile to 15 and was
orally bioavailable (%F = 7), acid 17 showed only minimal exposure in
rat. Replacing the imidazole scaffold of 15 with a pyridine in 23 further
improved oral bioavailability (%F = 23) while maintaining a low
clearance, short half-life, and low volume of distribution, despite high
Several compounds were evaluated in a rat pharmacokinetic (PK)
model, in an N-in-1 format (Table 6).25 The 12-membered imidazole
Table 3
Amine linker SAR in 12- and 13-membered imidazole macrocycles.
Compd #
n
Saturation/unsaturation
R
FXIa Ki
(nM)13
aPTT EC1.5x
(μM)14
PSAa
(Å2)
HLMb
(ring size)
t1/2 (min)
1
0 (12)
0 (12)
0 (12)
0 (12)
0 (12)
0 (12)
1 (13)
1 (13)
saturated
saturated
saturated
saturated
saturated
saturated
E-alkene
E-alkene
–
1.0
370
1.1
1.3
8.3
16
0.30
NTd
2.6
169
152
152
152
152
152
169
152
90
NTd
54
25
7
c
c
9
(S)-CF3
10
11
12
13
2
(R)-CF3
c
(R)-CHF2
2.0
e
(S)-CH3
5.2
H
–
6.2
6
0.16
53
0.27
> 40
41
29
c
14
(R)-CF3
a
b
c
d
e
PSA = polar surface area.
HLM = human liver microsome stability.
The absolute stereochemistry was assigned based on an X-ray co-crystal of 10 with FXIa (Fig. 1).
NT = not tested.
The configuration of the stereocenter is the same as for (R)-CF3 10 and (R)-CHF2 11, however the R designation changed to S based on the CH3 group which
altered the prioritization in the Cahn-Ingold-Prelog system.
3

T. Fang, et al.
Bioorganic&MedicinalChemistryLettersxxx(xxxx)xxxx
Fig. 1. X-ray crystal structure of 10 bound to Factor XIa with omit electron
density contoured at 3 rmsd (gray mesh). The red spheres depict water mole-
cules, the dotted lines depict hydrogen bonds, and ethylene diol is an artifact of
the flash-cooling procedure. (For interpretation of the references to colour in
this figure legend, the reader is referred to the web version of this article.)
Fig. 3. X-ray crystal structure of 15 bound to Factor XIa with omit electron
density contoured at 3 rmsd (gray mesh). The red spheres depict water mole-
cules, the dotted lines depict hydrogen bonds, and ethylene diol is an artifact of
the flash-cooling procedure. (For interpretation of the references to colour in
this figure legend, the reader is referred to the web version of this article.)
PSA (162 Å2). The ethyl ester 24 exhibited a low clearance, a longer
half-life, and a higher volume of distribution and good oral bioavail-
ability (%F = 41). In addition to improved oral bioavailability for the
amine-linked macrocycles, imidazole 16 and pyridine 24 were
>
1,000-fold selective against a panel of relevant serine proteases (Table
S1 in Supplementary Data Section).
All macrocycles were prepared using ring-closing olefin metathesis
(RCM) as the key macrocyclization step.27 The synthesis of macrocyclic
analogs in which the P2 prime region was modified (Table 2) follows a
synthetic sequence previously described by Corte et al.12 that is in-
cluded in the Supplementary Data (Schemes S1 and S2). The synthesis
of representative imidazole macrocycles with amine-linker 10, 15 and
17 (Tables 3 and 4) is described in Scheme 1. Condensation of aniline
1a12 with 1-ethoxy-2,2,2-trifluoroethanol in EtOH at elevated tem-
perature gave the corresponding aminal, which was then reacted with
allyl magnesium bromide which provided trifluoroethylamine 1b as a
1:1 mixture of diastereomers.28 Pretreatment of 1b with p-toluene-
sulfonic acid to form the imidazolium ion followed by macrocyclization
via RCM using Grubbs II catalyst gave imidazole macrocycle 1c as a
mixture of diastereomers.29 Hydrogenation of 1c over palladium on
carbon in the presence of TFA, followed by separation of diastereomers
Fig. 2. Overlay of X-ray crystal structures of 1 (purple) and 10 (yellow) in
Factor XIa. (For interpretation of the references to colour in this figure legend,
the reader is referred to the web version of this article.)
