A multidisciplinary study of 3-(β-D-glucopyranosyl)-5-substituted-1,2,4-triazole derivatives as glycogen phosphorylase inhibitors: Computation, synthesis, crystallography and kinetics reveal new potent inhibitors

Article abstract of DOI:10.1016/j.ejmech.2018.01.095

3-(β-D-Glucopyranosyl)-5-substituted-1,2,4-triazoles have been revealed as an effective scaffold for the development of potent glycogen phosphorylase (GP) inhibitors but with the potency very sensitive to the nature of the alkyl/aryl 5-substituent (Kun et al., Eur. J. Med. Chem. 2014, 76, 567). For a training set of these ligands, quantum mechanics-polarized ligand docking (QM-PLD) demonstrated good potential to identify larger differences in potencies (predictive index PI = 0.82) and potent inhibitors with K_i's < 10 μM (AU-ROC = 0.86). Accordingly, in silico screening of 2335 new analogues exploiting the ZINC docking database was performed and nine predicted candidates selected for synthesis. The compounds were prepared in O-perbenzoylated forms by either ring transformation of 5-β-D-glucopyranosyl tetrazole by N-benzyl-arenecarboximidoyl chlorides, ring closure of C-(β-D-glucopyranosyl)formamidrazone with aroyl chlorides, or that of N-(β-D-glucopyranosylcarbonyl)arenethiocarboxamides by hydrazine, followed by deprotections. Kinetics experiments against rabbit muscle GPb (rmGPb) and human liver GPa (hlGPa) revealed five compounds as potent low μM inhibitors with three of these on the submicromolar range for rmGPa. X-ray crystallographic analysis sourced the potency to a combination of favorable interactions from the 1,2,4-triazole and suitable aryl substituents in the GP catalytic site. The compounds also revealed promising calculated pharmacokinetic profiles.

Full text of DOI:10.1016/j.ejmech.2018.01.095

Accepted Manuscript
A multidisciplinary study of 3-(β-D-glucopyranosyl)-5-substituted-1,2,4-triazole
derivatives as glycogen phosphorylase inhibitors: Computation, synthesis,
crystallography and kinetics reveal new potent inhibitors
Sándor Kun, Jaida Begum, Efthimios Kyriakis, Evgenia C.V. Stamati, Thomas A.
Barkas, Eszter Szennyes, Éva Bokor, Katalin E. Szabó, George A. Stravodimos,
Ádám Sipos, Tibor Docsa, Pál Gergely, Colin Moffatt, Myrto S. Patraskaki, Maria
C. Kokolaki, Alkistis Gkerdi, Vassiliki T. Skamnaki, Demetres D. Leonidas, László
Somsák, Joseph M. Hayes
PII:
S0223-5234(18)30117-X
10.1016/j.ejmech.2018.01.095
EJMECH 10172
DOI:
Reference:
To appear in: European Journal of Medicinal Chemistry
Received Date: 28 December 2017
Revised Date: 26 January 2018
Accepted Date: 30 January 2018
Please cite this article as: Sá. Kun, J. Begum, E. Kyriakis, E.C.V. Stamati, T.A. Barkas, E. Szennyes,
É. Bokor, K.E. Szabó, G.A. Stravodimos, Áá. Sipos, T. Docsa, Pá. Gergely, C. Moffatt, M.S. Patraskaki,
M.C. Kokolaki, A. Gkerdi, V.T. Skamnaki, D.D. Leonidas, Láó. Somsák, J.M. Hayes, A multidisciplinary
study of 3-(β-D-glucopyranosyl)-5-substituted-1,2,4-triazole derivatives as glycogen phosphorylase
inhibitors: Computation, synthesis, crystallography and kinetics reveal new potent inhibitors, European
Journal of Medicinal Chemistry (2018), doi: 10.1016/j.ejmech.2018.01.095.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

ACCEPTED MANUSCRIPT
A multidisciplinary study of 3-(β- -glucopyranosyl)-5-
D
substituted-1,2,4-triazole derivatives as glycogen
phosphorylase inhibitors:
computation, synthesis, crystallography and kinetics
reveal new potent inhibitors
a
#
b,c
#
d
#
d
Sándor Kun, Jaida Begum,
Efthimios Kyriakis, Evgenia C.V. Stamati, Thomas A.
d
a
a
a
d
Barkas, Eszter Szennyes, Éva Bokor, Katalin E. Szabó, George A. Stravodimos, Ádám
e
e
e
f
d
d
Sipos, Tibor Docsa, Pál Gergely, Colin Moffatt, Myrto S. Patraskaki, Maria C. Kokolaki,
d
d
d
a
Alkistis Gkerdi, Vassiliki T. Skamnaki, Demetres D. Leonidas,* László Somsák,* Joseph
b
M. Hayes*
a
Department of Organic Chemistry, University of Debrecen, POB 400, H-4002 Debrecen,
Hungary
b
School of Physical Sciences & Computing, Division of Chemistry, University of Central
Lancashire, Preston PR1 2HE, United Kingdom
c
School of Chemistry, University of Leeds, Leeds, LS2 9JT, United Kingdom
d
Department of Biochemistry & Biotechnology, University of Thessaly, Biopolis, 41500
Larissa, Greece
e
Department of Medical Chemistry, Faculty of Medicine, University of Debrecen, Egyetem
tér 1, H-4032 Debrecen, Hungary
f
Health and Life Sciences, De Montfort University, Gateway House, Leicester. LE1 9BH,
United Kingdom
*
Corresponding authors
tel: +302410565278 (DDL); fax: +302410565291; email: ddleonidas@bio.uth.gr
tel: +3652512900 ext 22348 (LS); fax: +3652512744; e-mail: somsak.laszlo@science.unideb.hu
tel: +441772894334 (JMH); fax: +441772894981; email: jhayes@uclan.ac.uk
#
Equal contribution
1

