Lead optimisation of selective non-zinc binding inhibitors of MMP13. Part 2

Article abstract of DOI:10.1016/j.bmcl.2011.11.034

Directed screening has identified a novel series of non-zinc binding MMP13 inhibitors that possess good levels of activity whilst demonstrating excellent selectivity over related MMPs. A lead optimisation campaign has delivered compounds with enhanced MMP

Full text of DOI:10.1016/j.bmcl.2011.11.034

Bioorganic & Medicinal Chemistry Letters 22 (2012) 271–277
Contents lists available at SciVerse ScienceDirect
Bioorganic & Medicinal Chemistry Letters
journal homepage: www.elsevier.com/locate/bmcl
Lead optimisation of selective non-zinc binding inhibitors of MMP13. Part 2
Chris De Savi a,b, , Andrew D. Morley a, Ian Nash a, Galith Karoutchi a,b, Ken Page a, Attilla Ting a,b
,
⇑
Stefan Gerhardt c
a Respiratory and Inflammation Research Area, AstraZeneca, Mereside, Alderley Park, Macclesfield SK10 4TG, UK
b Oncology Innovative Medicines, AstraZeneca, Mereside, Alderley Park, Macclesfield SK10 4TG, UK
c Global Structural Chemistry, AstraZeneca, Mereside, Alderley Park, Macclesfield SK10 4TG, UK
a r t i c l e i n f o
a b s t r a c t
Article history:
Directed screening has identified a novel series of non-zinc binding MMP13 inhibitors that possess good
levels of activity whilst demonstrating excellent selectivity over related MMPs. A lead optimisation cam-
paign has delivered compounds with enhanced MMP13 potency, good selectivity and acceptable bio-
availability profiles leading to a predicted twice-a-day dosing regimen in man.
Ó 2011 Elsevier Ltd. All rights reserved.
Received 4 October 2011
Revised 4 November 2011
Accepted 6 November 2011
Available online 16 November 2011
Keywords:
MMP13
Selectivity
Non-zinc binder
Lead Optimisation
Osteoarthritis (OA) remains the most common form of arthritis
and is the leading cause of disability in elderly adults both in the Uni-
ted States and the United Kingdom. Treatment options are limited to
steroidal and non-steroidal anti-inflammatory drugs (NSAIDS) and
although providing symptomatic relief for pain and inflammation
in arthritis they have failed to modify the progression of the disease.
There is still a massive unmet medical need to develop disease mod-
ifying agents that reduce or reverse the cartilage destruction associ-
ated with OA. Cartilage is primarily constituted of chondrocytes and
an extracellular matrix that consists of aggrecan, collagen and
water.1 Collagen primarily renders tensile strength whilst aggrecan
is responsible for providing the elasticity and compressibility asso-
ciated with cartilage. Collagenase-3 (MMP13) has been shown to
be the key enzyme in the cleavage of type II collagen and believed
to play a pivotal role in the breakdown of cartilage in osteoarthritic
joints. In a previous communication we reported the identification
of a novel class of selective MMP13 inhibitors which bound to the
enzyme without interacting with the active site zinc atom.2
The profile and SAR of this series, exemplified by compound 1,
provided confidence to progress into lead optimisation and here
we describe the work undertaken during this phase to identify ana-
logues suitable for nomination as development candidates.
Although 1 possessed an encouraging profile, the focus was on
improving MMP13 potency, selectivity over MMP2 and further low-
ering in vitro metabolism of this series. SAR from previous studies2
and structural biology information suggested the (S)-3-alkynyl-3-
hydroxyquinuclidine motif was required for the observed potency
and selectivity profile in this series. Consequently it was decided
to keep this portion of the molecule fixed and concentrate primarily
on modifications to the benzamide motif to improve profiles
further.
Synthetically, replacement of the pyrimidinylmethylamide
motif was easily achieved, so SAR could be generated rapidly.
Key data is shown in Table 1.
