Arylsulphonyl hydroxamic acids: Potent and selective matrix metalloproteinase inhibitors

Article abstract of DOI:10.1016/S0960-894X(01)00259-1

A series of novel matrix metalloproteinase inhibitors is described in which selectivity between MMP and 'sheddase' activity has been achieved and which demonstrate potent in vivo activity in models of arthritis and cancer.

Full text of DOI:10.1016/S0960-894X(01)00259-1

Bioorganic & Medicinal Chemistry Letters 11 (2001) 1465–1468
Arylsulphonyl Hydroxamic Acids: Potent and Selective Matrix
Metalloproteinase Inhibitors
Andrew D. Baxter, Ranjev Bhogal, John Bird, John F. Keily, David T. Manallack,
John G. Montana, David A. Owen, William R. Pitt, Robert J. Watson*
and Ruth E. Wills
Celltech Chiroscience Ltd, Granta Park, Abington, Cambridge CB1 6GS, UK
Received 26 January 2001; accepted 11 April 2001
Abstract—A series of novel matrix metalloproteinase inhibitors is described in which selectivity between MMP and ‘sheddase’
activity has been achieved and which demonstrate potent in vivo activity in models of arthritis and cancer. # 2001 Elsevier Science
Ltd. All rights reserved.
The association between overexpression of matrix
metalloproteinases and diseases such as metastatic
tumours has long been established.1 Within our labora-
tories, development of inhibitors which mimic the nat-
ural enzyme subtrate led to the discovery of
BMS275291 1, a thiol which was found to be a potent
and broad spectrum inhibitor of matrix metalloprotein-
ases with no activity against shedding of ectodomain
proteins.2
Other research groups have postulated that inhibition of
MMP-1 may contribute to side effects6 as the enzyme
plays a crucial role in wound healing and tissue remo-
delling. In the current programme of work, we therefore
elected to design and prepare MMP inhibitors, which
spared MMP1 and ectodomain shedding.
The development of MMP inhibitors has been particu-
larly hampered by side effects of early clinical candi-
dates such as Marimastat 2,1b which causes
musculoskeletal pain and stiffness, limiting the max-
imum tolerable dose and necessitating treatment ‘holi-
days’ in patients. Given that there are both dose
limitations and a lack of efficacy of early MMP inhibi-
tors, there is an urgent need to provide compounds
that can be dosed at clinically useful levels without the
requirement to cut short the treatment. There is con-
siderable debate as to whether side effects arise from
inhibition of specific MMPs, or from the broad spectrum
of activity of early compounds against so-called sheddase
enzymes.1b With the observation that 1 was incapable of
generating tendinitic effects in animal models, in contrast
to compounds such as 2, we felt that sheddase inhibition
may play a part in MMP-inhibitor-related tendinitis.2ꢀ5
To assist in the identification of non-peptidic MMP
inhibitors, we employed computer-aided molecular
design (CAMD) techniques to search databases for
likely lead compounds. The UNITY database search-
ing package within Sybyl (Tripos, St Louis, USA) was
used for this purpose. Using the crystal structure of
*Corresponding author. Tel.:+44-1223-896515; fax:+44-1223-
896400; e-mail: bobwatson@celltechgroup.com
0960-894X/01/$ - see front matter # 2001 Elsevier Science Ltd. All rights reserved.
PII: S0960-894X(01)00259-1

