Synthesis, induced-fit docking investigations, and in vitro aldose reductase inhibitory activity of non-carboxylic acid containing 2,4-thiazolidinedione derivatives

Article abstract of DOI:10.1016/j.bmc.2008.04.072

In continuation of our studies, we here report a series of non-carboxylic acid containing 2,4-thiazolidinedione derivatives, analogues of previously synthesized carboxylic acids which we had found to be very active in vitro aldose reductase (ALR2) inhibit

Full text of DOI:10.1016/j.bmc.2008.04.072

Bioorganic & Medicinal Chemistry 16 (2008) 5840–5852
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry
journal homepage: www.elsevier.com/locate/bmc
Synthesis, induced-fit docking investigations, and in vitro aldose reductase
inhibitory activity of non-carboxylic acid containing
2,4-thiazolidinedione derivatives
a
a
b
b
Rosanna Maccari a, , Rosaria Ottanà , Rosella Ciurleo , Dietmar Rakowitz , Barbara Matuszczak ,
*
Christian Laggner c, Thierry Langer c
a Dipartimento Farmaco-chimico, Facoltà di Farmacia, Università di Messina, Polo Universitario dell’Annunziata, 98168 Messina, Italy
b Department of Pharmaceutical Chemistry, Institute of Pharmacy, University of Innsbruck, Innrain 52a, A-6020 Innsbruck, Austria
c Department of Pharmaceutical Chemistry, Institute of Pharmacy and Centre for Molecular Biosciences (CMBI), University of Innsbruck, Innrain 52c, A-6020 Innsbruck, Austria
a r t i c l e i n f o
a b s t r a c t
Article history:
In continuation of our studies, we here report a series of non-carboxylic acid containing 2,4-thiazolidin-
edione derivatives, analogues of previously synthesized carboxylic acids which we had found to be very
active in vitro aldose reductase (ALR2) inhibitors. Although the replacement of the carboxylic group with
the carboxamide or N-hydroxycarboxamide one decreased the in vitro ALR2 inhibitory effect, this led to
the identification of mainly non-ionized derivatives with micromolar ALR2 affinity. The 5-arylidene moi-
ety deeply influenced the activity of these 2,4-thiazolidinediones. Our induced-fit docking studies sug-
gested that 5-(4-hydroxybenzylidene)-substituted derivatives may bind the polar recognition region of
the ALR2 active site by means of the deprotonated phenol group, while their acetic chain and carbonyl
group at position 2 of the thiazolidinedione ring form a tight net of hydrogen bonds with amino acid res-
idues of the lipophilic specificity pocket of the enzyme.
Received 15 February 2008
Revised 23 April 2008
Accepted 30 April 2008
Available online 3 May 2008
Keywords:
2,4-Thiazolidinediones
Aldose reductase
Diabetes mellitus
Induced-fit docking
Ó 2008 Elsevier Ltd. All rights reserved.
1. Introduction
long-term complications is unavoidable, since glycaemic levels
cannot be completely normalized in individuals with diabetes. In
Diabetes mellitus (DM) is a common chronic metabolic disease,
which is always associated with severe degenerative complica-
tions, such as neuropathy, nephropathy, retinopathy, cataracts,
accelerated atherosclerosis, increased risk of myocardial infarction
and stroke. The onset of these pathologies is a dramatic event in
the course of both type 1 and type 2 diabetes (DM1 and DM2). In
fact, their prevention and control are still serious and challenging
therapeutic problems as they represent the leading causes of mor-
bidity and mortality for diabetic patients.1–3
Prolonged exposure to high concentrations of glucose is the
main factor responsible for the pathogenesis of secondary diabetic
complications. According to several predominant theories, this oc-
curs through a variety of metabolic mechanisms, such as increased
intracellular formation of advanced glycation-end products, abnor-
mal activation of protein kinase C, increased flux of glucose
through both the polyol and the hexosamine pathways, which lead
to microvascular and macrovascular damage in the long run.3–5 It
has been demonstrated that careful control of glycaemia can delay
the emergence of these harmful effects.1,6 However, the onset of
addition, in many patients affected by DM2, vascular complications
are already present at the time of diagnosis.3
The worldwide incidence of DM is very high and it is predicted
to rise from the current estimate of about 180 million to more than
360 million by 2030.7 Currently, there are few drugs that are able
to counteract the development of the associated pathologies.
Therefore, the need to search for new drug candidates in this field
appears to be critical.
The increased flux of glucose through the polyol pathway,
which occurs under conditions of hyperglycaemia in tissues pos-
sessing insulin-independent glucose transport (retina, lenses, kid-
ney, and nerve), is
a well-known factor implicated in the
development of secondary diabetic complications.4,8–10
Aldose reductase (EC 1.1.1.21, ALR2), a member of the aldo-keto
reductase superfamily, is the first and rate-limiting enzyme of the
polyol pathway. It catalyzes the NADPH-dependent reduction of
glucose to sorbitol which, in some tissues, is oxidized to fructose
by sorbitol dehydrogenase.
Under euglycaemic conditions, ALR2, which has low affinity for
glucose, produces little sorbitol. Indeed, a wide variety of alde-
hydes, especially medium-chain hydrophobic lipid-derived alde-
hydes and their glutathione adducts, are reduced more efficiently
* Corresponding author. Tel.: +39 90 6766406; fax: +39 90 355613.
E-mail address: rmaccari@pharma.unime.it (R. Maccari).
0968-0896/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmc.2008.04.072

R. Maccari et al. / Bioorg. Med. Chem. 16 (2008) 5840–5852
5841
than glucose by ALR2. This would suggest that the most common
role of ALR2 may be the removal of potentially toxic aldehydes
generated during lipid peroxidation.9,11
Most 2,4-TZDs 1–3 are effective in vitro ARIs, with micromolar
or submicromolar IC50 values.23–25 The biological data we have col-
lected so far have shown that the presence of an acetic chain on N-
3 (compounds 3) is related to the highest inhibition levels.
N-unsubstituted analogues 1 were also shown to be good ALR2
inhibitors. However, they were always less potent (from 10 to 100
times) than the corresponding acids 3. In addition, analogous
methyl esters 2, although devoid of any acidic functionality, were
shown to satisfy the minimal requisites for ALR2 inhibition, that
is, the formation of hydrogen bonds with Tyr48 and His110 and
the presence of a lipophilic aromatic moiety. In fact, some of them
produced appreciable inhibition levels, similar to those of N-
unsubstituted 2,4-TZDs 1.23,24
In contrast, under hyperglycaemic conditions, an excess of glu-
cose (more than 30% of the total metabolized amount) is converted
to sorbitol through the polyol pathway in tissues possessing insu-
lin-independent uptake of glucose, leading to intracellular accu-
mulation of this osmotically active polyol. At the same time,
NADPH and NAD+ deprivation causes changes in cellular redox
potentials, increased oxidative stress and reduced activity of other
NADPH-dependent enzymes, such as nitric oxide (NO) synthase
and glutathione reductase.9 Alterations of cytokine signalling and
kinase cascades are also involved.4,9,12–14 These metabolic and bio-
chemical changes result in inflammation, chronic vascular damage
and decrease in nerve conduction velocity, and have been recog-
nized as being related to the pathogenesis of diabetic complica-
tions.4,8,12–15
In continuing our search aimed at identifying new active ana-
logues, we designed the synthesis of non-carboxylic acid contain-
ing 5-arylidene-2,4-thiazolidinedione derivatives.
Indeed, carboxylic acids that are active as ARIs are potent in vi-
tro inhibitors. However, their effectiveness generally decreases in
vivo, probably because of their poor capability to penetrate key tar-
get tissues, in particular peripheral nerves.17,19,20 Thus, the devel-
opment of new low acidity ARIs, with pKa values higher than
those of carboxylic acids, is desirable because they may display
better bioavailability than their carboxylic acid counterparts.
In particular, we replaced the carboxylic acetic chain on N-3 of
(5-arylidene-2,4-dioxothiazolidin-3-yl)acetic acids (3) with the
isostere acetamide and N-hydroxyacetamide moieties, obtaining
(5-arylidene-2,4-dioxothiazolidin-3-yl)acetamides (4) and the
analogous N-hydroxyacetamides (5), respectively (Fig. 2). They fit-
ted the pharmacophoric model which we had previously defined
for 5-arylidene-2,4-thiazolidinediones effective as ARIs, consisting
of two H-acceptor groups and a lipophilic region.24
Our initial hypothesis was that the acetamide group of com-
pounds 4 might be able to bind the ALR2 active site via hydrogen
bonds, analogously to the acetate chain of esters 3.
On the other hand, the N-hydroxycarboxamide moiety (com-
pounds 5) can mimic monoanionic acidic groups. In addition, it is
known that it can serve as both hydrogen bond donor and acceptor,
as well as a metal chelating moiety,29 thereby playing an important
role in a variety of biologically active compounds.30
The substitution pattern of the 5-arylidene moiety of 2,4-
TZDs 4 and 5 was based on the previously acquired SARs. The
presence of an additional aromatic ring or an H-bond donor
group in this moiety was shown to enhance the ALR2 inhibitory
effect.23–25
We here report the synthesis and in vitro ALR2 inhibitory
activity of a series of 5-arylidene-2,4-thiazolidinediones 4 and
5. Additionally, induced-fit docking (IFD) studies were carried
out in order to gain new insight concerning the possible binding
modes of our compounds, which should also be useful to guide
the future synthesis of improved compounds. Furthermore, it
should serve as an evaluation of the scope and limitations of
the IFD method.
Therefore, ALR2 inhibition has been proposed as a pharmaco-
logical strategy to prevent or delay the onset and progression of
such pathologies, independently of glycaemia control.
In addition, other possible therapeutical applications of ALR2
inhibitors (ARIs) are currently emerging. In fact, recent studies
have suggested that ALR2 may be involved in several physiological
cellular functions, including growth and signalling.9,12 In particu-
lar, it has been demonstrated that ALR2 inhibition can suppress
the lipopolysaccharide-induced production of pro-inflammatory
cytokines, such as TNF-a, IL-6, IL-1b, IFN-c and inflammatory medi-
ators, such as NO and prostaglandin E2 (PGE2) in macrophages.13
TNF-a-induced overexpression of cyclooxygenase 2 together with
PGE2 production have also been prevented in human colon cancer
cells.16 As a consequence, ALR2 inhibition could be a novel ap-
proach to the chemoprevention of colon cancer as well as to the
therapeutic treatment of diseases associated with high levels of
cytokine expression, such as inflammation, rheumatoid arthritis
and sepsis.9,13,16
Although numerous ARIs have been reported over the last three
decades,14,17–20 many of them proved to be clinically inadequate,
because of pharmacokinetic problems, low efficacy or adverse
side-effects. Currently, epalrestat (Fig. 1) is the only ARI available
on the market21 and few molecules (such as fidarestat, Fig. 1) are
in late-stage clinical development.19,22
In particular, two main classes of orally active ARIs have been
identified so far: cyclic imides, mostly hydantoins (such as sorbinil,
fidarestat; Fig. 1) and carboxylic acids (such as zopolrestat, epalre-
stat; Fig. 1).14,17–20 In addition, 2,4-thiazolidinediones (2,4-TZDs)
have received attention as hydantoin bioisosteres that are poten-
tially devoid of hypersensitivity reactions apparently related to
the hydantoin ring. In fact, many 2,4-TZDs have been reported as
both antihyperglycaemic agents and ARIs, especially after the dis-
covery of 2,4-thiazolidinedione derivatives as oral antidiabetic
drugs.14,17–20
In the last few years, we have synthesised numerous 5-arylid-
ene-2,4-thiazolidinediones (1–3, Fig. 1) which were designed and
assayed as in vitro ARIs.23–25 They possess the structural requisites
considered critical for ALR2 inhibition: (a) an acidic function and/
or H-bond acceptor groups that are able to interact with a posi-
tively charged recognition region of the ALR2 active site formed
by Tyr48, His110, Trp111 and the nicotinamide ring of the cofactor
NADP+ and (b) an aromatic moiety, which can establish hydropho-
bic interactions with a lipophilic pocket of the enzyme lined by
Leu300 and Trp111.17–20,26–28 In addition, molecular docking simu-
lations into the ALR2 active site highlighted that phenolic or car-
boxylic groups in the 5-benzylidene moiety can favourably
interact either with amino acid residues lining this lipophilic pock-
et, such as Leu300, or with the positively charged recognition
region.25
2. Chemistry
(Z)-(5-Arylidene-2,4-dioxothiazolidin-3-yl)acetamides (4a–g)
were prepared by the reaction of the corresponding (Z)-5-arylid-
ene-2,4-thiazolidinediones (1), synthesised according to
a
reported procedure,23,24 with 2-chloroacetamide in refluxing ace-
tonitrile, in the presence of potassium carbonate (Scheme 1). The
synthesis of hydroxybenzylidene-substituted acetamides 4h–k
was carried out by adding 2-chloroacetamide dropwise to a solu-
tion of corresponding 2,4-TZDs 1 and potassium carbonate in
acetonitrile over 2 h, in order to prevent the nucleophilic attack
of hydroxyl group to chloroacetamide.