ACCEPTED MANUSCRIPT
Abstract: 3-(β-
D
-Glucopyranosyl)-5-substituted-1,2,4-triazoles have been revealed as an
effective scaffold for the development of potent glycogen phosphorylase (GP) inhibitors but
with the potency very sensitive to the nature of the alkyl/aryl 5-substituent (Kun et al., Eur. J.
Med. Chem. 2014, 76, 567). For a training set of these ligands, quantum mechanics-polarized
ligand docking (QM-PLD) demonstrated good potential to identify larger differences in
potencies (predictive index PI = 0.82) and potent inhibitors with K ’s < 10 µM (AU-ROC =
i
0
.86). Accordingly, in silico screening of 2335 new analogues exploiting the ZINC docking
database was performed and nine predicted candidates selected for synthesis. The compounds
were prepared in O-perbenzoylated forms by either ring transformation of 5-β-
D-
D-
D-
glucopyranosyl tetrazole by N-benzyl-arenecarboximidoyl chlorides, ring closure of C-(β-
glucopyranosyl)formamidrazone
with
aroyl
chlorides,
or
that
of
N-(β-
glucopyranosylcarbonyl)arenethiocarboxamides by hydrazine, followed by deprotections.
Kinetics experiments against rabbit muscle GPb (rmGPb) and human liver GPa (hlGPa)
revealed five compounds as potent low µM inhibitors with three of these on the
submicromolar range for rmGPa. X-ray crystallographic analysis sourced the potency to a
combination of favorable interactions from the 1,2,4-triazole and suitable aryl substituents in
the GP catalytic site. The compounds also revealed promising calculated pharmacokinetic
profiles.
Keywords: 1,2,4-triazole, C-β-D-glucopyranosyl derivatives, glycogen phosphorylase
inhibitors, QM/MM docking, kinetics, X-ray crystallography
2

ACCEPTED MANUSCRIPT
1
. Introduction
Type 2 diabetes (T2D) is a heterogeneous disease characterized by hyperglycemia. The
incidence of diabetes is on the rise globally with approximately 422 million people currently
affected and with ~90% of these cases T2D [1]. Current approaches to management of T2D
include modifications to diet, regular exercise and prescribed oral antihyperglyaemic drugs.
Adequate control of blood glucose levels is crucial to reduce the incidence of the long term
complications of T2D such as nephropathy, neuropathy and an increased risk of blindness and
cardiovascular disease [2]. However, current antihyperglycaemic drugs have somewhat
limited efficacy in this regard.
Glycogen phosphorylase (GP; EC 2.4.1.1) is a validated target for the development of new
T2D treatments having a direct influence on blood glucose levels through the glycogenolysis
pathway [3]. Structurally, GP exists as a homodimer consisting of two identical subunits, each
8
42 residues long. It is regulated allosterically by phosphorylation, has pyridoxal-5-phosphate
PLP) as co-factor, and exists in two interconvertible forms: the phosphorylated
predominantly active R state (GPa), and the unphosphorylated predominantly inactive T state
GPb). Seven different GP binding sites have been identified to date offering multiple
(
(
opportunities for modulation of enzymatic activity [4-6]: the catalytic, allosteric, new
allosteric, inhibitor, glycogen storage, benzimidazole [7] and quercetin binding site [8]. A
large number of synthetic and natural product GP inhibitors targeting the different binding
sites have been identified in recent years and solved structures of inhibitor bound GP
complexes have facilitated further structure based inhibitor design efforts [4, 5, 9]. GP
inhibitors have demonstrated considerable potential for T2D treatment in cellular models [10-
1
2] and in vivo [13-16], as well as promise against other conditions such as myocardial and
cerebral ischemias, and tumors [17-19].
3

ACCEPTED MANUSCRIPT
Of the different GP binding sites, the catalytic site has been explored most in terms of
inhibitor design efforts. The physiological inhibitor of GP is α-D-glucose (K = 1.7 mM) and
i
rational design of α and β-substitutions at the anomeric carbon have proved effective towards
increasing inhibitor potency [4, 5, 9, 20, 21]. β-Substitutions have proved particularly
effective by exploiting favorable interactions in the so-called β-cavity of the catalytic site, a
pocket lined by both polar and non-polar residues. Choice of the linker group which connects
the glucose moiety with different alkyl/aryl substituents is critical to inhibitor potency [4].
Among recently reported potent catalytic site inhibitors [22, 23], C-glucopyranosyl-1,2,4-
triazoles have been revealed as a novel scaffold (Table 1) for the design of potent GP
inhibitors [23, 24] with I and II exhibiting potencies of 0.41 and 0.67 µM, respectively. The
potencies, however, are very sensitive to the alkyl/aryl ligand substituent occupying the β-
cavity, with 50% of the compounds reported to date (X-XVIII) having K ’s > 100 µM to no
i
inhibition.
With this in mind, we report here the exploitation of the C-glucopyranosyl-1,2,4-triazole
skeleton (Table 1) for the design of new potent GP inhibitors. Using the previously reported
compounds with known inhibition constants (K ’s) as a training set (Table 1) [23, 24], we
i
have investigated different docking methods to reproduce the trends in binding affinities for
this congeneric series of compounds. Quantum mechanics/molecular mechanics (QM/MM)
docking in the form of quantum mechanics – polarised ligand docking (QM-PLD) with Glide
and QSite [25] was found to perform best as determined by a rigorous statistical analysis.
Then, exploiting the ZINC docking database (http://zinc.docking.org/) [26], different –R
groups (with a basic substructure phenyl) from commercially available acids, acid chlorides
and aldehydes led to the virtual screening of 2335 new C-glucopyranosyl-1,2,4-triazole
4

ACCEPTED MANUSCRIPT
analogues. Nine of the screened candidates were selected for synthesis and extensive kinetics
studies performed. Five of the nine candidates revealed potencies better than the threshold
employed for ’activity’ (K < 10 µM) with three compounds in the upper-nanomolar range for
i
rmGPa. These inhibitors are among the most potent catalytic site inhibitors discovered to
date. Crystallographic studies have been performed and these structural studies have revealed
the interactions that govern the observed potencies. Favorable pharmacokinetic profiles are
critical to the effectiveness of these inhibitors at the cellular level and in vivo. In this regard,
absorption, distribution, metabolism and excretion (ADME) predictions are additionally
reported and analyzed.
Table 1. Training set of 3-(β-D-glucopyranosyl)-5-substituted-1,2,4-triazoles with their
inhibition data towards rmGPb.[23]
N
OH HN
O
HO
R
HO
N
OH
R
K [µM]
i
R
K [µM]
i
I
0.41
X
-CH
2
OH
105
II
0.67
XI
111
CF3
NH2
CH3
III
IV
1.7
1.9
XII
-CH
3
499
OCH3
OCH3
XIII
518*
OCH3
OCH3
N
V
2.9
XIV
707
OH
5

ACCEPTED MANUSCRIPT
VI
VII
VIII
IX
7
XV
XVI
778
C(CH3)3
NH2
no inh.
at 625 µM
14
-C(CH )
3 3
NH2
33.5
39.7
XVII
XVIII
no inh.
at 625 µM
NO2
CH3
COOH
NO2
no inh.
at 625 µM
CH3
NO2
*
Calculated from the IC value by using a web-based tool.[27]
50
2
. Results & Discussion
2
.1 In Silico Selection of Synthetic Candidates
2
.1.1 Training Set Results
The training set of 3-(β-D-glucopyranosyl)-5-substituted-1,2,4-triazoles used in this study
represent a congeneric series of 18 compounds (Table 1) with a good spread of K values
i
ranging from 0.41 µM to no inhibition [23, 24]. ‘Activity’ was defined in terms of inhibitor
potency throughout. Performance of different docking methods, Glide-SP (standard-
precision), -XP (extra-precision) and QM-PLD with respect to the prediction of ligand
activities was analyzed in a statistical manner. The correlation between predicted and
experimental activities was measured using the Pearson (R ) and Spearman (R ) correlation
P
S
coefficients, which quantify the relative strength (and direction) of a relationship. Rp is
parametric and takes into account the absolute differences between experimental data (binding
6

