The chemosensitizing agent lubeluzole binds calmodulin and inhibits Ca2+/calmodulin-dependent kinase II

Article abstract of DOI:10.1016/j.ejmech.2016.03.045

An affinity capillary electrophoresis (ACE) method to estimate apparent dissociation constants between bovine brain calmodulin (CaM) and non-peptidic ligands was developed. The method was validated reproducing the dissociation constants of a number of wel

Full text of DOI:10.1016/j.ejmech.2016.03.045

Accepted Manuscript
2+
The chemosensitizing agent lubeluzole binds calmodulin and inhibits Ca
calmodulin-dependent kinase II
/
Claudio Bruno, Dr. Maria Maddalena Cavalluzzi, Ph. D, Maria Rosaria Rusciano,
Angelo Lovece, Antonio Carrieri, Riccardo Pracella, Giulia Giannuzzi, Lorenzo
Polimeno, Maurizio Viale, Maddalena Illario, Carlo Franchini, Giovanni Lentini
PII:
S0223-5234(16)30227-6
10.1016/j.ejmech.2016.03.045
EJMECH 8470
DOI:
Reference:
To appear in: European Journal of Medicinal Chemistry
Received Date: 10 July 2015
Revised Date: 16 March 2016
Accepted Date: 17 March 2016
Please cite this article as: C. Bruno, M.M. Cavalluzzi, M.R. Rusciano, A. Lovece, A. Carrieri, R. Pracella,
G. Giannuzzi, L. Polimeno, M. Viale, M. Illario, C. Franchini, G. Lentini, The chemosensitizing agent
2+
lubeluzole binds calmodulin and inhibits Ca /calmodulin-dependent kinase II, European Journal of
Medicinal Chemistry (2016), doi: 10.1016/j.ejmech.2016.03.045.
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
S
N
OH
N
F
F
O
N
lubeluzole

ACCEPTED MANUSCRIPT
The chemosensitizing agent lubeluzole binds calmodulin and inhibits Ca2+/calmodulin-
dependent kinase II
Claudio Bruno,a Maria Maddalena Cavalluzzi,a* Maria Rosaria Rusciano,b Angelo Lovece,a
Antonio Carrieri,a Riccardo Pracella,a Giulia Giannuzzi,c Lorenzo Polimeno,d Maurizio Viale,e
Maddalena Illario,b Carlo Franchini,a Giovanni Lentini.a
aDipartimento di Farmacia – Scienze del Farmaco, Università degli Studi di Bari ‘Aldo Moroʼ,
via E. Orabona 4, 70126 Bari, Italy.
bDipartimento di Biologia e Patologia Cellulare e Molecolare, Università Federico II, 80131
Napoli, Italy.
c Dipartimento di Bioscienze, Biotecnologie e Biofarmaceutica, Università degli Studi di Bari
ʻAldo Moroʼ, via E. Orabona 4, 70126 Bari, Italy.
dDipartimento dell’Emergenza e dei Trapianti di Organi, Policlinico - P.zza Giulio Cesare -
70124 Bari, Italy.
eIRCCS Azienda Ospedaliera Universitaria San Martino - IST Istituto Nazionale per la Ricerca
sul Cancro (IRCCS SMIST), U.O.C. Terapia Immunologia, L.go R. Benzi, 10, 16132 Genova,
Italy.
*Corresponding author:
Dr. M. M. Cavalluzzi, Ph. D.
Dipartimento di Farmacia – Scienze del Farmaco
1

ACCEPTED MANUSCRIPT
Università degli Studi di Bari ‘Aldo Moro’
via Orabona 4, 70125 Bari, Italy
phone number: +39 080 5442232
fax number: +39 080 5442724
E-mail address: mariamaddalena.cavalluzzi@uniba.it
Abstract
An affinity capillary electrophoresis (ACE) method to estimate apparent dissociation constants
between bovine brain calmodulin (CaM) and non-peptidic ligands was developed. The method
was validated reproducing the dissociation constants of a number of well-known CaM ligands. In
particular, the potent antagonist 125-C9 was ad hoc synthesized through an improved synthetic
procedure. The ACE method was successfully applied to verify CaM affinity for lubeluzole, a
well-known neuroprotective agent recently proved useful to potentiate the activity of anti-cancer
drugs. Lubeluzole was slightly less potent than 125-C9 (Kd = 2.9 ± 0.7 and 0.47 ± 0.06 µM
respectively) and displayed Ca2+/calmodulin-dependent kinase II (CaMKII) inhibition (IC50 = 40
± 1 µM). Possible binding modes of lubeluzole to CaM were explored by docking studies based
on the X-ray crystal structures of several trifluoperazine-CaM complexes. An estimated
dissociation constant in good agreement with the experimental one was found and the main
aminoacidic residues and interactions contributing to complex formation were highlighted. The
possibility that interference with Ca2+ pathways may contribute to the previously observed
chemosensitizing effects of lubeluzole on human ovarian adenocarcinoma and lung carcinoma
cells are discussed.
2

ACCEPTED MANUSCRIPT
Keywords: lubeluzole, calmodulin, affinity capillary electrophoresis, anti-cancer activity, human
carcinoma cells, voltage-gated sodium channels, docking.
3

