Chelator fragment libraries for targeting metalloproteinases

Article abstract of DOI:10.1002/cmdc.200900516

(Chemical Equation Presented) A chelator fragment library based on a variety of metal binding groups was screened against a metalloproteinase. Lead hits were identified and an expanded library of select compounds was synthesized, resulting in numerous high-affinity hits against several metalloprotein targets. The findings clearly demonstrate that chelators can be used to generate libraries suitable for fragment-based lead design (FBLD) directed at important metalloproteins.

Full text of DOI:10.1002/cmdc.200900516

DOI: 10.1002/cmdc.200900516
Chelator Fragment Libraries for Targeting Metalloproteinases
Arpita Agrawal,[a] Sherida L. Johnson,[b] Jennifer A. Jacobsen,[a] Melissa T. Miller,[a] Li-Hsing Chen,[b]
Maurizio Pellecchia,*[b] and Seth M. Cohen*[a]
Fragment-based lead design (FBLD), sometimes referred to as
fragment-based drug discovery (FBDD), is an increasingly im-
portant strategy for the discovery of biologically active com-
pounds.[1] FBLD generally uses libraries consisting of modest
collections (100–1000 compounds) of small molecular frag-
ments (MW<300 amu) that are screened against targets of in-
terest.[2] Although such fragments do not bind as tightly (Kd
values in the micro- to milli-molar range) as more complex
molecules, including those used in high-throughput screening
(HTS) approaches, they can provide ‘hits‘ that serve as efficient
starting structures for the development of potent inhibitors.
Fragments that bind to a target are identified, after which one
of two approaches is generally pursued: a) a single fragment
can be elaborated in order to obtain a tight binder; or b) mul-
tiple fragments binding at adjacent and distinct sites can be
connected by an appropriate linker to obtain a potent inhibi-
tor. Compared to HTS, FBLD is purported to have several ad-
vantages, including a more efficient exploration of chemically
diverse space and higher ligand efficiencies.
from a chelator fragment library (CFL), showing the usefulness
of these libraries in identifying novel leads for metalloprotein
inhibitors, including inhibitors of matrix metalloproteinases
(MMPs) and anthrax lethal factor (LF).[10]
Matrix metalloproteinases (MMPs) represent one of the most
well-established targets in the realm of metalloproteins. These
zinc(II)-dependent enzymes have been extensively studied, and
the development of MMP inhibitors (MMPi) has played an im-
portant role in the discovery of inhibitors for other zinc(II)-
dependent metalloproteins, such as anthrax lethal factor (LF),
histone deacetylases (HDACs), and others.[10] Indeed, the earli-
est application of FBLD to a metalloprotein target was directed
at MMP-3.[3] Therefore, we focused our preliminary library
screening efforts on MMPs, as a representative system for iden-
tifying fragments that would bind zinc(II) metalloproteins.
Based on widely-reported criteria for fragment libraries,[2] an in-
itial chelator fragment library (CFL-1) was assembled. A modest
library containing 96 structural cores was prepared from chela-
tors with two to four donor atoms for binding metal ions and
sufficient solubility for screening (Figure 1). The chelating
groups included picolinic acids, hydroxyquinolones, pyrimi-
Although the application of FBLD to metalloprotein targets
of medicinal interest has been described,[3] the design, synthe-
sis, and use of general fragment libraries based on
metal chelators for FBLD applications has not been
widely reported.[4–7] This is surprising in light of the
fact that several small-molecule chelators have been
shown to effectively inhibit metalloproteins.[8,9] Fur-
thermore, FBLD using metal-chelating moieties
should be particularly well suited for inhibitor discov-
ery against metalloproteins because: a) chelators
demonstrate binding affinities suitable for FBLD
screening; b) chelators provide a diverse range of
molecular platforms from which to develop lead
compounds; and c) the propensity for chelators to
bind metal ions allows for better prediction of their
probable binding position within a protein active site
in the absence of experimental structural data of the
complex. Herein, we describe the preparation of an
Figure 1. Representative compounds from chelator fragment library 1 (CFL-1). For the full
expanded chelator fragment library (eCFL) derived
list of compounds, see the Supporting Information (figure S1).
