Design and synthesis of novel lactate dehydrogenase a inhibitors by fragment-based lead generation

Article abstract of DOI:10.1021/jm201734r

Lactate dehydrogenase A (LDHA) catalyzes the conversion of pyruvate to lactate, utilizing NADH as a cofactor. It has been identified as a potential therapeutic target in the area of cancer metabolism. In this manuscript we report our progress using fragment-based lead generation (FBLG), assisted by X-ray crystallography to develop small molecule LDHA inhibitors. Fragment hits were identified through NMR and SPR screening and optimized into lead compounds with nanomolar binding affinities via fragment linking. Also reported is their modification into cellular active compounds suitable for target validation work.

Full text of DOI:10.1021/jm201734r

Article
pubs.acs.org/jmc
Design and Synthesis of Novel Lactate Dehydrogenase A Inhibitors
by Fragment-Based Lead Generation†
Richard A. Ward,* Claire Brassington, Alexander L. Breeze, Alessandro Caputo, Susan Critchlow,
Gareth Davies, Louise Goodwin, Giles Hassall, Ryan Greenwood, Geoffrey A. Holdgate, Michael Mrosek,
Richard A. Norman, Stuart Pearson, Jonathan Tart, Julie A. Tucker, Martin Vogtherr,‡ David Whittaker,
Jonathan Wingfield, Jon Winter, and Kevin Hudson
Oncology and Discovery Sciences iMEDs, AstraZeneca, Mereside, Alderley Park, Macclesfield, Cheshire, SK10 4TG, U.K.
S
* Supporting Information
ABSTRACT: Lactate dehydrogenase A (LDHA) catalyzes the conversion of pyruvate to lactate, utilizing NADH as a cofactor. It
has been identified as a potential therapeutic target in the area of cancer metabolism. In this manuscript we report our progress
using fragment-based lead generation (FBLG), assisted by X-ray crystallography to develop small molecule LDHA inhibitors.
Fragment hits were identified through NMR and SPR screening and optimized into lead compounds with nanomolar binding
affinities via fragment linking. Also reported is their modification into cellular active compounds suitable for target validation
work.
rate was relatively low at around 1%, when using a 30%
inhibition cutoff with 10 μM compound concentration. These
actives mainly grouped into small clusters or singletons. As a
primary deconvolution tool for the HTS output, we measured
the binding of compounds to purified recombinant human
LDHA protein using ligand-observed (1D) NMR. We failed to
confirm binding for any of the compounds tested; therefore,
the actives of initial interest were considered likely to be false
positives. One common mechanism for a significant number of
the false positives was suspected to be inhibition of the target
via silver, or other heavy metal impurities, in the compound
samples.10 This hypothesis was confirmed by the resynthesis of
some compounds without the use of reagents containing silver.
These examples were inactive upon testing. Although no way
forward was identified from the HTS output, the ligand-
observed NMR provided us with an assay suited for initiating a
fragment-based lead generation (FBLG) campaign.11,12 FBLG
approaches have been particularly successful in a number of
targets (including HSP9013−16 and BACE17−20) and across the
kinase target class more generally.21 One of the most cited
INTRODUCTION
■
In normal tissues, lactate generation is limited to anaerobic
conditions where oxygen levels are low. In contrast, cancer cells
preferentially convert glucose into lactate through glycolysis,
even under normal oxygen concentrations, a phenomenon
termed “aerobic glycolysis” or the Warburg effect.1,2 Lactate
dehydrogenase (LDH) is a tetrameric enzyme comprising
subunits encoded by the LDHA and LDHB genes, which can
be combined to generate five isoforms LDH1 (B4), LDH2
(B3A1), LDH3 (B2A2), LDH4 (B1A3), and LDH5 (A4).
LDHA is a key glycolytic enzyme that catalyzes the formation
of lactate from pyruvate, with LDHB favoring the backward
conversion of pyruvate from lactate. LDHA is frequently up-
regulated in clinical tumors, and high expression is often linked
to poor prognosis.3−7 In addition, inhibition of LDHA with
siRNA or small molecules has been reported to induce
oxidative stress and cell death.8,9 In contrast, LDHB is located
in heart muscle, which may indicate that achieving selectivity
over this enzyme would be desirable. It is thus thought that
LDHA could be a promising target for cancer therapy.
Our initial effort to identify inhibitors of LDHA utilized a
high-throughput screen (HTS) of our corporate compound
collection. Although active compounds were identified, the hit
Received: December 23, 2011
Published: March 14, 2012
© 2012 American Chemical Society
3285
dx.doi.org/10.1021/jm201734r | J. Med. Chem. 2012, 55, 3285−3306

Journal of Medicinal Chemistry
Article
Figure 1. Summary of FBLG strategy to be applied to LDHA. X-ray crystal structure shown is PDB code 1i10.25
examples of this approach is the discovery of ABT-263, a potent
inhibitor of Bcl-2, which has been reported by Abbott.22 Our
ligand-observed NMR capability provided an adequate
throughput for the primary screening of fragments, and we
developed a capability for follow-up confirmation and
characterization using protein-observed (2D) NMR on stable
isotope-labeled LDHA protein. However, measuring the
binding affinity of compounds by protein-observed NMR was
both time and resource intensive. It was apparent that a higher
throughput assay would be required, so a number of label free
approaches23 to measure affinity were evaluated. After
exploration of several options, the BIAcore assay24 gave an
acceptable balance of throughput, sensitivity, and data
consistency. As part of the HTS workup, we had also
established a soakable rat LDHA protein crystal structure
system. Although a human LDHA protein crystal system has
previously been published (PDB code 1i1025), we found the rat
crystal system to be more amenable to soaking experiments and
hypothesized that their high homology would make the rat
system a suitable surrogate (Figure S1, Supporting Information).
Human and rat LDHA sequences share 94% global se-
quence similarity using the Bestfit algorithm;26 the cofactor
and substrate binding sites, however, appear identical. We
therefore prosecuted a FBLG campaign against LDHA,
reasoning that the target may indeed be “druggable” or at
least “ligandable”27 but that our corporate collection might not
contain compatible chemical matter. Analysis of the binding site
of LDHA encouraged us that if we could find fragments
binding to the different subpockets of the active site (i.e.,
adenine, nicotinamide, and substrate regions), then we could
potentially link or merge multiple fragments (Figure 1). At this
stage we also made the decision that we would not attempt to
drive selectivity over LDHB, as the consequences of LDHB
inhibition are not fully understood and it would significantly
complicate the cascade of assays required. Sequence alignments
between LDHA and LDHB, again using the BestFit algorithm,
highlighted them to be 83% similar. However, the substrate and
cofactor pockets show significantly higher levels of homology
and the key catalytic residues are entirely conserved. Herein, we
report the identification of fragment hits and their modification
to produce cell active chemical probes.
that were either synthesized or purchased and subsequently
profiled in assays that could be used in an FBLG cascade.
Compound 1 has been described as an inhibitor of
Plasmodium falciparum LDH with selectivity over human
LDH and was synthesized using literature protocols.28 This
compound appears to be more active against human LDHB
than LDHA (IC50 = 7.8 μM against LDHB compared with
IC50 = 450 μM against LDHA), consistent with the literature
precedent for competitive inhibition against lactate but not
pyruvate.28 The binding of this compound to LDHA was
detected by NMR only when NADH or NAD+ was also present
in the assay. Weak binding (millimolar range) was observed in
the presence of NADH, and tight (slow exchange) binding in
the presence of 1 mM NAD+, in agreement with the previously
reported preference of the azole series for binding to the
LDH−NAD+ complex based on the results of kinetic analysis.
The binding mode of this compound to the rat LDHA−NAD+
complex was also confirmed by crystallography,29 which shows
it to bind in the substrate pocket in a manner similar to that
previously reported for the binding of the equivalent oxadiazole
to human LDHA−NAD+ (PDB code 1t2f28). Although we
were able to confirm binding and activity of compound 1, we
did not pursue this further, since compounds of this class show
limited opportunity for further elaboration. Compounds 230
and 38 have been disclosed as LDHA inhibitors; however, we
were unable to validate these compounds in our cascade as
specific inhibitors. Compound 2 was synthesized using
literature protocols30 and showed significant superstoichiomet-
ric and nonspecific binding. Insufficient sample was available for
testing in our enzyme assay. Compound 3 (known as FX-11)
was also synthesized using the published protocols;8 however,
denaturation of the protein was observed in our assays and we
were unable to generate a ligand bound crystal structure.
Compound 49 was synthesized using a modified version of the
literature protocol (synthesis S1, Supporting Information). We
were able to confirm specific binding of 4 to LDHA by 2D
NMR; precipitation of protein−ligand complex at higher ligand
concentrations hampered determination of an accurate Kd from
the chemical shift titration data, but our calculated value of ∼9
μM (which probably slightly underestimates the true Kd) is
similar to the Ki of 8.9 μM against NADH reported by the
1
authors. The pattern of LDHA backbone amide H,15N shifts
RESULTS AND DISCUSSION
induced by 4 indicated binding to the adenine subpocket of
the cofactor site rather than the nicotinamide/substrate region
as postulated by Granchi et al. BIAcore studies with com-
pound 4 demonstrated significant levels of nonspecific binding.
■
Assay Assessment. A survey of the literature identified a
number of chemical probes to assess in our assays (NMR,
BIAcore, and enzyme assay). Table 1 highlights compounds
3286
dx.doi.org/10.1021/jm201734r | J. Med. Chem. 2012, 55, 3285−3306

Journal of Medicinal Chemistry
Article
a
Table 1. Data Generated on Published LDHA Inhibitors
Kd (μM)
IC50 (μM)
compd
LDHA NMR
LDHA BIAcore
LDHA enzyme
LDHB enzyme
b
1
2
3
4
5
6
7
8
tight binding
ambiguous
ambiguous
no binding
ambiguous
ambiguous
ambiguous
weak binding
weak binding
weak binding
12.8
450
7.8
not tested
>500
>500
>500
>500
>500
132
not tested
not tested
not tested
not tested
not tested
not tested
9.4
c
>9.0
640
650
4000
tight binding
a
Data quoted are an average of at least two independent measurements, apart from NMR data which were derived from a single experiment.
b
c
Measurement by NMR detected in the presence of 1 mM NAD+. Kd is a lower limit estimate from 2D chemical shift titration data and is
influenced by poor solubility of protein/ligand complex.
Efforts to soak 4 into our rat LDHA crystal system identified
electron density in the adenine pocket, but a binding mode could
not be confirmed29 and the compound showed no activity in our
enzyme assay. Compounds containing the core structure of
compound 5 have been previously disclosed as weak inhibitors
of LDHA.31 We were able to confirm binding by NMR and
BIAcore, along with obtaining a crystal structure showing
binding to the adenine pocket.29 Compounds 4 and 5 therefore
appear not to be potent or specific enough inhibitors of LDHA
to show enzyme activity in our assay. Similarly, two fragments
of NADH, compounds 6 and 7, also show reproducible binding
and yet were inactive in the enzyme assay. The key compound
in validating our early cascade was compound 8 (tetrahydro-
NAD32). Compound 8 showed tight binding by 2D NMR; it
was therefore not possible to quantify the binding with a Kd
value. Compound 8 also showed binding in BIAcore (12.8 μM)
and was active in our enzyme assay with IC50 = 132 μM. This
apparent drop-off in activity is likely due to the competition of
3287
dx.doi.org/10.1021/jm201734r | J. Med. Chem. 2012, 55, 3285−3306

Journal of Medicinal Chemistry
Article
a
Table 2. Selected Examples from the Fragment Screen against LDHA
compd
LDHA NMR Kd (μM)
LE (NMR)
LDHA BIAcore Kd (μM)
LE (BIAcore)
0.16
LDHA enzyme IC50 (μM)
9
300
0.22
0.19
0.17
0.20
0.23
0.20
1600
>500
>500
>500
>500
>500
>500
10
11
12
13
14
1400
2400
1000
400
weak binding
3000
0.17
0.17
770
weak binding
weak binding
4200
a
Data quoted are an average of at least two independent measurements, apart from NMR data which were derived from a single experiment.
the compounds with NADH in the enzyme assay compared to
measuring their direct binding in BIACore. As observed for
compound 1, compound 8 also shows greater activity in our
LDHB enzyme assay (IC50 = 9.4 μM) than in LDHA. These
results obtained with 8 gave us confidence that were we to
initiate a fragment screen to identify small molecules that
bound tightly to the active site of LDHA and that were
competitive with NADH, these could be a starting point for
developing bona fide inhibitors.
Fragment Screening. Prior to the development of the
BIAcore assay, ligand-observed NMR was used to screen our
diverse library12,33 of approximately 1000 fragments to identify
chemical start points. Forty-four binders were identified (4.6%
hit rate), which gave early encouragement that the target was
indeed ligandable.27 However, of the 27 hits prioritized for
protein-observed NMR follow-up, we were only able to
quantify binding for 13 fragments. Of these, only 3 fragments
showed tighter binding than 1 mM with the range of affinities
measured by NMR being from 300 to 4200 μM. The binding of
a selection of these hits was measured in BIAcore to help
further validate the assay, with a priority for follow-up given to
fragments that showed consistent binding across both plat-
forms. A selection of fragments from this initial screen are
shown in Table 2, along with the NMR, BIAcore, and enzyme
activity data where generated.
different assays may be misleading. For example, the most
ligand efficient fragment according to NMR Kd data (13) did
not have a measurable affinity by BIAcore, although weak
binding was observed. This highlights the value of orthogonal
assays in the context of FBLG. All fragment hits for which we
obtained cocrystal structures with rat LDHA bound in the
adenine pocket, which is a relatively shallow and open pocket.
The binding modes of 12 and 14 to rat LDHA are shown in
Figure 2 to highlight the interactions made by the fragments.
The asymmetric unit of the rat LDHA crystal system used in
these studies contains a tetramer. In several cases, initial
electron density for bound compound was clearly visible in only
one of the four monomers, generally chain B. Unless indicated
otherwise, all figures and discussions of binding mode refer to
this monomer. The benzothiazole ring of 12 lies in the same
hydrophobic pocket as that occupied by the adenine ring of
NADH, between the side chains of Val-52 and Ile-115.
Fragment 12 is additionally held in place via a water-mediated
hydrogen bond between the thiazole nitrogen and the side
chain hydroxyl of Tyr-82 and via extensive hydrogen bond
interactions between the amide group and residues in the ribose
binding site and at the mouth of the phosphate channel (amide
NH to side chain carboxyl of Asp-51, and amide carbonyl to
backbone NH and CO of Gly-96). The quality of the electron
density indicates a degree of flexibility in the orientation of the
N-methyl group which is directed toward solvent.
It became evident from this initial screen that the ligand
efficiency (LE)34 was generally modest, which reinforced how
challenging it would be to develop potent small molecules
against the target. In this manuscript, we have used the simplest
measure of LE that is the pKd divided by the number of heavy
atoms. An additional observation was that although LE may be
a suitable method for prioritizing fragments, the actual LE value
varies depending on the assay, so comparing LE values across
Fragment 14 binds in the same region as 12, with the
chlorophenyl group packed against the side chain of Val-52.
The bound water previously observed in the complex with 12 has,
however, been displaced, allowing the tetrazole to directly interact
with the hydroxyl side chain of Tyr-82. The tetrazole ring of 14
extends into a region not occupied by 12, where it may make a
long-range interaction (>3.3 Å) with the side chain of Asn-122.
3288
dx.doi.org/10.1021/jm201734r | J. Med. Chem. 2012, 55, 3285−3306

