Identification of potent and selective hydantoin inhibitors of aggrecanase-1 and aggrecanase-2 that are efficacious in both chemical and surgical models of osteoarthritis

Article abstract of DOI:10.1021/jm501522n

A disintegrin and metalloproteinase with thrombospondin motifs-4 (ADAMTS-4) and ADAMTS-5 are zinc metalloproteases commonly referred to as aggrecanase-1 and aggrecanase-2, respectively. These enzymes are involved in the degradation of aggrecan, a key component of cartilage. Inhibitors of these enzymes could be potential osteoarthritis (OA) therapies. A series of hydantoin inhibitors of ADAMTS-4 and ADAMTS-5 were identified from a screening campaign and optimized through structure-based drug design to give hydantoin 13. Hydantoin 13 had excellent selectivity over other zinc metalloproteases such as TACE, MMP2, MMP3, MMP13, and MMP14. The compound also produced efficacy in both a chemically induced and surgical model of OA in rats.

Full text of DOI:10.1021/jm501522n

Evaluation Only. Created with Aspose.PDF. Copyright 2002-2021 Aspose Pty Ltd.
Article
pubs.acs.org/jmc
Identification of Potent and Selective Hydantoin Inhibitors of
Aggrecanase‑1 and Aggrecanase‑2 That Are Efficacious in Both
Chemical and Surgical Models of Osteoarthritis
Timothy B. Durham,* Valentine J. Klimkowski, Christopher J. Rito, Jothirajah Marimuthu, James L. Toth,
Chin Liu, Jim D. Durbin, Stephanie L. Stout, Lisa Adams, Craig Swearingen, Chaohua Lin,
Mark G. Chambers, Kannan Thirunavukkarasu, and Michael R. Wiley
Eli Lilly and Company, Lilly Corporate Center, Indianapolis, Indiana 46285, United States
S
* Supporting Information
ABSTRACT: A disintegrin and metalloproteinase with throm-
bospondin motifs-4 (ADAMTS-4) and ADAMTS-5 are zinc
metalloproteases commonly referred to as aggrecanase-1 and
aggrecanase-2, respectively. These enzymes are involved in the
degradation of aggrecan, a key component of cartilage. Inhibitors
of these enzymes could be potential osteoarthritis (OA)
therapies. A series of hydantoin inhibitors of ADAMTS-4 and
ADAMTS-5 were identified from a screening campaign and
optimized through structure-based drug design to give hydantoin
13. Hydantoin 13 had excellent selectivity over other zinc
metalloproteases such as TACE, MMP2, MMP3, MMP13, and
MMP14. The compound also produced efficacy in both a
chemically induced and surgical model of OA in rats.
an extracellular matrix (ECM) protein that forms cross-linked
rigid triple helices. These helices provide strength to the
cartilage. Aggrecan is an extensively glycosylated ECM protein
that provides compressibility. Aggrecan has two globular
structural domains (designated G1 and G2) at the N-terminus
and a third globular domain (G3) at the C-terminus.6 Between
the G2 and G3 domains, the protein is highly glycosylated, with
>100 glycosyl chains attached. The glycolsylated core of the
protein is highly hydrated. It is this high level of hydration that
provides the compressibility to cartilage.
The ECM proteins in cartilage are synthesized by
chondrocytes located within the cartilage matrix. In the normal
process of tissue remodeling, the ECM proteins are later
degraded by a number of proteases. One hypothesis for how
OA progresses is that the rate of ECM synthesis versus
degradation becomes unbalanced when an insult to the system
occurs. Thus, inhibitors of ECM proteases could provide an
effective means to halt or even reverse the progression of OA.
In this context, several proteases could be potential drug targets
due to their propensity to degrade collagen.7 Multiple members
of the matrix metalloprotease (MMP) family meet this criteria.8
MMPs are zinc metalloproteases. There are 28 MMPs known,
all of which have a high level of homology around the catalytic
site. The MMPs have been implicated not only in degradation
of collagen but also in the degradation of a variety other
INTRODUCTION
■
Osteoarthritis (OA) is a disease characterized by the
degradation of the joint cartilage leading to pain and loss of
joint function. The disease is widespread, with estimates
suggesting that in 2008 there were at least 27 million people in
the U.S. with some form of OA.1 Worldwide estimates place the
percentage of the population over the age of 60 with
symptomatic OA as 9.6% for men and 18% for women.2
Studies have found that radiographic OA of the knee in persons
45 years of age and older in the U.S. and Europe is widespread
and affects 14% of men and 23% of women.3 The societal
impact of this disease is significant, leading to a major economic
burden on the world’s healthcare systems and significant
compromises in quality of life.4 Currently, the main course of
treatment for OA patients involves symptom relief through the
use of NSAIDS and, when that becomes no longer effective,
joint replacement surgery. Unfortunately, there is no approved
treatment regimen that can directly affect the progression of
OA. Therefore, a disease-modifying OA drug (DMOAD)
would be a valuable asset to patients and care givers.
The joint contains both cartilage and synovial fluid. Synovial
fluid is a non-Newtonian fluid composed primarily of water,
hyaluronic acid, and the glycosylated peptide lubricin. Synovial
fluid is responsible for the transport of nutrients to and waste
products out of the joint space. This exchange process between
synovial fluid and plasma is believed to be diffusion-controlled.5
Cartilage is nonvascularized tissue composed of two major
components: type II collagen and aggrecan. Type II collagen is
Received: October 5, 2014
© XXXX American Chemical Society
A
dx.doi.org/10.1021/jm501522n | J. Med. Chem. XXXX, XXX, XXX−XXX