5842
R. Maccari et al. / Bioorg. Med. Chem. 16 (2008) 5840–5852
O
O
H
O
H
N
N
OH
HN
HN
O
CF3
O
F
F
N
N
N
S
O
CONH2
O
O
Sorbinil
Fidarestat (FID)
Zopolrestat
S
O
O
O
O
OH
N
OH
N
OH
N
F
S
S
S
N
O
F
CF3
F
Epalrestat
Lidorestat (3NA)
Tolrestat (TOL)
H
O
O
O
S
N N
O
O
N
O
N
O
OH
OH
O
O
O
S
S
O
O
O
O
OH
OH
O
O
Cl
PDB code ITA
PDB code ITB
PDB code 62P
O
S
R
O
O
S
N
1
2
3
R = H
R = CH2COOCH3
R = CH2COOH
N
OH
O
O
Ar
1-3
HO
3a
Figure 1. Chemical structures of some ARIs.
MW-irradiation, the treatment of acids 3 with ethylchloroformate
and the subsequent reaction of the intermediate anhydrides with
hydroxylamine produced compounds 5 in overall ten minutes
and in 40–65% yields. This MW-assisted procedure allowed a
marked decrease in the reaction times and a 10–15% increase in
yields and thus provided a very rapid and convenient method to
prepare N-hydroxyacetamides 5.
O
S
R
N
4
5
R = CH2CONH2
R = CH2CONHOH
O
Ar
4, 5
The structures of compounds 4 and 5 were assigned on the basis
of their analytical and spectroscopic data.
Figure 2. General structure of derivatives 4 and 5.
The 1H NMR spectra of compounds 4 showed a singlet at 4.19–
4.27 ppm, due to the resonance of the methylene group of the acet-
amide chain on N-3, which was diagnostic. Their 13C NMR spectra
In the synthesis of (Z)-(5-arylidene-2,4-dioxothiazolidin-3-yl)-
N-hydroxyacetamides (5) we applied a procedure reported by Red-
dy et al.31 The treatment of acids 3 with ethylchloroformate in
diethylether, in the presence of triethylamine, followed by the
addition of a freshly prepared solution of hydroxylamine in meth-
anol, provided the desired compounds 5 (Scheme 1). The starting
(Z)-(5-arylidene-2,4-dioxothiazolidin-3-yl)acetic acids (3) were
synthesised by the treatment of (Z)-2,4-TZDs 1 with methyl bro-
moacetate, using potassium carbonate as base, and subsequent
hydrolysis of methyl esters 2 in acidic medium, as already reported
(Scheme 1).24
showed
a triplet at 43.3–44.0 ppm and a singlet at 165.3–
166.2 ppm, attributable to the methylene group and to the
carbonyl group of the acetamide chain, respectively, which con-
tributed to the unambiguous assignment of their structure.
1H NMR spectra of N-hydroxyacetamides 5 displayed a singlet
at 4.19–4.63 ppm, due to the resonance of the methylene group
on N-3, and two broad singlets, exchangeable with D2O, in the
range from 9.01 to 11.0 ppm, attributable to the resonance of the
NHOH group protons. In 13C NMR spectra, the methylene group
on N-3 resonated at 41.5–46.2 ppm; in addition, the carbonyl
groups of the thiazolidinedione ring and of the N-hydroxyaceta-
mide chain resonated in the range between 162.5 and 168.6 ppm.
In an attempt to increase the yields of N-hydroxyacetamides 5
(30–50%) and to reduce reaction times (24–40 h), we carried out
the same synthesis by irradiation with microwaves (MW). Upon

R. Maccari et al. / Bioorg. Med. Chem. 16 (2008) 5840–5852
5843
O
S
O
S
CH2COOCH3
O
O
S
CH2COOH
O
H
O
O
H
a
N
b
N
N
c
N
+
Ar CHO
S
O
Ar
Ar
Ar
1
2
3
e, f
d
O
CH2CONHOH
O
CH2CONH2
O
N
N
S
S
O
Ar
Ar
5 a-g
4 a-k
Scheme 1. Reagents and conditions: (a) C5H11N, C2H5OH, D; (b) BrCH2COOCH3, K2CO3, acetone; (c) AcOH, HCl, D; (d) ClCH2CONH2, K2CO3, CH3CN, D; (e) ClCOOC2H5, (C2H5)3N,
diethyl ether, MW (max 300 W) 80 °C; (f) NH2OH, CH3OH, MW (max 300 W) 80 °C.
The monitoring of the course of the reaction of acids 3 with eth-
ylchloroformate, by means of 1H NMR spectra, also enabled the
intermediate anhydrides to be identified. A singlet at 4.60 ppm
was attributable to the resonance of the methylene group of the
acetic chain on N-3, slightly more deshielded than the correspond-
ing group of starting acids 3, whereas a quartet at circa 4.40 ppm
and a triplet at 1.40 ppm were due to the resonance of the ethyl
group of the anhydride.
investigate possible binding modes of a given ligand. This prob-
lem has been thoroughly discussed in a recent paper given by
Zentgraf et al.33 One of the lessons learned from their studies
is that the attempt to predict the possible binding mode of a gi-
ven ligand by simply comparing its substructure to those of
compounds with experimentally determined binding modes can
often fail, even amongst compounds that are structurally closely
related. The results from high-resolution X-ray crystal structures
suggest that both the hydantoin-based ALR2 inhibitors such as
sorbinil, which have a pKa of about 8.0, and a recently reported
sulfonyl-pyridazinone derivative (Fig. 1, PDB code 62P, com-
plexes 1Z89 and 1Z8A), which has a calculated pKa of about
9.3, are deprotonated at the binding site.38 This deprotonated
state may be stabilized by the nearby positive charges at NADP+
and Lys77, as well as by the hydrogen bonds donated by Trp111
and His110. Based on these results, we expect that our N-
hydroxyacetamides 5, which have calculated pKa values of about
8.7,39 are also binding in their deprotonated form, which was
supported by more conclusive poses as well as lower docking
score values in multiple docking runs (details not shown).
One method that tries to overcome binding site flexibility dur-
ing docking is the IFD protocol developed by Schrödinger, LLC.40
This method has already been successfully used to cross-dock tol-
restat into the binding site of 2ACR, which corresponds to the
closed sorbinil conformation.40 However, Zentgraf et al. reported
that they were not able to predict the correct binding modes of
their naphto[1,2-d]isothiazole acetic acid derivatives using the in-
duced-fit methodology.33
We chose to evaluate the IFD method to see how well this
method is able both to reproduce known binding conformations
and to predict new ones presented with different protein geom-
etries, to give possible explanations for the SARs and possible
binding modes of our compounds. With an ideal IFD protocol,
one could expect that, no matter which experimental structure
is used as an input geometry for the docking, the correct confor-
mations of both the ligand and the protein would always be
found. Previous studies, however, showed that this is not the
case. In the studies of Sherman et al.40 three residues whose
side-chains should be regarded as flexible during the IFD process
were manually selected for the ALR2 example. One may argue, of
course, that the knowledge about which side-chains have to
move was already available. In the case of ALR2, there are many
different flexible side chains surrounding the binding site, so
that we would have to include about five or six of them to sam-
ple all the movements we know to take place amongst the dif-
3. Induced-fit docking
Over the last few years, many X-ray crystal structures of ALR2
have been deposited in the PDB database.32 Experimental data
showed that the ALR2 binding site can be divided into two
sub-pockets, a catalytic pocket (also called the recognition re-
gion) and a specificity pocket. The catalytic pocket contains
one molecule of NADP+ and is usually occupied by an acidic moi-
ety of the bound ligand. Its conformation is rather stable in all
the various reported crystal structures. In contrast, the specificity
pocket can adopt different conformations, adapting itself to the
size of the bound ligands via movement of the loop region be-
tween Cys298 and Cys303 and the different rotameric states of
the surrounding amino acid side chains, especially of Leu300
which acts as a gate-keeper between the closed and open states.
Until recently, only three distinct conformations of the specificity
pocket were known, and were called by Zentgraf et al. as the
‘sorbinil conformation’, the ‘tolrestat conformation’ and the
‘IDD594 conformation’.33 The three different binding pocket con-
formations are shown in Figure 3, as exemplified by the three
PDB structures 1PWM (ligand fidarestat, Fig. 1, sorbinil confor-
mation),34 2FZD (ligand tolrestat, Fig. 1, tolrestat conforma-
tion),35 and 1Z3N (ligand lidorestat, Fig. 1, IDD594
conformation)36 that were used in this study. The same authors
reported two novel binding modes observed for two different
naphto[1,2-d]isothiazole acetic acid derivatives (ITA and ITB,
Fig. 1) exemplified by the structures with PDB codes 2NVC and
2NVD.33,37 While the ligands are found in different regions, the
overall conformation of the protein (especially the specificity
pocket) is similar to that of the sorbinil conformation (Fig. 3).
The most distinct differences are the rotation of Trp20 and the
subsequent breaking of the salt bridge between Lys21 and a
NADP+ phosphate group in 2NVD.
As stated above, the ALR2 binding site shows a high degree of
flexibility, which poses
a big challenge to those trying to

