Stem cells are the most sensitive screening tool to identify toxicity of GATA4-targeted novel small-molecule compounds

Article abstract of DOI:10.1007/s00204-018-2257-1

Safety assessment of drug candidates in numerous in vitro and experimental animal models is expensive, time consuming and animal intensive. More thorough toxicity profiling already in the early drug discovery projects using human cell models, which more closely resemble the physiological cell types, would help to decrease drug development costs. In this study we aimed to compare different cardiac and stem cell models for in vitro toxicity testing and to elucidate structure–toxicity relationships of novel compounds targeting the cardiac transcription factor GATA4. By screening the effects of eight compounds at concentrations ranging from 10?nM up to 30?μM on the viability of eight different cell types, we identified significant cell type- and structure-dependent toxicity profiles. We further characterized two compounds in more detail using high-content analysis. The results highlight the importance of cell type selection for toxicity screening and indicate that stem cells represent the most sensitive screening model, which can detect toxicity that may otherwise remain unnoticed. Furthermore, our structure–toxicity analysis reveals a characteristic dihedral angle in the GATA4-targeted compounds that causes stem cell toxicity and thus helps to direct further drug development efforts towards non-toxic derivatives.

Full text of DOI:10.1007/s00204-018-2257-1

Archives of Toxicology
https://doi.org/10.1007/s00204-018-2257-1
IN VITRO SYSTEMS
Stem cells are the most sensitive screening tool to identify toxicity
of GATA4-targeted novel small-molecule compounds
S. Tuuli Karhu1 · Mika J. Välimäki1,2 · Mikael Jumppanen3 · Sini M. Kinnunen1,2 · Lotta Pohjolainen1 ·
Robert S. Leigh1 · Samuli Auno3 · Gábor Földes4,5 · Gustav Boije af Gennäs3 · Jari Yli‑Kauhaluoma3 ·
Heikki Ruskoaho1 · Virpi Talman1
Received: 6 February 2018 / Accepted: 4 July 2018
© The Author(s) 2018
Abstract
Safety assessment of drug candidates in numerous in vitro and experimental animal models is expensive, time consuming
and animal intensive. More thorough toxicity profiling already in the early drug discovery projects using human cell models,
which more closely resemble the physiological cell types, would help to decrease drug development costs. In this study we
aimed to compare different cardiac and stem cell models for in vitro toxicity testing and to elucidate structure–toxicity rela-
tionships of novel compounds targeting the cardiac transcription factor GATA4. By screening the effects of eight compounds
at concentrations ranging from 10 nM up to 30 µM on the viability of eight different cell types, we identified significant cell
type- and structure-dependent toxicity profiles. We further characterized two compounds in more detail using high-content
analysis. The results highlight the importance of cell type selection for toxicity screening and indicate that stem cells repre-
sent the most sensitive screening model, which can detect toxicity that may otherwise remain unnoticed. Furthermore, our
structure–toxicity analysis reveals a characteristic dihedral angle in the GATA4-targeted compounds that causes stem cell
toxicity and thus helps to direct further drug development efforts towards non-toxic derivatives.
Keywords Toxicity screening · High-content analysis · Structure–toxicity relationship · Stem cells · Cardiomyocytes ·
Isoxazole derivatives
Introduction
Early drug discovery projects routinely use tumor-derived or
genetically immortalized cell lines and primary cells from
Electronic supplementary material The online version of this
experimental animals to assess toxicity of novel compounds
(Horvath et al. 2016). The cell type chosen for toxicity
screening may influence the results and a wrong choice can
lead to misjudgements in lead selection. Most often toxic
compounds are identified during preclinical drug develop-
ment, but sometimes, for example due to species differences,
they may only come up in the clinical phase or during post-
approval follow-up with detrimental results (Kerbrat et al.
2016; Suntharalingam et al. 2006; Wysowski et al. 2001).
More precise in vitro toxicity assessment using human cell
models that more closely resemble physiological cell types
is, therefore, urgently needed to improve the validity of
results and to reduce drug discovery costs and the use of
experimental animals (Sison-Young et al. 2012).
article (https://doi.org/10.1007/s00204-018-2257-1) contains
supplementary material, which is available to authorized users.
* Virpi Talman
virpi.talman@helsinki.fi
1
Drug Research Program and Division of Pharmacology
and Pharmacotherapy, Faculty of Pharmacy, University
of Helsinki, P.O. Box 56, 00014 Helsinki, Finland
2
Department of Pharmacology and Toxicology, Institute
of Biomedicine, University of Oulu, Oulu, Finland
3
Drug Research Program and Division of Pharmaceutical
Chemistry and Technology, Faculty of Pharmacy, University
of Helsinki, P.O. Box 56, 00014 Helsinki, Finland
4
National Heart and Lung Institute, Imperial College London,
London, UK
The recent progress in human stem cell-based models,
including the discovery of human induced pluripotent stem
5
Heart and Vascular Center, Semmelweis University,
Budapest, Hungary
Vol.:(0123456789)
1 3

