Glucose-based spiro-isoxazolines: A new family of potent glycogen phosphorylase inhibitors

Article abstract of DOI:10.1016/j.bmc.2009.08.060

A series of glucopyranosylidene-spiro-isoxazolines was prepared through regio- and stereoselective [3+2]-cycloaddition between the methylene acetylated exo-glucal and aromatic nitrile oxides. The deprotected cycloadducts were evaluated as inhibitors of mu

Full text of DOI:10.1016/j.bmc.2009.08.060

Bioorganic & Medicinal Chemistry 17 (2009) 7368–7380
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry
journal homepage: www.elsevier.com/locate/bmc
Glucose-based spiro-isoxazolines: A new family of potent glycogen
phosphorylase inhibitors
Mahmoud Benltifa a,b,c,d, Joseph M. Hayes e, Sébastien Vidal a,b,c,d, David Gueyrard a,b,c,d
,
e
e
e
Peter G. Goekjian a,b,c,d, Jean-Pierre Praly a,b,c,d, , Gregory Kizilis , Costas Tiraidis , Kyra-Melinda Alexacou ,
*
e
e
f
Evangelia D. Chrysina e, , Spyros E. Zographos , Demetres D. Leonidas , Georgios Archontis ,
*
Nikos G. Oikonomakos e,
a Université de Lyon, Institut de Chimie et Biochimie Moléculaires et Supramoléculaires associé au CNRS, UMR 5246, Laboratoire de Chimie Organique 2, Bâtiment Curien,
43 Boulevard du 11 Novembre 1918, F-69622 Villeurbanne, France
b Université Lyon 1, F-69622 Villeurbanne, France
c CNRS, UMR5246, Institut de Chimie et Biochimie Moléculaires et Supramoléculaires (ICBMS), Laboratoire de Chimie Organique 2, Bâtiment Curien,
43 Boulevard du 11 Novembre 1918, F-69622 Villeurbanne, France
d CPE-Lyon, F-69616 Villeurbanne, France
e Institute of Organic and Pharmaceutical Chemistry, National Hellenic Research Foundation, 48 Vassileos Constantinou Avenue, Athens 11635, Greece
f Department of Physics, University of Cyprus, PO Box 20537, CY1678 Nicosia, Cyprus
a r t i c l e i n f o
a b s t r a c t
Article history:
A series of glucopyranosylidene-spiro-isoxazolines was prepared through regio- and stereoselective
[3+2]-cycloaddition between the methylene acetylated exo-glucal and aromatic nitrile oxides. The depro-
tected cycloadducts were evaluated as inhibitors of muscle glycogen phosphorylase b. The carbohydrate-
Received 25 March 2009
Revised 6 August 2009
Accepted 11 August 2009
Available online 2 September 2009
based family of five inhibitors displays Ki values ranging from 0.63 to 92.5 lM. The X-ray structures of the
enzyme–ligand complexes show that the inhibitors bind preferentially at the catalytic site of the enzyme
retaining the less active T-state conformation. Docking calculations with GLIDE in extra-precision (XP)
mode yielded excellent agreement with experiment, as judged by comparison of the predicted binding
modes of the five ligands with the crystallographic conformations and the good correlation between
the docking scores and the experimental free binding energies. Use of docking constraints on the well-
defined positions of the glucopyranose moiety in the catalytic site and redocking of GLIDE-XP poses using
electrostatic potential fit-determined ligand partial charges in quantum polarized ligand docking (QPLD)
produced the best results in this regard.
Keywords:
Glycogen phosphorylase
1,3-Dipolar cycloaddition
Isoxazoline
X-ray crystallography
GLIDE docking calculations
Quantum polarized ligand docking (QPLD)
Ó 2009 Elsevier Ltd. All rights reserved.
1. Introduction
dominantly relaxed (R) state), and GPb is the non-phosphorylated
enzyme (low activity, low substrate affinity, predominantly tense
Type 2 diabetes is characterized by hyperglycemia resulting
from defects in insulin secretion, insulin action or both.1 The
chronic hyperglycemia of diabetes is associated with long-term
damage, dysfunction, and failure of various organs, especially the
eyes, kidneys, nerves, heart, and blood vessels.2 Glycogen phos-
phorylase (GP) catalyzes the breakdown of glycogen to glucose-
1-phosphate in the liver and in tissues fluctuating with energy
demands.3,4 Two interconvertible forms of GP exist: GPa refers to
the phosphorylated form (high activity, high substrate affinity, pre-
(T) state). Among the different therapeutic targets for the
treatment of type 2 diabetes, the design of inhibitors of GP is a
promising therapeutic strategy with potential pharmaceutical
applications in improving glycemic control in type 2 diabetes.5
Several inhibitors of GP have been reported in the literature6–9
and the glucose-based ones6,7 were shown to bind selectively at
the catalytic site of this enzyme (either GPa or GPb) in contrast
with other families of inhibitors interacting at other sites. Gluco-
pyranosylidene-spiro-heterocycles A–L (Fig. 1) display interesting
inhibitory activities in which glucopyranosylidene-spiro-hydan-
toin A appears as the most effective glucose analogue inhibitor.
Crystallographic studies10 of the GPb–A complex have shown that
A binds at the catalytic site mainly through a dense hydrogen bond
network. Most of these analogues exploit the formation of hydro-
gen bonds with the amino-acids of GP’s catalytic site and the
changes in inhibition are mostly related to a varying number of
* Corresponding authors. Tel.: +33 472 43 11 61; fax: +33 478 89 89 14 (J.-P. P.);
tel.: +30 210 7273851; fax: +30 210 7273831 (E.D.C.).
E-mail addresses: jean-pierre.praly@univ-lyon1.fr (J.-P. Praly), echrysina@eie.gr
(E.D. Chrysina).
Dr. Nikos G. Oikonomakos sadly passed away on the 31st of August 2008 during
the preparation of this manuscript. This paper is dedicated to his memory.
0968-0896/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmc.2009.08.060

M. Benltifa et al. / Bioorg. Med. Chem. 17 (2009) 7368–7380
7369
OH
OH
O
OH
O
O
O
HO
HO
HO
HO
HO
HO
H
N
H
N
OH
O
A Ki = 3.1
OH
O
B Ki = 5.1
OH
HN
O
S
NH
N
H
N
H
C Ki = 28.6
O
OH
OH
O
O
HO
HO
O
H
H
N
HO
HO
HO
HO
H
N
N
O
OH
O
O
OH
O
OH
O
X
N
N
H
N
H
R
D X = O Ki > 11 500
E X = S Ki > 10 000
F
Ki = 59.7
G R = Me Ki = 1 200
H R = OH Ki = 39
I R = NH2 Ki = 146
J R = NHAc Ki = 550
HO
HO
HO
HO
O
O
HN
NH
O
O
NH
O
N
H
O
OH
OH
K
Ki = 220 000
L Ki = 8 200
Figure 1. Representative glucopyranosylidene-spiro-heterocyclic inhibitors of rabbit muscle GPb and their Ki (
l
M).6,7
possible interactions. Nevertheless, none of these inhibitors occupy
the so-called b-pocket in close vicinity to the catalytic site, which is
lined by both polar and nonpolar groups and has no access to the
bulk solvent.11
We have previously reported the synthesis of a series of glu-
co-12 and galactopyranosylidene-spiro-oxathiazoles13 incorporat-
ing an aromatic moiety at the 3-position of the oxathiazole ring
(Fig. 2). Preliminary results showed that glucose-based 3-phenyl
analogues. All the inhibitors bind at the catalytic site of the enzyme
which retains its less active T-state conformation.
In earlier work, we demonstrated that GLIDE17,18 docking calcu-
lations could successfully reproduce the binding of glucose-based
inhibitors at the catalytic site of GP.19 With the aim to develop a
new computational protocol that can accurately predict experi-
mental binding properties, in the current work we have performed
molecular modeling studies of the GP–spiro-isoxazoline com-
plexes, using GLIDE and quantum polarized ligand docking20
(QPLD) from calculations in extra-precision (XP) mode. The impor-
tance of more accurate charges derived using quantum mechanics
(QM) or hybrid quantum mechanics/molecular mechanics (QM/
MM) methods has recently been highlighted.21,22 We compare
the docking results of QPLD-XP calculations with optimized elec-
trostatic potential (ESP) ligand atomic charges, to results obtained
by standard GLIDE-XP calculations with OPLS-AA(2001) force-
field23 ligand atomic charges. For the QPLD-XP runs, the top-
ranked poses from the standard GLIDE-XP calculations were
subjected to reassignment of ligand partial charges. This was done
by fitting to ESPs generated by single point energy calculations on
the ligand binding poses either in the free state, using QM (free
ligand; no receptor), or in the ‘field’ of the receptor using QM/
MM, accounting for receptor–ligand charge polarization effects.
The ligand poses with the new partial charges were then redocked
oxathiazole derivatives inhibited GPb with Ki values of 25.9
lM
(R = H) and 48.0
l
M (R = p-F).14 We therefore recently performed
the synthesis of a family of glucopyranosylidene-spiro-oxathiaz-
oles and also studied the inhibition of glycogen phosphorylase.15
Glucopyranosylidene-spiro-isoxazolines have attracted our atten-
tion as a novel class of inhibitors of GP.16 They were presumed to
preserve the main structural features of glucose-based GP inhibi-
tors such as the 4C1 ring conformation, with the oxygen and nitro-
gen atoms in the heterocyclic ring available to hydrogen bonding
with amino-acids. An important feature of the new compounds is
the presence of aromatic groups that occupy the b-pocket in the
enzyme.
Five new glucopyranosylidene-spiro-isoxazolines were synthe-
sized and tested for their inhibitory effect on GPb upon binding
by kinetic assays and X-ray crystallographic experiments. The Ki
values obtained varied from 0.63 to 92.5
l
M
for the new
glucose framework
hydrophobic
spiro-junction
group
OH
OH
O
O
HO
HO
HO
HO
S
3
OH
OH
O
O
N
2
N
R
1
3-phenyl glucopyranosylidene-spiro-oxathiazole
hydrogen bond
acceptors
Figure 2. Structure of 3-aryl glucopyranosylidene-spiro-oxathiazoles and expected interactions associated with the glucopyranosylidene-spiro-isoxazoline inhibitors.