ACCEPTED MANUSCRIPT
1. Introduction
Lubeluzole [(S)-1, Fig. 1] is a homochiral benzothiazole derivative that has shown
neuroprotective properties in preclinical models of ischemic stroke. Several possible mechanisms
have been hypothesized to justify its beneficial activity, including inhibition of glutamate release,
inhibition of glutamate-activated nitric oxide (NO) synthesis, and blockade of voltage-gated
calcium channels [1–4]. Also epigenetic control on NO synthase (NOS) and calmodulin (CaM)
activities has been hypothesized [2]. Recently, lubeluzole was proved to synergize with both
doxorubicin and paclitaxel on human ovarian adenocarcinoma A2780 and human lung
carcinoma A549 cells, respectively [5]. This activity might stem from either some of the above
mechanisms or the well-known lubeluzole voltage-gated sodium channel (VGSC) blocking
activity [6–8]. However, the synergistic effects of lubeluzole for both drugs were observed over a
wide concentration window (0.005–5 µM), the lowest limit being at least 40 times lower than
human plasma concentrations clinically relevant for the anti-ischemic activity [9] and more than
100 times lower than IC50 values for VGSC [6]. Thus, the possibility of concurrence of other
molecular mechanisms contributing to lubeluzole chemosensitizing activity has to be
hypothesized.
S
N
OH
N
F
F
O
N
(S)-1
Fig. 1. Structure of lubeluzole.
4

ACCEPTED MANUSCRIPT
CaM is a small ubiquitous acidic protein that acts as a Ca2+-sensor and modulates the
activity of several proteins [10]. Several anticancer agents bind to CaM [11], and a number of
CaM ligands act as chemosensitizing agents [12–14]. Finally, several CaM ligands inhibit
Ca2+/CaM-dependent kinase II (CaMKII) [15] and the basic biological function of CaMKII is
involved in several cancers [16–19]. This is why CaMKII inhibitors have been suggested as
possible anti-cancer agents [20]. Interestingly, several known CaM ligands, such as loperamide
and trifluoperazine (Fig. 2), share common structural features with lubeluzole. Therefore, we
wondered if lubeluzole might interfere with Ca2+ pathways by binding CaM, thus potentiating
the activity of anti-cancer agents. Several methodologies for the determination of dissociation
constants between CaM and small molecules have been proposed. They are based on NMR
[21,24], CD [22,24], fluorescence spectroscopy [23–25], gel mobility shift assays [24],
enzymatic competition assays [24], chromatography on CaM-sepharose gel [26], binding assays
with tritiated ligands [27,28], and isothermal titration calorimetry [15]. Most of the above
methods involve the synthesis of chromophoric or radioactive ligands [25,27,28], the preparation
of 13C [24], 15N [21], or chromophoric labeled CaM [23]. Several methods require large amounts
of CaM and do not allow accurate Kd determination [22,24,26]. Finally, biochemical assays
require enzymes, cofactors, and substrates [24] while isothermal titration calorimetry is quite
unpractical [15].
Compared with the above reported methods, affinity capillary electrophoresis (ACE) offers some
advantages in the determination of apparent dissociation constants, such as small amount of
protein and ligands, and short time for each analysis. ACE does not require high purity ligands,
since Kd values are based on migration times and not on peak integration. Finally, native protein
in a homogeneous aqueous medium is used. Thus, the experimental setting reasonably
5

ACCEPTED MANUSCRIPT
approximates the physiological environment of the protein and reliable affinity measures can be
obtained. Here we report an ACE method to estimate apparent dissociation constants between
CaM and non-peptidic, protonatable compounds. Since most of the available CaM ligands are
old molecules displaying affinity at the micromolar level, we decided to synthesize a recently
reported potent CaM antagonist, 125-C9 [15], as a reference compound, trying to improve the
early reported synthetic procedure. Once proven the affinity of lubeluzole for CaM, its possible
antagonist behaviour on CaMKII has been investigated and its possible binding mode has been
explored by docking simulations carried out on properly selected crystallographic structures of
CaM-antagonist complex.
2. Results and discussion
2.1. Ca2+/CaM antagonism
Since the amino acid sequence of CaM is completely invariant (100% identity) among all
vertebrates [29], we choose the relatively less expensive and widely used [11,22,43,53] bovine
brain CaM to develop our ACE assay. We got inspiration from a pioneering ACE method used to
measure binding constants of acidic 4-alkylbenzenesulfonamides to carbonic anhydrase B (CAB)
[30]. This method measures changes in CAB electrophoretic mobility (µ
ep) in relation to various
concentrations of charged ligands in the run buffer. The observed CAB µep variations were
caused by the changes in CAB charge (from Z to Z ± z, where Z and z indicate the charges of
CAB and the ligand, respectively), while the changes in CAB µ
ep caused by the changes in mass
were relatively small. However, the original framework was modified in order to make CaM
affinity assay reliability as high as possible. Most of the values reported in the literature for CaM
ligands are derived running experiments at pH 7.4, under the assumption that the latter may be
6

ACCEPTED MANUSCRIPT
assumed as ʻthe physiological pHʼ. Indeed, this statement holds for circulating and interstitial
fluids. However, when drugs supposed to operate in intracellular environments of solid tumors
are concerned, further aspects should be taken into account. First of all, intracellular pH (pHi)
may present deviations from neutrality, generally being more alkaline (>7.6) than that of normal
cells [31,32]. Furthermore, the relative permittivity (εr) of the aqueous medium lowers at the
protein surface [33]. Thus, protonatable CaM ligands should experiment microenvironmental
conditions unfavorable for protonation. Indeed, CaM antagonists widely protonated at pH 7.4
(e.g. trifluoperazine and W-7, Fig. 2) form only few, non-specific ionic interactions with CaM
[34], with the binding being mainly driven by the hydrophobic effect [35]. The above
considerations suggested the adoption of a relatively alkaline operational pH [8,35] roughly
centered on pKas range of the known CaM ligands used as the training set (Fig. 2 and Table S1).
Unfortunately, at these pH values the shift in CaM µep caused by the interaction of the selected
set of compounds (Table S1) with CaM was expected to be negligible. In fact, at pH 8.35 the
migration time of CaM, injected as a sample, was not influenced by the presence of various
concentrations of the ligands in the buffer. This is why we employed a reversed procedure in
which buffers containing different concentrations of CaM were used to evaluate the changes of
migration times of each ligand injected as a sample. A saturating concentration of CaCl2 was
used to activate CaM through the formation of Ca2+-CaM complex in conditions previously
reported as granting physiological Ca2+-binding properties of CaM [36]. Glycine/tris buffer was
preferred to borate buffer because the latter causes current decrease in the presence of CaCl2
(chelating conditions) [37]. In order to minimize the use of the relatively costly protein, the
partial filling technique, i.e. partial-filling ACE (PFACE), was used. Here, zones of protein are
used in place of the column being completely filled with the analyte, thereby reducing the
7