Archives of Toxicology
cell (hiPSC) technology (Takahashi et al. 2007; Yu et al.
2007), enables the production of almost any human cell type
in the laboratory. This provides a unique possibility to carry
out drug screening and toxicity studies using differentiated
human cells that cannot be readily obtained from living per-
sons, such as cardiomyocytes and neurons, and eliminates
the influence of interspecies differences. Additionally, pluri-
potent stem cells have been accepted as a validated model for
evaluating reproductive and developmental toxicity in vitro
(Seiler and Spielmann 2011).
et al. 1998). Additionally, the GATA4 consensus sites in
combination with an NKX2-5 binding element are essential
for stretch-activated B-type natriuretic peptide (BNP) tran-
scription and cardiomyocyte hypertrophy (Pikkarainen et al.
2003). We have recently reported novel compounds that tar-
get GATA4 and its interaction with NKX2-5 (Välimäki et al.
2017). The hit compound 3i-1000 [compound 3 in our previ-
ous report (Välimäki et al. 2017)] inhibits GATA4-induced
synergistic transactivation of a promoter sequence contain-
ing NKX2-5 binding sites and attenuates stretch-induced
cardiomyocyte hypertrophy. The compound also inhibits
pro-hypertrophic extracellular signal-regulated kinase 1/2
signalling in the hypoxic endocardial layer of the left ven-
tricle in vivo after systemic administration in heart-targeted
nanoparticles (Ferreira et al. 2017) and improves cardiac
function after myocardial infarction or angiotensin II-
induced cardiac hypertrophy in vivo (Kinnunen et al. 2018).
The aims of this study were (1) to compare different
cardiac and stem cell types in toxicity screening and (2) to
investigate the in vitro toxicity and structure–toxicity rela-
tionships of a new family of compounds targeting GATA4
and its interaction with NKX2-5. Toxicity of a structurally
conserved set of eight compounds (Fig. 1) was investigated
using various cardiac and non-cardiac cell models: H9c2
myoblast cell line derived from rat myocardium, primary
neonatal rat ventricular cardiomyocytes (NRVCs), primary
neonatal rat cardiac fibroblasts (CFs), mouse embryonic
fibroblasts (MEFs), mouse embryonic stem cells (mESCs),
mESC derivatives from day 5 embryoid bodies (D5EBs),
hiPSCs and hiPSC-derived cardiomyocytes (hiPSC-CMs).
The cardiac transcription factor GATA4 is a master regu-
lator of cardiogenesis (Pikkarainen et al. 2004). It regulates
cardiac development, cardiomyocyte hypertrophy, cardio-
myocyte survival and cardiac regeneration (Gupta et al.
2013; Kikuchi et al. 2010; Malek Mohammadi et al. 2017;
Pikkarainen et al. 2003; Rysä et al. 2010). GATA4 inter-
acts directly with other cardiac transcription factors, such as
NKX2-5 and TBX5, to cooperatively regulate gene expres-
sion (Ang et al. 2016; Kinnunen et al. 2015; Luna-Zurita
et al. 2016; Pikkarainen et al. 2004). GATA4 mutations
that affect its transcriptional activity or interactions with its
cofactors can lead to cardiomyopathies and congenital heart
defects (Ang et al. 2016; Garg et al. 2003; Li et al. 2013).
Cardiac transcription factors and their protein–protein inter-
actions can thus be considered attractive, yet challenging,
drug targets.
The activities of both GATA4 and NKX2-5 are enhanced
by hypertrophic stimuli (Kohli et al. 2011). GATA4 and
NKX2-5 act synergistically as crucial controllers of, for
instance, atrial natriuretic peptide (ANP) transcription (Lee
Fig. 1 Compound structures.
Chemical structures of the
active compound 3i-1000
and its derivatives analysed
herein. These molecules can
formally be characterized by
northern and southern parts
that are connected via a linker
unit as defined in the figure.
The northern part is typically
comprised of a para-phenylen-
ediamine, whereas the southern
part is typically comprised
of an ortho-aryl-substituted
heterocycle. The linker unit is a
two-atom spacer consisting of
a nitrogen atom and sp2 or sp3
carbon atom
1 3

