Binding of 1-benzopyran-4-one derivatives to aldose reductase: A free energy perturbation study

Article abstract of DOI:10.1016/S0968-0896(01)00408-4

The relative binding affinities to human aldose reductase (ALR2) of three new 7-hydroxy-2-benzyl-4H-1-benzopyran-4-one inhibitors were predicted by free energy perturbation (FEP) simulations. Molecular substitutions were specifically designed to investigate the role of hydrogen bonding at the active site of ALR2. Starting from the lead inhibitor 7-hydroxy-2-(4′-hydroxybenzyl)-4H-1-benzopyran-4-one, the 4′-hydroxyl was mutated to methyl and to trifluoromethyl, and an hydroxyl at position 8 was additionally introduced. Once synthesized and tested as inhibitors of ALR2, the compounds displayed variations of K_i that were in qualitative to quantitative agreement with the calculated relative free energies of binding. The results, discussed in terms of balance between free energies of solvation and free energies of binding to ALR2, elucidate the importance of hydrogen bonding with Thr113 and with Trp111 and cofactor, and provide a rationale to the observed differences in binding affinities.

Full text of DOI:10.1016/S0968-0896(01)00408-4

Bioorganic & Medicinal Chemistry 10 (2002) 1427–1436
Binding of 1-Benzopyran-4-one Derivatives to Aldose Reductase:
A Free Energy Perturbation Study
Giulio Rastelli,a,* Luca Costantino,a M. Cristina Gamberini,a Antonella Del Corso,b
Umberto Mura,b J. Mark Petrash,c Anna Maria Ferraria and Sara Pacchionia
aDipartimento di Scienze Farmaceutiche, Universita di Modena e Reggio Emilia, Via Campi, 183, 41100 Modena, Italy
b
`
Dipartimento di Fisiologia e Biochimica, Universita di Pisa, Via S. Maria, 55, 56100 Pisa, Italy
cDepartment of Ophthalmology and Visual Sciences, Washington University, School of Medicine, St. Louis, MI 63110, USA
Received 25 September 2001; accepted 16 November 2001
Abstract—The relative binding affinities to human aldose reductase (ALR2) of three new 7-hydroxy-2-benzyl-4H-1-benzopyran-4-
one inhibitors were predicted by free energy perturbation (FEP) simulations. Molecular substitutions were specifically designed to
investigate the role of hydrogen bonding at the active site of ALR2. Starting from the lead inhibitor 7-hydroxy-2-(40-hydroxy-
benzyl)-4H-1-benzopyran-4-one, the 40-hydroxyl was mutated to methyl and to trifluoromethyl, and an hydroxyl at position 8 was
additionally introduced. Once synthesized and tested as inhibitors of ALR2, the compounds displayed variations of Ki that were in
qualitative to quantitative agreement with the calculated relative free energies of binding. The results, discussed in terms of balance
between free energies of solvation and free energies of binding to ALR2, elucidate the importance of hydrogen bonding with
Thr113 and with Trp111 and cofactor, and provide a rationale to the observed differences in binding affinities. # 2002 Elsevier
Science Ltd. All rights reserved.
Introduction
Therefore, new ALR2 inhibitors not structurally related
to carboxylic acids and spyrohydantoins, and possibly
having pKa values higher than carboxylic acids, are
highly desirable.
Aldose reductase (ALR2; EC1.1.1.21) is an important
enzyme in the development of degenerative complica-
tions of diabetes mellitus, through its ability to reduce
excess d-glucose into d-sorbitol with the associated
conversion of NADPH to NADP+.1 Available experi-
mental data link D-glucose metabolism to long-term
diabetic complications such as cataract, neuropathy,
nephropathy, and retinopathy. Aldose reductase inhibi-
tors should therefore be able to safely prevent or arrest
the development of diabetic complications.1
In a recent publication, we reported that 1-benzopyran-
4-one derivatives are potent and selective inhibitors of
ALR2.5 The combination of ALR2 activity and selec-
tivity with respect to the closely related enzyme alde-
hyde reductase (ALR1) appears promising for the
development of drugs effective against diabetic compli-
cations.5 As an additional advantage, 1-benzopyran-4-
one derivatives are considerably less acidic than car-
boxylic acids: the 7-hydroxyl group of 1-benzopyran-4-
one, which turned out to be one of the most important
structural requirements for ALR2 inhibition,5 has a pKa
of 7.3.6
To date, two main classes of ALR2 inhibitors have been
reported, namely carboxylic acids and spirohydantoins.2,3
However, carboxylic acids possess in vivo activity lower
than spirohydantoins, probably because their acidic
nature results in poor pharmacokinetics. On the other
hand, spirohydantoins like Sorbinil, which are less
acidic than carboxylic acids and posses better pharma-
cokinetic properties, cause hypersensitivity reactions.2,4
In the crystal structures of the complexes between
ALR2 and Zopolrestat7 or Tolrestat8 as well as in pre-
vious molecular modeling studies,9À12 the carboxylic
acid derivatives bind ALR2 with the carboxylate inter-
acting with Tyr48 and His110, which are two key resi-
dues in binding and catalysis.13,14 Docking of 7-
hydroxy-benzopyran-4-one derivatives into the ALR2
*Corresponding author. Tel.: +39-059-205-5145; fax +39-059-205-
5131; e-mail: rastelli.giulio@unimo.it
0968-0896/02/$ - see front matter # 2002 Elsevier Science Ltd. All rights reserved.
PII: S0968-0896(01)00408-4

1428
G. Rastelli et al. / Bioorg. Med. Chem. 10 (2002) 1427–1436
Scheme 2. Free energy perturbation cycle.