7370
M. Benltifa et al. / Bioorg. Med. Chem. 17 (2009) 7368–7380
into the receptor using GLIDE-XP. We also investigated the effect
on accuracy due to introducing positional constraints on parts of
the ligand. The quality of generated docking conformations was ex-
pected to be strongly correlated with scoring.24–26 In this work, the
best results were obtained using QPLD-XP with electrostatic poten-
tial (ESP) fit-determined ligand partial charges calculated in the
‘field’ of the receptor with constraints on the well-defined gluco-
pyranose moiety atomic positions. An average RMSD (heavy
atoms) between superimposed crystallographic and top-ranked
ligand docking conformations of 0.691 Å was obtained using this
method, with an impressive correlation (R2 = 0.958) between the
docking GlideScores (GSs) and experimental binding free energies.
Overall, the results obtained were in excellent agreement with
those from kinetic and crystallographic studies both in terms of
experimental free binding energy and mode of binding for all five
ligands.
the cycloaddition reaction was demonstrated through a set of
NMR experiments (1H, 13C, COSY, HSQC, and HMBC), showing in
particular a signal resonating between 106 and 108 ppm in the
13C NMR spectrum corresponding to the anomeric carbon atom
for the desired regioisomer.16 The (R) configuration at the anomer-
ic center of 3a–e was established by NOE experiments which
showed consistently a contact between the methylene protons of
the isoxazoline ring and the H-2 proton of the sugar ring. This
observation demonstrates that the nitrile oxide is approaching
the acetylated exo-glycal from the bottom
a-face, as already ob-
served by other groups. The acetylated cycloadducts were then
deprotected under Zemplén conditions to afford the hydroxylated
derivatives 4a–e in nearly quantitative yields.
2.2. Enzyme kinetics
Inhibition constants (Ki) were measured against GPb for com-
pounds 4a–e (Fig. 3). Structure 4d with an aromatic ring bearing
a para-nitro group exhibited a drastically increased Ki value
2. Results and discussion
(92.5
ible molecules and in particular 4a (6.6
presence in 4e of a hydrophobic extended aromatic system such
as a 2-naphthyl moiety improved the Ki value to the sub- M range.
This value was among the lowest reported to date for the glucose-
based family of GPb inhibitors.6,7 The favorable influence of a prop-
erly oriented 2-naphthyl residue on binding to the enzyme has
been already observed for spiro-compounds,14,15 C-glucosyl het-
erocycles40,41 and N-glucosyl-N0-acyl ureas.6,7 In order to rational-
ize the inhibitory activities of 4a–e on an accurate structural
basis, crystallographic studies and docking calculations have been
carried out.
l
M) in contrast with electron rich, less bulky but more flex-
2.1. Synthesis of glucopyranosylidene-spiro-isoxazolines
lM). Interestingly, the
Although [3+2]-cycloaddition reactions27 have been applied to
carbohydrates for a long time,28 their synthetic potential for pre-
paring heterocycles by regio- and stereoselective formation of
covalent bonds with atom economy recently arise strong interest.
For example, sugar-based dipoles such as glycosyl azides,29–31 ni-
trile oxides32 or nitrones32 have permitted significant advances.
Unsaturated sugars and in particular exo-glycals33,34 proved to be
useful as dipolarophiles. Although the regio- and stereoselectivities
of these reactions have been well documented in the literature,
none of these studies reported the final deprotection or biological
evaluation of the cycloadducts.
We first designed an approach using benzyl protected exo-gly-
cals which provided the desired cycloadducts in good yields. How-
ever, the subsequent hydrogenolysis of benzyl ethers was not
compatible with the stability of the isoxazoline ring in which the
N–O bond was reduced by dihydrogen in the presence of a Pd cat-
alyst, thereby changing drastically the structure of the molecule.16
We therefore considered acetates as protecting groups for their
final removal without altering the spiro-bicyclic system created
during the cycloaddition reaction. The previously reported persily-
lated exo-glucal 135 was prepared using our recently reported
methodology36,37 and then converted into the acetylated exo-glu-
cal 238,39 in a one-pot procedure (Scheme 1). 1,3-Dipolar cycload-
ditions with a series of aromatic nitrile oxides, generated in situ
from the corresponding hydroximoyl chlorides and triethylamine,
afforded cycloadducts 3a–e in high yields. The regiochemistry of
l
2.3. X-ray crystallographic studies
The overall architecture of the native T-state GPb with the loca-
tion of the catalytic site is shown in Figure 4. Electron density dif-
ference maps (Fig. 5) clearly defined the position of each inhibitor
within the catalytic site, consistent with the kinetic results. The
mode of binding and the interactions that the glucopyranose moi-
ety makes with GPb, in all five inhibitors, is almost identical with
those previously described at 100 K from a 1.93 Å resolution crys-
tallographic experiment,19 except that the atomic positions of C1–
C6 and O5 of b-
4a–e structures (compared to those of
D
-glucopyranose are slightly changed in the GPb/
-glucose) and exhibit
a-D
shifts of ꢀ0.3–0.5 Å towards the b-pocket. All the inhibitors bound
at the catalytic site of the enzyme, by making direct and water-
mediated hydrogen bonds with the protein residues and, with
OTES
O
OAc
O
TESO
TESO
AcO
AcO
a, b
CH2
CH2
OTES
OAc
1
2
a Ar = p-MeO-Ph
b Ar = p-Me-Ph
c Ar = Ph
c
d Ar = p-O2N-Ph
e Ar = 2-Naphthyl
OH
OAc
O
O
HO
HO
AcO
AcO
d
OH
AcO
Ar
Ar
O
O
4a-e
3a-e
N
N
Scheme 1. Preparation of the glucopyranosylidene-spiro-isoxazoline inhibitors 4a–e. Reagents and conditions: (a) nBu4NF, THF, rt, 4 h; (b) C5H5N, Ac2O, rt, 16 h (87% overall);
(c) ArC(Cl)@NOH, Et3N, CH2Cl2, rt, 16 h (83–99%); (d) NaOMe, MeOH, rt, 3 h (78–99%). (TES = Et3Si).

M. Benltifa et al. / Bioorg. Med. Chem. 17 (2009) 7368–7380
7371
O6
OH
O5
O
C6
C4
O4
C5
HO
C11
C10
C14
C7
HO
C1
C2
O3
C9
OH
O2
C12
C3
C8
R
O
O7
N
N1
C13
O9
O
C15
C15
CH3
H
N
N2
CH3
O
R
Ki
O8
O
O8
4a
4b
4c
4d
6.6 0.3 μM
7.9 0.1 μM
19.6 0.2 µM
92.5 4.2 μM
O6
OH
O5
O
C6
C4
C12 C13
O4
C5
HO
C11
C10
C18
C7
C14
HO
C1
C2
O3
C9
C8
OH
O2
C3
C16
C15
O
O7
N
N1
C17
4e
Ki = 0.63 0.02 μM
Figure 3. Chemical structures of glucopyranosylidene-spiro-isoxazoline 4a–e compounds showing the numbering system used for X-ray crystallographic studies and the
kinetic parameters (Ki values) for inhibition of GPb.
the exception of 4b, by inducing minor movements of the side
chains of residues 282–287, of the 280s loop. The structural results
also suggest a distortion of the five-membered ring where C1 is
ꢀ15° from the 4-atoms (O7–N1–C8–C7) plane. Below, we describe
briefly the GPb/4a–e interactions (shown in Fig. 6) at the catalytic
site. We start with compound 4c, which carries the simplest group
(-H) at the 4-position of the phenyl-spiro-isoxazoline moiety.
2.3.1. Compound 4c
There are no direct hydrogen bonding interactions between the
atoms of the substituent introduced at C1 and the protein (Fig. 6c).
O7 is hydrogen bonded to Asp283 OD2, Tyr573 OH, and Lys574 NZ
through Wat214, and also to Gly135 N and Asp283 OD1 through
another water molecule (Wat60). Also nitrogen atom N1 forms
water-mediated interactions with Asn283 OD1 (through
Wat290), and also with Glu88 OD2, Gly134 N, Gly137 N (through
Wat80) (Table S10). The phenyl-spiro-isoxazoline analogue, on
binding to GPb, makes a total of 14 hydrogen bonds and 79 van
der Waals interactions, of which 11 are nonpolar/nonpolar con-
tacts to Gly135, Leu136, and His341. In the T-state the catalytic site
is partially blocked by the 280s loop, comprising residues 282–287
and in the GPb–4c complex the 280s loop retains its closed posi-
tion by inducing only minor shifts of the side chains of Asn282,
Asp283, and Asp284. Compound 4c (Ki = 19.6 0.2
lV) binds al-
most 87 times more tightly than
a-D
-glucose (Ki = 1.7 0.1 mV),11
possibly because of the additional interactions of the spiro-isoxaz-
oline group with the protein.
Figure 4. A schematic diagram of the GPb dimeric molecule, viewed down the
molecular dyad. One subunit is colored in green and the other in blue. The position
is shown for the catalytic site. The catalytic site, marked by (1R)-30-(2-naphthyl)-
spiro[1,5-anhydro-D
-glucitol-1,50-isoxazoline] (4e) (shown in red), is buried at the
2.3.2. Compound 4b
Introduction of a nonpolar methyl group in the 4-position of the
phenyl-spiro-isoxazoline in compound 4b increased the affinity for
center of the subunit and is accessible to the bulk solvent through a 15 Å long
channel. The inhibitor 4e upon binding to the enzyme, promotes the less active T-
state through stabilisation of the closed position of the 280s loop (shown in white).

7372
M. Benltifa et al. / Bioorg. Med. Chem. 17 (2009) 7368–7380
Figure 5. 2Fo ꢁ Fc electron density maps of 4a (a), 4b (b), 4c (c), 4d (d) and 4e (e) structures bound at the catalytic site of GPb (from measurements at 293 K or, for 4b and 4e
at 100 K). The maps are contoured at 1.0
level. Electron density maps were calculated by using REFMAC42 before incorporating ligand coordinates.
r
GPb by 2.5-fold (Ki = 7.9 0.1
l
V) compared to 4c. Similar to GPb–
10 mM Bes, pH 6.7, 30% (v/v) DMSO for 3.5 h (Table S2), prior to
flash-cooling in the nitrogen steam (100 K). DMSO, which in solu-
tion promotes the R-state structure, was tested previously as a
cryoprotectant in an experiment where native T-state GPb crystals
were allowed to equilibrate for ꢀ30 min with a buffered solution
(pH 6.7) containing 10 mM Bes and 30% (v/v) DMSO, and a com-
plete data set was collected at a maximum resolution of 2.3 Å.45
The Fo ꢁ Fc and 2Fo ꢁ Fc electron density maps indicated no binding
of DMSO at the catalytic or at other regulatory binding sites,
although in the 100 K DMSO GPb structure, the main-chain B-fac-
tor values for residues 283–285 were increased, compared to the
room temperature T-state GP structure, from 26.7 to 48.1 Å2.45 In
the present cryogenic experiment with 4b (1.80 Å resolution) and
4e (1.86 Å resolution), 16 and 15 molecules of DMSO were identi-
fied with average B-factor values of 40.4 and 43.0 Å2, respectively,
but none of these molecules bound at the catalytic site of the en-
zyme. On the contrary, cryogenic experiments with GPb using glyc-
erol as a cryoprotectant45–47 and GPa48 have shown that glycerol
binds at the catalytic site and competes with glucose analogues, thus
preventing the formation of complexes. In the present work, a paral-
lel investigation has been carried out with a single GPb crystal
soaked for 2 h in a buffered solution (pH 6.7) containing 10 mM
Bes and 30% (v/v) DMSO and data collected to 1.93 Å resolution.49
The structural results showed that the residue Asn284 was not sup-
4c binding, O7 and nitrogen N1 atoms make water-mediated inter-
actions with Tyr573, Lys574, and Gly135 (through Wat458), and
with Asp283, Glu88, Gly134, Ile136, and Gly137 (through
Wat895) (Fig. 6b). 4b, on binding to the enzyme, makes a total of
14 hydrogen bonds and 91 van der Waals interactions, 14 of which
are nonpolar/nonpolar contacts to Gly135, Leu136, His341, and
Asp283, which undergoes a conformational change. There is a shift
in the 280s loop; the major shifts (compared to the GPb-
a-D-glu-
cose complex) for C atoms are for residues Asn282 (1.1 Å),
a
Asp283 (1.4 Å), Asn284 (1.4 Å), Phe285 (2.1 Å), Phe286 (1.2 Å),
and Glu287 (0.7 Å), and these shifts lead to increased contacts be-
tween Asp283 and 4b phenyl ring atoms C10-C14 (Table S8). The
4-methyl group (C15) makes additional van der Waals contacts
with His341 CE1 and NE2 (distance = 3.5 Å) and four water mole-
cules (Wat282, Wat417, Wat896, Wat897).
It appears that upon binding of 4b at the catalytic site, the 280s
loop shifts to accommodate the 4-phenyl-spiro-isoxazoline moi-
ety. Similar shifts of the 280s were observed previously on binding
of N-benzoyl-N0-b- -glucopyranosyl urea43 and b-
D D-glucopyrano-
syl bismethoxyphosphoramidate44 to the catalytic site of GPb in
order to avoid steric clashes with Asn284.
The crystallographic binding experiment with 4b was per-
formed by soaking a single crystal of GPb with 30 mM 4b in