Archives of Toxicology
Cell viability and toxicity were assessed using the lactate
dehydrogenase (LDH) assay and the 3-(4,5-dimethylthia-
zol-2-yl)-2,5-diphenyltetrazolium (MTT) bromide assay to
investigate necrosis and mitochondrial redox metabolism,
respectively. Two compounds were analysed in more detail
in hiPSCs and hiPSC-CMs using high-content analysis
(HCA). The results indicate that the cell types selected for
toxicity screening have a major effect on the results. In addi-
tion, the data reveal structural features causing stem cell
toxicity in the 3i-1000 family of compounds.
compounds 1a–f were synthesized via O-(benzotriazol-1-
yl)-N,N,N′,N′-tetramethyluronium hexafluorophosphate
(HBTU)-mediated amide coupling of various carboxylic
acids (2a–f) and aminopyrazole 3a or aniline 3b (Fig. 2a).
The yields were moderate to good (49–73%). There can be
at least two reasons for slightly lower than expected yields
of amides: first, the nucleophilicity of the aniline-type start-
ing materials such as compound 3b is reduced compared to
aliphatic amines; and second, the heterocyclic carboxylic
acid starting materials (2a–f) are sterically hindered.
The used carboxylic acids were either commercially
available (2c–f) or synthesized via [3+2] dipolar cycloaddi-
tion reactions (2a, 2b) starting from affordable and commer-
cially available aldehydes (4a, 4b) and alkyne 5 (Fig. 2b).
The regiochemistry of the isoxazole-forming reaction is
dominated by the electrophilicity of the alkyne (Houk et al.
1973). Synthesis of the phenyl-substituted isoxazolecarbox-
ylic acid 2a started from benzaldehyde 4a by converting
it to the oxime 6a in the presence of hydroxylammonium
chloride and pyridine. The resulting oxime 6a was subjected
to Oxone®-mediated cycloaddition reaction (Han et al. 2014)
in the presence of ethyl 2-butynoate 5 and a catalytic amount
of potassium chloride. Hydrolysis of the resulting ester 7a
with aqueous sodium hydroxide gave the phenyl-substituted
isoxazolecarboxylic acid 2a.
Materials and methods
Compounds and reagents
The lead compound 3i-1000 was purchased from Pharma-
tory Ltd (Oulu, Finland). All other compounds targeting
GATA4 and its interaction with NKX2-5 were synthesized
at the Division of Pharmaceutical Chemistry and Technol-
ogy, Faculty of Pharmacy, University of Helsinki (Finland),
as described below. Cell culture media, supplements and
reagents were purchased from Gibco (Thermo Fisher Scien-
tific, Paisley, UK) unless otherwise stated. Dulbecco’s modi-
fied Eagle medium (DMEM), bovine serum albumin, insu-
lin–transferrin–sodium selenite media supplement, sodium
pyruvate, pancreatin, 4′,6-diamidino-2-phenylindole (DAPI)
and all reagents used in cytotoxicity assays were purchased
from Sigma-Aldrich (Steinheim, Germany). Collagenase
type II was purchased from Worthington Biochemical Cor-
poration (Lakewood, NJ, USA). Growth Factor-Reduced
Matrigel® was acquired from Corning (Bedford, MA, USA),
and small-molecule inhibitors Y-27632, CHIR99021 and
Wnt-C59 were bought from Tocris Bioscience (Bristol,
UK). Leukemia inhibitory factor (LIF), as well as gelatin
and fetal bovine serum (FBS) used in mESC and MEF cul-
tures, were from Merck Millipore. 5-Bromo-2′-deoxyuridine
(BrdU) was purchased from Abcam (Cambridge, UK). The
primary antibodies used in immunofluorescence stainings
were: polyclonal rabbit anti-Ki67 (Abcam), monoclonal rat
anti-BrdU (Abcam), and monoclonal mouse anti-α-actinin
(Sigma-Aldrich). The secondary antibodies used were all
purchased from Life Technologies (Eugene, OR, USA):
Alexa Fluor 488 goat anti-mouse lgG, Alexa Fluor 546
donkey anti-rabbit lgG, and Alexa Fluor 647 goat anti-rat
lgG. CellEvent™ Caspase-3/7 Green Detection Reagent,
MitoTracker® Orange CMTMRos probe, as well as Hoechst
33342 were from Invitrogen (Carlsbad, CA, USA).
For the synthesis of the 2-methyloxazole-substituted
isoxazolecarboxylic acid 2b a slightly different approach
was used. The corresponding oxime 6b was prepared from
2-methyloxazole-4-carbaldehyde 4b by treatment with
hydroxylammonium chloride in the presence of sodium
acetate. The carboxylic acid 2b was prepared from the
oxime 6b in the presence of ethyl 2-butynoate 5 using
[hydroxy(tosyloxy)iodo]benzene (HTIB)-mediated cycload-
dition reaction in water followed by hydrolysis of the formed
ester 7b with aqueous sodium hydroxide (Raihan et al.
2010). Water has been reported to be a good solvent for
dipolar cycloaddition reactions (Butler and Coyne 2010;
Kesornpun et al. 2016). Both of the prepared carboxylic
acids (2a, 2b) were used in the final amide coupling step
(Fig. 2a) without further purification.
Preparation of the N,N-disubstituted amine 1g (Fig. 2c)
is based on the initial imine (8) formation between isoxa-
zolyl amine 9 and 4-(diethylamino)benzaldehyde 10 in the
presence of acetic acid and anhydrous Na2SO4 followed by
reduction of the carbon–nitrogen double bond with NaBH4.
The resulting amine 11 was coupled with propionyl chloride
in the presence of 4-(dimethylamino)pyridine (DMAP) to
give the desired amine 1g in good yield.
Synthesis of compounds
Cell cultures
The synthesis of compounds (1a–g) is described in Fig. 2
COS-1 cells were used to investigate the compound effects
on GATA4–NKX2-5 synergistic gene activation using a
and details are provided in Supplementary Information. The
1 3