5844
R. Maccari et al. / Bioorg. Med. Chem. 16 (2008) 5840–5852
Figure 3. Comparison of different binding modes reported for ALR2 protein–ligand complexes, exemplified by the complexes used in this study. Graphics were created with
PyMOL.46 Top left: 1PWM (ligand fidarestat, ‘sorbinil conformation’),34 top right: 2FZD (ligand tolrestat, ‘tolrestat conformation’),35 middle left: 1Z3N (ligand lidorestat, ‘IDD594
conformation’),36 middle right: 2NVC (ligand code ITA),37 bottom left: 2NVD (ligand code ITB).37 The important neighboring amino acids are labelled; the labels of groups lying
further in the back have lower contrast. The cofactor NADP+, placed behind the ligands in this view, is represented in line-mode.
ferent experimental structures. This would, of course, enlarge the
binding site and thus increase the chance of the ligands being
wrongly placed along the pocket walls. However, we had repre-
sentative examples for five different binding site conformations

R. Maccari et al. / Bioorg. Med. Chem. 16 (2008) 5840–5852
5845
Table 1
available, and we knew that Leu300 exhibits the biggest contri-
In vitro bovine lenses ALR2 inhibitory activity of 2,4-thiazolidinediones 4 and 5
bution to the flexibility of the binding site. We thus decided to
perform IFD into these different binding site geometries with a
few selected ligands, together with the native ligands of all the
selected protein ligand complexes, mutating only Leu300 during
the initial docking run. The native ligands were used for cross-
docking comparison, to see how well the different starting struc-
tures could transform into the different geometries and thus
accommodate the ligands of the other structures in a conforma-
tion similar to their experimentally determined ones. We also
wanted to use the results from these compounds as a guideline
to decide whether there is a measure that is an indicator of the
correct binding mode, for example, a lower GlideScore value.
We selected the complexes 1PWM, 2FZD, 1Z3N, 2NVC and
2NVD, which are all derived from human ALR2 and have a high
experimental resolution (0.92, 1.08, 1.04, 1.65 and 1.55 Å,
respectively).
The ligands we chose for docking were the corresponding li-
gands fidarestat, tolrestat, lidorestat, the compounds with the
PDB codes ITA and ITB, and six compounds selected from our
own 2,4-TZDs: 3a, 4g, 4i, 5b, 5e, and 5g. Compounds 4g and 4i
were chosen as representatives for the acetamides 4; 4g is a close
yet inactive derivative of 4i which, in turn, is the most effective out
of the acetamides 4, 5b, 5e and 5g are different examples of the N-
hydroxyacetamides which all show high ALR2 affinity, yet bear
aromatic residues of different shapes, sizes and polarities; finally,
compound 3a (Fig. 1, IC50 = 0.15 lM, previously reported by us)24
was included as an acetic acid derivative bearing the same aro-
matic residue as 4i and 5g.
O
S
CH2COR
N
O
Ar
a
Compound
R
Ar
IC50
4a
4b
4c
4d
4e
NH2
NH2
NH2
NH2
NH2
NH2
NH2
NH2
NH2
NH2
NH2
NHOH
NHOH
NHOH
NHOH
NHOH
NHOH
NHOH
3-OC6H5-C6H4
4-OC6H5-C6H4
4-C6H5-C6H4
1-Naphthyl
8% (12.5 lM)
n.i.
18% (12.5 lM)
32% (12.5 lM)
3% (12.5 lM)
10% (12.5 lM)
n.i.
16% (12.5 lM)
6.52 (4.40–9.66)
31% (12.5 lM)
n.i.
12.7 (11.0–14.6)
3.67 (3.38–3.99)
7.34 (5.30–10.16)
20.3 (16.5–25.0)
4.20 (3.88–4.54)
39% (25 lM)
1.79 (1.47–2.18)
1.41 (1.12–1.79)
2-Naphthyl
4f
3-OCH3-C6H4
4-OCH3-C6H4
3-OH-C6H4
4g
4h
4i
4-OH-C6H4
4j
3-OCH3,4-OH-C6H3
3-OH,4-OCH3-C6H3
3-OC6H5-C6H4
4-OC6H5-C6H4
4-OCH2C6H5-C6H4
4-C6H5-C6H4
1-Naphthyl
4k
5a
5b
5c
5d
5e
5f
5g
Sorbinil
2-Naphthyl
4-OH-C6H4
n.i., no inhibition (0% inhibition at 12.5 lM dose).
a
IC50 (lM) (95% C.L.) or% inhibition at the given concentration.
The selected PDB structures were processed in Maestro (Schrö-
dinger, LLC)41 using the Protein Preparation Wizard.
with compound 4h producing only 16% inhibition at 12.5 lM dose.
Moreover, replacing the hydrogen atom of both 3- and 4-hydroxy
group of compounds 4h and 4i with a methyl group (compounds
4f and 4g) or a phenyl ring (compounds 4a and 4b) always led to
less effective or inactive derivatives; this effect was particularly
evident in the case of 4-subsituted derivatives 4b and 4g, which
were totally inefficacious (Table 1). The insertion of a methoxy
group in position 3 of compound 4i also led to a less active
analogue (4j, 31% inhibition at 12.5 lM dose), whereas the 5-(3-hy-
droxy-4-methoxybenzylidene)-substituted isomer (4k) proved to
be ineffective at 12.5 lM dose.
Since our compounds are assayed with ALR2 from bovine
lenses, we also made a homology model using the Structure Predic-
tion tool of Prime accessible from the Maestro GUI. The sequence of
ALR2 from Bos taurus (P16116), which shows 82% identity and 90%
similarity with human ALR2, was downloaded from the ExPASy
homepage,42 and a homology model based on 2FZD, including both
tolrestat and NADP+ as ligands, was created. Comparison with the
obtained model showed that within an 8.0 Å radius of the binding
site, the only sequence differences could be found at positions 297
(bovine: Ala, human: Val) and 301 (bovine: Val, human: Leu).
It may be hypothesised that the appreciable inhibitory activity
of 4i is related to the ability of the phenolic group (calculated
pKa = 9.30)39 to bind the polar region of the ALR2 active site,
whereas the acetamide chain might face amino acid residues of
the lipophilic pocket (such as Leu300), thus making 4i assume a
180° rotated orientation, different from the one originally ex-
pected. This hypothesis is backed up by the results from our dock-
ing studies, reported below, which indicated that the deprotonated
phenol group of the 4-hydroxybenzylidene moiety of 4i can form
hydrogen bonds with Tyr48 and His110 in the catalytic centre of
ALR2, whilst the acetamide chain can interact with amino acid res-
idues of the specificity pocket of the enzyme.
The replacement of the acetamide chain on N-3 with the N-
hydroxyacetamide one led to 2,4-TZDs 5 and provided a general
significant increase in the ALR2 inhibitory effect (Table 1).
In particular, N-hydroxy-2-[5-(4-hydroxybenzylidene)-2,4-
dioxothiazolidin-3-yl]acetamide (5g) exhibited an inhibitory effect
(IC50 = 1.79 lM) which was fourfold greater than that of analogue
4i and very similar to that of sorbinil (Table 1).
4. Results and discussion
4.1. Aldose reductase inhibition
2,4-TZDs 4 and 5 were evaluated for their ability to inhibit
the in vitro reduction of D,L-glyceraldehyde by partially purified
ALR2 from bovine lenses; sorbinil was used as a reference drug
(Table 1).
Out of acetamides 4, [5-(4-hydroxybenzylidene)-2,4-dioxothiaz-
olidin-3-yl]acetamide (4i) displayed an interesting micromolar IC50
value (6.52 lM; Table 1), very close to those of the previously as-
sayed5-(4-hydroxybenzylidene)-2,4-thiazolidinedioneandthecor-
responding acetic acid methyl ester (IC50 = 8.96 and 6.18 lM,
respectively).24 In contrast, analogues 4a–g, bearing an additional
aromatic ring or a methoxy substituent in the 5-arylidene moiety,
produced no or only modest ALR2 inhibition. In particular, (5-naph-
thalen-1-ylmethylene-2,4-dioxothiazolidin-3-yl)acetamide
(4d)
N-Hydroxy-2-[2,4-dioxo-5-(4-phenoxybenzylidene)thiazolidin-
3- yl]acetamide (5b) and 5-naphthalen-1-ylmethylene analogue 5e
also displayed activity in the low micromolar range (IC50 = 3.67
and 4.20 lM, respectively). In contrast, the displacement of the
4-phenoxy group to position 3 of the 5-benzylidene ring (5a), as
well as the insertion of a methylene group (5c), gave a two–three-
broughtabout32% inhibitionat12.5 lMdose, provingtobe moreac-
tive than the corresponding methyl ester (12% inhibition at 50 lM
dose),24 whereas its 5-naphthalen-2-ylmethylene isomer 4e was
shown to be 10 times less effective at the same dose (Table 1).
The displacement of the hydroxy group of 4i to position 3 of the
5-benzylidene ring markedly decreased inhibitory effectiveness,

