Bivalent enzyme inhibitors discovered using dynamic covalent chemistry

Article abstract of DOI:10.1002/chem.201201507

A bivalent dynamic covalent chemistry (DCC) system has been designed to selectively target members of the homodimeric glutathione-S-transferase (GST) enzyme family. The dynamic covalent libraries (DCLs) use aniline-catalysed acylhydrazone exchange between bivalent hydrazides and glutathione-conjugated aldehydes and the bis-hydrazides act as linkers to bridge between each glutathione binding site. The resultant DCLs were found to be compatible and highly responsive to templating with different GST isozymes, with the best results coming from the M and Schistosoma japonicum (Sj) class of GSTs, targets in cancer and tropical disease, respectively. The approach yielded compounds with selective, nanomolar affinity (K_i=61 nM for mGSTM1-1) and demonstrates that DCC can be used to simultaneously interrogate binding sites on different subunits of a dimeric protein. Library powers: A bivalent dynamic covalent chemistry (DCC) system has been designed to selectively target members of the homodimeric glutathione-S-transferase (GST) enzyme family. The approach (see scheme) yielded compounds with selective, nanomolar affinity and demonstrates that DCC can be used to simultaneously interrogate binding sites on different subunits of a dimeric protein. Copyright

Full text of DOI:10.1002/chem.201201507

FULL PAPER
DOI: 10.1002/chem.201201507
Bivalent Enzyme Inhibitors Discovered Using Dynamic Covalent Chemistry
Alexandra J. Clipson,[b] Venugopal T. Bhat,[a] Iain McNae,[b] Anne M. Caniard,[b]
Dominic J. Campopiano,[b] and Michael F. Greaney*[a]
Abstract: A bivalent dynamic covalent
chemistry (DCC) system has been de-
signed to selectively target members of
the homodimeric glutathione-S-trans-
ferase (GST) enzyme family. The dy-
namic covalent libraries (DCLs) use
aniline-catalysed acylhydrazone ex-
change between bivalent hydrazides
and glutathione-conjugated aldehydes
and the bis-hydrazides act as linkers to
bridge between each glutathione bind-
ing site. The resultant DCLs were
found to be compatible and highly re-
sponsive to templating with different
GST isozymes, with the best results
coming from the M and Schistosoma
japonicum (Sj) class of GSTs, targets in
cancer and tropical disease, respective-
ly. The approach yielded compounds
with selective, nanomolar affinity (Ki =
61 nm for mGSTM1-1) and demon-
strates that DCC can be used to simul-
taneously interrogate binding sites on
different subunits of a dimeric protein.
Keywords: drug design · dynamic
covalent chemistry · enzymes · hy-
drazones · inhibitors
Introduction
conventional methods,[5] and applying it to targets that
would challenge standard drug discovery approaches. We
describe one such approach here, in which we use DCC to
discover selective bivalent inhibitors of glutathione-S-trans-
ferase (GST) isozymes, including the most potent inhibitor
to date for the M-class of this enzyme.
The GSTs are a superfamily of enzymes responsible for
conjugating the cellular thiol glutathione (GSH) to an array
of endogenous and xenobiotic electrophiles (Scheme 1).[6]
The use of dynamic covalent chemistry (DCC) to discover
new binding motifs for protein targets is a growing area of
research.[1,2] The DCC experiment involves an assembly of
small molecules, termed a dynamic combinatorial library
(DCL), in dynamic exchange through a reversible chemical
reaction. In the presence of a target, such as a protein, the
composition of the DCL will adjust to minimise the global
free energy of the system. With suitable design parameters,[3]
this adaptive behaviour can favour the strongest binding
protein component(s) at the expense of the weaker ones.
The approach is thus distinct from conventional structure-
based drug design (SBDD) in that it requires no structural
information and the synthesis event is accompanied by
a qualitative binding assay.
Pioneering work in biological DCC has featured protein
targets that were necessarily well understood in terms of
medicinal chemistry structure–activity relationships (SAR),
with DCLs producing hit structures largely comprising
known binding motifs.[4] This was required to correlate the
observed behaviour of DCLs in terms of amplifications with
binding affinities. The future development of protein-direct-
ed DCC will depend on exploiting its complementarity to
Scheme 1. GSTs and GSH conjugation. Electrophiles include, for exam-
ple, xenobiotics, quinones (dopamine metabolism), prostaglandins, 4-hy-
droxyalkenals (lipid peroxidation), base propenals (DNA breakdown),
chemotherapeutic agents.
[a] Dr. V. T. Bhat, Prof. M. F. Greaney
School of Chemistry, University of Manchester
Oxford Rd., Manchester, M13 9PL (UK)
E-mail: Michael.Greaney@manchester.ac.uk
The majority of GSTs exist as homodimers containing a con-
served GSH binding site and a hydrophobic substrate bind-
ing region at the interface of the two monomers (Figure 1).
As GSH conjugation is central to phase II metabolism, mod-
ulation of GST isozyme activity has been identified in
a range of therapeutic areas, for example, oncology (multi-
drug resistance and cell proliferation),[7] asthma[8] and in-
flammation.[9] Parasite GSTs are also drug targets in tropical
[b] Dr. A. J. Clipson, Dr. I. McNae, Dr. A. M. Caniard,
Dr. D. J. Campopiano
School of Chemistry, University of Edinburgh
Kingꢀs Buildings, West Mains Rd., Edinburgh, EH9 3JJ (UK)
Supporting information for this article is available on the WWW
under http://dx.doi.org/10.1002/chem.201201507.