1466
A. D. Baxter et al. / Bioorg. Med. Chem. Lett. 11 (2001) 1465–1468
neutrophil collagenase (MMP-8, PDB code 1mnc) the
location of a carboxylate group as a potential zinc bin-
der was defined. This simply employed the atomic co-
ordinates of three of the atoms of the hydroxamate
group (carbonyl carbon and oxygen atoms, and the
nitrogen atom) onto which the carboxylate moiety was
superimposed. Another atomic position was also
defined corresponding to the ‘first’ carbonyl oxygen
position on the inhibitor that interacts with the main-
chain amide NH group of Leu 181. Distance con-
straints were applied between the two molecular
fragments and the crystal structure of MMP-8 was
used to define regions of steric occlusion. The search
results were examined ‘on screen’ to refine the lists of
compounds to be screened against MMPs 1, 8 and 9
at a concentration of 20 mM. Hits were then followed
up in order to obtain IC50 values against these
enzymes.
Michael addition of 4-mercaptobenzophenone 9 with
the acrylic acid 10 proved capricious under a variety of
conditions, however the derived bromide 11 was found
to couple rapidly and in high yield under very mild
conditions. Oxone oxidation of the crude product
allowed us to isolate the acid 7 in high yield by crystal-
lisation. Conversion of this to the desired hydroxamic
acid was carried out by coupling with TBDMS pro-
tected hydroxylamine, followed by deprotection using
dry HCl to give 8 in good overall yield.
Reduction of 8 using borohydride gave efficient conver-
sion to the alcohol 12, while the use of triethyl silane
gave the diaryl methane 13.
The hydroxamic acid 8 was found to possess extremely
poor aqueous solubility (<0.001 mg/mL) and we there-
fore prepared the analogous pyridophenone 14 by a
similar route to that shown in Scheme 1, achieving a
significant improvement in solubility. The N-oxide 15
was also isolated as some oxidation with oxone was
observed during the synthesis.
The SAR by NMR work of Fesik et al. at Abbott7
addressed a similar issue, finding small molecules which
bound into 15N-labelled stromelysis and optimising the
linkage between them.
After completion of our screening exercise, a number of
modestly potent inhibitors were identified, of which we
selected the ester 3 as a starting point for a medicinal
chemistry lead optimisation exercise.
It was quickly found that simplification of the aromatic
hydrophobe to a benzophenone did not delete inhibi-
tory activity and that a carboxylic acid was equipotent
to the ester. A three-atom linker to the carboxylate gave
a modest increase in potency, however, 4 remained only
weakly active at around 5 mM.
Scheme 1. Reagents: (i) HBr, AcOH; (ii) Et3N, DMF; (iii) Oxone,
MeOH, water; (iv) EDC, H2NOTBS, DCM; (v) HCl, ether, DCM; (vi)
NaBH4, EtOH; (vii) TFA, Et3SiH.
Table 1. Assay results for compounds 1–5 and 9–19 (IC50, nanomo-
lar)
Replacement of the ether with a thioether linker
reduced activity, however the sulphone 5 was accom-
modated.
Compounds
MMP1
MMP2
MMP3
MMP8 MMP9
3
7
8
12
13
14
15
7000
20,000
2,000
25,000
>30,000
10,000
20,000
>10,000
>10,000
1200
100
2500
>2000
100
3000
300
15
3000
300
3
6000
400
5
2500
150
15
Seeking additional binding interactions at the alpha
position, we incorporated the phenpropyl a chain used
to good effect in the Bayer compound 6 to give us the
sulphone 7. We then sought to increase binding affinity
to the enzyme active site zinc atom, by preparation of
the hydroxamic acid 8 as shown in Scheme 1.
800
4
1700
200
3
50
350
80
150

A. D. Baxter et al. / Bioorg. Med. Chem. Lett. 11 (2001) 1465–1468
1467
We were encouraged to find that the pyridophenone 14
retained excellent potency against MMPs 2, 3, 8 and 9,
and a superior selectivity profile with respect to MMP1
when compared to compound 8. The derived N-oxide
was significantly less potent.
It is well known that access to the S10 pocket can lead to
efficient and selective MMP inhibitors and that this
pocket in MMP-1 is truncated by arginine 214.8 MMPs
2, 3, 8 and 9 all have much longer S10 pockets than
MMP-1. To test whether the lower efficacy of 8 in inhi-
biting MMP-1 relative to these other MMPs can be
explained by the shorter S10 of the former enzyme, a
docking study was carried out. To this end, 8(R) was
docked into a crystal structure of MMP-8 (PDB code
1mnc) and of MMP-1 (PDB code 1hfc) using the pro-
gram GOLD (CCDC, Cambridge, UK). In order for
this program to be able to reproduce the correct binding
modes of the similar MMP inhibitors as found in crystal
structures, two distance constraints of 2.2 A are
required to anchor the two hydroxamate oxygens to
the catalytic zinc ion and a hydrogen bond constraint is
needed between a sulphone oxygen and a protein back-
bone NH (leucine 181 in both MMP-1 and 8).
The predicted binding mode in MMP-8 (shown in Fig.
1) puts the arylketone moiety of 8 in the S10 pocket. In
MMP-1, this substituent is predicted to bind along the
prime side substrate binding groove and not in the S10
pocket. These results are consistent with the hypothesis
that 8 has little activity at MMP-1 due to its inability to
bind in the S10 pocket.
Figure 1. Predicted binding mode of 8 in MMP-8.
Having achieved good potency against the enzymes of
interest and good selectivity versus MMP1 with com-
pounds 8 and 14, we compared their activity in a range
of shedding events against Marimastat 2 (Table 2).5
Both of our compounds were inactive in these assays,
while Marimastat 2 gave significant activity at micro-
molar concentrations. We were thus satisfied that our
compounds possessed the desired in vitro inhibitory
profile, sparing MMP1 and sheddases and potent
against the enzymes of interest.
Compounds were initially screened against MMPs 1, 2,
3, 8 and 9 as shown in Table 1, using previously repor-
ted fluorimetric assays.5 The acid 5 was modestly
potent, with selectivity over MMP1, while the addi-
tional substituent in 7 conferred a significant increase in
potency versus MMPs 2, 8 and 9.
As anticipated from its poor aqueous solubility, 8 was
found to have very poor oral bioavailability in the rat,
while 14 gave significant plasma levels, up to 300 ng/mL
following oral dosing at 10 mg/kg.9
Low nanomolar potency was achieved with the hydrox-
amic acid 8 with a selectivity ratio of 500 between
MMPs 1 and 2. Reduction of the ketone had a
significant detrimental effect, with the alcohol 12
being 500-fold less active against MMP9 than the
ketone 8. The diarylmethane 13 had intermediate
potency.
Whittaker and co-workers have reported that only
broad spectrum MMP inhibitors showed efficacy in
their B16 melanoma model and postulated that compe-
titors’ compounds may act as broad spectrum agents at
the doses used in animal models.1b With this in mind,
we were intrigued to investigate whether we could
Table 2. IC50 values (mM) against shedding of ectodomain proteins from stimulated PBMCs
Compound
TNFa
L-Selectin
TNF RII
IL1 RII
IL6-R
2
8
14
2
5
1
10
>100
37
10
>100
>100
>100
>100
>100
>100
>100
>100