5846
R. Maccari et al. / Bioorg. Med. Chem. 16 (2008) 5840–5852
Table 3
fold decrease in potency (Table 1). Analogous to what was
observed for acetamides 4, the replacement of the 5-naphthalen-
1-ylmethylene moiety (5e) with the 5-naphthalen-2-ylmethylene
one (5f) resulted in a significant decrease in potency.
Lowest GlideScore values obtained for cross-docking the ligands of five different X-
ray crystal complexes with the IFD method
Ligand 3-letter code
Protein structure used for IFDa
Comparing the ALR2 inhibitory effects of N-hydroxyacetamides
5 with those of the corresponding acetamides 4 and carboxylic
acids 3 showed an order of activity 3 > 5 > 4. Compounds 5 were
also generally more effective inhibitors (from 3 to more than 10
times) than N-unsubstituted analogues 1,24 probably due to the
potential capability of the N-hydroxyacetamide chain on N-3 to
interact with the biological target through a network of H bonds.
However, the order of activity observed in vitro might change in
vivo. In fact, the calculated pKa values of N-hydroxyacetamides 5
ranged from 8.64 to 8.75.39 Thus, at physiological pH values,
compounds 5 are expected to be prevalently in non-ionized form
(calculated percentage of undissociated form 95% to 96%). In con-
trast, carboxylic acids 3 are almost totally in their anionic form
(calculated pKa values in the range between 2.57 and 3.00).39 The
pKa values of compounds 5 were also higher than those of their
counterparts 1; the latter (calculated pKa = 7.60)39 are expected
to undergo ionisation in 38–39% percentage at physiological pH
values. Therefore, derivatives 5 should be able to cross biological
barriers better than previously assayed compounds 1 and 3.
1PWM
1Z3N
2FZD
2NVC
2NVD
FID
3NA
TOL
ITA
ITB
À11.904
À11.517
À7.917
À10.580
À12.051
À9.357
À10.709
À11.932
À10.659
À11.527
À11.669
À11.008
À12.039
À8.806
À9.929
À8.937
À6.172
À10.806
À9.001
À11.232
À11.575
À11.372
À11.317
À12.684
À11.970
a
Values for ligands that were docked into the protein of their original complex
structure are written in italics.
to decide which conformation to choose. Interestingly, the two
top-ranked poses of fidarestat showed a high RMSD of over 5.0,
with the hydantoin ring wrongly placed near the backbone of
Leu301 and Ser302. The same error was observed when we docked
neutral fidarestat into the binding site.
As can be seen from Table 3, in four out of five cases (all com-
plexes except 2NVD) the native ligands obtained the lowest mini-
mum GlideScore value when docked into their original complexes.
We thus believe that a low GlideScore value could be a good indi-
cator in choosing what poses might describe the bioactive confor-
mation of our own compounds. Interestingly, IFD of the native
ligand into 2NVD also gave a ligand pose that was similar to the
experimental one. No correlation with the correct poses could be
observed for the IFDS, a composite score that is calculated by add-
ing together the GlideScore value and 5% of the Prime energy from
the refinement calculations.
The docking results for our six compounds are summarized in
Table 4. Visual inspection of the top-ranked poses for every li-
gand–protein combination showed that, in contrast to the cross-
docking of the crystal structure ligands, the compounds often
had similar poses in different proteins. Phenols 3a, 4i and 5g have
their lowest GlideScore value when docked into 2NVC, followed by
structurally similar 1PWM, whilst compounds 5b and 5e had their
lowest GlideScore values when docked into 2FZD, suggesting that
the hydrophobic moieties of these compounds are able to open
the lipophilicity pocket. A detailed discussion of possible binding
modes follows. This is based on reasonable pharmacophoric inter-
actions, including the knowledge from the experimental data avail-
able which show that the presence of a weakly acidic group at the
4.2. Docking results
Tables 2 and 3 summarize the results from the cross-docking of
the X-ray crystal structure ligands. As can be seen from Table 2, de-
spite the flexibility of the binding site during IFD, in four out of five
cases the lowest RMSD values could be obtained when the
compounds were docked into their original binding site, whilst
acceptable results (RMSD 6 2.0) were also found in a few cases of
cross-docking. In 18 out of 25 cases, the pose with the lowest
InducedFitDockingScore (IFDS) value was also the one with the
lowest GlideScore value (both GlideScore and IFDS have negative
values, with a lower value indicating better results). In the remain-
ing seven cases, lower RMSD was found five times by IFDS, and two
times by GlideScore. In our opinion, this difference is too small to
show a real advantage of using IFDS over GlideScore. In only three
cases was the highest ranked pose also the one with the lowest
RMSD value. These results show that whilst this method is able
to find acceptable conformations close to the ligands’ binding
poses, both IFDS and GlideScore do not present a reliable measure
Table 2
RMSD values obtained for cross-docking the ligands of five different X-ray crystal complexes with the IFD methoda
Ligand 3-letter code
FID
Calculated RMSD
Protein structure used for IFDb
2FZD
1PWM
1Z3N
2NVC
2NVD
IFDS
GS
Minimum
5.22
5.22
0.67
5.13
5.13
4.42
5.50
5.50
2.66
1.75
1.75
1.44
6.88
6.88
5.45
3NA
TOL
ITA
ITB
a
IFDS
GS
Minimum
1.51
1.31
1.04
0.70
0.70
0.67
2.00
2.00
1.94
3.12
3.12
2.37
5.49
7.11
3.56
IFDS
GS
Minimum
1.48
1.48
1.48
1.20
1.20
0.78
1.79
3.23
1.76
4.18
4.70
1.65
5.52
5.52
3.47
IFDS
GS
Minimum
1.46
7.19
0.98
4.80
4.80
4.37
7.16
7.16
1.76
0.88
0.88
0.75
6.83
6.80
5.56
IFDS
GS
Minimum
1.41
1.41
1.41
5.48
5.34
4.49
4.08
4.08
4.08
7.79
7.79
1.22
1.65
1.65
1.12
RMSD values are shown for the pose with lowest InducedFitDockingScore (IFDS), lowest GlideScore (GS) and minimum RMSD value amongst all poses (minimum).
Values for ligands that were docked into the protein of their original complex structure are written in italics.
b