Archives of Toxicology
Fig. 2 Synthesis of com-
pounds. a Synthesis of amides.
Reagents and conditions: (i)
HBTU, DIPEA, DMF, rt, 1 d,
49–73%. b Synthesis of the
carboxylic acid intermediates.
Reagents and conditions: (ii)
H2NOH·HCl, pyridine, EtOH;
(iii) H2NOH·HCl, NaOAc;
(iv) Oxone®, KCl, H2O; (v)
HTIB, H2O. c Synthesis of the
N,N-disubstituted amide via
imine formation. Reagents and
conditions: (vi) AcOH, Na2SO4
(anhydr.) rt, overnight, 74%;
(vii) NaBH4, MeOH, THF, rt, 2
d, 37%; (viii) EtCOCl, DMAP,
pyridine, rt, 3 d, 74%
luciferase reporter assay (Kinnunen et al. 2015). Toxicity
profiling was carried out in eight cell types: the H9c2 myo-
blast cell line, NRVCs, CFs, mESCs, MEFs, mESC deriva-
tives from day 5 embryoid bodies (D5EBs), hiPSCs and
hiPSC-CMs. All cell cultures were maintained at 37 °C in a
humidified atmosphere of 5% carbon dioxide (CO2) and the
regular growth medium was used for compound exposures.
Cell culture methods and cardiomyocyte differentiation from
human pluripotent stem cells are described in Supplemen-
tary Information.
Children’s Hospital) (Arceci et al. 1993). The luciferase
reporter plasmid pGL3-3xHA containing three Nkx2-5
high-affinity binding sites has been described previously
(Kinnunen et al. 2015). To create mouse Nkx2-5 express-
ing pMT2-Nkx2-5, the cDNA was cloned from pFLCI-
Nkx2-5 (clone D330001C20, Imagenes), which was a
kind gift from R. Kerkelä (University of Oulu, Finland),
with primers containing the EcoRI restriction site (for-
ward: 5′-ATATATGAATTCGTCGCCACCATGTTCCCC
AG-3′ and reverse 5′-ATATATGAATTCCTACCAGGC
TCGGATGCCGTGC-3′) and ligated to pMT2 EcoRI
site. The correct clone was identified by digestion with
specific restriction enzymes and the sequence was veri-
fied by DNA sequencing (DNA Sequencing and Genom-
ics Laboratory, Institute of Biotechnology, University of
Helsinki, Finland).
Plasmids
The empty expression plasmid MT2 (pMT2) and mouse
Gata4 expressing pMT2-Gata4 plasmids were gifts
from D. B. Wilson (Department of Pediatrics, St. Louis
1 3

Archives of Toxicology
2010). Force fields suitable for small molecules (MMFF94x
and OPLS-AA) were applied for the molecule parameteriza-
tions and energy minimizations. Moreover, the default set-
tings were employed to score and rank the conformational
databases. The lowest energy conformation was selected to
define the torsion angle between the ring planes in the south-
ern part of the compound. Mogul v.1.7.2 software utiliz-
ing knowledge-based library of molecular geometries from
the Cambridge Structural Database (CSD) was applied to
provide precise experimentally derived information about
preferred ring system geometries of 3i-1000 and 3i-1047
(Bruno et al. 2004).
Luciferase assays
COS-1 cells grown on 96-well plates were transfected with
the luciferase reporter plasmid pGL3-3xHA (100 ng/well)
and equal amounts of expression vectors pMT2-Gata4 and
pMT2-Nkx2-5 (total of 50 ng/well) using Fugene 6 (Pro-
mega) at a reagent:DNA ratio of 3:1. Cells transfected with
either pMT2, pMT2-Gata4 or pMT2-Nkx2-5 were used to
confirm synergistic transactivation of the reporter gene in
all experiments. After 6 h of transfection, the cells were
treated with compounds or vehicle (0.1% dimethyl sulfox-
ide, DMSO) and the luminescence was measured 24 h later
using Neolite Reporter Gene Assay System (PerkinElmer,
Turku, Finland) according to manufacturer’s instructions and
Victor2 plate reader (PerkinElmer, Turku, Finland). Each
compound was tested in three technical replicates in at least
two independent experiments.
Automated fluorescence microscopy
and high‑content analysis
For proliferation studies, the cells were exposed to the test
compounds for 24 h and 10 µM BrdU was added to the cul-
ture medium for the last 1 h (hiPSCs) or 24 h (hiPSC-CMs)
before fixation with 4% paraformaldehyde (PFA) for 15 min
at room temperature (rt). The cells were then permeabilized
with 0.1% Triton X-100 for 10 min and DNA was hydro-
lysed with 2 M hydrochloric acid for 30 min at rt followed
by neutralization with 0.1 M sodium borate (pH 8.5) for
30 min. Non-specific binding sites were blocked with 4%
FBS in phosphate-buffered saline (PBS) for 45 min at rt,
whereafter the cells were incubated with anti-Ki67 (1:250)
and anti-BrdU (1:250) antibodies for 60 min at rt. Cells were
then washed 3×5 min with PBS and incubated with Alexa
Fluor-conjugated secondary antibodies (1:200) and DAPI
(5 µg/ml) for 45 min at rt. In hiPSC-CM cultures, staining
with primary antibody against α-actinin (at 1:600) was addi-
tionally used to identify myocytes. The plates were imaged
and analysed with CellInsight CX5 High Content Screening
Platform (Thermo Scientific) using a 10× objective (Olym-
pus UPlanFL N 10×/0.3). For quantification, the cells were
first identified based on DAPI fluorescence, which defined
the nuclear area for quantifications. In hiPSC-CM cultures,
non-myocytes were excluded based on absence of α-actinin
staining and nuclear area. The thresholds for α-actinin fluo-
rescence intensity and nuclear area were set manually in
each experiment to allow optimal exclusion of non-myo-
cytes. In hiPSC-CMs cultures, the data were collected only
from α-actinin positive cells. The thresholds for Ki67+ and
BrdU+ cells were also set manually in each experiment as
described previously (Mioulane et al. 2012) to adjust for
slight variation in staining intensity.
Cytotoxicity assays
The cells were exposed to the test compounds for 24 h and
cytotoxicity was quantified with MTT and LDH assays as
described previously (Talman et al. 2011). For the deter-
mination of cell membrane integrity, 50 µl of medium
from each well was transferred onto a new 96-well plate
followed by addition of 50 µl of substrate solution contain-
ing 1.3 mM β-nicotinamide adenine dinucleotide, 660 µM
iodonitrotetrazolium, 54 mM l(+)-lactic acid, 280 µM phen-
azine methosulphate and 0.2 M Tris–HCl (pH 8.0). After a
30-min incubation, the reaction was stopped by adding 50 µl
of 1 M acetic acid to each well and absorbance was meas-
ured at 490 nm using Victor2 plate reader (PerkinElmer,
Turku, Finland). Spontaneous LDH release was measured
from untreated cells, maximal LDH release from cells
lysed with 0.9% Triton X-100 and background absorbance
was measured from wells without cells. After background
extraction, cytotoxicity was calculated as follows: cytotox-
icity (%)=[(sample−spontaneous LDH release)/(maximal
LDH release−spontaneous LDH release)]×100. For quan-
tification of metabolic activity, 3-(4,5-dimethylthiazol-2-yl)-
2,5-diphenyltetrazolium bromide (MTT) was added to the
cells at a final concentration of 0.5 mg/ml followed by 2 h
incubation in cell culture conditions. The medium was then
aspirated and formazan crystals were solubilized in DMSO.
Absorbance was measured at 550 nm and absorbance at
650 nm was subtracted as background.
To study mitochondrial function using high-content
analysis, the cells were plated and treated as in proliferation
studies, except that MitoTracker Orange CMTMRos solu-
tion (400 nM) was added to the cells for the last 30 min at
37 °C. The accumulation of the MitoTracker probe used is
dependent upon mitochondrial membrane potential allowing
Computational chemistry
Commercial modeling package MOE 2014.09 (Chemical
Computing Group Inc., Montreal, Canada; http://www.
chemcomp.com) with LowModeMD module was utilized
to generate small-molecule conformation databases (Labute
1 3