A wide variety of amide groups were explored including aro-
matic, heteroaromatic, cyclic and acyclic variants. Pyridines (2c
and 2k) showed good selectivity ratios versus MMP2 (>300-fold)
whereas directly linked pyridine (2e) was far less potent in
MMP13 leading to a poor MMP2 selectivity ratio. A wider explo-
ration of meta substituents on the aryl group was undertaken and
N
O
an emerging correlation of MMP13 pIC50 versus
p
(pi)4 of this
Compound 1
OH
N
H
substituent was seen (Fig. 1). 3,4-Dimethoxybenzyl substituted
compound (2l) showed very good MMP13 potency (pIC50 8.0),
excellent selectivity versus MMP14 (>1000-fold) but poorer
MMP2 selectivity. Methylation of the amide nitrogen generally
led to MMP13 potency decreases (2a compared with 2b).
Saturated cyclohexylmethyl amide (2q) was less potent than its
N
N
O
⇑
Corresponding author. Tel.: +44 1625 513873.
E-mail address: chris.desavi2@astrazeneca.com (C. De Savi).
0960-894X/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2011.11.034

272
C. De Savi et al. / Bioorg. Med. Chem. Lett. 22 (2012) 271–277
Table 1
MMP13, MMP2 potency and in vitro human Clint data of amide analogues of compound 1
N
O
R
2
OH
O
a
a
Compd
R
MMP13 pIC50
7.1
MMP2 pIC50
5.2
clogP
Hu Micsb Clint
10.2
N
H
2a
2.8
N
N
N
2b
2c
6.6
7.9
<5.0
5.4
2.8
3.1
Me
N
H
18.8
26.1
N
2d
7.7
5.7
4.2
N
H
N
N
2e
2f
6.5
7.2
<5.0
5.1
3.5
4.3
<2.0
<2.0
N
H
N
H
N
H
2g
2h
7.5
7.1
5.1
2.7
4.3
12.5
10.9
O
S O
N
H
<5.0
O
S
N
H
2i
2j
7.2
7.2
5.1
3.4
3.8
8.6
9.9
O
O
O
N
H
<5.0
N
N
N
H
2k
2l
7.5
8.0
<5.0
5.9
3.3
4.0
5.6
2.7
O
O
O
N
H
O
N
H
2m
2n
7.3
7.4
5.1
5.2
4.3
4.3
4.8
O
O
16.1
N
H
O

C. De Savi et al. / Bioorg. Med. Chem. Lett. 22 (2012) 271–277
273
Table 1 (continued)
a
a
Compd
R
MMP13 pIC50
MMP2 pIC50
5.2
clogP
Hu Micsb Clint
N
H
2o
7.0
4.5
F
N
2p
2q
7.1
7.0
5.1
5.3
2.6
5.0
2.7
2.9
O
Me
N
N
H
2r
7.6
7.0
<5.0
<5.0
4.4
3.2
N
H
H
N
2s
2t
H
N
7.3
6.7
5.2
4.7
3.7
2.4
8.9
3.9
H
N
2u
Me
a
For general assay procedures see Ref. 2; values mean of n = 2 experiments.
b
Human microsome metabolism intrinsic clearance (l
L/min/mg).3
Figure 2. Rat Heps Clint versus measured logD (R2 = 0.72, N = 20).
Figure 1. MMP13 pIC50 versus pie of meta position (R3).
benzyl analogue (2f) whereas the directly linked cyclohexyl group
(2r) gave a modest improvement in MMP13 potency and good
selectivity versus MMP2. The most lipophilic ligand efficient
compounds (MMP13 pIC50–clogP)5 were 2g, 2p and 2c (LLE 4.8,
4.6 and 4.8, respectively). Furthermore simple methyl amide 2u
was also ligand efficient. Substitution on the benzylic position
generally led to significant increases in potency (pyridyl amide
2a versus methyl substituted pyridyl amide 2c, pIC50 7.1 and
7.9, respectively). Generally, the human microsomal clearance
for the quinuclidine amide series was low and, while in vitro
rat hepatocyte clearance was also low, it correlated strongly with
measured logD ( Fig. 2, R2 = 0.72, RMSE = 1.98, N = 20). This
correlation provided a crude, yet simple ranking tool to advance
compounds for further in vivo profiling in the lead optimisation
campaign.