M. Benltifa et al. / Bioorg. Med. Chem. 17 (2009) 7368–7380
7373
Figure 6. Stereo diagrams showing interactions between 4a (a), 4b (b), 4c (c), 4d (d) and 4e (e) and protein in the vicinity of the catalytic site (from measurements at 293 K or,
for 4b and 4e at 100 K). The hydrogen bonding pattern between ligands, protein residues and water molecules (w) in the catalytic site is represented by dotted lines.
ported in the electron density map, and little density was present to
R2 = 0.958
-8.4
-8.6
-8.8
-9
-9.2
-9.4
-9.6
-9.8
-10
suggest a new position. Therefore, a clear conclusion cannot be
drawn from the experimental and computational results (vide infra)
whether the significant shifts in residues 282–287 observed in the
GPb–4b structure arise from steric clashes of protein residues with
4b or are correlated with the presence of 30% DMSO in the cryogenic
crystallographic experiment. Most evidence does point to the latter
but a combination of both factors is also possible.
2.3.3. Compound 4a
-9
-8
-7
-6
-5
To improve the biological activity of 4b we replaced the –CH3 in
the 4-position of the phenyl-spiro-isoxazoline with a methoxy sub-
stituent (–OCH3), compound 4a, based on the expectation that the
addition of a polar group would induce additional hydrogen bond-
ing interactions between the methoxy O8 and His341 NE2 and the
water molecules in the vicinity of b-pocket. The new analogue was
ΔG experimental (kcal/mol)
Figure 7. Correlation between GlideScores (GSs) and the experimental binding free
energies obtained using the QPLD-XP docking algorithm with glucopyranose
positional constraints and ESP fit ligand charges (in ‘field’ of receptor) for the 4a–
e ligands studied.
a slightly better inhibitor than 4b, with a Ki value of 6.6 0.3 lV, a
result not as good as we expected. The structure of compound 4a in
complex with GPb was determined at 1.85 Å resolution, which re-
vealed the detailed interactions of the ligand within the catalytic
site. The water-mediated hydrogen bonding interactions of O7
and N1 with Asp283, Tyr573, Lys574, Gly135, Glu88, Gly134, and
Gly137 are maintained (Tables S3 and S4, Fig. 6a). In addition to
these interactions, the methoxy O8 group forms a hydrogen bond
with Ala383 O through Wat289, and also participates in a complex

7374
M. Benltifa et al. / Bioorg. Med. Chem. 17 (2009) 7368–7380
hydrogen bonding network (through Wat222 and Wat102) (Table
S4). The crystal structure of GPb–4a shows that O8 is not hydrogen
bonded to His341 NE2, although O8 is at a hydrogen bonding dis-
tance from His341 NE2 (3.3 Å). The program CONTACT (CCP4) does
not list this interaction due to a slight deviation of the allowed an-
gle value for this bond (111°). As in the case of 4b, the –CH3 group
(C15) makes van der Waals contacts with Asn282 O, Phe285 O,
His341 CE1 and four water molecules (Wat194, Wat204, Wat274,
Wat289), and it seems that this is not a favorable environment
for a methyl group. The GPb–4a superimposes exactly with GPb–
4c with respect to the position of the phenyl group, except that
Wat276 (in GPb–4c structure) shifts by ꢀ1.4 Å (Wat274 in GPb–
4a structure) to avoid clashes with the –CH3 group. Although the
network of interactions is enhanced when the methoxy group is
present, the affinity is only slightly improved (as compared with
4b). Docking results replicate this finding and reveal a decreased
Lipo-EvdW descriptor contribution to the docking GlideScore (vide
infra: Docking Results and Fig. 8) within the b-cavity for a para-
substituted phenyl with –OMe compared to –Me in 4b.
makes a total of 16 hydrogen bonds and exploits 90 van der Waals
interactions (10 nonpolar/nonpolar, 25 polar/polar, and 55 nonpo-
lar/polar). Compound 4d appears to form an improved network of
interactions, however it exhibits poorer affinity than the lead mol-
ecule (4c); this may be due to the changes observed in the water
structure and again as in the case of 4a, that a polar para-phenyl
substituent is non-optimal with regard to non-bond interactions
as defined by the docking Lipo-EvdW contribution (vide infra:
Docking Results and Fig. 8). There were also indications in both
2Fo ꢁ Fc and Fo ꢁ Fc electron density difference maps for weak
binding of compound 4d at the inhibitor site, which is formed by
two hydrophobic aromatic residues, Phe285 and Tyr613, and is lo-
cated at the surface of the protein molecule; the site has been
shown to bind purines, nucleosides, nucleotides, and heterocyclic
compounds.6,7,45 The electron density at the inhibitor site was
not well resolved to indicate the binding mode of the ligand. In a
crystallographic experiment with 2 mM 4d (17 h soaking of GPb
crystals in 10 mM Bes, 2% DMSO buffer, pH 6.7) (data not shown),
electron density maps indicated strong binding of 4d at the cata-
lytic site, but very poor electron density at the inhibitor site. It
seems therefore, that the catalytic site is indeed the primary bind-
ing site for 4d.
2.3.4. Compound 4d
Introduction of a nitro group in the 4-position of the phenyl-
spiro-isoxazoline, compound 4d, showed that the affinity of this
analogue for GPb, compared to that of the lead compound (4c),
2.3.5. Compound 4e
was significantly reduced (Ki = 92.5 4.2
l
M), although one would
Compound 4e, the most active spiro-isoxazoline derivative, re-
vealed a 2700-fold improvement in enzymic activity, when com-
expect that the addition of a polar group would have the opposite
effect. Oxygen atoms O8 and O9 of the 4-nitro group make water-
mediated hydrogen bonding interactions with residues Glu88,
Arg292, Tyr280, Asn282, Asp339, His341, and Ala383 (Table S12
and S13, Fig. 6d). Structural comparison between GPb–4d and
GPb–4c complex structures showed that the atomic positions of
C9–C14 of the phenyl group in the GPb–4d structure are shifted
ꢀ0.3–0.6 Å towards Asn282 O and Glu88 OE1. The new position
of the 4-nitrophenyl group creates more space for the ligand to
be accommodated at the b-pocket, causing displacement of two
water molecules (Wat214 and Wat290 in GPb–4c structure) to
avoid close contacts with C14, and also shifts of another two water
molecules; Wat294 (in GPb–4c structure) shifted ꢀ0.7 Å (Wat272
in GPb–4d structure) and Wat276 (in GPb–4c structure) shifted
ꢀ1.2 Å (Wat260 in GPb–4d structure). The 4-nitro-spiro-isoxazo-
line analogue 4d on binding at the catalytic binding site of GPb
pared to
a-D-glucose (Ki = 1.7 0.1 mV) and is one of the most
potent known inhibitors of muscle GPb6,7 that binds at the catalytic
site. The structure of compound 4e was determined at 1.86 Å res-
olution, in the presence of 30% DMSO, under cryogenic conditions
(100 K). In contrast to 4b, the structural results with 4e show that
the ligand interacts directly with main-chain or/and sidechain of
Asn282, Asn284, Phe285, and indirectly with the carboxylates
OD1 and OD2 of Asp283, holding the 280s loop in its closed T-state
conformation and therefore competing with any possible effects of
the allosteric effector DMSO on the TR allosteric equilibrium. The
water-mediated hydrogen bonding interactions of O7 with
Asp283, Tyr573, Lys574, and Gly135, and those of N1 with
Glu88, Gly134, and Gly137 are also maintained in the GPb–4e
complex structure (Tables S15 and S16) (Fig. 6e). Overall, 4e makes
a total of 15 hydrogen bonds and 111 van der Waals interactions
(17 nonpolar/nonpolar, 15 polar/polar, 79 polar/nonpolar) (Table
S17), of which 40 are contacts to the 2-naphthyl group. Superposi-
tion of the 4c and 4e compounds bound at the catalytic site shows
that the 2-naphthyl group of 4e is inclined ꢀ11° with respect to the
phenyl group of 4c in order to minimize close contacts with
His341; also some rearrangement of the water structure in the
vicinity of the b-pocket involved Wat294 (in the GPb–4c structure)
which was displaced, and water molecules Wat276 (Wat409 in the
GPb–4e structure), Wat226 (Wat404 in the GPb–4e structure) and
Wat198 (Wat333 in the GPb–4e structure) which shifted ꢀ1.2, 0.5,
and 1.1 Å, respectively, to optimize contacts with the 2-naphthyl
group. The inhibition constant of 4e for rabbit muscle GPb is
-7
Lipo-EvdW
-5
H-bond
SiteMap
LowMW
RotPenal
-3
-1
1
4e
4a
4b
4c
4d
Ki = 0.63 0.02
l
V, ꢀ30 times decreased than the lead compound
4c, indicating a difference in the binding free energy between 4e
and 4c of about 2.1 kcal/mol. A major contribution to the difference
in the binding affinity between 4e and lead compound 4c may re-
sult from the extended interactions of the 2-naphthyl group with
protein atoms and the water hydrogen-bonding network in the
vicinity of the b-pocket. The output from docking calculations is
more revealing in this aspect.
-6.92 -5.77 -6.15 -5.72 -5.84
-2.26 -2.72 -2.19 -2.34 -2.17
-0.35 -0.36 -0.36 -0.36 -0.36
-0.7 -0.83 -0.94 -1 -0.73
0.57 0.63 0.69 0.75 0.58
Lipo-EvdW
H-bond
SiteMap
LowMW
RotPenal
Figure 8. Breakdown of the individual XP term contributions to the GlideScores
(GSs) for binding of each ligand obtained using QPLD-XP docking with glucopyr-
anose positional constraints and with ESP fit charges calculated in the ‘field’ of the
receptor. The best GS is the most negative, hence large negative contributions are
the most favorable. The highest to lowest scoring ligands are displayed in order
from left (4e) to right (4d).
2.4. Docking results
The GPb enzyme was prepared for docking using the coordi-
nates of GPb:4e co-crystallized complex with Schrödingers ‘Protein