ACCEPTED MANUSCRIPT
amounts of sample required for the binding assay. In particular, we used flow-through partial-
filling affinity capillary electrophoresis (FTPFACE) [38] to estimate the dissociation constants of
well-known protonatable CaM ligands (Fig. 2) considered as a training set. While in PFACE the
capillary is partially filled with a plug of ligand and receptor, and electrophoresed, in this
technique we introduced a plug of CaM solubilized in the run buffer and then a smaller plug of
the ligands. Thus, applying the voltage, the plug of basic samples flowed through the acidic CaM
plug (Fig. S1). Scatchard analysis of the changes in ligand migration times relative to the non
interacting peak of EOF, as a function of the CaM concentration, yields Kd values obtained
repeating the experiments two or three times.
As an example, the superimposition of six electropherograms obtained for trifluoperazine is
shown in Fig. 3A. The lower electropherogram corresponds to trifluoperazine run in the absence
of CaM. The migration time of the sample was about 5 minutes. Going up, electropherograms of
trifluoperazine in the presence of growing concentrations of CaM are shown. Trifluoperazine
peak (indicated by an arrow) slows down with increasing concentration of CaM. Scatchard plots
from a triplicate experiment are shown in Fig. 3C.
8

ACCEPTED MANUSCRIPT
S
N
CF3
N
O
Cl
N
N
H
N
OH
N
trifluoperazine
loperamide
prenylamine
Cl
S
N
NH2
Cl
N
H
O S
O
N
HN
fendiline
chlorpromazine
W-7
N
O
H
N
S
N
Cl
N
N
N
N
N
N
promethazine
clozapine
125-C9
Fig. 2. Structures of well-known protonatable CaM ligands.
9

ACCEPTED MANUSCRIPT
A
B
C
Fig. 3. A) Superimpositions of electropherograms of trifluoperazine alone (A) and in the presence of growing
concentrations of CaM (B, 7.5 mM CaM; C, 15 mM CaM; D, 30 mM CaM; E, 45 mM; F, 60 mM CaM). B)
Binding curve of trifluoperazine-CaM. C) Scatchard plots from triplicate experiment on trifluoperazine-CaM
binding.
For the training set ligands (Fig. 2), Kd values were obtained averaging the values obtained
from the Scatchard plots of at least two data sets and were in agreement with those reported in
the literature (Table 1 and S2). We found that the higher the apparent Kd of the ligands (entries
10

ACCEPTED MANUSCRIPT
6–8), the greater the standard deviation of the measurement. It is noteworthy that the method was
applicable to ligands in complex mixtures: loperamide and chlorpromazine affinity values were
acquired from direct analysis on their samples obtained from their corresponding
pharmaceuticals.
The method was applied to verify CaM affinity for lubeluzole which displayed a Kd value of 2.9
± 0.7 µM.
Table 1.
Comparison of apparent Kd values obtained through the FTPFACE method for selected CaM ligands and
corresponding Kd, or IC50, or Ki values reported in the literature.
Entry
compound
FTPFACE Kd values (ꢀM)a Kd (ꢀM) IC50 (ꢀM)
trifluoperazine
prenylamine
fendiline
loperamide
chlorpromazine
promethazine
3.4 ± 0.1
1.80 ± 0.06
2.9 ± 0.1
3.5 ± 0.5
3.1 ± 0.5
50 ± 20
12 ± 4
61 ± 23
2.9 ± 0.7
5a
–
–
–
1
2
3
4
5
6
7
8
9
0.5–0.7a
0.5–0.7a
–
–
–
12c
8d
60d
W-7
7.2a
13.6e
clozapine
lubeluzole
125-C9
–
–
20d
–
0.47 ± 0.06
0.85b
–
10
a see ref 53; b see ref 15; c see ref 54; d see ref 55; e see ref 56.
The CaMKII inhibitory effect of lubeluzole was then evaluated in vitro following a previously
reports assay [39] and, as shown in Table 2, it exerted an inhibitory activity slightly higher than
that of the well-known CaMKII inhibitor W-7 and quite comparable with the novel Ca2+/CaM
antagonist 125-C9. Thus, it is conceivable that lubeluzole, similarly to W-7 and 125-C9, inhibits
CaMKII binding competitively with CaM.
11

ACCEPTED MANUSCRIPT
Table 2
Inhibition of CaMKII activity by W-7, 125-C9, and lubeluzolea
Compounds
IC50 ± SEM (µM)
W-7
50 ± 2b
125-C9
35 ± 2c
40 ± 1
lubeluzole
a
CaM was used at 5 µΜ, values represent the mean of three
independent determinations; b 100 µM (at 0.5 µM CaM), see ref 57;
c 40.4 µM (at 2 µM CaM), see ref 15.
2.2. Chemistry
2.2.1. Lubeluzole [(S)-1] preparation
Lubeluzole [(S)-1] was prepared following a synthetic route based on the hydrolytic kinetic
resolution reported in our previous work [5], as outlined in scheme 1. Unlike the early reported
procedure, the glycidyl ether (RS)-4 was synthesized following the Williamson procedure. The
first attempt performed by reacting 2 with (RS)-3 in NaOH solution [40] gave the desired product
in acceptable
yield (45%) but the parasitic formation of 1-chloro-3-(3,4-
difluorophenoxy)propan-2-ol was observed too. Although the latter compound is a useful
intermediate in the synthesis of lubeluzole [41], we tried to optimized the reaction conditions in
order to improve (RS)-4 yield. Thus, modifying a literature procedure [42], the reaction was
performed in CH3CN and in the presence of Cs2CO3 giving (RS)-4 in 88% yield, thus improving
the overall yield of lubeluzole from 30% [5] to 38%. The attempt to perform the Williamson
reaction under microwave irradiation at 130 °C for 20 min failed.
12