In vivo profiling of historical compound 1 showed that although
its iv rat PK profile was encouraging, oral dosing gave poor bio-
availability in rat (F = 2%). This increased non-proportionally with
dose suggesting saturation of processes limiting oral exposure.
Caco-2 monolayer experiments showed compound 1 to have an
A to B Papp value of 6.7 and a B to A Papp value of 33, hinting at po-
tential efflux. Furthermore, heterocyclic amide 2p also showed
poor absorption in rat (F = 3%; A to B Papp = 8.2; B to A Papp = 53)
hinting at potential efflux issues for the entire series. Interestingly,
truncating the molecule right back to the parent N-methylamide,
2u, resulted in a compound with good bioavailability in rat and im-
proved Caco-2 permeability (Table 2).
Although potency was reduced, the overall MMP selectivity pro-
file was unaffected and ligand efficiency was improved compared
to 1.

274
C. De Savi et al. / Bioorg. Med. Chem. Lett. 22 (2012) 271–277
Table 2
In vitro and in vivo pharmacokinetic profile of 2u vs compound 1
Parameter
2u
1
a
MMP13 pIC50
6.7
7.0
a
MMP2,9,12,14 pIC50
clogP/logD
Mol wt
4.7/<4/<4/<4
2.4/2
376.5
>1600
24/26
3.9/2.7
<5 (4/4)
1.9/1.8
454.5
810
—/23
Solubilityc
l
M
lM
Hu/Rat Prot binding (%free)
Hu Micsb/Rat Heps Clint
<2/3.2
d
e
Cyp inhib pIC50
Rat PK Cl/Vdss/t
<5 (5/5)
<5 (4/5); 5.1 (2C19)
17/2.4/2.4
f
16/2.4/1.9
54 (0.7)
23/34
½
Rat F% (Fabs
)
2 (0.03) to 39 (0.51)h
6.7/33
Caco-2g AB/BA Papp
a,b
See footnotes from Table 1.
c
Thermodynamic solubility data.6
Rat Sprague-Dawley hepatocyte metabolism intrinsic clearance(l
L/min/106 cells).7
Inhibition of cytochrome P450 isoforms: 1A2, 2C9, 2C19, 2D6 and 3A4.
d
e
f
Compounds dosed at 2 mg/kg iv, n = 2 animals.
g
Apparent permeability (Papp cm/s  10À6
) in human cancer colon cells at
Figure 3. MMP13 pIC50 versus core substructure (red—amides; yellow—sulfona-
10
lM/pH7.4. AB—apical to basolateral direction; BA—basolateral to apical direc-
mides); lines attaching direct matched pairs.
tion (efflux).
h
Compounds dosed at 2 and 50 mg/kg po, n = 6 animals.
The crystal structure of 1 bound to MMP13 had been solved2
and revealed interactions for both the amide oxygen and NH, how-
ever the SAR study described above suggested the amide substitu-
ent was not contributing to good ligand efficiency, therefore amide
isosteres possessing alternative H-bonding vectors were evaluated
to explore other regions of the MMP13 protein and hopefully result
in MMP13 potency increases (Table 3).
The SAR was broadly as expected with acid (3), alcohol (4) and
tertiary amides (7) all being weaker MMP13 inhibitors than amide
2u. Linking through nitrogen to the biaryl ether (9 and 10) also
shows reduced potency, presumably from an inability to form
the key H-bonding interactions with the MMP13 protein.2 Sulfur
linkers, however, show good MMP13 inhibitory profiles, with sul-
fonamide 13 showing equivalent potency to 1 along with a supe-
rior selectivity profile. Unfortunately, attempts to expand the SAR
of both 12 and 13 were unsuccessful with the majority of matched
pair analogues prepared in the sulfonamide series showing signif-
icant reductions in MMP13 potency versus the corresponding
amides ( Fig. 3). The only matched pair compound showing
increased potency in the sulfonamides series was compound 13,
the simple methyl sulfonamide which was moderately more po-
tent than methyl amide 2u.