M. Benltifa et al. / Bioorg. Med. Chem. 17 (2009) 7368–7380
7375
Table 1
Comparison of docking performance using default OPLS-AA(2001) forcefield ligand charges (GLIDE-XP) or electrostatic potential (ESP) fit ligand charges (QPLD-XP) and different
docking constraints with respect to reproduction of the crystallographic ligand binding conformationsa
Ligand
GLIDE-XP
QPLD-XP
QPLD-XP
OPLS-AA(2001) charges
ESP fit charges (gas phase)
ESP fit charges (field receptor)
Constraints
None
Core
(i) Core
(i) Core
(i) Core
(ii) O(H)
(ii) O(H)
(ii) O(H)
4a
4b
4c
4d
4e
0.655
1.210
0.435
0.506
1.502
0.655
0.476
0.430
0.499
1.498
0.654
0.477
0.436
0.498
1.498
0.656
0.477
0.554
0.671
1.503
0.555
0.473
0.434
0.499
1.498
Average
0.862
0.712
0.713
0.772
0.692
a
Measured as root-mean-square-distance (RMSD) in Å between superimposed heavy atoms from the top-ranked binding poses and the crystallographic ligand confor-
mations. Docking positional constraints were on the glucopyranose moiety: (i) core constraints (1 Å) on the glucopyranose ring atoms; (ii) positional constraints (radius 1 Å)
on the O(H) oxygen atoms based on their positions in the 4e complex.
Preparation Wizard’ as described in Section 4.5. In our previous
benchmark study,19 we found that the inclusion of cavity waters,
albeit crucial in flexible receptor–ligand molecular dynamics
(MD) calculations, was not important in rigid receptor docking cal-
culations. Based on these results, no crystallographic waters were
retained in the current study. Different docking algorithms were
probed for their effectiveness in reproducing crystallographic and
kinetic data. Results for the superimposition of the top-ranked li-
gand binding poses on the crystallographic conformations for these
docking algorithms (GLIDE-XP and QPLD-XP) using different gluco-
pyranose positional constraints and ligand partial charges are
shown in Table 1, while the correlation between the docking GSs
and experimental binding free energies are shown in Table 2.
For the GLIDE-XP docking runs using OPLS-AA(2001) partial
charges and no docking constraints, it appears that excellent corre-
lation (R2 = 0.878) for the docking GSs—experimental binding free
energies has been attained. However, superimposing the top-
ranked ligand docking poses for ligands 4a–e and their crystallo-
graphic binding conformations, we see that the results for two of
the ligands (4b and 4e) are not within the desired accuracy thresh-
old with RMSDs (heavy atoms) >1.0 Å. In the case of 4b the crystal-
lographic conformation was ranked as the second best pose with a
corresponding docking score (GS) = ꢁ9.04. This could be attributed
to the rigid-docking protocol, which ignores the protein conforma-
tional changes and shift in 280’s loop occurring upon binding of 4b.
The top-ranked pose had an almost identical score (GS = ꢁ9.05)
with the crystallographic conformation but was significantly dif-
ferent to the crystallographic conformations (RMSD (heavy
atoms) = 1.210 Å (Table 1)). This is in contrast to a clear-cut dis-
tinction between the GSs for correct and incorrect binding poses
compared to crystallography for all other ligands. On imposing glu-
copyranose positional constraints in the docking calculations, the
crystallographic ligand binding conformation was now ranked first
and excellent correlation for docking GS versus experimental bind-
ing free energy for all ligands (including 4b) obtained, indicating
that the receptor conformation from the 4e complex is also consis-
tent with favorable 4b binding. For 4e, although the RMSD is
1.506 Å, the larger RMSD is due to a flip in an otherwise close to
superimposable naphthyl moiety. Nevertheless, the incorrect top-
ranked binding pose for 4b (RMSD = 1.210 Å) shows that an excel-
lent correlation with relative binding affinities does not always
constitute a proof of successful binding prediction. Application of
docking positional constraints on the glucopyranose moiety en-
sured that the ligands adopted their correct and well-defined posi-
tions in the catalytic site and gave rise to better predicted ligand
binding conformations. Its application also ensured greater com-
putational efficiency in searching for correct binding poses and
not unrealistic glucopyranose positions/orientations. Two types
of positional constraints were used: core constraints on the gluco-
pyranose 6-membered ring positions with a tolerance of 1 Å from
their reference positions which was the 4e ligand in the prepared
4e complex; and positional constraints (denoted O(H) in Tables 1
and 2) on the glucopyranose oxygen atoms which were restricted
to spheres with radius 1 Å around their reference positions. Once
these constraints were applied, more accurate ligand binding con-
formations were obtained using both GLIDE-XP with OPLS-
AA(2001) charges and QPLD-XP where the GLIDE-XP poses were
redocked using ESP fit charges. There was not a significant differ-
ence in performance of the algorithms GLIDE-XP and QPLD-XP
using docking constraints with respect to reproduction of the li-
gand crystallographic conformations with average RMSDs ranging
between 0.692 and 0.772 Å. QPLD-XP using ESP fit charges in the
‘field’ of the receptor performed best (RMSD = 0.692 Å) in this re-
gard. Inspection of the top-ranking poses reveals the expected glu-
copyranose moiety hydrogen bonds are close to intact (elongated
in a few cases) for all ligands. This was also the case in our previous
GLIDE-XP docking calculations.19 With respect to the correlation
between top-ranked docking scores and experimental binding free
energies, there is a more significant difference in performance be-
tween GLIDE-XP and QPLD-XP using constraints. While GLIDE-XP
docking with OPLS-AA(2001) ligand charges produces good corre-
lations (R2 = 0.782; 0.722), the correlation obtained redocking the
GLIDE-XP ligand poses with ESP fit charges calculated in the gas
phase using QM (R2 = 0.848) or in the ‘field’ of the receptor using
QM/MM (R2 = 0.958) is excellent (Fig. 7).
Table 2
Comparison of docking performance using the different algorithms, ligand charges
and docking constraints with respect to correlation between the docking GlideScores
(GSs) and experimental binding free energies for ligands 4a–e
Docking
algorithm
Ligand charges
Constraintsa
Correlation
coefficient (R2)
GLIDE-XP
GLIDE-XP
GLIDE-XP
QPLD-XP
QPLD-XP
OPLS-AA(2001)
OPLS-AA(2001)
OPLS-AA(2001)
ESP fit (gas phase)
ESP fit (‘field’ of receptor)
—
Core
Core + O(H)
Core + O(H)
Core + O(H)
0.878
0.782
0.722
0.848
0.958
GLIDE uses a crude explicit water model for modeling solvation
effects.17,18 Our results show that this model is indeed very effec-
tive. However, caution here should be used, as the pseudo-explicit
water molecules are placed after the initial docking and therefore
a
Constraints on glucopyranose moiety—core constraints (1 Å) on the glucopyr-
anose ring atoms; positional constraints (radius 1 Å) on the O(H) oxygen atoms
based on their positions in the 4e complex.

7376
M. Benltifa et al. / Bioorg. Med. Chem. 17 (2009) 7368–7380
may not be the best tool for predicting conserved water molecules.
In terms of overall performance, the QPLD-XP method with con-
straints on the glucopyranose moiety was best in terms of correct
binding conformations and GS—experimental binding free energy
correlation. At least for this small ligand set tested, the ESP fit
charges calculated in the field of the receptor, yielded superior re-
sults to the free ligand gas phase ESP fit charges which also gave
very good results. Additional tests with larger ligands sets, how-
ever, are necessary to confirm the transferability of this assertion.
Nevertheless, standard GLIDE-XP with OPLS-AA(2001) charges and
docking constraints on the glucopyranose moiety appears a fast-
track method for screening larger databases of ligands, the top
poses which then can be redocked more accurately using QPLD-
XP. It should also be noted that as docking is limited by conforma-
tional sampling, it is important for greater accuracy to use a num-
ber of different ligand input conformations.
3. Conclusion
In conclusion, a new series of spiro-isoxazoline compounds
were designed, synthesized and tested for inhibition and crystallo-
graphic binding to GPb. All the inhibitors bound at the catalytic site
of the enzyme, promoting the less active T-state conformation
through an extended network of hydrogen bond interactions with
protein residues in the vicinity of the same site. They all induced
only minor shifts upon binding except 4b that causes more pro-
found changes to avoid steric clashes with Asn284. Compound 4e
was found to be the most potent inhibitor among the five com-
pounds with a Ki value of 0.63 M that may be attributed to both
the rigidity of the spiro and 2-naphthyl groups that undergo little
loss of conformational entropy on binding and to the extended net-
work of interactions formed. The rigid receptor docking calcula-
tions with glucopyranose positional constraints accurately
reproduced ligand crystallographic binding conformations and
produced excellent correlation between top-ranked ligand docking
score and experimental binding free energies. Although good re-
sults were obtained with OPLS-AA(2001) ligand partial charges,
there was a significant improvement using ESP fit ligand charges
in the QPLD-XP calculations. For the latter, ESP fit charges in the
‘field’ of the receptor yielded superior results compared to those
calculated for the free ligand in the gas phase. Our ligand data
set was admittedly small and, therefore, the success levels of the
methods must be treated with some scepticism as regards to their
application to larger ligand data sets and standalone predictability
of ligand binding affinities. Nevertheless, the small data set here
allowed us to analyze results in more detail and served as another
benchmark for developing an optimum protocol for docking at the
catalytic site of GP, carrying on from our previous work.19 A trade
off between accuracy versus computational expense must be con-
sidered in the choice between GLIDE-XP using default OPLS-AA
(2001) atomic partial charges or the more computationally expen-
sive QPLD-XP method for future applications. However, the
methods employed here can investigate the docking conformations
and relative binding affinities of a series of GP complexes with
good accuracy and significantly less computational effort, com-
pared to full-scale molecular dynamics free-energy simulations.52
2.4.1. Breakdown of individual GlideScore (GS) descriptor
contributions to ligand binding affinities
Given the accuracy of the docking results obtained, we decided
to decipher the descriptor components of GSs for each ligand into
their individual contributions using XP-Visualizer, to obtain useful
information or clues for the design of more potent inhibitors. For
this purpose, the GSs of most successful docking algorithm and
method were used: QPLD-XP with ESP fit ligand charges calculated
in the ‘field’ of the receptor and employing docking constraints on
the glucopyranose moiety positions. Correlation of R2 = 0.958 be-
tween docking scores and experimental binding free energies
was obtained (Fig. 7). The break down of descriptor information
is depicted graphically and in tabular form in Figure 8.
GlideScore is an empirical scoring function17 based on Chem-
Score,50 but modified and extended. The XP descriptors of rele-
vance used to score our set of ligands were Lipo-EvdW, H-bond,
SiteMap, LowMW, and RotPenal. The Lipo-EvdW descriptor is the
ChemScore lipophilic pair term and fraction of the total receptor-li-
gand van der Waals energy; H-bond is the ChemScore H-bonding
pair term; SiteMap scores the ligand/receptor non-H-bonding po-
lar/hydrophobic interactions following generation of hydropho-
bic–hydrophilic surfaces; LowMW rewards for ligands with low
molecular weight (MW); and RotPenal penalizes for ligand rotable
bonds.51
Lipo-EvdW is the major contributor to the final GSs, as clearly
indicated in Figure 8. The term is highest for the ligands 4e
(ꢁ6.92) and 4b (ꢁ6.15) with nonpolar substituents in the b-cavity,
naphthyl and p-Me-phenyl, respectively. para-Substitution of the
phenyl groups with polar substituents is not favorable with much
lower Lipo-EvdW contributions for 4a (ꢁ5.77) and 4d (ꢁ5.84), and
close to that for an unsubstituted phenyl moiety (4c (ꢁ5.72)). The
H-bond contribution for 4a with the para–OMe phenyl susbtituent
is highest (ꢁ2.72) forming a H-bond with Phe285 and close to a di-
rect H-bond with His341, but 4d with a –NO2 para-phenyl substi-
tuent is poorest (ꢁ2.17) and forms only water-mediated H-bonds
as seen in the crystal structure (vide supra). The SiteMap term is
approximately identical across the set of ligands. For the LowMW
term, although ligand 4e is penalized with respect to the other li-
gands for its bulkier/heavier naphthyl substituent, the gains else-
where are clear to be seen. Finally, for the RotPenal term, the
penalty contribution for restricting rotation of the (para-substi-
tuted) phenyl and naphthyl moieties is lowest (ꢀ+0.57) for 4e
(naphthyl rotation) and the p-NO2 phenyl moiety in 4d, and high-
est for the more flexible unsubstituted phenyl moiety (+0.75) of 4b.
Overall, the ligand 4e binds strongest due to its nonpolar and rigid
aromatic ring moiety. On this basis, substituted naphthyl groups or
more extended rigid aromatic groups should be investigated as
putative more potent inhibitors. Work in this area is currently un-
der way.
4. Experimental section
4.1. General methods
Thin-layer chromatography (TLC) was carried out on aluminum
sheets coated with Silica Gel 60 F254 (Merck). TLC plates were in-
spected by UV light (k = 312 nm) and developed by treatment with
a mixture of 10% H2SO4 in EtOH/H2O (1:1 v/v) followed by heating.
Silica gel column chromatography was performed with GeduranÒ
Silica Gel Si 60 (40–63 lm) purchased from Merck (Darmstadt,
Germany). Preparative reversed phase chromatography (RP-18)
was performed using a 15 ꢂ 150 mm column of fully endcapped
Silica Gel 100 C18 (>400 mesh, Fluka). 1H and 13C NMR spectra were
recorded at 23 °C using Bruker Advance DRX300 or DRX500 spec-
trometers with the residual solvent as the internal standard. The
following abbreviations are used to explain the observed multiplic-
ities: s, singlet; d, doublet; dd, doublet of doublet; ddd, doublet of
doublet of doublet; t, triplet; td, triplet of doublet; q, quadruplet;
m, multiplet; br, broad; p, pseudo. Structure elucidation was de-
duced from 1D and 2D NMR spectroscopy which allowed, in most
cases, complete signal assignments based on COSY, HSQC, and
HMBC correlations. NMR solvents were purchased from Euriso-
Top (Saint Aubin, France). HRMS (LSIMS) mass spectra were re-
corded in the positive mode using a Thermo Finnigan Mat 95 XL
spectrometer. MS (ESI) mass spectra were recorded in the positive