Archives of Toxicology
selective staining of active mitochondria. Following fixation
and permeabilization, hiPSC-CMs were in addition stained
with anti-α-actinin (Alexa Fluor 488) and DAPI. Cells were
identified and hiPSC-CMs discriminated from non-myocytes
as described above. MitoTracker intensity corresponding to
active mitochondria was analysed in the perinuclear area
defined by a 3-pixel ring or a 5-pixel ring around the nucleus
in hiPSCs and hiPSC-CMs, respectively. The fluorescence
intensity threshold was set manually in each experiment
based on the assumption that 2% of control hiPSC-CMs and
5% of control hiPSCs were MitoTracker negative.
3i-1000, 3i-1120 and 3i-1148 exhibited an inhibitory effect
on GATA4- and NKX2-5-induced gene activation, whereas
compounds 3i-1047, 3i-1165 and 3i-1229 had only a weak
effect and compounds 3i-1051 and 3i-1228 were ineffective
(Fig. 3).
The GATA4‑targeted compounds do not induce cell
necrosis
When exposed to the test compounds for 24 h, none of
the cell types investigated exhibited a significant necrotic
response as measured by LDH release into the culture
medium (Supplementary Fig. S1). More specifically, com-
pound-induced cytotoxicity was always less than 15% of the
maximal LDH release regardless of the compound and the
cell type.
To study caspase activation, the cells were exposed to
the test compounds for 4 h and then incubated with 3–5 µM
CellEvent™ Caspase-3/7 Green Detection Reagent solution
in PBS with 5% FBS for 50 min and stained with Hoechst for
10 min at 37 °C. Cells were fixed with 4% PFA and scanned
with CellInsight using the 10× objective, or alternatively,
the staining solution was replaced with fresh 5% FBS in
PBS followed by live cell imaging with CellInsight. Hoechst
fluorescence was used to identify the cells and define the
nuclear area. The intensity of green fluorescence within the
nucleus was quantified as a measure of caspase 3/7 activity.
The fluorescence intensity threshold for caspase positive and
caspase negative cells was set manually in each experiment
based on the assumption that 5% of control hiPSCs exhibited
caspase 3/7 activity.
Toxicity of GATA4‑targeted compounds is cell
type‑specific
In the MTT assay, the effects of individual compounds on
cell viability differed substantially between the cell types
studied as exemplified by compounds 3i-1000, 3i-1047,
3i-1051 and 3i-1229 (Fig. 4). For all compounds tested, a
comparison between hiPSCs and hiPSC-CMs is presented
in Fig. 5 and for other cell types in Supplementary Figure
S2. Overall, cardiomyocytes, fibroblasts, and H9c2 cells
were the most resistant cell types, as none of the compounds
induced significant toxicity in these cells. The two fibroblast
types (CFs and MEFs) exhibited almost identical responses
to the small molecules: none of the compounds had a con-
siderable effect on CF viability, while only 3i-1120 at 30 µM
induced modest toxicity in MEFs (Supplementary Fig. S2;
Statistics
Results are expressed as mean+ SEM from at least three
independent experiments. Data were analysed using IBM
SPSS Statistics 24 software. Homogeneity of variances
was tested using Levene’s test. If the data met the assump-
tions of the test, one-way ANOVA and a Tukey HSD post
hoc test were carried out; otherwise Welch ANOVA and a
Games–Howell post hoc test were performed. Additionally,
factorial ANOVA was carried out for hiPSC viability data.
Differences at the level of P<0.05 were considered statisti-
cally significant.
Results
Effects on GATA4–NKX2‑5 interaction
A luciferase reporter assay with an artificial promoter
containing three high-affinity binding sites for NKX2-5
was used to evaluate the effects of the compounds on
GATA4–NKX2-5 transcriptional synergy (Kinnunen
et al. 2015). We have previously shown that while GATA4
alone induces only minimal activation of this promoter,
it strongly potentiates NKX2-5-mediated transactivation
(Kinnunen et al. 2015; Välimäki et al. 2017). Compounds
Fig. 3 Effects of the small-molecule compounds on GATA4- and
NKX2-5-induced synergistic gene activation. Luciferase reporter
activity was normalized to vehicle (DMSO) control and is presented
as mean+SEM (n=2–3, apart from 3i-1000; n=13)
1 3