Attention was next turned back to the truncated alkyl and sat-
urated cyclic amide series, as they appeared to have broader scope
to improve potency and overall profiles than the benzyl amide ana-
logues (reduced clogP and molecular weight and increased LLE).
One key SAR observation from the benzyl amide series was appar-
ent: hydrophilic meta substituents led to favourable MMP13 po-
tency profiles.
A hypothesis was generated to incorporate
hydrophilic substituents into both the 3 and 4 positions of cyclo-
hexyl amide 2r, which should allow access to protein areas that
are hydrophilic based on modelling.8 This also provided an oppor-
tunity to explore broader SAR, whilst reducing overall clogP of the
series (Table 4).
The structurally simple achiral piperidine derivatives all main-
tained modest MMP13 potency despite varied substitution on
nitrogen. N-Methyl piperidine was a particularly soluble com-
pound compared to the majority of aryl amide compounds pre-
pared and also had low in vitro clearance in both human
microsomes and rat hepatocytes. Interesting SAR was also noted
around the regiochemistry of cyclic sulfones with modest increases
in MMP13 potency seen for sulfonepyran-3yl 21 (isomer 1) and 22
(isomer 2) versus the achiral sulfonepyran-4yl 20. Despite exten-
sive small molecule crystallisation efforts we have been unable
to unequivocally assign the absolute stereochemistry associated
with the sulfonepyranyl substituents nonetheless the two pure
diastereomers 21 and 22 were separated by chiral HPLC.9 The rings
of the sulfone diastereomers 21/22 are in very close proximity to
the hydrophobic leucine 184 of MMP13 suggesting this could be
picking up van der Waals interaction with the protein and there
is a water molecule within H-bonding distance to the sulfone oxy-
gen in the in-house crystal structure10 which may also contribute
to the increase in potency observed (see Fig. 4). The majority of
compounds were also inactive against MMP9, MMP12 and
MMP14. Compound 22 had the best overall cross-MMP selectivity
profile of all the compounds made in the project. The most lipo-
philic ligand efficient compounds were compounds 16, 20 and 22
(LLE 6.4, 6.4 and 6.2, respectively). A selected number of com-
pounds were profiled in vivo and data is presented for compounds
14, 21 and 22. Interestingly, significant pharmacokinetic differ-
ences were seen between the chirally pure diastereomers 21 and
22. Compound 21 showed low clearance in rat versus its stereoiso-
mer 22 (Clp = 17 and 51 mL/min/kg, respectively). Oral
Table 3
Amide replacements of compound 2u
N
R
OH
O
a
Compd
R
MMP13 pIC50
clogP
LLEb
2u
3
4
5
6
7
8
9
10
11
12
13
CONHMe
CO2H
CH2OH
CO2Et
CONH2
CONMe2
CONHOMe
NHCOMe
NHSO2Me
SOMe
6.7
5.3
<5.0
6.0
6.2
5.8
5.6
<5.0
5.6
6.4
6.5
7.0
2.4
1.1
2.6
4.2
2.2
2.1
2.5
2.7
2.5
2.1
2.0
2.4
4.3
4.2
-
1.8
4
3.7
3.1
-
3.1
4.3
4.5
4.6
SO2Me
SO2NHMe
a
See footnotes from Table 1.
