In vitro and in silico determination of glutaminyl cyclase inhibitors

Article abstract of DOI:10.1039/c9ra05763c

Alzheimer's disease (AD) is the most common form of neurodegenerative disease currently. It is widely accepted that AD is characterized by the self-assembly of amyloid beta (Aβ) peptides. The human glutaminyl cyclase (hQC) enzyme is characterized by assoc

Full text of DOI:10.1039/c9ra05763c

RSC Advances
View Article Online
View Journal | View Issue
PAPER
In vitro and in silico determination of glutaminyl
cyclase inhibitors†
Cite this: RSC Adv., 2019, 9, 29619
a
*
Phuong-Thao Tran,
Nguyen Thanh Tung
‡
Van-Hai Hoang,‡b Jeewoo Lee,c Tran Thi Thu Hien,d
e
fg
*
and Son Tung Ngo
Alzheimer's disease (AD) is the most common form of neurodegenerative disease currently. It is widely
accepted that AD is characterized by the self-assembly of amyloid beta (Ab) peptides. The human
glutaminyl cyclase (hQC) enzyme is characterized by association with Ab peptide generation. The
development of hQC inhibitors could prevent the self-aggregation of Ab peptides, resulting in impeding
AD. Utilizing structural knowledge of the hQC substrates and known hQC inhibitors, new heterocyclic
and peptidomimetic derivatives were synthesized and were able to inhibit the hQC enzyme. The
inhibiting abilities of these compounds were evaluated using
a fluorometric assay. The binding
mechanism at the atomic level was estimated using molecular docking, free energy perturbation, and
quantum chemical calculation methods. The predicted log(BBB) and human intestinal absorption values
indicated that these compounds are able to permeate the blood–brain barrier and be well-absorbed
through the gastrointestinal tract. Overall, 5,6-dimethoxy-N-(3-(5-methyl-1H-imidazol-1-yl)propyl)-1H-
benzo[d]imidazol-2-amine (1_2) was indicated as a potential drug for AD treatment.
Received 25th July 2019
Accepted 13th September 2019
DOI: 10.1039/c9ra05763c
rsc.li/rsc-advances
Introduction
Alzheimer's disease (AD) is known to be one of the most critical
types of dementia, affecting several million people worldwide.
There are ca. 10 million patients arising annually.1 The nancial
aDepartment of Pharmaceutical Chemistry, Hanoi University of Pharmacy, Hanoi,
Vietnam. E-mail: thaotp119@gmail.com
bInstitute of Research and Development, Duy Tan University, Da Nang 550000,
cost of treatment and Medicare for AD patients is rising
rapidly.1 AD is strongly associated with the self-aggregation of
amyloid beta (Ab) peptides,2,3 which are a heterogeneous
mixture of peptides having different solubility, stability, and
biological and toxic properties.3,4 These peptides are generated
from the single-pass transmembrane amyloid precursor protein
(APP) by several proteases at different sites. Several species of Ab
peptides have been observed, whereas Ab42 and Ab40 form the
majority.2 The self-assembly of Ab peptides produces several
products including random coils, oligomers, photobrils,
brils, and plaques.5–8 This process results in the impairment of
memory function and the loss of neurons and leads to synaptic
dysfunction.9–11
Pyroglutamate Ab peptides (AbpE) were found in the self-
aggregation of Ab peptides and act as initiators for Ab accu-
mulation. In fact, AbpE are only found in AD brains and
constitute approximately 50% of the total Ab.12 Their formation
is a multistep process requiring the loss of two or ten amino
acids to expose the N-terminal glutamate at the third or eleventh
position, followed by intramolecular dehydration of the exposed
glutamate. The Ab formation process results in the loss of three
or six charges for Ab3pE or Ab11pE, respectively, leading not
only to higher hydrophobicity but also to more rapid forming of
b-sheet structures and, thus, greater stability and aggregation
Vietnam
cLaboratory of Medicinal Chemistry, College of Pharmacy, Seoul National University,
Seoul, Korea
dVietnam University of Traditional Medicine, Hanoi, Vietnam
eInstitute of Materials Science, Vietnam Academy of Science and Technology, Hanoi,
Vietnam
fLaboratory of Theoretical and Computational Biophysics, Ton Duc Thang University,
Ho Chi Minh City, Vietnam. E-mail: ngosontung@tdtu.edu.vn
gFaculty of Applied Sciences, Ton Duc Thang University, Ho Chi Minh City, Vietnam
† Electronic supplementary information (ESI) available: Include Scheme 3,
chemistry, additional computational approach. Additional gures involve the
input of quantum chemical calculation to estimate detail of coordination link
and force constant between Zn(II) and its ligands, the input of quantum
chemical calculation to estimate atomic charges of the hQC active site
including Zn(II) via RESP approach, the time dependence of RMSD of soluble
complex over mimic time, the distribution of coordination links between Zn(II)
and its ligands, the modeling of the quantum chemical calculation the
interaction energy between compound 1_2 and the hQC active site including
Zn(II), and the modeling of the quantum chemical calculation the interaction
energy between compound 1_2 and the hQC active site without Zn(II).
Additional table about comparison of molecular docking results of available
inhibitors on hQC using Autodock Vina and Autodock4.2, the force constant
between Zn(II) and its ligands, detail of available compounds, and the binding
free energy of inhibitors 2_x to hQC enzyme. Additional data of 1H NMR & MS
spectra of the compounds. See DOI: 10.1039/c9ra05763c
‡ Contributed equally to the work.