[a] A. Agrawal, J. A. Jacobsen, M. T. Miller, Prof. S. M. Cohen
Department of Chemistry and Biochemistry, University of California, San
Diego, 9500 Gilman Drive, La Jolla, CA 92093 (USA)
Fax: (+1)858-822-5598
dines, hydroxypyrones, hydroxypyridinones and salicylic acids,
in addition to other compounds that are well-established com-
ponents of metalloprotein inhibitors, such as hydroxamic acids
and sulfonamides (Figure 1). The CFL-1 library was screened
against MMP-2, at a fragment concentration of 1 mm, using a
fluorescent based assay (see Supporting Information).[11] From
the library, 31 compounds were found to be weak “hits”
(>50% inhibition); this high hit rate is consistent with the fact
that all fragments are known metal chelators. Several of the
hits were known from previous studies,[8,9] thus validating the
E-mail: scohen@ucsd.edu
[b] Dr. S. L. Johnson, L.-H. Chen, Prof. M. Pellecchia
Cancer Research Center and Infectious and Inflammatory Disease Center
Burnham Institute for Medical Research
10901 North Torrey Pines Rd, La Jolla, CA 92037 (USA)
Fax: (+1)858-713-9925
E-mail: mpellecchia@burnham.org
Supporting information for this article is available on the WWW under
http://dx.doi.org/10.1002/cmdc.200900516.
ChemMedChem 2010, 5, 195 – 199
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
195

MED
screening approach as positive controls. Importantly, several
fragments identified have not been widely reported for use in
metalloprotein inhibitors (figure S2, Supporting Information).
Preliminary screening results obtained from CFL-1 prompted
the development of an eCFL-1 geared toward the inhibition of
specific zinc(II) metalloproteins. The components of eCFL-1
were based on one of the chelator hits (3-hydroxy-1,2-dime-
thylpyridine-4(1H)-thione) from CFL-1 against MMP-2. The ex-
panded library was used to demonstrate that hits from the
parent library (CFL-1) could be further elaborated to more ad-
vanced leads, just as in a traditional hit maturation approach.
Derivatives of 3-hydroxy-1,2-dimethylpyridine-4(1H)-thione
(3,4-HOPTO), which consists of an O,S atom donor set, and its
O,O analogue 3-hydroxy-1,2-dimethylpyridine-4(1H)-one (3,4-
HOPO), were used for the generation of eCFL-1 (figure S3, Sup-
porting Information).
these reactions ranged between 3–50%. The 3,4-HOPO frag-
ments outnumber those obtained for 3,4-HOPTO owing to
their ease of isolation, as the 3,4-HOPO fragments generally
precipitated from the reaction mixture upon cooling. In con-
trast, the 3,4-HOPTO reactions resulted in mixtures (as shown
by TLC) and in most cases required flash silica column chroma-
tography to isolate (see Experimental Section for details). Nev-
ertheless, the resulting small fragment library (87 components)
was suitable for screening in a 96-well plate format (see Sup-
porting Information). The fragments in eCFL-1 possessed
chemical properties that are consistent with reported FBLD
libraries (MW between 139 and 353 amu with only 8 fragments
>300 amu; number of H-bond donors and acceptors between
0 and 2).[2]
The eCFL-1 library was screened, using commercially-avail-
able fluorescence-based assays, against four zinc(II)-dependent
metalloprotein targets: gelatinase-1 (MMP-2), gelatinase-2
(MMP-9), stromelysin-1 (MMP-3), and anthrax lethal factor (LF).
All fragments were tested at a concentration of 50 mm; com-
pounds that showed >50% inhibition of metalloenzyme activi-
ty were classified as a hit against that enzyme (Figure 2). Dose
response IC50 values for the resulting hits were then deter-
mined (Table 1).