LLE (Lipophilic ligand efficiency)5 values are calculated as pIC50-clogP.
b

C. De Savi et al. / Bioorg. Med. Chem. Lett. 22 (2012) 271–277
275
Table 4
Hydrophilic saturated amide analogues of compound 2r
N
O
OH
R
O
Solubilityc
M)
b
d
Compds
R
MMP13
pIC50
MMP2
pIC50
MMP9
pIC50
MMP12
pIC50
MMP14
pIC50
clogP Hu Clint
L/min/mg)
Rat Clint
L/min/106
a
a
a
a
a
(
l
(
l
(l
cells)
N
14
15
7.1
7.3
5.1
5.1
<5.0
<5.0
5.1
<5.0
<5.0
2.42
1.62
2.6
<2
>2000
91
N
O
S
N
<5.0
2.9
2.5
3.7
<2
<2
O
N
N
O
N
16
17
7.4
7.6
5.3
5.2
<5.0
<5.0
<5.0
<5.0
<5.0
<5.0
1.02
1.61
O
N
N
O
N
O
N
18
19
7.3
7.1
5.4
5.1
<5.0
<5.0
<5.0
<5.0
<5.0
<5.0
1.82 <2.0
N
N
2.75
0.62
4.7
2.5
N
S
O
O
20
21
7.0
7.4
<5.0
5.3
<5.0
<5.0
<5.0
<5.0
<5.0
<5.0
N
1
2
1.75 <2.0
2.9
160
S
N
N
O
O
O
22
23
8.0
7.1
5.6
5.1
<5.0
<5.0
<5.0
<5.0
<5.0
<5.0
1.75 <2.0
1.20 <2.0
3.4
<2
120
160
S
O
S
N
O
a–d
See footnotes from Tables 1 and 2.
bioavailability for 21 was good whereas no bioavailability was seen
for compound 22 (rat F = 35% and 0%, respectively). The difference
in clearance for these two stereoisomers could not be explained by
their in vitro clearance profiles alone, since both had low in vitro
Clint values in rat hepatocytes.
In contrast, methyl piperazine 14 showed moderate clearance
in rat (Clp = 16 mL/min/kg), good volume of distribution
(Vdss = 12.6 L/kg) leading to a high half-life of 10 h. Unfortunately
bioavailability in rat was very low (F = 2%). However, due to the
excellent iv half-life and good solubility associated with compound
14, it was advanced further in vivo, along with compound 21 into
dog pharmacokinetic experiments (Table 5).
This observation consequently built confidence in our ability to
predict human clearance accurately from similar data. Values for
Vdss in both rat and dog were moderate to high, as might be ex-
pected for moderately basic compounds, meaning that human Vdss
was likely to be similar and that minimisation of human clearance
was less likely to be an issue. Previous testing of MMP13 inhibitors
in a Guinea pig PD model of cartilage degradation had shown that
P3-fold free MMP13 enzyme potency was necessary at Cmin to en-
sure clinical efficacy.11 Consequently the human pharmacokinetic
half-life (t ) was a key parameter to achieve clinical efficacy. Met-
½
abolic clearance for both compounds, as predicted from in vitro
microsomal/hepatocyte data, was moderate to low. Given this,
despite the high expected human Vdss, any non-metabolic clear-
ance (notably renal excretion) could reduce half-life unacceptably.
In vivo clearance in both rat and dog could be predicted well for
both compounds from in vitro hepatocyte and microsomal data.

276
C. De Savi et al. / Bioorg. Med. Chem. Lett. 22 (2012) 271–277
Figure 4. Structure of diastereomers 21/22 bound to MMP13.10 H-bonds between ligand, MMP13 and the water molecule are indicated by yellow dotted lines and distance
(in Å) between the water molecule and acetohydroxamic acid (in buffer) by purple dotted line.