Journal of Medicinal Chemistry
Article
Figure 2. Binding mode of fragments 12 (left) and 14 (right) in the rat LDHA crystal structure, showing final 2Fo − Fc electron density for 12 and
14 (blue, 1.0σ level). Selected nearby water molecules and protein residues are shown. Hydrogen bonding interactions with the protein are indicated
as dashed lines. All figures were prepared using PyMol.35
This binding mode is made possible by a flip of the Phe-118
side chain into an alternative rotamer. A number of fragments
with acidic functionality, such as compound 14, which we had
expected to bind in the substrate pocket, were shown to occupy
the adenine pocket upon soaking into rat LDHA crystals. The
binding of acidic fragments in the adenine pocket had also been
observed for compounds 4 and 5, observations that were also
supported by NMR data.
Our crystal system contains high concentrations of malonate
(a substrate mimic), and this was observed to occupy the
substrate binding site. We were therefore concerned that
certain fragments with affinity for the substrate/nicotinamide
pocket in solution might not be able to compete with
endogenous malonate in the crystal system and thus were
biased toward binding in the promiscuous adenine groove. We
were unable to identify alternative crystallization conditions,
most probably as a result of malonate binding at the substrate
site and thus stabilizing the conformation of the flexible active
site loop, and also binding to additional sites at the monomer
interfaces and thus stabilizing the tetramer structure. Attempts
to replace malonate with alternative precipitants in the buffers
used for crystal soaking experiments showed citrate and sulfate
to be compatible with maintenance of crystal integrity; however,
the bound malonate at the substrate site was then replaced by
citrate or sulfate.
then informed our wider testing using the higher throughput
BIAcore assay. The NMR assay continued to be used in an
orthogonal manner to confirm the binding of key compounds
from BIAcore.
In total, around 350 analogues of the screening hits were
tested in the BIAcore assay. The screening of these analogues
initially investigated the effect of minor changes to the core
scaffold and subsequently the effect of adding groups that
might grow toward the nicotinamide pocket. A selection of
analogues identified from these searches are shown in Table 3.
Compound 15 was identified by looking for the minimal
number of heavy atoms on the scaffold required to show
binding in the BIAcore assay, which was an important
parameter to establish with this technology now being the
primary screening tool. Compound 16 contained a group with
the potential to occupy the phosphate channel. No clear
improvement in potency was observed, likely because of the
phosphate channel being charged and solvent exposed.38
Compound 17 provided the most notable increase in affinity,
while maintaining the LE of compound 15, through the
addition of a lipophilic group. We were unable, however, to
obtain an unambiguous binding mode for 17 in our rat LDHA
crystal structure system. Compound 18 was subsequently
tested in BIAcore, and although less potent than 17, a complex
crystal structure was successfully obtained (Figure 3).
We searched for analogues for a number of the confirmed
fragment hits, focusing particularly on compounds 9, 10, 11,
and 12. Of the hits we identified, these looked the most
promising with respect to binding affinity, LE, and chemical
feasibility. We had a concern with compound 13 in that it
appeared to require the phenol group for the interaction with
Tyr-82.36 We believed the identification of a start point in the
substrate pocket would potentially require an acidic fragment.
We did not, therefore, wish to start with an acidic fragment
(such as compound 14) in the adenine pocket. The additional
attraction of compounds such as 9−12 was that the crystal
structures gave an indication of how the molecules could be
grown toward the nictonamide region of the pocket. This
helped focus analogue selection, as there appeared limited
opportunities to pick up other interactions in the open ade-
nine groove, although it does provide an opportunity to ex-
ploit the solvent exposed region of the pocket to modulate
physicochemical properties. We initially identified a number
of near-neighbors of each fragment from our corporate col-
lection and commercial databases37 for testing by NMR which
The benzothiazole core of 18 binds identically to that of the
parent fragment 12, maintaining the water-mediated interaction
between the thiazole nitrogen and the side chain hydroxyl of
Tyr-82 and maintaining interactions through the amide group
with the side chain of Asp-51 and the backbone of Gly-96. The
carboxylic acid extends toward the phosphate binding region,
picking up interactions with the side chain of Arg-98 and the
backbone of Ala-29 (via water). Flexibility in the binding of this
substituent is observed, as indicated by the poorer quality of
the electron density for this region and variation in the
modeled orientations between the four different molecules
in the tetramer. The S-allyl substituent is not fully visible in
the electron density, although its presence was confirmed by
liquid chromatography−mass spectrometry (LC−MS) analysis
of the samples used in the soaking experiment. The S-allyl
has been modeled in a variety of conformations to reflect the
disorder indicated by the lack of density while maintaining a
preferred torsion angle around the S−CH2 bond. Given the
lack of density for the S-allyl substituent, the modest increase in
potency observed for 18 (BIAcore Kd = 130 μM) compared
3289
dx.doi.org/10.1021/jm201734r | J. Med. Chem. 2012, 55, 3285−3306

Journal of Medicinal Chemistry
Article
a
Table 3. Evolution of the Adenine Pocket Fragment 12 by Analogue Screening To Yield Compounds 17 and 18
compd
LDHA BIAcore Kd (μM)
LE (BIAcore)
LDHA NMR Kd (μM)
LE (NMR)
0.2
LDHA enzyme IC50 (μM)
12
15
16
17
18
770
790
500
40
0.21
0.22
0.16
0.21
0.l8
1000
not tested
540
>500
>500
>500
>500
>500
0.18
0.16
0.18
92
130
170
a
Data quoted are an average of at least two independent measurements, apart from NMR data which were derived from a single experiment.
directed fragment set to target this region, exploiting a number
of different data sources (Figure 4).
Figure 3. X-ray crystal structure of compound 18 bound in rat LDHA,
showing final 2Fo − Fc electron density for 18 (blue, 1.0σ level).
Selected nearby water molecules and protein residues are shown.
Hydrogen bonding interactions with the protein are indicated as
dashed lines.
Figure 4. Summary of the combined nicotinamide/substrate pocket
directed screening set.
with 16 (500 μM) is most likely driven by lipophilic inter-
actions between the sulfur and Phe-118. In addition, we also
monitored the ligand lipophilicity efficiency (LLE)39−41 of our
fragments, here defined as pKd subtracted by cLogP (version
4.3). Although 15 and 17 have acceptable LE values (0.22 and
0.21, respectively), the LLE of the two compounds is relatively
poor (0.83 and 0.64) reflecting the lipophilic nature of the
adenine pocket. We were prepared to work with such fragments
despite poor LLE, since we hypothesized that their relatively
high lipophilicity would eventually be balanced out by a re-
quirement for hydrophilic interactions in the substrate pocket.
We also suspected that there would be scope to modulate
properties through substitution at the solvent-exposed region of
the adenine pocket. Our primary aim, therefore, was to find an
“anchor” in the adenine pocket from which we could grow.
Targeting the Substrate Pocket. Having identified
potential chemical start points bound in the adenine pocket,
we turned our attention to fragments binding in the
nicotinamide and/or substrate pockets. Results from the
screening of our diverse fragment set indicated that finding
hits for these pockets would be challenging, so we constructed a
One method used to identify potential fragments for
screening was fragmentation of compounds described in the
literature as NADH mimics, for example, inhibitors of targets
such as poly-ADP-ribose polymerases (PARPs)42,43 and
nicotinamide phosphoribosyltransferase (NAMPRT).44,45 Sub-
structural searching was used to include fragments with a
pendent amide directly attached to an aromatic ring as potential
mimics of the key interaction of NADH in the nicotinamide
pocket. To target the substrate pocket, we identified a set of
acidic fragments through substructural searches, focusing on
carboxylic acid groups directly attached to an aromatic or
aliphatic five- or six-membered ring. Fragments of this type
could target the substrate pocket in a similar manner as
compound 1 and yet provide the potential for further
elaboration. Furthermore, our in-house LDHA crystal struc-
tures showed malonate (19, Figure 5) binding in the substrate
pocket. In contrast to oxamate, which is bound in a subset of
published vertebrate LDH structures,46−49 malonate provides a
potential vector for further elaboration. Oxamate is a flat
molecule, and it was less clear from inspection of the binding
mode how this template could be exploited to grow out of the
3290
dx.doi.org/10.1021/jm201734r | J. Med. Chem. 2012, 55, 3285−3306

Journal of Medicinal Chemistry
Article
A key result for the project was the identification of
compounds 20 and 21, which were selected because they
contained the malonate substructure and also ranked favorably
in the virtual screen against the “apo” structure. Compound 20
has improved affinity and LE compared to 21; however,
because 20 contains a bromine atom, LE can be misleading, as
the increase in molecular weight and lipophilicity due to this
single atom is large. This is highlighted if you compare their
LLE values based on the BIAcore data, where 20 has a LLE of
2.1 while 21 has a more favorable LLE of 2.8. Compounds 20
and 21 were soaked into crystals of rat LDHA (Figure 6) and
Figure 5. Chemical structure of malonate (19) and oxamate, present
in the substrate pocket of in-house LDHA structures.
substrate pocket, similar to the situation reported for com-
pound 1. Malonate has the central methylene group as a
potential growth point. We were able to detect the binding of
malonate to LDHA by NMR with Kd = 5 mM, giving a
respectable LE of 0.33. Fragments were therefore added to the
screening set which contained malonate and oxamate
substructures.
Finally, two protein-based virtual screens were performed
using a published human muscle LDH structure, which has
NADH and oxamate bound (PDB code 1i1025). In one of these
virtual screens, both NADH and oxamate were removed
(referred to as “apo”), while in the second only NADH was
removed (referred to as “substrate bound”). Fragments were
docked using Glide50 in standard precision mode, with the top
ranked hits visually assessed and prioritized for testing.
Considered fragments were also filtered by property and
visual assessment. Although the use of rigid property-based cut-
offs was avoided, the molecules generally conformed to a cLogP
(version 4.3) between −1 and 3.5, fewer than 25 heavy atoms,
4 hydrogen bond donors, and 6 hydrogen bond acceptors with
no structural alert flags from our in-house filtering protocol.12
Exceptions to the filtering were made for compounds that were
likely to be ionized or where there was a particularly strong
rationale for inclusion, for example, compounds that were
identified from multiple sources. In total, 695 fragments were
identified by the above methods. However, there was significant
overlap, leaving 450 unique fragments for testing.
The targeted screen was run at a single compound
concentration of 2.5 mM using the BIAcore assay and identified
40 hits as clear binders (8.9% hit rate), with fragments being
identified from all subsets. The majority of these initial hits,
however, did not give reproducible Kd values in the BIAcore
concentration response follow-up. Subsequent analysis showed
that a significant proportion of these compounds also bound to
acid-denatured LDHA, suggesting that their binding was
nonspecific. A set of compounds showing binding only to
folded LDHA were prioritized for crystallography, two
examples of which are shown in Table 4.
Figure 6. X-ray crystal structure of compound 20 (cyan) overlaid with
21 (purple), bound in the substrate pocket of LDHA. Selected protein
residues are shown. Hydrogen bonding interactions with the protein
are indicated as dashed lines.
shown to bind in such a way as to occupy both the substrate
and nicotinamide pockets.
Compounds 20 and 21 bridge the nicotinamide and
substrate pockets with the diacid binding in the substrate
pocket in a manner similar to that observed for malonate alone.
The 4-bromobenzyl and 3,4-dimethoxybenzyl groups overlap
with the area occupied by the nicotinamide ring, placing the
4-substituent where the oxygen of the ribose ring would
otherwise be located. The oxygen atoms of the diacid moiety
are involved in polar interactions with the side chains of
residues lining the substrate binding pocket: Arg-105, Asn-137,
Arg-168, His-192, and Thr-247 and in chains B and C, Gln-99.
The linker oxygen of 20 contacts the side chain of Thr-247,
while a flip at the prochiral central carbon of the diacid group in
21 shifts the equivalent carbon ∼0.6 Å away from Thr-247.
This movement results in a subtle shift of the benzyl ring
overlaying the methyl group of the p-methoxy substituent in 21
onto the position occupied by the p-bromo of 20. These two
groups are then oriented such that they may form polar
a
Table 4. Selected Hits from Output of Directed Fragment Screening against the Nicotinamide/Substrate Pocket
compd
LDHA BIAcore Kd (μM)
LE (BIAcore)
LDHA NMR Kd (μM)
LE (NMR)
LDHA enzyme IC50 (μM)
20
21
210
280
0.24
0.20
470
0.22
0.17
>500
>500
1000
a
Data quoted are an average of at least two independent measurements, apart from NMR data which were derived from a single experiment.
3291
dx.doi.org/10.1021/jm201734r | J. Med. Chem. 2012, 55, 3285−3306