In order to prepare a reasonable number of 3,4-HOPO and
3,4-HOPTO derivatives, a robust microwave-assisted synthetic
procedure was developed (scheme S1, Supporting Informa-
tion).[12] Using reaction conditions adapted from Orvig et al.,[13]
~2.2 equiv of amine and 1 equiv of hydroxypyrone (or hydroxy-
pyrothione) heterocycle were combined in HCl and irradiated
in a microwave synthesizer, generating the desired products in
<2 min. These conditions were used to produce 63 3,4-HOPO
fragments and 24 3,4-HOPTO fragments. Isolated yields for
The 87-fragment library generated a total of 17 hits against
MMP-2, 18 hits against MMP-9, 5 hits against MMP-3, and 10
Figure 2. Lead hits identified against zinc(II)-dependent metalloenzymes from a expanded library (eCFL-1) of 3,4-HOPO and 3,4-HOPTO fragments. 3,4-HOPO
fragments are noted in italics.
196
www.chemmedchem.org
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2010, 5, 195 – 199

HOPTO fragment (94G5) was a hit against all four metallo-
enzymes tested. Four fragments were found to only be hits
against LF (94B11, 94D8, 94E9, 94F10).
Table 1. IC50 values of lead fragments against metalloenzyme targets.[a]
Compd
IC50[b] [mm]
MMP-3
Ligand Efficiency[c]
LF
MMP-2
MMP-9
[kcalmolÀ1
]
The most potent fragment identified against all the metallo-
enzymes was 94G5, which is comprised of a 3,4-HOPTO chela-
tor and a biphenyl moiety. This fragment showed a particularly
notable IC50 value of 3 mm against anthrax LF. Although a
handful of very potent (<0.5 mm) LF inhibitors have been re-
ported,[17] the activity of the 94G5 fragment against LF rivals
many of the described inhibitors.[4,14,15,18–21] Interestingly, the bi-
phenyl substituent found in 94G5 is common to other LF[18,22]
and MMP inhibitors,[3,23] although it is quite possible that the
biphenyl group in 94G5 does not occupy the same subsite as
this moiety in these previously reported inhibitors. Nonethe-
less, this finding validates the ability of the chelator fragment
library approach to identify potent structures that are consis-
tent with established chemical motifs for these metalloprotein
targets.
94B11
94C7
94D8
94E8
94E9
94E12
94F5
94F6
94F7
94F8
94F9
94F10
94F12
94G4
94G5
94G6
94G7
94G8
94G10
94G11
94G12
94H2
94H3
34
ꢁ
17
ꢁ
41
34
ꢁ
ꢁ
34
ꢁ
13
ꢁ
33
20
12
10
24
23
ꢁ
ꢁ
ꢁ
ꢁ
33
ꢁ
ꢁ
ꢁ
ꢁ
ꢁ
17
ꢁ
33
21
12
9
20
21
ꢁ
8
23
2
36
18
19
31
41
30
30
34
–
–
–
0.46
–
0.33
0.38
0.42
0.41
0.38
0.43
–
0.41
0.42
0.37
0.33
0.37
0.38
0.39
0.65
0.37
0.39
0.60
ꢁ
ꢁ
ꢁ
28
39
ꢁ
ꢁ
ꢁ
ꢁ
ꢁ
4
ꢁ
ꢁ
38
ꢁ
ꢁ
ꢁ
13
ꢁ
23
32
ꢁ
3
ꢁ
ꢁ
38
21
ꢁ
ꢁ
ꢁ
4
34
1
ꢁ
17
30
35
35
29
ꢁ
The development of effective inhibitors of LF has proven to
be particularly difficult.[24] Comparing the results obtained here
against LF with those obtained from traditional HTS ap-
proaches further highlights the significance of these chelator
fragment libraries. Table 2 compares the five most potent LF
hits obtained from eCFL-1 to the top hits obtained from two
previously reported HTS screens performed on libraries that
each contained at least 10000 compounds.[14,15] These results
show that eCFL-1, with only 87 components, generated hits
that are comparable or superior in potency, ligand efficiency
and drug-likeness to those obtained from labor- and cost-
intensive HTS screens. Specifically, compared to the typical HTS
campaigns reported, the proposed FBLD approach provided:
1) a much higher hit frequency (11% for eCFL-1 versus
<0.25% for the HTS libraries), 2) ligand efficiencies[25] equal to
or higher than those observed for the HTS hits; and 3) com-
pounds that are generally more drug-like given their initial se-
lection based on general drug-like criteria. In fact, the low-
molecular weight of the fragments identified in eCFL-1 are still
amenable to further elaboration (<307 amu), unlike most of
the HTS hits,[14,15] which are already near the conventional MW
limits for a small-molecule therapeutic (Table 2).[26] Overall, this
comparison shows that chelator-based fragment libraries are a
much more efficient and effective way to explore the relevant
chemical space for metalloprotein targets.