Archives of Toxicology
Fig. 4 Cell type-dependent toxicity of GATA4-targeted compounds.
The results show the effects of a 3i-1000, b 3i-1047, c 3i-1051, and
d 3i-1229 on the viability of all eight cell types tested. Cell viabil-
ity was determined with the MTT assay after a 24-h compound expo-
sure. Results are expressed as mean+SEM (n=3–4). ***P<0.001
vs. DMSO; **P<0.01 vs. DMSO; *P<0.05 vs. DMSO (one-way
ANOVA followed by Tukey’s HSD or Welch ANOVA followed by
Games–Howell)
not statistically significant). Also the two cardiomyocyte
types, hiPSC-CMs and NRVCs, acted comparably: none
of the compounds caused toxicity in either of the cardio-
myocyte types. However, 3i-1047 induced a 51% increase
in hiPSC-CM metabolic activity (P=0.033), which was not
observed in NRVCs (Fig. 4b). The mechanism(s) of this
increase remain to be established, but it might be due to
increased cell proliferation, higher mitochondrial content or
increased metabolic activity.
as well as a moderate, but not statistically significant
(P = 0.088) reduction in metabolic activity of D5EBs
at ≥ 10 µM (Fig. 4a). As for 3i-1047, smaller 38–45%
reductions in the viability of mESCs (P = 0.100), D5EBs
(P=0.612) and hiPSCs (P=0.021) were observed, and this
was apparent only at the highest concentration of 30 µM
(Fig. 4b). These cell types were also sensitive to compounds
3i-1120, 3i-1148 and 3i-1229 (Fig. 5b; Supplementary Fig.
S2). In general, the hiPSCs were the most sensitive cell
type based on the IC50 values (3i-1000, 1.46 µM; 3i-1120,
0.60 µM; 3i-1148, 0.30 µM, and 3i-1229, 2.07 µM) and on
the maximal toxicity induced at the highest concentration
tested. The effects of both treatment and dose, as well as the
treatment–dose interaction were all statistically significant
(P<0.001, factorial ANOVA).
Unlike in cardiomyocytes, fibroblasts and H9c2 cells,
significant toxicity was observed with several of the test
compounds in mESCs, hiPSCs and D5EBs (Fig. 4, Sup-
plementary Fig. S2). As an example, 3i-1000 induced pro-
found reductions in cell viability in mESCs (56% at 30 µM,
P = 0.006) and hiPSCs (93–94% at ≥ 3 µM, P < 0.001),
1 3