Journal of Medicinal Chemistry
Article
interactions with the backbone carbonyl of Thr-94. The
m-methoxy substituent of 21 is relatively solvent exposed.
Inspection of the complex structures indicates that removal
of the bromine atom in 20 would be required to allow growth
along the phosphate channel. We were therefore encouraged by
the data on compound 21, in which the para substituent is a
methoxy group, potentially offering such a growth opportunity.
Only a limited number of analogues could subsequently be
tested in the BIAcore assay, because of the relative scarcity of
the malonate “warhead” in our corporate collection. A small set
of fragments related to compounds 20 and 21 from which one
of the acidic groups had been removed or replaced were tested;
however, the binding affinity was too low for reliable detection
of binding for these compounds. We were aware that the
diacidic nature of the malonate motif would cause significant
challenges in terms of cell membrane permeability and ADME
properties,51 despite providing a tight binding chemical start
point in the substrate pocket. We took the decision to retain
the diacid until we had built additional potency elsewhere in the
molecule, at which point we would be in a better position to
assess the full impact of modifying or removing one of the
acidic functionalities.
acidic group, which would need to be removed or modified to
allow further growth.
Having exhausted all the relevant analogues from our
corporate collection, we took the decision to synthesize
compounds in order to grow the fragments toward each
other. To achieve this aim, we synthesized sets of compounds
growing from the adenine pocket (A1 and A2, Scheme 1)
and from the substrate pocket, S1 and S2 (Schemes 2 and 3).
a
Scheme 2. Synthesis of Substrate Targeted Library S1
a
(i) NaOMe, MeOH, rt; (ii) diethyl 2-chloromalonate, 70 °C, 1−3
days; (iii) LiOH, THF, water, rt, 16 h.
We did not expect to make significant increases in potency in
this growing stage, as the phosphate pocket is highly solvent
exposed and charged. Our aim was to maintain ligand efficiency
where possible and to close the distance between the fragments.
Amide couplings provided compounds for libraries A1 and A2
from the previously reported52 aminobenzothiazole 36 and
commercially available compound 37.
Fragment Growing. An overlay of the structures of com-
pounds 12 and 20 in complex with rat LDHA (Figure 7) showed
Libraries S1 and S2 were accessed from the corresponding
substituted phenols, benzaldehydes, or benzyl halides. Niederl
and Roth53 reported the alkylation of sodium phenolates with
halomalonate esters. In accordance with their findings, we
found that use of diethyl 2-chloromalonate was optimal, as it
gave exclusively monophenoxymalonates 38 (Scheme 2),
although reaction was usually incomplete, especially for
electron rich phenols. The more reactive diethyl 2-bromomal-
onate led to formation of diphenoxymalonates via self-
bromination. We obtained products from a diverse range of
phenols but found that hydroxy heterocycles and aliphatic
alcohols were unsuccessful. Hydrolysis under basic conditions
subsequently gave the library S1 diacids. Because of intra-
molecular H-bonding, the first hydrolysis to the monoacid
was significantly faster (reaction time of ∼30 min at ambient
temperature) than the second hydrolysis (reaction time of
∼2 days).
Figure 7. Overlay of the complex structures of compounds 12
(purple) and 20 (cyan) with rat LDHA, showing the shortest distance
separating the fragments.
that they were separated by a significant distance (over 7 Å)
that would need to be reduced before we could consider
linking. Compound 18 extends toward the phosphate channel,
but it does not close the overall gap between the two fragments
(Figure S2, Supporting Information) and the substituent
directed toward the nicotinamide pocket already contains an
Knoevenagel condensation of aromatic aldehydes and
malonate esters has been reported54 catalyzed by amine tipped
dendrimers (Scheme 3). We found that polymer-supported
trisamine resin together with molecular sieves was also effective
a
Scheme 1. Synthesis of Adenine Pocket Libraries
a
(i) RCO2H, HATU, DIPEA, DMF, rt, 16 h.
3292
dx.doi.org/10.1021/jm201734r | J. Med. Chem. 2012, 55, 3285−3306

Journal of Medicinal Chemistry
Article
a
Scheme 3. Synthesis of Substrate Targeted Library S2
a
(i) Diethyl malonate, PS-trisamine, 4 Å molecular sieves, EtOH, 50 °C, 1−3 days; (ii) H2, Pd−C, EtOH; (iii) LiOH, water, THF, rt, 16 h; (iv)
diethyl malonate, NaH, THF; (v) Meldrum’s acid, H2O, 70 °C, 4 h; (vi) NaBH4, AcOH, DCM, rt, 1 h; (vii) LiOH, H2O, 120 °C, 1.5 h.
as a catalyst for this reaction. This gave good results on a
number of substrates, with the advantage that the catalyst could
be removed by filtration. Especially in the case of electron-
deficient aldehydes, the temperature and number of equivalents
of malonate must be carefully controlled to avoid the further,
irreversible reaction of the alkylidene products 39 with excess
malonate. Catalytic hydrogenation gave alkyl malonates 40,
which were hydrolyzed to afford malonic acids for library S2.
Another approach to these compounds was developed via
alkyl-substituted Meldrum’s acids. Benzaldehydes were con-
densed with Meldrum’s acid under the aqueous, uncatalyzed
conditions of Bigi55 to yield the alkylidines 41. Reduction to the
alkyl derivatives 42 was effected with NaBH4/AcOH, followed
by basic hydrolysis to the malonic acids. Where Knoevenagel
synthesis was not feasible, alkyl malonates 40 were formed via
alkylation of diethyl malonate. Dialkylation was unavoidable,
but the required materials could be separated by chromatography.
The libraries based on the template S1 were less chemically
tractable than for S2. The final hydrolysis step often led to
partial cleavage of the C−O bond to the malonate position.
For libraries A1 and A2, protein−ligand docking was used to
focus the selection of a set of commercially available reagents to
groups that grew in the preferred direction; however, as we
were not confident that the binding mode prediction (or
scoring) would be reliable in this region of the protein, it was
used as a method of prioritization, as opposed to rigid filtering.
An issue we quickly identified when synthesizing library A1 was
the relatively poor solubility of these compounds, which
prevented us from measuring accurate Kd values for many of
these molecules. The decision was taken to abandon library A1
and switch the scaffold to the less lipophilic and more soluble
library scaffold A2, keeping the substituents the same. This
change, however, decreased the affinity of the library template
by around an order of magnitude, comparing compounds 16
(Kd of 500 μM by NMR) against 17 (40 μM). As part of our
exploitation of template A2, we synthesized approximately 150
compounds.
were obtained were bound with the benzothiazole core in the
same position as the parent fragment, 15; however, out of all
the amide side chains, only the urea group in 22 (Table 5) was
directed toward the phosphate channel. Although no more
potent than compound 15, compound 22 was identified as an
attractive hit, as the urea group made specific interactions with
the protein that appeared to “cut the corner” of the NADH
binding mode, thereby avoiding the charged phosphate pocket
and bringing our fragments within 3.5 Å at closest approach
(Figure 8). The compound also offered clear opportunities for
further growth and optimization.
The benzothiazole core of 22 binds in the adenine pocket as
described previously for 12 and 18, while the urea group ex-
tends into the phosphate channel, where it is anchored by
multiple direct and water-mediated interactions. The amino
groups make a bidentate interaction with the backbone car-
bonyl of Gly-96, while the urea carbonyl makes water mediated
contacts to Gly-28, Ala-29, Val-30 and Gly-31 (backbone NH),
and Thr-94 (side chain hydroxyl). Dual soaks with 20 plus 22
or with 21 plus 22 confirmed that both fragments could bind
simultaneously and that the binding modes were not perturbed
(Figure S3, Supporting Information).
We targeted an immediate increase in potency by modifying
compound 22 with the S-propyl group to give compound
23 (Table 5), in the hope that it might allow us to observe
enzyme activity, helping us to further validate our approach.
A significant increase in affinity was observed in the BIAcore
assay (Kd = 35 μM for 23 compared with Kd = 770 μM for 22),
consistent with previous SAR; however, this appeared not to be
sufficient to result in measurable enzyme activity. Compound
24 was synthesized to confirm the hypothesis derived from
visual analysis of the crystal structure that the urea group could
be further substituted to grow into the nicotinamide pocket.
Encouragingly, when the BIAcore data for 24 were compared to
22 (Kd = 160 μM versus Kd = 770 μM) a modest improvement
in LE from 0.16 to 0.18 was observed, indicating further growth
is tolerated. The complex crystal structure showed an identical
binding mode for 24 and 22 (Figure S4, Supporting
Information).
BIAcore measurements highlighted 10 compounds from
library A2 that showed an improved (or at least broadly
consistent) affinity over the early compounds based on the
same template. All of these compounds were soaked into rat
LDHA crystals in order to determine the orientation of the
substituent on the amide. Compounds for which structures
Seventy compounds were synthesized across templates S1
and S2. Analysis of the crystal structures of compounds 20 and
21 in complex with rat LDHA resulted in meta- or para-
substituents being prioritized in order to grow the fragments
3293
dx.doi.org/10.1021/jm201734r | J. Med. Chem. 2012, 55, 3285−3306

Journal of Medicinal Chemistry
Article
a
Table 5. Selected Binders from the Adenine and Substrate Pocket Libraries
compd
LDHA BIAcore Kd (μM)
LE (BIAcore)
LDHA NMR Kd (μM)
LE (NMR)
LDHA enzyme IC50 (μM)
15
22
23
24
25
790
770
35
0.22
0.16
0.20
0.18
0.18
not tested
binding
binding
not tested
240
>500
>500
>500
>500
>500
160
1100
0.23
a
Data quoted are an average of at least two independent measurements, apart from NMR data which were derived from a single experiment.
linker on the position of the individual fragments was moni-
tored computationally, along with the agreement of the linker
geometry with experimental data in the Cambridge Structural
Database56 (CSD). Compound 26 was the first linked com-
pound (represented by core 1) made as a one-off compound to
test our hypothesis. Gratifyingly, 26 produced a breakthrough
increase in affinity (Kd = 0.13 μM) while maintaining LE
(Table 6).
The synthesis of linked compounds required substrate
and adenine pocket fragments containing pendent amine
groups. To access the malonate-bearing fragment for core 1
(Scheme 4), aldehyde 43 was formed by Mitsunobu coupling
of 4-hydroxybenzaldehyde and N-Boc-aminoethanol. 43 was
subjected to our previous Knoevenagel conditions; subsequent
hydrogenation of the alkylidine product 44 led to malonate 45.
Treatment of 45 with HCl under anhydrous condtions
removed the Boc group to give 46 while leaving the malonate
esters intact. We found that 46 in free-base form was unstable
to polymerization on storage; it was either formed and used
immediately or stored in salt form.
The adenine pocket unit was formed via amide coupling of
37 and acid 47 (Scheme 5). The resulting amide 48 was
deprotected with TFA to give 49. Our initial strategy for
constructing the urea linker was via the active phenyl carbamate
50 (Scheme 6), which was reacted with the malonate-bearing
amine 46 to afford the fully linked molecule 51. Basic ester
hydrolysis afforded the final compound 26.
Figure 8. Binding mode of compound 21 (purple), superimposed with
the crystal structure of fragment 22 (cyan) and the NADH bound
1i1025 structure (pink).
along the phosphate channel toward the adenine pocket. We
included substituents such as methoxy, amides, sulfonamides,
and amines to provide additional growth points should the
group be tolerated by the protein. The libraries based on the
template S2 were both more chemically tractable than those for
S1 and consistently showed an improved affinity, resulting in us
favoring the carbon-linked malonate template. Compound 25
(Table 5) was one of the most interesting examples, with
Kd = 1100 μM. This compound was designed to assess
whether the p-methoxy functionality in 21 could be used as a
growth vector. Compound 25 also showed binding by NMR,
with Kd = 240 μM.
Fragment Linking. With the gap between the two
fragments now reduced to less than 3 Å (Figure S4, Supporting
Information), we decided to link the two molecules 24 and 25.
We used the crystal structure of compound 21 to inform us of
the likely binding mode of 25, as the crystal structure of 25 was
not available at the time. Multistep chemistry routes meant that
we required a strong rationale for making specific linked
compounds. Possible target compounds with a range of linkers
were modeled in the active site of LDHA with input from
medicinal, structural, and synthetic chemists. The effect of the
Core 2 required building block 59, which involved a longer
route (Scheme 7). Boc protection of commercially available
amino acid salt 52 followed by esterification gave 53, which was
converted to aldehyde 55 via alcohol 54. The Knoevenagel
reaction of 55 was problematic; the alkylidine malonate pro-
duct 56 was prone to reaction with a second equivalent of
malonate to give 57. This occurred more readily than for core
1 analogue 44; in that case the p-oxygen presumably moder-
ated the electrophilicity of the aldehyde carbon. Formation of
impurity 57 was irreversible, and it could not be separated
3294
dx.doi.org/10.1021/jm201734r | J. Med. Chem. 2012, 55, 3285−3306

Journal of Medicinal Chemistry
Article
a
Table 6. BIAcore and Enzyme Activity Data on Linked Compounds
a
Data quoted are an average of at least two independent measurements.
from 56 by chromatography. It could be controlled by reducing
the excess of malonate used in the reaction; however, this led to
incomplete reactions and lower yields. Our best solution was to
take this inseparable mixture of product and starting material
through to the subsequent hydrogenation step. Any remaining
55 was reduced to the corresponding toluene derivative which
could then be removed from product 58 by chromatography.
Finally, acidic deprotection of the Boc group gave the required
amine 59. This compound was even more prone to
polymerization than 46 and was also used immediately or
stored in salt form.
The starting point for the oxymalonate linked compounds
was 4-hydroxyphenethylamine. Reaction with phenyl chlor-
oformate under aqueous basic conditions gave clean conversion
to the active carbamate 60 (Scheme 8). Coupling with amine
49 gave the linked phenol 61 which was able to displace
3295
dx.doi.org/10.1021/jm201734r | J. Med. Chem. 2012, 55, 3285−3306