M. Benltifa et al. / Bioorg. Med. Chem. 17 (2009) 7368–7380
7377
mode using a Thermo Finnigan LCQ spectrometer. Optical rotations
were measured using a Perkin–Elmer polarimeter.
(320 lL, 2.30 mmol) was treated according to method A. The resi-
due was purified by flash silica gel column chromatography (PE/
EtOAc, 7:3) to afford 3b (138 mg, 95%) as a pale yellow foam.
4.1.1. [3+2]-Cycloaddition of aromatic nitrile oxides to exo-
glucals (Method A)
Rf = 0.22 (PE/EtOAc, 7:3);
½
a 2D0 = +57 (c 1, CH2Cl2); 1H NMR
ꢃ
(300 MHz, CDCl3) d 2.02 (s, 3H, CH3), 2.05 (s, 3H, CH3), 2.06 (s,
3H, CH3), 2.07 (s, 3H, CH3), 2.39 (s, 3H, PhCH3), 3.31 (d, 1H,
J = 17.8 Hz, H-4a), 3.41 (d, 1H, J = 17.8 Hz, H-4b), 4.05 (d, 1H,
J = 10.8 Hz, CH2OAc), 4.28 (m, 1H, CH2OAc), 4.30 (m, 1H, H-7),
A solution of exo-glucal 2 (0.3 mmol) and a-chloroarylaldoxime
(1.5 mmol, 5 equiv) in CH2Cl2 (5 mL) was stirred at rt under Argon.
Triethylamine (2.25 mmol, 7.5 equiv) was dissolved in CH2Cl2
(5 mL) and slowly added dropwise in 5 h with a syringe pump.
The reaction was stirred at rt for an additional 12 h then the sol-
vent was evaporated. The residue was purified by flash silica gel
column chromatography to afford the desired cycloadducts 3a–e.
0
5.21 (dd, 1H, J = 9.6 Hz, J = 9.9 Hz, H-8), 5.42 (d, 1H, J = 10.1 Hz,
H-10), 5.55 (dd, 1H, J = 10.1 Hz, J = 9.6 Hz, H-9), 7.22 (d, 2H,
J = 8.1 Hz, H-ar), 7.54 (d, 2H, J = 8.1 Hz, H-ar); 13C NMR (75 MHz,
CDCl3) d 20.50 (CH3), 20.53 (CH3), 20.57 (CH3), 20.64 (CH3), 21.4
(PhCH3), 43.4 (C-4), 61.4 (CH2OAc), 67.8 (C-8), 69.1 (C-10), 69.3
(C-7), 71.6 (C-9), 106.8 (C-5), 125.3 (C-ar), 126.7 (s, 2C, CH-ar),
129.5 (s, 2C, CH-ar), 141.3 (C-ar),157.6 (C-3), 169.5, 169.7, 170.2,
170.5 (4s, OCOCH3); MS (ESI) m/z = 477.8 [M+H]+, 500.0 [M+Na]+,
954.6 [2M+H]+, 976.6 [2M+Na]+; HRMS (ESI) m/z = C23H28NO10
[M+H]+ calcd 478.1713, found 478.1718.
4.1.2. Zemplén deacetylation (Method B)
A solution of the acetyl-protected carbohydrate derivatives 3a–
e (0.15 mmol) and NaOMe (50 lL, 1 M in MeOH) in MeOH (3 mL)
was stirred at rt for 3 h. The solution was then neutralized to pH
5 with a cation exchange resin (DOWEX 50WX2, H+ form). The re-
sin was filtered off and washed with MeOH (3 ꢂ 10 mL) then the
filtrate was evaporated off. The residue was purified by flash silica
gel column chromatography to afford the desired hydroxylated
carbohydrate derivatives 4a–e.
4.2.4. (5R,7R,8R,9S,10R)-8,9,10-Tris(acetoxy)-7-[(acetoxy)methyl]-
3-(phenyl)-1,6-dioxa-2-azaspiro[4,5]dec-2-ene (3c)
A
solution composed of
2
(51 mg, 0.15 mmol),
a
-chloro-
L,
benzaldoxime (115 mg, 0.74 mmol) and triethylamine (155
l
4.2. Syntheses of glucose-based inhibitors
1.10 mmol) was treated according to method A. The residue was
purified by flash silica gel column chromatography (PE/EtOAc, 3:2)
to afford 3c (67 mg, 99%) as a white solid. Rf = 0.25 (PE/EtOAc, 3:2);
4.2.1. 3,4,5,7-Tetra-O-acetyl-2,6-anhydro-1-deoxy-D-gluco-
hept-1-enitol (2)39
mp = 55–56 °C (CH2Cl2/EP);
½
a 2D0 = +60 (c 1, CHCl3); 1H NMR
ꢃ
A solution of 3,4,5,7-tetra-O-triethylsilyl-2,6-anhydro-1-deoxy-
(300 MHz, CDCl3) d 2.00 (s, 3H, CH3), 2.03 (s, 3H, CH3), 2.04 (s, 3H,
CH3), 2.06 (s, 3H, CH3), 3.31 (d, 1H, J = 17.8 Hz, H-4a), 3.41 (d, 1H,
J = 17.8 Hz, H-4b), 4.02 (d, 1H, J = 11.1 Hz, CH2OAc), 4.27 (m, 1H,
CH2OAc), 4.34 (m, 1H, H-7), 5.19 (t, 1H, J = 9.7 Hz, J = 9.7 Hz, H-8),
5.42 (d, 1H, J = 10.1 Hz, H-10), 5.54 (dd, 1H, J = 10.1 Hz, J = 9.7 Hz,
H-9), 7.42(m, 3H, H-ar), 7.63(m, 2H, H-ar);13C NMR(75 MHz, CDCl3)
d 20.53 (CH3), 20.56 (CH3), 20.6 (CH3), 20.7 (CH3), 43.3 (C-4), 61.4
(CH2OAc), 67.7 (C-8), 69.1 (C-10), 69.4 (C-7), 71.5 (C-9), 106.9 (C-
5), 126.8 (s, 2C, CH-ar), 128.2 (C-ar), 128.8 (s, 2C, CH-ar), 130.9 (C-
ar), 157.7 (C-3), 169.5, 169.7, 170.2, 170.6 (4s, OCOCH3); MS (ESI)
m/z = 463.9 [M+H]+, 486.0 [M+Na]+, 948.7 [2M+Na]+; HRMS (ESI)
m/z = C22H25NO10Na [M+Na]+ calcd 486.1376, found 486.1378.
D
-gluco-hept-1-enitol37 (526 mg, 0.83 mmol) in tetrabutylammo-
nium fluoride (4.16 mL, 1 M in THF) was stirred at room tempera-
ture for 4 h. The solution was then diluted with pyridine (4 mL)
and acetic anhydride (2 mL). The reaction mixture was stirred at
rt for 16 h. The solution was poured into iced water (100 mL)
and the aqueous layer extracted with EtOAc (3 ꢂ 150 mL). The or-
ganic layers were combined, dried (MgSO4) and evaporated. The
residue was purified by flash silica gel column chromatography
(PE/EtOAc, 7:3) to afford 239 (250 mg, 87%) as a colorless syrup.
4.2.2. (5R,7R,8R,9S,10R)-8,9,10-Tris(acetoxy)-7-[(acetoxy)methyl]-
3-(4-methoxyphenyl)-1,6-dioxa-2-azaspiro[4,5]dec-2-ene (3a)
A solution composed of 2 (55 mg, 0.16 mmol),
methoxybenzaldoxime (89 mg, 0.48 mmol) and triethylamine
(100 L, 0.72 mmol) was treated according to method A. The resi-
a
-chloro-4-
4.2.5. (5R,7R,8R,9S,10R)-8,9,10-Tris(acetoxy)-7-[(acetoxy)methyl]-
3-(4-nitrophenyl)-1,6-dioxa-2-azaspiro[4,5]dec-2-ene (3d)
l
A solution composed of 2 (43 mg, 0.12 mmol),
benzaldoxime (124 mg, 0.62 mmol) and triethylamine (130
a
-chloro-4-nitro-
L,
due was purified by flash silica gel column chromatography (PE/
l
EtOAc, 3:2) to afford 3a (65 mg, 83%) as a white solid. Rf = 0.23
0.93 mmol) was treated according to method A. The residue was
purified by flash silica gel column chromatography (PE/EtOAc, 3:2)
to afford 3d (59 mg, 94%) as a pale yellow solid. Rf = 0.25 (PE/EtOAc,
(PE/EtOAc, 3:2); mp = 68–69 °C (CH2Cl2/EP);
½
a 2D0
= +75 (c 1,
ꢃ
CHCl3); 1H NMR (300 MHz, CDCl3) d 2.00 (s, 3H, CH3), 2.04 (s, 3H,
CH3), 2.05 (s, 3H, CH3), 2.06 (s, 3H, CH3), 3.29 (d, 1H, J = 17.7 Hz,
H-4a), 3.38 (d, 1H, J = 17.7 Hz, H-4b), 3.83 (s, 3H, OCH3), 4.03 (dd,
1H, J = 2.0 Hz, J = 12.6 Hz, CH2OAc), 4.27 (dd, 1H, J = 3.7 Hz,
J = 12.6 Hz, CH2OAc), 4.34 (ddd, 1H, J = 2.0 Hz, J = 3.7, J = 10.1 Hz,
H-7), 5.19 (dd, 1H, J = 9.5 Hz, J = 10.1 Hz, H-8), 5.41 (d, 1H,
J = 10.0 Hz, H-10), 5.53 (dd, 1H, J = 10.0 Hz, J = 9.5 Hz, H-9), 6.91
(d, 2H, J = 8.9 Hz, H-ar), 7.58 (d, 2H, J = 8.9 Hz, H-ar); 13C NMR
(75 MHz, CDCl3) d 20.58 (CH3), 20.59 (CH3), 20.6 (CH3), 20.7
(CH3), 43.5 (C-4), 55.4 (OCH3), 61.4 (CH2OAc), 67.8 (C-8), 69.1 (C-
10), 69.2 (C-7), 71.6 (C-9), 106.7 (C-5), 114.2 (s, 2C, CH-ar), 120.6
(C-ar), 128.5 (s, 2C, CH-ar), 157.3 (C-3), 161.6 (C-ar), 169.6, 169.8,
170.3, 170.6 (4s, OCOCH3); MS (ESI) m/z = 493.9 [M+H]+, 516.0
[M+Na]+, 986.6 [2M+H]+, 1008.7 [2M+Na]+; HRMS (ESI) m/
z = C23H27NO11Na [M+Na]+ calcd 516.1481, found 516.1488.
3:2); mp = 119–120 °C (CH2Cl2/EP); ½a D20
ꢃ
= +58 (c 1, CHCl3); 1H NMR
(300 MHz, CDCl3) d 2.00 (s, 3H, CH3), 2.03 (s, 3H, CH3), 2.04 (s, 3H,
CH3), 2.05 (s, 3H, CH3), 3.33 (d, 1H, J = 17.9 Hz, H-4a), 3.44 (d, 1H,
J = 17.9 Hz, H-4b), 4.05 (d, 1H, J = 10.8 Hz, CH2OAc), 4.26 (m, 1H,
CH2OAc), 4.32 (m, 1H, H-7), 5.21 (dd, 1H, J = 9.4 Hz, J = 10.0 Hz, H-
8), 5.43 (d, 1H, J = 10.1 Hz, H-10), 5.54 (dd, 1H, J = 10.1 Hz,
J = 9.4 Hz, H-9), 7.81 (d, 2H, J = 8.9 Hz, H-ar), 8.26 (d, 2H, J = 8.9 Hz,
H-ar); 13C NMR (75 MHz, CDCl3) d 20.50 (CH3), 20.52 (CH3), 20.57
(CH3), 20.64 (CH3), 42.8 (C-4), 61.3 (CH2OAc), 67.6 (C-8), 69.1 (C-
10), 69.7 (C-7), 71.4 (C-9), 107.8 (C-5), 124.1 (s, 2C, CH-ar), 127.7
(s, 2C, CH-ar), 134.2 (C-ar), 148.9 (C-ar),156.2 (C-3), 169.5, 169.7,
170.1, 170.5 (4s, OCOCH3); MS (ESI) m/z = 508.8 [M+H]+, 531.0
[M+Na]+, 1038.6 [2M+Na]+; HRMS (ESI) m/z = C22H24N2O12Na
[M+Na]+ calcd 531.1227, found 531.1224.
4.2.3. (5R,7R,8R,9S,10R)-8,9,10-Tris(acetoxy)-7-[(acetoxy)methyl]-
4.2.6. (5R,7R,8R,9S,10R)-8,9,10-Tris(acetoxy)-7-[(acetoxy)methyl]-
3-(4-toluyl)-1,6-dioxa-2-azaspiro[4,5]dec-2-ene (3b)
3-(2-naphthyl)-1,6-dioxa-2-azaspiro[4,5]dec-2-ene (3e)
A solution composed of 2 (105 mg, 0.30 mmol),
a-chloro-4-
A solution composed of 2 (100 mg, 0.29 mmol),
a-chloro-2-
methybenzaldoxime (260 mg, 1.52 mmol) and triethylamine
naphthaldoxime (296 mg, 1.40 mmol) and triethylamine (300
lL,

