The translocator protein ligands as mitochondrial functional modulators for the potential anti-Alzheimer agents

Article abstract of DOI:10.1080/14756366.2021.1900158

Small molecule modulators of mitochondrial function have been attracted much attention in recent years due to their potential therapeutic applications for neurodegenerative diseases. The mitochondrial translocator protein (TSPO) is a promising target for such compounds, given its involvement in the formation of the mitochondrial permeability transition pore in response to mitochondrial stress. In this study, we performed a ligand-based pharmacophore design and virtual screening, and identified a potent hit compound, 7 (VH34) as a TSPO ligand. After validating its biological activity against amyloid-β (Aβ) induced mitochondrial dysfunction and in acute and transgenic Alzheimer’s disease (AD) model mice, we developed a library of analogs, and we found two most active compounds, 31 and 44, which restored the mitochondrial membrane potential, ATP production, and cell viability under Aβ-induced mitochondrial toxicity. These compounds recovered learning and memory function in acute AD model mice with improved pharmacokinetic properties.

Full text of DOI:10.1080/14756366.2021.1900158

Journal of Enzyme Inhibition and Medicinal Chemistry
ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/ienz20
The translocator protein ligands as mitochondrial
functional modulators for the potential anti-
Alzheimer agents
TaeHun Kim, Mohammad N. Morshed, Ashwini M. Londhe, Ji W. Lim, Ha E.
Lee, Suengmok Cho, Sung J. Cho, Hayoung Hwang, Sang M. Lim, Jae Y. Lee,
Jiyoun Lee & Ae N. Pae
To cite this article: TaeHun Kim, Mohammad N. Morshed, Ashwini M. Londhe, Ji W. Lim, Ha E.
Lee, Suengmok Cho, Sung J. Cho, Hayoung Hwang, Sang M. Lim, Jae Y. Lee, Jiyoun Lee & Ae N.
Pae (2021) The translocator protein ligands as mitochondrial functional modulators for the potential
anti-Alzheimer agents, Journal of Enzyme Inhibition and Medicinal Chemistry, 36:1, 831-846, DOI:
10.1080/14756366.2021.1900158
To link to this article: https://doi.org/10.1080/14756366.2021.1900158
© 2021 The Author(s). Published by Informa
UK Limited, trading as Taylor & Francis
Group.
View supplementary material
Submit your article to this journal
View related articles
Published online: 23 Mar 2021.
Article views: 472
View Crossmark data
Full Terms & Conditions of access and use can be found at
https://www.tandfonline.com/action/journalInformation?journalCode=ienz20

JOURNAL OF ENZYME INHIBITION AND MEDICINAL CHEMISTRY
2021, VOL. 36, NO. 1, 831–846
https://doi.org/10.1080/14756366.2021.1900158
RESEARCH PAPER
The translocator protein ligands as mitochondrial functional modulators for the
potential anti-Alzheimer agents
a,b,c
a,b
a,d
TaeHun Kima,b , Mohammad N. Morshed
, Ashwini M. Londhe
, Ji W. Lim
, Ha E. Leea,b, Suengmok Cho
ꢀ
ꢀ
ꢀ
ꢀ
, Sung J. Chof, Hayoung Hwangf, Sang M. Lima,b , Jae Y. Leea,d, Jiyoun Leeg and Ae N. Paea,b,d
e
aConvergence Research Center for Diagnosis, Treatment and Care System of Dementia, Korea Institute of Science and Technology, Seoul,
b
Republic of Korea; Division of Bio-Medical Science and Technology, KIST School, Korea University of Science and Technology, Seoul, Republic
c
d
of Korea; Center for Advanced Research in Sciences (CARS), University of Dhaka, Dhaka, Bangladesh; KHU-KIST Department of Converging
e
Science and Technology, Kyung Hee University, Seoul, Republic of Korea; Department of Food Science and Technology, Pukyong National
University, Pusan, Republic of Korea; New Drug Development Center, Daegu-Gyeongbuk Medical Innovation Foundation (DGMIF), Daegu,
Republic of Korea; Department of Global Medical Science, Sungshin University, Seoul, Republic of Korea
f
g
ABSTRACT
ARTICLE HISTORY
Received 28 October 2020
Revised 26 February 2021
Accepted 2 March 2021
Small molecule modulators of mitochondrial function have been attracted much attention in recent years
due to their potential therapeutic applications for neurodegenerative diseases. The mitochondrial translo-
cator protein (TSPO) is a promising target for such compounds, given its involvement in the formation of
the mitochondrial permeability transition pore in response to mitochondrial stress. In this study, we per-
formed a ligand-based pharmacophore design and virtual screening, and identified a potent hit com-
pound, 7 (VH34) as a TSPO ligand. After validating its biological activity against amyloid-b (Ab) induced
mitochondrial dysfunction and in acute and transgenic Alzheimer’s disease (AD) model mice, we devel-
oped a library of analogs, and we found two most active compounds, 31 and 44, which restored the
mitochondrial membrane potential, ATP production, and cell viability under Ab-induced mitochondrial tox-
icity. These compounds recovered learning and memory function in acute AD model mice with improved
pharmacokinetic properties.
KEYWORDS
Alzheimer’s disease;
mitochondrial function
modulators; translocator
protein ligand
1. Introduction
(DWm)11. More importantly, several TSPO ligands showed neuro-
protective effect by decreasing microglial activation in a rat model
of neurodegeneration, suggesting that TSPO may serve as a
potential therapeutic target12. In particular, it has been reported
that TSPO ligands may act on the formation of the mitochondrial
permeability transition pore (mPTP), which causes mitochondrial
swelling and cell death by allowing the passage of ions and pro-
teins between the mitochondria and the cytosol13,14. Although,
the exact structure of the mPTP still remains elusive, many studies
indicate that the mPTP is a multiprotein complex composed of
Alzheimer’s disease (AD) is a late-onset, progressive, age-depend-
ent neurodegenerative disorder, characterised clinically by the
impairment of cognitive functions and changes in behaviour and
personality1. Among several hypothesis regarding the AD patho-
genesis, mitochondrial dysfunction has been reported as an early
event observed in AD2. Specifically, recent studies suggested that
amyloid-b (Ab) has deleterious effects on mitochondrial function
and may contribute to respiratory chain impairment, neuronal
apoptosis, and generation of reactive oxygen species (ROS) in
AD3,4. While the mechanisms underlying Ab-induced mitochon-
drial stress is still under debate, emerging evidences indicate that
the maintenance of mitochondrial function may provide a poten-
cyclophilin
D (CypD), the voltage-dependent anion channel
1(VDAC1), and TSPO. Compounds targeting CypD15,16 and TSPO14
have demonstrated neuroprotective activity by blocking the for-
mation of the mPTP, specifically in Ab-induced mitochondrial dys-
function17,18. In addition, the administration of TSPO ligands
recovered DWm and ATP production, disrupted by the treatment
of Ab in vitro19, but also improved cognitive function of transgenic
AD model mice20. Another recent study also suggest that TSPO
ligands may inhibit ROS generation by blocking the interaction
between TSPO and VDAC121, supporting that TSPO is a viable tar-
tial treatment option for AD5,6
.
As a therapeutic strategy against AD, several mitochondrial
proteins have been highlighted as promising drug targets7,8. In
particular, the 18 kDa translocator protein (TSPO) is an evolution-
ally conserved protein located in the outer mitochondrial mem-
brane (OMM), and is known to play crucial roles in
steroidogenesis, apoptosis, immune response, and Ca2þ homeo-
stasis9,10. TSPO is also involved in mitochondrial functions such as
get for the treatment AD22,23
In this work, we aim to find novel TSPO ligands that can
.
ATP production, ROS generation, and mitochondrial polarisation restore mitochondrial function in cellular and animal models of
CONTACT Ae N. Pae
anpae@kist.re.kr
Convergence Research Center for Diagnosis, Treatment and Care System of Dementia, Korea Institute of Science and
Technology, Seoul, 02792, Republic of Korea; Jiyoun Lee
Republic of Korea
jlee@sungshin.ac.kr
Department of Global Medical Science, Sungshin University, Seoul, 01133,
ꢀ
These authors contributed equally to this work.
Supplemental data for this article can be accessed here.
ß 2021 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use,
distribution, and reproduction in any medium, provided the original work is properly cited.