Table 4
Substituted amine linker SAR in the 12-membered imidazole macrocycle.
Compd #
R
FXIa Ki
(nM)13
aPTT EC1.5x
(μM)14
PSAa
(Å2)
HLMb
t1/2 (min)
15
16
17
18
19
20
21
CO2Me
CO2Et
0.46
0.22
0.67
0.31
0.61
2.0
0.55
0.86
0.87
0.36
0.54
1.1
178
178
189
172
172
178
166
10
7
CO2H
70
14
10
4
C(O)NMe2
CH2OH
CH2CO2Me
–
24
25
8
a
b
PSA = polar surface area.
HLM = human liver microsome stability.
4

T. Fang, et al.
Bioorganic&MedicinalChemistryLettersxxx(xxxx)xxxx
Fig. 4. (A) X-ray crystal structure of 16 bound to Factor XIa with omit electron density contoured at 3 rmsd (gray mesh). The red spheres depict water molecules, the
dotted lines depict hydrogen bonds, and ethylene diol is an artifact of the flash-cooling procedure. (B). van der Waals contact between the ethyl portion of the ester
and small pocket formed by residues 39–41. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this
article.)
by preparatory reverse phase hplc, afforded macrocycle 1d. Global
deprotection with aqueous 5 N HCl provided the corresponding amine,
which was coupled with acid 1k to give 10. Alternatively Ullmann
coupling of bromide 1e12 with (R)-2-aminopent-4-enoic acid provided
acid 1f which was converted to methyl ester 1g under basic conditions
with iodomethane. Macrocyclization of 1g via RCM using Grubbs II
catalyst with microwave heating gave macrocycle 1h. Hydrogenation of
alkene in 1h yielded 1i. Deprotection of 1i as described above followed
by coupling with activated carboxylic ester 1j provided methyl ester 15.
Compound 1i was converted to acid 17 in a three step sequence which
involved deprotection of the Boc and SEM groups, hydrolysis of the
methyl ester under basic conditions, and finally coupling with 1j. The
synthesis for all-carbon imidazole macrocycle 21 was carried out by an
alternative sequence and can be found in Scheme S3 (Supplementary
Data).
Table 5
Substituted amine linker SAR in the 12-membered pyridine macrocycle.
Compd #
R
FXIa Ki
(nM)13
aPTT EC1.5x
(μM)14
PSAa
(Å2)
HLMb
t1/2 (min)
22
23
24
25
CF3
16
13
136
162
162
173
95
CO2Me
CO2Et
CO2H
3.4
2.5
2.6
3.4
6.1
7.1
111
30
99
The syntheses of representative pyridine-based macrocycles with
amine-linker 22, 24 and 25 (Table 5) is described in Scheme 2. Similar
to the imidazole synthesis, aniline 2a16c was condensed with 1-ethoxy-
2,2,2-trifluoroethanol, followed by addition of allyl magnesium
a
b
PSA = polar surface area.
HLM = human liver microsome stability.
Table 6
Pharmacokinetic profile for selected 12-membered macrocyclic compounds.
Entry
Cl
t1/2
Vdss
F
AUCa
PSAd
(Å2)
Dose
PBe rat
(%free)
RLMg
Caco-2
AB/BA
(nm/sec)
(mL/min/kg)
(h)
(L/kg)
(%)
(nM*h)
iv/po (mpk)
t1/2 (min)
1
10
1.0
0.5
1.4
1.8
1.5
1.0
1.2
3.6
0.3
1.2
< 1
< 1
9
24
169
130
152
178
178
189
162
162
1.13/2.27
0.31/0.61
0.72/1.44
0.53/1.06
0.50/1.0
1.6
76
NTf
14
48
34
75
87
59
< 15/67h
< 15/205
NTf
3
53
1
NTf
10
15
16
17
23
24
12
0.4
307c
487b
478b
14b
0.2
3.0
2.8
7.7
5.2
4.9
0.3
8
< 0.2
< 0.2
1.4
< 15/103
< 15/133h
< 15/ < 15
< 15/213
< 15/101h
0.2
7
0.33
0.53
1.3
1
0.49/0.98
0.52/1.04
0.50/1.0
23
41
889b
1294
NTf
0.8
Compounds were dosed in rat in an N-in-1 format. Area under the curve (AUC) values were from the po dose.
