First dual AK/GSK-3β inhibitors endowed with antioxidant properties as multifunctional, potential neuroprotective agents

Article abstract of DOI:10.1016/j.ejmech.2017.06.017

The manuscript deals with the design, synthesis and biological evaluation of novel benzoxazinone-based and indole-based compounds as multifunctional neuroprotective agents. These compounds inhibit human adenosine kinase (hAK) and human glycogen synthase kinase 3 beta (hGSK-3β) enzymes. Computational analysis based on a molecular docking approach underlined the potential structural requirements for simultaneously targeting both proteins’ allosteric sites. In silico hints drove the synthesis of appropriately decorated benzoxazinones and indoles (5a-s, and 6a-c) and biochemical analysis revealed their behavior as allosteric inhibitors of hGSK-3β. For both our hit 4 and the best compounds of the series (5c,l and 6b) the potential antioxidant profile was assessed in human neuroblastoma cell lines (IMR 32, undifferentiated and neuronal differentiated), by evaluating the protective effect of selected compounds against H₂O₂ cytotoxicity and reactive oxygen species (ROS) production. Results showed a strong efficacy of the tested compounds, even at the lower doses, in counteracting the induced oxidative stress (50 μM of H₂O₂) and in preventing ROS formation. In addition, the tested compounds did not show any cytotoxic effect determined by the LDH release, at the concentration range analyzed (from 0.1 to 50 μM). This study allowed the identification of compound 5l, as the first dual hAK/hGSK-3β inhibitor reported to date. Compound 5l, which behaves as an effective antioxidant, holds promise for the development of new series of potential therapeutic agents for the treatment of neurodegenerative diseases characterized by an innovative pharmacological profile.

Full text of DOI:10.1016/j.ejmech.2017.06.017

Accepted Manuscript
First dual AK/GSK-3β inhibitors endowed with antioxidant properties as
multifunctional, potential neuroprotective agents
Simone Brogi, Anna Ramunno, Lida Savi, Giulia Chemi, Gloria Alfano, Alessandra
Pecorelli, Erika Pambianchi, Paola Galatello, Giulia Compagnoni, Federico Focher,
Giuseppe Biamonti, Giuseppe Valacchi, Stefania Butini, Sandra Gemma, Giuseppe
Campiani, Margherita Brindisi
PII:
S0223-5234(17)30463-4
10.1016/j.ejmech.2017.06.017
EJMECH 9512
DOI:
Reference:
To appear in: European Journal of Medicinal Chemistry
Received Date: 10 March 2017
Revised Date: 7 June 2017
Accepted Date: 8 June 2017
Please cite this article as: S. Brogi, A. Ramunno, L. Savi, G. Chemi, G. Alfano, A. Pecorelli, E.
Pambianchi, P. Galatello, G. Compagnoni, F. Focher, G. Biamonti, G. Valacchi, S. Butini, S. Gemma,
G. Campiani, M. Brindisi, First dual AK/GSK-3β inhibitors endowed with antioxidant properties as
multifunctional, potential neuroprotective agents, European Journal of Medicinal Chemistry (2017), doi:
10.1016/j.ejmech.2017.06.017.
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
Graphical Abstract

ACCEPTED MANUSCRIPT
1
First Dual AK/GSK-3β Inhibitors Endowed with
Antioxidant Properties as Multifunctional, Potential
Neuroprotective Agents
a,1
b,1
c
a
a
d
Simone Brogi, Anna Ramunno, Lida Savi, Giulia Chemi, Gloria Alfano, Alessandra Pecorelli,
d
b
c
c,*
c
Erika Pambianchi, Paola Galatello, Giulia Compagnoni, Federico Focher, Giuseppe Biamonti,
Giuseppe Valacchi, Stefania Butini, Sandra Gemma, Giuseppe Campiani, Margherita Brindisi
d,e
a,*
a, *
a
a
a
European Research Centre for Drug Discovery and Development (NatSynDrugs) and Department of
Biotechnology, Chemistry, and Pharmacy, Università degli Studi di Siena via Aldo Moro 2, 53100
Siena, Italy.
b
Dipartimento di Farmacia/DIFARMA, Università degli Studi di Salerno, via Giovanni Paolo II 132,
8
4084 Fisciano, Salerno, Italy.
c
Istituto di Genetica Molecolare, CNR, via Abbiategrasso 207, 27100 Pavia, Italy
d
Department of Animal Science, North Carolina State University, NC Research Campus, PHHI
building, 600 Laureate Way, Kannapolis 28081 – NC USA
e
Department of Life Sciences and Biotechnology, University of Ferrara, Via Borsari 46, 441212,
Ferrara, Italy
Keywords
Antioxidant agents, Glycogen synthase kinase 3 beta (GSK-3β), Adenosine Kinase (AK), dual
inhibitors, molecular docking
Corresponding authors:
*
*
Federico Focher, Tel. +390382546352, email: focher@igm.cnr.it
Stefania Butini, Tel. +390577234161, email: butini3@unisi.it
Sandra Gemma, Tel. +390577234326, email: gemma@unisi.it
*
1
These authors contributed equally to this work