ACCEPTED MANUSCRIPT
free energies) and predictions (docking score); R describes the non-parametric relationship
S
between the activity ranks of the ligands in the two datasets. Rp is a better measure for
absolute predictions (binding affinities), with R more appropriate for relative ranking[28, 29]
S
and therefore assumes priority over Rp in this work. Additionally, we considered the
‘predictive index’ (PI) [30]. This metric provides an important additional measure for the
relative ranking of ligand potencies by using a weight that depends on the difference between
the experimental activities of ligands. Therefore, the PI reflects that a good model should be
able to differentiate ligands which have larger differences in potencies, whereas if potencies
are similar, the weight will be low. PI values range from -1 (predictions always wrong) to +1
(predictions always right), with 0 corresponding to predictions that are completely random.
Finally, the area under the ROC (AU-ROC) curve metric was calculated for each docking
model representing the probability of active compounds being ranked earlier than inactive
compounds (range 0-1). For AU-ROC, we used a threshold of K < 10 µM to define activity,
i
leading to n = 6 actives in the set of N = 18 compounds. While a smaller dataset alters the
shape of the null distributions (Figure 1), it does not limit its application to studies of this
type, which has also included the CSAR benchmark exercises [29]. The statistical metrics
used in our analysis are described in more detail in the Supplementary Information.
7

ACCEPTED MANUSCRIPT
Figure 1. Comparison of AU-ROC null distributions for a set of n = 6 actives in a set of 18
ligands (this study) versus a representative set of n = 10 actives in a set of 1000 ligands used
in larger scale virtual screening. Distributions were created by random selection of ligands
and calculation of AU-ROC for each of these selections using the program R [31]. With a
smaller pool of ligands, random selection of the best or worst subset is more likely, hence the
wider and shorter distribution on the left. Statistically significant models, however, will still
be indicated by p-values < 0.05.
The results for statistical analysis of docking performance are shown in Table 2. Improvement
of results with increasing ‘accuracy’ of the docking algorithms was observed from Glide-SP
(
Rp = 0.59, R = 0.65, PI = 0.70) to –XP (Rp = 0.73, R = 0.73, PI = 0.77). A high AU-ROC
S
S
value of 0.89 for Glide-XP was obtained indicating the efficiency of the algorithm to
recognize inhibitors with K ’s less than the 10 µM threshold value. Superior performance of
i
QM/MM docking approaches has previously been reported [32-35], including for the GP
catalytic site [34]. QM-PLD involved reparametrization of the ligand atomic charges in the
field of the receptor using single point QM/MM calculations, with the ligand representing the
QM region and described using B3LYP/LACVP* [36-40]. Using these new electrostatic
potential (ESP) fit ligand charges, ligands were redocked using Glide-XP. For QM-PLD,
while the Rp (0.71) and AU-ROC (0.86) values were similar to those from Glide-XP, metrics
reflecting performance with respect to relative ranking of activities were improved for QM-
PLD (R = 0.77, P.I. = 0.82). A PI value of 0.82 is comparable to a PI value of 0.84 obtained
S
for a congeneric series of 16 p38 MAP kinase protein using computationally expensive
thermodynamic integration (TI) free energy simulations [30]. Overall, while QM-PLD did
produce reasonable correlation between predicted and experimental values for scores and
affinity, respectively (Pearson Rp = 0.71), and relative ranking of potencies (Spearman Rs =
0
.77), it demonstrated better potential to identify larger differences in potencies (predictive
index PI = 0.82) and potent inhibitors with K ’s < 10 µM (AU-ROC = 0.86). This suggested
i
the potential of QM-PLD to correctly propose potent 3-(β-D-glucopyranosyl)-5-substituted-
1
,2,4-triazole analogues at the screening stage. Probabilities (p-values) which reflect the
8

ACCEPTED MANUSCRIPT
robustness of the predictions were all << 0.05 (Table 2) and validate their statistical
significance.
Table 2. For the training set of eighteen 3-(β-
D
-glucopyranosyl)-5-substituted-1,2,4-triazoles (Table
1
), statistical analysis of the agreement between docking and experimental activities as described in
a
the text.
Pearson Correlation
Spearman Correlation
Method
RP
p-value
0.020
0.002
0.003
t-value
RS
p-value
0.008
0.003
0.001
t-value
PI
p-value
0.0029
0.0007
0.0002
AU-ROC
0.76
p-value
0.0352
0.0035
0.0064
Glide-SP
Glide-XP
QM-PLD
0.59
0.73
0.71
2.64
3.83
3.64
0.65
0.73
0.77
3.11
3.80
4.37
0.70
0.77
0.82
0.89
0.86
a
The best values for each metric are highlighted in bold. For the correlation coefficients R and R , p-values are
p
s
derived from the t-distribution (t-values) for n - 2 degrees of freedom (13).
2
.1.2 Screening Set Results
Based on the training set results, virtual screening of 2335 new 3-(β-D-glucopyranosyl)-5-
substituted-1,2,4-triazoles was performed using QM-PLD. As all training set ligands were
predicted to preferentially bind in one tautomeric state of the 1,2,4-triazole (tautomer shown
in Table 1) forming hydrogen bond interactions with His377 O, only this tautomeric state was
considered in the calculations. This favorable interaction was later confirmed in our
crystallographic studies (vide infra). Nine candidates (target compounds 10a-f and 10h-j as
per Scheme 1 and Table 3) were selected for synthesis based on the novelty of their chemical
structure, docking scores, predicted interactions with GPb, synthetic viability and commercial
availability of starting materials. 10g was not screened but was of academic interest.
2
.2 Syntheses
The target compounds were prepared by adapting our methods for the synthesis of C-
glucopyranosyl 1,2,4-triazoles (Scheme 1) [23, 24, 41-43]. Synthetic studies were started by
transformations of tetrazole 1 [44] since this method proved to be the most versatile one in
previous investigations (Route A) [23]. Thus, N-benzyl arenecarboxamides 5 were treated
9