Table 5
Profiles of advanced compounds 14 and 21
N
N
N
O
O
1
O
OH
OH
S
N
N
O
O
O
Parameter
14
21
a
MMP13 IC50
0.074
l
M
0.037 lM
>10/5.1/>10/>10/>10/>10/>10 lM
>100/>10/>10
1.4
a
MMP 1,2,3,9,12,14,19 IC50
ADAM-TS 1,4,5a
LogD
>10/7.3/>10/>10/>8/>10/>10
>100/>10/>10
1
lM
lM
lM
Solubilityc
(
l
M)
>2000
160
Hu/Rat/Dog % free
41/73/47
2.6/<2
25/27/33
<2/<2
h
Hu Micsb/Heps Clint
e
Cyp inhib pIC50
>10
l
M (5/5)
>10
>100
116
lM (5/5)
lM
lM
hERG IC50
>100
lM
PLDi EC50
8
l
M
Caco-2 (@10
l
M)g
0.3 Â 10À6 AB/7.4 Â 10À6 BA
1.3 Â 10À6 AB/43.5 Â 10À6 BA
17 mL/min/kg/4.8 h/5.9 L/kg
13 mL/min/kg/5.0 h/3.9 L/kg
35/21 (46/28)
Rat PKf Cl/t /Vdss
16 mL/min/kg/10.1 h/12.6 L/kg
3.6 mL/min/kg/23 h/6.6 L/kg
2/47 (3/77)
½
Dog PKj Cl/t /Vdss
½
Rat/Dog F% (Fabs
Human pPK Cl/t /Vdss
)
2.2 mL/min/kg/43 h/7.8 L/kg
133 b.i.d
0.290
4.5 mL/min/kg/11 h/3.9 L/kg
217 b.i.d
0.200
½
Human pDTM Dose (mg/day)
Human pFree Cmax M)
(l
a–g
See footnotes from Table 2.
h
i
Human hepatocyte metabolism intrinsic clearance Clint (l
L/min/106 cells). Clint reported is mean of six separate experiments.
12
Phospholipidposis EC50
(l
mol/L). See Ref.
.
j
Compounds dosed at 1 mg/kg iv, n = 4 animals and 1 mg/kg po, n = 4 animals.
Elimination of both compounds 14 and 21 via this route was tested
in a renal excretion study in the rat as a surrogate for humans and
very little, if any, renal elimination (<1% of total clearance) was
observed. Consequently, it was assumed that renal elimination in
humans would also be negligible; the human t data quoted here
½
is the result of predicted human Vdss based on PBPK modelling
using both rat and dog data plus metabolic clearance predicted
from human hepatic microsomal and hepatocyte data. Finally,

C. De Savi et al. / Bioorg. Med. Chem. Lett. 22 (2012) 271–277
277
bioavailability and therefore fraction absorbed (Fabs) was moderate
to poor for compound 21 (Fabs = 28% and 46% in dog and rat,
respectively) but relatively consistent, whereas Fabs was acceptable
for compound 14 in dog (50%), but very poor in rat.
Hargreaves (AZ) for crystallisation. And final thanks to John G.
Cumming (AZ) for technical suggestions and proof-reading the
manuscript. We would like to thank both Paul Turner and Andy
Stanier for compound synthesis.
For human dose predictions, the average of the two values
across rat and dog was used for each compound (37% and 26%
for compounds 21 and 14, respectively). Clearly, human absorption
is the most significant risk for compound 14 with a possibility that
it will be as bad as that observed in rat. For compound 21 the mag-
nitude of the likely efficacious dose itself is the biggest risk factor.
However, this is driven principally by a clearance prediction based
on an undetectable metabolic rate in human hepatocytes; an assay
better able to quantify very low Clint values is likely to cause the
dose estimate to be adjusted to a lower level.
As the combination of free fraction, potency and human half-life
was not quite acceptable for once daily administration for either
compound, the use of a BID dosing regimen is expected to mini-
mise total daily dose and free Cmax, thereby reducing the risk of
off-target toxicity.
Supplementary data
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.bmcl.2011.11.034.
References and notes
1. Kiani, C.; Chen, L.; Wu, Y. J.; Yee, A. J.; Yang, B. B. Cell Res. 2002, 12, 19.
2. De Savi, C.; Morley, A. D.; Ting, A.; Nash, I.; Karabelas, K.; Wood, C. M.; James,
M.; Norris, S. J.; Karoutchi, G.; Rankine, N.; Hamlin, G.; MacFaul, P. A.; Ryan, D.;
Baker, S. V.; Hargreaves, D.; Gerhardt, S. Bioorg. Med. Chem. Lett. 2011, 21, 4215.