ACCEPTED MANUSCRIPT
2
Abstract
The manuscript deals with the design, synthesis and biological evaluation of novel benzoxazinone-
based and indole-based compounds as multifunctional neuroprotective agents. These compounds
inhibit human adenosine kinase (hAK) and human glycogen synthase kinase 3 beta (hGSK-3β)
enzymes. Computational analysis based on a molecular docking approach underlined the potential
structural requirements for simultaneously targeting both proteins’ allosteric sites. In silico hints drove
the synthesis of appropriately decorated benzoxazinones and indoles (5a-s, and 6a-c) and biochemical
analysis revealed their behavior as allosteric inhibitors of hGSK-3β. For both our hit 4 and the best
compounds of the series (5c,l and 6b) the potential antioxidant profile was assessed in human
neuroblastoma cell lines (IMR 32, undifferentiated and neuronal differentiated), by evaluating the
2 2
protective effect of selected compounds against H O cytotoxicity and reactive oxygen species (ROS)
production. Results showed a strong efficacy of the tested compounds, even at the lower doses, in
counteracting the induced oxidative stress (50 µM of H O ) and in preventing ROS formation. In
2
2
addition, the tested compounds did not show any cytotoxic effect determined by the LDH release, at the
concentration range analyzed (from 0.1 to 50 µM). This study allowed the identification of compound
5
l, as the first dual hAK/hGSK-3β inhibitor reported to date. Compound 5l, which behaves as an
effective antioxidant, holds promise for the development of new series of potential therapeutic agents
for the treatment of neurodegenerative diseases characterized by an innovative pharmacological profile.

ACCEPTED MANUSCRIPT
3
1
. Introduction
Oxidative stress (OS) stems from an imbalance between the increase of the levels of reactive oxygen
species (ROS), with respect to basal conditions, and the detoxification efficiency of the biological
system in repairing eventual resulting damage [1]. OS is implicated in many pathologies and in several
neurodegenerative diseases including Alzheimer’s disease (AD), Parkinson’s disease (PD), multiple
sclerosis (MS) [2, 3], and amyotrophic lateral sclerosis (ALS) [4]. Although the etiology of these
diseases is not fully elucidated, a series of seminal evidences have indicated OS as one of the potential
hallmark common to different neurodegenerative diseases [5]. Chronic OS causes neuronal cell death
by different mechanisms: moderate OS induces an apoptotic mechanism while acute OS causes
necrosis [6]. For these reasons the identification of compounds able to prevent or lower cellular ROS
formation is an intriguing task potentially endowing compounds with neuroprotective properties [7, 8].
Accordingly, the development of agents able to prevent neuronal cell death by inhibiting one or more
enzymes involved in the neurodegenerative processes is a challenging task [9]. This is especially true
for those diseases that are currently only symptomatically treated and for which a disease-modifying
approach is highly desirable [10, 11].
Further, for the treatment of complex central nervous system (CNS) disorders and neurodegeneration
where disease progression involves a wide number of biological systems, the development of multi-
target directed ligands (MTDLs) is particularly advantageous [12-14] with respect to selective
compounds (acting in a one-drug one-target fashion) [15]. In this view, efficient targeting of different
pathways of a given disease by using only one molecule may improve therapeutic effectiveness [12,
1
4-16]. To support this hypothesis, besides a fine-tuning modulation of multiple G protein–coupled
receptors (GPCRs) for attaining neuroprotection [17], dual enzyme inhibitors such as β-site amyloid
precursor protein cleaving enzyme (BACE-1)/Glycogen Synthase Kinase 3β (GSK-3β) [18],
cyclooxygenase-2 (COX-2)/5-lipoxygenase (5-LOX) [19], µ-calpain/cathepsin B [20], fatty acid amide

ACCEPTED MANUSCRIPT
4
hydrolase (FAAH)/monoacylglycerol lipase (MAGL) have also been explored and/or proposed for the
treatment of different brain disorders [21].
Human adenosine kinase (hAK, EC.2.7.1.20) and human glycogen synthase-3β (hGSK-3β, EC
2
.7.11.26) are both involved in neurodegenerative disorders (such as AD and PD), and in the
modulation of OS. Following our strong interest in the development of MTDLs [22] for different
neurological [14, 23-25] and neurodegenerative diseases [11, 13, 26-30] we decided to explore the
generation of dual inhibitors of hAK and hGSK-3β.
hGSK-3β is a constitutively active serine-threonine protein kinase, which plays a key role in several
physiological processes ranging from glycogen metabolism to gene transcription. Its abnormal activity
has been associated with some brain and neurodegenerative disorders such as schizophrenia, bipolar
disorder, AD and PD [31]. Concerning AD, hGSK-3β over-activity accounts for memory impairment,
tau hyperphosphorylation, increased amyloid beta (Aβ) production and local plaque-associated
microglial-mediated inflammatory responses [32]. It was demonstrated that OS was able, in cultures of
rat cortical neurons, to significantly increase tau phosphorylation by elevating the activity of GSK-3β
[
33]. It has been proposed that, either interaction of the Aβ with GSK-3β, or the Aβ mediated OS and
ROS overproduction are able to activate GSK-3β. On the other hand, different studies have concluded
that an increased functioning of GSK-3β may promote the β-secretase mediated Aβ formation. In any
case, inhibition of GSK-3β may result beneficial for the treatment of AD [33]. More recently, hGSK-3β
was also discovered to play a role in several cellular processes associated with the pathogenesis of PD,
including the accumulation of α-synuclein aggregates, OS and mitochondrial dysfunction [34].
Furthermore, in neuronal hippocampal cell lines it was demonstrated that inhibition of GSK-3β is
involved in the control of OS [35]. The seminal studies that clearly evidenced the link between GSK-3β
and the modulation of OS have prompted the development of a series of GSK-3β inhibitors for the
treatment of neurodegenerative diseases. Nevertheless it must be considered that over-inhibition of