ACCEPTED MANUSCRIPT
O
O
ii
OH
O
O
F
F
F
F
F
F
O
i
+
Cl
(RS)-3
4
2
(RS)-
(S)-5
S
N
N
iii
NH
6
S
N
OH
N
F
F
O
N
(S)-1
Scheme 1. Reagents and conditions: (i) Cs2CO3, MeCN, reflux; (ii) (R,R)-(salen)CoIII(OAc), H2O, room temp; (iii)
Yb(OTf)3, anhyd CH2Cl2, room temp.
2.2.2. 125-C9 preparation
125-C9 was prepared by modifying a literature procedure [15] which started from the acylation
of N,N'-dibenzylethylenediamine (7) with 3-methoxybenzoyl chloride to give N-benzyl-N-[2-
(benzylamino)ethyl]-3-methoxybenzamide (8, scheme 2). Considering that the low yield
reported in the literature for 8 (19%) was caused by the formation of 9 as a side product, we
optimized the experimental conditions in order to enhance the monoacylation product yield.
Thus, the reaction was performed at 0 °C for 2 h by using DBU as a base and 8 was obtained in
51% yield. Furthermore, by submitting 9 to a partial hydrolysis with 12 M HCl under microwave
irradiation, the overall yield of 8 increased to 60%. In the next step, 8 was reacted with N,N-
dimethylglycine under microwave irradiation, in the presence of EEDQ, and the intermediate 10
was obtained in 65% yield (versus 40% [15]). Finally, reduction of 10 gave 125-C9 (82%)
which was converted into its hydrochloride salt with gaseous HCl.
13

ACCEPTED MANUSCRIPT
ii
O
O
H
N
H
N
N
i
N
N
N
H
+
O
O
O
O
7
8
9
iii
N
N
O
N
N
N
N
iv
O
O
O
125-C9
10
v
125-C9.HCl
Scheme 2. Reagents and conditions: (i) 3-methoxybenzoyl chloride, DBU, anhyd CH2Cl2, 0 °C, 2 h; (ii) 12 M HCl,
THF, MW, 120 °C, 20 min; (iii) N,N-dimethylglycine, EEDQ, CHCl3, MW, 100 °C, 15 min; (iv) LiAlH4, anhyd
THF, reflux, 20 h; (v) Et2O, gaseous HCl.
2.3. Molecular modelling: binding site mapping and docking of lubeluzole to CaM
The accessibility of CaM molecular surface was envisaged through a proper investigation
carried out in order to characterize the multiple binding clefts of this biological target. It is
indeed true that CaM binds diverse agonists with different stechiometry, as proved by X-ray data
of more than one CaM-trifluoperazine complex [10,43,44]. To achieve this topic FPOCKET (ver
2.0) [45], a protein pocket prediction algorithm based on Voronoi tessellation was selected as a
tool for our binding pocket characterization and ranking. Beside a simple visualization, that
14

ACCEPTED MANUSCRIPT
might be at glance the first interpretation of a protein binding surface, FPOCKET also scores the
detected clefts with some straight parameters (i.e. extension, hydrophobicity, polarity, and
charge) which might be useful for a deeper understanding of the chemical and geometrical cliche
of a molecular target.
In our study four main clefts, with quite different extension, and located between the
symmetrical crevice of the α-helices bundle of CaM, were identified on the molecular surface of
CaM, as it can be perceived in Fig. 4.
Fig.4. Binding site mapping of CaM surface as achieved by FPOCKET. The four P1, P2, P3 and P4 pockets are
depicted in red, magenta, blue, and green, respectively.
15

ACCEPTED MANUSCRIPT
In details, the P1 cleft reveals the largest surface as well as the highest pocket score,
according to the FPOCKET scoring method (Table 3). It is interesting to note that, as observed in
the X-ray structure of the 1:4 complex, P1 is completely fulfilled by a trifluoperazine molecule
and also partially by the tricyclic moiety of a second trifluoperazine, suggesting that this might
be the high affinity, or at least the primary, CaM ligand-binding site. As long as this evidence is
concerned a secondary accessible crevice P2, accepting a third antagonist molecule, is located, as
well as P1, near the two ionic pinches, comprising Glu7 and 127, and Glu54 and 84 respectively.
It might be argued that these two distinct cavities might be mandatorily occupied by ligands in
order to successfully hamper CaM from it elongated dumb-bell to a compact globular, and
therefore inactive, conformation. To get further insights into CaM antagonism, we decided to
investigate further by docking the most promising compound of our data set, lubeluzole [(S)-1],
on the target structure.
Table 3
FPOCKET scoring
P1
P2
P3
P4
937.935
9.731
real volume (Å3)
pocket score
druggability score
1163.630
31.144
0.659
322.450
16.342
0.781
241.576
15.730
0.840
0.811
Due to the complexity and diversity of CaM binding surface, in a first issue we performed
blind dockings as a preliminary step useful for the identification of a suitable, and eventually
multiple, binding mode. Overall the three-dimensional structure of CaM-trifluoperazine complex
shows a globular-shaped protein with a large and central cavity, most likely occupied by
16