832
T. KIM ET AL.
AD. Previously developed TSPO ligands mostly consist of either a 2.2. Virtual screening using an external library
benzodiazepine or a steroid scaffold, because TSPO was initially
Virtual screening was carried out by combining two HipHop
thought to be expressed only in the periphery with a propensity
to bind benzodiazepines such as diazepam24,25. Given that the
structures of the bacterial and mouse TSPO have been identified
very recently26,27, we believe that this new structural information
will allow us to find TSPO ligands with a novel scaffold that can
be readily optimised for further therapeutic development. To this
end, we performed a ligand-based pharmacophore modelling, fol-
lowed by a virtual screening as described in Figure 1. We identi-
fied several hit compounds, and performed in vitro mitochondrial
functional assays to evaluate biological activity. We further opti-
mised these hit compounds via a similarity search and in vitro
assays, and found two most active compounds, 7 (VH034) and 14
(VH062). Next, we designed and synthesised a library of TSPO
ligands based on the structures of the two compounds. We eval-
uated biological activities of the synthesised ligands in vitro and
in vivo to identify novel TSPO ligands that can recover neuronal
cells from Ab-induced mitochondrial dysfunction, but also improve
cognitive impairment in AD model mice. We also performed phar-
macokinetic analysis and docking studies to validate the activity
of the identified ligands.
Refine models and one shape-added pharmacophore model to
obtain new compounds with desired activity profiles. The com-
mercial library Asinex (AsinexGold, 229,398 compounds,
AsinexPlatinum, 125,231 compounds, Asinex, Moscow, Russia,
www.asinex.com) and ChemDiv (693,042 compounds, ChemDiv,
Inc., San Diego, CA, USA, www.chemdiv.com) have been utilised
for virtual screening. The best pharmacophore model was used
for the virtual screening experiment, by selecting the fast flexible
database search option. From the initial screening, compounds
with similar functional and spatial properties defined in 3D
pharmacophore query were selected on the basis of their fit val-
ues. Compounds obtained by fit values were further filtered by
visual inspection. The selected compounds were purchased for
biological assays.
2.3. Synthesis
All reagents were obtained from commercial sources and used
without further purification. All reactions were performed under a
nitrogen atmosphere in oven-dried glassware. Reactions were
monitored by analytical thin-layer chromatography (TLC) plates
(Merck, catalog no. 1.05715) with spots visualised by UV light
(k ¼ 254 nm) or using a KMnO4 solution. Solvents were evaporated
using a rotary evaporator under a reduced pressure of ꢁ50 mBar.
The reaction products were purified by flash column chromatog-
raphy using silica gel 60 (Merck, catalog nos. 1.07734). Melting
points were determined using an OptiMelt melting point appar-
2. Materials and methods
2.1. Generation and validation of common feature
pharmacophore model
Six TSPO ligands were collected from Integrity, and four TSPO atus (Stanford Research System, Inc.) in open capillary tube with-
1
ligands from the Integrity database of Prous (www.prous.com) out correction. H (300 or 400 MHz) and 13C (75 or 100 MHz) NMR
were selected for HipHop pharmacophore analysis. The conform- spectra were recorded using tetramethylsilane (TMS) as the
internal standard. Chemical shifts (d) are reported in parts per mil-
lion (ppm) values relative to TMS, and the coupling constants (J)
are reported in hertz (Hz). The purity (ꢂ95%) of the samples was
determined by analytical HPLC using a Waters E2695 system with
SunFire C18 column (4.6 mm ꢃ 150 mm; 5 mm). HPLC data were
recorded using the parameters as follows: H2O/MeCN, 90/10 ! 0/
100 in 20 min, þ3 min isocratic, flow rate of 1.0 ml/min, k ¼ 254
and 280 nm. High–resolution mass spectra (HRMS) were recorded
on an LTQ Orbitrap (Thermo Electron Corporation) instrument.
Reaction yields are for purified products. Detailed experiment pro-
cedures and characterisation of all synthesised compounds can be
found in the Supporting Information.
ational models of selected compounds having up to 250 conform-
ers were built using the “best conformer generation” method with
a 20 kcal/mol energy cut-off. HipHop pharmacophore models were
derived by comparing a set of conformational models and a num-
ber of three-dimensional configurations of chemical features
shared among the training set molecules. The parameter settings
of Maximum Omitted Features, Misses, and Complete Misses were
varied to generate multiple hypotheses in which some com-
pounds may or may not fit all features. Three chemical functions
with hydrogen bonding acceptor, hydrogen bonding donor, and
hydrophobic group were used as the pharmacophore features.
The hypothesis generation process in Catalyst was returned 10
possible pharmacophore hypotheses having different arrange-
ments of constituent features or ranking scores. After deleting the
redundant hypotheses that have the same chemical characteristics
and nearly the same distances between these functions, diverse
2.4. Cell culture
HT-22 (mouse hippocampal cells) cells were grown in Dulbecco’s
configurations of hypotheses were selected according to ranking Modified Eagle’s Medium (DMEM, GIBCO) supplemented with 10%
scores and fitting scores. The best hypothesis was determined by (vol/vol) FBS and antibiotics (100 mg/mL penicillin/streptomycin
comparing the highest fit value of training set compounds.
mix) in a humidified atmosphere at 37 ꢄC with 5% CO2.
Figure 1. Virtual screening process to identify the best hit compound.

JOURNAL OF ENZYME INHIBITION AND MEDICINAL CHEMISTRY
833
2.5. Preparation of Ab1–42
diphenyltetrazolium bromide, Sigma-Aldrich), incubated for 2 h at
37 ꢄC, and were treated with 135 mL of MTT solubilising solution
(10% Triton-X 100 in isopropanol with 0.1 M HCl) for 2 h at 37 ꢄC.
The % values of cell viability were determined by using the optical
density (OD) values at 560 nm, and normalised by taking the
vehicle control (5 mM of DMSO) as 0%.
To prepare a homogenous solution of monomeric Ab, Ab1–42 pep-
tide (American Peptide) was mixed with 1 ml of 1,1,1,3,3,3 hexa-
fluoro-2-propanol (HFIP; Sigma–Aldrich) and vortexed gently.
Then, the solution was aliquoted into microcentrifuge tubes and
lyophilised to prepare the Ab1–42 samples as dried peptide films.
The prepared Ab1–42 samples were stored at ꢅ80 ꢄC and used
immediately for cell-based assays to minimise possible aggrega-
tion of Ab itself. Before the use of Ab1–42 samples, the amount of
protein was quantified by the BCA protein assay (Pierce BCA pro-
tein assay kit; Thermo Scientific). The prepared Ab1–42 samples
were suspended with anhydrous dimethyl sulfoxide (DMSO) to a
concentration of 5 mM, and used for each cell-based assay.
2.9. CYP inhibition assay
Human CYP450 (CYP1A2, 2D6, 2C9, 2C19, and 3A4) activities were
R
determined by using the VividV CYP450 screening kit (Invitrogen,
Madison, WI, USA) in a clear 96-well plate. Five positive controls
were used as 10 mM solutions in MeCN for five different CYP450
enzymes: CYP1A2 (a-naphthoflavone), CYP2D6 (quinidine), CYP2C9
(sulfaphenazole), CYP2C19 (lansoprazole), and CYP3A4 (ketocon-
azole). Each sample (test compounds, positive inhibition controls,
and solvent controls) was added into each well with the Master
2.6. JC-1 assay
Thirty thousand HT-22 cells per well were seeded into a clear 96-
well plate (FALCON) at 200 mL per well one day prior to assay.
750 mM of JC-1 (Stratagene) in DMSO stock solution was dissolved
into phenol red-free Opti-MEM (GIBCO) medium to make final
concentration of 7.5 mM JC-1 per well. Medium was removed
from the plate, and 100 mL per well of JC-1 was added. The sam-
ple plates were incubated for 1 h and 15 min at 37 ꢄC and washed
twice with 100 mL per well of PBS. Subsequently, cells were treated
with a 25 mL solution of each compound at 10 mM in Opti-MEM
and incubated at 37 ꢄC for 10 min followed by addition of a 25 mL
of Ab (American peptide, 1-42) solution at 10 mM. Fluorescence
was measured at every one hour for three hours at ex/em
530 nm/580 nm (“red”) and ex/em 485 nm/530 nm (“green”) by
R
Pre-Mix [CYP450 BACULOSOMESV Reagent (recombinant human
CYP450 isozyme and rabbit NADPHP450 reductase) and
Regeneration System (3.3 mM glucose-6-phosphate and 0.3 U/ml
glucose-6-phosphate dehydrogenase in 100 mM potassium phos-
phate, pH 8.0)]. The mixture was incubated for 5 min at 37 ꢄC, and
R
the VividV CYP substrates and 0.1 mM NADPþ buffer were added
to initiate the CYP450 enzyme reaction. The % remaining activities
of CYP450 were measured after 20 min by using a fluorescent
R
plate reader (FlexstationV 3, Molecular Devices, USA).
2.10. Measurement of liver microsomal stability
R
using the FlexstationV 3 (Molecular Devices, USA) reader. The ratio
Liver microsomes (human, dog, rat, and mouse; 0.5 mg/mL) were
incubated with 1 mM of each test compound in the presence of
potassium phosphate buffer (PPB) for 5 min at 37 ꢄC. A solution of
the NADPH regeneration system (#44332000, Oriental Yeast Co.,
Japan) was added to begin metabolic reactions of liver micro-
somes. The resulting mixtures were incubated for 30 min at 37 ꢄC,
and the reaction was terminated by the addition of a cold MeCN
solution containing an internal standard (chlorpropamide). The
mixture was separated by centrifugation (14,000 rpm for 5 min at
4 ꢄC), and the supernatant was analysed by LC-MS/MS (Shimadzu
Nexera XR system and TSQ vantage; Thermo Scientific).
of green to red fluorescence was calculated and normalised by
taking the percent changes using vehicle control as 100%.
2.7. Luciferase-based ATP assay
Ten thousand HT-22 cells per well were seeded into a clear 96-
well plate (FALCON) one day prior to assay. Medium was removed
from the plate, and cells were treated with a 25 mL solution of
each compound at 10 mM and incubated at 37 ꢄC for 10 min fol-
lowed by the addition of a 25 mL of Ab (American peptide) solu-
tion at 10 mM. Cells were incubated at 37 ꢄC for 7 h and washed
twice with PBS. Cells were lysed by using 1% Triton-X 100 in TBST
buffer solution and the protein concentrations were determined
by using the BCA protein determination kit (Thermo scientific). An
equal amount of cell lysates from each well was plated into a
white 96-well plate (NUNC) and the amount of ATP generated in
each sample was determined by using the ATP determination kit
(Molecular Probes, USA) containing D-luciferin and luciferase. The
% inhibition value was calculated by measuring luminescence
2.11. hERG inhibition assay
For automated patch-clamp NPC-16 Patchliner (Nanion
€
Technologies, Munchen, Germany) recordings, CHO-K1 Tet-On
hERG cells (IonGate Biosciences GmbH, Frankfurt, Germany) were
plated into 100-mm cell culture dishes. Whole-cell currents were
recorded with the intracellular solution containing: 50 mM KCl,
60 mM KF, 10 mM NaCl, 2 mM MgCl2, 20 mM EGTA and 10 mM
HEPES (pH 7.2), and with the extracellular solution containing:
140 mM NaCl, 4 mM KCl, 2 mM CaCl2, 1 mM MgCl2, 5 mM Glucose
and 10 HEPES (pH 7.4). To assist stable seal formations, the seal
enhancer containing: 80 mM NaCl, 3 mM KCl, 35 mM CaCl2, 10 mM
MgCl2 and 10 mM HEPES (pH 7.4) was used only at the seal forma-
tion step. Prior to the whole-cell recordings, the external seal
enhancing solution was exchanged to the extracellular solution
described above. hERG channel currents were recorded using the
parallel EPC-10 patch-clamp amplifiers (HEKA Elektronik,
Lambrecht/Pfalz, Germany), and low-pass filtered (10 kHz) with a
4-pole Bessel filter. Cell suspension and patch solutions were auto-
R
from each sample (FlexstationV 3, Molecular Devices, USA), and
comparing the ATP levels of the vehicle control treated with Ab
as a negative control. Cell viability was also calculated based on
the ATP levels of each sample without the treatment of
Ab solution.
2.8. MTTassay
To a clear 96-well plate (FALCON), the cultured HT-22 cells were
seeded in a number of 5,000 per well 24 h prior to the assay. A
solution of each test compound (5 mM, 25 mL) was added to each
well. After 24 h of incubation at 37 ꢄC, the cells were treated with matically added onto the four recording wells in the microfabri-
a
10 mL of MTT solution (3–(4,5-dimethylthiazol-2-yl)-2,5- cated disposable chip (NPC-16 Chip, Nanion Technologies,