ACCEPTED MANUSCRIPT
with SOCl to give imidoyl chlorides which, without isolation and purification, were reacted
2
with 1 in boiling xylene to result in the corresponding fully protected 1,2,4-triazole
derivatives 6aa, ba, and c-e. Removal of the benzyl protecting groups by catalytic
hydrogenation furnished 7a-d which were further deprotected to 10a-d by the Zemplén
protocol. Protective group cleavage of 6d and 6e was also effected in a reversed order to give
9
d and 9e under Zemplén conditions. Catalytic hydrogenolysis of 9e yielded 10e, however, N-
benzyl cleavage failed from 9d even at higher temperature and pressure in a sealed tube. In
order to get the sulfamoylated derivatives 10f and g, tosyl amidrazone 2 was acylated [24, 43]
by 4-sulfamoylbenzoyl chlorides to 7f and g which, on Zemplén transesterification, produced
1
0f and g, respectively (Route B). None of these methods proved suitable to prepare triazoles
h-j which could, however, be obtained via Route C [42]. Thus, anhydro-aldonic acid 3 was
7
converted to the corresponding acid chloride 4 [45] which was reacted with thioamides to
give the N-(β- -glucopyranosylcarbonyl)thioamides 8h-j. Ring closure of compounds 8 by
D
hydrazine hydrate furnished 1,2,4-triazoles 7h-j which were then O-debenzoylated by the
Zemplén method to the target compounds 10h-j.
Scheme 1. Synthesis of the target C-glucopyranosyl 1,2,4-triazoles 10.
1
0

ACCEPTED MANUSCRIPT
Route A
Route B
Route C
NH
N
N
OBz
O
OBz
O
N
OBz NH
O
O
BzO
BzO
BzO
BzO
BzO
BzO
NHNHTs
X
OBz
OBz
OBz
1
2
3 X = OH
ii
4
X = Cl
O
Cl
ii
i
iii
S
Bn
Ar
NHBn
iv
Ar
N
N
5
H N
2
Ar
N
N
OBz
O
OBz
O
OBz HN
O
O
S
v
vi
BzO
BzO
Ar
BzO
BzO
Ar
N
BzO
BzO
N
N
H
Ar
OBz
OBz
OBz
Bn
6
7
8
vii
vii
N
OH
HN
N
OH
N
viii
O
O
Ar
N
HO
HO
Ar
HO
HO
N
OH
OH
10
Bn
9
Reagents and conditions: i) anhydr. m-xylene at boiling temp.; ii) SOCl at boiling temp.; iii) 1.
2
ArCOCl, anhydr. CHCl , pyridine 0 °C → rt, 2. Bu NF, THF at boiling temp.; iv) anhydr. CH CN,
3
4
3
pyridine, rt; v) H (1 atm), Pd(C), anhydr. EtOAc or THF, both at boiling temp.; vi) NH NH ·H O,
2
2
2
2
pyridine, rt; vii) cat. NaOMe, anhydr. MeOH, CHCl , rt; viii) H (1 atm), Pd(C), anhydr. MeOH,
3
2
EtOAc at boiling temp.
Table 3: Structure of the aromatic moieties (Ar) for the compounds in Scheme 1 and the
corresponding yields of the reactions.
Isolated yield
Ar
Route
6
7
8
9
10
aa
a
A
A
63
-
-
-
-
-
62
-
-
-
-
-
76
-
ba
b
A
A
64
-
77
-
-
77
c
A
A
37
64
86
60
-
-
-
52
d
97
94 (from 7)
e
f
A
B
46
-
-
-
-
81
-
81
75
47
1
1

ACCEPTED MANUSCRIPT
g
B
C
-
-
42
70
-
-
-
56
90
h
71
i
C
C
-
-
65
74
58
81
-
-
70
90
j
2
.3 Enzyme Kinetics
The inhibition constant value (K ) of the synthesized compounds for human liver glycogen
i
phosphorylase a (hlGPa) together with the values for rmGPa (to compare the effect of the
compounds in the liver and muscle) and rmGPb (to validate structural data) are summarized
in Table 4. As the data shows for rmGPb, inhibitors 10a-d are almost equipotent (K ’s 1.19 –
i
5
.05 µM) while 10h is somewhat less potent (K = 11.50 µM) and close to our threshold value
i
for ‘activity’ (K < 10 µM). Compounds 10e, 10f and 10i are moderate inhibitors (K ’s 20 –
i
i
9
8.2 µM) while 10j and 10g do not show any significant inhibition. However, there were
some solubility issues for 10j, while 10g was not one of our predicted compound. From the Ki
values it seems that completely non-polar substituents like fluorene (10c) or biphenyl (10d)
increase the potency more than more polar ones like carboxy-biphenyl (10a) and carboxy-
naphthyl (10b), or at least the effect of the polar carboxylate groups on potency is neutral. Of
significant interest is the fact that the change of the position of the second phenyl ring at the
biphenyl substituent from the meta position (10e) to the para (10d) leads to an approximately
2
8-fold increase in potency. Compounds 10a (K = 0.98 µM), 10b (K = 0.78 µM) and 10c (K
i i i
=
0.84 µM) show a preference for rmGPa with sub-micromolar activity observed, while 10d
displays similar potency for all three enzymes.
1
2

ACCEPTED MANUSCRIPT
Table 4. Inhibition of glycogen phosphorylases by the synthesized compounds and crystallographic numbering.
Ki (µM)
Code
Compounds
rmGPb
rmGPa
hlGPa
1
1
1
0a
0b
0c
4.42 ± 0.17
0.98 ± 0.04 2.67 ± 0.12
0.78 ± 0.05 3.88 ± 0.13
0.84 ± 0.07 2.02 ± 0.33
2.09 ± 0.18 2.23 ± 0.08
5.05 ± 0.40
1.19 ± 0.13
2.38 ± 0.19
1
1
0d
0e
a
b
b
68.2
N. m.
N. m.
a
b
b
1
1
0f
98.2
N. m.
N. m.
N. m.
N. m.
b
b
0g
no inh. at 625 µM
1
1
1
0h
0i
11.50 ± 0.23
3.38 ± 0.28 8.91 ± 0.44
b
b
20 ± 0.95
>1.5 mM
N. m.
N. m.
N. m.
N. m.
b
b
0j
a
Calculated from the IC value by using a web-based tool [27].
5
0
b
Not measured.
2
.4 X-ray Crystallography
1
3