Archives of Toxicology
Fig. 5 Effects of the test compounds on a hiPSC-CM and b hiPSC
viability. Cell viability was determined with the MTT assay after a
24-h compound exposure. Results are expressed as mean+SEM
(n=3–4). ***P<0.001 vs. DMSO; **P<0.01 vs. DMSO; *P<0.05
vs. DMSO (one-way ANOVA followed by Tukey’s HSD or Welch
ANOVA followed by Games–Howell)
0° to 19°. In this context, the ability of the compounds
to reach both the coplanar orientation and the global low
energy conformation is defined by steric factors linked to
the internal energy of the compounds. For instance, steric
hindrance in the 3i-1000 family of compounds, with two
C–H bonds (ortho position) in the six-membered phenyl
ring, enforces larger dihedral angles into the ring system
due to overlapping electron clouds and associated increase
in internal energy. However, in case of compounds with a
five-membered ring (3i-1047 family), the critical molecular
area is less crowded and allows the compounds to adopt a
periplanar, and in the case of 3i-1228, almost a coplanar
orientation without extra cost in energy. Moreover, because
of the presence of a heteroatom in the five-membered ring of
the 3i-1047 family compounds, an additional intramolecular
hydrogen bond is formed and associated favorably to the
lowest energy conformations.
Structure–toxicity relationships
When comparing the effects of different compounds within
each individual cell type, the compounds can be grouped
into two main classes: toxic (3i-1000, 3i-1120, 3i-1148 and
3i-1229) and non-toxic (3i-1047, 3i-1051, 3i-1165, and
3i-1228) compounds. Based on the hiPSC toxicity data,
3i-1148 can be considered as the most toxic compound with
an IC50 of 0.30 µM (Fig. 5b). Structurally, all of the deriva-
tives of 3i-1000, despite the fact that they all possess unique
structural modification in the southern, middle or northern
part of the compound, were classified as toxic compounds
based on their effects on stem cell and progenitor cell viabil-
ity. Surprisingly, toxicity was preserved despite the extensive
structural modifications in the middle and northern parts
(3i-1120 and 3i-1229, respectively). These findings suggest
that the southern part of the compound is predominantly
responsible for the toxic outcome in stem cells.
Parallel conformational evaluation of ring orientations
in the southern part was carried out with knowledge-based
approach relying on data derived from small-molecule crys-
tal structures. Conformational analysis with Mogul (Cam-
bridge Crystallographic Data Center) provides experimen-
tally validated approximation of the specific torsion angle
for the ring systems in the southern part (Bruno et al. 2004).
The data suggest that the 3i-1047 compound family with a
five-membered ring in the southern part adopts a signifi-
cantly flatter ring geometry in comparison to the six-mem-
bered ring systems in the 3i-1000 compound family. This is
in agreement with the conformation analysis measured by
force field methods and correlates directly with stem cell
toxicity observed with the 3i-1000 family of GATA4-tar-
geted compounds.
To characterize the structural determinants of toxicity,
a force field-based conformational search was carried out
(Labute 2010). The analysis suggests that the torsional
angles for the ring system geometry in the southern part
are distinct for the 3i-1000 and 3i-1047 compound fami-
lies (Fig. 6). Two different force fields (MMFF94x and
OPLS-AA) suitable for small organic compounds were
applied for the parametrization of the molecule structures.
The six-membered ring bound to isoxazoles of the 3i-1000
family showed consistent torsional angles from 28° to 51°
for the ring geometry with both tested force field settings,
whereas, the five-membered ring bound to isoxazoles of
the 3i-1047-family preferred a flatter orientation of the
ring system in the southern part with values ranging from
1 3

Archives of Toxicology
Fig. 6 Structure-based toxicity relay on the consistent conformational
geometry identified in the southern part of the 3i-1000 and related
compounds. a Test compounds were classified into two structural
categories omitting a five-membered or a six-membered ring bound
to the isoxazole (3i-1000 or 3i-1047 families). b Force field-based
calculations (MMFF94x) revealed family-correlated conformations
for the representative compounds 3i-1000 and 3i-1047. c Knowledge-
based conformational analysis with Mogul (Cambridge Structural
Database) suggests unique set of torsion angles for both compound
families
3i-1000 to represent the more toxic compounds and
3i-1047 as a non-toxic representative. HiPSCs and
hiPSC-CMs were selected for HCA assays as representa-
tives of sensitive and resistant cell types, respectively.
High‑content analysis of cell viability
Based on the LDH and MTT results, two compounds
were chosen for HCA of cell viability and proliferation:
1 3