R. Maccari et al. / Bioorg. Med. Chem. 16 (2008) 5840–5852
5847
Table 4
poses, the phenoxyphenyl residue of 5b has lipophilic contacts
with the surrounding side chain residues of Phe115, Phe122,
Leu124 and Leu300, whilst the 1-naphthyl group of 5e is sur-
rounded by the side chain residues of Phe115, Phe122, Val130
and Leu300. Furthermore, when docking 5e into 1Z3N, we found
a pose similar to that of lidorestat, with the naphthyl residue form-
ing p–p interactions with Trp111, and additional lipophilic contacts
with Phe115, Phe122 and Leu300.
Lowest GlideScore values obtained for induced-fit docking of six selected compounds
into five different protein conformations from X-ray crystal complexesa
Compound
Protein structure used for IFD
1PWM
1Z3N
2FZD
2NVC
2NVD
3a
4g
4i
5b
5e
5g
À12.156
À8.739
À11.751
À7.966
À7.423
À11.687
À11.394
À8.943
À11.312
À9.433
À12.934
À9.473
À11.365
À7.356
À10.066
À9.521
À8.472
À10.362
À10.643
À8.901
À11.285
À11.472
À11.572
À11.065
À12.416
À8.575
À10.259
À11.926
The program Ligandscout v2.0,43 which allows 3D pharmaco-
phores from macromolecule/ligand complexes to be rapidly de-
À10.701
À10.992
rived in
a fully automated way, was used to visualize the
a
For each ligand, the lowest GlideScore values are written in italics.
pharmacophoric interactions of 3a and 5b, respectively, with the
binding site in both 3D and 2D depiction (Fig. 4, middle and bottom
row). The detected interactions fit well with our observations dis-
cussed above.
catalytic centre is an important factor. Possible binding modes are
also based on a low GlideScore value, but within the limits stated
above.
As seen from both the cross-docking results and the poses re-
ceived from docking our own compounds, there were no big struc-
tural changes within the overall structure of the chosen binding
pocket conformations with the settings we selected (temporarily
mutating only the side chain of Leu300 to Ala, scaling of the van
der Waals radii for the receptor atoms by 70%). One might thus
wonder whether it is reasonable at all to use an induced-fit dock-
ing method in this case, or whether an ensemble docking approach,
using a number of different rigid binding sites, would also have
been successful. In an attempt to answer this question, we docked
5b into the rigid binding pocket of 2FZD, using Glide with default
settings and no constraints. The best-ranked pose obtained from
this experiment showed a GlideScore value of À6.76, in contrast
to À11.472 obtained in the IFD experiment where the same bind-
ing pocket was used. Additionally, no interactions with the cata-
lytic centre were found by any pose from this run. Since the bent
pose obtained during the IFD run of 5b into 2FZD was not retrieved
in any other IFD experiment, we would have simply missed this
reasonable pose when using only rigid binding pockets.
Regarding the validity of using structural information of human
vs. bovine aldose reductase, we came to the following conclusion:
whilst both Leu300 and Ser302 play an important role in the binding
oftheligands, thesidechainofLeuresp. Val301pointsawayfromthe
binding site. We obtained no results from the docking into the hu-
man structures that would suggest an important role of this residue
in ligandbinding. The side chain of the proteinat position 297is even
further away from both experimental and modeled ligand poses.
Moreover, compoundsdocked into the homology model gave results
similar to those of docking into 2FZD. We thus conclude that com-
paring binding and modeling studies between bovine and human
ALR2 proteins is a valid approach.
The inspection of the highest ranked docking poses gave similar
results for 5-(4-hydroxybenzylidene) derivatives 3a, 4i and 5g
(exemplified by 3a, Fig. 4 left side). All of these compounds appear
to bind in a mode that is flipped in comparison to the one originally
proposed. In particular, for compound 3a, we received a pose that
is quite different from the one that we had suggested previously.24
Thus, the deprotonated phenol group incorporates the catalytic
centre with the usual hydrogen bonds donated by Tyr48 and
His110, whilst the acetic acid side chain forms a tight net of hydro-
gen bonds with the surrounding backbone NH groups of Leu301
and Ser302 as well as the hydroxyl group of Ser302. In addition,
the carbonyl group at position 2 of the thiazolidinedione ring forms
a hydrogen bond with the backbone NH groups of Leu300 and
Ala299. Finally, a weak hydrogen bond between the thiol group
of Cys298 and the ring sulfur atom of the ligands is suggested.
The carboxylic acid 3a, which showed the highest activity amongst
the three compounds, is also able to establish the highest number
of hydrogen bonds by interacting with every backbone NH group
from Ala299 to Ser302 and also with the hydroxyl group of
Ser302. It should be noted that similar interactions with Ser302
can be seen in 2NVC, where the acetic acid side chain of the ligand
that is connected via the ester group is placed at a similar position
(Fig. 3). Although no positive charge is located in close vicinity of
the backbone binding area between Ala299 and Ser302, the multi-
ple donated hydrogen bonds should be well able to stabilize the
negative charge of the carboxylates. These results also provide a
possible explanation as to why the phenol derivative 4i is the only
compound from the amide series which shows such a high affinity,
whilst other amides that have a highly active twin in the N-hydrox-
yacetamide series (4b/5b; 4d/5e) or in the acetic acid series23–25
(4a–d, 4f, 4h, 4j, and 4k) are virtually inactive or barely active. Fur-
thermore, if compounds of series 4 were to bind to the catalytic
centre via the amide group, then small alterations at the phenyl
ring as seen for compounds 4f, 4g, 4h, 4j and 4k should not have
such a big impact on the binding, since these groups would then
point outwards from the binding site. A similar correlation is also
seen in our acetic acid methyl ester series reported earlier, where
the corresponding compound (Ar = 4-OH-C6H4, IC50 = 6.18 M)24 is
one of the few derivatives with micromolar affinity.
5. Conclusion
The in vitro evaluation of (5-arylidene-2,4-dioxothiazolidin-3-
yl)acetamides (4) and analogous N-hydroxyacetamides (5) high-
lighted that the replacement of the carboxylic anionic head of acids
3 withthe carboxamideor N-hydroxycarboxamide groupproduces a
general decrease in the in vitro ALR2 inhibitory effect, whereas in
comparison with N-unsubstituted analogues 1, the insertion of the
N-hydroxyacetamide chain on N-3 (compounds 5) gave a general
improvement in activity.
The second investigated acetamide derivative 4g, which was
inactive in our assays, gave the highest GlideScore values (Table
4) with its lowest score lying at 1.999 or more above the lowest
score for the other investigated compounds.
Ourdockingstudies suggest thatthe appreciable inhibitoryactiv-
ity of 4i, which stands out amongst compounds 4, is related to the
capability of the phenolic group to bind the polar region of the
ALR2 active site.
For the more lipophilic compounds 5b and 5e, 5-(4-phenoxyb-
enzylidene)- and 5-naphthalen-1-ylmethylidene substituted,
respectively, we received a second type of pose that suggests inter-
actions with the lipophilic selectivity pocket in a way similar to
that of tolrestat (exemplified by 5b, Fig. 4 right side). In these
Out of our novel non-carboxylic acid bearing 2,4-thiazolidinedi-
one derivatives 4 and 5 the ones that exhibited micromolar ALR2
binding affinity will be used as starting points for a future drug dis-
covery programme aimed at enhancing their activity profiles and
evaluating their in vivo effectiveness.

5848
R. Maccari et al. / Bioorg. Med. Chem. 16 (2008) 5840–5852
Figure 4. Depiction of the two major types of binding modes predicted for our ligands, based on the lowest GlideScore values in Table 4, exemplified by 3a docked into 2NVC
(left side),37 and 5b docked into 2FZD (right side).35 Top: Docked poses displayed in PyMOL,46 same representation as in Figure 3. Middle: The same poses visualized in
LigandScout,43 with pharmacophore features detected by the software. Features showing pharmacophoric interactions between ligand and binding site: red arrows:
hydrogen bond acceptors, red star: negative ionizable/charged, purple ring and arrows: aromatic plane, yellow spheres: hydrophobic. The cofactor NADP+, placed behind the
ligands in this view, is represented in stick-mode. Bottom: 2D depiction of interactions detected by LigandScout with the surrounding amino acids and the cofactor. HIE:
histidine (HIS) with hydrogen on the epsilon nitrogen, as assigned by Maestro during the protein preparation step.
We have elaborated a procedure for induced-fit docking of
compounds into the binding pocket of ALR2, taking into account
the diverse complex structures available as well as the current
knowledge regarding the protonation state of the ligands. The
promising results give us a possible explanation of our binding
data, especially for the 5-(4-hydroxybenzylidene) derivatives
from our different series. Our IFD studies indicate that these lat-
ter can bind to the catalytic centre via the deprotonated phenol
group and that they can form a number of hydrogen bonds with
the protein backbone near Ser 302 via the acetic chain. A bind-
ing mode that is closer to the previously assumed one,24 where
the carboxylic or hydroxamic acid residue is located at the rec-
ognition region and the arylidene residue is placed at the hydro-
phobic specificity pocket of the enzyme, is suggested for
compounds with a more lipophilic and non-acidic arylidene res-
idue. We think that the working procedure presented herein has
expanded the knowledge about the applicability of induced-fit
docking methodology. It has also aided the modelling of this
well-known target protein which, due to the flexibility of its
binding site, is still highly challenging.