ACCEPTED MANUSCRIPT
The rmGPb-ligand complexes of the most potent inhibitors from the kinetics studies (10a-d,
1
0h) were determined using X-ray crystallography so as to decipher the interactions
responsible for their potency. All compounds show similar potency for hlGPa and rmGPb
and, the active sites in hlGP and rmGP are identical in terms of sequence and structure
therefore conclusions based on structural data obtained with rmGPb are directly applicable to
hlGPa.
The 2Fo-Fc and Fo-Fc electron density maps revealed that all five inhibitors were bound at
the active site and clearly defined the position of each atom of the ligands (Figure 2). The
superposition of the structures of free rmGPb and the rmGPb-inhibitor complexes over well-
defined residues (18– 49, 262–312, 326–829) gave r.m.s.d. values of 0.15, 0.20, 0.17, 0.16,
and 0.19 Å for the 10a, 10b, 10c, 10d, and 10h, complexes, respectively, indicating that the
binding of the inhibitors at the catalytic site did not trigger any major conformational change
on the overall protein structure.
1
4

ACCEPTED MANUSCRIPT
Figure 2. The REFMAC weighted 2Fo-Fc electron density maps of the bound ligands at the
catalytic site contoured at 1.0 σ before the incorporation of the ligand molecules in the
refinement process. The final models of the inhibitors are also shown.
Each of the inhibitors binds at the catalytic site by anchoring the glucose moiety at a location
previously observed for α-D-glucose and other glucose-based inhibitors [20, 46-50]. There the
glucose moiety engages in hydrogen bonding (Table 5) and van der Waals interactions almost
identical to those that have been previously observed for similar glucose analogues.
Furthermore, three conserved water molecules mediate hydrogen bond interactions between
the glucopyranose moiety of each ligand and residues Asp283, Tyr573, Lys574, Thr671,
Ala673, Thr676, and the phosphate group of the cofactor PLP.
Table 5: Potential hydrogen bond interactions of inhibitors with rmGPb residues at the catalytic
site in the crystal. Hydrogen bonds are reported if the distance is less than 3.4 Å and the angle at
a
the donor group is above 100°. Numbers shown are distances in Å.
Inhibitor atom
Protein atoms
Asn284 (ND2)
Try573 (OH)
Glu672 (OE2)
Water (O)
10a
3.1
3.0
3.0
2.9
2.8
2.6
3.3
3.1
3.1
2.9
2.6
2.7
2.8
2.7
3.0
3.2
3.1
2.7
10b
3.1
3.1
3.2
3.0
2.8
2.6
3.2
3.1
3.2
2.9
2.6
2.7
2.8
2.7
3.0
-
10c
3.1
3.0
3.2
2.9
2.8
2.6
3.2
3.1
3.2
2.8
2.6
2.7
2.8
2.7
2.9
-
10d
3.0
3.1
3.2
3.0
2.8
2.7
3.2
3.1
3.2
2.8
2.6
2.7
2.7
2.7
2.9
-
10h
2.6
3.1
3.3
2.8
2.8
2.7
3.3
3.1
3.3
2.8
2.6
2.7
2.8
2.9
2.8
-
O2’
Water (O)
Glu672 (OE2)
Ala673 (N)
Ser674 (N)
Gly675 (N)
Gly675 (N)
Water (O)
O3’
O4’
O6’
His377 (ND1)
Asn484 (ND2)
His377 (O)
Water (O)
N2
N5
Glu287 (N)
Gly288 (N)
Water (O)
O17
-
-
-
-
-
-
-
-
1
5

ACCEPTED MANUSCRIPT
Water (O)
Water (O)
3.3
-
-
-
-
-
-
-
-
2.7
3.0
3.3
2.8
-
Water (O)
-
-
-
-
Asn282 (ND2)
Water (O)
3.1
-
-
-
-
O18
-
-
-
Water (O)
2.8
21
-
-
-
Total
19
15
15
15
a
Atom numbering is as displayed in Table 4.
The triazole linker participates in a hydrogen bond interaction with the main chain oxygen of
His377 (Table 4), as was predicted by the modelling, and in water-mediated hydrogen bond
interactions with the main chain nitrogen atoms of Gly135 and Leu136 through a conserved
water molecule in all rmGPb ligand complexes. Similar interactions have been previously
observed with other glucose derived inhibitors bound to rmGPb [46, 51], and the increased
inhibitory potency of the 1,2,4-triazole compounds was ascribed to the hydrogen-bond
forming capacity of the heterocyclic nitrogen atoms to the main chain oxygen of His377 and
to water-mediated hydrogen bonding interactions with the sidechain atoms of Asp283 and the
main chain atoms of Leu136.
In total, 10a, 10b, 10c, 10d, and 10h engage in 116, 92, 90, 92, and 81 van der Waals
interactions with protein residues at the active site of rmGPb, respectively (Figure 3). The
phenyl ring A (c.f. Table 4) in 10a, 10b, 10c, and 10d is involved in van der Waals
interactions with protein residues Glu88 and Asn284 while phenyl ring B (c.f. Table 4) is
involved in van der Waals interactions with Asn282, Phe285, Phe286, and Arg292. These
interactions have been also observed with other 1,2,4-triazole inhibitors and they were the
source of the significant potency displayed by them [46, 51]. In addition the naphthyl ring of
1
0h engages in van der Waals interactions with Glu88, Leu136, Asn283, and Ala383.
1
6

ACCEPTED MANUSCRIPT
The carboxyl group of 10a is involved in hydrogen bond interactions with the main chain
atoms of Glu287 and Gly288 and the side chain of Asn282, while it also participates in water-
mediated interactions with Lys289, Arg292, and Glu385 (Figure 3). The carboxyl group of
1
0b participates in hydrogen bond interactions with the side chain of Asn282 and water-
mediated interactions with Tyr280, Glu287, Gly288, Glu296, Arg292, and Glu385 (Figure 3).
All inhibitors except 10h bind similar at the active site and do not seem to cause any
significant conformational change. The binding of 10h triggers a significant shift of the 280s
loop (residues 282-289). The r.m.s. distance for all atoms of these residues between the
rmGPb-10c and the rmGPb-10h complex structures is 1.0 Å with Asn282, Asn284, Phe285
and Glu287 being the residues with the greatest difference (Figure 4). This conformational
change may arise from the different orientation of the naphthyl ring in 10h with respect to the
other four inhibitors studied here. The energy cost associated with this conformational change
may also offer an explanation for the lower potency displayed by this compound in
comparison to the potency of the other four inhibitors.
1
7

ACCEPTED MANUSCRIPT
Figure 3. Diagrams of the binding of 10a (a), 10b (b), 10c (c), 10d (d), and 10h (e) at the active site of
rmGPb. Van der Waals interacting protein atoms are shown bigger and in meshed spheres.
1
8