Chem. Eur. J. 2012, 00, 0 – 0
ꢁ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
1
&
ÞÞ
These are not the final page numbers!

nations with immediate qualitative information on their
binding affinity.
Results
We designed our DCL using aniline-catalysed acylhydrazone
exchange[14] as the reversible reaction. We have recently
shown that aniline catalysis permits this reaction to be used
at pH values greater than 6, enabling its use with biomolec-
ular targets, such as the GSTs.[13b,15] The physicochemical
and structural features of the acylhydrazone linkage have
proven highly versatile in DCC, being exemplified across
a range of systems and applications.[16] Crucially, the quater-
nary structure of the dimeric enzyme remains intact in the
presence of aniline, enabling ligand selection to take place
from the DCL. We chose the three nitro-substituted benzal-
dehydes A1–A3 as our ligands. The compounds are ana-
logues of the chloro-2,4-dintrobenzene (CDNB) structure,
an archetypal weak-binding substrate of all GSTs.[11a,17] Our
linkers L1–L4 were designed to have differing lengths and
lipophilicity, thus maximising the opportunity to discrimi-
nate across the H-sites of different GST homodimers.
We constructed a DCL using the seven components A1–
A3 and L1–L4 (Scheme 2), potentially forming a mixture of
24 homo- and hetero-bis-acylhydrazones (plus additional
mono-acylhydrazones). Under reaction conditions with ani-
line (5 mm) in ammonium acetate buffer (0.1m) containing
15% by volume DMSO, pH 6.4 at 258C, we observed equili-
bration in 6 h. Pleasingly, we were able to resolve the major-
ity of bis-acylhydrazone components using two HPLC col-
umns in series (Figure 3). Identification of each peak was
done by deconvolution in tandem with LC-MS analysis.
The DCL was then templated with four separate GST iso-
zymes: mGSTM1-1, hGSTP1-1, SjGST and mGSTA4-4
(Figure 4). Best responses were obtained for the M and Sj
isozymes, with the most notable amplification being product
3 (A1-L3-A1), amplified over 600% by the M-class protein
(Figure 5). Smaller amplifications were observed corre-
sponding to hydrazones containing the hydrazide linker L2,
the shortest of the linkers, and a single compound was am-
plified (2, >100%) containing the longest linker, L4. For
the SjGST isoform several peaks were amplified to a similar
extent over the blank equilibrium. With the exception of
A1-L3-A1, all of the other amplifications were due to acyl-
hydrazones containing the hydrazide linker L2.
Figure 1. Structure of hGSTM1-1 that illustrates the H- and the G-sites
(grey: monomer 1; yellow: monomer 2; green: G-site; red: H-site); PDB
ID: 1XWK; the Figure was generated by using PyMOL.
diseases, such as schistosomiasis and malaria, where they
play a pivotal role in the life cycle of the parasite.[10]
The development of GST inhibitors is hampered by the
plasticity of the H-site, which is large and has evolved to
accept a range of small molecules in a variety of different
conformations.[11] Thus, despite the GSTs being amenable to
X-ray crystallography, conventional SBDD methods have
yielded relatively little in the way of potent, isoform-specific
inhibitors. A notable exception is Atkinsꢀ reports of bivalent
inhibitors for the GSTa and p classes.[12]
As part of a multivalent strategy, non-selective GST inhib-
itors, such as ethacrynic acid, were appended to either end
of a peptidic linker designed to span across the dimer inter-
face. Using molecular modelling and split-and-pool combi-
natorial chemistry, the linker could be optimised to afford
potent inhibitors with excellent selectivity between the
GSTa and p isozymes (Figure 2). We have previously stud-
Figure 2. Atkinsꢀ bivalent GSTA1-1 inhibitor, 1; KD =42 nm with 100-fold
selectivity over GSTP1-1.
ied GST inhibition by DCC and discovered selective, mono-
The DCL was generally unresponsive to the mGSTA4-4
isozyme as a template, with some small amplifications being
noted for components containing L2. The hGSTP1-1 isoform,
by contrast, appeared to interfere with the DCL equilibrium
for certain components. Bishydrazones containing two A1
units were degraded by the enzyme, with monohydrazones
being favoured by the enzyme (see the Supporting Informa-
tion). As a result, this isozyme was not used further in tem-
plating studies.
valent G-site binders with low micromolar affinity.[13]
A
DCC approach based around a bivalent strategy has exciting
potential both for the discovery of more potent and selec-
tive GST inhibitors, as well as the development of new con-
cepts for protein-directed DCC in general. The DCL would
produce bivalent molecules that span relatively large surface
areas across two protein subunits—a new development that
contrasts with existing DCL systems directed by proteins.
The DCL design would feature linking molecules and G-site
binders, affording the opportunity to discover novel combi-
With the amplified components identified, it was necessa-
ry to establish that the observed amplifications were due to
&
2
&
www.chemeurj.org
ꢁ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 0000, 00, 0 – 0
ÝÝ
These are not the final page numbers!

Bivalent Enzyme Inhibitors
FULL PAPER
cation would be suppressed.
This experiment is complicated
in the case of the GSTs first by
the lack of any potent inhibitors
that are commercially available,
and second by the propensity of
such inhibitors that are known
to bind in varying locations
across the large H-site. A varie-
ty of non-selective, weak-bind-
ing inhibitors are known, how-
ever, and their effect on the
DCL was examined (Figure 6).
Benzyl
isothiocyanate
(BITC) binds in the H-site of
GSTs and is known to covalent-
ly modify the active site.[18] Sul-
fasalazine binds in the H-site
and is a competitive inhibitor.