Scheme 1. Substituents explored with free energy perturbation.
structure and molecular mechanics calculations of the
complexes predicted that the dissociated 7-hydroxyl
hydrogen bonds to Tyr48 and His110, thus resembling
the carboxylate function of previously known inhibi-
tors.5 In Figure. 1, a selection of aminoacidic residues
interacting with one of the most active inhibitors of
this class [7-hydroxy-2-(40-hydroxybenzyl)-4H-1-benzo-
pyran-4-one]5 is presented. In addition to the hydrogen
bonds established between the dissociated 7-hydroxyl
and Tyr48 and His110, the 40-hydroxyl hydrogen
bonds to Thr113. The 2-(40-hydroxylbenzyl) sub-
stituent, in fact, is able to bind the so-called ‘specifi-
city pocket’, in line with previous crystallographic7,8
and modeling9À12 studies of ALR2–inhibitor com-
plexes. These studies showed that conformational
changes upon binding of suitable hydrophobic-aro-
matic inhibitors lead to the opening of this additional
pocket, which is lined by Trp111 and Leu300, and
that inhibitors that bind this pocket are selective for
ALR2 with respect to ALR1, as indeed is the case of the
7-hydroxy-2-benzyl-4H-1-benzopyran-4-one derivatives
previously reported.5
closely match the inhibitory activity trend observed
experimentally.19
Here, two substitution sites were chosen for modifica-
tion (Scheme 1). In the structure of the complex, the
40OH acts as a hydrogen bond donor to the hydroxyl of
Thr113 (Fig. 1). To investigate the importance of the 40-
hydroxyl as a hydrogen bond donor to Thr113, the
hydroxyl was mutated to methyl (hydrophobic) and to
trifluoromethyl (hydrophobic and weak hydrogen bond
acceptor) (Scheme 1). Secondly, an additional hydroxyl
group at position 8 was introduced because it could
provide two additional hydrogen bonds, one with the
amino hydrogen of Trp111 and one with the carbonyl
oxygen of the nicotinamide of NADP+ (Fig. 1).
Removal of the 40-hydroxyl by mutation to hydro-
phobic substituents reduces desolvation energy and
should facilitate binding. In contrast, introduction of a
hydroxyl group at the 8 position provides stronger
interactions with ALR2 which should compensate for
the higher desolvation cost.
To compare the predicted and the experimental free
energies of binding, the derivatives reported in Scheme 1
were synthesized and tested as inhibitors of human ALR2.
In the present work, we predict the effects of structural
modifications of the inhibitor 7-hydroxy-2-(40-hydroxy-
benzyl)-4H-1-benzopyran-4-one using free energy per-
turbation15 (FEP) calculations. In many cases,
predictions based on this methodology led to free
energy differences of binding that were in qualitative to
quantitative agreement with experiment.16À18 More-
over, when applied systematically to a series of protein–
inhibitor complexes, FEP predictions were found to
Results
The FEP methodology was applied to calculate both
solvation free energy differences in water and binding
free energy differences to ALR2 between the analogues
reported in Scheme 1. We used the thermodynamic
perturbation method20À22 to calculate the free energy
differences.
The thermodynamic cycle relevant to this study is
shown in Scheme 2. The free energies of binding to the
protein of the two related inhibitors IA and IB, ÁGA and
ÁGB, are difficult to calculate and are determined
experimentally. ÁGsolv corresponds to mutating IA into
IB in water, while ÁGALR2 corresponds to mutating IA
into IB while bound to ALR2 in water. Since the ther-
modynamic cycle is closed and free energy is a state
function, the relative binding free energy is
ÁÁGbind=ÁGBÀÁGA=ÁGALR2ÀÁGsolv. While ÁGsolv
and ÁGALR2 represent physically unrealizable processes,
these two simulations are computationally manageable.
Figure 1. Residues of the active site of ALR2 that interact with the
reference inhibitor 7-hydroxy-2-(40-hydroxybenzyl)-4H-1-benzopyran-
4-one.
In the present work, the reference inhibitor (IA) was 7-
hydroxy-2-(40-hydroxybenzyl)-4H-1-benzopyran-4-one

G. Rastelli et al. / Bioorg. Med. Chem. 10 (2002) 1427–1436
1429
Table 1. Calculated relative solvation and binding free energies (kcal/mol) between 7-hydroxy-2-(40-hydroxybenzyl)-4H-1-benzopyran-4-one
(40OH) and its analogues
Mutation
Run
Length (ps)
160
ÁGsolv
<ÁG>solv
5.64Æ0.07
ÁGALR2
<ÁG>ALR2
5.49Æ0.15
40OH!40CH3
Forward
Reverse
Forward
Reverse
Forward
Reverse
5.57Æ0.06
À5.71Æ0.02
5.73Æ0.04
5.34Æ0.07
À5.63Æ0.14
5.99Æ0.10
400
600
5.79Æ0.06
5.72Æ0.02
6.21Æ0.21
6.29Æ0.14
À5.84Æ0.03
5.70Æ0.03
À6.42Æ0.07
6.43Æ0.20
À5.74Æ0.02
À6.15Æ0.18
40OH!40CF3
Forward
Reverse
Forward
Reverse
Forward
Reverse
160
400
600
10.40Æ0.51
À9.80Æ0.11
9.99Æ0.11
10.10Æ0.30
10.09Æ0.10
10.09Æ0.22
10.16Æ0.33
À10.09Æ0.20
11.95Æ0.04
10.13Æ0.04
11.64Æ0.31
11.58Æ0.39
À10.19Æ0.16
9.87Æ0.25
À11.33Æ0.15
11.96Æ0.02
À10.31Æ0.29
À11.19Æ0.27
8H!8OH
Forward
Reverse
Forward
Reverse
Forward
Reverse
160
400
600
À1.83Æ0.01
1.97Æ0.08
À2.46Æ0.08
2.83Æ0.22
À2.16Æ0.07
2.69Æ0.29
À1.90Æ0.07
À2.64Æ0.18
À2.43Æ0.27
À0.60Æ0.01
0.57Æ0.04
À0.59Æ0.02
À0.41Æ0.11
À0.65Æ0.12
À0.29Æ0.02
0.52Æ0.01
À0.53Æ0.01
À0.76Æ0.04
(40OH), and the starting IA–ALR2 structure was taken
from our previous work.5 The free energy changes upon
mutating the inhibitors are given in Table 1. Both for-
ward and reverse simulations were performed, and the
resulting ÁG values were averaged. Simulations of dif-
ferent lengths, ranging from 160 to 600 ps, were done to
check convergence of the free energy results. To com-
pare with experimental results, the new derivatives were
synthesized, and their experimental inhibition constants
(Ki) of human ALR2 were determined. The Ki values,
the values of ÁGA and ÁGB inferred from the Ki’s, and
Table 2. Inhibition constants (Ki) against human aldose reductase,
and comparison of experimental (ÁÁGexpt) and calculated (ÁÁGbind)
free energy differences of binding relative to 40OH
hALR2
a
b
c
Ki
(mM)
ÁGexpt
ÁÁGexpt
ÁÁGbind
40OH
40CH3
40CF3
8OH
0.18Æ0.08
1.17Æ0.08
2.95Æ0.01
À9.26Æ0.28
À8.14Æ0.08
À7.59Æ0.01
À6.84Æ0.08
1.12Æ0.29
1.67Æ0.28
2.42Æ0.29
0.57Æ0.14
1.49Æ0.45
1.78Æ0.30
10.4Æ0.68
aExperimental free energies of binding calculated from the equation
ÁG=ÀRTln 1/Ki, and expressed in kcal/mol.
bExperimental free energy differences of binding relative to the refer-
ence compound 40OH.
the experimental (ÁÁGexpt) and calculated (ÁÁGbind
free energy differences of binding are reported in Table 2.