834
T. KIM ET AL.
€
Munchen, Germany). To obtain the inhibitory constants, hERG tail sequence of arm entries was recorded. The arm entry of the
currents were evoked by repolarizing steps to ꢅ50 mV for 500 ms mouse was counted when the all four limbs were within an arm.
preceded by a 500-ms depolarisation potential of þ20 mV at a A set of entries into all three arms is considered as a spontaneous
alternation. The spontaneous alternation (%) was calculated by
the following equation:
holding potential of ꢅ80 mV with a 20-s sweep interval. Whole-
cell currents were acquired and digitised at 5 kHz using
Patchmaster (HEKA Elektronik, Lambrecht/Pfalz, Germany). The
extracellular solution was exchanged to the extracellular solution
containing each blocker via four pipette tips of NPC-16 Patchliner
using a 4-fold volume of solution (40 mL) with a speed of 4 mL/s,
and the exchanged blocker solution was applied for 100–200 s to
the patch-clamped cells until blocker binding had reached equilib-
rium by monitoring hERG tail currents. Whole-cell recordings were
analysed using the Patchmaster/Fitmaster (HEKA Elektronik,
Lambrecht/Pfalz, Germany), IGOR Pro (WaveMetrics Inc., Portland,
OR, USA), and the GraphPad Prism 4.0 (GraphPad Software, Inc.,
La Jolla, CA, USA) software.
Spontaneous alternation ð%Þ
ꢀ
ꢁ
number of alternations
¼ 100ꢃ
ð
Þ
total number of arm entries – 2
2.15. Contextual fear conditioning test
To the Tg mice of the AD model (n ¼ 7 per group), each test com-
pound (30 mg/kg) was orally administrated daily for 1 month. Each
mouse was placed in a fear-conditioning chamber (Courlbourn,
USA) for 90 s before giving an electric shock. Trials of training
were performed using the chamber equipped with a fear condi-
tioning system (FreActimetrics, USA). The training was performed
by giving a conditional stimulus (CS) of 75 dB sound for 20 s fol-
lowed by an unconditional stimulus (US) of electric foot shock
(0.5 mA) for the last 2 s in CS. After an additional stay for 1 min,
the mouse was returned to its home cage. Fear conditioning test
was conducted 24 h after the training. The mouse was placed in
the same chamber for 5 min without presentation of CS. Freezing
behaviour was considered as the complete absence of any move-
ment except for respiration and heartbeat. Freezing response was
measured by the fear conditioning system without application of
CS or US.
2.12. Pharmacokinetic study
Each test compound at a dose of 10 mg/kg was administered
intravenously or orally to male Sprague-Dawley (SD) rats. Blood
samples collected via carotid artery at several time points after
administration of each compound were stored at ꢅ70 ꢄC until LC-
MS/MS analysis. To determine the plasma concentration of each
compound, a 0.1 mg/mL of internal standard in 100 mL of MeCN
was added to 50 mL of plasma samples. After vortex-shaking and
centrifugation at 10,000 rpm for 10 min, 5 mL of the supernatant
R
was analysed by LC-MS/MS [HP1100V HPLC system (Agilent, Santa
R
Clara, CA) combined with API3200V triple-quadrupole mass spec-
trometer (Applied Biosystems-SCIEX, Concord, Canada)]. The HPLC
chromatographic separation was performed on a reversed-phase
2.16. Surface plasmon resonance (SPR) measurements
R
XterraV C18 column (Waters Corporation, Milford, MA) using gradi-
To determine the binding affinities of the selected compounds,
the KD values were measured by using Biacore T200 optical bio-
sensor system (GE Healthcare). The recombinant human TSPO
(18 kDa, Origene Technologies, Inc., No.: TP320107) was covalently
immobilised on a CM5 chip (GE Healthcare) at 25 ꢄC via standard
amine-coupling protocols in 10 mM sodium acetate (pH 4.5) at a
flow rate of 10 mL/min for 1500s (immobilised protein densities:
5,000–6,000 response unit (RU)). The tested compounds were ini-
tially prepared at 10 mM in DMSO, and serially diluted in PBS-P
buffer (10 mM NaH2PO4, 150 mM NaCl, pH 7.4) to 1% DMSO. For
each tested compound, the kinetic values were measured using a
five-point concentration series, 0.001, 0.01, 0.5, 1, and 5 mM.
Multiple blank samples of running buffer alone (0 mM of tested
compounds) were incorporated in every measurement. The fil-
tered PBS-P buffer containing 1% DMSO was used as a running
buffer during the SPR experiments. To measure the kinetic values,
a series of different concentrations were injected over the immo-
bilised chip at a flow rate of 50 mL/min (contact time: 120 s, dis-
sociation time: 300 s), followed by the regeneration process at a
flow rate of 50 mL/min (contact time: 120 s, stabilisation time: 10 s).
The KD value of each compound was calculated by Biacore T200
evaluation software (GE Healthcare) after the standard solvent cor-
rection process.
ent elution consisted of 0.1% formic acid and 0.1% formic acid in
90% MeCN at a flow rate of 0.3 ml/min.
2.13. Animals
The acute AD mice model was prepared by administration of Ab1-
solution (10% DMSO/90% PBS) to male ICR mice (6 weeks old,
42
30 ꢅ 33 g) via intracerebroventricular (ICV) injection (Ab1-42 dose:
500 pmol/mouse) as described previously28. The wild type mice
(B6C3F1) and double APP/PS1 transgenic mice (Tg AD mice
model; APPswe/PSEN1dE9) were purchased from Jackson
Laboratory (Bar Harbour, ME, USA). The Tg AD model mice were
11 months old at the beginning of the behaviour test, and were
housed for 1 month in a room under controlled temperature and
fed ad libitum. The behaviour tests were conducted during day-
time in air-controlled and soundproof experiment room. The ani-
mal experiments were abided by the guidelines of the
Institutional Animal Care and Use Committee of Korea Institute of
Science and Technology.
2.14. Y-Maze spontaneous alternation test
To the mice of the acute AD model (n ¼ 7 per group), each test
compound (test compound in 20% cyclodextrin, 30 mg/kg) was
intravenously injected daily for 6 days. The Y-maze spontaneous
alternation test was performed 1 day after the administration of
each test compound. The Y maze was made of three equally
2.17. Docking study
A homology model of human TSPO was prepared using Discovery
studio 2017 R2 client software1. The human TSPO sequence
spaced black plastic arms (40 L ꢃ 10 W ꢃ 12 H cm) positioned at (UniProt ID: P30536) was taken from UniProt database. The high-
an equal angle. Each mouse was placed at the end of one arm resolution crystal structure of the translocator protein 18 kDa
and allowed to move freely through the maze for 8 min, and the (RCSB ID: 4UC1) from Rhodobacter sphaeroides (Sequence