The crystal structure of sulfasal-
zine bound to hGSTP1-1 also
shows that the pyridinyl ring
binds in a shallow pocket on
the surface of the protein.[19]
Ethacrynic acid is a non-specific
inhibitor of GSTs widely used
in GST studies, and binds in the
H-site. There is evidence that
ethacrynic acid can bind in
more than one orientation
within this hydrophobic binding
pocket.[20] Similarly, the GSH-
conjugate of CDNB, DNP–
GSH, is believed to bind in
a different position to CDNB
Scheme 2. DCL design for bivalent inhibitor synthesis; GS=S-glutathionyl. Our DCL design uses the aniline-
catalysed reversible hydrazone formation between three aldehydes (A1–A3) and four bivalent hydrazides
(L1–L4).
within
the
hydrophobic
pocket.[19] A final inhibitor, ci-
bacron blue, which binds to the
ligand in site of GSTs,[19] was
found to disrupt the back-
ground DCL equilibrium and
could not be used as a control.
Initial control experiments
with smaller DCLs established
that most of the inhibitors (at
200 mm) had little effect on the
enzyme-templated DCL distri-
bution. DNP–GSH, however,
was effective in suppressing am-
plification in a dose-dependent
Figure 3. DCL consisting of linkers L1–L4 and aldehydes A1–A3.
fashion.
Adding
increasing
amounts of DNP–GSH to the
a genuine templative effect from the GST active site. First,
the addition of bovine serum albumin (BSA) to the DCL
had little effect on the equilibrium composition (see the
Supporting Information). Second, we examined the effect of
adding GST inhibitors to the DCL mixture to see if amplifi-
main DCL reduced the amplification of A1-L3-A1 (peak 3)
and A1-L4-A1 (peak 2) in a dose-dependent fashion (see
the Supporting Information). The amplification of peak
numbers 11, 14, 17 and 21, corresponding to hydrazones not
containing a terminal GSH residue, remained unaffected by
Chem. Eur. J. 2012, 00, 0 – 0
ꢁ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
&
3
&
ÞÞ
These are not the final page numbers!

M. F. Greaney et al.
Figure 6. Known inhibitors of GSTs used to block the targetꢀs binding
pocket in DCL studies.
containing GSH were binding in a different location within
the enzyme, other inhibitors, such as ethacrynic acid and
CDNB, were also added to the DCL. Unfortunately, the in-
hibitors eluted in the middle of the library of bis-acylhydra-
zones, making it very difficult to compare the GST-templat-
ed library with the blank.
To assess the biological activity of the amplified compo-
nents we separately synthesised a selection of symmetrical
bis-acylhydrazones for assay, to identify trends between am-
plification and binding affinity. The compounds A1-L1-A1,
A1-L2-A1, A1-L3-A1 and A1-L4-A1 containing the GS-
tagged aldehyde A1 were strongly responsive to mGSTM1-
1 in the DCL, and include the strongest amplified com-
pound (A1-L3-A1). Compounds containing the L2 linker
were also amplified to a lesser extent, whereas the remain-
ing hydrazones containing L1 and L4 were not selected in
the DCL and could serve as
Figure 4. DCL targeted with four different GST isoforms: A) blank equi-
librium; B) mGSTM1-1; C) hGSTP1-1; D) SjGST; E) mGSTA4-4. The
principal amplified peaks for mGSTM1-1 and SjGST are highlighted with
asterisks.
the addition of DNP–GSH. This could indicate that the ex-
ternal inhibitor was blocking a GSH-binding interaction for
the A1-containing ligands. To see if those compounds not
useful negative controls. The
compounds were tested in the
chloro-2,4-dinitrobenzene
(CDNB) assay, a standard pro-
tocol that measures the inhibi-
tion of GSH conjugation to
CDNB by SNAr reaction.[17]
Table 1 shows the IC50 data for
a subset of symmetrical bis-
acylhydrazones containing alde-
hyde component A1.
The IC50 data correlated well
with the amplifications ob-
served in the DCL in the case
of mGSTM1-1. The product
Figure 5. The changes in DCL composition are represented as %differences of peak area over blank equilibri-
um. Error bars represent standard deviation from two experiments.
&
4
&
www.chemeurj.org
ꢁ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 0000, 00, 0 – 0
ÝÝ
These are not the final page numbers!

Bivalent Enzyme Inhibitors
FULL PAPER
Table 1. Table of IC50 data of DCL products with different GST isoforms
measured by using the CDNB assay.
inhibited the enzyme at 342 and 500 mm, respectively, and
the assembled bis-acylhydrazone product was active at
50 nm (10000-times the potency of the isolated hydrazide
linker). More specifically, varying the linker length and
structure proved an effective strategy for discriminating be-
tween each GST isozyme;[12b] the weak, non-specific inhibi-
tor A1 became a potent, selective inhibitor when linked
with L2 (SjGST), L3 (GSTM1-1) and L4 (GSTP1-1), respec-
tively.
Hydrazone
IC50 [mm]
hGSTP1-1
mGSTM1-1
SjGST]
mGSTA4-4
A1-L1-A1
A1-L2-A1
A1-L3-A1
A1-L4-A1
aldehyde A1
hydrazide L3
1.207
0.337
0.050
0.413
341.7
126.5
11.81
13.45
0.356
3.471
0.252
0.989
1.800
265.6
–
>100
>100
>100
>100
>500
–
ꢀ500
>500
>500
The inhibition data for mGSTA4-4 shows that each of the
hydrazone products had little effect on the enzyme activity,
in line with the minimal templating effect of the enzyme on
the DCL equilibrium. This made the mGSTA4-4 isoform an
excellent negative control, demonstrating that changes in
the distribution of hydrazone products was due to specific
binding site interactions within the mGSTM1-1 isoform and
not due to any non-specific effects from the introduction of
the hydrophobic protein to the DCL.