This journal is © The Royal Society of Chemistry 2019
RSC Adv., 2019, 9, 29619–29627 | 29619

View Article Online
RSC Advances
Paper
propensity of the AbxpE. Moreover, newly formed lactam rings
are more stable under aminopeptidase medium than is full
length Ab.13 All of these factors result in AbpE acting as a seed
species for toxic Ab structural aggregation in the early stage of
AD.14
The human glutaminyl cyclase (hQC) is a zinc-dependent
enzyme that catalyzes the intramolecular cyclization of the N-
terminal glutamine residue into pyroglutamic acid-liberating
ammoniac. It is a key enzyme for some hormones post-
translation, protecting them from proteolytic degradation.15,16
It also converts N-truncated Ab into AbpE, a more toxic Ab
species.17 Therefore, hQC is regarded as an initiator agent for
pathological Ab accumulation, and inactive hQC can prevent
and treat AD. Some kinds of hQC inhibitors were discovered
with imidazole, triazole or benzimidazole as zinc-binding
motifs. Most of these inhibitors mimicked two or three amino
acids of the sequence of N-truncated Ab, NH2–Glu–Phe–Arg.
They all contain zinc-binding group, a hydrogen-bonding donor
and an aromatic group to interact with Phe325 in the pocket as
a critical amino acid for potent binding.18–22 Recently, a new
Fig. 2 (A) Urea/thiourea property issues and phase I metabolism
restrict BBB penetration; (B) structure of [11C]PBD150; (C) diagram of
binding pocket was reported with an orientation opposite to design new hQC inhibitors.
that of the Phe325 pocket, based on the discovery of SEN177.
Crystal studies of the SEN177–hQC complex showed two inter-
discovery.26 Moreover, thiourea group is unstable, it can be
metabolized by avin monooxygenase (FMO) to reactive species
which can alkylate proteins and nucleic acids.27 Therefore, they
are limited in their ability to penetrate through BBB and in in
vitro–in vivo correlation. In addition, compound [11C]PDB150,
the thiourea-containing hQC inhibitor with the potential
actions between the enzyme and the inhibitor in terms of
inhibitor orientation: the pyridine ring with Trp207 and the
uorine atom of 2-uoropyridine with the hydrogen atom of
His330 (ref. 23) (Fig. 1). However, up to now, only PQ-912 is
being studied in clinical trials and has completed a phase IIa
trial.24
In the present work, both computational and experimental
studies were performed to design two new potential inhibitory
hQC series based on structural knowledge of the hQC substrates
and known hQC inhibitors.19–22,25 To the best of our knowledge,
most potential hQC inhibitors contain substituted urea or
thiourea scaffolds. The potency of thiourea derivatives are
greater than that of corresponding ureas.18 Both thiourea and
urea types contribute not only two hydrogen bonding donors
(HBD) and three hydrogen bonding acceptors (HBA) but also
exible bonding (Fig. 2A). It looks like that these properties
cause blood–brain barrier (BBB) penetration issue based on
Hassan Pajouhesh's suggestion for CNS drug design and
Scheme 1 Synthetic scheme for 1_x series compounds. Conditions and
reagents: (a) KSCN, AcOH, Br2, 0 ꢀC to r.t., 12 h; (b) isoamyl nitrate, CuCl2,
MeCN, 0 ꢀC to r.t., 12 h; (c) Pd(OAc)2, P(Cy)3, NaOt-Bu, DMA, 100 ꢀC, 12 h;
(d) H2, Pd/C, MeOH/THF, r.t.; (e) 1-(3-isothiocyanatopropyl)-5-methyl-1H-
imidazole, THF/DMF (1/1), r.t., 2 h then EDC.HCl, 70 ꢀC, 1 h; (f) p-TSA, EtOH,
reflux, o.n.; (g) LAH, THF, 0 ꢀC to r.t., 1 h; (h) CBr4, Ph3P, THF, r.t., 2 h; (i) 5-
methyl-1-trityl-1H-imidazole, MeCN, reflux, 24 h then MeOH, THF, reflux,
Fig. 1 (A) Structure of the hQC substrate; (B) hQC inhibitor scaffold for o.n.; (k) N2H4$H2O, EtOH, r.t., o.n.; (n) 3,4-dimethoxyaniline, CDI, DCM-
the 1st pocket; (C) structure of SEN177.
DMF, r.t., o.n.; (m) PBr3, reflux, 2 h.
29620 | RSC Adv., 2019, 9, 29619–29627
This journal is © The Royal Society of Chemistry 2019

View Article Online
Paper
RSC Advances
inhibitory efficacy, Ki value of 60 nM (16), was determined to reported.29 The used buffer consisted of 25 mM HEPES (Sigma),
impossible permeation the BBB (Fig. 2B).28 Aer these consid- with pH 7.0, adjusted with HCl. The reaction mixture contained
erations, we designed the new hQC inhibitory series without 25 mL of substrate (0.4 mM), 50 mL of the test compound (0.016,
urea and thiourea scaffolds. In the rst series, thiourea groups 0.08, 0.4 and 2 mM), 25 mL of pGAPase (0.2 unit). Aer incuba-
were replaced by a bioisostere or a fused heterocyclic ring. In tion in 96-well black plates (Greiner, Austria) for 10 min at
particular, we also study the effect of rigid structures on the 37 ꢀC, the reaction was started by adding 50 mL of hQC solution
region of inhibitor–enzyme interaction. The second series was (0.04 mg mLꢂ1). The excitation/emission wavelengths at 380/
designed based on urea/thiourea scaffold removal combining 460 nm were correlated with the amount of ACM, the last
two amino acid residues of the terminal substrate. Be contained product of the enzymatic process.
a necessary nutrient for the brain and amino acid that they
would be present in brain uid through positive transport by
protein pumps (Fig. 2C). The hQC inhibiting abilities of these
compounds were determined using a uorometric assay with
a uorogenic substrate. The binding mechanisms of the imid-
Initial shape of glutaminyl cyclase and inhibitors
The structure of hQC was obtained from the Protein Data Bank
with ID: 3PBB30 referring to the previous work.31 The 3D-
structural inhibitors were built utilizing Gabedit package,32
azole derivatives to the hQC enzyme were then evaluated using
and these structures were optimized in the gas phase using
a combination of molecular docking, alchemical free energy,
quantum chemical calculations with hybrid functional B3LYP
at the 6-31G(d,p) level of the basic set.