7378
M. Benltifa et al. / Bioorg. Med. Chem. 17 (2009) 7368–7380
2.15 mmol) was treated according to method A. The residue was
purified by flash silica gel column chromatography (PE/EtOAc,
3:1) to afford 3e (140 mg, 94%) as a pale yellow foam. Rf = 0.28
J = 10.0 Hz, H-8), 3.60 (d, 1H, J = 9.7 Hz, H-10), 3.69–3.77 (m, 4H,
CH2OH H-4b H-9), 3.83 (ddd, 1H, J = 2.3 Hz, J = 4.7, J = 10.0 Hz, H-
7), 7.44 (m, 3H, H-ar), 7.69 (m, 2H, H-ar); 13C NMR (75 MHz,
CD3OD) d 44.2 (C-4), 62.4 (CH2OH), 71.4 (C-8), 73.2 (C-100), 75.9
(C-7), 76.2 (C-9), 110.9 (C-5), 127.8 (s, 2C, CH-ar), 129.9 (s, 2C,
CH-ar), 130.6 (C-ar),131.6 (C-ar), 159.6 (C-3); MS (ESI) m/
z = 296.0 [M+H]+, 318.0 [M+Na]+, 612.9 [2M+Na]+; HRMS (ESI) m/
z = C14H17N1O6Na [M+Na]+ calcd 318.0954, found 318.0951.
(PE/EtOAc, 3:1); ½a D20
ꢃ
= +39 (c 1, CH2Cl2); 1H NMR (300 MHz, CDCl3)
d 2.01 (s, 3H, CH3), 2.04 (s, 3H, CH3), 2.04 (s, 3H, CH3), 2.05 (s, 3H,
CH3), 3.43 (d, 1H, J = 17.7 Hz, H-4a), 3.54 (d, 1H, J = 17.7 Hz, H-4b),
4.05 (m, 1H, CH2OAc), 4.25 (m, 2H, H-7 CH2OAc), 5.22 (t, 1H,
J = 9.7 Hz, J = 9.7 Hz, H-8), 5.46 (d, 1H, J = 10.1 Hz, H-10), 5.57 (dd,
1H, J = 10.1 Hz, J = 9.7 Hz, H-9), 7.50 (m, 2H, H-ar), 7.90 (m, 5H,
H-ar); 13C NMR (75 MHz, CDCl3): d = 20.48 (CH3), 20.51 (CH3),
20.55 (CH3), 20.62 (CH3), 43.2 (C-4), 61.4 (CH2OAc), 67.7 (C-8),
69.1 (C-10), 69.4 (C-7), 71.5 (C-9), 107.0 (C-5), 123.0 (CH-ar),
125.7 (C-ar), 127.5, 127.6, 127.8, 128.3, 128.6 (5s, 6C, CH-ar),
132.7 (C-ar), 134.2 (C-ar), 157.8 (C-3), 169.5, 169.7, 170.2, 170.5
(4s, OCOCH3); MS (ESI) m/z = 513.9 [M+H]+, 536.1 [M+Na]+,
1026.7 [2M+H]+, 1048.7 [2M+Na]+; HRMS (ESI) m/z = C26H27NO10-
Na [M+Na]+ calcd 536.1533, found 536.1535.
4.2.10. (5R,7R,8R,9S,10R)-7-(Hydroxymethyl)-3-(4-nitrophenyl)-
1,6-dioxa-2-azaspiro[4,5]dec-2-ene-8,9,10-triol (4d)
A solution of 3d (51 mg, 0.10 mmol) was treated according to
method B. The residue was purified by flash silica gel column
chromatography (PE/EtOAc, 3:2 then EtOAc then EtOAc/MeOH
9:1) to afford 4d (34 mg, 99%) as a pale yellow solid. Rf = 0.30
(EtOAc/MeOH, 4/1); mp = 181–182 °C (MeOH/Et2O); ½a D20
ꢃ
= +46
(c 0.5, MeOH); 1H NMR (500 MHz, CD3OD, 55 °C, with CD3OH as
internal standard at 4.84 ppm) d 3.71 (d, 1H, H-4a), 3.81 (dd,
1H, J = 9.4 Hz, J = 9.6 Hz, H-8), 3.96 (d, 1H, J = 9.3 Hz, H-10),
4.05–4.12 (m, 3H, CH2OH H-4b H-9), 4.14 (dd, 1H, J = 2.4 Hz,
J = 12.0 Hz, CH2OH), 4.19 (ddd, 1H, J = 9.6 Hz, J = 2.4 Hz, J = 4.8,
H-7), 8.27 (d, 2H, J = 8.6 Hz, H-ar), 8.62 (d, 2H, J = 8.6 Hz, H-ar);
13C NMR (125 MHz, CD3OD, 55 °C) d 43.8 (C-4), 62.6 (CH2OH),
71.6 (C-8), 73.4 (C-10), 76.19, 76.21 (2s, 2C, C-7 C-9), 111.9 (C-
5), 125.0 (s, 2C, CH-ar), 128.7 (s, 2C, CH-ar), 136.9 (C-ar), 150.3
(C-ar), 158.2 (C-3); MS (ESI) m/z = 341.0 [M+H]+, 363.0 [M+Na]+;
HRMS (ESI) m/z = C14H17N2O8Na [M+H]+ calcd 341.0985, found
341.0987.
4.2.7. (5R,7R,8R,9S,10R)-7-(Hydroxymethyl)-3-(4-methoxyphenyl)-
1,6-dioxa-2-azaspiro[4,5]dec-2-ene-8,9,10-triol (4a)
A solution of 3a (61 mg, 0.12 mmol) was treated according to
method B. The residue was purified by flash silica gel column chro-
matography (PE/EtOAc, 3:2 then EtOAc then EtOAc/MeOH 9:1) to
afford 4a (40 mg, 98%) as a white solid. Rf = 0.34 (EtOAc/MeOH,
4:1); mp = 214–215 °C (MeOH/Et2O); ½a D20
ꢃ
= +105 (c 0.5, MeOH);
1H NMR (500 MHz, CD3OD, 55 °C, with CD3OH as internal standard
at 4.84 ppm) d 3.60 (d, 1H, H-4a), 3.75 (dd, 1H, J = 8.9 Hz, J = 9.2 Hz,
H-8), 3.89 (d, 1H, J = 9.8 Hz, H-10), 3.99–4.12 (m, 5H, H-4b CH2OH
H-7 H-9), 4.15 (s, 3H, OCH3), 7.28 (d, 2H, J = 8.1 Hz, H-ar), 7.93 (d,
2H, J = 8.1 Hz, H-ar). The reported data16 were referenced with
CD2HOD as internal standard at 3.31 ppm; 13C NMR (125 MHz,
CD3OD, 55 °C) d 44.6 (C-4), 56.0 (OCH3), 62.6 (CH2OH), 71.7 (C-8),
73.3 (C-10), 75.9 (C-9), 76.4 (C-7), 110.7 (C-5), 115.4 (s, 2C, CH-
ar), 123.1 (C-ar), 129.5 (s, 2C, CH-ar), 159.5 (C-3), 163.2 (C-ar);
MS (ESI) m/z = 326.0 [M+H]+, 348.0 [M+Na]+, 672.9 [2M+Na]+,
997.6 [3M+Na]+; HRMS (ESI) m/z = C15H20N1O7 [M+H]+ calcd
326.1240, found 326.1243.
4.2.11. (5R,7R,8R,9S,10R)-7-(Hydroxymethyl)-3-(2-naphthyl)-1,6-
dioxa-2-azaspiro[4,5]dec-2-ene-8,9,10-triol (4e)
A solution of 3e (130 mg, 0.25 mmol) was treated according to
method B. The residue was purified by flash silica gel column chro-
matography (PE/EtOAc, 3:2 then EtOAc then EtOAc/MeOH 9:1) to
afford 4e (68 mg, 78%) as a white solid. Rf = 0.50 (EtOAc/MeOH
4:1); mp = 196–198 °C;
(500 MHz, CD3OD, 60 °C, with CD2HOD as internal standard at
3.31 ppm) 3.46 (d, 1H, J = 17.4 Hz, H-4a), 3.47 (dd, 1H,
½
a 2D0 = +68 (c 0.33, MeOH); 1H NMR
ꢃ
d
4.2.8. (5R,7R,8R,9S,10R)-7-(Hydroxymethyl)-3-(4-toluyl)-1,6-di-
oxa-2-azaspiro[4,5]dec-2-ene-8,9,10-triol (4b)
J = 9.2 Hz, J = 9.7 Hz, H-8), 3.64 (d, 1H, J = 9.7 Hz, H-10), 3.71–3.88
(m, 5H, CH2OH H-4b H-7 H-9), 7.53 (m, 2H, H-ar), 7.89 (m, 4H,
H-ar), 8.05 (s, 1H, H-ar); 13C NMR (125 MHz, CD3OD, 60 °C) d
44.3 (C-4), 62.7 (CH2OH), 71.7 (C-8), 73.4 (C-10), 76.0 (C-7), 76.4
(C-9), 111.1 (C-5), 124.2, 127.9 (2s, 2C, CH-ar), 128.2 (C-ar),
128.4, 128.5, 128.9, 129.6 (4s, 5C, CH-ar), 134.7 (C-ar), 135.8 (C-
ar), 159.8 (C-3); MS (ESI) m/z = 346 [M+H]+, 712.9 [2M+Na]+; HRMS
(ESI) m/z = C18H20NO6 [M+H]+ calcd 346.1291, found 346.1291.
A solution of 3b (135 mg, 0.30 mmol) was treated according to
method B. The residue was purified by flash silica gel column chro-
matography (PE/EtOAc, 3:2 then EtOAc then EtOAc/MeOH 9:1) to
afford 4b (81 mg, 93%) as a pale yellow solid. Rf = 0.31 (EtOAc/
MeOH, 9:1); mp = 200–202 °C; ½a D20
ꢃ
= +73 (c 1, MeOH); 1H NMR
(300 MHz, CD3OD, with CD2HOD as internal standard at
3.31 ppm) d 2.36 (s, 3H, PhCH3), 3.44 (d, 1H, J = 17.6 Hz, H-4a),
3.44 (dd, 1H, J = 9.3 Hz, J = 9.5 Hz, H-8), 3.58 (d, 1H, J = 9.8 Hz, H-
10), 3.67–3.85 (m, 5H, CH2OH H-4b H-7 H-9), 7.24 (d, 2H,
J = 8.1 Hz, H-ar), 7.56 (d, 2H, J = 8.1 Hz, H-ar); 13C NMR (75 MHz,
CD3OD): d 21.5 (PhCH3), 44.3 (C-4), 62.4 (CH2OH), 71.4 (C-8),
73.2 (C-10), 75.9 (C-7), 76.2 (C-9), 110.7 (C-5), 127.7 (C-ar), 127.8
(s, 2C, CH-ar), 123.6 (s, 2C, CH-ar), 141.2 (C-ar),159.6 (C-3); MS
(ESI) m/z = 310.0 [M+H]+, 332.0 [M+Na]+, 640.9 [2M+Na]+; HRMS
(ESI) m/z = C15H20NO6 [M+H]+ calcd 310.1291 found 310.1292.
4.3. Enzyme kinetics
Rabbit muscle GPb was isolated, purified, recrystallized, and as-
sayed as described.53,54 Kinetic experiments were performed in the
direction of glycogen synthesis in the presence of constant concen-
trations of glycogen (0.2% w/v), AMP (1 mM), various concentra-
tions of Glc-1-P (2–20 mM) and inhibitors, in 30 mM imidazole
buffer (pH 6.8), 60 mM KCl, 0.6 mM dithiothreitol, 0.6 mM EDTA,
and 2% DMSO.55
4.2.9. (5R,7R,8R,9S,10R)-7-(Hydroxymethyl)-3-(phenyl)-1,6-di-
oxa-2-azaspiro[4,5]dec-2-ene-8,9,10-triol (4c)
4.4. X-ray crystallography
A solution of 3c (17.6 mg, 0.04 mmol) was treated according to
method B. The residue was purified by flash silica gel column chro-
matography (PE/EtOAc, 3:2 then EtOAc then EtOAc/MeOH 9:1) to
afford 4c (11 mg, 97%) as a pale yellow solid. Rf = 0.36 (EtOAc/
Native T-state GPb crystals, grown in the tetragonal lattice,
space group P43212,56 were soaked with either 20 mM 4a (for
10 h), 20 mM 4c (for 5 h, or 20 mM 4d (for 10 h) in a buffered solu-
tion (10 mM Bes, pH 6.7, 10–20% DMSO), prior data collection. Sim-
ilarly, data for GPb–4e and GPb–4b complexes were collected from
single crystals soaked with either 6 mM 4e (for 5 h) or 30 mM 4b in
MeOH, 4:1); mp = 195–197 °C (MeOH/Et2O); ½a D20
ꢃ
= +69 (c 0.63,
MeOH); 1H NMR (300 MHz, CD3OD, with CD2HOD as internal stan-
dard at 3.31 ppm) d 3.35 (d, 1H, H-4a), 3.45 (t, 1H, J = 9.3 Hz,