ACCEPTED MANUSCRIPT
antagonists. Two domains are arranged symmetrically so that hydrophobic surfaces lie opposite
each other forming a tunnel, with acidic residues of the α-helices from both domains lying
outside the same tunnel in negatively-charged patches [44]. This entire conformational space was
therefore sampled by means of AUTODOCK (ver. 4.2) [46].
Because of both extended target area and considerable ligand flexibility, it was impossible to
retrieve cluster poses with a valuable number of members, so as selection criterion for the
supposed binding mode we chose the one endowing the best free energy of binding (FEB) to
carried out further and more refined dockings (see Methods section) arising a plausible binding
mode for lubeluzole [(S)-1]. The best FEB pose placed lubeluzole [(S)-1] into the largest and
highest pocket score P1 pocket, but nonetheless the same ligand might explore alternative
binding most likely referred to P2 pocket.
In particular, while an intramolecular hydrogen bond redirects the difluorophenoxy rings
towards one EF domain, the benzothiazole fragment is deeply buried in a highly hydrophobic
moiety, having its aromatic ring patched through an edge-to-face π-π stacking to the side chain of
Phe92, and arising other favorable Van der Waals with Leu105, Met109, Met124 (see Fig. 5).
The basic centre is anchored to the carboxy group of Glu127 while the hydroxyl groups
stabilizes the complex making a charged reinforced hydrogen bond to negatively charged head of
Glu7. This same evidence is not achieved in the binding of lubeluzole [(S)-1] to P2, which might
therefore correspond to a secondary, or at least less relevant, binding site.
17

ACCEPTED MANUSCRIPT
Fig.5. Binding mode of lubeluzole to CaM. Side chains of Phe92, Leu105, Met109, and Met124 are displayed to
help interpretation, with C-alpha atoms being coloured in blue.
Overall this results into a free energy of binding (FEB) = –6.27 kcal/mol, corresponding to an
estimated pKi = 4.60 quite similar to the experimental pKd value. It’s interesting to note that the
chair binding conformation of the piperidine ring reproduces quite well the X-ray structure [47],
so there should be no interconversion phenomenon of the aliphatic cycle in the binding site.
It has been shown that the two hydrophobic binding pockets of C- and N-terminal domains,
formed only when Ca2+ ions are bound to CaM, are the key recognition sites for both inhibitors
and target enzymes [44], included Ca2+/calmodulin-dependent kinase II (CaMKII) [48]. This
evidence could explain the antagonist behaviour of lubeluzole on CaMKII.
18

ACCEPTED MANUSCRIPT
3. Conclusions
A FTPFACE method was developed and validated by application to reference set of potent CaM
ligands, among which the potent antagonist 125-C9 was ad hoc synthesized through an improved
synthetic procedure. The affinity values obtained for most of the studied compounds (Table S1)
were in agreement with those reported in the literature. These results demonstrate that the
method proposed might be used for routine affinity analyses to individuate potent CaM ligands.
The major advantages of this method are its efficiency, the possibility of testing both CaM and
ligands without labelling them, and finally the possibility of testing ligands without previous
separation from excipients. The method herein described was successful in disclosing a new
potent CaM ligand―lubeluzole. It was slightly less potent than 125-C9 (Kd = 2.9 ± 0.7 and 0.47
± 0.06 µM, respectively) and performed as a Ca2+/calmodulin-dependent kinase II (CaMKII)
inhibitor too. Docking studies based on the X-ray crystal structures of several trifluoperazine-
CaM complexes have been performed in order to explore possible binding modes of lubeluzole
to CaM and the main aminoacidic residues and interactions contributing to complex formation
were highlighted. An estimated dissociation constant in good agreement with the experimental
one was found. Thus, the ability of lubeluzole to antagonize CaM activities might contribute to
its observed chemosensitizing properties. However, given the difference observed between
potencies displayed in the above and the previously reported activities, further investigations are
required in order to achieve new insights into the complex mechanism behind lubeluzole
chemosensitizing action.
4. Experimental section
4.1. Capillary electrophoresis
19

ACCEPTED MANUSCRIPT
4.1.1. Apparatus
Capillary electrophoresis experiments were performed using a P/ACE MDQ Beckman
instrument (Palo Alto, CA, USA), equipped with a diode-array spectrophotometric detector. The
fused-silica capillary was 60 cm length (50 cm to the detector) and 50 µm i.d., and was
purchased from Quadrex corporation (Woodbridge, CT, U.S.A). The temperature of the capillary
was 25 °C. The samples were injected by pressure: 5 psi for 30 s for CaM, followed by 0.5 psi
for 5 s for ligands. The applied voltage was 20 kV. Before each run, the capillary was
conditioned for 3 min with 0.1 M NaOH, for 3 min with H2O, and for 5 min with the run buffer.
Detection was performed at 220 nm.
4.1.2. Chemicals
Bovine brain CaM was purchased from Calbiochem (Merck KGaA, Darmstadt, Germany), W-7
hydrochloride [N-(6-aminohexyl)-5-chloronaphthalene-1-sulfonamide], clozapine [8-chloro-11-
(4-methyl-1-piperazinyl)-5H-dibenzo[b,e][1,4]-diazepine], promethazine hydrochloride [10-[2-
(dimethylamino)propyl]phenothiazine hydrochloride], trifluoperazine dihydrochloride [10-[3-(4-
methylpiperazin-1-yl)propyl]-2-(trifluoromethyl)-10H-phenothiazine
dihydrochloride],
tris(hydroxymethyl)aminomethane (Tris), glycine, and HPLC grade ethanol were from Sigma-
Aldrich (Milwaukee, WI, USA); 0.1 M sodium hydroxide was from J. T. Baker (Deventer, The
Netherlands); sodium hydrogen phosphate, sodium dihydrogen phosphate, sodium borate, and
deionized water (conductivity ≤0.1 µS/cm) were from Carlo Erba (Milan, Italy). Loperamide
hydrochloride
[4-(p-chlorophenyl)-4-hydroxy-N,N-dimethyl-α,α-diphenyl-1-
piperidinebutyramide hydrochloride] (Imodium® gelcaps) was from Johnson & Johnson S.p.a.,
chlorpromazine hydrochloride [3-(2-chloro-10H-phenothiazin-10-yl)-N,N-dimethyl-propan-1-
20