)
cCalculated free energies of binding relative to 40OH, expressed as the
difference between <ÁG>ALR2 and <ÁG>solv of the longer (600 ps)
simulations of Table 1.
Solvation free energies
The solvation free energies of 40CH3 relative to 40OH
are 5.64Æ0.07, 5.79Æ0.06 and 5.72Æ0.02 kcal/mol for
the 160, 400 and 600 ps simulations, respectively (Table
1). The free energies of relative solvation calculated in
the forward (40OH!40CH3) and reverse (40CH3!40OH)
directions are fairly similar, giving very low hysteresys.
The positive sign of ÁGsolv indicates that the 40CH3
derivative is less solvated than the reference 40OH
derivative, in agreement with a polar hydroxyl being
mutated to a methyl.
they can, the trifluoromethyl substituent would be
expected to be more solvated than the methyl, in con-
trast with the present results. Even though the AMBER
force-field parameterized by Cornell et al.25 does not
contain an explicit hydrogen bonding term, hydrogen
bonds in proteins, nucleic acids and organic molecules
in such a ‘minimalist’ force-field are consistently descri-
bed by the electrostatic and van der Waals terms.25À29 In
this respect, the ability of fluorine to act as a potential
hydrogen bond acceptor is, indeed, guaranteed by the
fact that fluorine is negatively charged in our inhibitors
(Scheme 3) as well as in the CF4 and CHCF3 molecules
used to derive the fluorine parameters for AMBER.30
Therefore, interaction with a hydrogen bond donor
would lead, in principle, to a favourable electrostatic
interaction energy. With this in mind, the finding that
the trifluoromethyl substituent is significantly more
hydrophobic than the methyl substituent can likely be
attributed to effects other than electrostatics, such as
van der Waals effects. Indeed, only rarely is fluorine
seen to act as a hydrogen bond acceptor,23 and water–
octanol partition coefficients indicate that the tri-
fluoromethyl substituent is more hydrophobic than the
The predicted changes in solvation free energy of 40CF3
relative to 40OH are 10.10Æ0.30, 10.09Æ0.10 and
10.09Æ0.22 kcal/mol for the 160, 400 and 600ps simu-
lations, respectively (Table 1). Compared with the
40OH!40CH3 perturbation, the 40OH!40CF3 mutation
gives remarkably higher values of ÁGsolv. This differ-
ence is estimated to be about 4.4 kcal/mol. Therefore,
the trifluoromethyl substituent turns out to be sig-
nificantly more hydrophobic than the methyl sub-
stituent of our lead inhibitors. Actually, whether
fluorine and fluoroalkyl substituents can or cannot act
as hydrogen bond acceptors is a matter of debate.23,24 If

1430
G. Rastelli et al. / Bioorg. Med. Chem. 10 (2002) 1427–1436
Scheme 3. Partial charges of the 40OH, 40CH3, 40CF3 and 8OH inhibitors.
methyl (p constants being 0.88 for trifluoromethyl and
0.56 for methyl), in line with the free energy differences
of solvation found here.
which is less stable than that of the reference inhibitor
40OH. Indeed, the 40CH3 derivative costs 5.72 kcal/mol
less to desolvate but looses 6.29 kcal/mol in the complex
with ALR2 as compared to 40OH, resulting in a ÁÁGbind
of 0.57 kcal/mol (Table 2). Therefore, the 40CH3 deriva-
tive is predicted to bind less effectively to ALR2. The
orientation of 40CH3 into the ALR2 binding site is
depicted in Fig. 2 (a), where the structure has been over-
layed with that of 40OH for comparison. While both inhi-
bitors form hydrogen bonds with Tyr48 and His110,
only 40OH hydrogen bonds to Thr113. As a result of
unfavorable interactions between the 40-methyl and the
Thr113 hydroxyl, the inhibitor and the Thr113 residue are
both displaced compared to 40OH [Fig. 2(a)].
Finally, an additional hydroxyl group was introduced
at position 8 (8H!8OH). As expected, the 8OH deri-
vative is more solvated than the reference compound.
The changes in free energy of solvation are
À1.90Æ0.07, À2.64Æ0.18 and À2.43Æ0.27 kcal/mol
for the 160, 400 and 600ps simulations, respectively
(Table 1).
ALR2 binding free energies
The free energy differences obtained from the mutations
into the solvated binding site of aldose reductase
(ÁGALR2) are also given in Table 1.
The free energy differences obtained after mutating
40OH into 40CF3 in ALR2 are 10.13Æ0.04, 11.64Æ0.31
and 11.58Æ0.39 kcal/mol for the 160, 400 and 600 ps
simulations, respectively (Table 1). The 40OH!40CF3
mutation is largely disfavored both in water and in
ALR2. As for the 40CH3 mutation, the balance between
free energies of 40CF3 in the two different environments
suggests that the gain in desolvation energy is offset by
the formation of a less stable complex with ALR2. The
The predicted free energy changes upon mutating 40OH
into 40CH3 (40OH!40CH3) in ALR2 are 5.49Æ0.15,
6.21Æ0.21 and 6.29Æ0.14 kcal/mol for the 160, 400 and
600 ps simulations, respectively (Table 1). According to
the positive sign of these free energies, the 40CH3 deriva-
tive is predicted to form an ALR2–inhibitor complex

G. Rastelli et al. / Bioorg. Med. Chem. 10 (2002) 1427–1436
1431
value of ÁÁGbind obtained from the longer simulation
is 1.49 kcal/mol (Table 2). Again, the 40CF3 derivative
is predicted to be less active than the 40OH reference
compound. In the complex, 40CF3 binds similarly to
40CH3 [Fig. 2(b)]. Owing to the bulkier trifluoromethyl
substituent, however, the inhibitor and Thr113 are fur-
ther displaced as compared to 40CH3, such that the
distance between the substituent and the hydroxyl
oxygen of Thr113 increases of about 0.3 A. Compared
to 40-CH3, the remarkable destabilization observed in
the 40CF3 free energy simulations partly arises from
steric hindrance with Thr113, as indeed is the case of
40-CH3, and partly from even more non-com-
plementary interactions with the oxygen of the Thr113
hydroxyl. In fact, the fluorine atoms have a partial
negative charge and interaction with the hydroxyl oxy-
gen is largely disfavoured. In this respect, we note that
if the Thr113 hydroxyl is rotated of 180ꢀ, it turns to a
hydrogen bond donor instead of an acceptor. This
situation would partly alleviate unfavorable interactions
with the trifluoromethyl substituent, which could accept
a hydrogen bond from the hydroxyl. Even if the hydro-
gen bond acceptor capability of organic fluorine is
questionable23,24 and the free energy difference of sol-
vation of CF3 indicates that this substituent is hydro-
phobic, favorable electrostatic interactions between the
negatively charged fluorine atoms and the hydroxyl
hydrogen would, at least, help stabilizing the complex.