R. Maccari et al. / Bioorg. Med. Chem. 16 (2008) 5840–5852
5849
6.2.4. (Z)-2-(5-Naphthalen-1-ylmethylene-2,4-dioxothiazolidin-
3-yl)acetamide (4d)
6. Experimental
6.1. Chemistry
Yield 77%; mp 296–298 °C; 1H NMR (DMSO-d6): d 4.27 (s, 2H,
NCH2); 7.35 and 7.58 (2 br s exchangeable with D2O, 2H, NH2);
7.64–8.14 (2m, 7H, arom); 8.58 (s, 1H, CH); 13C NMR (DMSO-d6):
d 43.8 (NCH2); 123.9, 126.2, 127.1, 127.5, 128.1, 129.5, 130.9 (CH
arom); 125.4 (5-C); 131.6 (CH methylidene); 130.7, 131.5, 133.8
(Cq arom); 165.4, 167.5, 168.0 (C@O); Anal. (C16H12N2O3S) C, H, N.
Melting points were recorded on a Kofler hot-stage appara-
tus and are uncorrected. TLC controls were carried out on pre-
coated silica gel plates (F 254 Merck). Elemental analyses (C, H,
N), determined by means of a C. Erba mod. 1106 elem. Ana-
lyzer, were within 0.4% of theory. 1H and 13C NMR spectra
were recorded on a Varian 300 magnetic resonance spectrome-
ter (300 MHz for 1H and 75 MHz for 13C). Chemical shifts are
given in d units (ppm) relative to internal standard Me4Si and
refer to DMSO-d6 solutions. Coupling constants (J) are given in
hertz (Hz). 13C NMR spectra were determined by Attached Pro-
ton Test (APT) experiments and the resonances were always
attributed by proton-carbon heteronuclear chemical shift corre-
lation. A microwave oven Discover (CEM) was used to carry out
microwave-assisted reactions. Unless stated otherwise, all mate-
rials were obtained from commercial suppliers and used with-
out further purification.
6.2.5. (Z)-2-(5-Naphthalen-2-ylmethylene-2,4-dioxothiazolidin-
3-yl)acetamide (4e)
Yield 22%; mp 266–269 °C; 1H NMR (DMSO-d6): d 4.26 (s, 2H,
NCH2); 7.35 and 7.77 (2 br s exchangeable with D2O, 2H, NH2);
7.61–8.10 (3m, 7H, arom); 8.25 (s, 1H, CH); 13C NMR (DMSO-d6):
d 43.8 (NCH2); 122.0 (5-C); 126.4, 127.7, 128.2, 128.6, 129.3,
129.5, 131.6 (CH arom); 131.0, 133.2, 133.9 (Cq arom); 133.7 (CH
methylidene); 165.8, 167.3, 167.6 (C@O); Anal. (C16H12N2O3S) C,
H, N.
6.2.6. (Z)-2-[5-(3-Methoxybenzylidene)-2,4-dioxothiazolidin-3-
yl]acetamide (4f)
Yield 63%; mp 210–212 °C; 1H NMR (DMSO-d6): d 3.81 (s, 3H,
OCH3); 4.23 (s, 2H, NCH2); 7.12 (d, J = 8.9 Hz, 1H, arom); 7.22 (m,
2H, arom); 7.50 (dd J = 8.9 and 8.9 Hz, 1H, arom); 7.34 and 7.75
(2 br s exchangeable with D2O, 2H, NH2); 7.93 (s, 1H, CH); 13C
NMR (DMSO-d6): d 43.3 (NCH2); 55.3 (OCH3); 115.5, 116.6, 121.9,
130.5 (CH arom); 121.5 (5-C); 133.2 (CH methylidene); 134.2,
159.6 (Cq arom); 165.3, 166.8, 167.1 (C@O); Anal. (C13H12N2O4S)
C, H, N.
6.2. General method for the synthesis of 2-(5-arylidene-2,4-
dioxothiazolidin-3-yl)acetamides (4a–k)
A mixture of 5-arylidene-2,4-thiazolidinedione 1 (17 mmol)
and potassium carbonate (1.93 g, 14 mmol) in acetonitrile was
refluxed for 45 min. Then
a solution of 2-chloroacetamide
(1.87 g, 20 mmol) in acetonitrile was slowly added and the mix-
ture was refluxed for 24–72 h. After evaporation of the solvent
under reduced pressure, the crude solid was washed with water,
until the pH value of the filtrate was neutral, and then with ace-
tonitrile, to give pure acetamide 4.
Compounds 4h–k were synthesised as described above, adding
a solution of 2-chloroacetamide (1.59 g, 17 mmol) in acetonitrile
dropwise to a solution of compound 1 (17 mmol) and potassium
carbonate (1.93 g, 14 mmol) in acetonitrile for 2 h.
6.2.7. (Z)-2-[5-(4-Methoxybenzylidene)-2,4-dioxothiazolidin-3-
yl]acetamide (4g)
Yield 32%; mp 263–264 °C; 1H NMR (DMSO-d6): d 3.84 (s, 3H,
OCH3); 4.23 (s, 2H, NCH2); 7.13 (m, 2H, arom); 7.61 (m, 2H, arom);
7.34 and 7.74 (2 br s exchangeable with D2O, 2H, NH2); 7.92 (s, 1H,
CH); 13C NMR (DMSO-d6): d 43.8 (NCH2); 56.2 (OCH3); 115.6, 132.7
(CH arom); 125.9 (5-C); 133.8 (CH methylidene); 118.5, 161.7 (Cq
arom); 165.9, 167.3, 167.6 (C@O); Anal. (C13H12N2O4S) C, H, N.
6.2.1. (Z)-2-[2,4-Dioxo-5-(3-phenoxybenzylidene)thiazolidin-3-
yl]acetamide (4a)
Yield 64%; mp 238–239 °C; 1H NMR (DMSO-d6): d 4.19 (s, 2H,
NCH2); 7.06–7.57 (3m, 9H, arom); 7.33 and 7.76 (2 br s exchange-
able with D2O, 2H, NH2); 7.94 (s, 1H, CH). 13C NMR (DMSO-d6): d
43.8 (NCH2); 119.4, 119.7, 120.8, 124.6, 125.2, 130.7, 131.5 (CH
arom); 122.6 (5-C); 133.0 (CH methylidene); 135.2, 156.2, 158.0
(Cq arom); 165.5, 167.1, 167.3 (C@O); Anal. (C18H14N2O4S) C, H, N.
6.2.8. (Z)-2-[5-(3-Hydroxybenzylidene)-2,4-dioxothiazolidin-3-
yl]acetamide (4h)
Yield 49%; mp 251–253 °C; 1H NMR (DMSO-d6): d 4.22 (s, 2H,
NCH2); 6.90 (dd, J = 7.8 and 1.5 Hz, 1H, arom); 7.01 (s, 1H, arom);
7.08 (d, J = 7.8 Hz, 1H, arom); 7.34 (dd, J = 7.8 and 7.8 Hz, 1H,
arom); 7.31 and 7.74 (2 br s exchangeable with D2O, 2H, NH2);
7.86 (s, 1H, CH); 9.90 (br s exchangeable with D2O, 1H, OH); 13C
NMR (DMSO-d6): d 43.8 (NCH2); 116.6, 118.6, 122.0, 131.1 (CH
arom); 121.7 (5-C); 134.0 (CH methylidene); 134.7, 158.5 (Cq
arom); 165.8, 167.3, 167.6 (C@O); Anal. (C12H10N2O4S) C, H, N.
6.2.2. (Z)-2-[2,4-Dioxo-5-(4-phenoxybenzylidene)thiazolidin-3-
yl]acetamide (4b)
Yield 49%; mp 198–200 °C; 1H NMR (DMSO-d6): d 4.19 (s, 2H,
NCH2); 7.07–7.23 (2m, 5H, arom); 7.30 and 7.71 (2 br s exchange-
able with D2O, 2H, NH2); 7.40–7.45 (m, 2H, arom); 7.65 (m, 2H,
arom); 7.90 (s, 1H, CH); 13C NMR (DMSO-d6): d 43.7 (NCH2);
118.8, 120.3, 125.1, 130.7, 132.9 (CH arom); 120.0 (5-C); 133.1
(CH methylidene); 128.0, 155.6, 159.7 (Cq arom); 165.8, 167.3,
167.5 (C@O); Anal. (C18H14N2O4S) C, H, N.
6.2.9. (Z)-2-[5-(4-Hydroxybenzylidene)-2,4-dioxothiazolidin-3-
yl]acetamide (4i)
Yield 37%; mp 268–270 °C; 1H NMR (DMSO-d6): d 4.19 (s, 2H,
NCH2); 6.90 (m, 2H, arom); 7.48 (m, 2H, arom); 7.29 and 7.69 (2
br s exchangeable with D2O, 2H, NH2); 7.82 (s, 1H, CH); 10.35 (br
s exchangeable with D2O, 1H, OH); 13C NMR (DMSO-d6): d 43.8
(NCH2); 117.0, 133.1 (CH arom); 117.2 (5-C); 134.2 (CH methyli-
dene); 124.4, 160.8 (Cq arom); 166.0, 167.4, 167.7 (C@O); Anal.
(C12H10N2O4S) C, H, N.
6.2.3. (Z)-2-(5-Biphenyl-4-ylmethylene-2,4-dioxothiazolidin-3-
yl)acetamide (4c)
Yield 24%; mp 244–245 °C; 1H NMR (DMSO-d6): d 4.24 (s, 2H,
NCH2); 7.33 and 7.69 (2 br s exchangeable with D2O, 2H, NH2);
7.41–7.54 (2m, 3H, arom); 7.73–7.77 (m, 4H, arom); 7.87 (m, 2H,
arom); 8.01 (s, 1H, CH); 13C NMR (DMSO-d6): d 44.0 (NCH2);
121.7 (5-C); 127.5, 128.2, 129.2, 129.9, 131.6 (CH arom); 133.6
(CH methylidene); 132.6, 139.4, 142.9 (Cq arom); 166.2, 167.9,
168.0 (C@O); Anal. (C18H14N2O3S) C, H, N.
6.2.10. (Z)-2-[5-(4-Hydroxy-3-methoxybenzylidene)-2,4-
dioxothiazolidin-3-yl]acetamide (4j)
Yield 42%; mp 243–245 °C; 1H NMR (DMSO-d6): d 3.82 (s, 3H,
OCH3); 4.20 (s, 2H, NCH2); 6.93 (d, J = 8.7 Hz, 1H, arom); 7.11 (d,
J = 8.7 Hz, 1H, arom); 7.21 (s, 1H, arom); 7.30 and 7.71 (2 br s
exchangeable with D2O, 2H, NH2); 7.84 (s, 1H, CH); 10.02 (br s