Evaluation Only. Created with Aspose.PDF. Copyright 2002-2021 Aspose Pty Ltd.
Journal of Medicinal Chemistry
Article
biomolecules important to disease and normal physiological
processes.9 However, broad spectrum MMP inhibitors have
been unsuccessful in clinical applications for a variety of
indications due to musculoskeletal syndrome (MSS), a
reversible fibromyalgia that causes pain and swelling in the
joints.10
ADAMTS-4 (a disintegrin and metalloproteinase with
thrombospondin motifs-4) and ADAMTS-5 are commonly
referred to as aggrecanase-1 and aggrecanase-2, respectively.
ADAMTS-4 and ADAMTS-5 are zinc metalloproteases that,
unlike the MMPs, are known to have a specific and primary role
in aggrecan degradation. Both aggrecanases cleave the
Glut373−Ala374 peptide bond located between the G1 and
G2 domains.11 This cleavage is the first step in aggrecan
degradation. Since aggrecan is responsible for cartilage’s
compressibility, aggrecan loss would be expected to increase
the mechanical trauma that the joint undergoes during
movement, leading to structural damage and inflammation. In
support of this hypothesis, ADAMTS-5−/− mice show
protection against both chemical and surgical models of
OA.12 Furthermore, mice genetically modified to express a
form of aggrecan that is resistant to cleavage by ADAMTS-5 are
also protected against models of OA compared to WT mice.13
Interestingly, ADAMTS-4−/− mice do not show protection
against OA models, suggesting that, in mice, ADAMTS-5 is the
more relevant protease for aggrecan cleavage.14 However, in
human chrondrocytes, siRNA transfection to decrease ex-
pression of either ADAMTS-4 or ADAMTS-5 resulted in
reduced aggrecan cleavage, suggesting that, in humans, both
enzymes may be important in aggrecan degradation.15
The above data point to ADAMTS-4 and ADAMTS-5 as
being interesting potential targets for DMOAD research.
However, the active sites of ADAMTS-4 and ADAMTS-5
share significant similarity with the MMPs in that all of the
enzymes contain a HExxHxxGxxH zinc-binding motif. Most of
the previously reported inhibitors of ADAMTS-4 and
ADAMTS-5 do not show significant selectivity over the other
MMPs or come from the chemical classes of hydroxamic acids
or sulfonamides that can be challenging platforms for the
development of oral drugs.16 Therefore, our goal was to identify
nonhydroxamate/sulfonamide inhibitors of ADAMTS-4 and
ADAMTS-5 with suitable potency and pharmacokinetics to
determine if chronic and selective inhibition of these enzymes
would result in modification of the degenerative process in OA
disease models similar to the phenotypes observed in knockout
animals.
Figure 1. Inhibitors of ADMTS-4 and ADAMTS-5.
challenging. Therefore, we decided to use a modified enzyme
assay run in the presence of 50% rat plasma as a tool for
assessing the effects of nonspecific protein binding; compound
1 had modest activity in the presence of rat plasma. It also had
low oral exposure in rats, necessitating the need to further
optimize pharmacokinetic performance in rodents.
We were able to co-crystallize compound 1 with hADAMTS-
4 (Figure 2), which allowed us to clearly understand the mode
of ligand binding. The hydantoin ring forms critical polar
interactions with the protein at three of its five positions. The
imide nitrogen of the hydantoin ring interacts directly with the
catalytic zinc. In addition, the hydantoin ring forms four
hydrogen-bonding interactions with structural water, the
carbonyl of Gly331, and the carboxyl of Glu362. The amide
linker also forms hydrogen bonds with Leu330 and Pro393.
The 4-chlorophenoxy ring is directed into the hydrophobic S1′
pocket, whereas the C5 methyl group of the hydantoin is
directed toward the S1 region.
Upon examination of the X-ray co-crystal structure of
hydantoin 1, we anticipated that an improvement in compound
potency could be achieved through fusion of the phenoxy acetyl
group into a benzofuran, thereby limiting the degrees of
conformational freedom in the ligand. We anticipated that this
RESULTS AND DISCUSSION
■
To initiate our research effort, we carried out a biochemical
screen of 80 000 structurally diverse compounds to identify
potential starting points for inhibitor development. From that
screening campaign, we identified hydantoin 1 as an interesting
dual inhibitor (Figure 1). Hydantoin 1 showed modest
inhibition of hADAMTS-4 and hADAMTS-5 enzymes (IC50
of 74 and 350 nM, respectively) as well as modest selectivity
over several zinc metalloproteases (Table 1). Furthermore,
hydantoin 1 and all of its subsequent analogues described
herein had essentially equipotent activity (<2-fold shift) against
rADAMTS-5.
Because the joint space contains large biopolymers (vide
supra), we wanted to understand if nonspecific protein binding
could limit the functional activity of our inhibitors. However,
developing a routine assay in synovial fluid was technically
B
dx.doi.org/10.1021/jm501522n | J. Med. Chem. XXXX, XXX, XXX−XXX