1468
A. D. Baxter et al. / Bioorg. Med. Chem. Lett. 11 (2001) 1465–1468
Table 3. In vivo evaluation of test compounds
3. Bird, J. Gordon Research Conference: Matrix Metallopro-
teinases, New London, NH, 1999.
4. Montana, J. G.; Baxter, A. D. Curr. Opin. Drug Discov.
Res. 2000, 3, 353.
Compound
Adjuvant arthritis10
B16-F10 mouse
melanoma11
5. Bird, J.; Montana, J. G.; Wills, R. E.; Baxter, A. D.; Owen,
D. A. PCT Patent Application WO9839024, 1998; Chem
Abstr. 1998, 129, 225791.
Marimastat
8
14
50%@10 mg/kg
40%@10 mg/kg
50%@3 mg/kg
40%@30 mg/kg
40%@30 mg/kg
80%@30 mg/kg
6. Shalinsky, D. R. Curr. Opin. Oncol., Endocr. Metabol.
Investigat. Drugs 1999, 1, 473.
7. Hadjuk, P. J.; Sheppard, G.; Nettesheim, D. G.; Olejnnic-
zak, E. T.; Shuker, S. B.; Meadows, R. P.; Steinman, D. H.;
Carrera, Jr, G. M.; Marcotte, P. A.; Severin, J.; Walter, K.;
Smith, H.; Gubbins, E.; Simmer, R.; Holzman, T. F.; Morgan,
D. W.; Davidsen, S. K.; Summers, J. B.; Fesik, S. W. J. Am.
Chem. Soc. 1997, 119, 5818.
8. Borkakoti, N. In Structure-Based Ligand Design; Guber-
nator, K., Bohm, H.-J., Eds.; Wiley-VCH: Weinheim, Ger-
many, 1998; pp 73–88.
9. Plasma concentration of compound 13 was determined by
addition of a standard solution of 8 (20 mL of a 2.5 mg/mL
solution) to 50 mL of plasma, which was then diluted to
1100 mL. This was loaded onto a C18 SPE cartridge, washed
with water and 30% methanol/water, then eluted with 400 mL
acetonitrile. 100 mL of the eluent was injected onto a 5 mM
C18 Hypersil BDS 100ꢁ4.6 mm column, eluted with aceto-
nitrile/0.1% formic acid into a Turbo Ionspray interface on a
Sciex API300 mass specrometer.
10. Adjuvant arthritis was induced in female Sprague–Dawley
rats by intradermal injection of Mycobacterium butyricum in
mineral oil into the left hind paw. Foot volume was monitored
over days 1–22 and compounds were dosed orally from days
12–22. Percent inhibition was determined at day 21 relative to
untreated control.
achieve in vivo activity with our highly selective com-
pounds.
We were pleased to find that, in in vivo models of
arthritis and cancer, 8 was equipotent to Marimastat,
while 14 gave significantly better inhibition in both
models (Table 3).
In summary, we have identified potent and selective
MMP inhibitors, which have provided us with excellent
leads for optimisation. Further work in this area will be
reported at a later date.
References and Notes
1. (a) Heath, E. I.; Grochow, L. B. Drugs 2000, 59, 1043. (b)
Whittaker, M.; Floyd, C. D.; Brown, P.; Gearing, A. J. H.
Chem. Rev. 1999, 99, 2735. (c) Cheng, M.; De, B.; Almstead,
N. G.; Pikul, S.; Dowty, M. E.; Dietsch, C. R.; Dunaway,
C. M.; Gu, F.; Hsieh, L. C.; Janusz, M. J.; Taiwo, Y. O.;
Natchus, M. G.; Hudlicky, T.; Mandel, M. J. Med. Chem.
1999, 42, 5426.
2. Baxter, A. D.; Bird, J.; Bannister, R.; Bhogal, R.; Manal-
lack, D. T.; Watson R. J.; Owen, D. A.; Montana, J. G.;
Henshilwood, J.; Jackson, R. C. In Cancer Drug Discovery and
Development: MatrixMetalloproteinase Inhibitors in Cancer
Therapy; Clandennin, N. J., Appelt K., Eds.; Humana:
Totowa, NJ, 2000; Vol. 8, pp 193–221.
11. The compounds were evaluated in a B16 lung melanoma
model of metastasis. Briefly, B16-F10 cells were maintained in
culture until required. These were administered intravenously
to C57BL6 mice (day 0) and 14 days later these animals were
killed and the lungs removed. These tissues were fixed and the
number of colonies present on the surface of the lungs was
counted. Test compounds were administered 4 and 2 h prior to
inoculation of the mice with B16-F10 cells and at 1, 6, 24, 48
and 72 h thereafter.