and quantum chemical calculations. In particular, molecular
docking was employed to predict binding poses between the
trial inhibitors and hQC. The binding process was then rened
using conventional molecular dynamics (MD) simulations, and Molecular docking
the physical insights provided by these processes were claried
The binding poses and affinities trials of inhibitors for the hQC
using the free energy perturbation (FEP) method. The inuence
enzyme were estimated using Autodock4.2 packages.33 In
of Zn2+ to the ligand-binding affinity was characterized through
particular, the input le for the molecular docking simulations
quantum chemical calculation. The good agreement between
was prepared using AutodockTools 1.5.6.33 The docking grid
the computations and experiments indicates that 5,6-dime-
was chosen as the center of the hQC active site, with the size of
thoxy-N-(3-(5-methyl-1H-imidazol-1-yl)propyl)-1H-benzo[d]
imidazol-2-amine (1_2) is able to be used as a drug with high
potential in preventing AD.
˚
60 ꢁ 60 ꢁ 60 and the spacing of 0.375 A. The genetic algorithm
(GA) run was chosen as 50, with the population size set to 300.
The GA number of evaluations was set as 25 000 000, with the
number of generations at 27 000. The Zn2+ atom was also
involved during the docking simulation as mention in the
Fig. 4.
Materials and methods
Synthetic experiments
All chemical reagents were commercially available and used
directly without any purication. Silica gel column chroma-
tography was performed on silica gel 60 of 230–400 mesh from
Merck. 1H NMR spectra were recorded on a JEOL JNM-LA 300 at
300 MHz by Bruker Analytik or on a DE/AVANCE Digital 400 at
400 MHz by Bruker Analytik, with Me4Si as a reference standard.
Mass spectra were recorded on a VG Trio-2 GC-MS instrument
and a 6460 Triple Quad LC-MS instrument. All nal compounds
were assessed for purity by high performance liquid chroma-
tography (HPLC) on Agilent 1120 Compact LC (G4288A) system
via the following conditions. Column: Agilent TC-C18 column
(4.6 mm ꢁ 250 mm, 5 mm). Mobile phase A: 0.1% TFA in MeOH,
mobile phase B: 0.1% TFA in water (v/v) in 30 min. Wavelength:
254 nM. Flow: 0.7 to 1.0 mL minꢂ1. According to the HPLC
analyses, all nal compounds showed a purity of $95%. In
addition, detailed information of inhibitors was described in
the supplementary (ESI le†).
Atomistic molecular dynamics simulation
The hQC protein including Zn2+ ion was parameterized using
the AMBER99SB-ILDN force eld.34 In particular, the active site
of hQC including Zn(II) (Fig. S1 and S2†) was parameterized by
using a Python Base Metal Center Parameter Builder (MCPB.py)
package.35 The force constant and detail of coordination link
between Zn(II) and its ligands were determined as described in
Fig. S1 and Table S2.† The atomic charges was estimated via the
restrained electrostatic potential (RESP) method.36 The RESP
calculation was carried out by using quantum chemical calcu-
lation using B3LYP functional at 6-31G(d) level of basic set in
gas phase (Fig. S2†). The protonation state of hQC enzyme was
predicted via H++ server with pH condition was selected at 7.4.37
Moreover, binding poses between inhibitors and the hQC
enzyme were generated using Autodock4.2 (ref. 33) as
mentioned above. The inhibitors were represented by employ-
ing the general amber force eld (GAFF),38 wherein the atomic
charges were assigned by the restrained electrostatic potential
(RESP) method through quantum chemical calculations using
Bioactive assay
The synthesized compounds were evaluated for hQC inhibition the MP2 functional at the 6-31G(d,p) level of the basic set.36 The
by a uorometric assay with a uorogenic substrate, Glu- complexes were put into a dodecahedron periodic boundary
AMC$HBr (L-glutamine 7-amido-4-methylcoumarin, BACHEM, condition box with the volume of ꢃ497.30 nm3. Water mole-
Switzerland), and an auxiliary enzyme, pyroglutamyl peptidase cules were represented using the TIP3P water model.39 Na+ ions
(pGAPase, 50 units, Qiagen, Germany), as our previously were added to neutralize the soluble system.
This journal is © The Royal Society of Chemistry 2019
RSC Adv., 2019, 9, 29619–29627 | 29621

View Article Online
RSC Advances
Paper
The simulation parameters were obtained by referring to CNS mPO was calculated followed the previously published
previous publications.40 In particular, GROMACS 5.1.3 (ref. 41) method.49 The blood–brain barrier (BBB) crossing ability of
with GPGPU (general-purpose computing on graphics process- a ligand was predicted using the PreADME protocol,50 referring
ing units) acceleration was employed to simulate the complexed previous studies.40,51 The human intestinal absorption capacity
hQC-ligand in solution. The non-covalent bond pair cut-off was of a ligand was also estimated using the same application.50
of 0.9 nm. The electrostatic potential was treated by employing
the fast smooth particle-mesh Ewald electrostatic method with
a cut-off of 0.9 nm.42 The effect of van der Waals (vdW) inter-
actions of each pair was in the range of 0.9 nm. The rst step of
the simulation was energy minimization through the steepest
Results and discussion
Chemistry
descent method. Then, 500 ps of positional restraint simulation
in NVT and NPT ensembles followed. A conventional MD
simulation with a length of 50 ns was carried out to relax the
complex to the stable state. The system coordinates were
monitored every 10 ps. There were 3 independent MD trajec-
tories performed. The last snapshots of the conventional MD
To synthesize trial compounds for our study, we prepared 1-(3-
isothiocyanatopropyl)-5-methyl-1H-imidazole following the re-
ported literature19 (Scheme 3 of the ESI le†) to couple with
other intermediates as described in Scheme 1. Specically, 2-
aminobenzothiazole Ib was prepared from 3,4-dimethoxyani-
line under acidic conditions and using the active reagent
simulations were used in the initial free energy calculation with bromine. Then, the amino group of 2-aminobenzothiazole was
the FEP method.43
converted to chloride Ic under a Sandmeyer-like reaction for use
in the last step, which was a Buchwald–Hartwing reaction to
Free energy calculation
provide compound 1_1. Next, the diamino derivative IIb was
introduced thiocyanate to form an o-amino thiourea interme-
diate, which was cyclized under EDC conditions to afford
1_2.10,11 In the case of the unfused thiazole ring 1_3, the amino
group of IIIa was also converted to chloride IIIb under diazo-
nium conditions for aromatic nucleophilic substitution with Ia
to yield intermediate IIIc. This compound was reduced under
LAH, and Apple conditions were applied to generate a bromide
derivative, which was coupled with trityl protected 4(5)-imid-
azole to provide 1_3. The last compound 1_4, a 2-amino oxa-
diazole ring, was produced from the hydrazide carboxamide
intermediate IVd under dehydrated cyclization conditions.