The inhibition mechanism of the bivalent inhibitors for
mGSTM1-1 was further characterised by using the CDNB
competition assay and the three best binding bis-acylhydra-
zones A1-L2-A1, A1-L3-A1 and A1-L4-A1. The data show
that the hydrazones were non-competitive partial inhibitors
of CDNB (Figure 7A and B). The Ki values were in close
agreement with IC50 measurements for A1-L3-A1 and A1-
L4-A1, meeting the Cheng–Prusoff criteria of IC50 =Ki for
non-competitive inhibitors (Table 2).[21]
with the greatest amplification was A1-L3-A1 (IC50 =50 nm)
an almost tenfold greater inhibition of GST activity relative
to other hydrazone products. Gratifyingly, excellent isoform
specificity was observed, with A1-L3-A1 showing approxi-
mately 20-fold greater inhibitory activity for mGSTM1-
1 over SjGST and 270-fold greater inhibitory activity over
hGSTP1-1. Both A1-L2-A1 and A1-L4-A1 showed smaller
amplifications in the DCL, and inhibited mGSTM1-1 at the
sub-micromolar level. Compound A1-L1-A1 was reduced in
concentration in the DCL and, accordingly, had the weakest
activity at 1.2 mm.
The data for SjGST was less well correlated between IC50
and amplification. Linker L1 was again selected against in
the DCL, consistent with DCL distribution, and proved the
weakest inhibitor (IC50 =3.5 mm). Compound A1-L2-A1,
however, was the most potent (IC50 =0.25 mm) despite being
amplified to a lesser extent than A1-L3-A1 (IC50 =0.99 mm).
Amplification is not always correlated to binding affinity,
a phenomenon identified by Severin[3a] and Sanders et al.[3b,c]
that arises from DCLs equilibrating to the global energy
minimum. The dynamic exchange of DCL components
means this global equilibration can be at the expense of any
one local energy minima for a particular protein-binder
complex, even the strongest one. For the SjGST-templated
library, the majority of amplified peaks (peaks 4, 11, 14, 17
and 21) contained the L2 linker as a common component. It
is possible that competition for this shared building block
reduced the A1-L2-A1 amplification factor relative to A1-
L3-A1 (peak 3), despite the stronger affinity of A1-L2-A1
for SjGST.
The IC50 data for hGSTP1-1 could not be related to the
DCL composition due to the lack of clear equilibration. The
best inhibitor was A1-L4-A1 with an IC50 value of 0.36 mm,
similar to that for mGSTM1-1 (0.41 mm), and considerably
more potent than the hydrazones containing shorter linkers.
These data, taken with the amplification of monohydrazones
in the hGSTP1-1 DCL, are consistent with the linkers L1–
L3 not being long enough to effectively span the two G-sites
in hGSTP1-1.[12]
The principle of multivalent binding was validated in gen-
eral for all GST isoforms—there was a large increase in in-
hibitory activity when the two units of aldehyde A1 were
connected by a linker, regardless of the length or shape of
the linker, compared to the single aldehyde unit. This was
strikingly illustrated for A1-L3-A1, the most potent inhibi-
tor of mGSTM1-1, whereby the two components A1 and L3
Table 2. Binding data for DCL components with mGSTM1-1.
Hydrazone
IC50 [nm]
Ki[a] [nm]
KD [nm]
A1-L1-A1
A1-L2-A1
A1-L3-A1
A1-L4-A1
1207
337
50
–
–
525
107
–
646Æ80.7
61Æ4.1
634Æ23.8
413
[a] Standard error shown for Ki data taken from three replicates.
Isothermal calorimetry (ITC) was then used to determine
the binding stoichiometry of the most potent compound A1-
L3-A1. The results gave a binding stoichiometry of 0.4 with
the GST monomer, supporting our hypothesis that the
linker is binding across the dimeric GST structure (Fig-
ure 7C). Finally, we examined the effect of the GS moiety
on the aldehyde with respect to binding affinity. The sym-
metrical compounds A2-LX-A2 and A3-LX-A3 were syn-
thesised and their IC50 values determined by using the
CDNB assay. As shown in Table 3, the removal of gluta-
Table 3. IC50 data for homo-acylhydrazone linkers with mGSTM1-1.[a]
Hydrazide linker
A1-LX-A1 IC50 A2-LX-A2 IC50 A3-LX-A3 IC50
mGSTM1 [mm] mGSTM1 [mm] mGSTM1 [mm]
aldehyde no linker 341.7
–
>500
L1
L2
L3
L4
1.21
–
–
3.38
106.4
>100
0.337
0.050
0.413
2.02
9.32
49.89
[a] Data were acquired by using the CDNB assay.
Chem. Eur. J. 2012, 00, 0 – 0
ꢁ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
&
5
&
ÞÞ
These are not the final page numbers!