ꢁ
ꢁ
ꢁ
36
[a] IC50 values of >50 mm are designated with an ꢁ. Standard deviations
from triplicate measurements for all values listed are <10%. Ligand effi-
ciencies for some fragments are also provided. [b] Gelatinase-1 (MMP-2),
gelatinase-2 (MMP-9), stromelysin-1 (MMP-3), and anthrax lethal factor
(LF). [c] Calculated for fragments against MMP-9.
hits against LF (Table 1). Considering the overall modest
number of compounds screened (87 in eCFL-1), this result is
quite remarkable. Hits with similar affinities have generally
been obtained with much greater effort by testing libraries
consisting of >10000 compounds in HTS campaigns.[14,15] Be-
tween the two different fragment groups in eCFL-1, the 63 3,4-
HOPO fragments produced only one hit against MMP-2
(94G12), two hits against MMP-9 (94G6, 94G12) and three hits
against LF (94B11, 94D8, 94F10). There were no 3,4-HOPO hits
identified against MMP-3. Of the 24 3,4-HOPTO fragments
screened, 16 hits were found against MMP-2, 16 hits against
MMP-9, 5 hits against MMP-3, and 7 hits against LF. The signifi-
cant difference in the number of hits generated from the
smaller 3,4-HOPTO set of fragments against these metallo-
enzymes highlights the importance the chelating group plays
in these libraries. The O,S-chelator of the 3,4-HOPTO fragments
produced at least five or more hits against each enzyme, de-
spite being the smaller subset (~28%) of eCFL-1.
With only one exception (94C7), all of the hits against MMP-
2 were also found to be hits against the other gelatinase,
MMP-9. The identification of common hits for MMP-2 and
MMP-9 is consistent with the high degree of structural homol-
ogy between the catalytic domains of these enzymes. A review
of the literature shows that most MMP-9 inhibitors are also
potent against MMP-2, but the reverse is not necessarily true
due to the slightly larger S1’ pocket present in MMP-2.[16] Only
five fragments, all based on the 3,4-HOPTO chelator, were
shared hits across the three MMPs. Among these, only one 3,4-
In summary, we have presented a new approach to frag-
ment libraries based on known small-molecule chelators and
derivatives of these metal-binding molecules. The chelator
fragment library (CFL-1) produced several hits when screened
against a matrix metalloproteinase. Two chelators from this
library were used to prepare a modest expanded library of hy-
droxypyridinone and hydroxypyridinethione fragments that
were screened against four zinc(II)-dependent metalloprotein
targets. This screen revealed a high hit rate for hydroxypyri-
dinethione fragments and generated leads with good activity
against all the metalloenzymes, including anthrax LF. The re-
sults obtained show that these chelator libraries are particular-
ly well suited for the application of FBLD against metallopro-
tein targets. Indeed, our findings demonstrate that this
ChemMedChem 2010, 5, 195 – 199
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemmedchem.org
197

MED
stirred at 658C for 5 min prior to one microwave cycle of 1.5 min
at 1658C, 250 psi (maxP), and 300 W. The product was precipitated
from solution at 48C overnight, isolated by vacuum filtration, and
dried in vacuo.
Table 2. Comparison of leads identified for anthrax LF by chelator-based
FBLD (this work) versus large library (ꢀ10000 compounds) HTS findings.