The second series, peptidomimetic, was started with 2 amino
acids, phenylalanine Va and alanine, forming the new peptide
bond Vb (Scheme 2). The protected amino group Vb was
exposed to free amine Vc under acidic conditions for N-alkyl-
ation, which was activated prior by a sulfonyl group, to obtain
the intermediate Ve.12 Next, an imidazole group was introduced
for Ve, followed by removing the sulfonyl group through strong
nucleophilic substitution to afford the nal compound 3_1.
Last, compound 3_3 was prepared from 3_1 under ammonium
conditions.
Free energy calculations were carried out using the FEP
method,43 since it is one of the most accurate methods to date.44
The double annihilation-binding free energy method was per-
formed to determine the binding free energy between two
biomolecules, referring to a previous study.45 16 values of the
coupling parameter l were employed to modify the Hamilto-
nian over the computations. In particular, l was chosen as 0.00,
0.10, 0.20, 0.35, 0.50, 0.65, 0.80, and 1.00 to change the elec-
trostatic interactions, and l was chosen as 0.00, 0.10, 0.25, 0.35,
0.50, 0.65, 0.75, 0.90, and 1.00 referring the previous study.44
The total free energy change during the Hamiltonian alteration
was calculated using Bennet's acceptance ratio method.46 The
difference between the free energy changes during two
processes was determined by demolishing the ligand from the
solvated complex and isolating the systems corresponding to
the absolute binding free energy of the ligand to the enzyme.
Structural analysis
The surface charge distribution of proteins was estimated using
the Adaptive Poisson–Boltzmann Solver (APBS) package.47 The
sidechain (SC) contact was computed when the distances
between non-hydrogen atoms were smaller than 0.45 nm. The
hydrogen bond (HB) was calculated when the distance between
the donor (D) and acceptor (A) was smaller than 0.35 nm and
the angle between A–H–D was larger than 135ꢀ according to the
previous work.48 It should be noted that H indicates a hydrogen
atom.
Estimated physicochemical properties, mPO score, log(BBB)
and HIA values
The open bioactivity prediction online-site Molinspiration Scheme 2 Synthetic route for 3_x series compounds. Conditions and
reagents: L-alanine methyl ester hydrochloride, HOBt, EDC, TEA, MC,
(https://www.molinspiration.com/cgi-bin/properties) was used
for the topological polar surface area (TPSA) prediction. log P
and log D were predicted using Medchem Designer 5.5 Simu-
lation Plus. The Online-site Chemicalize of ChemAxon was used
r.t., 12 h; (b) TFA, TIPS, DMC, r.t., o.n.; (c) 2-nitrobenzenesulfonyl
chloride, TEA, DMC, 0 ꢀC to r.t., 4 h; (d) 1,3-dibromopropane, Cs2CO3,
DMF, 60 ꢀC, 6 h; (e) 5-methyl-1-trityl-1H-imidazole, MeCN, reflux,
24 h then MeOH, THF, reflux, o.n.; (f) thiophenol, K2CO3, MeCN, r.t.,
for pKa prediction (https://chemicalize.com/#/calculation). The o.n.; (g) NH3, MeOH, r.t., o.n.
29622 | RSC Adv., 2019, 9, 29619–29627
This journal is © The Royal Society of Chemistry 2019

View Article Online
Paper
RSC Advances
important role for potency, thus, the decreasing inhibitory
activities of thiazole and oxadiazole compounds can be associ-
ated with the lacking of the hydrogen bonding donor at the 1-N.
Surprisingly, the HBD at 1-N containing compounds, benzo-
thiazole (1_1) or benzimidazole (1_2) and peptidomimetics
(3_1, 3_3), were also less potent than 1. It leads to a hypothesis
that the structural modication causes to change the position
and key interaction of these inhibitors in enzymatic active site.
In particular, compound 1_2 (IC50 ¼ 0.11 ꢄ 0.03 mM) appeared
to be the most potent in this group, and it exhibited 2-folds
more potent than 1_1 (IC50 ¼ 0.30 ꢄ 0.09 mM), supporting the
fact that the hydrogen atom of –NH in benzimidazole (1_2)
might play a important role in favorable binding interaction,
but further investigations are required to conrm this state-
ment as the available data is inconclusive.