ACCEPTED MANUSCRIPT
Figure 4. Superposition of the rmGPb-10h complex structure (maroon) and the free structure (grey).
The small differences in the inhibitory potency between inhibitors bearing a carboxyl group
(10a and 10b) and the unsubstituted non-polar ligands 10c and 10d could be attributed to the
interactions of the carboxylates. Bulk desolvation effects will be associated with the
negatively charged carboxylate going from the free unbound state to the bound state in the
catalytic site. Additionally, the carboxyl groups upon binding hydrogen-bond a water
molecule placing it in close proximity (~3.3 Å) to the phenyl ring of Tyr280 (10a complex) or
Phe286 (~3.7 Å; 10b complex), constituting unfavorable interactions. These unfavorable
interactions are, however, counterbalanced by the hydrogen bond interactions of the carboxyl
groups (O17 and O18, Table 5). In the 10c and 10d complexes no extra waters are brought
into the catalytic site since they do not have any polar groups. Comparing 10c and 10d, their
aromatic groups form 29 and 32 van der Waals interactions, respectively, at the rmGPb active
site, indicating that they do not have significant differences in their protein interactions and
hence they are almost equipotent. The better potency of inhibitor 10a with respect to 10b for
rmGPb (and hlGPa) could be attributed to two additional hydrogen bond interactions of one
of its carboxyl oxygen atoms with the main chain amides of Glu287 and Gly288 (Table 5).
2
.5 Pharmacokinetics Predictions
1
9

ACCEPTED MANUSCRIPT
Early monitoring of the pharmacokinetic profiles of lead compounds is recommended to help
avoid potential for failure in later stage drug development trials. The absorption, distribution,
metabolism and excretion (ADME) properties of our synthesized compounds were calculated
using QikProp 3.5 [25], the results of which are shown in Table 6. An orally active drug
should have no more than one violation of Lipinski’s ‘rule of five’ [52] while for Jorgensen’s
‘rule of three’ [53, 54] more drug-like molecules have fewer violations. All inhibitors except
the least potent sulfamoyl derivative (10f) have just 0 or 1 violation of Lipinski’s ‘rule of
five’. Considering Jorgensen’s ‘rule of three’, there was a maximum of one violation for the
-1
ligands. While the Caco-2 cell permeability (> 22 nm s ) appears adequate for most ligands
-1
(
~ 60 nm s ), it is flagged for 10a, 10b and the sulfamoyl derivative 10f; the sensitive
lipophilicity/solubility balance of β-D-glucose analogues with heterocyclic linkers has
previously been highlighted [55, 56]. Polar surface areas (PSAs), however, are within the
range of 95% of known drugs (10f excepted) although they are close to or above (10a and
2
1
0b) Veber at al.’s [57] recommended value of < 140 Å for oral bioavailability. With respect
to the carboxylic acid moiety of 10a and 10b, a wide variety of endogenous substances such
as amino acids and triglycerides possess this group and it is often part of the pharmacophore
of diverse classes of therapeutic agents with more than 450 of carboxylic acid-containing
drugs currently marketed worldwide [58]. A diminished ability to passively diffuse across
biological membranes is more relevant in the context of central nervous system (CNS) drug
discovery [58]. All other properties (Table 6) are satisfactory for the studied ligands. The
predicted degree of binding to human serum albumin (log K ) which affects bioavailability
hsa
is -1.15 – -0.50 and within the range for 95% of known drugs (-1.5 – 1.5), while the log BB
values (-3.6 – -2.1) predict that the inhibitors are unlikely to cross the blood-brain barrier.
3
. Conclusions
2
0

ACCEPTED MANUSCRIPT
Docking is widely used to distinguish active from inactives chemotypes in virtual screening
studies. Here, GP inhibitor screening in the form of QM/MM docking (QM-PLD) has been
applied to identify novel 3-(β-D-glucopyranosyl)-5-substituted-1,2,4-triazole derivatives. The
advantage of a computationally driven (QM/MM docking) approach to candidate selection is
highlighted, with eight out of the nine selected compounds selected for synthesis revealing
either moderate or strong activity (K ’s < 100 µM) compared to only nine of eighteen in
i
previous studies (Table 1). Five of the nine candidates had K ’s < 10 µM, our defined
i
threshold for ‘activity’, with 10a-d and 10h potent low µM inhibitors (rmGPa, hlGPa) and
1
0a-c on the high nM range for rmGPa. X-ray crystallography studies of the protein-ligand
complexes revealed that the addition of polar groups such as carboxyl to the bulky aromatic
ligand substituents in the β-cavity (although it offers additional hydrogen bond interactions
with protein residues) does not lead to improved potency since this group also attracts water
molecules that are placed in close proximity of hydrophobic residues. Structural studies
indicate that improvement of the potency might otherwise arise by the replacement of the
carboxyl groups with another non-polar chemical entity. Carbohydrate-based compounds and
triazoles have proven to be important classes of compounds in the treatment or development
of new clinical candidates for a number of conditions [59, 60]. The 3-(β-D-glucopyranosyl)-5-
substituted-1,2,4-triazoles studied here are predicted to have drug-like potential with only
permeability flagged as a potential issue to efficacy. However, the compounds have in general
(for six of the nine synthesized) better predicted Caco-2 cell permeabilities (by a factor of 3)
than previously reported glucose analogues [10, 55, 56] and accordingly are to be considered
in further follow up experiments (cellular and in vivo) for efficacy.
2
1

ACCEPTED MANUSCRIPT
Table 6. Results of ADME property predictions for the different tautomers of the 3-(β-D-glucopyranosyl)-5-substituted-1,2,4-triazole inhibitors studied in this work
[b]
[b]
2 [c]
PSA [Å ]
[d]
[e]
Inhibitor
Lipinskis’s Rule of Five and Violations (V)
Jorgensen’s Rule of Three and Violations (V)
log Khsa
log BB
[f]
HBD
[g]
HBA
-1 [h]
Caco-2 [nm s ]
[i]
M [Da]
r
log P(o/w)
V
log S
NMP
V
2
(<140 Å )
(
<500)
(≤5)
6
(≤10)
10
10
8
(<5)
(>22)
1.6*
1.6*
61.9
61.7
61.8
2.9*
57.1
61.8
59.7
(>-5.7)
(<7)
6
1
0a
0b
0c
4
27.4
01.4
95.4
83.4
83.4
86.4
57.4
07.4
31.4
-0.2
-0.7
0.6
1
1
0
0
0
2
0
0
0
-3.9
-3.4
-4.2
-3.9
-3.9
-2.6
-3.4
-4.3
-4.5
1
1
1
0
0
1
0
1
1
185.1
185.0
135.6
135.6
135.6
201.0*
137.0
135.6
137.0
-0.85
-0.94
-0.43
-0.50
-0.50
-1.15
-0.61
-0.38
-0.30
-3.5*
-3.3*
-2.2
1
4
6
6
1
3
3
3
3
3
4
4
5
7
1
0d
0e
5
8
0.5
6
-2.2
1
5
8
0.5
6
-2.2
1
0f
7*
5
11
8
-2.8*
-0.1
0.8
6
-3.6*
-2.1
1
0h
6
1
0i
0j
5
8
7
-2.2
1
5
8
1.0
7
-2.2
[j]
Range
130-725
0-6
2-20
-2.0-6.5
-
<25 poor;
00 great
>
-6.5-
0.5
1-8
-
7-200
-1.5-1.5
-3.0-1.2
5
[a] ADME data were calculated as described in the text using Qikprop 3.5; predicted properties outside the range for 95% of known drugs are indicated with an asterisk (*). [b]
Rules as listed in the columns, with any violations of the rules highlighted in italics. [c] PSA represents the van der Waals (polar) surface areas of N and O atoms; recommended
2
PSA < 140 Å according to Veber et al. [57]. [d] log Khsa: predicted binding to human serum albumin. [e] log BB: the predicted blood-brain barrier coefficient. [f] Number of
hydrogen bond donors. [g] Number of hydrogen bond acceptors. [h] Caco-2 cell permeability. [i] Number of primary metabolites. [j] Range for 95% of known drugs [61].
22