JOURNAL OF ENZYME INHIBITION AND MEDICINAL CHEMISTRY
835
Figure 2. Representative TSPO ligands used as a training set to generate the pharmacophore model.
Identity: 33.7%, sequence similarity: 49.7%) was used as a tem-
plate for the homology model generation. Docking studies were
performed with XP docking methodology implemented in
€
Maestro, Schrodinger software2. Ligands were prepared for the
energy-minimised structures using OPLS_2005 force field. Proteins
were prepared using the protein preparation wizard. Hydrogens
were added and water molecules beyond 5 Å from hetero groups
were removed. The prepared ligands were docked using the extra
precision (XP) method and flexible ligand sampling. Docking
images were prepared using Discovery studio software.
3. Results and discussion
3.1. Generation of a pharmacophore model
We generated common feature pharmacophore models by using six
R
known TSPO ligands (compounds 1–6) from the IntegrityV database
of Prous (Figure 2). These compounds showed potent TSPO binding
affinities within the nanomolar range: 2–(2-(4-fluorophenyl)-1H-indol-
3-yl)-N,N-dihexylacetamide (FGIN-1–27, Ki ¼ 3.25 nM)29, N,N-dibutyl-
2–(6,8-dichloro-2–(4-chloro-phenyl)imidazo[1,2-a]pyridin-3-yl)acetamide
(AGN-PC-00A3VH, Ki ¼ 2.68 nM)30, N-(4-chloro-2-phenoxyphenyl)-N-
(2-isopropoxybenzyl)acetamide (DAA1097, IC50 ¼ 0.92 nM)31, N-(2,5-
dimethoxybenzyl)-N-(5-fluoro-2-phenoxyphenyl)acetamide (DAA1106,
IC50 ¼ 1.6 nM)31, 7-chloro-5–(4-chlorophenyl)-1-methyl-1,3-dihydro-2H-
benzo[e][1,4]diazepin-2-one (Ro5-4864, Ki ¼ 1.02 nM)32, and N-(sec-
Figure 3. The ligand-based pharmacophore model mapping with FGIN-1–27 (1).
(C atom, black; N atom, blue; O atom, red; F atom, cyan). The pharmacophore
features are drawn as colour-coded spheres: H-bond acceptor, green; hydrophobic
feature, light blue; and ring aromatic feature, blue.
each run. The final pharmacophore model was generated by dif-
fering the aromatic ring and hydrophobic feature options as
described in Figure 3. The final pharmacophore contains a hydro-
gen bond acceptor, an aromatic ring and four hydrophobic fea-
tures. Compound FGIN-1–27 overlapped with the model, showing
the oxygen atom map with a hydrogen bond acceptor feature,
the fluoro-benzene part with a hydrophobic aromatic feature, and
the benzene of the indole and n-alkyl chains of amide part with
four hydrophobic features.
butyl)-1–(2-chlorophenyl)-N-methylisoquinoline-3-carboxamide
11195, IC50 ¼ 1.1 nM)33.
(PK-
We applied HipHop pharmacophore module34,35 which does
not use any activity data of the training set compounds; instead,
it generates hypotheses by taking each training set compound as
a reference to accommodate structural features of all the com-
pounds. Since this method assumes that all the compounds in the
training set are equally important, all of the chemical features in
each compound will be considered to build a hypothesis space.
Next, in order to get more common features, the pharmacophore
3.2. Virtual screening and in vitro assays
We screened commercially available libraries from Asinex
models were refined by removing one training set compound in (AsinexGold, 229,398 compounds; AsinexPlatinum, 125,231

836
T. KIM ET AL.
compounds) and ChemDiv (693,042 compounds) against our activities (>40%) (Table 1). Thus, we chose these two compounds
pharmacophore model (Figure 3). Through the BEST flexible search
of the databases, 47,268 compounds were found as initial hits from
the HipHop pharmacophore model by the fit value (3.50 out of 6.00).
Among them, 56 compounds (VH001–VH056; Supplementary Table
S1) were manually selected on the basis of their structural diversity
and functional groups that are crucial for activity such as the number
of heteroatoms, aromatic/non-aromatic ring, n (n¼ 5, 6, 7)-membered
ring, polycyclic ring, alkyl substituent, and their positions.
and further evaluated their biological activity by measuring cell
viability and reactive oxygen species (ROS) production. The effects
on cell viability of each compound under Ab-induced cytotoxicity
were determined by measuring the metabolic activity of NAD(P)H-
dependent cellular oxidoreductases using the 3–(4,5-dimethylthia-
zol-2-yl)-2,5-diphenyl-tetrazolium bromide (MTT) dye. To measure
cellular ROS levels, we used a fluorescence indicator, 20,70-dichloro-
fluorescein diacetate (DCFDA), and calculated the percent inhib-
ition of ROS generation by applying the same method used for
MTT assays. Compound 7 inhibited Ab-induced ROS generation
(97% at 5 mM) almost to the level of the untreated control (Table
1). On the other hand, compound 8 did not appear to reduce ROS
generation, which is likely due to the toxicity of the compound
itself (34% of cell viability from the MTT assay). On the basis of
the in vitro activity, we selected compound 7 as the best hit from
the initial screening library.
The selected 56 compounds were purchased and tested by
in vitro cell-based assays. Four different cell-based assays (JC-1,
MTT, ATP production, and cellular ROS assays) were carried out
sequentially. First, the JC-1 assay was performed to screen all 56
compounds, VH001–VH056 (Supplementary Table S1). Since mito-
chondrial membrane depolarisation is one of the distinct signs of
mPTP opening, the mitochondrial uptake of a fluorescent JC-1 dye
(5,59,6,69-tetrachloro-1,19,3,39-tetraethylbenzimidazolocarbocya-
nine iodide) was measured to evaluate the effects of each com-
pound on the mitochondrial membrane potential (DWm). HT22
cells, a mouse hippocampal cell line, were treated with 5 mM of
Ab, incubated with each compound at the same concentration of
5 mM, and subsequently treated with JC-1. On the basis of changes
in the red/green fluorescence intensity ratio between the cells
treated with Ab and control, the percent recovery of DWm for
each compound was calculated and normalised to the DWm of
healthy cells as 100%. Two neuroprotective compounds, piracetam
and cyclosporine A (CsA), were also tested for comparison,
because these compounds have demonstrated neuroprotective
activity by restoring Ab-induced mitochondrial dysfunction36 or by
inhibiting prolonged opening of the mPTP37.
Based on the JC-1 assay results measured from piracetam (60%
at 5 mM) and CsA (55% at 5 mM), we selected compounds exhibit-
ing greater than 50% of the percent recovery values for additional
in vitro assays. Among 56 compounds, 23 compounds were
selected and further tested for the luciferase-based ATP assay. It
should be noted that the two reference compounds, piracetam
and CsA, appeared to demonstrate differing activity, 127% for
piracetam and ꢅ46% for CsA; it is not clear whether the effect on
DWm can be correlated with ATP production. Compounds 7
3.3. Additional similarity search and in vitro activity of the
identified hit compounds
To explore biological activities of structurally similar compounds
with 7, a similarity search was carried out by using 7 as a struc-
tural template. A total of 12 compounds (9–20; Table 2) were
selected from a commercial database and purchased for in vitro
assays. These newly selected compounds possess the same tricyc-
lic thienopyrimidine core; compounds 9–12 contain an alkyl chain
connected via the sulphur atom of the tricyclic core ring, while
compound 13 bears a polar amide group at the same position.
Compounds 14–16 contain a lipophilic aromatic group at the
same sulphur atom. In compounds 17–20, the sulphur side chain
is replaced with a carbonyl group, and both the two nitrogens in
the pyrimidine core have aromatic substituents, adding additional
hydrophobic groups compared with compounds 9–16.
Compounds with a thioalkyl group (9–12) are less effective in
recovering DWm than compound 7, whereas compounds with an
additional aromatic side chain with no sulphur atom in the core
(17–20) have no ATP recovering effect, suggesting that the aro-
(VH034) and 8 (VH050) showed the most potent recovery matic side chain connected through the sulphur atom is
Table 1. In vitro cell-based assay results of 7 and 8.
MTT
Compound
Recovery of DWm (%)a
ATP Recovery (%)a
Recovery (%)a
Viability (%)b
ROS Inhibition (%)
7 (VH034)
8 (VH050)
Piracetam
CsA
78%
92%
60%
55%
42%
45%
127%
ꢅ46%
28%
ꢅ184%
29%
95%
34%
132%
90%
97%
ꢅ10%
129%
ND
61%
aRecovery percent values were calculated by normalising each measured value against the untreated control; for example, 0% (Ab-induced
damaged condition) and 100% (normal condition in the absence of Ab).
bHT22 cell viability after 24 h of the treatment with 5 lM of each compound.
ND: Not determined.