To explore this possibility, the molecular dynamics
trajectory of the longer simulation was examined, and
showed that the Thr113 hydroxyl does not rotate of
180ꢀ during molecular dynamics. Rather, the hydroxyl
firmly hydrogen bonds to the backbone carbonyl of
Thr113 in the simulations [Fig. 2(b)]. Therefore,
Thr113 can act solely as a hydrogen bond acceptor, and
a hydrogen bond with trifluoromethyl cannot be
formed.
Finally, the free energy changes upon mutating 8H into
8OH in ALR2 are À0.59Æ0.02, À0.41Æ0.11 and
À0.65Æ0.12 kcal/mol for the 160, 400 and 600 ps
simulations, respectively (Table 1). While free energy
differences are negative, that is the 8OH derivative is
predicted to bind ALR2 more efficiently than the refer-
ence molecule, the net gain in free energy of binding to
ALR2 is overcome by a higher desolvation cost. The
value of ÁÁGbind is 1.78 kcal/mol, leading to the pre-
diction of a significant loss of binding. The 8-hydroxyl
does, indeed, hydrogen bond to Trp111 and to the car-
bonyl oxygen of the nicotinamide, as initially predicted
[Fig. 2(c)]. However, hydrogen bonds with these two
residues are not as strong as with water, as free energies
predict. In addition, some steric constraints between the
8-hydroxyl and the protein must be present in this case,
since the 8-hydroxyl slightly displaces (of about 0.35 A)
the nearby Trp111 residue as compared to the reference
compound [Fig. 2(c)].
Synthesis and biological evaluation
7-Hydroxy-2-(40 -trifluoromethylbenzyl)-4H-1-benzo-
pyran-4-one 2 (40-CF3 in Scheme 1) was synthesized
starting from 2 - hydroxy - 4 - methoxy(methoxy-
carbonyl)acetophenone 2a by reaction with 4-tri-
fluoromethylphenyl acetylchloride in presence of
Mg(OEt)2 (Scheme 4) following a general procedure.31
The methyl 7-methoxy-2-(40-trifluoromethylbenzyl)-4H-
Figure 2. Selection of aminoacids that interact with 40CH3 (a, cyan),
40CF3 (b, magenta) and 8OH (c, green). The structure of 40OH (yel-
low) is superimposed in each case, for comparison. Only polar hydro-
gens are shown, and water molecules are omitted.

1432
G. Rastelli et al. / Bioorg. Med. Chem. 10 (2002) 1427–1436
1-benzopyran-4-one-3-carboxylate 2b thus obtained was
saponified using HBr 48%, then decarboxylated by
heating in quinoline. The subsequent treatment with
BBr3 allowed us to obtain the desired compound 2.
Conceptually, the relative binding of two related inhi-
bitors to ALR2 (ÁÁGbind) can be analyzed as the sum
of two components, the free energy difference of solva-
tion of the ligands in water (ÁGsolv), and the free energy
difference of binding to the protein (ÁGALR2). The dif-
7,8-Dihydroxy-2-(40-hydroxybenzyl)-4H-1-benzopyran-4-
one 3 (8-OH in Scheme 1) and 7-hydroxy-2-(40-methyl-
benzyl)-4H-1-benzopyran-4-one 4 (40-CH3 in Scheme 1)
were synthesized by the reaction between the appro-
priate 2-hydroxyacetophenone (3a, 4a) and methyl 4-
methoxyphenylacetate or methyl 4-methylphenylacetate
respectively in presence of NaH/pyridine (Scheme 5);
the intermediate 1,3-diketone was then cyclized to 3b or
4 by treatment with hydrochloric acid/acetic acid; 3b
was then converted to 3 by treatment with BBr3.
ference between these two free energies (ÁÁGbind)
determines the overall effectiveness of binding of the
two related inhibitors. On interpretative grounds, while
the ligand is fully solvated by water molecules before
binding to the protein, water molecules are displaced
(all or in part) and replaced with protein residues when
the ligand approaches the binding site of the host.
Therefore, a cost in free energy of solvation must be
payed by the ligand to desolvate, and a gain in free
energy of binding to the protein is obtained after for-
mation of the complex. Figure 3 illustrates schemati-
cally the predictions of the relative free energies of
binding that arise from the present study. In the fig. the
black arrows (which correspond to ÁÁGbind) are the
differences between the ÁGALR2 and the ÁGsolv white
arrows. A prediction of a less active inhibitor implies
that the black arrow is directed toward the positive free
energy axis, while, on the reverse, a more active inhi-
bitor would have a black arrow toward the negative
axis.
The inhibition constants (Ki) of the compounds were
determined as described in the experimental part. The
40CH3, 40CF3 and 8OH derivatives proved to be less active
than the reference 40OH derivative, in agreement with pre-
dictions from FEP. The Ki values are reported in Table 2.
Discussion
In the present work, we have predicted the relative bind-
ing of three derivatives of 7-hydroxy-2-(40-hydroxy-
benzyl)-4H-1-benzopyran-4-one to ALR2 by molecular
dynamics and free energy perturbation calculations.