ACCEPTED MANUSCRIPT
4
. Experimental
4
.1 Computational details
4
.1.1 Protein Preparation
The initial setup of the GPb receptor for docking was performed using Schrodinger’s “Protein
Preparation Wizard” [25] and the solved co-crystallized complex with N-(2-naphthyl)-N´-(β-
D
-glucopyranosyl)urea [5]. Water molecules within 5 Å of the native ligand were initially
retained but deleted for subsequent docking. Bond orders were assigned and hydrogen atoms
added, with protonation states for basic and acidic residues based on residue pK values at
a
normal pH (7.0) calculated using PROPKA [62]. Subsequent optimization of hydroxyl
groups, histidine protonation states and C/N atom “flips”, and side-chain O/N atom flips of
Asn and Gln was based on optimizing hydrogen bonding patterns. The phosphate in
pyridoxal-phosphate (PLP) was assigned in monoanionic form. Finally, an “Impref”
minimization of the GPb complex was performed using the OPLS-AA(2005) force field [63]
to remove steric clashes and bad contacts but with heavy atoms constrained to within 0.3 Å
(RMSD) of their crystallographic positions.
4
.1.2 Ligand Preparation
Training and screening set ligands were prepared using a combination of Maestro, LigPrep
.5 and CombiGlide 3.8 [25]. For creation of the screening set database ligands consisting of
diverse –R groups, “purchasable” acids (R-COOH), acid chlorides (R-COCl) and aldehydes
R-CHO) were downloaded from the ZINC docking database [26]. The following criteria
were used as filters so that the final proposed ligands contained aryl substituents and were
drug-like’:[52, 57] MWs ≤ 200, 224, and 240 Da with benzoyl chloride, benzoic acid and
benzaldehyde used as substructures, respectively; 0 ≤ rotatable bonds ≤ 4. Using CombiGlide,
the resulting reagents were ‘reacted’ with the 3-(β- -glucopyranosyl)-1,2,4-triazole ‘core’ to
3
(
‘
D
2
3

ACCEPTED MANUSCRIPT
yield the final 3-(β-
D
-glucopyranosyl)-5-substituted-1,2,4-triazoles. 2335 distinct ligands
constituted the screening set. LigPrep 3.5 generation of tautomers and favorable ionization
states at pH = 7 ± 2 of the ligands resulted in three distinct tautomers of the triazole moiety (H
on each of three N atoms); however, with only one triazole tautomer favored for the binding
of all training set ligands to GPb (H on N2), only this tautomeric state was considered for
docking of the screening set ligands.
4
.1.3 Docking Calculations
Training Set Ligands: Flexible-ligand docking calculations were performed using Glide 6.8 in
both standard- (SP) and extra-precision (XP) modes, as well as with QM-PLD [25, 64]. In the
docking calculations with Glide, the shape and properties of the GPb catalytic site were first
mapped onto grids with dimensions of 26.7 × 26.7 × 26.7 Å centred on the native co-
crystallized ligand. Positional constraints on the four glucose -OH oxygen atoms were applied
(radius 1.5 Å) [65]. Standard parameters were otherwise applied for the Glide-SP and XP
docking calculations. Post-docking minimization of the ligand poses was performed (with
strain correction) and a maximum of 3 poses per ligand saved. Poses were ranked by ’docking
score’. For the QM-PLD calculations, atomic partial charges of the output docking poses from
Glide XP docking were reparametrized in the ‘field’ of the receptor (ESP fit charges) using
QSite 6.8 [25] and single point energy QM/MM calculations. The ligands were then redocked
using Glide-XP.
Virtual Screening: Docking of the screening set ligands was performed using QM-PLD and
with the same settings as for the training set ligands described above.
4
.1.4 ADME Property Predictions
QikProp from Schrödinger [25] was used for the ADME property predictions. Specifically,
Lipinski’s rules of five [52] (MW < 500, log P < 5, HBA ≤ 10, and HBD ≤ 5) and Jorgensen’s
2
4

ACCEPTED MANUSCRIPT
rule of three [53, 54] (log S > -5.7, Caco-2 permeability > 22 nm/s, number of primary
metabolites < 7) were applied to predict oral bioavailability. Calculations were performed in
normal mode and are reported for the 1,2,4-triazole binding most stable tautomer (Table 1),
with test calculations here and from previous studies [56] on β-D-glucopyranosyl 1,2,4-
triazole derivatives showing that the tautomeric state of the triazole has limited effect on the
property predictions.
4
4
.2 Synthesis
.2.1 General methods
Anhydrous toluene, m-xylene, CHCl , EtOAc, CH CN (P O ), THF (Na, benzophenone),
3
3
2
5
pyridine (KOH), MeOH (Mg) were prepared by distillation from the indicated drying agent.
Reagents were purchased from commercial suppliers (Sigma-Aldrich, TCI, Alfa Aesar) and
used without further purification. TLC was carried out on precoated plates (DC-Alurolle
Kieselgel 60, F , Merck), and the spots were visualized under UV light and by gentle
2
54
heating. Column chromatography was performed on silica gel (Kieselgel 60, 63-200 µm,
1
13
Molar Chemicals). H, C NMR spectra were recorded on Bruker DRX 360 and Bruker DRX
1
13
4
00 spectrometers. TMS ( H) or resonance peak of the solvent ( C) was used as reference.
Mass spectra were obtained by a Bruker micrOTOF-Q instrument. Microanalyses were
performed on an Elementar Vario Micro Cube. Optical rotations were determined with a
Perkin-Elmer 241 polarimeter at rt.
4
.2.2 General procedure I for the synthesis of 4-benzyl-3-(2,3,4,6-tetra-O-benzoyl-β-
D-
glucopyranosyl)-5-substituted-1,2,4-triazoles (6)
The solution of an N-benzyl-arenecarboxamide (5, 2 equiv.) in SOCl (5 mL/mmol) was
2
heated at boiling temp. for 2 h then the excess of SOCl was removed under diminished
2
2
5