M. Benltifa et al. / Bioorg. Med. Chem. 17 (2009) 7368–7380
7379
10 mM Bes, pH 6.7, 30% DMSO, prior to mounting in a loop and
flash-cooling in the nitrogen stream at 100 K. Diffraction data were
collected from single crystals at EMBL-Hamburg outstation (Beam-
line X13) and SRS, Daresbury Laboratory (PX 9.6) at 293 K for 4a,
4c, and 4d and 100 K for 4b and 4e. Data reduction and integration
followed by scaling and merging of the intensities obtained was
performed with Denzo and Scalepack, respectively, as imple-
mented in HKL suite.57
Crystallographic refinement of the complexes was performed
by maximum-likelihood methods using REFMAC.42 The starting
model employed for the refinement of the complex was the
100 K structure of the native T-state GPb determined at 2.0 Å res-
olution.10 2Fo ꢁ Fc and Fo ꢁ Fc electron density maps were calcu-
lated and visualized using the program for molecular graphics
‘O’.58 Ligand models of 4a–e compounds were fitted to the electron
density maps after adjustment of their torsion angles. Alternate cy-
cles of manual rebuilding with ‘O’ and refinement with REFMAC
improved the quality of the models.
of 0.5 Å for heavy atoms and O–H hydrogen atoms) within
100 kJ/mol of the global minimum conformation were saved for
each ligand, and these conformations then clustered using XClus-
ter71 into five sets of conformations based on atomic coordinate
RMS differences. The ‘leading’ or lowest energy conformation
was taken from each of the five clusters. The resulting five confor-
mations/ligand were then taken, plus the original minimized con-
formation for each ligand, and these six conformations were used
for docking into the previously prepared 4e receptor. Docking
accuracy is limited by sampling so that use of an ensemble of con-
formations for each ligand will help reduce errors associated with
inadequate sampling.
For the GLIDE docking runs, both the extra-precision (XP)17,18
mode with standard OPLS-AA(2001) charges (standard GLIDE-
XP) and quantum mechanics—polarized ligand docking (QPLD-
XP) were performed.20 In the GLIDE-XP docking runs, standard
parameters were applied including van der Waals scaling for non-
polar atoms (by 0.8) to include modest ‘induced fit’ effects, with
up to 3 poses per ligand input conformation saved. For QPLD-
XP docking runs, the output docking poses from GLIDE-XP dock-
ing with OPLS-AA(2001) partial charges were used to obtain elec-
The stereochemistry of the protein residues was validated by
PROCHECK.59,60 Hydrogen bonds and van der Waals interactions
were calculated with the program CONTACT as implemented in
CCP459 applying a distance cut off 3.3 Å and 4.0 Å, respectively.
Protein structures were superimposed using LSQKAB.59 The figures
trostatic potential (ESP) fit ligand charges using the program
20
QSITE
and the QM/MM method in the ‘field’ of the receptor,
61
were prepared with the program MOLSCRIPT and rendered with
and for comparison, ESP fit ligand charges in the gas phase (free
ligand; no receptor) using QM. Single point energy calculation
using the B3LYP72–74 density functional theory (DFT) method
and 6–31G* basis sets75–77 on the ligand (QM part) and the
OPLS-AA forcefield for the receptor (MM part) were used for this
purpose. The ligands were then redocked into the catalytic site
using GLIDE in XP mode and the new ligand charges. Binding con-
straints on the glucopyranose were applied in the docking calcu-
lations as described in Section 2.4 and Tables 1 and 2. The
obtained ligand poses were considered conformationally distinct
if their RMSD (heavy atoms) was larger than 0.5 Å. The quality
of the docking results was assessed by comparing (superimpos-
ing) the top-ranked pose for each method with the native X-ray
conformation from the GPb–ligand complexes, and measurement
of the correlation between the GlideScores obtained and experi-
mental binding free energies. Docked ligand binding conforma-
tion results were considered ‘accurate’ when the RMSD of the
ligand heavy atom positions in the top-ranked pose and the crys-
tallographic conformation was less than 1.0 Å.
Raster3D62 and with the program MG 63,64 Coordinate sets for com-
.
parisons were 2PYD (100 K GPb-a-D-glucose complex) or 100 K
T-state native GPb (code 2GPN). The coordinates of the new struc-
tures have been deposited with the RCSB Protein Data Bank (http://
www.rcsb.org/pdb) with codes 2QRG (GPb–4a complex), 2QRQ
(GPb–4b complex), 2QRH (GPb–4c complex), 2QRM (GPb–4d com-
plex), and 2QRP (GPb–4e complex).
4.5. Docking calculations
Docking calculations were performed using the program GLIDE
4.5.51 The initial setup of the GPb enzyme for docking of the ligands
was performed using Schrodingers’ ‘Protein Preparation Wizard’,
starting from the X-ray crystal structure of the GPb–4e complex.
Residual DMSO molecules along with all crystallographic H2O mol-
ecules were deleted, bond orders assigned and hydrogen atoms
added. The initial assignments of protonation states for basic and
acidic residues and tautomeric states was based on pKa at their
normal pH (7.0). However, subsequent optimization of hydroxyl,
histidine protonation states and C/N atom ‘flips’, and sidechain
O/N atom ‘flips’ of Asn and Gln was based on optimizing hydrogen
bonding patterns, so that the final assignments were checked on
visual inspection of the protein. In particular, all final His residues
were assigned as neutral, either in a HIE or HID state. The phos-
phate in PLP was used in its mono-anion form. Finally, an ‘Impref’
minimization of the GPb–4e complex was performed using the
OPLS-AA (2001) force field to remove steric clashes and bad con-
tacts. At the end of the minimization, the root-mean-square devi-
ation (RMSD) of all heavy atoms was within 0.3 Å of the
crystallographic positions.
In the docking calculations with GLIDE, the shape and proper-
ties of the catalytic binding site were mapped onto grids with
dimensions of 27 Å ꢂ 27 Å ꢂ 27 Å, centered on the 4e ligand. The
five ligands 4a–e were extracted from their crystal complex struc-
tures and prepared for docking using Schrodingers’ Maestro and
the BUILD module, followed by minimization (500 steps, TNCG)
using the OPLS-AA (2001) force field and MacroModel.65 The min-
imized structures (500 steps, TNCG) were then subjected to a con-
formational search in bulk water solvent (GB/SA continuum
model).66–68 1000 steps of the mixed mode Monte Carlo Multiple
Minima/Low-Mode conformational search (MCMM/LMCS) algo-
rithm were used.69,70 Unique conformations (based on an RMSD
Breakdown of the contributions to GlideScores using the XP-
Visualizer module was performed for the most successful docking
method (QPLD-XP, ESP fit charges calculated in receptor ‘field’, and
constraints on the glucopyranose moiety positions). This allowed
us to analyze and compare the contributions of the different
descriptor terms for the individual ligands.
Acknowledgements
This work was supported by the EU Marie Curie Early Stage
Training (EST) contract no MEST-CT-020575; a Marie Curie Host
Fellowships for the Transfer of Knowledge (ToK) contract no
MTKD-CT-2006-042776; the Greek GSRT Scientific and Technolog-
ical cooperation between Greece and USA (2004–2007); the Hel-
lenic General Secretariat for Research and Technology (GSRT),
Ministry of Development, through the program ‘Excellence in
Research Institutes (2nd cycle)’ to N.G.O.; the EMBL-Hamburg out-
station under FP6 ‘Structuring the European Research Area Pro-
gramme’ contract no RII3/CT/2004/5060008, and SRS Daresbury
Laboratory (contract IHPP HPRI-CT-1999-00012). Financial support
by University Lyon 1, CNRS, the Région Rhône-Alpes (funds from
MIRA Recherche 2001 & 2006 and stipends to M.B.) and the Euro-
pean Union (FP6-IP 503467) is gratefully acknowledged. Support
provided by the French Ministry of Foreign Affairs (PAI Balaton