Compd[a]
IC50/Ki
MW
DG
Ligand
Efficiency
[kcalmolÀ1
[mm][b] [amu] [kcalmolÀ1
]
3,4-HOPTO derivatives:
A solution of 3-hydroxy-2-methyl-4H-
]
pyran-4-thione (thiomaltol; 200 mg, 1.41 mmol) in HCl (0.38m,
2 mL) in a 10 mL microwave reaction vessel was treated with
amine (3.1 mmol, 2.2 equiv). The heterogenous reaction mixture
was sealed and stirred at 658C for 5 min prior to one microwave
cycle of 1.5 min at 1658C, 250 psi (max P), and 300 W. The reaction
typically produced a yellow solution with a black, oily residue at
the bottom of the tube. The reaction mixture was extracted with
CH2Cl2 (2ꢁ) and the organic phase was dried (MgSO4), vacuum fil-
tered and evaporated in vacuo to give a dark residue. Purification
by silica column chromatography (CH2Cl2 or 0–2% MeOH in
CH2Cl2) gave the desired product.
94D8
94F8
94F10
94G5
94G10
17/4.3
13/3.3
23/5.9
3/0.8
231
259
292
307
249
À7.4
À7.5
À7.2
À8.4
À7.2
0.43
0.42
0.33
0.38
0.42
21/5.4
1.7/0.6 218
3.9/3.8 289
À8.6
0.53
0.34
À7.4
Acknowledgements
We thank Dr. Yongxuan Su (U.C.S.D.) for assistance with mass
spectrometry measurements. This work was supported by the
University of California, San Diego and the National Institutes of
Health (R01 HL080049–04 and R03 AI070651–02 to S.M.C; R01
AI059572 and U01 AI070494 to M.P.).
5.9/2.1 392
À7.8
À8.4
0.30
0.25
0.8/0.8 510
Keywords: chelators · fragment-based lead design · libraries ·
metalloproteins · zinc
[1] P. J. Hajduk, J. Med. Chem. 2006, 49, 6972.
[2] M. Congreve, G. Chessari, D. Tisi, A. J. Woodhead, J. Med. Chem. 2008,
51, 3661.
[3] E. T. Olejniczak, P. J. Hajduk, P. A. Marcotte, D. G. Nettesheim, R. P. Mead-
ows, R. Edalji, T. F. Holzman, S. W. Fesik, J. Am. Chem. Soc. 1997, 119,
5828.
1.7/1.6 549
À7.9
0.23
[4] V. Patel, R. Mazitschek, B. Coleman, C. Nguyen, S. Urgaonkar, J. Cortese,
R. H. Barker, Jr., E. Greenberg, W. Tang, J. E. Bradner, S. L. Schreiber, M. T.
Duraisingh, D. F. Wirth, J. Clardy, J. Med. Chem. 2009, 52, 2185.
[5] S. M. Sternson, J. C. Wong, C. M. Grozinger, S. L. Schreiber, Org. Lett.
2001, 3, 4239.
[a] Compounds identified in this study shown by alphanumeric code,
compounds identified in reference HTS studies shown by structure
[b] See Supporting Information for the determination of Ki values.
[6] M. Flipo, T. Beghyn, J. Charton, V. A. Leroux, B. P. Deprez, R. F. Deprez-
Poulain, Bioorg. Med. Chem. 2007, 15, 63.
approach produces, with a relatively modest screening effort,
high-quality hits that are superior to those obtained from tradi-
tional HTS campaigns where random libraries of tens of thou-
sands of compounds are tested. These findings clearly demon-
strate the potential of chelator-based libraries for targeting
metalloproteins. Such libraries should be useful for identifying
lead structures against a variety of metalloenzymes, well
beyond the zinc(II)-dependent targets reported here. Efforts
along these lines are underway and will be reported in due
course.
[7] J. M. Salvino, R. Mathew, T. Kiesow, R. Narensingh, H. J. Mason, A. Dodd,
R. Groneberg, C. J. Burns, G. McGeehan, J. Kline, E. Orton, S.-Y. Tang, M.
Morrisette, R. Labaudininiere, Bioorg. Med. Chem. Lett. 2000, 10, 1637.
[8] F. E. Jacobsen, J. A. Lewis, S. M. Cohen, J. Am. Chem. Soc. 2006, 128,
3156.