M. F. Greaney et al.
thione from the products resulted in a 100-fold decrease in
inhibition for the longer linkers L3 and L4, demonstrating
the importance of glutathione for binding affinity. The loss
of glutathione for the shortest linker L2 was better tolerat-
ed, with a sixfold reduction in inhibition, suggesting that
these shorter compounds were binding in a different mode
to the longer linkers. Assaying A2-L2-A2, the most potent
bivalent inhibitor without glutathione, against CDNB again
showed a non-competitive, partial inhibitor profile with Ki =
1.98
N
Discussion
DCC has been used to identify novel bivalent inhibitors of
GSTs, with the most potent compound (A1-L3-A1) inhibit-
ing mGSTM1-1 with nanomolar activity and excellent iso-
form specificity over three other isoforms studied. Inhibition
of the M-class of GSTs is relevant to oncology, where resist-
ance to anticancer therapies has been correlated with al-
tered expression levels of this enzyme (along with
GSTP1).[7,22] Resistance mechanisms are twofold: GSTM1
can catalyse GSH conjugation to anticancer agents, and it
can also down-regulate the p38-MAPK pro-apoptotic path-
way.[23]
In this second mechanism, GSTM1 has been shown to
bind to apoptosis signal-regulating kinase 1 (ASK1), which
is a MAP kinase that activates both the JNK and p38 signal-
ling pathways leading to apoptosis. GSTM1 acts as an inhibi-
tor of ASK1, which dissociates in response to cell stress re-
sulting in the activation of ASK1. Literature reports of M-
class GST inhibitors are sparse, with the natural product cur-
cumin being the most potent compound described to date
(IC50 =0.3 mm for GSTM1-1).[24]
To better understand the role of the selected linker L3 in
spanning the two G-sites in the GSTM1-1 homodimer, we
constructed an overlay model based on the crystal structure
of human glutathione-S-transferase M1A-1A (PDB ID:
1XWK) containing the ligand DNP–GSH from Listowsky
and co-workers.[25] The model shows clearly that the L3
linker can comfortably span the dimer interface without in-
troducing any unfavourable contacts (Figure 8A). The dis-
tance between the two GS moieties in the crystal structure
(15.4 ꢂ) and the equivalent in the model (15.5 ꢂ) allows the
GS moiety to nearly perfectly overlay in both binding sites
within the dimer maintaining all contacts with the protein
(Figure 8B). The slightly longer L4 and the shorter L2 and
L1 linkers have correspondingly weaker affinity in their hy-
drazone adducts, requiring different linker conformation or
protein movement to allow the GS moiety to maintain its
binding mode in both G-sites.
The presence of glutathione at the terminal of each linker
was of particular importance for binding affinity for the
longer linkers A1-L3-A1 and A1-L4-A1. The glutathione
was less important for linkers made from L2, with all hydra-
zones featuring L2 giving some amplification in the GST-
templated DCLs. The activity of A1-L2-A1 against the M, P
Figure 7. Assay of most potent compound, A1-L3-A1, with mGSTM1-1.
A) Michaelis–Menten, and B) Lineweaver–Burk plots. The data best fit
to a non-competitive (partial) equation. Standard errors are from three
repeat experiments. C) ITC data: the top panel shows heat effects associ-
ated with the injection of ligand into the cell containing GST. The lower
panel shows the binding isotherm, generated from the integrated heats in
the top panel, and the best fitted curves (see the Supporting Information
for ITC data).
&
6
&
www.chemeurj.org
ꢁ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 0000, 00, 0 – 0
ÝÝ
These are not the final page numbers!

Bivalent Enzyme Inhibitors
FULL PAPER
Figure 9. A1-L2-A1 binding data relative to monovalent hydrazones pre-
viously prepared.
they bound into the H-site. ITC was used to determine
whether, due to the short linker length, two molecules of
A1-L2-A1 were in fact binding to the GST dimer—one in
each active site. Unfortunately, the results were inconclusive
(n=0.7) with the measurements being complicated by gella-
tion effects from the compound at higher concentration (see
the Supporting Information).
Figure 8. A) and B) A1-L3-A1 overlaid onto GSTM1-1 crystal structure
containing GS-DNB bound at each G-site. The protein surface across the
dimer interface is shown with the two halves of the dimer coloured
yellow and red, respectively. The GS-DNB moiety of the crystal structure
is shown in stick form with white carbons and the modelled A1-L3-A1
ligand is shown with cyan carbons. The ligand is seen to span the dimer
interface whilst introducing no unfavourable contacts.
Conclusion
The versatility of DCC in discovering novel binding partners
for proteins has been demonstrated, with the identification
of a novel, isoform specific, bivalent inhibitor of mGSTM1-
1. The success of the study rests on the ability of aniline to
catalyse acylhydrazone exchange in the presence of proteins,
without compromising tertiary and quaternary protein struc-
ture. The lead compound, A1-L3-A1, was amplified from
a 24-membered library, and is the most potent (and selec-
tive) inhibitor described to date for the M-class of GSTs.
DCC has a natural overlap with multivalent, fragment-type
approaches to ligand discovery given that it is predicated on
the assembly of small, weakly-binding fragments, in situ.
Our study shows how this element of DCL design may be
used very successfully in the discovery of potent enzyme in-
hibitors. We have also demonstrated that the addition of an
external enzyme inhibitor to an equilibrating DCL can pro-
vide additional insights to simply switching amplification off.
The large GST binding region interrogated by our DCLs en-
abled us to select a range of inhibitors that bind in different
areas of the enzyme, and examine their effect (or lack of
one) on the amplification of best-binding components. Final-
ly, by designing DCLs to probe larger protein surface areas
across two 28 kDa protein monomers, we have shown that
the approach is suited to targets that are less tractable to
and SjGST is somewhat surprising. The L2 linker should not
be long enough to span the distance between the two G-
sites without substantial distortion of the dimer structure,
suggesting a different, possibly monovalent, binding mode.