Substitution of the polar 40OH with hydrophobic sub-
stituents like 40CH3 and 40CF3, which resulted in less
solvated compounds, did not led to compounds that
Scheme 4. Synthetic route for 7-hydroxy-2-(40-trifluoromethylbenzyl)-4H-1-benzopyran-4-one.
Scheme 5. Synthetic route for 7-hydroxy-2-(40-methylbenzyl)-4H-1-benzopyran-4-one and 7,8-dihydroxy-2-(40-hydroxybenzyl)-4H-1-benzopyran-4-
one.

G. Rastelli et al. / Bioorg. Med. Chem. 10 (2002) 1427–1436
1433
Experimental
Computational methods
The AMBER 532 program was used for all molecular
mechanics (MM), molecular dynamics (MD) and free
energy perturbation (FEP) calculations. The all-atom
force field by Cornell et al.25 was used. During MD and
FEP simulations, all bond lenghts were constrained
using the SHAKE33 algorithm, allowing a time step of
2.0 fs. Solute and solvent were coupled to a constant
temperature heat bath with a coupling constant of 0.2
ps to maintain a temperature of 300 K. A residue based
cutoff of 10 A was employed. Pairlists were updated
every 10 time steps. In FEP, the entire inhibitor was
treated as the perturbed group. No intraperturbed
group contributions to the free energy were calculated.
The starting coordinates for FEP calculations were
taken from our previous structure of the inhibitor 7-
hydroxy-2-(40-hydroxybenzyl)-4H-1-benzopyran-4-one
bound in the active site of aldose reductase.5 This
structure was originally obtained starting from the
coordinates of the human ALR2 holoenzyme34 with
inhibitors docked into the substrate binding site. The
ternary complexes were energy-minimized and equili-
brated using MD, as previously described.5,10 This
structure has the open additional hydrophobic pocket
lined by Trp111 and Leu300 (whose relevance is dis-
cussed in refs 5, 8 and 10) and is very similar to the
structure of aldose reductase that binds Zopolrestat,
whose coordinates are still not completely available
(only the backbone atoms and the inhibitor are present
in the Protein Data Bank, entry code 1MAR7). In the
structure of the protein, all Lys and Arg residues were
positively charged, while Glu and Asp residues were
negatively charged; the dH, eH or protonated forms of
histidines were assigned on the basis of favorable inter-
actions with their environment.
Figure 3. Schematic diagram showing the predictions of relative sol-
vation and binding free energies that arise from the study. An ‘ideal’
prediction is presented.
bind better to the enzyme (first case in Figure 3). While
40CH3 costs 5.72 kcal/mol less to desolvate as compared
to 40OH, it is predicted to bind ALR2 6.29 kcal/mol less
favorably than 40OH. Overall, since the loss of binding
to ALR2 offsets the gain in desolvation, the 40CH3 is
predicted to be less active than the 40OH. The calculated
value of ÁÁGbind is 0.57 kcal/mol, which compares
qualitatively with the experimental value of 1.12 infer-
red from the Ki’s (Table 2). Similar reasonings hold for
the 40CF3 derivative, which is 10.09 kcal/mol less sol-
vated than the 40OH, and 11.58 kcal/mol less apt to
bind ALR2 than 40OH. The predicted ÁÁGbind of 1.49
kcal/mol for this compound is close to the experimental
value of 1.67 kcal/mol (Table 2).
Counterions were placed around the charged residues at
the surface of the enzyme-inhibitor complex in order to
have a total charge of À1. The À1 charge was chosen so
as to have the same net charge in the enzyme and in the
water simulations (i.e., the negative charge of the dis-
sociated 7-hydroxyl of the inhibitor). The oxidized form
of NADPH (NADP+) was used in all simulations
because inhibitors preferentially bind this form.35 The
parameters for NADP+ were taken from our previous
simulations,5 and those of fluorine were taken from the
work of Gough et al.30 As regards the inhibitors, their
structures were completely optimized using AM1, and
charges were calculated with electrostatic potential fits
to a 6-31G* ab-initio wave function, followed by a RESP
fit.36,37 The partial atomic charges of the inhibitors are
reported in Scheme 3.
The results obtained for the 8OH derivative fall into the
second case in Figure 3, where the perturbed ligand is
both more solvated and more favourably bound to the
enzyme. In this case, a desolvation penalty of À2.43
kcal/mol offsets a gain in binding of À0.65 kcal/mol, so
that the 8OH derivative is predicted to be less active
than 40OH. The predicted ÁÁGbind of 1.78 kcal/mol is
in the ballpark of the experimental value of 2.42 kcal/
mol as inferred from the Ki’s (Table 2).
Finally, the third case in Fig. 3 can be regarded as the
‘ideal’ condition for the prediction of a more active
inhibitor, that is, the inhibitor is less solvated in water
and better bound to the protein. In this case, the two
factors sinergically enhance inhibition. This condition
can be achieved, for example, by introducing suitable
hydrophobic groups into the ligand, which decrease the
desolvation penalty and enhance favorable interactions
with hydrophobic residues. Unfortunately, none of the
derivatives here presented fall into this case. While the
substitution of the polar 40OH with hydrophobic sub-
stituents resulted in less solvated compounds, they did
not prove to bind better to the enzyme.
Three independent perturbations were carried out start-
ing from the reference inhibitor. In order to set up the
perturbations, dummy atoms were added for the grow-
ing atoms (i.e., 40OH!40CH3, 40OH!40CF3,
8H!8OH). These dummy atoms had the same bonding
parameters of the growing group, but zero charges and
zero van der Waals parameters.

1434
G. Rastelli et al. / Bioorg. Med. Chem. 10 (2002) 1427–1436
The active site of ALR2 containing the inhibitor and
NADP+ was solvated with a spherical cap of 1549
TIP3P38 water molecules within 28 A of the center of
mass of the inhibitor, and harmonic radial forces (1.5
kcal/mol A2) were applied to these waters to avoid
evaporation.
standard. J values are given in Hz. Microanalyses were
carried out in the Microanalysis laboratory of the
Dipartimento di Scienze Farmaceutiche, Modena Uni-
versity. Analyses indicated by the symbol of the ele-
ments were within Æ0.4% of the theoretical values.
TLC were performed on precoated silica gel F254 plates
(Merck). Silica gel (Merck, 70–230 Mesh) was used for
column chromatography.