JOURNAL OF ENZYME INHIBITION AND MEDICINAL CHEMISTRY
837
Table 2. In vitro cell-based assay results of 9–20.
MTT
Recovery of
ATP Recovery
Compound
R1
R2
DWm (%)
(%)
Recovery (%)
Viability (%)
ROS Inhibition (%)
7 (VH034)
9 (VH057)
10 (VH058)
11 (VH059)
12 (VH060)
13 (VH061)
78%
30%
52%
59%
1%
42%
28%
ND
45%
ND
ND
ND
95%
ND
87%
ND
ND
ND
97%
ND
ꢅ177%
ꢅ234%
ND
OMe
OEt
OEt
OEt
OMe
Me
Me
i-Bu
ND
49%
ND
ND
14%
n-Bu
66%
ND
14 (VH062)
OMe
77%
122%
92%
120%
136%
15 (VH063)
16 (VH064)
OEt
86%
74%
31%
38%
59%
32%
139%
135%
ꢅ4%
OMe
112%
17 (VH065)
18 (VH066)
19 (VH067)
20 (VH068)
72%
53%
73%
50%
ꢅ105%
ꢅ18%
ꢅ77%
ꢅ55%
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
Piracetam
CsA
60%
55%
127%
ꢅ46%
29%
61%
132%
90%
129%
NDa
ND: Not determined.
important for mitochondrial function. Overall, 14 (VH062) is the describe the additional SAR in the later section (3.6, com-
pound 27–62).
most potent compound in four different cell-based assays without
any noticeable cytotoxicity (Table 2). Specifically, compound 14
appeared to be highly effective in recovering ATP production and
metabolic function measured by MTT assays, comparable to the
levels of untreated cells. Considering that 14 has a smaller R1 sub-
stituent (methoxy versus ethoxy group), and highly electron-with-
drawing aromatic substituents in the R2 benzamide ring (difluoro
versus dimethoxy group) compared with 7, the SAR at these two
3.4. In vivo activity of the most active hit compounds, 7
(VH034) and 14 (VH062)
We evaluated in vivo activity of 7 and 14 in an acute AD mouse
model. The acute AD model mice were generated by the adminis-
tration of Ab1-42 peptide (500 pmol) according to the previously
positions need to be investigated more thoroughly. We will reported method28, and each compound (30 mg/kg daily) was

838
T. KIM ET AL.
Figure 4. Recovery of Y-maze spontaneous alternations in an acute AD model by the treatment of compounds 7 or 14 (30mg/kg daily, 6 days, i.p.) Each group was
ꢀ
evaluated by comparing with the piracetam-treated group (30mg/kg daily, 6 days, i.p.). Data are expressed as a mean SEM (n ¼ 7 per group): p < 0.05.
Figure 5. Recovery of Y-maze spontaneous alternations in transgenic (Tg; APPswe/PSEN1dE9 2x TG) AD model mice by the treatment of 7. Each group was evaluated
ꢀ
by comparing with the piracetam-treated group (30mg/kg daily, 1 month, PO). Data are expressed as a mean SEM (n ¼ 7 per group): p < 0.05.
administered intraperitoneally for 6 days. Piracetam (30 mg/kg show higher number of total freezing, which is an index of fear
daily) was also administered in parallel for comparison. The Y-
maze spontaneous alternation tests were performed to measure
the recovery of learning and memory deficit in acute AD model
mice38,39. After 6 days of administration, we found that compound
7 alleviated spatial learning and memory deficits significantly,
restoring 76% of Ab-induced memory impairments in acute AD
model mice (Figure 4). In contrast, compound 14 did not show
any comparable activity, and the piracetam-treated mice only
recovered 20% of spatial learning and memory deficits. We believe
that the treatment of 7 was beneficial for ameliorating Ab-induced
cognitive impairments.
To further validate the therapeutic effects of compound 7, we
performed the Y-maze tests by using a transgenic (Tg) mouse
model of AD (APPswe/PSEN1dE9 2xTG) aged 11 months.
Compound 7 was administrated to the Tg AD model mice for
1 month (30 mg/kg daily) and the Y-maze tests were performed at
12 months of age. As shown in Figures 5 and 7 restored 50% of
spatial learning and memory deficits, whereas piracetam recov-
ered 33% of the deficits. Overall, based on the results of the Y-
maze behaviour tests, compound 7 showed greater recovery than
piracetam in both transgenic and acute AD model mice.
learning and memory of rodents; therefore, the number of total
freezing responses under fear-relevant stimuli was measured with
the compound-treated Tg AD model mice. As shown in Figure 6,
the treatment of 7 restored cognitive function more effectively
compared to piracetam-treated mice, which corresponds to the
observations from the Y-maze tests. Specifically, the piracetam-
treated Tg mice were not affected by the treatment (percent
recovery
¼
ꢅ2%), whereas 7–treated mice demonstrated
a
remarkable improvement in cognitive function (62%). Given the
excellent in vivo activity of compound 7, as well as the significant
in vivo recovery effect, we decided to design new derivatives
based on the core structure of compounds 7 and 14 to optimise
the biological activity.
3.5. Synthesis of the designed ligands
Compounds 7, 14 and their derivatives were prepared as shown
in Scheme 1. Starting from commercially available piperidone
compounds 21a-b, the tetrahydrothieno pyridine compounds
22a-b were generated via Gewald reaction41. Subsequently, the
amino group of compounds 22a-b were converted to an isothio-
In addition, we assessed recovery effects of compound 7 by
performing contextual fear conditioning tests.40 Healthy mice cyanate group42,43. The addition of p-alkoxy anilines to the

JOURNAL OF ENZYME INHIBITION AND MEDICINAL CHEMISTRY
839
Figure 6. Contextual fear conditioning tests in wildtype and transgenic AD model mice. Each group was evaluated by comparing with the piracetam-treated group
ꢀ
(30mg/kg daily, 1 month, PO). Data are expressed as a mean SEM (n ¼ 7 per group): p < 0.05.
Figure 7. Recovery of Y-maze spontaneous alternations in acute AD model mice by treating each compound. Each group was evaluated by comparing with the pirace-
ꢀ
tam-treated group (30mg/kg daily, 6 days, i.p.). Data are expressed as a mean SEM (n ¼ 7 per group): p < 0.05.
isothiocyanate, followed by a cyclisation reaction yielded com- additional substituent (R2) at the amide nitrogen while diversified
pounds 25aa-b44. Finally, compounds 7, 14 and their derivatives
the benzamide substituents (R3) as described in Table 3.
Most derivatives showed excellent DWm percent recovery val-
ues close to or greater than 70%, except compounds 37, 41, and
42. These compounds with low activity have relatively small sized
(less than three carbons) alkyl side chains or a single alkyl chain at
the amide nitrogen, suggesting that the size and hydrophobicity
of the substituent at this position is crucial for cellular activity.
Compounds with an aromatic ring (31–36, 44–47) and a longer
alkyl chain (39–40) at the same position generally showed better
activity than compounds with a piperidine group (28–30), also
indicating that overall hydrophobicity of the molecule is important
for membrane-associated activity. The percent recovery values of
DWm were enhanced with increasing the length of alkyl chain, in
the order of methyl (compound 37; 45%) < ethyl (compound 38;
65%) < n-propyl (compound 39; 79%) < n-butyl (compound 40;
80%). In addition, the di-alkyl substituted (R2 and R3) compounds
37–40 showed greater recovery than mono-alkyl substituted (R3
only) compounds 41–43 from the JC-1 assay. When these R2 and
R3 alkyl groups were replaced with a constrained heterocycle
were obtained via SN2 reaction of compounds 25aa-b with 2-
bromo-N-substitued acetamide 26. Additionally, to study SAR of
different functional groups (R) at the position of a pendent alkoxy
phenyl ring of 7 and 14, compounds 58–62 were designed and
synthesised as illustrated in Scheme 2. To synthesise compounds
58–62, we used various alkylamines to diversify R groups (24c-g)
instead of using the alkoxy anilines.
3.6. In vitro activity of compounds 27–62
On the basis of our initial cell-based assays (Table 2), we found
that the sulphur-linked benzamide group is particularly important
for in vitro activity while an alkoxy substituent in the aromatic
side chain also affected biological activity to some extent. To fur-
ther explore SAR of these specific positions, we have designed
and synthesised additional compounds and tested for in vitro
activity. Specifically, we fixed the aromatic alkoxy substituent as
either a methoxy (27) or an ethoxy (28–47) group; we added an (compounds 44–47), they demonstrated consistently potent