Biological activity
As mentioned above, the inhibiting abilities of the trial
compounds toward the hQC protein were estimated using
a binding assay, referring the previous study.29 Obtained results
are described in Table 1. Compounds with extremely low half-
maximal inhibitory concentrations (IC50 > 10 000 nM) are not
shown in Table 1 for clarity, since these compounds insigni-
cantly inuenced the hQC enzyme. All six compounds showing
in Table 1 emerge as potential inhibitors of the hQC protein,
with IC50 values ranging from 0.64 to 0.11 mM. However, when
comparison to reference thiourea compound 1,19 all of six
compounds showed a drop of potency from 3.7 to 22 times, even
though, they contained the similar structure to thiourea motifs,
such as: benzothiazole (1_1), benzimidazole (1_2), or the bio-
isosteres: 2-aminothiazole (1_3) and 2-amino oxadiazole (1_4).
As the previous study on SAR of thiourea derivatives (16), the
hydrogen bonding donor (HBD) at 1-N of thiourea played an
Molecular docking
The molecular docking approach is oen employed to rapidly
and roughly estimate the binding pose and binding affinity of
a ligand to a protein.52 Autodock Vina53 and Autodock4 (ref. 33)
are popularly employed to solve these problems. Although
Autodock Vina was indicated as providing a higher accuracy
with lower CPU consumption,53 Autodock4 showed a higher
performance in this study (described in ESI le†). Autodock4
was thus employed to determine the three-dimensional struc-
ture of the hQC-inhibitor complexes. The obtained binding
poses of ligands with the hQC protein are described in Fig. 3. All
of the ligands completely bound to the experimental active site
of the hQC enzyme. Normally, the docked complex adopts an
Table 1 IC50 values for inhibition of the hQC enzyme by compounds
1_x and 3_x
Cpd
Structure
IC50 (mM)
1_1
0.30 ꢄ 0.09
1_2
1_3
1_4
0.11 ꢄ 0.03
0.32 ꢄ 0.10
0.42 ꢄ 0.06
3_1
0.64 ꢄ 0.09
3_3
0.45 ꢄ 0.04
Fig. 3 Docked conformations of potential inhibitors to the active site
of the hQC enzyme (PDB ID: 3PBB) were obtained from molecular
docking simulations employing Autodock4.2 package.33 The Zn2+ ion
was mentioned as blue sphere.
117
0.029 ꢄ 0.004
This journal is © The Royal Society of Chemistry 2019
RSC Adv., 2019, 9, 29619–29627 | 29623

View Article Online
RSC Advances
Paper
appropriate conformation that can be used as an initial struc-
ture for atomistic MD simulations. Furthermore, the docking
energies are shown in Table 2. Although the correlation between
the predicted and experimental affinities was not very high
since the molecular docking method did not consider the
dynamics of the complexes, the coefficient was acceptable, with
the value of Rdock ¼ 0.54. The obtained results were thus rened
using all-atom MD simulations.
Renement of docking results using MD simulations
As mentioned above, the molecular docking approach contains
some limitations. The docked conformation was used as the initial
shape for the all-atom MD simulation. The MD simulation was
performed according to the description in the Materials & methods
section. Six complexes were independently mimicked three times
through MD simulations. The time dependence of the RMSD of the
six complexes is described in Fig. S3 of the ESI le.† The systems
mostly reached equilibrium states aer approximately 20 ns. The
coordination links between Zn(II) and its ligands are also stable aer
ꢃ20 ns (Fig. S4†). The last snapshots of MD simulations were used
as the initial conformations for FEP simulations.54 Moreover, six
systems involving isolated ligands in solution were also simulated
over three independent MD simulations of 2 ns in length to provide
stable snapshots of the solvated ligand system for FEP calculations.
The demolishing free energy of the ligands from two systems
including the solvated complex and ligand was estimated using the
BAR method.46 The different free energy is the binding free energy of
the ligand to the enzyme, shown in Table 2. A high correlation
coefficient between the experiments and FEP simulations was ob-
tained, with the value of R ¼ 0.92 (Fig. 4). On average, the theoretical
binding free energy (hDGFEPi z ꢂ7.30 kcal molꢂ1) was slightly
underestimated compared to that in the experimental study
(hDGEXPi z ꢂ8.89 kcal molꢂ1). Here, it is noted that the IC50 was
assumed as equal to the inhibition constant ki. This assumption
probably caused the difference between the calculated and experi-
mental values. Additionally, the difference may come from the
inaccuracy of simulating the interaction energy between the
component molecules including protein, ligand and solution
Fig. 4 The correlation between the experimental and computational
binding free energies.
the binding process of the ligand to the hQC enzyme. A compound
forming a larger vdW interaction with the hQC target probably has
a stronger binding affinity.
molecules.55,56 Furthermore, the vdW interaction free energy
vdW
DG
(zꢂ6.68 kcal molꢂ1 on average) dominates over the
FEP
Coulomb interaction free energy DGcFoEuP (zꢂ0.63 kcal molꢂ1) during
Table 2 Binding affinity of potential inhibitors. DEdock was provided
from Autodock Vina. The absolute binding free energy DGFEP was
obtained by the FEP method. DGEXP was approximately determined
when we assumed that IC50 was equal to the inhibition constant (ki).
The unit is of kcal molꢂ1
cou
FEP
vdW
FEP
System
DEdock
DG
DG
DGFEP
DGEXP
1_1
1_2
1_3
1_4
3_1
3_3
ꢂ5.1
ꢂ5.7
ꢂ5.7
ꢂ5.5
ꢂ4.9
ꢂ5.6
ꢂ0.08
ꢂ5.78
1.92
2.71
0.22
ꢂ9.64
ꢂ6.91
ꢂ9.20
ꢂ6.72
ꢂ3.88
ꢂ3.73
ꢂ9.72 ꢄ 0.41
ꢂ12.68 ꢄ 0.45
ꢂ7.28 ꢄ 0.47
ꢂ4.01 ꢄ 0.40
ꢂ3.66 ꢄ 0.41
ꢂ6.49 ꢄ 0.46
ꢂ8.94
ꢂ9.57
ꢂ8.92
ꢂ8.75
ꢂ8.50
ꢂ8.68
Fig. 5 The binding pose of compound 1_2 with the hQC enzyme
obtained over 50 ns of MD simulations. (A) The critical residues of hQC
form sidechain and hydrogen bond contacts with compound 1_2; (B)
the charged surface of the hQC enzyme in the binding mode with
compound 1_2.