It should be noted, however, that the IC50 value of A1-L2-
A1 was very favourable in comparison to other monovalent
inhibitors based on the A1 scaffold. Previous work from our
laboratory has described acylhydrazone GST inhibitors
formed from DCLs using A1 and monovalent acylhydra-
zides designed to explore the hydrophobic binding pock-
et.[13b] The best of the compounds from that work had IC50
values of 50 mm for hGSTP1-1 and 20 mm for SjGST (2 and
3, Figure 9). This is a fivefold reduction in inhibition for
hGSTP1-1 and a tenfold reduction for SjGST when com-
pared to bivalent inhibitor A1-L2-A1. Compounds 2 and 3
showed little inhibitory effect against mGSTM1-1 with IC50
values being greater than 100 mm.
The monovalent acylhydrazones were competitive inhibi-
tors of CDNB, and molecular docking studies indicated that
Chem. Eur. J. 2012, 00, 0 – 0
ꢁ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
&
7
&
ÞÞ
These are not the final page numbers!

M. F. Greaney et al.
Chem. 2006, 14, 3275–3284; e) S. Zameo, B. Vauzeilles, J.-M. Beau,
Eur. J. Org. Chem. 2006, 5441–5444.
conventional structure-based drug discovery methods.
Future work will look to extrapolate these techniques to the
study of protein–protein interactions.
[5] DCMS (dynamic combinatorial mass spectrometry) has been used
to discover novel binding fragments via disulfide formation and pro-
tein mass spectrometry. Seminal work: D. A. Erlanson, A. C. Braist-
ed, D. R. Raphael, M. Randal, R. M. Stroud, E. M. Gordon, J. A.
Wells, Proc. Natl. Acad. Sci. USA 2000, 97, 9367–9372; recent exam-
ples: a) E. C. Y. Woon, M. Demetriades, E. A. L. Bagg, W. Aik,
S. M. Krylova, J. H. Y. Ma, M. Chan, L. J. Walport, D. W. Wegman,
K. N. Dack, M. A. McDonough, S. N. Krylov, C. J. Schofield, J. Med.
Chem. 2012, 55, 2173–2184; b) B. M. R. Liꢅnard, R. Huting, P. Las-
saux, M. Galleni, J.-M. Frere, C. J. Schofield, J. Med. Chem. 2008,
51, 684–688; c) B. M. R. Liꢅnard, N. Selevsek, N. J. Oldham, C. J.
Schofield, ChemMedChem 2007, 2, 175–179; d) S. A. Poulsen, J.
Am. Soc. Mass Spectrom. 2006, 17, 1074–1080.
Experimental Section
HPLC conditions for analysis of DCL libraries: Column, Luna 5m
C18(2), 50 mmꢃ2.0 mm, and Luna 5m C18(2), 30 mmꢃ4.6 mm, in se-
quence; flow rate, 1 mLminÀ1; wavelength, 254 nm; temperature, 308C;
gradient H2O/MeCN (0.01% TFA) from 95 to 80% over 15 min, then to
60% over a further 15 min, and finally to 5% over 5 min.
DCL blank equilibrium; aniline catalysis of reversible hydrazone forma-
tion: The four bishydrazide linkers L1–L4 (4ꢃ2.4 mL, 10 mm, DMSO), al-
dehyde A1 (2.4 mL, 10 mm, water), aldehydes A2-A3 (2ꢃ2.4 mL, 10 mm,
DMSO) and aniline (15 mL, 0.1m, DMSO) were added to a mixture of
DMSO (15.6 mL, 15% by vol.) and ammonium acetate buffer (252.6 mL,
100 mm, pH 6.4). The DCL was allowed to incubate at room temperature
with occasional shaking, and was monitored periodically by HPLC to es-
tablish the blank composition until the relative populations of the hydra-
zones became constant.
[6] J. D. Hayes, J. U. Flanagan, I. R. Jowsey, Annu. Rev. Pharmacol.
Toxicol. 2005, 45, 51–88.
[7] Review: D. M. Townsend, K. D. Tew, Oncogene 2003, 22, 7369–
7375. Recent examples: a) K. Tsuboi, D. A. Bachovchin, A. E.
Speers, T. P. Spicer, V. Fernandez-Vega, P. Hodder, H. Rosen, B. F.
Cravatt, J. Am. Chem. Soc. 2011, 133, 16605–16616; <lit b>J. Son,
J. J. Lee, J. S. Lee, A. Schuller, Y. T. Chang, ACS Chem. Biol. 2010,
5, 449–453; c) L. Federici, C. Lo Sterzo, S. Pezzola, A. Di Matteo, F.
Scaloni, G. Federici, A. M. Caccuri, Cancer Res. 2009, 69, 8025–
8034; d) W. H. Ang, L. J. Parker, A. De Luca, L. Juillerat-Jeanneret,
C. J. Morton, M. Lo Bello, M. W. Parker, P. J. Dyson, Angew. Chem.
2009, 121, 3912–3915; Angew. Chem. Int. Ed. 2009, 48, 3854–3857;
e) W.-S. Li, W. S. Lam, K.-C. Liu, C.-H. Wang, H.-C. Chang, Y.-C.
Jen, Y. T. Hsu, S. S. Shivatare, S.-C. Jao, Org. Lett. 2010, 12, 20–23.
[8] Selected examples: a) I. Romieu, M. Ramirez-Aguilar, J. J. Sienra-
Monge, H. Moreno-Macꢆas, B. E. del Rio-Navarro, G. David, J.
Marzec, M. Hernꢇndez-Avila, S. London, Eur. Respir. J. 2006, 28,
953–959; b) W. D. Carroll, W. Lenney, P. W. Jones, R. C. Strange, F.