Evaluation Only. Created with Aspose.PDF. Copyright 2002-2021 Aspose Pty Ltd.
Journal of Medicinal Chemistry
Article
a
Table 1. Biochemical Activity and Rat Oral Exposure of ADAMTS-4 and ADAMTS-5 Inhibitors
b
rat plasma
hADAMTS-4 hADAMTS-5
hADAMTS-5 IC50
(nM)
hMMP2
hMMP3
hMMP13
hMMP14
hTACE
2 h plasma 8 h Plasma
c
c
compd
IC50 (nM)
IC50 (nM)
IC50 (nM) IC50 (nM) IC50 (nM) IC50 (nM) IC50 (nM) (ng/mL)
(ng/mL)
1
74
110
12
17
8
350
150
22
19
17
9
620
2400
420
1700
330
110
530
35
24 000
ND
50 000
ND
7000
ND
>50 000
ND
ND
430
50
d
7
ND
ND
ND
ND
ND
1300
80
8
2100
6100
1800
7500
7100
11 000
5
3000
20 000
1900
38 000
64 000
69 000
3
720
3500
5400
ND
<200
ND
d
9
3100
930
18 000
870
10
11
10 000
ND
4300
2200
ND
5
4200
4700
30 000
1
>100 000
8600
d
12
22
4
19
4
ND
ND
90
13
14
86 000
150
10 000
7
1100
1900
1
1
18
840
a
b
All values are relative IC50’s. The IC50 data for ADAMTS-4 and ADAMTS-5 are mean values generated from at least n = 3. Rat plasma IC50’s were
c
generated using hADAMTS-5 under standard assay conditions run in the presence of 50% rat plasma. Plasma levels are mean values generated in n
= 3 rats after a 10 mg/kg PO dose. Compounds were extracted from plasma using acetonitrile and quantified by MS. Data shown is for the
d
racemate. ND, not determined.
π-system. Benzofuran 7 showed improved potency against
ADAMTS-5 with no loss in ADAMTS-4 potency (Figure 1 and
Table 1).
We next sought further potency improvement and hoped
that we could enhance the pharmacokinetic performance of the
molecules through optimization of the P1 substituent. We also
saw the P1 substituent as a promising position to modulate
physical properties such as solubility and polarity to fine-tune
the molecules’ performance characteristics. Synthesis of
analogues of compound 1 was straightforward and began
with introduction of the P1 substituent through addition of
either a Grignard reagent or organolithium species to the
Weinreb amide of Boc-glycine to produce aminoketone 3
(Scheme 1). We next carried out a Bucherer−Burge reaction to
give hydantoin 4.17 The hydantoins were easily resolved via
chiral chromatography to the more active stereo isomer 5.
Deprotection to give amine 6 and subsequent amide formation
produced hydantoins 7−13 (Figure 1).
Figure 2. X-ray co-crystal structure of hydantoin 1 with hADAMTS-4.
Zinc is shown as a blue sphere. Resolution = 1.24 Å.
As the size of the P1 substituent was increased from methyl
(7) to phenyl (8), we saw a 10× improvement in biochemical
potency against ADAMTS-4 and ADAMTS-5 (Table 1).
However, compound 8 had low oral exposure in rats and
suffered from limited selectivity against other metalloproteases.
should also improve the stereoelectronic interactions between
the His361 imidazole ligating the catalytic zinc and the ligand
a
Scheme 1. Synthesis of Hydantoins 7−13
a
Conditions: (a) M-R (2 equiv; M = MgBr, MgCl, or Li), THF, −78 °C to ambient temperature, 20−96%; (b) KCN, (NH4)2CO3, MeOH/water
1:1, heat, 23−85%; (c) chiral chromatography; (d) HCl·dioxane, MeOH, quant.; (e) PyBrop, DIPEA, appropriate benzofuran carboxylix acid, DMF,
25−66% over three steps.
C
dx.doi.org/10.1021/jm501522n | J. Med. Chem. XXXX, XXX, XXX−XXX