ꢂ2.76
29624 | RSC Adv., 2019, 9, 29619–29627
This journal is © The Royal Society of Chemistry 2019

View Article Online
Paper
RSC Advances
Table 3 Predicted blood–brain barrier permeation and human intestinal absorption obtained using the PreADME application
2
˚
System
MW
log P
log D
pKa
pTSA (A )
mPO
log(BBB)
HIA (%)
1
1_1
1_2
1_3
1_4
3_1
3_3
CNS drugs
344.44
332.42
315.38
344.43
329.36
372.47
357.46
<450
2.25
2.91
2.12
3.02
2.17
1.33
0.43
<5
2.11
2.80
1.98
2.92
2.10
0.96
0.56
1–4
7.33
7.08
7.86
7.07
7.07
8.79
8.49
<10
60.35
61.21
77.00
51.21
87.25
85.26
99.48
<90
5.44
5.43
5.50
5.36
5.78
5.02
4.44
>4
ꢂ1.38
ꢂ0.24
ꢂ0.40
ꢂ1.24
ꢂ1.13
ꢂ1.44
ꢂ1.17
ꢂ2 to 1
94.72
90.74
89.30
97.15
95.94
94.27
84.95
Through the combination of theoretical and experimental of 3_1 and 3_3 are slightly larger than the suggestion value. pKa
results, compound 1_2 was indicated as the most potent value of all 1_x series compounds are below 8.0, indicating that
inhibitor of the hQC enzyme. In particular, although other they avoid possible interactions with P-gp.58 For mPO score, six
compounds formed strong vdW-interaction free energy and very compounds pass the recommendation score for CNS drug (mPO
weak electrostatic interaction energy with the hQC enzyme, the score > 4). In general, 1_x series and reference compound 1 are
cou
best inhibitor 1_2 adopted a balance of electrostatic (DG
¼
in of the rule of CNS drugs, conrmed their estimations in BBB
penetration below. However, PDB150, a 1 derivative, was
FEP
vdW
FEP
ꢂ5.78 kcal molꢂ1) and vdW interaction free energies (DG
¼
ꢂ6.91 kcal molꢂ1), as described in Table 2. In detail, compound demonstrated to impossible permeation the BBB,28 indicating
1_2 formed two HBs to residue E202 of the hQC protein by two the same result of 1 on BBB penetration study. Therefore,
hydrogen atoms of NH in thiourea motif (Fig. 5A). The new log(BBB) was employed for more BBB information. Moreover,
interaction of 2-N with enzyme caused its conformational the PreADME package50 was employed to predict the BBB-
change and can explain the similar activity of 1_1, 1_3 and 1_4, permeating ability, termed log(BBB). A compound is able to
and weaker than 1_2 on hQC. Consequently, 14 residues were cross the BBB when the log(BBB) is within the range from ꢂ2 to
found that they adopt SC contacts with the ligand (Fig. 5A). 1.59 The obtained results of the log(BBB) for the six compounds
Moreover, the charged surface of the hQC enzyme was analyzed are mentioned in Table 3. The predicted log(BBB) values ranged
using the APBS package, and the obtained result is shown in from ꢂ1.44 to ꢂ0.24, indicating that all of these compounds are
Fig. 5B. The active site of the hQC enzyme mostly adopted able to cross the BBB. Inhibitor 1_1 is the compound best
a negatively charged surface. The compound 1_2 ts well into capable of penetrating the BBB with the log(BBB) of ꢂ0.24.
the active site, although the net charge of the compound is zero. Meanwhile, inhibitor 1_2 also shows the ability to penetrate the
This may explain why the replacement of the S atom in BBB impressively, with the log(BBB) of ꢂ0.44. And, the low
compound 1_1 could signicantly increase the binding affinity log(BBB) of 1 (ꢂ1.38) can explain for low predicted BBB pene-
of compound 1_2 (Table S3†). Moreover, the quantum chemical tration. Moreover, the human intestinal absorption (HIA)
calculations were performed to roughly evaluate the effect of ion capability was also evaluated to predict the percentage of a drug
Zn2+ on the binding affinity of compound 1_2 to the hQC candidate absorbed through the gastrointestinal tract. The ob-
enzyme. The details of computations were mentioned in the ESI tained results indicated that all six inhibitors can be used as an
le.† The computational modeling was described in Fig. S2 and oral drug, with HIA values mostly larger than 89% (Table 3).
S3 of the ESI le.† The different energy between two cases, when
Zn2+ is appearance and disappearance, is approximate of
4.19 kcal molꢂ1, occupying ꢃ13% of total interaction energy.
Conclusions
Obtained results indicate that the strong inuence of Zn2+ on
ligand-affinity. However, the Zn2+ adopts repulsive force to the
ligand 1_2 indicating that the total net charge of the potential
ligand should not be positive. Overall, designing a new inhibitor
for the hQC enzyme should be carefully considered to address
this problem.