Child, M. K. Whyte, R. A. Primhak, A. A. Fryer, Clin. Exp. Allergy
2005, 35, 1155–1161; c) F. Sampsonas, M.-A. Archontidou, E. Salla,
K. Karkoulias, G. Tsoukalas, K. Spiropoulos, Allergy Asthma Proc.
2007, 28, 282–286.
Enzyme-templated DCL assay: Protein, GST (to a final concentration of
20 mm, for mGSTM1-1; 152.8 mL, 1.1 mgmLÀ1, in potassium phosphate
buffer 0.1m, pH 6.8) or BSA (to a final concentration of 20 mm, 199.5 mL,
2.0 mgmLÀ1, in 0.9% aqueous NaCl solution containing sodium azide),
the four hydrazide linkers L1–L4 (4ꢃ2.4 mL, 10 mm, DMSO), aldehyde
A1 (2.4 mL, 10 mm, water), aldehydes A2-A3 (2ꢃ2.4 mL, 10 mm, DMSO)
and aniline (15 mL, 0.1m, DMSO) were added to a mixture of DMSO
(15.6 mL, 15% by vol.) and ammonium acetate buffer (to a total volume
of 255 mL, for mGSTM1-1; 99.8 mL, 100 mm, pH 6.4). The DCL was al-
lowed to incubate at room temperature, with occasional shaking, for
48 h. HPLC analysis was performed and the traces were compared with
the blank equilibrium.
[9] B. Xue, Y. Wu, Z. Yin, H. Zhang, S. Sun, T. Yi, L. Luo, FEBS Lett.
2005, 579, 4081–4087.
[10] S.-C. Jao, J. Chen, K. Yanga, W.-S. Lia, Bioorg. Med. Chem. 2006,
14, 304–318.
Acknowledgements
[11] a) B. Mannervik, U. H. Danielson, Crit. Rev. Biochem. Mol. Biol.
1988, 23, 283–337; b) D. Sheehan, G. Meade, V. M. Foley, C. A.
Dowd, Biochem. J. 2001, 360, 1–16.
[12] a) R. P. Lyon, J. J. Hill, W. M. Atkins, Biochemistry 2003, 42, 10418–
10428; b) D. Y. Maeda, S. S. Mahajan, W. M. Atkins, J. A. Zebala,
Bioorg. Med. Chem. Lett. 2006, 16, 3780–3783; c) S. Mahajan, L.
Hou, R. Paranji, D. Maeda, J. Zebala, W. M. Atkins, J. Am. Chem.
Soc. 2006, 128, 8615–8625.
We thank CRUK, EastChem and the EPSRC for funding (studentship
for A.C., scholarships to V.T.B. and A.M.C., leadership fellowship to
M.F.G., respectively) and the EPSRC mass spectrometry service at the
University of Swansea. Dr. Elizabeth Blackburn is thanked for assistance
with ITC measurements and Dr. Scott Baxter for assistance with protein
purification.
[13] a) B. Shi, R. Stevenson, D. J. Campopiano, M. F. Greaney, J. Am.
Chem. Soc. 2006, 128, 8459–8467; b) V. T. Bhat, A. M. Caniard, T.
Luksch, R. Brenk, D. J. Campopiano, M. F. Greaney, Nat. Chem.
2010, 2, 490–497.
[14] a) E. H. Cordes, W. P. Jencks, J. Am. Chem. Soc. 1962, 84, 826–831;
b) A. Dirksen, S. Dirksen, T. M. Hackeng, P. E. Dawson, J. Am.
Chem. Soc. 2006, 128, 15602–15603.
[15] For alternative nucleophilic catalysis of hydrazone formation, see:
a) A. R. Blanden, K. Mukherjee, O. Dilek, M. Loew, S. L. Bane,
Bioconjugate Chem. 2011, 22, 1954–1961; b) S. R. Beeren, M. Pittel-
kowz, J. K. M. Sanders, Chem. Commun. 2011, 47, 7359–7361.
[16] a) G. R. L. Cousins, S.-A. Poulsen, J. K. M. Sanders, Chem.
Commun. 1999, 1575–1576; b) T. Bunyapaiboonsri, O. Ramstrçm, S.
Lohmann, J.-M. Lehn, L. Peng, M. Goeldner, ChemBioChem 2001,
2, 438–444; c) R. L. E. Furlan, Y.-F. Ng, S. Otto, J. K. M. Sanders, J.
Am. Chem. Soc. 2001, 123, 8876–8877; d) S. L. Roberts, R. L. E.
Furlan, G. R. L. Cousins, J. K. M. Sanders, Chem. Commun. 2002,
938–939; e) T. Bunyapaiboonsri, H. Ramstrçm, O. Ramstrçm, J.
Haiech, J.-M. Lehn, J. Med. Chem. 2003, 46, 5803–5811; f) O. Ram-
strçm, S. Lohmann, T. Bunyapaiboonsri, J.-M. Lehn, Chem. Eur. J.
[1] a) V. T. Bhat, M. F. Greaney, in Dynamic Combinatorial Chemistry:
In Drug Discovery, Bioorganic Chemistry, and Materials Science
(Ed.: B. L. Miller), Wiley, New York, 2009, pp. 43–82; b) O. Ram-
strçm, L. Amorim, R. Caraballo, O. Norberg, in Dynamic Combina-
torial Chemistry, (Eds.: J. N. H. Reek, S. Otto), Wiley, New York,
2009, pp. 109–150; c) M. Hochgꢄrtel, J.-M. Lehn, in Fragment-Based
Approaches in Drug Discovery (Eds.: W. Jahnke, D. A. Erlanson),
Wiley, New York, 2006, pp. 341–364.