ACCEPTED MANUSCRIPT
5
GSK-3β might be deleterious since, by activating Wnt signaling, it would induce tumorigenesis
through tumor necrosis factor toxicity. Accordingly, since it was demonstrated that optimal inhibition
levels should be applied [36] we investigated an indirect modulation of GSK-3β by developing
allosteric modulators.
hAK, which phosphorylates the nucleoside adenosine (Ado) to Ado monophosphate (AMP), plays an
important role in modulating both intracellular and extracellular concentrations of Ado, which exerts
autocrine and paracrine activity on cell physiology, via interaction with Ado receptors in the vicinity of
its release site [37, 38]. It is relevant to the purpose of developing neuroprotective agents that Ado
exerts a cytoprotective effect by enhancing the activities of antioxidant enzymes (e.g., superoxide
dismutase, catalase and glutathione peroxidase) thus reducing the level of ROS [39]. On these bases
inhibition of Ado catabolic enzymes such as AK may allow to elicit these beneficial effect on OS.
Recent studies report that abnormal Ado concentrations are involved in some diseases affecting central
and peripheral nervous systems (CNS, PNS) such as schizophrenia, epilepsy, AD, Huntington’s
disease, ALS, and MS, where the fine-tuning of Ado concentration by hAK inhibition appears
physiologically relevant [40, 41].
The role of the inhibition of both enzymes for the treatment of neurodegenerative disorders, and their
converging efficacy in the control of OS suggested hAK and hGSK-3β as suitable targets for the
development of innovative MTDLs endowed with neuroprotective properties. A further support to the
rational for developing dual inhibitors stemmed from the analysis of the structure of both enzymes.
We recently described the development of allosteric inhibitors for hAK [42, 43] while selective
allosteric modulators of hGSK-3β [44], by targeting sites distinct from the ATP binding site, have been
described by others [45-47].
Aiming at developing the first hAK/hGSK-3β dual inhibitors with potential neuroprotective properties,
our analysis started from the investigation of the chemical structure of known allosteric inhibitors of

ACCEPTED MANUSCRIPT
6
hAK or hGSK-3β exemplified by NSD438 (1) [42], VP0.7 (2), [45] and benzothiazinone 3 [47] (Figure
) for retrieving the common structural determinants for allosteric inhibition of both enzymes. A
1
preliminary biological test to assess hGSK-3β inhibition potency on a selected panel of our in house
database allowed us to select compound 4 [48] (Figure 1) as a 16 µM hGSK-3β inhibitor. Starting from
4
and taking into account our knowledge on the allosteric modulation of hAK, by combining synthetic,
molecular modeling and biological efforts we identified the benzoxazinone as a promising scaffold for
hAK/hGSK-3β dual inhibition (compounds 5a-s, Figure 1 and Table 1). In more details, a medicinal
chemistry approach was based on bioisosteric ring contraction of the seven membered ring of the
pyrrolobenzoxazepine system of 1 and the transformation of the keto functionality in the lactam of the
benzoxazinones 5a-s with a bicyclic scaffold that mimics that of compound 1. An aromatic system
attached to the lactam nitrogen, mainly represented by an N-benzyl-substituted system as in 3, could
serve as mimetic of the pyrrole-fused system of 1. A further ring contraction of compounds 5a-s was
achieved with the indole-based inhibitors (6a-c). These newly conceived scaffolds also allowed
improving synthetic feasibility and chemical tractability by removal of chiral centers. All the developed
compounds were investigated as dual inhibitors of hAK and hGSK-3β.
To characterize the potential neuroprotective profile of the developed compounds (4, 5c,l and 6b), the
cellular effects on cytoprotection and OS modulation were evaluated on the human neuroblastoma cell
line IMR 32. As observed none of the compounds showed a cytotoxic effect as determined by LDH
release. In addition, compounds 4, 5c,l and 6b were effective in preventing ROS production after H O
2
2
treatment in IMR 32 cells. Among this new class of compounds derivative 5l stood out and represents
the first hAK/hGSK-3β dual inhibitor reported to date which is also able to counteract OS in
neuroblastoma cell lines. This compound will pave the way to the discovery of a new class of
multifunctional drugs as potential neuroprotective agents.

ACCEPTED MANUSCRIPT
7
Me
O
S
OH
N
O
O
Me
H
S
N
N
N
H
10
O
O
O
N
Me
1. NSD438
2. VP0.7
3
S
N
4
O
IC50 hGSK-3 = 16 M
IC50 hAK > 50 M
Title compounds
X, Y = H
X = 1-morpholin, Y = H
X = F, Y = H
2-naphthyl
1-naphthyl
2-naphthyl
p-tolyl
X, Y = -OCH O-
Ph
2
R1
X
Y
O
R1
O
N
R2
Z
N
O
H
R2
removal of
carbonyl
Me
Bn, p-sustitued Bn
PhSO2
(±)-5a-s
6a-c
N(Et)2
PhCO
=
Favourable SAR
Unfavourable SAR
Me, Et
Figure 1. Outline of the rational design of hAK/hGSK-3β dual allosteric inhibitors 5a-s and 6a-c
structures are defined in Table 1).
(