Journal of Medicinal Chemistry
Article
a
Scheme 4. Synthesis of Substrate Template for Core 1 Fragment Linking
a
(i) DIAD, PPh3, THF; (ii) diethyl malonate, PS-trisamine, 4 Å molecular sieves, EtOH, 50 °C, 80%; (iii) H2,Pd−C, EtOH, rt, 16 h, 95%; (iv) TFA,
DCM, rt, 16 h, 99%.
a
Scheme 5. Synthesis of Adenine Template for Fragment Linking
a
(i) HATU, DIPEA, DMF, rt, 16 h, 99%; (ii) TFA, DCM, 18 h, 93%.
a
Scheme 6. Fragment Linking of Core 1 Compounds
a
(i) Phenyl chloroformate, pyridine, THF, 0 °C to rt, 16 h, 39%; (ii) amine 46, NMP, rt, 3 days, 53%; (iii) LiOH·H2O, H2O, EtOH, rt, 2 days, 48%.
dimethyl 2-chloromalonate under basic conditions to give
oxymalonate diester 62. Hydrolysis to 63 was much quicker
than for core 2 because of the electron-withdrawing effect of
the oxygen. The reactions had to be stopped quickly after
completion, as prolonged exposure to base led to cleavage of
the malonate group.
Some of our urea-linked compounds, such as 27 and 32,
were synthesized via the phenyl carbamates of the malonate-
bearing amines 46 and 59; however, we encountered lower
yields, having issues with homocoupling and the formation of
byproduct. An improved linking approach was developed for
the synthesis of subsequent urea-linked molecules, for example,
29 (Scheme 9). Reaction of amine 49 with 1 equiv of N,N′-
carbonyldiimidazole (CDI) gave the active intermediate 64 as a
stable solid which could be stored. This was reacted with the
malonate-bearing amine 59 to give, after subsequent hydrolysis,
compound 29 cleanly and in good yield.
so the ethyl ester 66 was converted to the corresponding
benzyl ester 68 via acid 67. The malonate ester moiety was
connected via Knoevenagel condensation as previously. As
with 56, overreaction forming the bis-malonate adduct was a
serious issue in the synthesis of 69, so yields were generally
low. Hydrogenation accomplished simultaneous reduction of
the double bond and removal of the benzyl group to give 70.
This was a relatively stable compound, although air oxidation
to acid was observed in some of the preceding aldehyde-
bearing intermediates.
The linked compound 71 was formed by a simple amide
coupling to amine 49 (Scheme 11). With no vulnerable groups
in the molecule, hydrolysis to 33 was straightforward.
An enzyme-complexed structure of 26 was obtained using
the rat LDHA crystal system (Figure 9a). This confirmed the
binding mode predicted using the structures of the two
separate fragments (21 and 24) and modeling work on the
linker region. In agreement with our earlier structural data for
21 and 24, the linking group appeared to take a more direct
route through the phosphate channel than is the case for
NADH (Figure 9b), thereby both avoiding the need for a
phosphate isostere and minimizing molecular size. The
potency gained from joining fragments 24 (Kd = 160 μM)
The acid 70 required for core 3 was synthesized starting from
4-phenylbutanoic acid (Scheme 10) and was converted to the
ethyl ester 65 under acid catalysis. Duff reaction with
hexamethylenetetramine introduced the aldehyde group
selectively into the 4-position to give 66 in 31% yield (lit.
57%).57 An orthogonal protecting group strategy was needed,
3296
dx.doi.org/10.1021/jm201734r | J. Med. Chem. 2012, 55, 3285−3306

Journal of Medicinal Chemistry
Article
a
Scheme 7. Synthesis of Substrate Template for Carbon-Linked Core 2 Compounds
a
(i) Boc2O, NaOH, EtOH, H2O, rt, 3 days, 43%; (ii) CDI, DCM, rt, 2 h; (iii) MeOH, rt, 16 h, 84% (two steps); (iv) LiAlH4, THF, 0 °C, 2 h, 99%;
(v) MnO2, DCM, rt, 4 days, 88%; (vi) diethyl malonate, PS-trisamine, 4 Å molecular sieves, EtOH, 60 °C, 16 h, 33%; (vii) H2, 5% Pd−C, EtOH, rt,
16 h, 75%; (viii) HCl in 1,4-dioxane, DCM, rt, 16 h, quant.
a
Scheme 8. Synthesis of Substrate Template for Oxygen-Linked Core 2 Compounds
a
(i) Phenyl chloroformate, NaHCO3, 1,4-dioxane, water, 4 h, quant; (ii) amine 49, NMP, 70 °C, 24 h, 56%; (iii) dimethyl 2-chloromalonate, K2CO3,
DMF, rt, 2 days, 28%; (iv) NaOH, H2O, THF, rt, 20 min, 79%.
and 25 (1100 μM) into a linked molecule with Kd = 0.13 μM
(26) was over 3 orders of magnitude. Critically, for this link-
ing process the LE is broadly maintained at 0.19. There are
limited examples in the literature where fragment linking has
given improvements of, or even maintained, the ligand ef-
ficiency of the initial fragments,58−63 highlighting the
challenges inherent in this approach.
assay and was immediately optimized by exploiting our prior
knowledge of SAR to replace the methyl on the benzothiazole
with an S-propyl to give compound 27. This resulted in a
∼3-fold improvement in affinity as measured by BIACore.
Compound 28 was made to probe the impact of removing one
of the carboxylic acids from a linked compound. We observed a
decrease in affinity of more than an order of magnitude com-
pared to 26 (Kd = 3.2 μM for 28 versus Kd = 0.13 μM for 26)
along with a reduction in LE.
Apart from tetrahydro-NAD (8), compound 26 was the first
compound synthesized that showed activity in our enzyme
3297
dx.doi.org/10.1021/jm201734r | J. Med. Chem. 2012, 55, 3285−3306

Journal of Medicinal Chemistry
Article
a
Scheme 9. Improved Method for Fragment Linking of Carbon-Linked Core 2 Compounds
a
(i) 1,1-carbonyl diimidazole, DCM, rt, 2 h, 60%; (ii) amine 59·HCl, NMP, 24 h, 56%; (iii) NaOH, H2O, rt, 2 days, 76%.
a
Scheme 10. Synthesisis of Substrate Template for Core 3 Fragment Linking
a
(i) conc H2SO4, EtOH, reflux, 20 h, quant; (ii) hexamethylenetetramine, TFA, reflux, 16 h, 31%; (iii) LiOH·H2O, THF, water, rt, 16 h, 91%; (iv)
CDI, imidazole, rt, 3 h; (v) BnOH, rt, 16 h, 81% (two steps); (vi) diethyl malonate, PS-trisamine, 4 Å molecular sieves, toluene, reflux, 3 days, 54%;
(vii) H2, 5% Pd−C, EtOAc, rt, 16 h, quant.
a
Scheme 11. Fragment Linking of Core 3 Compounds
a
(i) HATU, DIPEA, DMF, rt, 16 h, 61%; (ii) NaOH, H2O, EtOH, rt, 2 days, 79%.
We subsequently made a number of other linker groups in
our search for improved affinity and LE and to broaden our
knowledge of the SAR. Core 2, with a shorter linking group,
looked particularly promising from a modeling perspective. The
BIAcore data showed a modest improvement in binding affinity
for one of the core 2 equivalents made (Kd = 0.13 μM for 26
compared with Kd = 0.069 μM for 29), although the binding
affinity of the second pair, compounds 27 (Kd = 0.046 μM) and
31 (Kd = 0.06 μM), remained broadly unchanged. However,
core 2 compounds 29 and 31 had improved enzyme activity
over their core 1 equivalents (26 and 27, respectively). Core 3
was made to assess the impact of replacing a nitrogen of the
urea with carbon, converting the group into an amide. This
resulted in a further increase in potency when compared with
core 1 (compounds 33 and 34). Compound 34 was both the
most tightly binding compound by BIAcore (Kd = 0.008 μM)
and the most potent compound in the enzyme assay (IC50
=
0.27 μM).
Across the data set in Table 6, replacement of methyl with
S-propyl gave a clear increase in the binding affinity in two of
the three matched pairs (26 compared with 27 and 33 compared
with 34) with LE in general maintained. For the third pair (29
and 31) although binding affinity is largely unchanged, compound
31 (IC50 = 0.29 μM) was more potent in the enzyme assay than
29 (IC50 = 1.9 μM).
The early trend from our fragment data suggested that the
carbon-linked malonate compounds in general bound more
tightly than the oxygen-linked compounds and therefore had
3298
dx.doi.org/10.1021/jm201734r | J. Med. Chem. 2012, 55, 3285−3306

Journal of Medicinal Chemistry
Article
observed for some examples require further consolidation by
testing with more physiologically relevant end points.
Examples of linked compounds with alternative cores (29
and 33) were soaked into the rat LDHA crystals and
subsequently cocrystallized with human LDHA. They main-
tained a consistent overall binding mode as modeled, spanning
the substrate, nicotinamide, and adenine pockets (Figure 11).
Figure 11. Binding mode of compounds 29 (purple) and 33 (pink)
compared with 26 (cyan).
Figure 9. (a) X-ray crystal structure of 26 (cyan), overlaid with parent
fragments (21 and 24, purple and pink, respectively). (b) X-ray crystal
structure of 26 (cyan), overlaid with NADH as bound in 1i1025 (pink).
Protein and water atoms are omitted for clarity.
Comparison of the binding modes of 26, 29, and 33 to rat
LDHA demonstrates the subtle effects on linker conformation
resulting from alterations in the length and chemistry of the
linker between the adenine and substrate binding site motifs.
The linked compounds recapitulate the binding modes
observed for the parent fragments, 21 and 24; however, in
order to accommodate the longer linker in 26, the benzyl
malonate of 26 is bound slightly deeper into the substrate
pocket and rotated by ∼8° with respect to that in 29 and 33.
An overlay of 26, 29, and 33 shows good agreement for the
adenine and urea moieties, after which the path taken by the
linker in 26 diverges from that of 29 and 33. Electron density
for this part of the linker region is consistent with a greater
degree of torsional flexibility in 26 than in the shorter linkers
of 29 and 33. These binding modes were confirmed to be
identical in human LDHA (Figure S5, Supporting Informa-
tion), thus validating our exploitation of the more robust rat
LDHA structural system as a surrogate.
improved LE. This was also observed for the linked
compounds. The carbon-linked versus oxygen-linked affinity
was investigated in core 2. The O-linked compounds bound less
tightly when comparing 29 (Kd = 0.069 μM) with 30 (1.2 μM)
and 31 (Kd = 0.06 μM) with 32 (0.18 μM). This pattern was
also observed in the enzyme activity of these compounds. In
general, there is good agreement between the BIAcore binding
data and enzyme assay for the linked compounds, with R2 = 0.86
(Figure 10). Note that only linked compounds showing
confirmed binding by BIAcore are included in Figure 10.
Series Optimization. We had successfully designed novel
LDHA inhibitors that bound tightly and showed potent enzyme
activity in our assays. However, no cellular activity was seen for
these compounds, which we assumed was due to the diacid
moiety preventing cell penetration. To identify compounds
with activity in cells, we synthesized and tested a set of mono-
and diesters based on the diacid structures in Table 6. Some
had already been made as precursors, while others were pre-
pared by transesterification. We expected that this modification,
which masks the acidic functionality, would allow the
compounds to penetrate cell membranes. Subsequently, the
cleavage of the esters into the active diacid form should occur
once the compounds were inside the cell.65 As we did not know
what level of ester stability would be optimal, we synthesized
both dimethyl and diethyl ester equivalents of the linked
compounds, dimethyl esters being intrinsically less chemically
stable than diethyl esters. Larger and more sterically hindered
ester groups were considered to further modulate ester stability.
However, the physical properties of these compounds would be
expected to be poor. In order to achieve activity in cells the
Figure 10. Correlation between BIAcore data (pKd) and enzyme
activity (pIC50) for linked compounds. Line of best fit was calculated
using least-squares regression.
A selection of compounds was tested in affinity-based LDHB
assays.64 Preliminary indications of selectivity over LDHB
3299
dx.doi.org/10.1021/jm201734r | J. Med. Chem. 2012, 55, 3285−3306

Journal of Medicinal Chemistry
Article
Figure 12. Binding affinity and cell activity comparison for carbon linked malonate diacids and diesters. The molecule substructure (diacid, diethyl,
or dimethyl ester) is labeled on the x-axis. The y-axis shows the BIAcore binding pKd for diacids and the cellular pIC50 for diesters. Data are the
average of at least two experiments.
a
esters should cleave quickly once across the cell membrane;
conversely, highly unstable esters may break down before enter-
ing the cell. This is particularly important when considering
how these compounds may behave in more physiologically
relevant environments. We assessed the cellular activity of
compounds by measuring their effect on lactate levels in the
culture media of SKBR3 cells. We did not see any consistent
activity with monoesters. In contrast, cellular activity was obser-
ved with a range of diesters, the most potent of which had
IC50 = 0.5 μM.
Scheme 12. Synthesis of Compound 35
a
(i) NaOMe, MeOH, rt, 5 h, 90%.
Figure 12 shows the BIAcore pKd to cell pIC50 relationship
for compounds that contain the carbon linked malonate group.
Three different substructures are plotted across the x-axis: the
diacid, the diethyl ester, and the dimethyl ester. On the y-axis
are BIAcore pKd data for the diacids or the cellular activity
(pIC50) if the compound is an ester. The lines indicate examples
of these scaffolds where the acid-to-ester structural change is the
only difference. This compound set shows a good BIAcore to cell
relationship and improved cellular activity of the dimethyl ester
when compared to the diethyl ester, which may be driven by the
increased rate of ester hydrolysis.
We selected compound 35 (Scheme 12), among other
diester examples, for further study as chemical probes. 35 was
synthesized by transesterification of 71 with NaOMe;
Knoevenagel reaction of aldehydes with dimethyl malonate
gave almost exclusively bis-addition and was of no synthetic
use. This specific compound had a reasonable combination of
activity in cells (IC50 = 4.8 μM) and satisfactory overall physical
properties. 35 is the dimethyl ester equivalent of compound 33,
which was one of the tightest binders synthesized with Kd =
0.093 μM by BIAcore and IC50= 0.5 μM in our enzyme assay.
CONCLUSIONS
■
We have reported on our FBLG approach against LDHA, which is
an emerging target in cancer metabolism. Fragment-based
approaches may prove to be a valuable way to identify inhibitors
against this target, this thinking is further supported by Moses
et al. recently publishing their work using a fragment approach.81
As part of our strategy, we successfully identified fragments binding
in the adenine and substrate regions of the active site. These
fragments were then grown toward each other through iterative
structure-based design and then linked to significantly increase
potency while maintaining ligand efficiency, which has been shown
to be a major challenge in the field of FBLG. The identified diacid
lead compounds were converted into their respective esters at
which point cellular activity was observed. To our knowledge these
compounds represent the first potent LDHA inhibitors that have
been fully validated through biophysical, enzymatic, and cellular
assays along with binding mode elucidation by X-ray crystallog-
raphy. Although these compounds require further optimization
to deliver a compound suitable for in vivo experiments, such as
modification/removal of the diacid functionality, the reported
chemical equity presents significant progress toward this goal.
3300
dx.doi.org/10.1021/jm201734r | J. Med. Chem. 2012, 55, 3285−3306