7380
M. Benltifa et al. / Bioorg. Med. Chem. 17 (2009) 7368–7380
34. Taillefumier, C.; Chapleur, Y. Chem. Rev. 2004, 104, 263.
35. Rajanbabu, T. V.; Reddy, G. S. J. Org. Chem. 1986, 51, 5458.
36. Bourdon, B.; Corbet, M.; Fontaine, P.; Goekjian, P. G.; Gueyrard, D. Tetrahedron
Lett. 2008, 49, 747.
n°11 061PE) is also acknowledged. We would furthermore like to
thank Schrödinger, particularly Jianxin Duan for technical support
and useful discussions regarding the computational work.
37. Gueyrard, D.; Haddoub, R.; Salem, A.; Bacar, N. S.; Goekjian, P. G. Synlett 2005,
520.
Supplementary data
38. Toth, M.; Kover, K. E.; Benyei, A.; Somsak, L. Org. Biomol. Chem. 2003, 1, 4039.
39. Toth, M.; Somsak, L. J. Chem. Soc., Perkin Trans. 1 2001, 942.
40. Benltifa, M.; Vidal, S.; Fenet, B.; Msaddek, M.; Goekjian, P. G.; Praly, J.-P.;
Brunyánszki, A.; Docsa, T.; Gergely, P. Eur. J. Org. Chem. 2006, 4242.
41. Tóth, M.; Kun, S.; Bokor, É.; Benltifa, M.; Tallec, G.; Vidal, S.; Docsa, T.; Gergely,
P.; Somsák, L.; Praly, J.-P. Bioorg. Med. Chem. 2009, 17, 4773.
42. Murshudov, G. N.; Vagin, A. A.; Dodson, E. J. Acta Crystallogr., Sect. D 1997, 53,
240.
43. Oikonomakos, N. G. Curr. Prot. Pept. Sci. 2002, 3, 561.
44. Chrysina, E. D.; Kosmopoulou, M. N.; Kardakaris, R.; Bischler, N.; Leonidas, D.
D.; Kannan, T.; Loganathan, D.; Oikonomakos, N. G. Bioorg. Med. Chem. 2005, 13,
765.
Supplementary data (data collection and refinement statistics
for compounds 4a–e, hydrogen bonding interactions, water hydro-
gen-bonding network and Van der Waals interactions for com-
pound 4a–e) associated with this article can be found, in the
online version, at doi:10.1016/j.bmc.2009.08.060.
References and notes
45. Tsitsanou, K. E.; Oikonomakos, N. G.; Zographos, S. E.; Skamnaki, V. T.;
Gregoriou, M.; Watson, K. A.; Johnson, L. N.; Fleet, G. W. J. Protein Sci. 1999, 8,
741.
46. Mitchell, E. P.; Withers, S. G.; Ermert, P.; Vasella, A. T.; Garman, E. F.;
Oikonomakos, N. G.; Johnson, L. N. Biochemistry 1996, 35, 7341.
47. Zographos, S. E.; Oikonomakos, N. G.; Tsitsanou, K. E.; Leonidas, D. D.; Chrysina,
E. D.; Skamnaki, V. T.; Bischoff, H.; Goldmann, S.; Watson, K. A.; Johnson, L. N.
Structure 1997, 5, 1413.
48. Oikonomakos, N. G.; Tsitsanou, K. E.; Zographos, S. E.; Skamnaki, V. T.;
Goldmann, S.; Bischoff, H. Protein Sci. 1999, 8, 1930.
49. Leonidas, D. D.; Zographos, S. E.; Oikonomakos, N. G., unpublished results.
50. Eldridge, M. D.; Murray, C. W.; Auton, T. R.; Paolini, G. V.; Mee, R. P. J. Comput.
Aided Mol. Des. 1997, 11, 425.
1. Ross, S. A.; Gulve, E. A.; Wang, M. H. Chem. Rev. 2004, 104, 1255.
2. American Diabetes Association. Diabetes Care 2004, 27, S5.
3. Baker, D. J.; Timmons, J. A.; Greenhaff, P. L. Diabetes 2005, 54, 2453.
4. Lukacs, C. M.; Oikonomakos, N. G.; Crowther, R. L.; Hong, L. N.; Kammlott, R. U.;
Levin, W.; Li, S.; Liu, C. M.; Lucas-McGady, D.; Pietranico, S.; Reik, L. Prot. Struct.
Funct. Bioinf. 2006, 63, 1123.
5. Treadway, J. L.; Mendys, P.; Hoover, D. J. Exp. Opin. Invest. Drugs 2001, 10, 439.
6. Oikonomakos, N. G.; Somsak, L. Curr. Opin. Invest. Drugs 2008, 9, 379.
7. Somsák, L.; Czifrák, K.; Tóth, M.; Bokor, É.; Chrysina, E. D.; Alexacou, K.-M.;
Hayes, J. M.; Tiraidis, C.; Lazoura, E.; Leonidas, D. D.; Zographos, S. E.;
Oikonomakos, N. G. Curr. Med. Chem. 2008, 2933.
8. Baker, D. J.; Greenhaff, P. L.; Timmons, J. A. Exp. Opin. Ther. Patents 2006, 16, 459.
9. Henke, B. R.; Sparks, S. M. Mini-Rev. Med. Chem. 2006, 6, 845.
10. Gregoriou, M.; Noble, M. E. M.; Watson, K. A.; Garman, E. F.; Krulle, T. M.; De-
La-Fuente, C.; Fleet, G. W. J.; Oikonomakos, N. G.; Johnson, L. N. Protein Sci.
1998, 7, 915.
51.
GLIDE Version 4.5, User Manual, Schrödinger, LLC, New York, NY, 2007.
52. Archontis, G.; Watson, K. A.; Xie, Q.; Andreou, G.; Chrysina, E. D.;
Zographos, S. E.; Oikonomakos, N. G.; Karplus, M. Prot. Struct. Funct. Bioinf.
2005, 61, 984.
53. Melpidou, A. E.; Oikonomakos, N. G. FEBS Lett. 1983, 154, 105.
54. Oikonomakos, N. G.; Zographos, S. E.; Johnson, L. N.; Papageorgiou, A. C.;
Acharya, K. R. J. Mol. Biol. 1995, 254, 900.
11. Martin, J. L.; Veluraja, K.; Ross, K.; Johnson, L. N.; Fleet, G. W. J.; Ramsden, N. G.;
Bruce, I.; Orchard, M. G.; Oikonomakos, N. G.; Papageorgiou, A. C.; Leonidas, D.
D.; Tsitoura, H. S. Biochemistry 1991, 30, 10101.
55. Anagnostou, E.; Kosmopoulou, M. N.; Chrysina, E. D.; Leonidas, D. D.;
Hadjiloi, T.; Tiraidis, C.; Zographos, S. E.; Gyorgydeak, Z.; Somsak, L.;
Docsa, T.; Gergely, P.; Kolisis, F. N.; Oikonomakos, N. G. Bioorg. Med. Chem.
2006, 14, 181.
12. Praly, J.-. P.; Boye, S.; Joseph, B.; Rollin, P. Tetrahedron Lett. 1993, 34, 3419.
13. Elek, R.; Kiss, L.; Praly, J.-. P.; Somsak, L. Carbohydr. Res. 2005, 340, 1397.
14. Somsák, L.; Nagy, V.; Vidal, S.; Czifrák, K.; Berzsényi, E.; Praly, J.-. P. Bioorg. Med.
Chem. Lett. 2008, 18, 5680.
56. Oikonomakos, N. G.; Melpidou, A. E.; Johnson, L. N. Biochim. Biophys. Acta 1985,
832, 248.
57. Otwinowski, Z.; Minor, W. In Macromolecular Crystallography, Pt A, 1997; Vol.
276, p 307.
15. Nagy, V.; Benltifa, M.; Vidal, S.; Berzsényi, E.; Teilhet, C.; Czifrák, K.; Batta, G.;
Docsa, T.; Gergely, P.; Somsák, L.; Praly, J.-. P. Bioorg. Med. Chem. 2009, 17, 5696.
16. Benltifa, M.; Vidal, S.; Gueyrard, D.; Goekjian, P. G.; Msaddek, M.; Praly, J.-. P.
Tetrahedron Lett. 2006, 47, 6143.
17. Friesner, R. A.; Banks, J. L.; Murphy, R. B.; Halgren, T. A.; Klicic, J. J.; Mainz, D. T.;
Repasky, M. P.; Knoll, E. H.; Shelley, M.; Perry, J. K.; Shaw, D. E.; Francis, P.;
Shenkin, P. S. J. Med. Chem. 2004, 47, 1739.
18. Friesner, R. A.; Murphy, R. B.; Repasky, M. P.; Frye, L. L.; Greenwood, J. R.;
Halgren, T. A.; Sanschagrin, P. C.; Mainz, D. T. J. Med. Chem. 2006, 49, 6177.
19. Alexacou, K. M.; Hayes, J. M.; Tiraidis, C.; Zographos, S. E.; Leonidas, D. D.;
Chrysina, E. D.; Archontis, G.; Oikonomakos, N. G.; Paul, J. V.; Varghese, B.;
Loganathan, D. Prot. Struct. Funct. Bioinf. 2008, 71, 1307.
20. Schrödinger Suite 2008 QM-Polarized Ligand Docking protocol; GLIDE Version
4.5, Schrödinger, LLC, New York, NY, 2007; JAGUAR Version 7.5, Schrödinger, LLC,
New York, NY, 2008; QSITE Version 5.0, Schrödinger, LLC, New York, NY, 2008.
21. Cho, A. E.; Guallar, V.; Berne, B. J.; Friesner, R. J. Comput. Chem. 2005, 26, 915.
22. Raha, K.; Peters, M. B.; Wang, B.; Yu, N.; WollaCott, A. M.; Westerhoff, L. M.;
Merz, K. M. Drug Discovery Today 2007, 12, 725.
58. Jones, T. A.; Zou, J. Y.; Cowan, S. W.; Kjeldgaard, M. Phys. Rev. B 1991, 47, 110.
59. CCP4. Acta Crystallogr., Sect. D 1994, 50, 760.
60. Laskowski, R. A.; Macarthur, M. W.; Moss, D. S.; Thornton, J. M. J. Appl.
Crystallogr. 1993, 26, 283.
61. Kraulis, P. J. Appl. Crystallogr. 1991, 24, 946.
62. Merritt, E. A.; Bacon, D. J.. In Macromolecular Crystallography, Pt B; Academic
Press: San Diego, 1997; Vol. 277, p 505.
63. Potterton, E.; McNicholas, S.; Krissinel, E.; Cowtan, K.; Noble, M. Acta
Crystallogr., Sect. D 2002, 58, 1955.
64. Potterton, L.; McNicholas, S.; Krissinel, E.; Gruber, J.; Cowtan, K.; Emsley, P.;
Murshudov, G. N.; Cohen, S.; Perrakis, A.; Noble, M. Acta Crystallogr., Sect. D
2004, 60, 2288.
65. MacroModel, Version 9.5, Schrödinger, LLC, New York, NY, 2007.
66. Gallicchio, E.; Zhang, L. Y.; Levy, R. M. J. Comput. Chem. 2002, 23, 517.
67. Guvench, O.; Shenkin, P.; Kolossvary, I.; Still, W. C. J. Comput. Chem. 2002, 23,
214.
23. Jorgensen, W. L.; Maxwell, D. S.; TiradoRives, J. J. Am. Chem. Soc. 1996, 118, 11225.
24. Klebe, G. Drug Discovery Today 2006, 11, 580.
25. Velec, H. F. G.; Gohlke, H.; Klebe, G. J. Med. Chem. 2005, 48, 6296.
26. Zeragraf, M.; Steuber, H.; Koch, C.; La Motta, C.; Sartini, S.; Sotriffer, C. A.; Klebe,
G. Angew. Chem., Int. Ed. 2007, 46, 3575.
27. 1,3-Dipolar Cycloaddition Chemistry; Padwa, A., Ed.; Wiley: New York, 1984.
28. Cycloaddition Reactions in Carbohydrate Chemistry; Giuliano, R. M., Ed.;
American Chemical Society: Washington, DC, 1992.
68. Still, W. C.; Tempczyk, A.; Hawley, R. C.; Hendrickson, T. J. Am. Chem. Soc. 1990,
112, 6127.
69. Chang, G.; Guida, W. C.; Still, W. C. J. Am. Chem. Soc. 1989, 111,
70. Kolossvary, I.; Guida, W. C. J. Am. Chem. Soc. 1996, 118, 5011.
4379.
71. MacroModel XCluster, Version 9.5, Schrödinger, LLC, New York, NY, 2005.
72. Becke, A. D. J. Chem. Phys. 1993, 98, 5648.
73. Lee, C. T.; Yang, W. T.; Parr, R. G. Phys. Rev. B 1988, 37, 785.
74. Stephens, P. J.; Devlin, F. J.; Chabalowski, C. F.; Frisch, M. J. J. Phys. Chem. 1994,
98, 11623.
75. Francl, M. M.; Pietro, W. J.; Hehre, W. J.; Binkley, J. S.; Gordon, M. S.; Defrees, D.
J.; Pople, J. A. J. Chem. Phys. 1982, 77, 3654.
29. Santoyo-Gonzalez, F.; Hernandez-Mateo, F. Top. Heterocycl. Chem. 2007, 7, 133.
30. Fernandez-Bolanos, J. G.; Lopez, O. Top. Heterocycl. Chem. 2007, 7, 31.
31. Gyorgydeak, Z.; Thiem, J.; Horton, D. Adv. Carbohydr. Chem. Biochem. 2006, 60,
103.
32. Pellissier, H. Tetrahedron 2007, 63, 3235.
76. Hariharan, P. C.; Pople, J. A. Theor. Chim. Acta 1973, 28, 213.
77. Hehre, W.; Ditchfield, R.; Pople, J. A. J. Chem. Phys. 1972, 56, 2257.
33. Li, X.; Wang, R.; Wang, Y.; Chen, H.; Li, Z.; Ba, C.; Zhang, J. Tetrahedron 2008, 64,
9911.