5850
R. Maccari et al. / Bioorg. Med. Chem. 16 (2008) 5840–5852
exchangeable with D2O, 1H, OH); 13C NMR (DMSO-d6): d 43.7
(NCH2); 56.1 (OCH3); 114.8, 116.7, 124.9 (CH arom); 117.5 (5-C);
134.3 (CH methylidene); 125.0, 148.6, 150.3 (Cq arom); 165.8,
167.3, 167.9 (C@O); Anal. (C13H12N2O5S) C, H, N.
arom); 133.3 (CH methylidene); 128.1, 155.6, 159.7 (Cq arom);
162.8, 167.5, 168.2 (C@O); Anal. (C18H14N2O5S) C, H, N.
6.3.3. (Z)-2-[5-(4-Benzyloxybenzylidene)-2,4-dioxothiazolidin-
3-yl]-N-hydroxyacetamide (5c)
Yield 65%; mp 164–165 °C; 1H NMR (DMSO-d6): d 4.63 (s, 2H,
NCH2); 5.20 (s, 2H, OCH2); 7.20 (m, 2H, arom); 7.37–7.47 (m, 5H,
arom); 7.62 (m, 2H, arom); 7.97 (s, 1H, CH); 9.02 and 10.82 (2 br
s exchangeable with D2O, 2H, NH and OH); 13C NMR (DMSO-d6):
d 46.2 (NCH2); 70.3 (OCH2); 118.0 (5-C); 116.6, 128.2, 128.5,
129.3, 133.2 (CH arom); 134.9 (CH methylidene); 126.1, 137.2,
160.5 (Cq arom); 163.7, 166.3, 167.6 (C@O); Anal. (C19H16N2O5S)
C, H, N.
6.2.11. (Z)-2-[5-(3-Hydroxy-4-methoxybenzylidene)-2,4-
dioxothiazolidin-3-yl]acetamide (4k)
Yield 60%; mp 268–270 °C; 1H NMR (DMSO-d6): d 3.81 (s, 3H,
OCH3); 4.20 (s, 2H, NCH2); 7.06–7.15 (m, 3H, arom); 7.31 and
7.71 (2 br s exchangeable with D2O, 2H, NH2); 7.79 (s, 1H, CH);
9.54 (br s exchangeable with D2O, 1H, OH); 13C NMR (DMSO-d6):
d 43.7 (NCH2); 56.2 (OCH3); 113.1, 116.5, 124.2 (CH arom); 118.2
(5-C); 134.2 (CH methylidene); 126.1, 147.5, 151.0 (Cq arom);
166.0, 167.4, 167.8 (C@O); Anal. (C13H12N2O5S) C, H, N.
6.3.4. (Z)-2-(5-Biphenyl-4-ylmethylene-2,4-dioxothiazolidin-3-
yl)-N-hydroxyacetamide (5d)
6.3. General method for the synthesis of 2-(5-arylidene-2,4-
dioxothiazolidin-3-yl)-N-hydroxyacetamides (5a–g)
Yield 40%; mp 219–220 °C; 1H NMR (DMSO-d6): d 4.21 (s, 2H,
NCH2); 7.40–7.52 (m, 4H, arom); 7.72–7.76 (m, 3H, arom); 7.83
(m, 2H, arom); 7.98 (s, 1H, CH); 9.05 and 10.98 (2 br s exchange-
able with D2O, 2H, NH and OH); 13C NMR (DMSO-d6): d 41.5
(NCH2); 120.7 (5-C); 126.8, 127.5, 128.2, 129.1, 130.8 (CH arom);
132.8 (CH methylidene); 132.0, 138.6, 142.0 (Cq arom); 163.5,
166.7, 167.6 (C@O); Anal. (C18H14N2O4S) C, H, N.
Method (A): A mixture of (5-arylidene-2,4-dioxothiazolidin-3-
yl)acetic acid 3 (14 mmol), ethyl chlorofomate (2.38 g, 22 mmol)
and triethylamine (2.55 g, 25.2 mmol) in diethylether was stirred
at 0 °C for 15 min. At the same time, a solution of hydroxylamine
hydrochloride (1.45 g, 21 mmol) in anhydrous methanol was
added to a solution of sodium hydroxide (0.84 g, 21 mmol) in
anhydrous methanol; the mixture was stirred at 0 °C for 15 min;
then the solid was removed and the filtrate was used as such.
The former mixture was added to this freshly prepared solution
of hydroxylamine and stirred at room temperature for 24–40 h.
After evaporation of the solvent in vacuo, the crude solid was
washed with diethylether; the solid was dried and then washed
with water to provide pure compound 5.
Method (B): A mixture of (5-arylidene-2,4-dioxothiazolidin-3-
yl)acetic acid 3 (14 mmol), ethyl chlorofomate (2.38 g, 22 mmol)
and triethylamine (2.55 g, 25.2 mmol) in diethylether in a hermet-
ically sealed vial was irradiated with microwaves (80 °C, max
150 psi, max 300 W) for 5 min. At the same time, a solution of
hydroxylamine hydrochloride (1.45 g, 21 mmol) in anhydrous
methanol was added to a solution of sodium hydroxide (0.84 g,
21 mmol) in anhydrous methanol; the mixture was stirred at
0 °C for 15 min; then the solid was removed and the filtrate was
used as such. The former mixture was added to the freshly pre-
pared solution of hydroxylamine and the mixture was irradiated
with microwaves (80 °C, max 150 psi, max 300 W) for 5 min. After
evaporation of the solvent in vacuo, the crude solid was washed
with diethylether; the solid was dried and then washed with
water to provide pure compound 5. Yields reported below refer
to method B.
6.3.5. (Z)-N-Hydroxy-2-(5-Naphthalen-1-ylmethylene-2,4-
dioxothiazolidin-3-yl)acetamide (5e)
Yield 53%; mp 218–219 °C; 1H NMR (DMSO-d6): d 4.24 (s, 2H,
NCH2); 7.61–7.73 (m, 3H, arom); 8.02–8.11 (m, 4H, arom); 8.58
(s, 1H, CH); 9.07 and 10.99 (2 br s exchangeable with D2O, 2H,
NH and OH); 13C NMR (DMSO-d6): d 42.2 (NCH2); 125.3 (5-C);
123.9, 126.1, 126.9, 127.4, 128.1, 129.4, 130.8 (CH arom); 131.5
(CH methylidene); 130.6, 131.4, 133.8 (Cq arom); 162.7, 167.8,
168.1 (C@O); Anal. (C16H12N2O4S) C, H, N.
6.3.6. (Z)-N-Hydroxy-2-(5-Naphthalen-2-ylmethylene-2,4-
dioxothiazolidin-3-yl)acetamide (5f)
Yield 46%; mp 286–287 °C dec.; 1H NMR (DMSO-d6): d 4.24 (s,
2H, NCH2); 7.58–7.70 (m, 3H, arom); 7.93–8.10 (m, 4H, arom);
8.18 (s, 1H, CH); 9.06 and 11.00 (2 br s exchangeable with D2O,
2H, NH and OH); 13C NMR (DMSO-d6): d 42.0 (NCH2); 124.4 (5-
C); 126.5, 127.6, 128.2, 128.5, 129.2, 129.4, 131.3 (CH arom);
132.2 (CH methylidene); 131.2, 133.2, 133.8 (Cq arom); 162.5,
167.9, 168.4 (C@O); Anal. (C16H12N2O4S) C, H, N.
6.3.7. (Z)-N-Hydroxy-2-[5-(4-Hydroxybenzylidene)-2,4-
dioxothiazolidin-3-yl]acetamide (5g)
Yield 63%; mp 155–157 °C. 1H NMR (DMSO-d6): d 4.19 (s, 2H,
NCH2), 6.96 (m, 2H, arom), 7.53 (m, 2H, arom), 7.88 (s, 1H, CH),
8.63, 9.01 and 10.82 (3 br s exchangeable with D2O, 3H, NH and
2 OH). 13C NMR (DMSO-d6): d 43.9 (NCH2); 117.0, 133.1 (CH arom);
124.3 (5-C); 134.2 (CH methylidene); 124.8, 160.8 (Cq arom);
166.0, 167.7, 168.6 (C@O); Anal. (C12H10N2O5S) C, H, N.
6.3.1. (Z)-N-Hydroxy-2-[2,4-Dioxo-5-(3-phenoxybenzylidene)-
thiazolidin-3-yl]acetamide (5a)
Yield 45%; mp > 230 °C 1H NMR (DMSO-d6): d 4.19 (s, 2H,
NCH2); 7.09-7.59 (3m, 9H, arom); 7.95 (s, 1H, CH); 9.03 and
10.86 (2 br s exchangeable with D2O, 2H, NH and OH); 13C NMR
(DMSO-d6): d 42.0 (NCH2); 119.4, 119.7, 120.8, 124.6, 125.2,
130.7, 131.5 (CH arom); 122.5 (5-C); 133.2 (CH methylidene);
135.2, 156.2, 158.1 (Cq arom); 162.6, 167.3, 168.0 (C@O); Anal.
(C18H14N2O5S) C, H, N.
6.4. Molecular modeling
All molecular modeling studies were performed using the
Schrodinger Suite 2007, including Maestro 8.0, Glide 4.5, and Prime
1.6,41 on a PC equipped with a 2.13 GHz Core 2 Duo processor and
1 GB of RAM, running Fedora Core 6 Linux.
The IFD protocol developed by Schrödinger, LLC.40 consists of
three main steps, that may be summarized as follows. First, an ini-
tial docking run is performed with the Glide docking module, dur-
ing which the van der Waals radii of both ligand and binding site
atoms are scaled (by 50% for both, if no side chains are selected
for removal, otherwise by 50% for the ligand and 70% for the
receptor), and user-selected side chains are temporarily mutated
6.3.2. (Z)-N-hydroxy-2-[2,4-Dioxo-5-(4-phenoxybenzylidene)-
thiazolidin-3-yl]acetamide (5b)
Yield 60%; mp 170 °C; 1H NMR (DMSO-d6): d 4.19 (s, 2H, NCH2);
7.09–7.11 (m, 4H, arom); 7.21 (m, 1H, arom); 7.43 (m, 2H, arom);
7.66 (m, 2H, arom); 7.93 (s, 1H, CH); 9.04 and 10.85 (2 br s
exchangeable with D2O, 2H, NH and OH); 13C NMR (DMSO-d6): d
42.0 (NCH2); 119.9 (5-C); 118.8, 120.4, 125.2, 130.8, 133.0 (CH