Journal of Medicinal Chemistry
Article
Buster (version 2.9.4, Global Phasing Ltd.) with intermediate rounds
of model building in Coot.71 The quality of the models was monitored
using MolProbity,72 Procheck,73−75 and Mogul.76 The final structures
have been deposited in the Protein Data Bank together with structure
factors and detailed experimental conditions (see Table S1, Supporting
Information for crystallographic statistics and PDB accession codes).
Additional Comments on Differences in Occupancy and Binding
Mode for Different Compounds among the Four Chains of the
LDHA Tetramer. Electron density for compounds 12 and 18 is best
defined in chain B and less well-defined in the other three chains. For
compound 14, additional electron density consistent with binding of
14 at a crystal contact was also noted; however, the quality of the
density was not sufficient to permit modeling of the compound at this
site. For compound 20, initial electron density for bound compound
was visible bridging the nicotinamide and substrate pockets of all four
chains of the tetramer in the crystal asymmetric unit; however, it was
more clearly defined in chains B and C than A and D. Interestingly, in
chains B and C the diacid portion of 20 binds in a manner identical to
that observed for malonate alone, while in chains A and D, where the
electron density for bound compound is poorer quality, the overlap is
less consistent. This is particularly clear in chain A, where residual
density consistent with bound malonate, 19 (∼30% occupancy), was
present. A difference in conformation of the Gln-99 side chain among
the four chains in the tetramer might provide an explanation. Steric
restrictions constrain 20 to bind in an identical manner to all four
chains, while malonate, being smaller, can adjust to the local
environment. For compounds 20 and 22 dual soak, consistent with
the results obtained for single soaks, compound 20 is only observed in
two of the four molecules in the asymmetric unit (chains B and C),
and in one of these (chain C) malonate is only partially displaced, so
both malonate and 20 are modeled at partial occupancy. For
compounds 21 and 22 dual soak, compound 21 is only observed in
two of the four molecules in the asymmetric unit (chains B and C).
For compound 26, positive difference density was observed on the
surface of the protein in the vicinity of the adenine pocket, consistent
with binding of additional copies of compound 26 at partial
occupancy. This density was best modeled as two alternative con-
formations of compound 26. Although this density was present in all
four copies of the LDHA monomer in the asymmetric unit, additional
copies of compound 26 have only been modeled in chains A, B, and C
where the density was strongest. We hypothesize that this additional
surface binding is an artifact of the high concentrations (2.5 mM) and
extended incubation times (72 h) used in the crystal soaking
experiment, as superstoichiometirc binding was not observed by
other techniques (BIAcore, ITC) or for other similar compounds
(compounds 29 and 33).
Nuclear Magnetic Resonance (NMR). Human LDHA(2−332)-
6His protein for 1D NMR was expressed and purified from E. coli cells
as described (methods S4, Supporting Information). For 1D (ligand-
observed) NMR studies, samples were prepared in 50 mM sodium
phosphate buffer, pH 7.0, 0.02% sodium azide. For 2D (protein-
observed) NMR studies, LDHA was uniformly labeled with 15N and
2H by growth in D2O-based minimal medium containing 15N-am-
monium chloride as the sole nitrogen source. Samples were prepared
in a buffer containing 100 mM HEPES, pH 7.5, 2 mM TCEP, 0.02%
sodium azide. For 1D studies, the concentration of LDHA was 40−48
μM monomer (10−12 μM tetramer), while for 2D studies, samples
containing 100 μM monomer (25 μM tetramer) were used. NMR
spectra were acquired on Bruker Avance 600 MHz spectrometers at
298 K, using a 5 mm triple-resonance HCN cryoprobe. Ligand binding
was detected using WaterLOGSY77 and T1ρ-filtered78 1D experiments
in combination with 2D 1H−15N TROSY-HSQC79 experiments to
provide binding-site and Kd information. NMR fragment screens were
carried out using mixtures of six compounds at 200 or 400 μM (final)
per compound. Active compounds in the mixtures were identified by
comparison with 1D reference spectra, and specific binding was
confirmed through addition of 1 mM NADH. Compounds whose
binding signal (WaterLOGSY or T1ρ-filtered) was appreciably reduced
in the presence of NADH were considered bona fide hits.
EXPERIMENTAL SECTION
■
X-ray Crystal Structures. Initial attempts to generate a soakable
crystal system using human LDHA were unsuccessful. Although two
different crystal forms of human LDHA were obtained, the active site
loop in both of these was held in an open conformation by crystal
contacts. Close inspection of the electron density revealed that the
substrate site in these crystals was occupied by either citrate (from the
crystallization buffer) or phosphate (from the purification buffer). In
addition, the crystals were very fragile, grew under acidic conditions
(pH 5.6), and diffracted to low resolution on an in-house source.
Attempts to soak cofactor (NADH) plus oxamate or fragments into
these crystals failed to yield complex crystal structures. Studies with rat
LDHA yielded a more robust crystal system in which the active site
loop adopted a closed conformation. These crystals grew under more
neutral conditions (pH 7.0), diffracted to high resolution (>2 Å) on an
in-house source, and were more amenable to soaking experiments. We
hypothesized that the high sequence similarity of the NADH pockets
would make the rat protein a suitable surrogate for human LDHA
(Figure S1, Supporting Information). In the later stages of the project,
extensive cocrystallization screening with a selection of more potent
compounds yielded conditions under which cocrystals of human
LDHA with linked compounds could be reliably produced, enabling us
to confirm the validity of the rat system as a structural surrogate.
Crystals of rat LDHA(2−332) were grown at 293K by hanging-
drop vapor diffusion by mixing purified protein at 20 mg/mL with a
precipitant consisting of 1.3−1.5 M sodium malonate, pH 7.0, 2%
glycerol in a 1:1 ratio to give a 4 μL drop. Drops were equilibrated
overnight and then seeded using a 1:2500 dilution of freshly prepared
seed stock (five crystals crushed in 10 μL of 1.6 M sodium malonate,
pH 7.0). Crystals appear within 3 days and continue to grow for a
further week. Complex structures were obtained by incubating these
crystals for periods of 1−4 days in either citrate or malonate based
soak buffer containing the compound of interest (2.5−10 mM) and
≤5% DMSO. Malonate based soak buffer (1.5 M sodium malonate,
pH 7.0, 10% glycerol) was generally used for compounds predicted to
bind at the adenine pocket, while citrate based soak buffer (1.3 M
trisodium citrate, 0.04 M HEPES, pH 7.0, 10% glycerol) was used for
compounds expected to bind at the substrate pocket. Crystals were
then flash cooled in a gaseous nitrogen stream at 100 K prior to data
collection. Rat LDHA(2−332) crystals belonged to the monoclinic
space group P21 with unit cell dimensions of (62.1 0.2) Å × (82.0
0.5) Å × (129.1 0.5) Å, β = (96.1 0.1)°. The asymmetric unit
comprises the biological tetramer.
Cocrystals of human LDHA(2−332)-6His with 33 were grown by
incubating purified protein at 27 mg/mL with 2 mM 33 (2% DMSO)
on ice for 30 min and then setting up a hanging-drop vapor diffusion
experiment at 293 K using a 1:1 ratio of reservoir (0.1 M MES, pH 6.5,
1.6 M ammonium sulfate, 10% v/v 1,4-dioxane) to protein +
compound to give a 2 μL drop. A narrow grid screen around these
conditions was subsequently used successfully to generate cocrystal
structures with several other compounds of similar chemotype.
Crystals were cryoprotected by quickly passing through a solution
containing 2.5 M ammonium sulfate, 0.1 M Tris-HCl, pH 6.5, 10%
glycerol and flash cooled in a cryostream at 100 K prior to data
collection. Cocrystals of human LDHA(2−332)-6His with 33
belonged to the monoclinic space group P21 with unit cell dimensions
of 64.6 Å × 137.4 Å × 84.1 Å, β = 102.1°. The asymmetric unit
comprises the biological tetramer.
Diffraction data were generally collected in-house using a Rigaku
FRE equipped with Osmic optics and a Saturn944 CCD detector at
100 K. Data for the rat LDHA−28 complex and the human LDHA−
36 structure were collected at the ESRF on beamlines ID23-EH1 and
ID29, respectively. Data processing was carried out using d*TREK66
or XDS67 as implemented within autoPROC.68 Data reduction and
structure solution by molecular replacement (initially using PDB code
1i1025 as a search model) were carried out using programs from the
CCP4 suite.69 Compounds were modeled into the electron density
using Flynn (version 1.1.1, Openeye Scientific Software, Inc.). The
protein−compound complex model was refined using Refmac70 and
3301
dx.doi.org/10.1021/jm201734r | J. Med. Chem. 2012, 55, 3285−3306

Journal of Medicinal Chemistry
Article
Surface Plasmon Resonance (SPR). A BIAcore 3000 or a
BIAcore S51 instrument (GE Healthcare) was used to detect binding
interactions using a direct binding assay format. Prior to activation, the
research grade CM5 chip surface was preconditioned using two 50 μL
injections each of 10 mM HCl, 50 mM NaOH, 0.1% SDS, and 0.085%
H3PO4 at a flow rate of 100 μL/min. LDHA was immobilized on the
sensor surface using standard amine coupling This was achieved
by activating the sensor surface using 7 min injections of a mixture of
11.5 mg/mL N-hydroxysuccinimide with 75 mg/mL 1-ethyl-3-(3-
dimethylaminopropyl)carbodiimide hydrochloride. Protein immobili-
zation was accomplished using a 10 min injection of hLDHA-6Lys-
6His (500 μg/mL) or hLDHA(2−332)-6His (100 μg/mL) in 20 mM
MES, pH 6.0 buffer. Remaining reactive esters were blocked using a
7 min injection of 1 M ethanolamine, pH 8.5, at a flow rate of 5 μL/min.
Immobilization levels typically were around 15 000 resonance units.
Reference flow cells were prepared without the protein, as well as with
a denatured protein surface. Denaturation was achieved by a 30 s
injection of 10 mM HCl at a flow rate of 30 μL/min. All binding
measurements were performed in 20 mM Na phosphate, 150 mM
NaCl, 0.005% v/v P20, 5% DMSO, pH 7.4, at a flow rate of 30 μL/min.
Compounds were injected over the active protein and two refer-
ence surfaces with at least 30 s association and dissociation times.
Solvent calibration and double referencing subtractions were made to
eliminate refractive index change and injection noise using Scrubber
2.0 (BioLogic Software). Surface regeneration was achieved using
dissociation for a time period allowing the response to return to
baseline. Control injections of a fixed, saturating NADH concentration
of 100 μM were interspersed with injections of compound to allow
monitoring of the functionality of the protein surface. SPR equilibrium
binding data, consisting of Req values from 8−10 point concentration
series, were analyzed by fitting a simple 1:1 binding model to yield
Rmax and Kd values using Grafit 6 (Erithacus Software).
assay, cells were harvested and resuspended in a phenol red free RPMI
1640 (Sigma, catalog no. R7509) medium supplemented with 0.1%
FCS and 1% (v/v) 200 mM L-glutamate. The cell number was
estimated to be ∼1 × 105 cells per mL. Compounds or controls were
dispensed into black clear-bottom tissue culture treated 384-well plates
(Costar, catalog no. 3712) as previously described. After the addition
of the compounds, 40 μL of cell suspension was added to each well
and the plates were transferred to a standard tissue culture incubator
for 4 h. Subsequently, 5 μL of culture medium was removed from each
well into a second black 384-well plate. The level of lactate present in
the 5 μL of culture medium was estimated using a lactate kit supplied
by Trinity Biotech (catalog no. 735-10). To each well 30 μL of
detection reagent was added, and plates were then stored in the dark
for 10 min before reading absorbance at 450 nm using an Envision
plate reader (Perkin-Elmer).
Calculation of IC50 Values. By use of the maximum and minimum
control wells as references for the 0% and 100% enzyme inhibition
points, it was possible to calculate the effect of each compound on the
activity of the enzyme. OriginLab software was used to estimate the
concentration of compound required to reduce the enzyme activity
to 50%.
Chemistry. General Statements. 1H NMR spectra were
measured on a Bruker DPX-400 Ultrashield. Chemical shifts δ are
reported relative to CDCl3 at 7.26 ppm or DMSO-d6 at 2.52 ppm as
an internal standard. LC−MS spectra were obtained on a Waters
2790/ZMD LC−MS system equipped with a Waters 3100 mass
spectrometer, a Waters Acquity photodiode array detector ranging
from 200 to 320 nm, and a Phenomenex Gemini C18 110A column,
50 mm × 2.00 mm, with 5 μm silica, using a 5 min gradient of 5−95%
MeCN in water. An amount of 0.1% (0.880 g/mL density) aqueous
NH3 was present as a basic modifier unless otherwise stated. All tested
1
compounds were purified to >95% purity as determined by H NMR
Biological Testing. Compound Preparation. Compounds
were dispensed into black low volume 384-well plates (Griener
Bio One, U.K., catalog no. 784076) using a Labcyte Echo 550
as previously described.80 The test range was dependent on the
stock concentration of compound but was typically across 12
points, using half log intervals with the exception of the last
point which was a whole log interval from the 11th point. Each
well was backfilled with the required volume of dimethyl
sulfoxide (DMSO) to ensure a final DMSO concentration of
1% after addition of all reagents. In addition to the compound
test wells each plate carried maximum and minimum controls.
A compound known to inhibit LDHA was used as a maximum
control. As a minimum control pure DMSO was added to wells.
Coupled Biochemical Enzyme Assay. After the addition of
compounds to the assay plates 6 μL/well substrate solution was
added followed by 6 μL/well enzyme solution. The substrate solution
contained the following: 0.2 mM sodium pyruvate (Sigma, U.K.,
catalog no. P8574); 50 mM Tris, pH7.5; 50 mM KCl; 0.17% (v/v)
Triton X100 (Sigma, U.K., catalog no. X100). The enzyme solution
(comprising purified full length human recombinant LDHA at a
working concentration of 0.012 μg/mL) contained 50 mM Tris, pH
7.5, 50 mM KCl, 0.17% (v/v) Triton X100, 0.06 mM NADH (Sigma,
U.K., catalog no. N8129). The NADH solution was made fresh, as
long-term storage can cause oxidation of the reduced reagent to
NAD+. Assay plates were incubated at room temperature (21 °C) for
60 min before the addition of 8 μL/well detection reagents. The
detection reagent contained 5.03 U/mL diaphorase (Worthington
Biochemcial Corp., NJ, U.S., catalog no. LS004330), 0.037 mM
resazurin (Sigma, U.K., catalog no. R7017) made up in 50 mM Tris,
pH 7.5, 50 mM KCl, 0.17% (v/v) Triton X100. Detection reagents
were allowed to develop for 10 min before the plates were read using a
Pherastar plate reader (BMG Labtech Ltd., U.K.). The reader required
an excitation filter at a wavelength of 485 nm and emission filter at a
wavelength of 540 nm.
and by HPLC total ion current over an array of wavelengths from 200
to 320 nm. NMR and mass spectra were run on isolated products and
were consistent with the proposed structures. Normal phase column
chromatography was performed on prepacked silica gel columns using
an ISCO CombiFlash system.
Synthesis of Compounds 33 and 35. Ethyl 4-Phenyl-
butanoate 65.56 Sulfuric acid (4.0 mL, 122 mmol) was added
to a solution of 4-phenylbutanoic acid (20.0 g, 122 mmol) in
ethanol (150 mL) and the mixture heated at reflux for 20 h.
The mixture was concentrated in vacuo and the residue dis-
solved in ethyl acetate (400 mL), washed with saturated
NaHCO3 solution (4 × 150 mL), brine (100 mL), dried over
MgSO4, and evaporated to dryness to afford ethyl 4-phenyl-
1
butanoate (23.11 g, 99%) as an orange liquid. H NMR (400
MHz, CDCl3, 30 °C) δ 1.25 (3H, t), 1.95 (1H, tt), 2.31 (2H, t),
2.65 (2H, t), 4.12 (2H, q), 7.16−7.21 (3H, m), 7.28 (2H, ddd).
Ethyl 4-(4-Formylphenyl)butanoate 66. Hexamethylenetetramine
(13.72 g, 97.84 mmol) was added to a solution of ethyl 4-
phenylbutanoate 65 (17.1 g, 88.94 mmol) in TFA (88 mL) at 22 °C.
The mixture was heated at 80 °C (internal temperature) under
reflux for 16 h. The mixture was concentrated in vacuo and the residue
partitioned between diethyl ether (150 mL) and saturated sodium
hydrogen carbonate solution (100 mL). The aqueous layer was
extracted with diethyl ether (3 × 100 mL), and the extracts were
combined with the organic layer. The combined extracts were dried
over MgSO4 and evaporated to dryness. The residue was purified by
flash silica chromatography. The elution solvent was 8% EtOAc in
isohexane. Pure fractions were evaporated to dryness to afford ethyl
4-(4-formylphenyl)butanoate (6.11 g, 31.2%) as a colorless oil. 1H NMR
(400 MHz, CDCl3, 30 °C) δ 1.26 (3H, t), 1.99 (2H, tt), 2.33 (2H, t),
2.74 (2H, t), 4.13 (2H, q), 7.35 (2H, d), 7.81 (2H, d), 9.98 (1H, s).
HPLC tR = 2.24 min.
4-(4-Formylphenyl)butanoic Acid 67. Lithium hydroxide hydrate
(4.17 g, 99.43 mmol) was added as a solution in water (100 mL) to a
solution of ethyl 4-(4-formylphenyl)butanoate 66 (7.30 g, 33.14
mmol) in THF (100 mL) and ethanol (40 mL) at 22 °C, and the
mixture was stirred for 16 h. The mixture was concentrated to a
volume such that all of the organic solvent had been removed, diluted
Measurement of Cellular LDHA Activity. The SKBR3 cells were
grown and passaged using RPMI 1640 (Sigma, R0883) medium
supplemented with 20% fetal calf serum (FCS) and 1% (v/v) 200 mM
L-glutamate (Invitrogen, catalog no. 25030-024). Prior to running the
3302
dx.doi.org/10.1021/jm201734r | J. Med. Chem. 2012, 55, 3285−3306