a
b
c
d
e
f
g
h
Vehicle for both iv and po: 70% polyethylene glycol 400; 20% water; 10% ethanol.
Vehicle for both iv and po: 70% propylene glycol; 20% water; 10% ethanol.
Compound was dosed in rat in a discrete format.
PSA = polar surface area.
PB = protein binding.
NT = not tested.
RLM = rat liver microsome stability.
Poor recovery in the Caco-2 permeability assay.
5

T. Fang, et al.
Bioorganic&MedicinalChemistryLettersxxx(xxxx)xxxx
Scheme 1. Reagents and conditions: (a) 1-ethoxy-2,2,2-trifluoroethanol, EtOH, 120 °C, 86%; (b) allyl magnesium bromide, THF, 0 °C to rt, 50%; (c) (R)-2-aminopent-
4-enoic acid, CuI, K2CO3, DMSO, 90 °C, 96%; (d) MeI, K2CO3, DMF, 71%; (e) pTsOH, DCM, reflux, then 40 mol% Grubbs II catalyst, 63%; (f) 40 mol% Grubbs II
catalyst, DCE, 120 °C (microwave), 30%; (g) H2, 10% Pd/C, TFA (2 eq.), EtOAc, for 1i 52%; (h) preparatory HPLC to separate diastereomers, 9%; (i) 5 N aq. HCl,
hydroxylamine, 75 °C, 49%; (j) 1k, EDC, HOBt, TEA, DMF, 50%; (k) 4 N HCl, cysteine, dioxane, 75 °C, 81%; (l) 1j, DIPEA, DMF, 37–42%; (m) 1 N NaOH, MeOH,
55 °C, 92%.
bromide which gave trifluoroethylamine 2b as a mixture of diaster-
eomers. Pyridine 2b was treated with p-toluenesulfonic acid to form the
pyridinium ion and then cyclized via RCM using Grubbs II catalyst to
give, following separation of the diastereomers by normal phase
chromatography, pyridine macrocycle 2d. Hydrogenation of 2d pro-
vided saturated macrocycle 2f. The Boc group in 2f was removed with
TFA, and the corresponding amine was coupled with 1j to give 22. Ester
24 and acid 25 analogs were prepared via a slightly modified sequence.
Scheme 2. Reagents and conditions: (a) 1-ethoxy-2,2,2-trifluoroethanol, pTsOH, EtOH, 120 °C, 54%; (b) allyl magnesium bromide, THF, −20 °C, 80%; (c) ethyl 2-
oxoacetate, maleic acid, then allyltributyltin, ACN, 22%; (d) pTsOH, DCM, reflux, then 40 mol% Grubbs II catalyst, for 2d, 18% after separation of diastereomers by
normal phase chromatography, for 2e, 60% as a mixture of diastereomers; (e) H2, 10% Pd/C, TFA, EtOH, for 2f 80%, for 2g 20% after separation of diastereomers by
normal phase chromatography; (f) 30% TFA in DCM, 87%; (g) 1j, DIPEA, DMF, 65–83%; (h) 4 N HCl in dioxane, 100%; (i) 1 N NaOH, MeOH, 55 °C, 100%.
6

T. Fang, et al.
Bioorganic&MedicinalChemistryLettersxxx(xxxx)xxxx
Condensation of 2a with ethyl 2-oxoacetate, followed by addition of
allyl tributyltin afforded amino ester 2c as a mixture of diastereomers.