The energy minimization was performed by optimizing
the water molecules first, keeping the rest of the system
frozen. In addition, 20 ps MD at 300 K were performed
on the water molecules only, in order to let the solvent
equilibrate around the protein and the inhibitor. Then,
energy minimization was continued by letting the inhi-
bitor, NADP+, all the protein residues within 12 A
from the inhibitor and all the water molecules free to
move. 10,000 steps of conjugate grandient minimization
were performed. Then, the minimized structure was
equilibrated with MD at 300 K for 250 ps. During the
first 10 ps of heating to 300 K, a 5 kcal/mol restraint
was applied on the inhibitor and the protein residues,
and then it was turned off.
The synthesis of 7-hydroxy-2-(40-hydroxybenzyl)-4H-1-
benzopyran-4-one 1 (40OH in Scheme 1) is described in
ref 5.
Methyl 7-methoxy-2-(40 -trifluoromethylbenzyl)-4H-1-
benzopyran-4-one 3-carboxylate (2b). To a suspension
of Mg(OEt)2 (0.41 g, 3.58 mmol) in anhyd EtOH (10
mL)
a solution of 2-hydroxy-4-methoxy(methoxy-
carbonyl)acetophenone 2a31 (0.81 g, 3.60 mmol) was
added. A white precipitate formed. EtOH was then
removed under reduced pressure, then anhyd benzene
was added (30 mL) followed by 4-trifluoromethyl-
phenylacetyl chloride (0.80 g, 3.60 mmol) dissolved in
benzene (10 mL). The mixture was heated at reflux for 3
h, then it was cooled and poured in 10% acetic acid in
water (100 mL) at 0 ꢀC. The suspension was extracted
with CHCl3 (3 ꢁ 20 mL), the organic layer was dried
(Na2SO4), the solvent removed under reduced pressure
and the residue purified by column chromatography
(cyclohexane/EtOAc 60:40). Yield 0.67 g (48%), mp 88–
The perturbations were carried out in a series of 40
windows with Ál=0.025. Simulations of increasing
lenght were performed to test for convergence of the
free energy results: 160ps (1000 steps of equilibration
and 1000 steps of data collection at each window
frame), 400 ps (2500 steps of equilibration and 2500
steps of data collection) and 600 ps (2500 steps of equi-
libration and 5000 steps of data collection). Simu-
lations were run both in the forward (l=1!0) and in
the reverse (l=0!1) direction, with 100 ps equilibra-
tion performed at l=0 before running the reverse
simulation.
89 ꢀC. H NMR d8.11 (1H, d, J=8.92), 7.58 (4H, m),
1
7.12 (1H, dd, J=8.92, J=2.38), 6.77 (1H, d, J=2.35),
4.14 (2H, s), 3.97 (3H, s), 3.90 (3H, s).
7-Methoxy-2-(40 -trifluoromethylbenzyl)-4H-1-benzo-
pyran-4-one-3-carboxylic acid (2c). A suspension of 2b
(0.30 g, 0.77 mmol) in HBr 48% (15 mL) was heated at
110–120 ꢀC for 14 h. After cooling, the precipitate was
filtered ꢀand washed with water. Yield 0.20 g (69%), mp
To obtain the structures of the 40CH3, 40CF3 and 8OH
complexes, the end-point (l=0) of the longer (600 ps)
ALR2-inhibitor perturbations were subjected to addi-
tional 100 ps MD, using the topology of the perturbed
inhibitors. During these additional 100 ps, structures
were collected every 0.2 ps and averaged every 10 ps,
using CARNAL. Then, the last 10 ps-averaged structure
was energy minimized with 10,000 steps of conjugate
gradient minimization.
1
179–80 C. H NMR d 14.00 (1H, s), 8.02 (1H, m), 7.68
(4H, m), 7.13 (2H, m), 4.46 (2H, s), 3.91 (3H, s).
7-Methoxy-2-(40 -trifluoromethylbenzyl)-4H-1-benzo-
pyran-4-one (2d). A solution of 2c (0.10 g, 0.26 mmol)
in quinoline (1.5 mL) was heated at 180 ꢀC for 10 min,
then cooled and poured in water (20 mL) containing
concd. HCl (3 mL). The solid thus obtained was filtered
and washed with water. Yield 0.070 g (80%), mp 182–
For mutations in water, the inhibitor was placed at the
center of a rectangular box of 1250 TIP3P water mole-
cules. The system was equilibrated at 300 K under peri-
185 ꢀC. H NMR d 7.91 (1H, d, J=9.07), 7.69 (4H, m),
1
odic boundary conditions, using
a
minimization/
7.04 (2H, m), 6.24 (1H, s), 4.15 (2H, s), 3.88 (3H, s).
equilibration simulation protocol similar to that above
described. Then, mutations were performed as already
described for aldose reductase.
7-Hydroxy-2-(40 -trifluoromethylbenzyl)-4H- 1-benzo-
pyran-4-one (2). To a solution of 2d (0.070 g, 0.2 mmol)
in anhyd CH2Cl2 (10 mL), under stirring at 0 ꢀC, a
solution of BBr3 (1M in CH2Cl2, 1.05 mL, 1.05 mmol)
in CH2Cl2 was added. The resulting solution was stirred
at rt for 24 h, then cooled again; water was added and
the precipitate was collected, washed with water and
purified by means of preparative TLC (CH2Cl2/CH3OH
97:3). Yield 0.005 g (7.5%), mp 162 ꢀC (dec.) (diethyl
Calculations were performed on an IBM-SP3 computer.
Chemistry
Melting points were determined on a Buchi 510 melting
1
point apparatus and are uncorrected. H NMR spectra
1
were recorded on a Bruker AC200 spectrometer (Centro
Interdipartimentale Grandi Strumenti, Modena Uni-
versity) in DMSO-d6 solution. Chemical shifts are
reported in ppm from tetramethylsilane as internal
ether/hexane). MS m/z 320 (M+); H NMR (CDCl3) d
8.07 (1H, d, J=8.73), 7.52 (4H, m), 6.98 (1H, dd,
J=8.76, J=2.13), 6.90 (1H, d, J=2.05), 6.15 (1H, s),
3.99 (2H, s). Anal. (C17H11F3O3) C, H.