Journal of Medicinal Chemistry
Article
with water (100 mL), and washed with diethyl ether (2 × 100 mL).
The aqueous layer was acidified to pH 2 with 2 M HCl and extracted
with ethyl acetate (3 × 150 mL). The combined extracts were washed
with saturated brine (50 mL), dried over Na2SO4, and evaporated to
afford impure 4-(4-formylphenyl)butanoic acid as a yellow solid. 4-(3-
Carboxypropyl)benzoic acid was present as a 8% impurity (from air
oxidation).
mixture contained 203 mg of 70, 0.60 mmol), 3-amino-N-(2-
methylbenzo[d]thiazol-6-yl)propanamide 49 (156 mg, 0.66 mmol),
and DIPEA (0.137 mL, 0.78 mmol) in DMF (10 mL). The mixture
was stirred at ambient temperature for 16 h. The mixture was
evaporated and the residue partitioned between ethyl acetate (70 mL)
and water (70 mL). The organic layer was washed sequentially with
0.1 M NH4Cl solution (70 mL), water (2 × 70 mL), and saturated
brine (30 mL), dried over Na2SO4, and evaporated. The residue was
purified by flash silica chromatography, and elution gradient was 1−5%
EtOH in DCM. Product-containing fractions were evaporated to
dryness to afford 76% pure product as a colorless gum, contaminated
with 23% N-(3-(2-methylbenzo[d]thiazol-6-ylamino)-3-oxopropyl)-4-
p-tolylbutanamide. The crude product was stirred in diethyl ether
for 16 h at 22 °C. The resulting white solid was collected by filtration,
giving 85% pure material. This was crystallized from EtOAc/isohexane
to give 92% pure material and recrystallized to give 95% pure diethyl
2-(4-(4-(3-(2-methylbenzo[d]thiazol-6-ylamino)-3-oxopropylamino)-
4-oxobutyl)benzyl)malonate (202 mg, 60.5%) as a white crystalline
solid. 1H NMR (400 MHz, DMSO, 30 °C) δ 1.11 (6H, t), 1.68−1.82
(2H, m), 2.07 (2H, t), 2.48−2.60 (4H, m), 2.76 (3H, s), 3.03 (2H, d),
3.24−3.30 (2H, m), 3.36 (2H, dt), 3.75 (1H, t), 3.99−4.15 (4H, m),
6.98−7.15 (4H, m), 7.51 (1H, dd), 7.81 (1H, d), 7.89 (1H, t), 8.40
(1H, s), 10.11 (1H, s). m/z (ES+), (M + H)+ = 554.3; 5 min acidic
method, HPLC tR = 2.25 min.
The impure solid was dissolved in DCM (200 mL) and sonicated.
The resulting suspension was filtered, removing the white solid
impurity 4-(3-carboxypropyl)benzoic acid (0.146 g). The filtrate was
evaporated to afford 4-(4-formylphenyl)butanoic acid (5.81 g, 91%) as
1
a pale yellow solid. H NMR (400 MHz, DMSO-d6, 30 °C) δ 1.84
(2H, tt), 2.24 (2H, t), 2.71 (2H, t), 7.44 (2H, d), 7.84 (2H, d), 9.98
(1H, s), 12.04 (1H, s). m/z (ES−), (M − H)− = 191.3; 5 min acidic
method, HPLC tR = 1.52 min.
Benzyl 4-(4-Formylphenyl)butanoate 68. 4-(4-Formylphenyl)-
butanoic acid 67 (3.00 g, 15.61 mmol) was dissolved in DCM (200
mL), and 1H-imidazole (0.213 g, 3.12 mmol) and 1,1′-carbonyl
diimidazole (3.04 g, 18.73 mmol) were added. The mixture was stirred
at 22 °C for 3 h. Then benzyl alcohol (6.47 mL, 62.43 mmol) was
added. The mixture was stirred at 22 °C for 16 h. The mixture was
evaporated, and the residue was purified by flash silica chromatog-
raphy. Elution gradient was 20% to 30% EtOAc in isohexane. Pure
fractions were evaporated to dryness to afford benzyl 4-(4-
1
formylphenyl)butanoate (3.57 g, 81%) as a colorless oil. H NMR
(400 MHz, CDCl3, 30 °C) δ 2.01 (2H, tt), 2.39 (2H, t), 2.73 (2H, t),
5.12 (2H, s), 7.28−7.40 (7H, m), 7.79 (2H, d), 9.97 (1H, s). m/z (ES+),
(M + Na)+ = 305.5; 5 min acidic method, HPLC tR = 2.68 min.
Diethyl 2-(4-(4-(Benzyloxy)-4-oxobutyl)benzylidene)malonate
69. A mixture of benzyl 4-(4-formylphenyl)butanoate 68 (500 mg,
1.77 mmol), diethyl malonate (0.296 mL, 1.95 mmol), PS-trisamine
(purchased from Novabiochem, 100 mg, 0.42 mmol), and powdered
activated 4 Å molecular sieves (200 mg) was suspended in toluene
(20 mL). The mixture was heated at 50 °C for 16 h. Additional diethyl
malonate (0.296 mL, 1.95 mmol), PS-trisamine (100 mg, 0.42 mmol),
and powdered activated 4 Å molecular sieves (200 mg, 1.77 mmol)
were added, and the mixture was stirred at 60 °C for a further 2 days.
A 2:1 mixture of product/starting aldehyde was observed. The mixture
was cooled to ambient temperature, filtered, and evaporated to
dryness. The residue was purified by flash silica chromatography.
Elution gradient was 15% to 25% EtOAc in isohexane. Complete
separation was not achieved. Product-containing fractions were
evaporated to afford a 83:17 mixture of diethyl 2-(4-(4-(benzyloxy)-
4-oxobutyl)benzylidene)malonate (405 mg, 53.9%) and starting
2-(4-(4-(3-(2-Methylbenzo[d]thiazol-6-ylamino)-3-oxopropylami-
no)-4-oxobutyl)benzyl)malonic Acid 33. Sodium hydroxide (11 mg,
0.27 mmol) was added as a solution in water (5 mL) to a solution of
diethyl 2-(4-(4-(3-(2-methylbenzo[d]thiazol-6-ylamino)-3-oxopropy-
lamino)-4-oxobutyl)benzyl)malonate 71 (38 mg, 0.07 mmol) in
ethanol (5 mL) at 22 °C, and the mixture was stirred for 16 h,
giving mostly monoacid. The mixture was concentrated to a volume
such that all of the ethanol had been removed, diluted with water
(5 mL), and stirred for a further 3 days at 22 °C. The mixture was
diluted with water (20 mL) and washed with diethyl ether (20 mL).
The aqueous layer was acidified to pH 2 with 2 M HCl and extracted with
ethyl acetate (4 × 20 mL). The combined organics were washed with
saturated brine (15 mL), dried over Na2SO4, evaporated and the
residue was triturated with diethyl ether to afford 2-(4-(4-(3-(2-
methylbenzo[d]thiazol-6-ylamino)-3-oxopropylamino)-4-oxobutyl)-
1
benzyl)malonic acid (27 mg, 79%) as a white crystalline solid. H
NMR (400 MHz, DMSO, 30 °C) δ 1.77 (2H, tt), 2.07 (2H, t), 2.45−
2.57 (4H, m), 2.76 (3H, s), 2.98 (2H, d), 3.36 (2H, dt), 3.52 (1H, t),
7.05 (2H, d), 7.09 (2H, d), 7.51 (1H, dd), 7.81 (1H, d), 7.89 (1H, t),
8.39 (1H, d), 10.11 (1H, s), 12.65 (2H, s). m/z (ES+), (M + H)+ =
498.5; HPLC tR = 1.65 min.
1
aldehyde 68, as a colorless gum. H NMR (400 MHz, CDCl3, 30 °C)
δ 1.29 (3H, t), 1.33 (3H, t), 1.98 (2H, tt), 2.37 (2H, t), 2.65 (2H, t),
4.30 (2H, q), 4.33 (2H, q), 5.11 (2H, s), 7.16 (2H, d), 7.37 (2H, d),
7.69 (1H, s). m/z (ES+), (M + Na)+ = 447.2; 5 min acidic method,
HPLC tR = 3.21 min.
Dimethyl 2-(4-(4-(3-(2-Methylbenzo[d]thiazol-6-ylamino)-3-oxo-
propylamino)-4-oxobutyl)benzyl)malonate 35. Sodium metal (8.3
mg, 0.36 mmol) was added to methanol (10 mL), and the mixture was
stirred until effervessence had ceased. Diethyl 2-(4-(4-(3-(2-methyl-
benzo[d]thiazol-6-ylamino)-3-oxopropylamino)-4-oxobutyl)benzyl)-
malonate 33 (100 mg, 0.18 mmol) was added and the mixture stirred
at 22 °C for 5 h. The mixture was quenched with saturated NH4Cl
solution (1 mL) and evaporated to a volume such that all of the
methanol had been removed. The residue was partitioned between
ethyl acetate (20 mL) and water (20 mL). The aqueous layer was
extracted with ethyl acetate (2 × 20 mL), and the extracts were
combined with the organic layer. The combined extracts were washed
with saturated brine (20 mL), dried over Na2SO4, and evaporated to
dryness. The residue was triturated with diethyl ether to afford
dimethyl 2-(4-(4-(3-(2-methylbenzo[d]thiazol-6-ylamino)-3-oxopro-
pylamino)-4-oxobutyl)benzyl)malonate (85 mg, 90%) as a white
4-(4-(3-Ethoxy-2-(ethoxycarbonyl)-3-oxopropyl)phenyl)butanoic
Acid 70. Impure diethyl 2-(4-(4-(benzyloxy)-4-oxobutyl)-
benzylidene)malonate 69 (contaminated with aldehyde 68, mixture
contained 352 mg, 0.83 mmol of 69) was dissolved in ethyl acetate
(25 mL), and the mixture was degassed and purged with nitrogen.
Then 5% palladium on carbon (70.6 mg, 0.03 mmol) was added and
the mixture degassed and purged with hydrogen. The mixture was
stirred under a balloon of hydrogen for 16 h. The mixture was
degassed and purged with nitrogen, filtered, and evaporated to afford a
87:13 mixture of 4-(4-(3-ethoxy-2-(ethoxycarbonyl)-3-oxopropyl)phenyl)-
butanoic acid and 4-p-tolylbutanoic acid (331 mg) as a colorless oil, which
1
was carried through to the next stage without further purification. H
NMR (400 MHz, DMSO-d6, 30 °C) δ 1.11 (6H, t), 1.72−1.81 (2H, m),
2.19 (2H, t), 2.51−2.59 (2H, m), 3.05 (2H, d), 3.78 (1H, t), 4.00−4.14
(4H, m), 7.07−7.15 (4H, m), 12.00 (1H, s). m/z (ES−), (M − H)− =
335.1; 5 min acidic method, HPLC tR = 2.29 min.
2-(4-(4-(3-(2-Methylbenzo[d]thiazol-6-ylamino)-3-oxopropylami-
no)-4-oxobutyl)benzyl)malonate 71. HATU (275 mg, 0.72 mmol)
was added dropwise as a solution in DMF (5 mL) to a solution of
impure 4-(4-(3-ethoxy-2-(ethoxycarbonyl)-3-oxopropyl)phenyl)-
butanoic acid 70 (contaminated with 4-(p-tolyl)butanoic acid; impure
1
crystalline solid. H NMR (400 MHz, DMSO, 30 °C) δ 1.68−1.83
(2H, m), 2.07 (2H, t), 2.45−2.57 (4H, m), 2.76 (3H, s), 3.04 (2H, d),
3.36 (2H, dt), 3.61 (6H, s), 3.80 (1H, t), 7.02−7.10 (4H, m), 7.51
(1H, dd), 7.81 (1H, d), 7.89 (1H, t), 8.40 (1H, d), 10.11 (1H, s). m/z
(ES+), (M + H)+ = 526.5; HPLC tR = 2.00 min.
3303
dx.doi.org/10.1021/jm201734r | J. Med. Chem. 2012, 55, 3285−3306