[2] P. T. Corbett, J. Leclaire, L. Vial, K. R. West, J.-L. Wietor, J. K. M.
Sanders, S. Otto, Chem. Rev. 2006, 106, 3652–3711.
[3] a) K. Severin, Chem. Eur. J. 2004, 10, 2565–2580; b) P. T. Corbett, S.
Otto, J. K. M. Sanders, Chem. Eur. J. 2004, 10, 3139–3143; c) P. T.
Corbett, J. K. M. Sanders, S. Otto, Chem. Eur. J. 2008, 14, 2153–
2166.
[4] Selected examples: a) I. Huc, J.-M. Lehn, Proc. Natl. Acad. Sci.
USA 1997, 94, 2106–2110; b) R. J. Lins, S. L. Flitsch, N. J. Turner, E.
Irving, S. A. Brown, Angew. Chem. 2002, 114, 3555–3557; Angew.
Chem. Int. Ed. 2002, 41, 3405–3407; c) S. Zameo, B. Vauzeilles, J.-
M. Beau, Angew. Chem. 2005, 117, 987–991; Angew. Chem. Int. Ed.
2005, 44, 965–969; d) S.-A. Poulsen, L. F. Bornaghi, Bioorg. Med.
&
8
&
www.chemeurj.org
ꢁ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 0000, 00, 0 – 0
ÝÝ
These are not the final page numbers!

Bivalent Enzyme Inhibitors
FULL PAPER
2004, 10, 1711–1715; g) J. Liu, K. R. West, C. R. Bondy, J. K. M.
Sanders, Org. Biomol. Chem. 2007, 5, 778–786; h) M. G. Simpson,
M. Pittelkow, S. P. Watson, J. K. M. Sanders, Org. Biomol. Chem.
2010, 8, 1173–1180; i) M. G. Simpson, M. Pittelkow, S. P. Watson,
J. K. M. Sanders, Org. Biomol. Chem. 2010, 8, 1181–1187; j) S. R.
Beeren, J. K. M. Sanders, J. Am. Chem. Soc. 2011, 133, 3804–3807;
k) S. R. Beeren, J. K. M. Sanders, Chem. Sci. 2011, 2, 1560–1567;
l) A. V. Gromova, J. M. Ciszewski, B. M. Miller, Chem. Commun.
2012, 48, 2131–2133; m) G. Deng, F. Li, H. Yu, F. Liu, C. Liu, W.
Sun, H. Jiang, Y. Chen, ACS Macro Lett. 2012, 1, 275–279.
[17] W. H. Habig, M. J. Pabst, W. B. Jakoby, J. Biol. Chem. 1974, 249,
7130–7139.
[21] Y.-C. Cheng, W. H. Prusoff, Biochem. Pharmacol. 1973, 22, 3099–
3108.
[22] C. C. McIlwain, D. M. Townsend, K. D. Tew, Oncogene 2006, 25,
1639–1648.
[23] Selected examples: a) S.-G. Cho, Y. H. Lee, H.-S. Park, K. Ryoo,
K. W. Kang, J. Park, S.-J. Eom, M. J. Kim, T. S. Chang, S.-Y. Choi, J.
Shim, Y. Kim, M.-S. Dong, M.-J. Lee, S. G. Kim, H. Ichijo, E.-J.
Choi, J. Biol. Chem. 2001, 276, 12749–12755; b) S. Dorion, H. Lam-
bert, J. Landry, J. Biol. Chem. 2002, 277, 30792–30797; c) K. Ryoo,
S.-H. Huh, Y.-H. Lee, K. W. Yoon, S.-G. Cho, E.-J. Choi, J. Biol.
Chem. 2004, 279, 43589–43594.
[24] R. Appiah-Opong, J. N. M. Commandeur, E. Istyastono, J. J. Bo-
gaards, N. P. E. Vermeulen, Xenobiotica 2009, 39, 302–311.
[25] Y. Patskovsky, L. Patskovska, S. C. Almo, I. Listowsky, Biochemistry
2006, 45, 3852–3862.
[18] L. A. Ralat, R. F. Colman, J. Biol. Chem. 2004, 279, 50204–50213.
[19] A. J. Oakley, M. Lo Bello, M. Nuccetelli, A. P. Mazzetti, M. W.
Parker, J. Mol. Biol. 1999, 291, 913–926.
[20] a) A. J. Oakley, J. Rossjohn, M. Lo Bello, A. M. Caccuri, G. Federi-
ci, M. W. Parker, Biochemistry 1997, 36, 576–585; b) A. J. Oakley,
M. Lo Bello, A. P. Mazzetti, G. Federici, M. W. Parker, FEBS Lett.
1997, 419, 32–36.
Received: April 30, 2012
Published online: && &&, 0000
Chem. Eur. J. 2012, 00, 0 – 0
ꢁ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
&
9
&
ÞÞ
These are not the final page numbers!

M. F. Greaney et al.
Drug Design
A. J. Clipson, V. T. Bhat, I. McNae,
A. M. Caniard, D. J. Campopiano,
M. F. Greaney* ............... &&&& — &&&&
Bivalent Enzyme Inhibitors Discov-
ered Using Dynamic Covalent Chemis-
try
Library powers: A bivalent dynamic
covalent chemistry (DCC) system has
been designed to selectively target
members of the homodimeric gluta-
thione-S-transferase (GST) enzyme
family. The approach (see scheme)
yielded compounds with selective,
nanomolar affinity and demonstrates
that DCC can be used to simulta-
AHCTUNGTRENNUNG
&
10
&
www.chemeurj.org
ꢁ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 0000, 00, 0 – 0
ÝÝ
These are not the final page numbers!