Amino ester 2c was subjected to RCM conditions which gave un-
saturated macrocycle 2e. The alkene in 2e was reduced as previously
described to give, following separation of diastereomers by normal
phase chromatography, saturated macrocycle 2g. The Boc group in 2g
was removed with 4 N HCl in dioxane, and the corresponding amine
was coupled with 1j to give 24. Compound 2g was converted to acid 25
via a three step sequence as was previously described for acid 17.
In conclusion, our efforts to improve the oral bioavailability of the
macrocyclic FXIa series focused on modifying the P2 prime, macro-
cyclic amide linker, and imidazole regions. Removing the methyl car-
bamate P2 prime resulted in a significant loss in FXIa affinity and did
not improve the oral bioavailability. The oral bioavailability of the
series was significantly improved by exploring SAR in the macrocyclic
linker and imidazole regions. Specifically, replacing the amide moiety
in the macrocyclic linker with amide isosteres led to the discovery of
the imidazole macrocycle with the amine linker which improved the
oral exposure in rat (10, 15 and 16). Finally, replacing the imidazole
scaffold with the pyridine ring further improved oral bioavailability
and afforded compound 24, a single-digit nanomolar FXIa inhibitors
with excellent selectivity against relevant blood coagulation enzymes.
Further improvements in potency for the pyridine-based macrocycles
will be described in subsequent work.
21; X-ray crystal structure of compound 24; Crystallographic data and
refinement statistics for X-ray structures of the FXIa complexes with
compounds 10, 15, 16, and 24) to this article can be found online at
https://doi.org/10.1016/j.bmcl.2020.126949.
References
1. GBD Collaborators. Lancet. 2018;392:1736.
2. Yeh CH, Hogg K, Weitz JI. Arterioscler Thromb Vasc Biol. 2015;35:1056.
3. Cohen AT, Harrington R, Goldhaber SZ, et al. Am Heart J. 2014;167:335.
4. Connolly DJ, Ezekowitz MD, Yusuf S, et al. N Engl J Med. 2009;361:1139.
5. Bane CE, Gailani D. Drug Discovery Today. 2014;19(9):1454.
6. (a) Lin J, Deng H, Jin L, et al. J Med Chem. 2006;49:7781
(b) Schumacher WA, Seiler SE, Steinbacher TE, et al. Eur J Pharmacol.
2007;570:167
(c) Wong PC, Crain EJ, Watson CA, Schumacher WAJ. Thromb Thrombolysis.
2011;32:129.
7. (a) Quan ML, Wong PC, Wang C, et al. J Med Chem. 2014;57:955
(b) Wong PC, Quan ML, Watson CA, et al. Thromb Thrombolysis. 2015;40:416
(c) Hu Z, Wong PC, Gilligan PJ, et al. ACS Med Chem Lett. 2015;6:590
(d) Wang X, Kurowski S, Wu W, et al. J Pharmacol Exp Ther. 2017;360:476
(e) Pinto DJP, Orwat MJ, Smith LM, et al. J Med Chem. 2017;60:9703.
8. (a) Buchanan MS, Carroll AR, Wessling D, et al. J Med Chem. 2008;51:3583
(b) Deng H, Bannister TD, Jin L, et al. Bioorg Med Chem Lett. 2006;16:3049.
9. (a) Hangeland JJ, Friends TJ, Rossi KA, et al. J Med Chem. 2014;57:9915
(b) Pinto DJ, Smallheer JM, Corte JR, et al. Bioorg Med Chem Lett. 2015;25:1635.
10. (a) Corte JR, Fang T, Hangeland JJ, et al. Bioorg Med Chem Lett. 2015;25:925
(b) Corte JR, Fang T, Pinto DJP, et al. Bioorg Med Chem. 2016;24:2257.
11. (a) Al-Horani RA, Gailani D, Desai UR. Thromb Res. 2015:379
(b) Argade MD, Mehta AY, Sarkar A, Desai UR. J Med Chem. 2014;57:3559
(c) Al-Horani RA, Ponnusamy P, Mehta AY, Gailani D, Desai UR. J Med Chem.