Journal of Medicinal Chemistry
Article
(6) Koukourakis, M. I.; Giatromanolaki, A.; Sivridis, E.; Gatter, K. C.;
Harris, A. L. Lactate dehydrogenase 5 expression in operable colorectal
cancer: strong association with survival and activated vascular
endothelial growth factor pathwaya report of the tumour angio-
genesis research group. J. Clin. Oncol. 2006, 24, 4301−4308.
(7) Koukourakis, M. I.; Giatromanolaki, A.; Sivridis, E.; Bougioukas,
G.; Didilis, V.; Gatter, K. C.; Harris, A. L. Lactate dehydrogenase-5
(LDH-5) overexpression in non-small-cell lung cancer tissues is linked
to tumour hypoxia, angiogenic factor production and poor prognosis.
Br. J. Cancer 2003, 89, 877−885.
(8) Le, A.; Cooper, C. R.; Gouw, A. M.; Dinavahi, R.; Maitra, A.;
Deck, L. M.; Royer, R. E.; Vander Jagt, D. L.; Semenza, G. L.; Dang,
C. V. Inhibition of lactate dehydrogenase A induces oxidative stress
and inhibits tumor progression. Proc. Natl. Acad. Sci. U.S.A. 2010, 107,
2037−2042.
(9) Granchi, C.; Roy, S.; Giacomelli, C.; Macchia, M.; Tuccinardi, T.;
Martinelli, A.; Lanza, M.; Betti, L.; Giannaccini, G.; Lucacchini, A.;
Funel, N.; Leon, L. G.; Giovannetti, E.; Peters, G. J.; Palchaudhuri, R.;
Calvaresi, E. C.; Hergenrother, P. J.; Minutolo, F. Discovery of N-
hydroxyindole-based inhibitors of human lactate dehydrogenase
isoform A (LDH-A) as starvation agents against cancer cells. J. Med.
Chem. 2011, 54, 1599−1612.
(10) Javed, M. H.; Waqar, M. A. Effect of various metals and
metabolites on the activity of lactate dehydrogenase. Pak. J. Pharmacol.
1993, 10, 7−14.
(11) Rees, D. C.; Congreve, M.; Murray, C. W.; Carr, R. Fragment-
based lead discovery. Nat. Rev. Drug Discovery 2004, 3, 660−672.
(12) Albert, J. S.; Blomberg, N.; Breeze, A. L.; Brown, A. J. H.;
Burrows, J. N.; Edwards, P. D.; Folmer, R. H. A.; Geschwindner, S.;
Griffen, E. J.; Kenny, P. W.; Nowak, T.; Olsson, L.; Sanganee, H.;
Shapiro, A. B. An integrated approach to fragment-based lead
generation: philosophy, strategy and case studies from AstraZeneca’s
drug discovery programmes. Curr. Top. Med. Chem. (Sharjah, United
Arab Emirates) 2007, 7, 1600−1629.
(13) Woodhead, A. J.; Angove, H.; Carr, M. G.; Chessari, G.;
Congreve, M.; Coyle, J. E.; Cosme, J.; Graham, B.; Day, P. J.;
Downham, R.; Fazal, L.; Feltell, R.; Figueroa, E.; Frederickson, M.;
Lewis, J.; McMenamin, R.; Murray, C. W.; O’Brien, M. A.; Parra, L.;
Patel, S.; Phillips, T.; Rees, D. C.; Rich, S.; Smith, D.; Trewartha, G.;
Vinkovic, M.; Williams, B.; Woolford, A. J. Discovery of (2,4-
dihydroxy-5-isopropylphenyl)-[5-(4-methylpiperazin-1-ylmethyl)-1,3-
dihydroisoindol-2-yl]methanone (AT13387), a novel inhibitor of the
molecular chaperone Hsp90 by fragment based drug design. J. Med.
Chem. 2010, 53, 5956−5969.
(14) Huth, J. R.; Park, C.; Petros, A. M.; Kunzer, A. R.; Wendt, M. D.;
Wang, X.; Lynch, C. L.; Mack, J. C.; Swift, K. M.; Judge, R. A.;
Chen, J.; Richardson, P. L.; Jin, S.; Tahir, S. K.; Matayoshi, E. D.;
Dorwin, S. A.; Ladror, U. S.; Severin, J. M.; Walter, K. A.; Bartley,
D. M.; Fesik, S. W.; Elmore, S. W.; Hajduk, P. J. Discovery and design
of novel HSP90 inhibitors using multiple fragment-based design
strategies. Chem. Biol. Drug Des. 2007, 70, 1−12.
(15) Brough, P. A.; Barril, X.; Borgognoni, J.; Chene, P.; Davies,
N. G. M.; Davis, B.; Drysdale, M. J.; Dymock, B.; Eccles, S. A.;
Garcia-Echeverria, C.; Fromont, C.; Hayes, A.; Hubbard, R. E.; Jordan,
A. M.; Jensen, M. R.; Massey, A.; Merrett, A.; Padfield, A.; Parsons, R.;
Radimerski, T.; Raynaud, F. I.; Robertson, A.; Roughley, S. D.;
Schoepfer, J.; Simmonite, H.; Sharp, S. Y.; Surgenor, A.; Valenti, M.;
Walls, S.; Webb, P.; Wood, M.; Workman, P.; Wright, L. Combining
hit identification strategies: fragment-based and in silico approaches to
orally active 2-aminothieno[2,3-d]pyrimidine inhibitors of the Hsp90
molecular chaperone. J. Med. Chem. 2009, 52, 4794−4809.
(16) Barker, J. J.; Barker, O.; Boggio, R.; Chauhan, V.; Cheng, R. K. Y.;
Corden, V.; Courtney, S. M.; Edwards, N.; Falque, V. M.; Fusar, F.;
Gardiner, M.; Hamelin, E. M. N.; Hesterkamp, T.; Ichihara, O.; Jones,
R. S.; Mather, O.; Mercurio, C.; Minucci, S.; Montalbetti, C. A. G. N.;
Muller, A.; Patel, D.; Phillips, B. G.; Varasi, M.; Whittaker, M.;
Winkler, D.; Yarnold, C. J. Fragment-based identification of Hsp90
inhibitors. ChemMedChem 2009, 4, 963−966.
ASSOCIATED CONTENT
* Supporting Information
Additional information on computational methodology, protein
purification, sequence analyses, protein−ligand crystal struc-
tures, and compound synthesis. This material is available free of
charge via the Internet at http://pubs.acs.org.
■
S
Accession Codes
†The following crystal structures have been deposited in the
RCSB Protein Data Bank: codes 4aj1, 4aj2, 4aj4, 4aje, 4aji, 4ajk,
4ajh, 4ajj, 4ajl, 4al4, 4ajh, 4ajo, and 4ajp.
AUTHOR INFORMATION
Corresponding Author
*Phone: +44 (0)1625 519045. E-mail: richard.a.ward@
astrazeneca.com.
■
Present Address
‡Merck KGaA, Frankfurter Strasse 250, D-64293 Darmstadt,
Germany.
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
The authors acknowledge Ann-Kathrin Schott, Isabelle Green,
and Jon Read for construct design; Graham Sproat for
molecular biology; Judith Stanway, Richard Ernill, and Linda
MacCallum for protein expression; Ian Hardern for protein
purification; Helen McMiken for crystallization screening and
initial crystallization optimization experiments; Judit Debreczeni,
Andrew Ferguson, Marie Fraser, and Joe Patel for structure
refinements; the ESRF MXpress team for data collection; Chris
De Savi, Ian Hollingsworth, Alice Hooper, Dave Perkins, Matt
Addie, Matt Grist, Jo Allen, and Kelly-Marie Johnson for
chemistry support; Nicky Whalley and Tom Dunkley for
compound profiling; Martin Watson for SPR and NMR
support; Clare Lane, Alison Hunter, and Kin Tam for
pharmacokinetic (PK) and physical property assessment.
ABBREVIATIONS USED
■
LDH, lactate dehydrogenase; LDHA, lactate dehydrogenase A;
LDHB, lactate dehydrogenase B; FBLG, fragment-based lead
generation; NMR, nuclear magnetic resonance; SPR, surface
plasmon resonance; HTS, high throughput screening; LE,
ligand efficiency; LLE, ligand lipophilicity efficiency
REFERENCES
■
(1) Warburg, O. On the origin of cancer cells. Science 1956, 123,
309−314.
(2) Kim, J.; Dang, C. V. Cancer’s molecular sweet tooth and the
Warburg effect. Cancer Res. 2006, 66, 8927−8930.
(3) Zhuang, L.; Scolyer, R. A.; Murali, R.; McCarthy, S. W.; Zhang,
X. D.; Thompson, J. F.; Hersey, P. Lactate dehydrogenase 5 expression
in melanoma increases with disease progression and is associated with
expression of Bcl-XL and Mcl-1, but not Bcl-2 proteins. Mod. Pathol.
2010, 23, 45−53.
(4) Koukourakis, M. I.; Giatromanolaki, A.; Winter, S.; Leek, R.;
Sivridis, E.; Harris, A. L. Lactate dehydrogenase 5 expression in
squamous cell head and neck cancer relates to prognosis following
radical or postoperative radiotherapy. Oncology 2009, 77, 285−292.
(5) Kolev, Y.; Uetake, H.; Takagi, Y.; Sugihara, K. Lactate
dehydrogenase-5 (LDH-5) expression in human gastric cancer:
association with hypoxia-inducible factor (HIF-1alpha) pathway,
angiogenic factors production and poor prognosis. Ann. Surg. Oncol.
2008, 15, 2336−2344.
3304
dx.doi.org/10.1021/jm201734r | J. Med. Chem. 2012, 55, 3285−3306