R. Maccari et al. / Bioorg. Med. Chem. 16 (2008) 5840–5852
5851
to alanine, to reduce steric clashes. In the second step, the replaced
residues are restored and side-chain prediction and minimization
are performed with the Prime module for a given number of struc-
tures (20 by default) of every ligand–protein complex obtained in
the first step. In the final step, the ligands are redocked into the in-
duced-fit structures from the previous stage that fall within a given
energy value of the lowest energy structure (default value 30 kcal/
mol).
The selected PDB structures were processed in Maestro (Schrö-
dinger, LLC)41 using the Protein Preparation Wizard, which per-
forms the following steps: assigning of bond orders, addition of
hydrogens, optimization of hydrogen bonds by flipping amino side
chains, correction of charges, and minimization of the protein com-
plex. All water molecules were removed from the protein struc-
ture. Automatically assigned bond orders, protonation states of
the ligands, and the orientation and tautomeric state of His110
were inspected manually and corrected where necessary. For easy
comparison of the results produced by docking into the different
complexes, the five selected complexes were aligned with the Pro-
tein Structure Alignment tool in Maestro.
0.25 M potassium phosphate buffer pH 6.8, 0.38 M ammonium sul-
fate, 0.11 mM NADPH and 4.7 mM -glyceraldehyde as substrate
D,L
in a final volume of 1.5 mL. All inhibitors were dissolved in DMSO,
the final concentration of this solvent in the reaction mixture was
1%. To correct for the non-enzymatic oxidation of the cofactor, the
rate of NADPH oxidation in the presence of all the components ex-
cept the substrate was subtracted from each experimental rate.
Each dose–effect curve was generated using at least three concen-
trations of inhibitor that caused an inhibition between 20% and
80%. Each concentration was tested in duplicate and IC50 values
as well as the 95% confidence limits (95% CL) were obtained by
using CalcuSyn software for dose effect analysis.45
Preliminarily, the solubility and stability of the tested com-
pounds were evaluated. They are soluble in a DMSO/H2O 1% mix-
ture and proved to be stable under the same experimental
conditions of the enzymatic assay.
Supplementary data
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.bmc.2008.04.072.
All acidic groups, including the weaker phenolic and hydroxa-
mic acids and the hydantoin residue of fidarestat, were prepared
in their deprotonated state. All ligands that were selected for dock-
ing were minimized using the OPLS_2005 force field.
References and notes
The selected ligands were then docked into each of the five pre-
pared protein binding sites, using the Induced-Fit Docking work-
flow accessible via the Maestro graphical interface. The pocket
was defined by selecting the ligand that is part of the respective
complex. In order to take into account the high flexibility of
Leu300, the side chain of this residue was changed to Ala in the ini-
tial docking step. The other settings were used at their default val-
ues, with Protein Prep constrained refinement turned on, van der
Waals radius scaling of 0.7 for the protein and 0.5 for the ligand,
standard precision to be used during both initial docking and final
redocking, a maximum of 20 poses to be carried forward for each
ligand after initial docking, refining of residues within 5.0 Å of
the ligand poses, optimization of the side chains, and redocking
into structures within 30.0 kcal/mol of the best and within the
top 20 overall structures.
1. Hanefeld, M.; Fischer, S.; Julius, U.; Schulze, J.; Schwanebeck, U.; Schmechel, H.;
Ziegelasch, H. J.; Lindner, J. Diabetologia 1996, 39, 1577.
2. Molitch, M. E.; DeFronzo, R. A.; Franz, M. J.; Keane, W. F.; Mogensen, C. E.;
Parving, H. H. Diabetes Care 2003, 26(Suppl. 1), S94.
3. Viberti, G. J. Diabetes Complications 2005, 19, 168.
4. Brownlee, M. Nature 2001, 414, 813.
5. Sheetz, M. J.; King, G. L. JAMA 2002, 288, 2579.
6. Eastman, R. C.; Siebert, C. W.; Harris, M.; Golden, P. J. Clin. Endocrinol. Metab.
1993, 77, 1105.
7. Wild, S.; Roglic, G.; Green, A.; Sicree, R.; King, H. Diabetes Care 2004, 27, 1047.
8. Yabe-Nishimura, C. Pharmacol. Rev. 1998, 50, 21.
9. Srivastava, S. K.; Ramana, K. V.; Bhatnagar, A. Endocr. Rev. 2005, 26, 380.
10. Oyama, T.; Miyasita, Y.; Watanabe, H.; Shirai, K. Diabetes Res. Clin. Pract. 2006,
73, 227.
11. Dixit, B. L.; Balendiran, G. K.; Watowich, S. J.; Srivastava, S.; Ramana, K. V.;
Petrash, J. M.; Bhatnagar, A.; Srivastava, S. K. J. Biol. Chem. 2000, 275, 21587.
12. (a) Ramana, K. V.; Friedrich, B.; Srivastava, S.; Bhatnagar, A.; Srivastava, S. K.
Diabetes 2004, 53, 2910; (b) Ramana, K. V.; Friedrich, B.; Tammali, R.; West, M.
B.; Bhatnagar, A.; Srivastava, S. K. Diabetes 2005, 54, 818.
Comparison of the cross-docking poses was performed by
RMSD calculation on the redocked native ligands and their corre-
sponding experimentally obtained conformations using the Super-
position tool, after import of the previously aligned protein
structures and without transformation of coordinates during
superposition.
13. Ramana, K. V.; Srivastava, S. K. Cytokine 2006, 36, 115.
14. Costantino, L.; Rastelli, G.; Cignarella, G.; Vianello, P.; Barlocco, D. Exp. Opin.
Ther. Patents 1997, 7, 843.
15. Suzen, S.; Buyukbingol, E. Curr. Med. Chem. 2003, 10, 1329.
16. Tammali, R.; Ramana, K. V.; Srivastava, S. K. Cancer Lett. 2007, 252, 299.
17. Costantino, L.; Rastelli, G.; Vianello, P.; Cignarella, G.; Barlocco, D. Med. Res. Rev.
1999, 19, 3.
18. Costantino, L.; Rastelli, G.; Gamberini, M. C.; Barlocco, D. Expert Opin. Ther.
Patents 2000, 10, 1245.
19. Miyamoto, S. Chem. Biol. Inform. J. 2002, 2, 74.
20. El-Kabbani, O.; Ruiz, F.; Darmanin, C.; Chung, R. P.-T. Cell. Mol. Life Sci. 2004, 61,
750.
6.5. Determination of the in vitro aldose reductase inhibition
NADPH, D,L-glyceraldehyde and dithiothreitol (DTT) were pur-
21. Castañer, J.; Prous, J. Drugs Future 1987, 12, 336.
22. Hotta, N.; Toyota, T.; Matsuoka, K.; Shigeta, Y.; Kikkawa, R.; Kaneko, T.;
Takahashi, A.; Sugimura, K.; Koike, Y.; Ishii, J.; Sukamoto, N. Diabetes Care 2001,
24, 1776.
23. Bruno, G.; Costantino, L.; Curinga, C.; Maccari, R.; Monforte, F.; Nicolò, F.;
Ottanà, R.; Vigorita, M. G. Bioorg. Med. Chem. 2002, 10, 1077.
24. Maccari, R.; Ottanà, R.; Curinga, C.; Vigorita, M. G.; Rakowitz, D.; Steindl, T.;
Langer, T. Bioorg. Med. Chem. 2005, 13, 2809.
chased from Sigma Chemical Co. DEAE-cellulose (DE-52) was ob-
tained from Whatman. Sorbinil was a gift from Prof. Dr. Luca
Costantino, University of Modena (Italy) and was used as standard.
All other chemicals were commercial samples of good grade. Calf
lenses for the isolation of ALR2 were obtained locally from freshly
slaughtered animals. The enzyme was isolated and purified by a
chromatographic procedure as previously described.44 Briefly,
ALR2 was released by carving the capsule and the frozen lenses
were suspended in potassium phosphate buffer pH 7.0 containing
5 mM DTT and stirred in an ice-cold bath for two hours. The sus-
pension was centrifuged for 30 min at 4000 rpm at 4 °C and the
supernatant was subjected to ion exchange chromatography on
DE-52. Enzyme activity was assayed spectrophotometrically on a
Cecil Super Aurius CE 3041 spectrophotometer by measuring the
decrease in absorption of the cofactor NADPH at 340 nm which
accompanies the oxidation of NADPH catalyzed by the enzyme.
The assay was performed at 37 °C in a reaction mixture containing
25. Maccari, R.; Ottanà, R.; Ciurleo, R.; Vigorita, M. G.; Rakowitz, D.; Steindl, T.;
Langer, T. Bioorg. Med. Chem. Lett. 2007, 17, 3886.
26. Urzhumtsev, A.; Tete-Favier, F.; Mitschler, A.; Barbanton, J.; Barth, P.;
Urzhumtseva, L.; Biellmann, J. F.; Podjarny, A.; Moras, D. Structure 1997, 5, 601.
27. El-Kabbani, O.; Wilson, D. K.; Petrash, J. M.; Quiocho, F. A. Mol. Vis. 1998, 4, 19.
28. Lee, Y. S.; Hodoscek, M.; Kador, P. F.; Sugiyama, K. Chem. Biol. Interact. 2003,
143–144, 307.
29. Vanjari, H.; Pande, R. J. Pharm. Biomed. Anal. 2003, 33, 783.
30. (a) Muri, E. M. F.; Nieto, M. J.; Sindelar, R. D.; Williamson, J. S. Curr. Med. Chem.
2002, 9, 1631; (b) Muri, E. M. F.; Nieto, M. J.; Williamson, J. S. Med. Chem. Rev.
2004, 10, 385; (c) Leoni, F.; Zaliani, A.; Bertolini, G.; Porro, G.; Pagani, P.; Pozzi,
P.; Dona, G.; Fossati, G.; Sozzoni, S.; Azam, T.; Bufler, R.; Fantuzzi, G.;
Goncharav, I.; Kim, S. H.; Pomerantz, B. J.; Reznikov, L. L.; Siegmund, B.;
Dinarello, C. A.; Mascagni, P. Proc. Natl. Acad. Sci. U.S.A. 2002, 99, 2995; (d) Clare,
B. W.; Scozzafava, A.; Supuran, C. T. J. Med. Chem. 2001, 44, 2253.

5852
R. Maccari et al. / Bioorg. Med. Chem. 16 (2008) 5840–5852
31. Reddy, A. S.; Kumar, M. S.; Reddy, G. R. Tetrahedron Lett. 2000, 41,
6285.
38. Steuber, H.; Zentgraf, M.; Podjarny, A.; Heine, A.; Klebe, G. J. Mol. Biol. 2006,
356, 45.
32. (a) Berman, H. M.; Westbrook, J.; Feng, Z.; Gilliland, G.; Bhat, T. N.; Weissig, H.;
Shindyalov, I. N.; Bourne, P. E. Nucleic Acids Res. 2000, 28, 235; (b) http://
www.rcsb.org/pdb.
39. Marvin Calculator Plugins, http://www.chemaxon.com.
40. Sherman, W.; Day, T.; Jacobson, M. P.; Friesner, R. A.; Farid, R. J. Med. Chem.
2006, 49, 534.
33. Zentgraf, M.; Steuber, H.; Koch, C.; La Motta, C.; Sartini, S.; Sotriffer, C. A.; Klebe,
G. Angew. Chem., Int. Ed. Engl. 2007, 46, 3575.
34. El-Kabbani, O.; Darmanin, C.; Schneider, T. R.; Hazemann, I.; Ruiz, F.; Oka, M.;
Joachimiak, A.; Schulze-Briese, C.; Tomizaki, T.; Mitschler, A.; Podjarny, A.
Proteins 2004, 55, 805.
35. Steuber, H.; Zentgraf, M.; Gerlach, C.; Sotriffer, C. A.; Heine, A.; Klebe, G. J. Mol.
Biol. 2006, 363, 174.
36. Van Zandt, M. C.; Jones, M. L.; Gunn, D. E.; Geraci, L. S.; Jones, J. H.; Sawicki, D.
R.; Sredy, J.; Jacot, J. L.; Dicioccio, A. T.; Petrova, T.; Mitschler, A.; Podjarny, A. D.
J. Med. Chem. 2005, 48, 3141.
41. Schrödinger Suite 2007; Schrödinger, LLC; New York, NY, USA, 2007.
42. (a) Gasteiger, E.; Gattiker, A.; Hoogland, C.; Ivanyi, I.; Appel, R. D.; Bairoch, A.
Nucleic Acids Res. 2003, 31, 3784–3788; (b) http://expasy.org/.
43. (a) Wolber, G.; Dornhofer, A.; Langer, T. J. Comput. Aided. Mol. Des. 2006, 20,
773; (b) Wolber, G.; Langer, T. J. Chem. Inf. Comput. Sci. 2005, 45, 160; (c)
LigandScout 2.0; Inte:Ligand GmbH, Vienna, Austria, 2007.
44. Costantino, L.; Rastelli, G.; Vescovini, K.; Cignarella, G.; Vianello, P.; Del Corso,
A.; Cappiello, M.; Mura, U.; Barlocco, D. J. Med. Chem. 1996, 39, 4396.
45. Chou, T. -C.; Hayball, M. P. CalcuSyn software version 1.1.1., Biosoft, Cambridge,
UK, 1996.
37. Steuber, H.; Zentgraf, M.; La Motta, C.; Sartini, S.; Heine, A.; Klebe, G. J. Mol. Biol.
2007, 369, 186.
46. PyMOL Molecular Graphics System v0.99, DeLano Scientific, Palo Alto, CA, USA,
2006.