G. Rastelli et al. / Bioorg. Med. Chem. 10 (2002) 1427–1436
1435
7,8-Dimethoxy-2-(40-methoxybenzyl)-4H-1-benzopyran-
4-one (3b). A solution of 2-hydroxy-3,4-dimethoxy-
acetophenone39 3a (1.00 g, 5.1 mmol) and methyl
4-methoxyphenylacetate (1.00 g, 5.6 mmol) in anhyd
pyridine (8 mL) was added dropwise to a well stirred sus-
pension of NaH (60% dispersion in mineral oil) (0.72 g,
18.0 mmol) in anhyd pyridine (8 mL). When the reac-
tion subsided, the mixture was heated at 90ꢀC for 15 min.
After cooling, the mixture was decomposed in 2 N HCl
and extracted with CH2Cl2 (3 ꢁ 25 mL). The combined
organic layers were washed with 1N HCl (2 ꢁ 30 mL)
and water (30 mL) and dried (Na2SO4); then the solvent
was removed under reduced pressure. Yield 1.08 g
preparations was assessed by SDS-PAGE according to
the method of Laemmli44 and gels were stained accord-
ing to the method of Wray et al.45
The assay for aldose reductase activity was performed
at 37 ꢀC using 4.7 mM d,l-glyceraldehyde as substrate,
in 0.25 M sodium phosphate buffer pH 6.8 containing
0.11 mM NADPH, 0.38 M ammonium sulfate and 0.5
mM EDTA. One Unit of enzyme activity is the amount
of enzyme that catalyzes the oxidation of 1 mmol of
NADPH/min.
The sensitivity of aldose reductase to different com-
pounds was tested in the above assay conditions in the
presence of the inhibitor dissolved at proper concentra-
tion in DMSO. The concentration of DMSO in the
assay mixture was kept constant at 1%. For the deter-
mination of Ki double reciprocal plots with d,l-glycer-
aldehyde as variable substrate at fixed concentration of
NADPH were obtained. For each inhibitor at least
three different concentrations were used.
(65%), mp 110–111 ꢀC. H NMR: 7.74 (1H, d, J=8.97),
1
7.34 (2H, m), 7.22 (1H, d, J=8.97), 6.94 (2H, m), 6.12
(1H, s), 3.97 (3H, s), 3.93 (2H, s), 3.77 (3H, s), 3.75 (3H, s).
7,8-Dihydroxy-2-(40-hydroxybenzyl)-4H-1-benzopyran-4-
one (3). To a stirred solution of 3b (0.50 g, 1.53 mmol)
in anhyd CH2Cl2 (30 mL) at 0 ꢀC under N2 atmosphere
was added a BBr3 solution in CH2Cl2 (1M, 9.20 mL,
9.20 mmol); the solution was left to stand at rt for 24 h,
then it was cooled again to 0 ꢀC, and water and ice were
added. The resulting precipitate was collected and
washed with water, then purified by column chromato-
graphy (CH2Cl2/CH3OH 94:6). Yield 0.20 g (45%), mp
263–265 ꢀC (CH3OH/acetone). 1H NMR d 9.45 (3H,
broad s), 7.34 (1H, d, J=8.67), 7.20 (2H, m), 6.90 (1H,
d, J=8.67), 6.75 (2H, m), 5.96 (1H, s), 3.87 (2H, s).
Anal. (C16H12O5) C, H.
Elemental analyses
Compd
Formula
C
Calcd/found
H
Calcd/found
2
3
4
C17H11F3O3
C16H12O5
C17H14O3
63.76/63.41
67.60/67.91
76.68/76.89
3.46/3.48
4.25/4.20
5.30/5.18
7-Hydroxy-2-(40-methylbenzyl)-4H-1-benzopyran-4-one
(4). A solution of 2-hydroxy-4-[2-(tetrahydropyranyl)-
oxy]acetophenone40 (1.0 g, 4.3 mmol) and methyl
4-methylphenyl acetate (0.77 g, 4.7 mmol) in anhyd
pyridine (8 mL) was added dropwise to a well stirred
suspension of NaH (60% dispersion in mineral oil) (0.52
g, 12.9 mmol) in anhyd pyridine (8 mL). When the
reaction subsided, the mixture was heated at 90 ꢀC for
15 min. After cooling, the mixture was decomposed in
2 N HCl and extracted with CH2Cl2 (3 ꢁ 25 mL). The
combined organic layers were washed with 1N HCl (2 ꢁ
30 mL) and water (30 mL) and dried (Na2SO4); then the
solvent was removed under reduced pressure. The resi-
due was then purified by means of column chromato-
graphy (cyclohexane/EtOAc 75:25). Yield 0.95 g (84%),
Acknowledgements
Financial support from Consiglio Nazionale delle
Ricerche (grant CNR 98.01911.CT03) is gratefully
acknowledged.
References and Notes
1. Yabe-Nishimura, C. Pharmacol. Rev. 1998, 50, 21.
2. Costantino, L.; Rastelli, G.; Cignarella, G.; Vianello, P.;
Barlocco, D. Exp. Opin. Ther. Pat 1997, 7, 843.
3. Costantino, L.; Rastelli, G.; Vianello, P.; Cignarella, G.;
Barlocco, D. Med. Res. Rev. 1999, 19, 3.
4. Eggler, J. F., Larson, E. R., Lipinski, C. A., Mylari, B. L.,
Urban, F. J. In A Perspective of Aldose Reductase Inhibitors.
Advances In Medical Chemistry. Jai: Greenwich, CT, 1993;
Vol. 2, p. 197.
mp 178–181 ꢀC (acetone/petroleum ether). H NMR d
1
10.69 (1H, s), 7.84 (1H, d, J=8.67), 7.21 (4H, m), 6.88
(1H, dd, J=8.69, J=2.24), 6.78 (1H, d, J=2.13), 6.01
(1H, s), 3.92 (2H, s), 2.29 (3H, s). Anal. (C17H14O3) C, H.
Enzyme inhibition assays
5. Costantino, L.; Rastelli, G.; Gamberini, M. C.; Vinson,
J. A.; Bose, P.; Iannone, A.; Staffieri, M.; Antolini, L.; Del
Corso, A.; Mura, U.; Albasini, A. J. Med. Chem. 1999, 42,
1881.
6. Costantino, L.; Rastelli, G.; Albasini, A. Eur. J. Med. Chem.
1996, 31, 693.
Recombinant human aldose reductase was expressed as
previously described.41 Human recombinant aldose
reductase were purified to electrophoretic homogeneity
through the same chromatographic steps as previously
described for the bovine enzyme.42 The pure enzymes
were stored at 4 ꢀC in 10 mM sodium phosphate buffer
pH 7.0 in the presence of 2 mM DTT.