2013;56:867.
Declaration of Competing Interest
12. Corte JR, Fang T, Osuna H, et al. J Med Chem. 2017;60:1060.
13. Ki values were obtained from purified human enzyme at 37 °C and were averaged
from multiple determinations (n ≥ 2), as described in Refs. 7a,12.
14. aPTT (activated partial thromboplastin time) in vitro clotting assay was performed in
human plasma as described in Refs. 7a,12. The reported EC1.5x values are the FXIa
inhibitor plasma concentrations which produce a 50% increase in the clotting time
relative to the clotting time in the absence of the inhibitor.
The authors declare that they have no known competing financial
interests or personal relationships that could have appeared to influ-
ence the work reported in this paper.
Acknowledgements
15. Ertl P. Polar surface area. In: Mannhold R, ed. Molecular Drug Properties. Measurement
and Prediction. Weinheim: Wiley-VCH Verlag GmbH & Co. KGaA; 2008:111–126.
16. (a) Corte JR, Yang W, Fang T, et al. Bioorg Med Chem Lett. 2017;27:3833
(b) Clark CG, Rossi KA, Corte JR, et al. Bioorg Med Chem Lett. 2019
(c) Corte JR, Pinto DJP, Fang T, et al. J Med Chem. 2019. https://doi.org/10.1021/
acs.jmedchem.9b01768.
The authors thank William R. Ewing, Mimi L. Quan, and Joanne M.
Smallheer for critical review of this manuscript; the Department of
Discovery Synthesis; Dr. Anuradha Gupta, Dr. Manivel Pitchai, and
Sankara Narayanan for the synthesis of intermediates at BMS Biocon
Research Center; the Department of Lead Discovery & Optimization;
and Mojgan Abousleiman for in vitro work. Some of this research used
resources of the Advanced Photon Source, a U.S. Department of Energy
(DOE) Office of Science User Facility operated for the DOE Office of
Science by Argonne National Laboratory under Contract No. DE-AC02-
06CH11357. Use of the IMCA-CAT beamline 17-ID at the Advanced
Photon Source was supported by the companies of the Industrial
Macromolecular Crystallography Association through a contract with
Hauptman-Woodward Medical Research Institute.
17. Meanwell NA. J Med Chem. 2011;54(8):2529.
18. (a) Black WC, Bayly CI, Davis DE, et al. Bioorg Med Chem Lett. 2005;15:4741
(b) Gauthier JY, Chauret N, Cromlish W, et al. Bioorg Med Chem Lett. 2008;18:923.
19. In addition, the –CH(CF3)-NH-CH backbone angle is close to 120o observed with an
amide, and the C-CF3 bond is isopolar with the C=O. see Ref.18a
.
20. The absolute stereochemistry was assigned based on an X-ray co-crystal of 10 with
FXIa. PDB deposition number for 10 is 5QTV.
21. Dunitz JD, Taylor R. Chem Eur J. 1997;3:89.
22. Replacing the NH with a CH2 removed the H-bond interaction with Leu41.
23. PDB deposition number for 15 is 5QTW and 16 is 5QTX.
24. PDB deposition number for 24 is 5QTY.
25. He K, Qian M, Wong H, et al. J Pharm Sci. 2008;97:2568.
26. The corresponding 13-membered macrocycle 4 displayed a moderate clearance and
showed no exposure.
27. Van Lierop BJ, Lummiss JAM, Fogg DE. Ring-closing metathesis. In: Grela K, ed.
Olefin metathesis: theory and practice. Hoboken, New Jersey: John Wiley & Sons Inc.
; 2014:85–152.
Appendix A. Supplementary data
28. Gong Y, Kato K. J Fluor Chem. 2004;125:767.
Supplementary data (Table S1 listing the human serine protease
selectivity profile; Scheme S1 and S2 describing the synthesis of the
analogs in Table 2. Scheme S3 describing the synthesis of the compound
29. Chen Y, Dias HVR, Lovely CJ. Tetrahedron Lett. 2003;44:1379.
7