3. McGinnity, D. F.; Parker, A. J.; Soars, M.; Riley, R. J. Drug Metab. Dispos. 2000, 28,
1327.
4. Hansch, C.; Muir, R. M.; Fujita, T.; Maloney, P.; Geiger, E.; Streich, M. J. Am.
Chem. Soc. 1963, 85, 2817.
5. Leeson, P. D.; Springthorpe, B. Nat. Rev. Drug Disc. 2007, 6, 881.
6. Buttar, D.; Colclough, N.; Gerhardt, S.; MacFaul, P. A.; Phillips, S. D.; Plowright,
A.; Whittamore, P.; Tam, K.; Maskos, K.; Steinbacher, S.; Steuber, H. Bioorg. Med.
Chem. 2010, 18, 7486.
Compounds 14 and 21 show excellent selectivity over other
MMPs and ADAMTS enzymes tested, good to excellent solubility,
low in vitro clearance in both human microsomes and hepatocytes,
no hERG liability; they, are free from Cyp inhibition across five Cyp
isoforms and have excellent free levels. Due to the basic nature of
the quinuclidine moiety both compounds were tested in our in-
house phospholipidosis assay.12 Although both compounds are
weak inhibitors, compound 21 has a far superior profile with an
7. Soars, M. G.; Grime, K.; Sproston, J. L.; Webborn, P. J. H.; Riley, R. J. Drug Metab.
Dispos. 2007, 35, 859.
8. The sitemap for MMP13 was generated based on MMP13 crystal structure (PDB
ID: 2YIG) using the Maestro molecular modelling program. The model of 21
and 22 were built based on the crystal structure conformation of 1, then
optimised in Macromodel and overlaid into the MMP13 structure and site map
for hypothesis generation. Macromodel and Maestro were licensed from
Schrödinger, LGG (www.schrodinger.com).
9. Diastereomers 21 and 22 were separated by chiral HPLC. Full experimental for
the preparation and analytical data for both compounds 21 and 22 can be
found in Supplementary data.
10. An in house crystal structure of diastereoisomers 21/22 in MMP13 (2.2 Å) was
generated and deposited into the RCSB Protein data bank (PDB ID: 4a7b)
(http://www.pdb.org/pdb/home/home.do). It has a very similar binding mode
as 1 (PDB ID: 2YIG) for the common core of the compounds, with the same H-
bond interactions around the amide and hydroxyquinuclidine with the
proteins. The electron density around the sulfonepyran-3yl is less well
defined than the rest of ligand which indicate flexibility or movement
around this group. Acetohydroxamic acid (added as part of the protein work
up for crystallisation) was chelated to the zinc in the crystal structure.
11. Unpublished results.
EC50 >100 l
M. These could represent excellent candidates13 to test
the clinical hypothesis that non-zinc binding selective MMP13
inhibitors could protect cartilage from type-II collagen degradation
and be free from musculoskeletal effects seen for other Zn-binding
MMP13 inhibitors that have faltered in the clinic.
Acknowledgements
We would like to thank Linette Ruston (AstraZeneca), David
Berry (Medicines Evaluation, AZ) and Mark E. Light (University of
Southampton) for growing and solving small molecule X-ray crys-
tal structures for representative compounds. We would also like to
thank Liz Flavell (AZ) who produced the MMP13 protein and David
12. Morelli, J. K.; Buehrle, M.; Pognan, F.; Barone, L. R.; Fieles, W.; Ciaccio, P. J. Cell
Biol. Toxicol. 2006, 22, 15.
13. Due to internal disease area strategy re-prioritisation, no further pre-clinical/
clinical profiling of compounds 14 and 21 was undertaken.