In vitro and in silico studies indicated that compound 1_2 is
a high-potential inhibitor of the hQC enzyme for the prevention
of AD. Although almost ligands formed strong vdW and weak
electrostatic interactions with the hQC enzyme, compound 1_2
adopted a balance of the two interactions. E202 is the most
important residue, forming a critical hydrogen bond with the
ligand 5,6-dimethoxy-N-(3-(5-methyl-1H-imidazol-1-yl)propyl)-
1H-benzo[d]imidazol-2-amine (1_2), which probably controls
the binding process of the complex. Furthermore, replacing
a chemical element of compound 1_2 with a positively charged
chemical element could improve the binding affinity the
designed ligand, but the total net charge of the ligand should be
zero to reduce repulsive force from ion Zn2+. The compound 1_2
was predicted to be able to permeate the BBB, and it can be used
An AD drug candidate is critically required to permeate the
blood–brain barrier (BBB) to transfer from the circulatory
system into the nervous system.57 Five physicochemical prop-
erties of six compounds, including MW, log P, log D, pKa, pTSA,
and multiparameter optimization (mPO) were calculated and
compared with compound 1 and suggested CNS drugs (Table
3).58 Six compounds have all molecular weight below 450 dalton,
log P below 5, log D in 1-4 range and pKa below 10. Compounds
1_1 to 1_4 are also in CNS drugs range of pTSA, meanwhile pTSA
This journal is © The Royal Society of Chemistry 2019
RSC Adv., 2019, 9, 29619–29627 | 29625

View Article Online
RSC Advances
Paper
¨
as an oral drug, as a high HIA was estimated. An in vivo study 20 D. Ramsbeck, M. Buchholz, B. Koch, L. Bohme,
should be carried out to conrm these results.
T. Hoffmann, H.-U. Demuth and U. Heiser, J. Med. Chem.,
2013, 56, 6613–6625.
21 P.-T. Tran, V.-H. Hoang, S. A. Thorat, S. E. Kim, J. Ann,
Y. J. Chang, D. W. Nam, H. Song, I. Mook-Jung, J. Lee and
J. Lee, Bioorg. Med. Chem., 2013, 21, 3821–3830.
22 V. T. H. Ngo, V.-H. Hoang, P.-T. Tran, N. Van Manh, J. Ann,
E. Kim, M. Cui, S. Choi, J. Lee, H. Kim, H.-J. Ha, K. Choi,
Y.-H. Kim and J. Lee, Bioorg. Med. Chem., 2018, 26, 3133–
3144.
Conflicts of interest
There are no conicts to declare.
Acknowledgements
This work was supported by Vietnam National Foundation for
Science & Technology Development (NAFOSTED) grant #104.01- 23 C. Pozzi, F. Di Pisa, M. Benvenuti and S. Mangani, J. Biol.
2017.06.
Inorg Chem., 2018, 23, 1219–1226.
24 P. Scheltens, M. Hallikainen, T. Grimmer, T. Duning,
A. A. Gouw, C. E. Teunissen, A. M. Wink, P. Maruff,
J. Harrison, C. M. van Baal, S. Bruins, I. Lues and
N. D. Prins, Alzheimer's Res. Ther., 2018, 10, 107.
References
1 Alzheimer's_association, Alzheimer's disease facts and gures
2018, 2018.
´
´
25 M. Szaszko, I. Hajdu, B. Flachner, K. Dobi, C. Magyar,
} ´ ´
I. Simon, Z. Lorincz, Z. Kapui, T. Pazmany, S. Cseh and
2 M. P. Murphy and H. LeVine, J. Alzheimer's Dis., 2010, 19, 311.
3 D. J. Selkoe and J. Hardy, EMBO Mol. Med., 2016, 8, 595–608.
´
G. Dorman, Mol. Diversity, 2017, 21, 175–186.
4 J. Nasica-Labouze, P. H. Nguyen, F. Sterpone, 26 H. Pajouhesh and G. R. Lenz, NeuroRx, 2005, 2, 541–553.
´
O. Berthoumieu, N.-V. Buchete, S. Cote, A. De Simone, 27 D. M. Ziegler, Annu. Rev. Pharmacol. Toxicol., 1993, 33, 179–
A. J. Doig, P. Faller, A. Garcia, A. Laio, M. S. Li, 199.
S. Melchionna, N. Mousseau, Y. Mu, A. Paravastu, 28 A. F. Brooks, I. M. Jackson, X. Shao, G. W. Kropog,
S. Pasquali, D. J. Rosenman, B. Strodel, B. Tarus,
J. H. Viles, T. Zhang, C. Wang and P. Derreumaux, Chem.
Rev., 2015, 115, 3518–3563.
5 H. W. Querfurth and F. M. LaFerla, N. Engl. J. Med., 2010,
362, 329–344.
P. Sherman, C. A. Quesada and P. J. H. Scott,
MedChemComm, 2015, 6, 1065–1068.
29 S. Schilling, T. Hoffmann, M. Wermann, U. Heiser,
C. Wasternack and H.-U. Demuth, Anal. Biochem., 2002,
303, 49–56.
6 S. T. Ngo, H. M. Hung, D. T. Truong and M. T. Nguyen, Phys. 30 K.-F. Huang, S.-S. Liaw, W.-L. Huang, C.-Y. Chia, Y.-C. Lo,
Chem. Chem. Phys., 2017, 19, 1909–1919.
7 P. H. Nguyen, J. M. Campanera, S. T. Ngo, A. Loquet and
P. Derreumaux, J. Phys. Chem. B, 2019, 123, 6750–6756.
8 S. T. Ngo, P. Derreumaux and V. V. Vu, J. Phys. Chem. B, 2019,
123, 2645–2653.
Y.-L. Chen and A. H.-J. Wang, J. Biol. Chem., 2011, 286,
12439–12449.
31 V.-H. Hoang, P.-T. Tran, M. Cui, V. T. H. Ngo, J. Ann, J. Park,
J. Lee, K. Choi, H. Cho, H. Kim, H.-J. Ha, H.-S. Hong, S. Choi,
Y.-H. Kim and J. Lee, J. Med. Chem., 2017, 60, 2573–2590.
32 A.-R. Allouche, J. Comput. Chem., 2011, 32, 174–182.
9 D. J. Selkoe, Science, 2002, 298, 789–791.
10 U. Sengupta, A. N. Nilson and R. Kayed, EBioMedicine, 2016, 33 G. M. Morris, R. Huey, W. Lindstrom, M. F. Sanner,
6, 42–49.