ACCEPTED MANUSCRIPT
8
2
2
. Results and Discussion
.1. Chemistry
The compounds 5a-s and 6a-c were synthesized as described in Schemes 1-4. Briefly, condensation of
-aminophenol 7 with α-bromonapthyl ester 8a or 8b in the presence of K CO gave the benzoxazinone
2
2
3
derivatives 9a,b that were N-alkylated with the appropriate benzyl bromide (Scheme 1) to afford the
compounds 5a-g (Table 1). Reduction of 9a with BH SMe in THF gave the corresponding
benzoxazine 10, which was reacted with benzoyl chloride or benzenesulfonyl chloride to obtain 5h and
i respectively (Scheme 1). The benzoxazinones 13a-e (Table 1) were prepared from O-alkylation of
appropriate 2-nitrophenol derivatives 5j-s with α-bromonapthyl ester 8a or 8b or α-bromophenyl ester
c to provide the corresponding intermediates 12a-e (Scheme 2). Subsequent reduction of 12a-e
3
2
5
8
afforded the corresponding aniline derivatives which spontaneously cyclized to benzoxazinone
analogues 13a-e. Finally, 13a-e were N-alkylated with the appropriate benzyl/alkyl halides to furnish
the title compounds 5j-s (Table 1). The synthesis of 6-nitrobenzo[1,3]dioxol-5-ol 11c was
accomplished in three steps, as depicted in Scheme 3. In detail, commercially available sesamol 14 was
converted in its acetyl analogue 15, that was nitrated with HNO /AcOH [49] to give the acetic acid 6-
3
nitrobenzo[1,3]dioxol-5-yl ester 16. Then, after hydrolysis of 16 with 20% (v/v) H SO , the 6-
2
4
nitrobenzo[1,3]dioxol-5-ol 11c was obtained in good yield.
The synthesis of indole-based compounds (6a-c, Scheme 4) started with isatine 17 which, after
treatment with sodium hydride, generated the sodium isatinate intermediate which was subjected to a
Grignard reaction with the appropriate aryl bromide magnesium salt for obtaining the intermediate
carbinols (18a,b). These latter were immediately dehydroxylated by means of a tin(II)chloride
reduction and the obtained indolinones (19a,b) were treated with acetyl chloride or diethylcarbamoyl
chloride providing the esters 6a,b and the carabamoyl analogue 6c.

ACCEPTED MANUSCRIPT
9
Scheme 1. Reagents and conditions: a) K CO , dry DMF, 100 °C; b) Cs CO , dry DMF, rt, R Br; c)
2
3
2
3
2
BH •SMe , dry THF, rt, 12 h; d) benzoyl chloride (for 5h) or benzenesulfonyl chloride (for 5i) TEA,
3
2
dry dichloromethane (DCM), rt, 12 h.
Scheme 2. Reagents and conditions: a) K CO , dry DMF, rt, 12 h; b) morpholine, MeCN, reflux [50];
2
3
c) Fe/NH Cl/THF/H O, 100 °C (for 13a,b) and SnCl /EtOH 80 °C (for 13c-e); d) Cs CO DMF,
4
2
2
,
2
3,
appropriately p-substituted benzylbromide, rt.

ACCEPTED MANUSCRIPT
1
0
2 3 2 4
Scheme 3. Reagents and conditions: a) Ac O, NaOH, 0 °C to rt; b) HNO c/AcOH, rt; c) 20% H SO .
Scheme 4. Reagents and conditions: a) NaH (60% slurry in mineral oil), p-tolylmagnesium bromide (1
M in THF) (for 18a) or 1-naphthylmagnesium bromide (for 18b), dry THF, from -15 °C to rt, then rt 14
h; b) SnCl , AcOH, HCl, 115 °C, 2.5 h; c) 2,6-lutidine and AcCl (for 6a,b) or diethylcarbamoyl
2
chloride (for 6c), dry DCM, from 0 °C to rt, then rt 16 h.
2
.2. Enzymatic studies, molecular modelling, and structure-activity relationships (SAR)
2
.2.1. Enzymatic inhibition
The inhibition effect of our compounds on hGSK-3β and hAK activity was examined over a range of
inhibitors’ concentrations in the assay conditions reported in Methods section, where labeled ATP (10
µM) or Ado (1 µM) are in their respective K concentrations. Table 1 shows the IC values of the
M
50
compounds on both enzymes.
Since our aim was to rationally design molecules potentially able to inhibit both enzymes, we have
tested on hAK the compounds that showed the best inhibition potential against hGSK-3β (Table 1).

ACCEPTED MANUSCRIPT
1
1
Table 1. Inhibitory activity of compounds 4, (±)-5a-s, and 6a-c against hAK and hGSK-3β (IC , µM).
5
0
5
a-s
6a-c
hAK
hGSK-3βa
IC50 (µM)
a
Cpd
R1
-
R2
-
X
-
Y
-
Z
-
IC50 (µM)
b
b
(
%)
(%)d
c
1
4
1.2
50
33)
NT
>
(
16
d
> 50
(±)-5a
NT
(
48)
d
(
±)-5b
NT
40
50
(
±)-5c
5.4
H
H
O
d
> 50
(±)-5d
NT
(
46)
d
> 50
40)
(±)-5e
NT
(
(
±)-5f
31.6
32.2
8.1 ± 1.8
MeO
(±)-5g
H
H
F
H
H
O
8.7 ± 1.8
> 50
d
(±)-5h
NT
(
25)
H,H
(
±)-5i
±)-5j
±)-5k
57.4
24.7
> 50
4.1 ± 0.3
21.2 ± 6.3
10.1 ± 2.8
(
H
H
O
O
(