840
T. KIM ET AL.
Scheme 1. Synthesis of 5,6,7,8-tetrahydropyrido[4’,3’:4,5]thieno[2,3-d]pyrimidin-4(3H)-one derivatives 7, 14, and 27–57. Reagents and conditions: (a) ethyl 2-cyanoace-
tate, Sulphur, Et3N, EtOH, reflux, 4 h, 96–100%, (b) 1,1’-thiocarbonylbis(pyridin-2(1H)-one), THF, reflux, 4 h, 80–86%, (c) substituted aniline, EtOH, reflux, 12 h, 84–93%,
(d) (i) 10% NaOH/MeOH (1:3), reflux 3 h, (ii) 2 N HCl, 93–98%, (e) 2-bromo-N-substitued acetamide 26, Et3N, MeCN, reflux, 5 h, 33–98%.
Scheme 2. Synthesis of 5,6,7,8-tetrahydropyrido[4’,3’:4,5]thieno[2,3-d]pyrimidin-4(3H)-one derivatives 58–62. Reagents and conditions: (a) substituted amine, EtOH,
reflux, 12 h, 57–99%, (b) (i) 10% NaOH/MeOH (1:3), reflux 3 h, (ii) 2 N HCl, 56–96%, (c) 2-bromo-N-(3,4-dimethoxyphenyl)acetamide 26, Et3N, MeCN, reflux,
5 h, 50–90%.
activity (78–83%), indicating that a bulky hydrophobic residue at (28–29, 37–43) generally showed little to no effects on both
this specific position is important for recovering Ab-induced mito- assays except compounds 30 and 40 which has relatively bulky
chondrial membrane depolarisation.
side chains. We think that the aromatic substituents are likely to
We further assessed in vitro activity of the compounds demon- act as a place holder or a steric blocker for specific interactions.
strated over 60% of recovery in the JC-1 assays by performing the Compound 27, which only differs the positions of difluoro-sub-
MTT assay and the ATP production assay. Six compounds (30–31, stituents from compound 14, did not show any effect on ATP
44–47) showed superior recovery activities in ATP production recovery or MTT assays, suggesting that the difluoro-substituents
compared to 7. Compounds without aromatic substituents might be involved in hydrogen-bonding or hydrogen-bonding

JOURNAL OF ENZYME INHIBITION AND MEDICINAL CHEMISTRY
841
Table 3. In vitro assay results of compounds 27–47.
MTT
Compound
R2
R3
Recovery of DWm
ATP recovery
Recovery
Viability
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
7
H
H
H
H
H
3,4-difluorophenyl
2-(Piperidin-1-yl)ethyl
3-(Piperidin-1-yl)propyl
1-Methylpiperidin-4-yl
Benzyl
Benzyl
Benzyl
4-Chlorobenzyl
4-Chlorophenyl
4-Methylbenzyl
Methyl
Ethyl
n-Propyl
73%
78%
69%
68%
81%
83%
86%
82%
82%
82%
45%
65%
79%
80%
51%
51%
73%
83%
78%
82%
81%
73%
ꢅ15%
ꢅ10%
ꢅ22%
60%
59%
7%
ꢅ4%
ꢅ22%
ꢅ10%
6%
18%
8%
35%
20%
18%
2%
ND
5%
92%
71%
89%
98%
109%
88%
104%
97%
107%
96%
ND
Methyl
Ethyl
H
H
7%
40%
38%
35%
ND
12%
15%
33%
ND
ND
1%
63%
54%
80%
86%
42%
H
Methyl
Ethyl
n-Propyl
n-Butyl
H
H
H
77%
99%
107%
ND
26%
32%
ND
n-Butyl
n-Propyl
Isopropyl
n-Butyl
ND
ND
ꢅ27%
79%
97%
113%
86%
94%
95%
4-Phenylpiperazin-1-yl
9%
4-Benzylpiperazin-1-yl
4-Phenylpiperidin-1-yl
4-Benzylpiperidin-1-yl
ꢅ31%
28%
31%
28%
ND: Not determined.
induced conformational changes. Compounds with a constrained viability (0–5%), suggesting that the additional lipophilic substitu-
heterocycle connecting the R2 and R3 positions (44–47) appeared ent may help sustain the membrane potential temporarily, but
does not seem to affect overall mitochondrial function.
to restore ATP production significantly, while piperidinyl analogs
(46, 47) are more effective in recovering both ATP production and
metabolic activity than piperazinyl analogs (44–45), again suggest-
ing that hydrogen-bonding interactions might be involved in this
binding region.
3.7. Assessment of drug-like properties of the
selected compounds
Next, we tested the second set of compounds, 48–57, that
were designed to have an N-methyl group replacing the aceta-
mide group at the 7-position of the tetrahydropyrido moiety in 7.
As described in Table 4, compounds in this series showed slightly
less of DWm recovery, while demonstrated better results in ATP
recovery and MTT assay, compared to their acetamide counter-
parts. We think that the replacement of the acetamide to the
methyl may have affected electron-density of the nitrogen with lit-
tle effect on overall hydrophobicity or conformation. The ATP pro-
duction assay and the MTT assay results indicated that
compounds in this group generally showed favourable recovery
activity in both assays except compounds 53 and 54.
On the basis of the in vitro mitochondrial functional assays, we
selected a total number of 20 compounds with excellent recovery
effects. We determined inhibition of CYP450 isozyme, inhibition of
the hERG, and human liver microsomal stability to evaluate drug-
like properties of the selected compounds (Table 6). Compounds
31–32, 34, 36, 45–48, 52, 55, and 56 showed relatively high
inhibition (<75%) against at least one CYP450 isozyme.
Compounds with bi-substituted N-alkyl chains (Table 3, 32, 33,
39, 40, 45–47, 56) demonstrated poor stability in human liver
microsome, showing less than 10% of the compound remaining
intact after 30 min. Compounds 45–48 also demonstrated low
water solubility, thus the IC50 values of the hERG inhibition could
not be determined. Compounds 35 and 36 showed moderate
hERG inhibition, implying a potential toxicity issue. Compounds
50 and 51 are similar in their structures, in vitro activity and drug-
like properties, therefore we selected compound 50 as a represen-
The final group of compounds contained an alkyl chain in
place of the 4-ethoxyphenyl side chain in compound 7 (com-
pounds 58 2 62). As shown in Table 5, compounds in this series
demonstrated comparable recovery of DWm (60–79%) likely due
to their hydrophobicity; however, they generally showed moder- tative compound. Based on this toxicity profile, we selected com-
ate to poor recovery of ATP production (ꢅ7 to 23%) and cell pounds 30, 31, 44, 49 and 50, which demonstrated low to

842
T. KIM ET AL.
Table 4. In vitro assay results of compounds 48–57.
MTT
Compound
R1
R2
Recovery of DWm
ATP Recovery
Recovery
Viability
48
49
50
51
52
53
54
55
56
57
7
H
H
3,4-Dimethoxyphenyl
Benzyl
Benzyl
55%
61%
78%
74%
67%
65%
57%
66%
69%
73%
73%
62%
36%
32%
22%
30%
10%
1%
56%
50%
50%
42%
35%
42%
33%
39%
19%
ꢅ21%
2%
45%
16%
66%
28%
106%
99%
95%
107%
80%
89%
107%
127%
108%
110%
95%
Methyl
Ethyl
H
H
H
H
Benzyl
4-Chlorobenzyl
4-Chlorophenyl
4-Methylbenzyl
4-Methylphenyl
n-Propyl
n-Propyl
n-Butyl
n-Butyl
7 (76%) and piracetam (20%) (Figure 4). In contrast, compounds
30, 49, and 50 did not show significant activity. We concluded
that compounds 31 and 44 are the most active compounds based
on their in vitro and in vivo protective effect against Ab-induced
mitochondrial dysfunction.
Table 5. In vitro assay results of compounds 58–62.
3.9. Pharmacokinetic properties of compounds 31 and 44
We evaluated the pharmacokinetic profiles of the two most active
compounds 31 and 44 using SD male rats. As shown in Table 7,
compound 31 showed low bioavailability (8.0%) whereas com-
pound 44 demonstrated much higher bioavailability (78.4%).
However, given its relatively low liver microsomal stability (15.3%)
and short plasma half-life (80 min), compound 44 needs further
optimisation for oral administration. It is notable that the presence
of a constrained phenylpiperazine group at the amide bond
region (R3 and R4) improved bioavailability while reduced liver
microsomal stability compared to the initial hit compound 7
(Supplementary Tables S3 and S4).
MTT
Recovery
of DWm
ATP
Recovery
Compound
R
Recovery
Viability
58
59
60
61
62
7
Ethyl
60%
79%
66%
71%
68%
73%
23%
ꢅ1%
14%
ꢅ7%
15%
42%
4%
5%
1%
0%
0%
99%
112%
96%
97%
94%
95%
n-Propyl
Isobutyl
n-Butyl
Isopropyl
28%
moderate CYP inhibition against all tested isoforms, low hERG
inhibition, with moderate to high liver microsomal stability, for
further in vivo studies.
3.10. TSPO binding affinity measured by SPR
The direct binding affinities of the initial hit compounds 7 and 14
as well as compounds 31 and 44 were determined against recom-
binant human TSPO by performing surface plasmon resonance
(SPR) studies. As shown in Table 8 and Supplementary Figure S2,
all four compounds demonstrated specific binding in a dose-
dependent manner, having the KD values in a low micromolar
range. Although PK11195, a well-known TSPO ligand, seemed to
be much more potent (KD ¼ 1.1 0.45 nM) than our newly identi-
fied ligands, given the size and structural differences, we believe
that these compounds can serve as a good lead compound with
a novel scaffold for future optimisation through SAR studies.
These results supported that these compounds specifically interact
3.8. In vivo activity of the selected compounds
To assess the in vivo activity of the selected compounds, 30, 31,
44, 49 and 50, we administered each compound (30 mg/kg, daily)
to the acute AD model mice for 6 days. After the administration,
we evaluated the cognitive function of the mice by performing
the Y-maze spontaneous alternation test. As shown in Figure 7,
compounds 31 and 44 significantly recovered learning and mem-
ory deficits in the acute AD model mice. Both compounds 31 and
44 improved Ab-induced memory deficits almost to the level of
normal mice, demonstrating the recovery percent of 107% and
98%, respectively. These percent values are greater than those of with TSPO to exert their biological activity.