Journal of Medicinal Chemistry
Article
(17) Cheng, Y.; Ashton, K.; Bartberger, M.; Brown, J.; Bryan, C.;
Chen, K.; Esmay, J.; Fremeau, R. T.; Golden, J.; Harried, S.; Hickman,
D.; Hitchcock, S. A.; Huang, M.; Jordan, B.; Judd, T.; Lopez, P.; Louie,
S. W.; Luo, Y.; Mercede, S.; Michelsen, K.; Nixey, T.; Paras, N. A.;
Poon, S.; Powers, T. S.; Sickmier, A.; St. Jean, D.; Tegley, C.; Zhong,
W.; Wen, P.; Wahl, R. C.; Wood, S.; Xue, M.; Yang, B. From
Fragment-Based Lead Generation to 2-Aminoquinolines as Potent
beta-Secretase Inhibitors That Are Efficacious in Vivo. Presented at the
240th National Meeting of the American Chemical Society, Boston,
MA, Aug 22−36, 2010; MEDI-288.
(18) Albert, J. S. Fragment-Based Lead Generation and Structure-
Based Design for the Discovery of High Affinity Beta-Secretase
Inhibitors. Presented at Frontiers in CNS and Oncology Medicinal
Chemistry, ACS-EFMC, Siena, Italy, October 7−9, 2007; COMC-005.
(19) Albert, J. S. Discovery of High Affinity beta-Secretase Inhibitors
Using Fragment-Based Lead Generation and Structure-Based Design.
Presented at the 234th National Meeting of the American Chemical
Society, Boston, MA, Aug 19−23, 2007; MEDI-243.
(20) Albert, J. S.; Edwards, P. D.; Andisik, D.; Campbell, J. B.;
Congreve, M. S.; Carr, R.; Chessari, G.; Edfeldt, F.; Folmer, R. H. A.;
Koether, G. M.; Kolmodin, K.; Murray, C. W.; Olsson, L.; Patel, S.;
DeBeer, T. Fragment Based Lead Generation Approaches for
Inhibitors of beta-Secretase: Development of a Novel Series of
Isocytosine-Based Inhibitors. Presented at the 233rd National Meeting
of the American Chemical Society, Chicago, IL, Mar 25−29, 2007;
MEDI-311.
dehydrogenases: crystal structure of a ternary complex of porcine
cytoplasmic malate dehydrogenase, α-ketomalonate and tetrahydro-
NAD. J. Mol. Biol. 1999, 285, 703−712.
(33) Blomberg, N.; Cosgrove, D. A.; Kenny, P. W.; Kolmodin, K.
Design of compound libraries for fragment screening. J. Comput.-Aided
Mol. Des. 2009, 23, 513−525.
(34) Hopkins, A. L.; Groom, C. R.; Alex, A. Ligand efficiency: a
useful metric for lead selection. Drug Discovery Today 2004, 9, 430−
431.
(35) DeLano, W. L. The PyMOL Molecular Graphics System;
Schrodinger, LLC, 2003.
(36) Patel, J. U.; Prankerd, R. J.; Sloan, K. B. A prodrug approach to
increasing the oral potency of a phenolic drug. 1. Synthesis,
characterization, and stability of an O-(imidomethyl) derivative of
17β-estradiol. J. Pharm. Sci. 1994, 83, 1477−1481.
(37) Gubernator, K. (www.emolecules.com) The Comprehensive
Source of Commercially Available, In Stock Chemicals. Presented at
the 239th National Meeting of the American Chemical Society,
San Francisco, CA, Mar 21−25, 2010; CINF-152.
(38) Cheng, A. C.; Coleman, R. G.; Smyth, K. T.; Cao, Q.; Soulard,
P.; Caffrey, D. R.; Salzberg, A. C.; Huang, E. S. Structure-based
maximal affinity model predicts small-molecule druggability. Nat.
Biotechnol. 2007, 25, 71−75.
(39) Bembenek, S. D.; Tounge, B. A.; Reynolds, C. H. Ligand
efficiency and fragment-based drug discovery. Drug Discovery Today
2009, 14, 278−283.
(21) Gill, A. New lead generation strategies for protein kinase
inhibitors: fragment based screening approaches. Mini-Rev. Med. Chem.
2004, 4, 301−311.
(40) Edwards, M. P.; Price, D. A. Role of physicochemical properties
and ligand lipophilicity efficiency in addressing drug safety risks. Annu.
Rep. Med. Chem. 2010, 45, 381−391.
(22) Tse, C.; Shoemaker, A. R.; Adickes, J.; Anderson, M. G.; Chen,
J.; Jin, S.; Johnson, E. F.; Marsh, K. C.; Mitten, M. J.; Nimmer, P.;
Roberts, L.; Tahir, S. K.; Xiao, Y.; Yang, X.; Zhang, H.; Fesik, S.;
Rosenberg, S. H.; Elmore, S. W. ABT-263: a potent and orally
bioavailable Bcl-2 family inhibitor. Cancer Res. 2008, 68, 3421−3428.
(23) Cunningham, B. T.; Laing, L. G. Advantages and application of
label-free detection assays in drug screening. Expert Opin. Drug
Discovery 2008, 3, 891−901.
(24) Bilitewski, U. Protein-sensing assay formats and devices. Anal.
Chim. Acta 2006, 568, 232−247.
(25) Read, J. A.; Winter, V. J.; Eszes, C. M.; Sessions, R. B.; Brady,
R. L. Structural basis for altered activity of M- and H-isozyme forms of
human lactate dehydrogenase. Proteins: Struct., Funct., Genet. 2001, 43,
175−185.
(41) Orita, M.; Ohno, K.; Warizaya, M.; Amano, Y.; Niimi, T. Lead
generation and examples opinion regarding how to follow up hits.
Methods Enzymol. 2011, 493, 383−419.
(42) Karlberg, T.; Hammarstrom, M.; Schutz, P.; Svensson, L.;
Schuler, H. Crystal structure of the catalytic domain of human PARP2
in complex with PARP inhibitor ABT-888. Biochemistry 2010, 49,
1056−1058.
(43) Woon, E. C. Y.; Threadgill, M. D. Poly(ADP-ribose)polymerase
inhibition: Where now? Curr. Med. Chem. 2005, 12, 2373−2392.
(44) Chen, L.; Petrelli, R.; Felczak, K.; Gao, G.; Bonnac, L.; Yu, J. S.;
Bennett, E. M.; Pankiewicz, K. W. Nicotinamide adenine dinucleotide
based therapeutics. Curr. Med. Chem. 2008, 15, 650−670.
(45) Burgos, E. S.; Ho, M.; Almo, S. C.; Schramm, V. L. A
phosphoenzyme mimic, overlapping catalytic sites and reaction
coordinate motion for human NAMPT. Proc. Natl. Acad. Sci. U.S.A.
2009, 106, 13748−13753.
(46) Abad-Zapatero, C.; Griffith, J. P.; Sussman, J. L.; Rossmann, M. G.
Refined crystal structure of dogfish M4 apo-lactate dehydrogenase.
J. Mol. Biol. 1987, 198, 445−467.
(47) Dunn, C. R.; Wilks, H. M.; Halsall, D. J.; Atkinson, T.; Clarke,
A. R.; Muirhead, H.; Holbrook, J. J. Design and synthesis of new
enzymes based on the lactate dehydrogenase framework. Philos. Trans.
R. Soc. London, Ser. B 1991, 332, 177−184.
(26) Smith, T. F.; Waterman, M. S.; Fitch, W. M. Comparative
biosequence metrics. J. Mol. Evol. 1981, 18, 38−46.
(27) Edfeldt, F. N. B.; Folmer, R. H. A.; Breeze, A. L. Fragment
screening to predict druggability (ligandability) and lead discovery
success. Drug Discovery Today 2011, 16, 284−287.
(28) Cameron, A.; Read, J.; Tranter, R.; Winter, V. J.; Sessions, R. B.;
Brady, R. L.; Vivas, L.; Easton, A.; Kendrick, H.; Croft, S. L.; Barros,
D.; Lavandera, J. L.; Martin, J. J.; Risco, F.; Garcia-Ochoa, S.; Gamo,
F. J.; Sanz, L.; Leon, L.; Ruiz, J. R.; Gabarro, R.; Mallo, A.;
Gomez de las Heras, F. Identification and activity of a series of azole-
based compounds with lactate dehydrogenase-directed anti-malarial
activity. J. Biol. Chem. 2004, 279, 31429−31439.
(48) Hewitt, C. O.; Eszes, C. M.; Sessions, R. B.; Moreton, K. M.;
Dafforn, T. R.; Takei, J.; Dempsey, C. E.; Clarke, A. R.; Holbrook,
J. J. A general method for relieving substrate inhibition in lactate
dehydrogenases. Protein Eng. 1999, 12, 491−496.
(29) AstraZeneca, unpublished results.
(30) Sem, D. S.; Bertolaet, B.; Baker, B.; Chang, E.; Costache, A. D.;
Coutts, S.; Dong, Q.; Hansen, M.; Hong, V.; Huang, X.; Jack, R. M.;
Kho, R.; Lang, H.; Ma, C.; Meininger, D.; Pellecchia, M.; Pierre, F.;
Villar, H.; Yu, L. Systems-based design of bi-ligand inhibitors of
oxidoreductases filling the chemical proteomic toolbox. Chem. Biol.
2004, 11, 185−194.
(49) Swiderek, K.; Panczakiewicz, A.; Bujacz, A.; Bujacz, G.; Paneth,
P. Modeling of isotope effects on binding oxamate to lactic
dehydrogenase. J. Phys. Chem. B 2009, 113, 12782−12789.
(50) Friesner, R. A.; Banks, J. L.; Murphy, R. B.; Halgren, T. A.;
Klicic, J. J.; Mainz, D. T.; Repasky, M. P.; Knoll, E. H.; Shelley, M.;
Perry, J. K.; Shaw, D. E.; Francis, P.; Shenkin, P. S. Glide: a new
approach for rapid, accurate docking and scoring. 1. Method and
assessment of docking accuracy. J. Med. Chem. 2004, 47, 1739−1749.
(51) Irvine, J. D.; Takahashi, L.; Lockhart, K.; Cheong, J.; Tolan,
J. W.; Selick, H. E.; Grove, J. R. MDCK (Madin−Darby canine
kidney) cells: a tool for membrane permeability screening. J. Pharm.
Sci. 1999, 88, 28−33.
(31) Baker, B. R.; Lee, W. W.; Skinner, W. A.; Martinez, A. P.; Tong,
E. Potential anticancer agentsL. Non-classical antimetabolites. II.
Some factors in the design of exoalkylating enzyme inhibitors,
particulary of lactic dehydrogenase. J. Med. Pharm. Chem. 1960, 2,
633−657.
(32) Chapman, A. D. M.; Cortes, A.; Dafforn, T. R.; Clarke, A. R.;
Brady, R. L. Structural basis of substrate specificity in malate
3305
dx.doi.org/10.1021/jm201734r | J. Med. Chem. 2012, 55, 3285−3306

Journal of Medicinal Chemistry
Article
(52) Sidoova, E.; Odlerova, Z.; Volna, F.; Blockinger, G. Synthesis
and antimicrobial activity of 2-alkylthio-6-aminobenzothiazoles. Chem.
Zvesti 1979, 33, 830−836.
(53) Niederl, J. B.; Roth, R. T. Disproportionation in aryloxymalonic
acid syntheses. J. Am. Chem. Soc. 1940, 62, 1154−1156.
(54) Krishnan, G. R.; Sreekumar, K. First example of organocatalysis
by polystyrene-supported PAMAM dendrimers: highly efficient and
reusable catalyst for Knoevenagel condensations. Eur. J. Org. Chem.
2008, 4763−4768.
(55) Bigi, F.; Carloni, S.; Ferrari, L.; Maggi, R.; Mazzacani, A.; Sartori,
G. Clean synthesis in water. Part 2. Uncatalysed condensation reaction
of Meldrum’s acid and aldehydes. Tetrahedron Lett. 2001, 42, 5203−
5205.
(56) Bruno, I. J.; Cole, J. C.; Edgington, P. R.; Kessler, M.; Macrae, C.
F.; McCabe, P.; Pearson, J.; Taylor, R. New software for searching the
Cambridge Structural Database and visualizing crystal structures. Acta
Crystallogr., Sect. B: Struct. Sci. 2002, 58, 389−397.
(57) Mauleon Casellas, D.; Carganico, G.; Fos Torro, M. D. L. D.;
Garcia Perez, M. L.; Palomer Benet, A. Patent Application WO
9604267, 1996.
(58) Ichihara, O.; Barker, J.; Law, R. J.; Whittaker, M. Compound
design by fragment-linking. Mol. Inf. 2011, 30, 298−306.
(59) Barker, J. J.; Barker, O.; Courtney, S. M.; Gardiner, M.;
Hesterkamp, T.; Ichihara, O.; Mather, O.; Montalbetti, C. A. G. N.;
Muller, A.; Varasi, M.; Whittaker, M.; Yarnold, C. J. Discovery of a
novel Hsp90 inhibitor by fragment linking. ChemMedChem 2010, 5,
1697−1700.
(60) Hung, A. W.; Silvestre, H. L.; Wen, S.; Ciulli, A.; Blundell, T. L.;
Abell, C. Application of fragment growing and fragment linking to the
discovery of inhibitors of Mycobacterium tuberculosis pantothenate
synthetase. Angew. Chem., Int. Ed. 2009, 48, 8452−8456.
(61) Bohacek, R. S. Computational Methods for Fragment Linking.
Presented at the 236th National Meeting of the American Chemical
Society, Philadelphia, PA, Aug 17−21, 2008; COMP-022.
(62) Fattori, D.; Squarcia, A.; Bartoli, S. Fragment-based approach to
drug lead discovery: overview and advances in various techniques.
Drugs R&D 2008, 9, 217−227.
macromolecular crystallography. Acta Crystallogr., Sect. D: Biol.
Crystallogr. 2010, 66, 12−21.
(73) Vaguine, A. A.; Richelle, J.; Wodak, S. J. SFCHECK: a unified
set of procedures for evaluating the quality of macromolecular
structure-factor data and their agreement with the atomic model. Acta
Crystallogr., Sect. D: Biol. Crystallogr. 1999, 55, 191−205.
(74) Laskowski, R. A.; Rullmannn, J. A.; MacArthur, M. W.; Kaptein,
R.; Thornton, J. M. AQUA and PROCHECK-NMR: programs for
checking the quality of protein structures solved by NMR. J. Biomol.
NMR 1996, 8, 477−486.
(75) Pontius, J.; Richelle, J.; Wodak, S. J. Deviations from standard
atomic volumes as a quality measure for protein crystal structures.
J. Mol. Biol. 1996, 264, 121−136.
(76) Allen, F. H. The Cambridge Structural Database: a quarter of a
million crystal structures and rising. Acta Crystallogr., Sect. B: Struct. Sci.
2002, 58, 380−388.
(77) Dalvit, C.; Fogliatto, G.; Stewart, A.; Veronesi, M.; Stockman, B.
WaterLOGSY as a method for primary NMR screening: practical
aspects and range of applicability. J. Biomol. NMR 2001, 21, 349−359.
(78) Miyoshi, T. Slow dynamics of polymer crystallites revealed by
solid-state MAS exchange NMR. Kobunshi Ronbunshu 2004, 61, 442−
457.
(79) Zhu, G.; Xia, Y.; Nicholson, L. K.; Sze, K. H. Protein dynamics
measurements by TROSY-based NMR experiments. J. Magn. Reson.
2000, 143, 423−426.
(80) Turmel, M.; Itkin, Z.; Liu, D.; Nie, D. An innovative way to
create assay ready plates for concentration response testing using
acoustic technology. J. Lab. Autom. 2010, 15, 297−305.
(81) Moorhouse, A. D.; Spiteri, C.; Sharm, P.; Zloh, M.; Moses, J. E.
Targeting Glycolysis: a fragment based approach towards bifunctional
inhibitors of hLDH-5. Chem. Commun. 2011, 47, 230−232.
(63) Crisman, T. J.; Bender, A.; Milik, M.; Jenkins, J. L.; Scheiber, J.;
Sukuru, S. C. K.; Fejzo, J.; Hommel, U.; Davies, J. W.; Glick, M.
“Virtual fragment linking”: an approach to identify potent binders from
low affinity fragment hits. J. Med. Chem. 2008, 51, 2481−2491.
(64) AstraZeneca, unpublished results.
(65) Beaumont, K.; Webster, R.; Gardner, I.; Dack, K. Design of ester
prodrugs to enhance oral absorption of poorly permeable compounds:
challenges to the discovery scientist. Curr. Drug Metab. 2003, 4, 461−
485.
(66) Pflugrath, J. W. The finer things in X-ray diffraction data
collection. Acta Crystallogr., Sect. D: Biol. Crystallogr. 1999, 55, 1718−
1725.
(67) Kabsch, W. Software XDS for image rotation, recognition and
crystal symmetry assignment. Acta Crystallogr., Sect. D: Biol. Crystallogr.
2010, 66, 125−132.
(68) Vonrhein, C.; Flensburg, C.; Keller, P.; Sharff, A.; Smart, O.;
Paciorek, W.; Womack, T.; Bricogne, G. Data processing and analysis
with the autoPROC toolbox. Acta Crystallogr., Sect. D: Biol. Crystallogr.
2011, 67, 293−302.
(69) Bailey, S. The CCP4 suite: programs for protein crystallography.
Acta Crystallogr., Sect. D: Biol. Crystallogr. 1994, 50, 760−763.
(70) Murshudov, G. N.; Skubak, P.; Lebedev, A. A.; Pannu, N. S.;
Steiner, R. A.; Nicholls, R. A.; Winn, M. D.; Long, F.; Vagin, A. A.
REFMAC5 for the refinement of macromolecular crystal structures.
Acta Crystallogr., Sect. D: Biol. Crystallogr. 2011, 67, 355−367.
(71) Emsley, P.; Cowtan, K. Coot: model-building tools for
molecular graphics. Acta Crystallogr., Sect. D: Biol. Crystallogr. 2004,
60, 2126−2132.
(72) Chen, V. B.; Arendall, W. B. III; Headd, J. J.; Keedy, D. A.;
Immormino, R. M.; Kapral, G. J.; Murray, L. W.; Richardson, J. S.;
Richardson, D. C. MolProbity: all-atom structure validation for
3306
dx.doi.org/10.1021/jm201734r | J. Med. Chem. 2012, 55, 3285−3306