ACCEPTED MANUSCRIPT
1
2
(
±)-5l
±)-5m
±)-5n
13.6
50
6.4
>
(
8.1 ± 0.8
(12)
-
OCH O-
O
2
>
50
(
Me
Et
> 50
25.6
(12)
>
(
50
8)
(±)-5o
(
±)-5p
-OCH
2
O-
O-
O
O
13.7
6.9 ± 1.5
>
50
40)
50
10)
(±)-5q
50
(
>
(
> 50
(8)
> 50
(15)
-
OCH
2
(
±)-5r
Me
Et
>
(
50
5)
(
±)-5s
6
a
Me
Me
-
-
-
-
-
-
-
-
-
50
50
6
b
13.6
10
d
> 50
38)
6
c
N(Et)
2
NT
(
a
b
Each value is the mean of at least three determinations, standard error of the mean is ≤ 15%;
inhibition % when tested at 50 µM; from ref [40]; NT, not tested.
c
d
Earlier studies demonstrated that our pyrrolobenzoxazepine-based compounds, typified by 1, inhibit
hAK in a non-competitive manner. We also hypothesized and demonstrated by mutagenesis studies that
the specific interaction site of these compounds could be identified in hAK in an allosteric pocket close
to the active site [42, 43] and placed in close proximity of the enzyme surface.
In order to identify a potential binding site on hAK for the class of compounds 5 and 6, we carried out a
blind docking calculation considering hAK in its closed form (PDB ID: 1BX4) and both enantiomers of
5
l. For this computational analysis AutoDock [51] software has been used. AutoDock software is able
to perform the blind docking of compounds and to select the correct complexes based on energy

ACCEPTED MANUSCRIPT
1
3
without previous data regarding the binding site. The results of AutoDock in silico experiment are
reported in Figure 2. In particular the output regarding the (R)-5l is reported in Figure 2A, while the
output of (S)-5l enantiomer is reported in Figure 2B. The data analysis clearly evidences that the
favorite binding site, on hAK enzyme, for the lower energy poses of both enantiomers of 5l is the
allosteric site previously identified for the compound 1 [42, 43], while only few docked solutions in the
considered range (see the Experimental section for details) were found to be in different sites with
respect to the above-mentioned allosteric site. Moreover only docked poses possessing much less
favorable docking scores and predicted ligand affinities, when compared to those found for 5l into hAK
allosteric binding site, were found in the ATP binding site. Additionally, for both enantiomers, no
poses were found in the Ado binding site. According to our computational investigation, both the
binding sites for Ado and ATP, as well as other potential sites located at the surface of the enzyme,
appeared not suitable for 5l-hAK interactions. In fact, we retrieved not favorable docking scores and
predicted ligand affinities for these sites which only accommodated conformers with predicted activity
>
30 µM. Overall, we hypothesized the hAK allosteric site as the most reliable binding site for 5l and
its analogues.

ACCEPTED MANUSCRIPT
1
4
Figure 2. Blind docking output for (R)-5l (A) and (S)-5l (B) of (colored sticks) against hAK (light blue
surface). In the picture are highlighted the main identified binding sites. The picture was generated by
AutoDockTools.
Thus, in order to test their mechanism of action also on hGSK-3β, we assayed two compounds
representative of the series, namely 5c and 5l, at different concentrations of the phosphate donor ATP.
We found out that the compounds inhibit the enzyme by a non-competitive mechanism, interacting
with either the enzyme or the enzyme-substrate complex, decreasing the rate of the enzyme catalytic
activity, without modification of the dissociation constant for the substrate/enzyme complex. As shown
in Figure 3, by increasing the concentrations of the inhibitors the value of Vmax decreases to a new value
called Vapp max. In contrast with Vmax, K for ATP is not affected, remaining close to 10 µM at each
M
inhibitor concentration, indicating for both 5c (Figure 3, Panel A) and 5l (Figure 3, Panel B) a non-
competitive inhibition.

ACCEPTED MANUSCRIPT
1
5
Finally, through the formula K
i
i
= [I]/(Vmax/Vapp max -1), we confirm for 5c and 5l K value equal to IC50
values, namely 5.6 ± 1.0 and 6.3 ± 1.5 µM, respectively
Figure 3. Effect of increasing concentrations of 5c (Panel A) and 5l (Panel B) on hGSK-3β activity, at
3
3
different [ P]-ATP concentrations. 5c concentrations: (■) 0.0, (▲), 1.25, (▼) 2.5, (◆) 5.0, (●) 7.5
µM. 5l concentrations: (■) 0.0, (▲), 3.0, (▼) 6.0, (◆) 9.0, (●) 12.0 µM. Vmax and K
at each inhibitor
concentration were determined with Prism3 software through the formula Y = X* Vmax /(K +X).
M
M

ACCEPTED MANUSCRIPT
1
6
Based on the biological results and on enzyme computational analysis, taking into account the
pharmacophore similarity of 5a-s and 6a-c with 2 and 3, we hypothesized that the new compounds may
bind the allosteric site targeted by compound 2 [45] on hGSK-3β (see Figures S1-S3 of the
Supplementary Material).
Accordingly the designed compounds were subjected to molecular docking studies into the two
postulated allosteric sites to verify the relevant interactions responsible for the dual-target profile and
support structure–activity relationships studies (SARs) and optimization process. In general the results
of the docking studies were consistent with the biochemical studies (determination of the inhibition
potencies of the racemic compounds against hAK and hGSK-3β reported in Table 1).
A systematic and comprehensive SAR study of the developed compounds 5a-s and 6a-c started from
compound 4 (Ki hGSK-3β = 16 µM; Ki hAK > 50 µM). We investigated the potential substitutions
able to improve the binding of compound 4 within the allosteric cavities of the target enzymes
following our docking output indications (Figure S1A and B for (R)-4 and (S)-4, respectively). Based
on known hGSK-3β inhibitors’ structure the pyrrole moiety was removed and replaced by hydrophobic
bulky and flexible groups while the benzoxazinone system was decorated by different substituents. The
putative binding modes of the dual inhibitors were investigated using a computational protocol based
on the Induced Fit Docking (IFD) technique [24, 28, 52]. As better detailed in the experimental section
we used human enzymes for the docking protocols (hAK, PDB ID: 1BX4; hGSK-3β, PDB ID: 1HF8).
The IFD output obtained with the (S)- and (R)-enantiomers within hAK and hGSK-3β binding sites for
the best performing dual hAK/hGSK-3β inhibitors 5c and 5l is depicted in Figures 4 and 5 respectively.