JOURNAL OF ENZYME INHIBITION AND MEDICINAL CHEMISTRY
843
Table 6. CYP450, hERG liability, and human liver microsomal stability profiles of the selected compounds.
CYP450
(% remaining Activity at 10 lM)
hERG
(IC50, lM)
Human Liver Microsomal Stability
(% remaining after 30 min)
Compound
CYP1A2
CYP2D6
CYP2C9
CYP3A4
CYP2C19
30
31
32
33
34
35
36
39
40
44
45
46
47
48
49
50
51
52
55
56
119.7
143.1
117.3
150.5
160.6
120.0
123.6
143.1
103.0
92.0
110.5
84.4
81.5
35.8
95.3
118.5
88.1
91.7
82.0
92.6
113.8
98.9
95.1
79.8
117.3
107.0
72.3
64.0
57.8
43.1
30.0
61.4
36.6
59.2
89.2
93.0
51.1
30.9
32.8
32.9
64.0
50.4
61.0
33.6
84.4
14.9
6.7
15.6
33.7
99.6
66.4
105.8
86.9
43.0
44.5
31.0
18.3
49.3
56.6
15.0
70.7
17.3
51.8
38.9
72.2
89.1
81.2
66.2
20.7
28.1
36.5
44.1
42.6
23.0
30.9
107.6
70.0
23.3
35.7
56.8
72.0
92.7
64.0
57.4
28.3
23.2
16.5
15.0
20.3
90.0
47.9
52.5
16.1
31.2
30.5
12.30 1.25
8.80 2.36
6.53 2.46
2.77 1.18
4.13 1.39
30.00 10.2
22.00 12.7
5.87 1.18
2.43 0.48
8.90 0.22
ND
ND
ND
ND
15.00 4.24
10.60 0.97
10.00 3.60
8.60 0.83
5.63 1.75
12.90 3.03
99.1
32.7
8.1
9.6
0.0
0.1
0.0
0.0
1.5
15.3
4.3
6.6
2.2
63.4
87.9
73.1
82.0
67.6
49.1
1.3
97.3
98.7
117.2
69.2
71.9
ND: not determined (due to the low solubility).
Table 7. Pharmacokinetic parameters (mean SD) of compounds 31 and 44 after intravenous (n ¼ 3) and oral (n ¼ 3) adminis-
tration (10 mg/kg) to SD male rats.a
31
44
Intravenous
Oral
Intravenous
Oral
AUC0–1
(mg min/mL)
AUClast
(mg min/mL)
Terminal half-life (min)
Cmax (mg/mL)
330.66 93.61
26.59 22.32
1070.23 114.39
614.11 120.15
273.60 64.45
11.67 5.99
1058.16 109.82
487.49 87.29
173.83 32.38
448.10 27.60
0.04 0.02
80.17 9.34
199.34 134.70
2.11 0.91
–
–
Tmax (min)
–
48
–
72
(30 ꢁ 60)
(60 ꢁ 120)
CL (mL/mim/kg)
Vss (mL/kg)
F (%)
32.53 10.08
7736.8 1114.3
8.0
–
–
9.43 1.05
785.49 39.64
78.4
–
–
aAUC0–1, total area under the plasma concentration–time curve from time zero to time infinity; AUClast, total area under the
plasma concentration–time curve from time zero to last measured time; Cmax, peak plasma concentration; Tmax, time to reach
Cmax; CL, time-averaged total body clearance; Vss, apparent volume of distribution at steady state; F, bioavailability.
Table 8. TSPO binding affinity.a
3.11. Docking study
Compound
KD
To further investigate the binding mode of the selected com-
pounds, we performed docking studies for the two most active
compounds 7 and 44. We generated the human TSPO homology
model based on 4UC1 x-ray crystal structure as a template. The
binding orientation of compounds 7 and 44 appeared to overlap
inside the central pocket of the TSPO model as shown in Figure
8(A). The amide group of 7 (glide XP score¼ ꢅ11.27) formed a
hydrogen bond with Tyr57 from TM2 as shown in Figure 8(B). The
dimethoxy phenyl ring inserted between TM1 and TM2 showed a
pi-alkyl bond with Cys19 from TM1. The tricyclic fused core ring
compactly packed inside the binding pocket via strong pi stacked
interactions of Trp95 from TM3 and pi-alkyl interactions of Leu49,
Leu150, and His43. Two Pi-sulphur bonds were also observed
between Trp53 and the sulphur atom of the thienopiperidinone
ring. The sulphur atom from the alkyl chain also exhibited pi-sul-
phur interactions with Phe25 as shown in Figure 8(D).
7
14
31
44
0.63 0.33 lM
1.69 0.40 lM
1.25 0.40 lM
2.74 0.66 lM
1.1 0.45 nM
PK11195
aThe KD values are expressed as the mean SD from three independent
measurements.
the presence of a bulky phenylpiperazine group, the tricyclic core
ring of compound 44 moved towards the space between intracel-
lular loop1 residues such as Phe25, His46, and Leu49 interacting
via strong pi stacked, pi-sulphur and pi-alkyl contacts respectively
(Figure 8(C)). The phenylpiperazine side chain is pointed towards
TM5 while interacts with Phe99 from TM3 via a strong pi-stacked
interaction. The presence of pi-stacking interactions corroborates
the superior activity observed from compounds with an aromatic
ring compared to compounds with no aromatic side chains.
Several pi-alkyl interactions with Val154, Leu150, Leu49, and Trp95
In contrast, compound 44 (glide XP score¼ ꢅ11.20) fitted in
reverse orientation inside the binding cavity (Figure 8(A)) which
explains slightly lower binding affinity of 44 than 7. Because of were also observed. The sulphur atom from the alkyl chain

844
T. KIM ET AL.
Figure 8. Interactions of 7 and compound 44 with the human TSPO homology model. (A) Surface depiction of the TSPO binding site with the binding orientation
overlapped with 7 (red stick) and compound 44 (blue stick). Detailed view of the interactions between 7 (B) and compound 44 (C) with the human TSPO. TSPO is rep-
resented with cyan helices and the interacting amino acids are depicted with grey sticks. Corresponding 2D interactions of 7 (D) and compound 44 (E).
showed two pi-sulphur bonds with Trp53 and Trp95. The oxygen hydrophobic region is preferable for binding of the phenylpipera-
atom of the ethoxy group formed a hydrogen bond with Tyr57.
Trp53 showed pi stacked and pi sigma contacts with the ethoxy
phenyl ring as described in Figure 8(E).
zine group. Therefore, in spite of having similar scaffolds 7 and 44
demonstrate an opposite binding orientation in the dock-
ing study.
Overall, 7 and 44 have a similar scaffold and share the same
binding cavity. In addition, both molecules appeared to form a
hydrogen bond with Tyr57 residue. Docking score reveals that
they have the equal binding potential with TSPO protein, but they
may bind in the reverse orientation. The binding region between
TM2, TM3, and TM5 is hydrophobic due to the presence of resi-
4. Conclusion
In this study, we synthesized a series of TSPO ligands derived
from a ligand-based pharmacophore. We evaluated their in vitro
dues like Leu49, Trp95, Phe99, Leu150, and Val154. This and in vivo activity of mitochondrial functional recovery and