ACCEPTED MANUSCRIPT
amine hydrochloride] (Prozin®, oral drops) was from Lusofarmaco S.p.a., fendiline
hydrochloride [N-(3,3-diphenylpropyl)-α-methylbenzylamine hydrochloride] and prenylamine
hydrochloride [N-(1-methyl-2-phenylethyl)-3,3-diphenylpropan-1-amine hydrochloride] were
synthesized in house by reduction of the product obtained condensating 3,3-diphenylpropionic
acid (Sigma-Aldrich) and 1-phenylethylamine (Sigma-Aldrich) or between 3,3-
diphenylpropionic acid (Sigma-Aldrich) and 1-methyl-2-phenylethylamine, respectively.
4.1.3. Preparation of samples and buffer
All the aqueous solutions were prepared by using deionized water. The buffer used for ACE
experiments was prepared by adding a 380 mM glycine solution to 50 mM Tris until pH 8.35
was reached. In the resultant solution, CaCl2 was added to reach 3 mM concentration. This buffer
was used to prepare a 10 mg/ml (600 µM) stock solution of CaM, and later to opportunely dilute
it to 1–0.025 mg/ml (60–1.25 µM). 10 mg/ml stock solutions of the ligands and the neutral
marker (mesityl oxide) were prepared in absolute ethanol. For testing loperamide CaM affinity,
the content of one Imodium® gelcap containing 2 mg of loperamide was taken up with 200 µL of
absolute ethanol; this mixture was sonicated for 20 min and then filtered. In the case of
chlorpromazine, the 40 mg/ml solution of Prozin® was diluted 1:4 with absolute ethanol.
Afterwards the sample stock solutions were diluted 1:100 with water.
Phosphate and borate buffers for pKa determinations were prepared by combining appropriate
amounts of 0.033 M Na2HPO4 and 0.033 M NaH2PO4 solutions, or 0.025 M Na2B4O7 and 0.025
NaOH solutions to achieve the proper pH in the range 6.80−9.80.
All the solutions were filtered through a PTFE mesh 0.20 µM porous size (Advantec MFS,
Dublin, CA, USA) and stored at 4 °C until usage.
21

ACCEPTED MANUSCRIPT
4.2. Chemistry
All chemicals were purchased from Sigma–Aldrich or Lancaster at the highest quality
commercially available. Solvents were RP grade unless otherwise indicated. Yields refer to
purified products and were not optimized. The structures of the compounds were confirmed by
routine spectrometric and spectroscopic analyses. Only spectra for compounds not previously
described are given. Melting points were determined on a Gallenkamp apparatus in open glass
capillary tubes and are uncorrected. Infrared spectra were recorded on a Perkin–Elmer (Norwalk,
CT) Spectrum One FT spectrophotometer and band positions are given in reciprocal centimeters
13
(cm–1). 1H and C NMR spectra were recorded on a Varian Mercury-VX spectrometer (Varian
1
13
Inc., Palo Alto, CA, USA), operating at 300 and 75 MHz for H and C, respectively, or on a
Agilent Technologies 500 MHz (Varian Inc., Palo Alto, CA, USA), operating at 500 and 126
MHz for 1H and 13C, respectively, using CDCl3 as solvent, unless otherwise indicated. Chemical
shifts are reported in parts per million (ppm) relative to solvent resonance: CDCl3, δ 7.26 (1H
NMR) and δ 77.3 (13C NMR). J values are given in Hz. EIMS spectra were recorded on a
Hewlett-Packard 6890-5973 MSD gas chromatograph/mass spectrometer (Hewlett-Packard, Palo
Alto, CA, USA) at low resolution. ESI+/–/MS/MS analyses were performed with an Agilent 1100
series LC-MSD trap system VL Workstation (Agilent, Palo Alto, CA, USA). Elemental analyses
were performed with a Eurovector Euro EA 3000 analyzer. Chromatographic separations were
performed on silica gel columns by flash chromatography (Kieselgel 60, 0.040–0.063 mm,
Merck, Darmstadt, Germany). TLC analyses were performed on precoated silica gel on
aluminum sheets (Kieselgel 60 F254, Merck).
22

ACCEPTED MANUSCRIPT
4.2.1. Acylation of N,N’-dibenzylethylenediamine (8)
To an ice-cold solution of N,N'-dibenzylethylenediamine (7, 2.0 g, 8.33 mmol) and DBU (0.89 g,
5.86 mmol) in dry CH2Cl2 (30 mL), a solution of 3-methoxybenzoyl chloride (0.71 g, 4.18
mmol) in dry CH2Cl2 (15 mL) was added dropwise under N2 atmosphere. The reaction mixture
was stirred at 0 °C for 90 min and then the solid was filtered off. The filtrate was diluted with
CH2Cl2 (40 mL) and washed twice with 2 M NaOH (40 mL) and twice with brine (20 mL). The
organic phase was dried (anhydrous Na2SO4) and concentrated under vacuum. The residue was
purified by flash chromatography (EtOAc) to give 8 and 9.
4.2.1.1. N-benzyl-N-[2-(benzylamino)ethyl]-3-methoxybenzamide (8)
1
Yellowish oil (0.79 g, 51%); IR (neat): 3317 (NH), 1630 (C=O) cm–1; H NMR (300 MHz): δ
1.48 (br s, 1H, exch D2O), 2.90 (br s, 2H), 3.32 (br s, 2H), 3.59 (br s, 2H), 3.72 (br s, 3H), 4.56
(br s, 2H), 6.92 (br d, J = 8.5 Hz, 1H), 6.95 (br s, 1H), 6.99 (dd, J = 7.4, 1.1 Hz, 1H), 7.10–7.21
(m, 1H), 7.22–7.40 (m, 10H); 13C NMR (500 MHz): δ 44.8 (1C), 46.7 (1C), 48.2 (1C), 53.6
(1C), 55.2 (1C), 112.0 (1C), 115.6 (1C), 118.8 (1C), 127.0 (1C), 128.0 (1C), 128.4 (4C), 128.8
(4C), 129.6 (1C), 137.7 (2C), 140.2 (1C), 159.6 (1C), 172.2 (1C); ESI+/MS m/z: 397 [M + Na+];
ESI+/MS/MS m/z: 268 (100).
4.2.1.2. N,N'-1,2-ethanediylbis[N-(phenylmethyl)-3-methoxybenzamide] (9)
White solid (0.36 g, 33%); mp: 130–131 °C; IR (KBr): 1631 (C=O) cm–1; 1H NMR (300 MHz):
δ 3.65 (s, 6H), 3.76 (s, 4H), 4.72 (s, 4H), 6.89 (d, J = 8.1 Hz, 2H), 6.99 (s, 2H), 7.04 (d, J = 7.2
Hz, 2H), 7.19 (apparent t, 2H), 7.22–7.50 (m, 10H); 1H NMR (toluene-d8, 500 MHz): δ 3.25 (s,
6H), 3.58 (s, 4H), 4.69 (s, 4H), 6.70 (d, J = 6.4 Hz, 2H), 6.90–7.15 (m, 14H), 7.24 (s, 2H); 13C
23