ACCEPTED MANUSCRIPT
1
7
Although the enantiomers of each compound were found to mainly target conserved residues, we noted
the lack of a stereoselective mode of interaction with the putative allosteric binding sites with slight
differences in terms of number of contacts. This was in agreement with previous results on hAK [42,
4
3].
(R)-5c and (S)-5c bind hAK (Figure 4 panels A and B, respectively) by H-bonding the backbone of
F338 with the carbonyl group of the benzoxazinone scaffold. This latter also stacked with W75 and
H107 residues. The naphtyl moiety was able to establish hydrophobic interactions and a cation-π
stacking with the sidechain of K71.
Figure 4. IFD output of compound 5c: (R)-5c and (S)-5c into hAK (panels A and B; respectively) and
hGSK-3β (panels C and D; respectively). H-bonds are reported as black dotted lines. Pictures were
generated by PyMOL. Nonpolar hydrogen atoms were omitted for the sake of clarity.

ACCEPTED MANUSCRIPT
1
8
When 5c was docked into the allosteric site of hGSK-3β, although the binding modes of the two
enantiomers were slightly different, we noted similar pattern of residues involved in the interactions
with the compound. Furthermore, (R)-5c (Figure 4C) was accommodated within a cleft where the
contacts with R209 mainly govern the retrieved binding modes. Compound 5c established hydrophobic
interactions such as cation-π stacking, and also an H-bond between the bridged oxygen of the
benzoxazinone nucleus and the sidechain of R209. The benzoxazinone carbonyl group formed polar
contacts with the sidechain of T235. The naphthyl moiety formed a strong network of hydrophobic
interactions with H173 (π -π stacking) and P331. On the contrary, the docked pose of (S)-5c (Figure
4
D) was mainly characterized by hydrophobic contacts. Indeed, the benzoxazinone moiety of (S)-5c
was differently accommodated to strongly interact by a double π -π stacking with H173. This event
hindered the formation of the above-noted H-bonds with T235 and R209, in favor of more productive
hydrophobic interactions with the P331 sidechain. For this enantiomer the p-F-benzyl moiety was
deeply positioned into the allosteric site of hGSK-3β interacting with the loop containing residue R209,
by hydrophobic interactions (Figure 4D).

ACCEPTED MANUSCRIPT
1
9
Figure 5. IFD output of compound 5l: (R)-5l and (S)-5l into hAK (panel A and B; respectively) and
hGSK-3β (panel C and D; respectively). H-bonds are reported as black dotted lines. Pictures were
generated by PyMOL. Non polar hydrogen atoms were omitted for the sake of clarity.
Docking studies with compound 5l into the hAK allosteric site showed that it could establish a series of
polar and hydrophobic contacts. In particular, (R)-5l (Figure 5A) formed an H-bond with the residue
F338 (backbone) by the oxygen of the carbonyl group. The tricyclic system can establish a series of
relevant interactions such as a π-π stacking with W75 by its aromatic moiety, a relevant hydrophobic
interaction with H107 by its dioxole ring and an H-bond with the backbone of F338 by the oxygen of
the carbonyl group.
Moreover, the naphthyl group was able to establish hydrophobic interactions and a cation-π stacking
with the sidechain of K71. The enantiomer (S)-5l (Figure 5B) was differently accommodated allowing

ACCEPTED MANUSCRIPT
2
0
H-bonds with residues E78 and H107. Furthermore, it was able to strongly interact with the binding site
by an improved network of hydrophobic interactions with respect to (R)-5l. Besides the previously
described π-π stacking with W75 and hydrophobic interactions with H107, the naphthyl moiety
established a cation-π stacking with K71 and a π-π stacking with F338.
Regarding the docking outputs of 5l into hGSK-3β (Figure 5), we observed that the compound was able
to bind the allosteric cavity in a similar fashion to that already reported for compound 2, targeting a
series of key residues by hydrophobic and polar contacts [45]. Furthermore, after analyzing the docking
output of (R)- and (S)-5l enantiomers (Figure 5 panels C and D, respectively) we found very limited
differences. In particular, the main contacts of the compound were established with R209 (cation-π
stacking) by the tricyclic moiety. Other polar contacts could be formed by our ligand and the sidechain
or backbone of T235 and S236, respectively for both enantiomers. In addition, the naphthyl moiety
established hydrophobic interactions with H173 (π-π stacking), and P331.
Starting from 5a (Figure S1C and D for (R)-5a and (S)-5a, respectively), bearing a “naked” benzyl
group, the p-position of the benzyl system was explored with compounds 5b-f by introducing electron-
withdrawing (EWGs) or electron-donating groups (EDGs) such as Me, F, Cl, NO and OMe, also
2
characterized by different shapes and steric hindrance. Introduction of a Me group (5b), led to a lack of
activity against hGSK-3β (Figure S1E and F for (R)-5b and (S)-5b, respectively). Also the introduction
of a p-Cl (5d) (Figure S1G and H for (R)-5d and (S)-5d, respectively) or a p-nitro group (5e) (Figure
S1I and J for (R)-5e and (S)-5e, respectively) caused a dramatic fall of the activity. These data support
the hypothesis that small substituents at p-position such as H and F may be better tolerated by the
hGSK-3β allosteric binding site, and may promote a binding mode as that retrieved for 5c. Introduction
of a p-methoxy as in 5f led to a compound with a similar pattern of interactions as for 5c with the
allosteric binding site. In addition 5f is able to form, with the oxygen of p-methoxy group, an H-bond