Evaluation Only. Created with Aspose.PDF. Copyright 2002-2021 Aspose Pty Ltd.
Journal of Medicinal Chemistry
Article
Figure 3. X-ray co-crystal structures of compounds 10 (left) and 11 (right) with hADAMTS-4. Zinc is shown as a blue sphere. Resolution =1.62 and
1.60 Å, respectively.
Figure 4. Calculated ΔG vs dihedral angle for model hydantoin system. The graph shows ΔG versus dihedral angle for a model system of hydantoins
10 and 11. In the model, the benzofuran was replaced with a simple acetyl side chain. Key conformations are shown on the right, with the
corresponding dihedral angle given below. Global minima are indicated in bold.
Therefore, we set out to further improve our hydantoin
inhibitors by increasing their plasma activity, metalloprotease
selectivity, and oral exposure in rats. We hypothesized that
replacing the phenyl ring of hydantoin 8 with a heterocycle
could improve functional activity by lowering protein binding
as well as limiting oxidative metabolism, thus improving
exposure. To that end, we prepared the series of heterocyclic
replacements for the phenyl ring shown in Table 1
(compounds 9−12).
Replacement of phenyl with a pyridyl substituent to give
hydantoin 9 resulted in gains in metalloprotease selectivity.
Furthermore, replacement of the pyridine with five-membered
heterocycles greatly impacted the pharmacokinetic performance
of the hydantoins. Thiazole 10 had potent activity against both
hADAMTS-4 and hADAMTS-5, with a rat plasma IC50 of 330
nM. It also displayed exceptional oral exposure (Tables 1 and
2). Thiazole 10 has only modest selectivity over the other
metaloproteases shown in Table 1. Therefore, we sought to
improve our selectivity profile while maintaining good oral
exposure properties.
To that end, we next explored imidazole and pyrazole as
heterocyclic P1 groups. Compound 11 provided our most
encouraging results. Compound 11 not only improved plasma
potency relative to that of thiazole 10 but also showed
enhanced selectivity against the other zinc metalloproteases.
Furthermore, we were delighted to see that imidazole 11 still
maintained reasonable oral exposure.
The subtlety of the SAR within this limited set of compounds
for both plasma activity and metalloprotease selectivity is
intriguing. In comparing compound 10 to 11, we see that the
simple change from thiazole to imidazole leads to an increase in
metalloprotease selectivity. This is due to two factors. One is a
very small improvement in potency against ADAMTS-5 going
from compound 10 to 11. The second factor is a loss in
potency against our panel metalloproteases, which in the case
of MMP3 and MMP14 was greater than 10-fold. To better
understand the difference in compound performance, we
obtained X-ray co-crystal structures for compounds 10 and
11 in hADAMTS-4 (Figure 3). The X-ray data reveal that the
only significant change in binding between the two compounds
D
dx.doi.org/10.1021/jm501522n | J. Med. Chem. XXXX, XXX, XXX−XXX