7. Wilson, D. K.; Tarle, I.; Petrash, J. M,; Quiocho, F. A.
Proc. Natl. Acad. Sci. U.S.A. 1993, 90, 9847.
8. Urzhumtsev, A.; Tete-Favier, F.; Mitscler, A.; Barbanton,
J.; Barth, P.; Urzhumgtseva, L.; Biellmann, J. F.; Podjarny,
A. D.; Moras, D. Structure 1997, 5, 601.
Protein concentration was evaluated by the method of
Bradford.43 The electrophoretic homogeneity of enzyme

1436
G. Rastelli et al. / Bioorg. Med. Chem. 10 (2002) 1427–1436
9. Costantino, L.; Rastelli, G.; Vescovini, K.; Cignarella, G.;
Vianello, P.; Del Corso, A.; Cappiello, M.; Mura, U.; Bar-
locco, D. J. Med. Chem. 1996, 39, 4396.
10. Rastelli, G.; Vianello, P.; Barlocco, D.; Costantino, L.; Del
Corso, A.; Mura, U. Bioorg. Med. Chem. Lett. 1997, 7, 1897.
11. Rastelli, G.; Costantino, L. Bioorg. Med. Chem. Lett.
1998, 8, 641.
12. Rastelli, G.; Costantino, L.; Vianello, P.; Barlocco, D.
Tetrahedron 1998, 54, 9415.
13. Bohren, K. M.; Grimshaw, C. E.; Lai, C. J.; Harrison,
D. H.; Ringe, D.; Petsko, G. A.; Gabbay, K. H. Biochemistry
1994, 33, 2021.
28. Wang, W.; Donini, O.; Reyes, C. M.; Kollman, P. A.
Annu. Rev. Biophys. Biomol. Struct. 2001, 30, 211.
29. Cheatham, T. E., III.; Kollman, P. A. Annu. Rev. Phys.
Chem. 2000, 51, 435.
30. Gough, C. A.; DeBolt, S. E.; Kollman, P. A. J. Comput.
Chem. 1992, 13, 963.
31. Cushman, M.; Nagarathnam, D.; Burg, D. L.; Gealen,
R. L. J. Med. Chem. 1991, 34, 798.
32. Case, D. A.; Pearlman, D. A.; Caldwell, J. W.; Cheatman,
T. E., III.; Ross, W. S.; Simmerling, C. L.; Darden, T. A.;
Merz, K. M.; Stanton, R. V.; Cheng, A. L.; Vincent, J. J.;
Crowley, M.; Ferguson, D. M.; Radmer, R. J.; Seibel, G. L.;
Singh, U. C.; Weiner, P. K.; Kollman, P. A. AMBER 5; Uni-
versity of California: San Francisco, 1997.
14. Grimshaw, C. E.; Bohren, K. M.; Lai, C. J.; Gabbay,
K. H. Biochemistry 1995, 34, 14374.
15. Kollman, P. A. Chem. Rev. 1993, 93, 2395.
16. Fox, T.; Scanlan, T. S.; Kollman, P. A. J. Am. Chem. Soc.
1997, 119, 11571.
17. McCarrick, M. A.; Kollman, P. A. J. Comp-Aided Drug
Des 1999, 13, 109.
18. Reddy, M. R.; Erion, M. D. J. Am. Chem. Soc. 2001, 123,
6246.
19. Pearlman, D. A.; Charifson, P. S. J. Med. Chem. 2001, 44,
3417.
20. Bash, P. A.; Singh, U.; Langridge, R.; Kollman, P. A.
Science 1987, 236, 564.
21. McCammon, J. A. Science 1987, 238, 486.
22. Jorgensen, W. L.; Briggs, J. M. J. Am. Chem. Soc. 1989,
111, 4190.
23. Dunitz, J. D.; Taylor, R. Chem. Eur. J. 1997, 3, 89.
24. Howard, J. A. K.; Hoy, V. J.; O’Hagan, D.; Smith, G. T.
Tetrahedron 1996, 52, 12613.
25. Cornell, W. D.; Cieplak, P.; Bayly, C. I.; Gould, I. R.;
Merz, K. M., Jr.; Ferguson, D. M.; Spellmeyer, D. C.; Fox,
T.; Caldwell, J. W.; Kollman, P. A. J. Am. Chem. Soc. 1995,
117, 5179.
33. van Gunsteren, W. F.; Berendsen, H. J. C. Mol. Phys.
1977, 34, 1311.
34. Wilson, D. K.; Bohren, K. M.; Gabbay, K. H.; Quiocho,
F. A. Science 1992, 257, 81.
35. Ehrig, T.; Bohren, K. M.; Prendergast, F. G.; Gabbay,
K. H. Biochemistry 1994, 33, 7157.
36. Bayly, C. I.; Cieplak, P.; Cornell, W. D.; Kollman, P. A.
J. Phys. Chem. 1993, 97, 10269.
37. Cieplak, P.; Bayly, C. I.; Cornell, W. D.; Kollman, P. A.
J. Comput. Chem. 1995, 16, 1357.
38. Jorgensen, W. L.; Chandrasekhar, J.; Madura, J. D.;
Impey, R. W.; Klein, M. L. J. Chem. Phys 1983, 79, 926.
39. Baker, W. J. Chem. Soc 1941, 662.
40. Sogawa, S.; Nihro, Y.; Ueda, H.; Miki, T.; Matsumoto,
H.; Satoh, T. Biol. Pharm. Bull. 1994, 17, 251.
41. Petrash, J. M.; Harter, T. M.; Devine, C. S.; Olins, P. O.;
Bhatnagar, A.; Liu, S.; Srivastava, S. K. J. Biol. Chem. 1992,
267, 24833.
42. Del Corso, A.; Barsacchi, D.; Giannessi, M.; Tozzi, M. G.;
Camici, M.; Mura, U. Arch. Biochem. Biophys. 1990, 283, 512.
43. Bradford, M. M. Anal. Biochem. 1976, 72, 248.
44. Laemmli, U. K. Nature 1970, 227, 680.
26. Kollman, P. A. Accounts Chem. Res. 1996, 29, 461.
27. Wang, J.; Cieplak, P.; Kollman, P. A. J. Comput. Chem.
2000, 21, 1049.
45. Wray, W.; Boulikas, T.; Wray, W. P.; Hancock, R. Anal.
Biochem. 1981, 118, 197.