[9] D. T. Puerta, J. A. Lewis, S. M. Cohen, J. Am. Chem. Soc. 2004, 126, 8388.
[10] F. E. Jacobsen, J. A. Lewis, S. M. Cohen, ChemMedChem 2007, 2, 152.
[11] D. T. Puerta, M. O. Griffin, J. A. Lewis, D. Romero-Perez, R. Garcia, F. J. Vil-
larreal, S. M. Cohen, J. Biol. Inorg. Chem. 2006, 11, 131.
[12] T. Cablewski, A. F. Faux, C. R. Strauss, J. Org. Chem. 1994, 59, 3408.
[13] Z. Zhang, S. J. Rettig, C. Orvig, Inorg. Chem. 1991, 30, 509.
[14] S. L. Johnson, L. H. Chen, M. Pellecchia, Bioorg. Chem. 2007, 35, 306.
[15] I. A. Schepetkin, A. I. Khlebnikov, L. N. Kirpotina, M. T. Quinn, J. Med.
Chem. 2006, 49, 5232.
[16] Q.-X. A. Sang, Y. Jin, R. G. Newcomer, S. C. Monroe, X. Fang, D. R. Hurst,
S. Lee, Q. Cao, M. A. Schwartz, Curr. Top. Med. Chem. 2006, 6, 289.
[17] M. Forino, S. Johnson, T. Y. Wong, D. V. Rozanov, A. Y. Savinov, W. Li, R.
Fattorusso, B. Becattini, A. J. Orry, D. Jung, R. A. Abagyan, J. W. Smith, K.
Alibek, R. C. Liddington, A. Y. Strongin, M. Pellecchia, Proc. Natl. Acad.
Sci. USA 2005, 102, 9499.
Experimental Section
General synthetic procedure for eCFL-1
3,4-HOPO derivatives: A solution of 3-hydroxy-2-methyl-4H-pyran-
4-one (maltol; 200 mg, 1.58 mmol) in HCl (0.38m, 2 mL) in a 10 mL
microwave reaction vessel was treated with amine (3.5 mmol,
2.2 equiv). The heterogenous reaction mixture was sealed and
[18] A. Agrawal, C. A. F. de Oliveira, Y. Cheng, J. A. Jacobsen, J. A. McCam-
mon, S. M. Cohen, J. Med. Chem. 2009, 52, 1063.
198
www.chemmedchem.org
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2010, 5, 195 – 199

[19] S. L. Johnson, D. Jung, M. Forino, Y. Chen, A. Satterthwait, D. V. Rozanov,
A. Y. Strongin, M. Pellecchia, J. Med. Chem. 2006, 49, 27.
[20] M. M. D. Numa, L. V. Lee, C. C. Hsu, K. E. Bower, C. H. Wong, ChemBio-
Chem 2005, 6, 1002.
[21] R. G. Panchal, A. R. Hermone, T. L. Nguyen, T. Y. Wong, R. Schwarzen-
bacher, J. Schmidt, D. Lane, C. McGrath, B. E. Turk, J. Burnett, M. J.
Aman, S. Little, E. A. Sausville, D. W. Zaharevitz, L. C. Cantley, R. C. Lid-
dington, R. Gussio, S. Bavari, Nat. Struct. Mol. Biol. 2004, 11, 67.
[22] J. A. Lewis, J. Mongan, J. A. McCammon, S. M. Cohen, ChemMedChem
2006, 1, 694.
[23] A. Agrawal, D. Romero-Perez, J. A. Jacobsen, F. J. Villarreal, S. M. Cohen,
ChemMedChem 2008, 3, 812.
[24] S. L. Johnson, M. Pellecchia, Curr. Top. Med. Chem. 2006, 6, 317.
[25] A. L. Hopkins, C. R. Groom, A. Alex, Drug Discovery Today 2004, 9, 430.
[26] C. A. Lipinski, J. Pharmacol. Toxicol. Methods 2000, 44, 235.
Received: December 14, 2009
Published online on January 7, 2010
ChemMedChem 2010, 5, 195 – 199
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemmedchem.org
199