JOURNAL OF ENZYME INHIBITION AND MEDICINAL CHEMISTRY
845
discovered two initial hit compounds from virtual screening and 10. Papadopoulos V, Baraldi M, Guilarte TR, et al. Translocator
in vitro assays, and further synthesised a library of analogs. Two
most active compounds 31 and 44 demonstrated excellent recov-
ery of DWm, ATP production, and cell viability upon Ab-induced
cytotoxicity. These compounds also demonstrated neuroprotective
effects in AD model mice, improving learning and memory impair-
ment. Our SPR data and the docking study results indicated that
7 and 44 directly interact with the TSPO binding site, while they
may occupy the cavity in reverse orientation to each other despite
their structural similarities. Taken together, our findings suggest
that these compounds exert neuroprotective effects in vitro and
in vivo by restoring mitochondrial function through the specific
interaction with TSPO in the presence of Ab. We believe that
these novel TSPO ligands can serve as a promising lead for further
therapeutic development, and may provide an alternative strategy
to the current treatment options for AD.
protein (18kda): new nomenclature for the peripheral-type
benzodiazepine receptor based on its structure and molecu-
lar function. Trends Pharmacol Sci 2006;27:402–9.
11. Veenman L, Gavish M. The role of 18 kda mitochondrial
translocator protein (tspo) in programmed cell death, and
effects of steroids on tspo expression. Curr Mol Med 2012;
12:398–412.
12. Leaver KR, Reynolds A, Bodard S, et al. Effects of translocator
protein (18 kda) ligands on microglial activation and neur-
onal death in the quinolinic-acid-injected rat striatum. ACS
Chem Neurosci 2012;3:114–9.
13. Starkov AA, Beal FM. Portal to Alzheimer’s disease. Nat Med
2008;14:1020–1.
14. Rupprecht R, Papadopoulos V, Rammes G, et al. Translocator
protein (18 kda) (tspo) as a therapeutic target for neuro-
logical and psychiatric disorders. Nature Rev Drug Discov
2010;9:971–88.
Disclosure statement
15. Du H, Guo L, Fang F, et al. Cyclophilin d deficiency attenu-
ates mitochondrial and neuronal perturbation and amelio-
rates learning and memory in Alzheimer’s disease. Nat Med
2008;14:1097–105.
No potential conflict of interest was reported by the author(s).
16. Du H, Yan SS. Mitochondrial permeability transition pore in
Alzheimer’s disease: cyclophilin d and amyloid beta. Biochim
Biophys Acta 2010;1802:198–204.
17. Chelli B, Falleni A, Salvetti F, et al. Peripheral-type benzodi-
azepine receptor ligands: mitochondrial permeability transi-
tion induction in rat cardiac tissue. Biochem Pharmacol
2001;61:695–705.
Funding
This work was supported by the National Research Council of
Science and Technology grant by the Korea government (MSIP)
[No. CRC-15–04-KIST].
18. Krestinina OV, Grachev DE, Odinokova IV, et al. Effect of per-
ipheral benzodiazepine receptor (pbr/tspo) ligands on open-
ing of ca2þ-induced pore and phosphorylation of 3.5-kda
polypeptide in rat brain mitochondria. Biochemistry (Mosc)
2009;74:421–9.
ORCID
Suengmok Cho
Sang M. Lim
http://orcid.org/0000-0001-5951-5832
http://orcid.org/0000-0003-0419-8950
19. Kim YS, Jung SH, Park BG, et al. Synthesis and evaluation of
oxime derivatives as modulators for amyloid beta-induced
mitochondrial dysfunction. Eur J Med Chem 2013;62:71–83.
20. Barron AM, Garcia-Segura LM, Caruso D, et al. Ligand for
translocator protein reverses pathology in a mouse model
of Alzheimer’s disease. J Neurosci 2013;33:8891–7.
21. Gatliff J, East D, Crosby J, et al. Tspo interacts with vdac1
and triggers a ROS-mediated inhibition of mitochondrial
quality control. Autophagy 2014;10:2279–96.
22. Werry EL, Bright FM, Piguet O, et al. Recent developments in
tspo pet imaging as a biomarker of neuroinflammation in
neurodegenerative disorders. Int J Mol Sci 2019;20:3161.
23. Barresi E, Robello M, Costa B, et al. An update into the
medicinal chemistry of translocator protein (tspo) ligands.
Eur J Med Chem 2021;209:112924.
24. Venneti S, Wang G, Wiley CA. The high affinity peripheral
benzodiazepine receptor ligand daa1106 binds to activated
and infected brain macrophages in areas of synaptic degen-
eration: implications for pet imaging of neuroinflammation
in lentiviral encephalitis. Neurobiol Dis 2008;29:232–41.
25. Chen MK, Guilarte TR. Translocator protein 18 kda (tspo):
molecular sensor of brain injury and repair. Pharmacol Ther
2008;118:1–17.
References
1. Anand R, Gill KD, Mahdi AA. Therapeutics of Alzheimer’s dis-
ease: past, present and future. Neuropharmacology 2014;76:
27–50.
2. Kumar A, Singh A. Ekavali. A review on Alzheimer’s disease
pathophysiology and its management: an update.
Pharmacol Rep 2015;67:195–203.
3. Tanzi RE, Bertram L. Twenty years of the Alzheimer’s disease
amyloid hypothesis: a genetic perspective. Cell 2005;120:
545–55.
4. LaFerla FM, Green KN, Oddo S. Intracellular amyloid-beta in
Alzheimer’s disease. Nat Rev Neurosci 2007;8:499–509.
5. Readnower RD, Sauerbeck AD, Sullivan PG. Mitochondria,
amyloid beta, and Alzheimer’s disease. Int J Alzheimers Dis
2011;2011:1–5.
6. Eckert GP, Renner K, Eckert SH, et al. Mitochondrial dysfunc-
tion-a pharmacological target in Alzheimer’s disease. Mol
Neurobiol 2012;46:136–50.
7. Santos RX, Correia SC, Wang X, et al. Alzheimer’s disease:
diverse aspects of mitochondrial malfunctioning. Int J Clin
Exp Pathol 2010;3:570–81.
8. Moreira PI, Carvalho C, Zhu X, et al. Mitochondrial dysfunc-
tion is a trigger of Alzheimer’s disease pathophysiology.
Biochim Biophys Acta 2010;1802:2–10.
26. Li F, Liu J, Zheng Y, et al. Crystal structures of translocator
protein (tspo) and mutant mimic of a human polymorphism.
Science 2015;347:555–8.
9. Chua SW, Kassiou M, Ittner LM. The translocator protein as a
drug target in Alzheimer’s disease. Expert Rev Neurother
2014;14:439–48.
27. Guo YZ, Kalathur RC, Liu Q, et al. Protein structure. Structure
and activity of tryptophan-rich TSPO proteins . Science 2015;
347:551–5.

846
T. KIM ET AL.
28. Cho SM, Kim HV, Lee S, et al. Correlations of amyloid-b con- 36. Kurz C, Ungerer I, Lipka U, et al. The metabolic enhancer
centrations between CSF and plasma in acute Alzheimer
mouse model. Sci Rep 2014;4:6777.
29. Kozikowski AP, Ma D, Brewer J, et al. Chemistry, binding
piracetam ameliorates the impairment of mitochondrial
function and neurite outgrowth induced by beta-amyloid
peptide. Br J Pharmacol 2010;160:246–57.
affinities, and behavioral properties of a new class of 37. Halestrap AP, Pasdois P. The role of the mitochondrial per-
“antineophobic” mitochondrial dbi receptor complex (mdrc)
ligands. J Med Chem 1993;36:2908–20.
meability transition pore in heart disease. Biochim Biophys
Acta 2009;1787:1402–15.
30. Trapani G, Laquintana V, Denora N, et al. Structure-activity 38. Riedel G, Wetzel W, Reymann KG. Computer-assisted shock-
relationships and effects on neuroactive steroid synthesis in
a series of 2-phenylimidazo[1,2-a]pyridineacetamide periph-
reinforced y-maze training: a method for studying spatial
alternation behaviour. Neuroreport 1994;5:2061–4.
eral benzodiazepine receptors ligands. J Med Chem 2005;48: 39. Oddo S, Caccamo A, Shepherd JD, et al. Triple-transgenic
292–305.
31. Okuyama
model of Alzheimer’s disease with plaques and tangles:
intracellular Abeta and synaptic dysfunction. Neuron 2003;
39:409–21.
S,
Chaki
S,
Yoshikawa
R,
et
al.
Neuropharmacological profile of peripheral benzodiazepine
receptor agonists, daa1097 and daa1106. Life Sci 1999;64: 40. Fanselow MS, Poulos AM. The neuroscience of mammalian
1455–64. associative learning. Annu Rev Psychol 2005;56:207–34.
32. Das SK, Sikpi MO, Skolnick P. Characterization of [3h]ro5- 41. Abaee MS, Cheraghi S. Efficient three-component gewald
4864 binding to “peripheral” benzodiazepine receptors in
guinea pig alveolar type ii cells. Biochem Pharmacol 1987;
36:2221–4.
reactions under et3n/h2o conditions. J Sulfur Chem 2014;35:
261–9.
42. Kim S, Yi KY. 1, 1’-thiocarbonyldi-2, 2’-pyridone. A new use-
ful reagent for functional group conversions under essen-
tially neutral conditions. J Organic Chem 1986;51:2613–5.
33. Rosenberg N, Rosenberg O, Weizman A, et al. In vitro mito-
chondrial effects of pk 11195, a synthetic translocator pro-
tein 18 kDa (TSPO) ligand, in human osteoblast-like cells. J 43. Shi Y, Sitkoff D, Zhang J, et al. Design, structure-activity rela-
Bioenerg Biomembr 2011;43:739–46.
tionships, x-ray crystal structure, and energetic contributions
of a critical p1 pharmacophore: 3-chloroindole-7-yl-based
factor xa inhibitors. J Med Chem 2008;51:7541–51.
34. Adane L, Bharatam PV, Sharma V. A common feature-based
3d-pharmacophore model generation and virtual screening:
identification of potential pfdhfr inhibitors. J Enzyme Inhib 44. Ahmed EK. Synthesis of some new pyrido [40, 30: 4, 5] thieno
Med Chem 2010;25:635–45.
[2, 3-d] pyrimidines and related fused heterocycles.
35. Xing RJ, Wang J, Pan L, Cheng MS. A selective pharmaco-
phore model for beta(2)-adrenoceptor agonists. Molecules
2009;14:4486–96.
Phosphorus Sulfur Silicon Relat Elem 2003;178:1–16.