R. K. Belew, D. S. Goodsell and A. J. Olson, J. Comput.
Chem., 2009, 30, 2785–2791.
34 K. Lindorff-Larsen, S. Piana, K. Palmo, P. Maragakis,
J. L. Klepeis, R. O. Dror and D. E. Shaw, Proteins: Struct.,
Funct., Bioinf., 2010, 78, 1950–1958.
11 J. Zhao, R. Nussinov and B. Ma, J. Biol. Chem., 2017, 292,
18325–18343.
12 Y.-M. Kuo, M. R. Emmerling, A. S. Woods, R. J. Cotter and
A. E. Roher, Biochem. Biophys. Res. Commun., 1997, 237,
188–191.
35 P. Li and K. M. Merz, J. Chem. Inf. Model., 2016, 56, 599–604.
13 A. P. Gunn, C. L. Masters and R. A. Cherny, Int. J. Biochem. 36 C. I. Bayly, P. Cieplak, W. Cornell and P. A. Kollman, J. Phys.
Cell Biol., 2010, 42, 1915–1918.
Chem., 1993, 97, 10269–10280.
14 F. Chiti and C. M. Dobson, Annu. Rev. Biochem., 2017, 86, 27– 37 R. Anandakrishnan, B. Aguilar and A. V. Onufriev, Nucleic
68. Acids Res., 2012, 40, W537–W541.
15 W. H. Busby, G. E. Quackenbush, J. Humm, W. W. Youngblood 38 J. Wang, R. M. Wolf, J. W. Caldwell, P. A. Kollman and
and J. S. Kizer, J. Biol. Chem., 1987, 262, 8532–8536. D. A. Case, J. Comput. Chem., 2004, 25, 1157–1174.
16 W. H. Fischer and J. Spiess, Proc. Natl. Acad. Sci. U. S. A., 39 W. L. Jorgensen, J. Chandrasekhar, J. D. Madura,
1987, 84, 3628–3632.
R. W. Impey and M. L. Klein, J. Chem. Phys., 1983, 79, 926–
935.
40 S. T. Ngo, X.-C. Luu, N. T. Nguyen, V. V. Vu and
H. T. T. Phung, PLoS One, 2018, 13, e0204026.
17 S. Schilling, T. Hoffmann, S. Manhart, M. Hoffmann and
H.-U. Demuth, FEBS Lett., 2004, 563, 191–196.
18 M. Buchholz, U. Heiser, S. Schilling, A. J. Niestroj, K. Zunkel
and H.-U. Demuth, J. Med. Chem., 2006, 49, 664–677.
´
41 M. J. Abraham, T. Murtola, R. Schulz, S. Pall, J. C. Smith,
¨
19 M. Buchholz, A. Hamann, S. Aust, W. Brandt, L. Bohme,
B. Hess and E. Lindahl, SowareX, 2015, 1–2, 19–25.
T. Hoffmann, S. Schilling, H.-U. Demuth and U. Heiser, J. 42 T. Darden, D. York and L. Pedersen, J. Chem. Phys., 1993, 98,
Med. Chem., 2009, 52, 7069–7080.
10089–10092.
29626 | RSC Adv., 2019, 9, 29619–29627
This journal is © The Royal Society of Chemistry 2019

View Article Online
Paper
RSC Advances
43 R. W. Zwanzig, J. Chem. Phys., 1954, 22, 1420–1426.
44 S. T. Ngo, D. T. Truong, N. M. Tam and M. T. Nguyen, J. Mol.
Graphics Modell., 2017, 76, 1–10.
drug-like properties, EuroQSAR 2002 Designing Drugs and
Crop Protectants: processes, problems and solutions,
Blackwell Publishing, Massachusetts, USA, 2003, pp. 418–
420.
45 S. T. Ngo, S.-T. Fang, S.-H. Huang, C.-L. Chou, P. D. Q. Huy,
M. S. Li and Y.-C. Chen, J. Chem. Inf. Model., 2016, 56, 1344– 51 S. T. Ngo and M. S. Li, Mol. Simul., 2013, 39, 279–291.
1356.
52 M. W. Chang, C. Ayeni, S. Breuer and B. E. Torbett, PLoS One,
2010, 5, e11955.
46 C. H. Bennett, J. Comput. Phys., 1976, 22, 245–268.
47 N. A. Baker, D. Sept, S. Joseph, M. J. Holst and 53 O. Trott and A. J. Olson, J. Comput. Chem., 2010, 31, 455–461.
J. A. McCammon, Proc. Natl. Acad. Sci. U. S. A., 2001, 98, 54 L. Tran, S. T. Ngo and M. T. Nguyen, Chem. Phys. Lett., 2018,
10037–10041.
48 S. T. Ngo, H. T. Thu Phung, K. B. Vu and V. V. Vu, RSC Adv., 55 H. Zhang, C. Yin, Y. Jiang and D. van der Spoel, J. Chem. Inf.
2018, 8, 41705–41712. Model., 2018, 58, 1037–1052.
49 T. T. Wager, X. Hou, P. R. Verhoest and A. Villalobos, ACS 56 H. Zhang, Y. Jiang, Z. Cui and C. Yin, J. Chem. Inf. Model.,
Chem. Neurosci., 2010, 1, 435–449. 2018, 58, 1669–1681.
696, 55–60.
50 S. K. Lee, I. H. Lee, H. J. Kim, G. S. Chang, J. E. Chung and 57 P. Garg and J. Verma, J. Chem. Inf. Model., 2006, 46, 289–297.
K. T. No, The PreADME approach: web-based program for 58 Z. Rankovic, J. Med. Chem., 2015, 58, 2584–2608.
rapid prediction of physico-chemical, drug absorption and 59 D. E. Clark, J. Pharm. Sci., 1999, 88, 815–821.
This journal is © The Royal Society of Chemistry 2019
RSC Adv., 2019, 9, 29619–29627 | 29627