ACCEPTED MANUSCRIPT
2
1
with the backbone of G210. This binding mode accounted for a similar docking score as well as the
activity against hGSK-3β (Figure S1K and L for (R)-5f and (S)-5f, respectively).
Further, for exploring a different region of the benzoxazinone scaffold, 2-naphthyl regioisomer was
evaluated (5g vs 5c) and this modification was well tolerated by both enzymes (Figure S1M and N for
(
R)-5g and (S)-5g, respectively).
Shifting the lactam carbonyl group to the exocyclic position (5h) resulted in a complete loss of potency
indicating the need of a specific geometry of the bicyclic system for correct interaction (Figure S1O
and P for (R)-5h and (S)-5h, respectively). On the other hand, when the amide carbonyl group of 5h is
replaced by a phenylsulfone group (5i), this latter is able to mimic the interaction with T235 and/or
S236, already described for 5c (Figure 4), 5f, and 5g. As expected, the binding mode of 5i accounted
for a similar inhibitory potency against hGSK-3β as found for 5c. Moreover, also for 5i the rings of the
central core established a cation-π stacking with R209 (Figure S1Q and R for (R)-5i and (S)-5i,
respectively).
Consistent with the positive contribution given by the small F substituent seen in 5c, we inserted a F
atom at C7 of the benzoxazinone moiety. This modification was well tolerated and for 5j we registered
a good inhibition profile for both enzymes paralleled by a favorable docking score (Figure S2A and B
for (R)-5j and (S)-5j, respectively). The introduction of a morpholine (5k) at C7 of the heterocyclic
system caused a slight decrease of activity against both hGSK-3β and hAK, due to the poorly tolerated
steric hindrance at this position (Figure S2C and D for (R)-5k and (S)-5k, respectively).
Since the introduction of small substituents at C7 was tolerated, our next docking outputs prompted us
to modify C7 and C8 of the benzoxazinone system. By adding a dioxole moiety to the scaffold of 5a,
we obtained one of our best performing compounds (5l) in terms of computational and biological
results.

ACCEPTED MANUSCRIPT
2
2
Based on these results we decided to investigate this tricyclic scaffold by introducing the same
substitutions as in 5a. As already observed, the introduction of a 2-naphthyl moiety (5p) implied minor
changes in the inhibitory activity, and preserved a favorable binding mode into the hGSK-3β allosteric
site (Figure S2E and F for (R)-5p and (S)-5p, respectively). On the contrary, the addition of the p-F on
the benzyl group (5m) did not led to the positive increment as registered for the first series of
compounds (Figure S2G and H for (R)-5m and (S)-5m, respectively).
In order to explore the hydrophobic substitution at 4 position of the tricyclic system we replaced the
benzyl moiety with the smaller Me (5n) (Figure S2I and J for (R)-5n and (S)-5n, respectively) and Et
(
5o) (Figure S2K and L for (R)-5o and (S)-5o, respectively) substituents. The absence of the aromatic
group in the central core determined the loss of a favorable 3D-arrangement in both enzymes with a
consequent loss of activity of both compounds (5n and 5o).
Since the 2-naphthyl substitution was well tolerated at 2 position, we explored the effect of the
simplification of the aromatic system by removing one of the aromatic rings. The phenyl substituted
analogue (5q) showed a different accommodation into the hGSK-3β allosteric binding site (Figure S2M
and N for (R)-5q and (S)-5q, respectively). When compared to the pose obtained for 5l we observed
that, due to the smaller size of the hydrophobic moiety, the key H-bonds were lost establishing
hydrophobic contacts with the side chain of the R209. The presence of the phenyl moiety at C2 was
then explored in combination with the replacement of the benzyl group, with Me (5r) (Figure S2O and
P for (R)-5r and (S)-5r, respectively) and Et (5s) (Figure S2Q and R for (R)-5s and (S)-5s,
respectively), as done for 5n and 5o. This led to compounds characterized by different binding modes
and by a relevant decrease of inhibitory activity against hGSK-3β.
With the aim of simplifying the central scaffold we developed indole-based analogues devoid of a
chiral centre but still able to establish the same contacts identified for the benzoxazinone-based
compounds. Consistent with our hypothesis analogue (6b) bearing a 1-naphthyl substituent showed a

ACCEPTED MANUSCRIPT
2
3
satisfactory binding mode into the hGSK-3β allosteric binding site (Figure S3A-C for 6a-c,
respectively), confirmed by biological assays. This compound is able to stack with the R209 and to
make polar contacts with the backbone of Y234, V208 and G210, sharing the pharmacophoric
elements, previously described for 5c, for interacting in both enzymes. This binding mode accounted
for the best IC against both hAK and hGSK of the indole-based series.
5
0
2
.3. Cytotoxicity evaluation and studies for evaluating antioxidant potential
Since OS is one of the main causes of neurodegeneration and is a relevant issue in a variety of
neurological diseases we investigated the therapeutic potential of our dual inhibitors of hAK/hGSK-3β.
To this end we decided to test the efficacy of compounds 4, 5c,l and 6b in an acute model of OS
induced by H O treatment in both, differentiated and not differentiated IMR 32 cells.
2
2
Our test started by engaging the selected compounds in preliminary studies for determining their
cytotoxicity profile by using IMR 32 cells. Accordingly, IMR 32 cells were treated with increasing
concentrations of the tested compounds and cell vitality was then assessed by LDH release (see
Supplementary Material for further details). Preliminary studies allowed us to define the correct cell
density to be used which was chosen on the base of cellular confluence 24 h after the seeding. The
optimal number of cell seeded was 100,000 cells/well, as shown Figure S4 of the Supplementary
Material. The cytotoxicity assessment (Figure 6) clearly showed that no cytotoxic effect among the 6
different doses tested was observed. Of note is that compound 4, at the doses of 10 and 50 µM, clearly
affect the cellular clusters formation (Figure S5) although 4 was found devoid of toxicity to the cells.
No significant effects on IMR 32 cellular morphology were observed after compounds treatment, as
reported Figure S5.