ACCEPTED MANUSCRIPT
pressure. The residue was dissolved in anhydrous toluene and evaporated in order to remove
the traces of the reagent. 5-(2,3,4,6-Tetra-O-benzoyl-β-D-glucopyranosyl)tetrazole (1, 1
equiv.) in anhydrous m-xylene (15 mL/mmol) was added and the mixture was refluxed until
disappearance of the tetrazole (TLC, 4:1 toluene-AcOH). After evaporation of the solvent the
residue was purified by column chromatography.
4
.2.3 General procedure II for the synthesis of 3-(2,3,4,6-tetra-O-benzoyl-β-D-
1
glucopyranosyl)-5-substituted-1,2,4-triazoles (7) from N -tosyl-C-(2,3,4,6-tetra-O-
benzoyl-β-
D
-glucopyranosyl)formamidrazone (2)
-glucopyranosyl)formamidrazone (2) was dissolved
1
N -Tosyl-C-(2,3,4,6-tetra-O-benzoyl-β-
D
in anhydrous CHCl (15 mL/mmol) and anhydrous pyridine (3.0 equiv.) was added. The
3
mixture was cooled in an ice bath, and the solution of an aroyl chloride (3.0 equiv.) in
anhydrous CHCl (5 mL/mmol) was added dropwise over 15 minutes. Subsequently, the
3
mixture was stirred at rt and monitored by TLC (3:1 EtOAc-hexane). After total consumption
of the starting material (2 days) the mixture was diluted with CHCl and extracted with water.
3
The organic phase was dried over MgSO and concentrated under diminished pressure. The
4
obtained residue was dissolved in THF (20 mL/mmol) and TBAF (1 M solution in THF, 2.0
equiv.) was added. The reaction mixture was boiled for 24 h and then it was concentrated and
purified by column chromatography.
4
.2.4 General procedure III for the preparation of N-(2,3,4,6-tetra-O-benzoyl-β-D-
glucopyranosylcarbonyl)thioamides (8)
A mixture of C-(2,3,4,6-tetra-O-benzoyl-β-D-glucopyranosyl)formic acid (3) and SOCl (10
2
mL/mmol) was boiled under reflux for two hours, then the excess of the reagent was removed
under diminished pressure. Last traces of SOCl were removed by co-evaporations with dry
2
2
6

ACCEPTED MANUSCRIPT
toluene. The resulting acid chloride (4) was dissolved in anhydrous CH CN (15 mL/mmol),
3
and a solution or a suspension of the corresponding thioamide (1.2 equiv.) and pyridine (1.2
equiv.) in anhydrous CH CN (30-120 mL/mmol depending on the solubility of the thioamide)
3
was added dropwise over 20 minutes. The reaction mixture was stirred at rt, and monitored by
TLC (1:1 hexane-EtOAc). After complete conversion the solvent was removed and the
residue was purified by column chromatography.
4
.2.5 General procedure IV for the preparation of 3-(2,3,4,6-tetra-O-benzoyl-β-
D
-
-
glucopyranosyl)-5-substituted-1,2,4-triazoles (7) from N-(2,3,4,6-tetra-O-benzoyl-β-
D
glucopyranosylcarbonyl)thioamides (8)
To a solution of an N-(2,3,4,6-tetra-O-benzoyl-β-D-glucopyranosylcarbonyl)thioamide (8) in
anhydrous pyridine (10 mL/mmol) hydrazine monohydrate (1.2 equiv.) was added, and the
reaction mixture was stirred at rt. After disappearance of the starting material (monitored by
TLC, 1:1 hexane-EtOAc) the solvent was removed and the residue was purified by column
chromatography.
N-Benzyl and O-benzoyl protecting groups were removed by standard methods to give the
test compounds 10. Detailed procedures for deprotections as well as compound
characterization are described in the Supporting Information.
4
.3 Kinetics
Rabbit muscle GPb (rmGPb) was purified from rabbit skeletal muscle following the protocol
developed by Fischer & Krebs [66] with slight modifications [67] while rmGPa was produced
by phosphorylation of rmGPb [51]. Human liver GPa (hlGPa) was produced following a
previously described protocol [51]. Kinetic studies were performed at 30 °C in the direction
of glycogen synthesis using 3 µg/mL of the rabbit muscle enzymes, or 1 µg/mL hlGPa in a 30
2
7

ACCEPTED MANUSCRIPT
mM imidazole/HCl buffer (pH 6.8) containing 60 mM KCl, 0.6 mM EDTA, and 0.6 mM
dithiothreitol, using constant concentrations of glycogen (0.2% w/v), AMP (1 mM; only for
the rmGPb experiments), and various concentrations of Glc-1-P and inhibitors. Initial
velocities were calculated from the pseudo-first order rate constants measuring the release of
orthophosphate ions spectrophotometrically [68] at five time-intervals. For the calculation and
statistical evaluation of the kinetic parameters, the non-linear regression program Grafit [69]
was employed.
4
.4 X-ray Crystallography
rmGPb-inhibitor complexes were formed by diffusion of 10 mM solution of the inhibitors in
the crystallization media supplemented with DMSO (15 %, v/v) in preformed rmGPb crystals
[
70] at room temperature for 3 hours prior to data collection. X-ray diffraction data were
collected using synchrotron radiation on station I04 (λ=0.9795 Å) at Diamond Synchrotron
Radiation Source in Oxford, U.K. at room temperature on a Pilatus 6M detector. To avoid
crystal radiation damage the crystal was translated five times. Crystal orientation, integration
of reflections, inter-frame scaling, partial reflection summation, and data reduction was
performed by the program XDS [71]. Scaling and merging of intensities were performed by
1
/2
Aimless [72] and the optimum resolution was selected based on the CC criterion [73].
Crystallographic refinement of the complexes was performed by maximum-likelihood
methods using REFMAC [74] with starting model the structure of the native T state rmGPb
complex determined at 1.9 Å resolution (Leonidas et al, unpublished results). Ligand
molecule coordinates and topologies were constructed using JLigand [75] and they were fitted
to the electron density maps by adjusting of their torsion angles. A summary of the data
processing and refinement statistics for the inhibitor complex structures is given in Table 1.
The refinement was validated by the PDB_REDO server [76]. As there were more than 5
reflections per atom available, both an isotropic and an anisotropic B-factor model were
2
8