ACCEPTED MANUSCRIPT
NMR (300 MHz): δ 40.7 (2C), 52.5 (2C), 55.4 (2C), 111.9 (2C), 116.2 (2C), 119.1 (2C), 127.4
(2C), 127.9 (2C), 129.0 (6C), 129.9 (2C), 137.0 (2C), 137.6 (2C), 159.8 (2C), 172.6 (2C);
ESI+/MS m/z: 531 [M + Na+]; ESI+/MS/MS m/z: 268 (100); Anal. Calcd for
.
(C32H32N2O4 0.33H2O): C, 74.69; H, 6.40; N, 5.44. Found: C, 74.86; H, 6.32; N, 5.52.
4.2.2. Hydrolysis of 9
A solution of 9 (0.23 g, 0.45 mmol) in THF (4 mL) and 12 N HCl (10 mL) was stirred for 20 min
at 120 °C in a microwave reactor. When the reaction was completed, the mixture was cooled and
NaOH pellets were added. The aqueous phase was extracted three times with EtOAc and the
combined organic phases were dried (anhydrous Na2SO4) and concentrated under reduced
pressure. Purification of the residue by flash chromatography (EtOAc/petroleum ether 1:1, then
MeOH/EtOAc 1:9) gave 0.10 g (62%) of 8.
4.2.3. N-benzyl-N-{2-[benzyl(N,N-dimethylglycyl)amino]ethyl}-3-methoxybenzamide (10)
A solution of 8 (0.30 g, 0.80 mmol), N,N-dimethylglycine hydrochloride (0.11 g, 0.80 mmol) and
EEDQ (0.24 g, 0.96 mmol) in CHCl3 (20 mL) was stirred for 15 min at 100 °C in a microwave
reactor. After evaporation of the solvent, the residue was taken up with EtOAc, washed with 2 N
NaOH, and then with brine. The organic phase was dried over anhydrous Na2SO4 and
concentrated under reduced pressure. The residue was purified by flash chromatography
(EtOAc/petroleum ether 9:1, then EtOAc) to give 0.24 g (65%) of 10 as a yellow oil: IR (neat):
1
1635 (C=O) cm–1; H NMR (toluene-d8, 500 MHz): δ 2.10 (s overlapping solvent resonance,
6H), 2.18 (s, 2H), 3.22 (s, 2H), 3.33 (s, 3H), 3.56 (s, 2H), 4.60 (s, 2H), 4.76 (s, 2H), 6.73 (br d, J
= 7.3 Hz, 1H), 6.86 (br s, 1H), 6.94–7.20 (m, 11H), 7.31 (br s, 1H); 13C NMR (500 MHz): δ 41.3
24

ACCEPTED MANUSCRIPT
(1C), 42.0 (1C), 45.7 (2C), 50.3 (1C), 52.6 (1C), 55.2 (1C), 61.8 (1C), 111.9 (1C), 115.8 (1C),
118.8 (1C), 126.7 (1C), 127.1 (1C), 128.6 (4C), 128.9 (4C), 129.7 (1C), 136.9 (2C), 137.3 (1C),
159.6 (1C), 170.9 (1C), 172.2 (1C); ESI+/MS m/z: 482 [M + Na+]; ESI+/MS/MS m/z: 482 (100).
4.2.4. N,N'-dibenzyl-N-[2-(dimethylamino)ethyl]-N'-(3-methoxybenzyl)ethane-1,2-diamine (125-
C9)
To a stirred solution of 10 (0.24 g, 0.52 mmol) in dry THF (15 mL), LiAlH4 (0.20 g, 5.2 mmol)
was added under N2 atmosphere. The reaction mixture was heated at reflux for 20 h, then it was
cooled with an ice bath and quenched by the careful addition of cold water until the end of gas
evolution. The residue was removed by filtration and the filtrate additioned with water. The
aqueous phase was acidified with 2 M HCl, washed with EtOAc, then made alkaline with NaOH
pellets and extracted with EtOAc. The combined organic phases were dried (anhydrous Na2SO4)
1
and concentrated under reduced pressure to afford 125-C9 (0.20 g, 89%) as light-brown oil: H
NMR (500 MHz): δ 2.13 (s, 6H), 2.31 (dd, J = 8.6, 6.1 Hz, 2H), 2.49 (dd, J = 8.6, 6.1 Hz, 2H),
2.54–2.60 (m, 2H), 2.61–2.67 (m, 2H), 3.54 (s, 4H), 3.55 (s, 2H), 3.78 (s, 3H), 6.76 (dd, J = 7.6,
2.2 Hz, 1H), 6.91 (d, J = 7.8 Hz, 1H), 6.93 (s, 1H), 7.18–7.34 (m, 10H), 7.37 (br d, J = 5.4 Hz,
13
1H); C NMR (500 MHz): δ 45.8 (2C), 51.2 (1C), 52.2 (1C), 52.3 (1C), 55.1 (1C), 57.6 (1C),
58.79 (1C), 58.83 (1C), 59.4 (1C), 112.2 (1C), 114.1 (1C), 121.0 (1C), 126.76 (1C), 126.79 (1C),
128.10 (2C), 128.14 (2C), 128.7 (2C), 128.8 (2C), 129.1 (1C), 139.5 (1C), 139.7 (1C), 141.6
(1C), 159.6 (1C); MS (70 eV) m/z (%) 373 (M+ – 58, 42), 91 (100).
25