ACCEPTED MANUSCRIPT
2
4
In summary, none of the tested molecules showed any cytotoxic effect on the adopted cellular model in
the wide range of doses chosen (from 0.1 µM to 50 µM). These results allowed us to perform further
experiments to evaluate their potential antioxidant properties against H O treatment.
2
2
Figure 6. Cytotoxic effect of compounds 4, 5c,l and 6b. Values represent average ± SD for three
experiments in triplicate. Data are expressed as arbitrary unit LDH release as compared to the
maximum release of LDH from Triton X-100- treated cells.

ACCEPTED MANUSCRIPT
2
5
Following our plans we then evaluated the cytotoxicity of different doses of H O (25, 50 and 100 µM)
2
2
(
2 2
Figure S6A). Further experiments were performed with 50 µM of H O , since it was the highest dose
that did not affect dramatically cells morphology and viability, indeed with 100 µM treatment the LDH
release paralleled the one of Triton X-100 (100% of cells damage) (Figures S6A and S6B).
Further, the levels of ROS by DCF probe were determined. As shown in Figure S7, we could observe
that after 1 h of H O treatment, ROS production linearly increased in a dose and time dependent
2
2
manner (Figure S7).
As shown in Figure 7, all the compounds were able to significantly prevent ROS formation. Notably is
that this effect was evident already at the lower tested doses (0.1, 0.5 and 1 µM), demonstrating a good
protection towards oxidative damage. This efficacy at lower doses may be due to the synergism
obtained by inhibition of both enzymes. It is possible to assume, that the lack of dose linearity is due to
the fact that the protective mechanism induced by these compounds reaches its maximum effect already
at the lower dose (0.1 µM) confirming the high efficacy of the compounds.

ACCEPTED MANUSCRIPT
2
6
Figure 7. Prevention of ROS formation by the selected compounds (4, 5c,l and 6b). Values represent
average ± SD for three experiments in triplicate. *p < 0.05 versus untreated sample (0 µM).
Based on these encouraging data, similar tests were performed in neuronal IMR 32 differentiated cells
by a 12 days treatment with bromodeoxyuridine (BrdU). As shown in Figure 8, all the compounds were
able to significantly prevent ROS production at very low doses (0.1-0.5 and 1 µM) after 50 µM H O
2
2
treatment. These outcomes, which demonstrated the good efficiency of the compounds 4, 5c, 5l, and 6b
in preventing ROS formation also in differentiated neuronal cells, indicate these compounds as a
possible new therapeutic strategy to prevent oxidative stress damage in neurodegenerative pathologies.

ACCEPTED MANUSCRIPT
2
7
Figure 8. Compounds 4, 5c,l and 6b prevent ROS formation (DCFDA Assay). Values represent
average ± SD for three experiments in triplicate. *p < 0.05 versus untreated sample (0 µM).
3
. Conclusion
In summary, we have described herein the rational design of the first MTDL prototypes for dual
inhibition of hAK/hGSK-3β reported to date. For the most promising compounds of the series (4, 5c,l
and 6b) their ability to counteract OS in neuronal cells was also experimentally measured. The
described medicinal chemistry approach led to a novel series of compounds sharing the
pharmacophoric requirements for interacting with both allosteric sites of the hAK/hGSK-3β enzymes.
Non-competitive mechanism of action was ascertained for 5c and 5l in hGSK-3β. Subsequently, the
most interesting dual hAK/hGSK-3β inhibitors (4, 5c,l and 6b) proved to be devoid of cytotoxicity

ACCEPTED MANUSCRIPT
2
8
towards the neuroblastoma cell line IMR 32 and were engaged in biological assays to evaluate their
ability to counteract ROS formation. Among the tested compounds, compound 5l stood out as a good
MTDL potentially endowed with therapeutic potential for neurodegenerative diseases. In fact 5l, one of
the best dual hAK/hGSK-3β inhibitors of the series, is able to counteract OS and preventing ROS
formation in neuronal cells devoid of any cytotoxicity. The dual hAK/hGSK-3β inhibition approach
herein described may lead to the development of innovative antioxidant/neuroprotective agents
potentially useful for the treatment of neurological disorders.
4
4
4
. Methods
.1. Chemistry
.1.1. General Procedures
1
13
Melting points were taken on a Gallenkamp melting point apparatus and are uncorrected. H and C
NMR spectra were recorded on a Bruker Avance 300 MHz, a Varian 300 MHz, or a Bruker 400 MHz
spectrometer. Splitting patterns are described as singlet (s), doublet (d), triplet (t), quartet (q),
quintuplet (qt), and broad (br). Chromatographic separations were performed on silica gel (Kieselgel
4
0, 0.040-0.063 mm, Merck). Reactions and products mixtures were routinely monitored by thin-layer
chromatography (TLC) on Merck 0.2 mm precoated silica (60 F254) aluminum sheets, with
visualization by irradiation with a UV lamp. ESIMS and HRESIMS were carried out by a Thermo
Finningan LCQ Deca XP Max ion-trap mass spectrometer equipped with Xcalibur software, and an
LTQ Orbitrap XL mass spectrometer (Thermo FisherScientific, San Jose, CA, USA) operated in
positive ion mode, respectively. The Orbitrap mass analyzer was calibrated according to the
manufacturer’s directions using a mixture of caffeine, methionine-arginine-phenylalanine-alanine-