ACCEPTED MANUSCRIPT
4.2.5. N,N'-dibenzyl-N-[2-(dimethylamino)ethyl]-N'-(3-methoxybenzyl)ethane-1,2-diamine
hydrochloride (125-C9.3HCl)
125-C9 (0.20 g, 0.46 mmol) was dissolved in dry Et2O and treated with gaseous HCl for a few
seconds to give a white solid, which was recrystallized from abs EtOH/Et2O to afford 0.11 g of
1
white crystals (44%): mp 190–192 °C; H NMR (CD3OD, 500 MHz): δ 2.87 (s, 6H), 3.48 (br s,
2H), 3.50–3.70 (m, 6H), 3.83 (s, 3H), 4.16 (br s, 2H), 4.38 (s, 2H), 4.43 (s, 2H), 7.02 (dd, J =
8.3, 2.4 Hz, 1H), 7.12 (d, J = 7.3 Hz, 1H), 7.24 (s, 1H), 7.36 (apparent t, 1H), 7.40–7.64 (m,
10H); 13C NMR (500 MHz): δ 42.5 (2C), 48.2 (2C), 54.7 (2C), 57.8 (3C), 58.0 (1C), 116.0 (1C),
116.3 (1C), 123.2 (1C), 128.7 (1C), 128.9 (1C), 129.08 (1C), 129.14 (4C), 130.1 (1C), 130.2
(1C), 130.8 (1C), 131.4 (4C), 160.4 (1C); ESI+/MS m/z: 432 [M + H+]; ESI+/MS/MS m/z: 205
(100); Anal. Calcd for (C28H37N3O.3HCl.H2O): C, 60.16; H, 7.57; N, 7.52. Found: C, 60.50; H,
7.47; N, 7.77.
4.3. CaMKII activity assay
CaMKII activity was tested on Autocamtide in the presence of the tested compounds. In a first
reaction step, active recombinant full-length CaMKII (Signal Chem, La Jolla) was incubated for
30 min at 30 °C with 1 mmol/L CaCl2 and 5 µmol/L CaM in 50 µL of a reaction mixture (50
mmol/L HEPES pH 7.5, 10 mmol/L MgCl2, 0.5 mmol/L dithiothreitol (DTT), 100 nmol/L
microcystin, 0.1 mmol/L non-radiolabeled ATP) [49]. In a second reaction step, a 10 µL aliquot
from the first reaction was then incubated with 25 mM EGTA, 0.2 µCi/µl of Easy Tides
Adenosine 5’-triphosphate [γ32P]-ATP (Perkin Elmer) and 0.5 mM Autocamtide [50] in the
presence of the tested compounds (at different concentrations) in order to determine the effects
26

ACCEPTED MANUSCRIPT
of the compounds on CaMKII activity on its substrate Autocamtide. AntCaNtide at a
concentration of 5 µM was used as positive control of CaMKII inhibition. The reaction was
carried out for 30 min at 30 °C, then 20 µL aliquots of the reaction mixture were spotted onto
Whatman P-81 phosphocellulose paper. EGTA was added to quantify CaMKII autonomous
activity. Dried filters were counted on a Beckman LS 6000 scintillation counter.
4.4. Molecular modeling
The 2.0 Å resolution Ca2+/CaM-trifluoperazine 1:4 complex X-ray structure (entry code 1LIN)
was taken from the Protein Data Bank and passed to the Protein Preparation Wizard
implemented in the MAESTRO software package [MacroModel, version 9.9, Schrödinger, LLC,
New York, NY, 2012]. The complex was stripped out of solvent and ligands and all the
hydrogens were added and their positions refined; the point charges for each atom were
calculated according to AMBER force field [51]. As a result, protein affinity maps were
calculated with AUTOGRID as follow: in the first blind docking a 80×80×80 0.5 Å spaced cubic
box was built around the center of mass of the protein, while for the refinement dockings a more
spaced (0.375 Å) cage was indeed considered.
The lubeluzole X-ray (CSD code) was downloaded from the Cambridge Structural Database and
submitted to the energy minimization and electrostatic charges calculation according to the
quanto-mechanical method AM1 implemented in MAESTRO. The indirect method to manage
the ring as a fully flexible entity during the AUTODOCK conformation search was applied [52].
This protocol converts the cyclic ligand into its corresponding acyclic form by removing a bond.
27

ACCEPTED MANUSCRIPT
Bonds between two identical atom types (C-C) and resulting in shorter chains were preferred to
keep the calculation simple and improve the quality of the final results.
Due to the conformational freedom of the ligand, the maximum number of energy evaluations
was set to 25 millions. Lamarckian genetic algorithm (LGA) was carried out for 100 and 500
runs during the blind and refinement docking respectively. In the first 100 runs the tran0, quat0
and dihe0 were set to random, while in the latest 500 runs the same values were set to the values
suggested by the same LGA for the best FEB pose.
Acknowledgments
The authors thank the Ministero dell’Istruzione, dell’Università e della Ricerca (MIUR) (Project
PRIN 2010FPTBSH_005 “NANO Molecular tEchnologies for Drug delivery – NANOMED”)
for financial help. The authors are indebted to dr. Ottavio Ungaro for his invaluable help.
28