Archives of Toxicology
To compare HCA-based cell viability analysis with
the more conventional MTT assay, cell viability was
assessed using a mitochondrion stain (MitoTracker),
whose accumulation in mitochondria is dependent on
mitochondrial membrane potential. Correspondingly to
the MTT results with hiPSCs, while 3i-1047 did not affect
MitoTracker staining, 3i-1000 impaired mitochondrial
function in hiPSCs at 10 µM concentration, as reflected
by a 5.7-fold (P = 0.001) increase in the percentage of
MitoTracker negative cells as compared to DMSO-treated
cells (Supplementary Fig. S3). Accordingly, the average
MitoTracker intensity in the perinuclear area reduced
substantially in cells exposed to 10 µM 3i-1000 (Sup-
plementary Fig. S3). Of note, the cell death induced by
3i-1000 at 10 µM also reduced hiPSC cell numbers by
98%. Normalized cell densities are shown in Supplemen-
tary Figure S3. The compounds had no effect on the pro-
portion of MitoTracker negative hiPSC-CMs, and induced
a slight increase in the average MitoTracker intensity in
hiPSC-CMs (Supplementary Fig. S3). This increased
MitoTracker staining may be due to increased numbers
of mitochondria in hiPSC-CMs, which would be in line
with the results for 3i-1047 in the MTT assay.
High‑content analysis of stem cell proliferation
In hiPSCs, the compound 3i-1000 decreased the proportion
of Ki67 positive (P=0.001) and BrdU-positive (P<0.001)
cells significantly at 10 µM, but not at 1 µM (Fig. 8),
and this was paralleled by decreased average intensity of
nuclear Ki67 and BrdU stainings (Supplementary Fig. S4).
The compound also induced a considerable 77% decrease
(P=0.062) in cells with high total DNA staining intensity
representing cells in G2 phase of the cell cycle (Fig. 8) and
a 66% decrease in the average total intensity of DNA stain-
ing (Supplementary Fig. S4). In contrast, 3i-1047 (at 1 or
10 µM) had no effect on the percentage of Ki67+, BrdU+
or high DNA cells (Fig. 8) or on the corresponding average
staining intensities (Supplementary Fig. S4). Cell numbers
were significantly reduced in response to 3i-1000 treatment
as shown in Fig. 8a and Supplementary Figure S3.
The toxic GATA4‑targeted compounds induce
apoptosis in stem cells
Finally, to elucidate the mechanism of cell death caused
by the toxic compounds in hiPSCs, caspase-3/7 activation
was analysed using HCA. To minimize the cell loss caused
by cell death and detachment, a shorter 4-h exposure time
was used. The compound 3i-1000 induced a 3.2-fold ( 0.5;
SEM) increase in cells positive for the fluorescent caspase
reporter (P=0.009), whereas 3i-1047 had no effect on cas-
pase-3/7 activity (Fig. 9). The findings are in agreement with
the MTT and LDH test results, which showed no signs of
necrosis, yet significant reductions in cell viability after 24 h
exposure to 3i-1000, and thus suggest the caspase-dependent
apoptosis as a mechanism contributing to stem cell death
caused by the toxic GATA4-targeted compounds.
High‑content analysis of hiPSC‑CM proliferation
The effects of 3i-1000 and 3i-1047 on the proliferation of
hiPSC-CMs were analysed using Ki67 and BrdU immu-
nostainings and HCA. At the concentration of 30 µM both
3i-1000 and 3i-1047 induced 1.7- and 1.5-fold increases,
respectively, in the percentage of Ki67+ cells compared
to DMSO-treated cells (Fig. 7), paralleling with substan-
tial increases in the average intensity of nuclear Ki67
staining (Supplementary Fig. S4), indicating that the
compounds may promote cell cycle activation. This was,
however, not linked to a similar increase in the propor-
tion of BrdU+ hiPSC-CMs. The compound 3i-1047 had
no effect on the percentage of BrdU+ cells (Fig. 7) or on
the average intensity of nuclear BrdU staining (Supple-
mentary Fig. S4). Interestingly, there was a tendency for
3i-1000 to induce a decrease (43%, P = 0.233) in BrdU+
hiPSC-CMs as compared to DMSO (Fig. 7) and in the
average intensity of nuclear BrdU staining (Supplemen-
tary Fig. S4). Furthermore, neither 3i-1000 nor 3i-1047
caused significant changes in the percentage of cells with
high total intensity of DNA staining compared to DMSO
(Fig. 7) or in the average total intensity of DNA stain-
ing (Supplementary Fig. S4), indicating that the propor-
tion of cells in the G2 phase of the cell cycle was not
affected. Taken together, the compounds did not have a
consistent inhibitory or promoting effect on hiPSC-CM
proliferation.
Discussion
Toxicity is a major cause of failure in drug development pro-
jects (Kola and Landis 2004). In general, toxic compounds
are identified already during lead compound screening;
however, sometimes they only come up in clinical trials. As
the monetary cost of failure increases steeply with progres-
sion towards clinical trials (Paul et al. 2010), more efficient
toxicity screening in early phases could improve research
and development productivity noticeably. Considering the
preclinical phase of drug development, the use of more pre-
cise cell models for cytotoxicity screening prior to in vivo
testing would not only improve the validity of the in vivo
efficacy data, but also would reduce the use of experimen-
tal animals. Therefore, in this study we sought to compare
different cardiac and stem cell types of rodent and human
origin for cytotoxicity screening. Additionally, our aim was
1 3

Archives of Toxicology
Fig. 7 Effects of compounds 3i-1000 and 3i-1047 on cell prolif-
eration in hiPSC-CMs. After a 24-h compound exposure and a con-
comitant BrdU-loading (24 h), the cells were fixed and stained for
α-actinin, DNA (DAPI), Ki67 and BrdU. Imaging and analysis was
carried out using CellInsight High Content Analysis Platform. a Rep-
resentative images from four independent experiments are shown.
Adjustments of individual color channels to enhance brightness and
contrast were made identically to all representative images. The
quantifications for the proportion of b Ki67-positive cells, c BrdU-
positive cells, and d cells with high DNA staining are expressed as
mean+SEM (n=4). No statistically significant differences were
observed compared to DMSO (one-way ANOVA followed by Tuk-
ey’s HSD)
to characterize the in vitro cytotoxicity of a selected set of
novel cardiac transcription factor-targeted compounds to
identify toxicity-related molecular structures.
by carrying out cytotoxicity screening in eight differ-
ent cell types. The northern part of the 3i-1000 family
of compounds exhibits a major contribution to inhibition
of GATA4–NKX2-5 synergy (Välimäki et al. 2017). Fur-
thermore, our preliminary data indicated that modifica-
tions in the southern part may affect the toxicity of the
compounds in some cell types. Compounds for the study
According to our previous study, the lead compound
3i-1000 exhibited no toxicity in NRVCs at concentrations
up to 50 µM (Välimäki et al. 2017). Here we investigated
in detail the safety of the compound and its derivatives
1 3

Archives of Toxicology
Fig. 8 Effects of compounds 3i-1000 and 3i-1047 on cell prolif-
eration in hiPSCs. After a 24-h compound exposure, the cells were
loaded with BrdU (1 h) and subsequently fixed and stained for DNA
(DAPI), Ki67 and BrdU. Imaging and analysis was carried out using
CellInsight High Content Analysis Platform. a Representative images
from three independent experiments are shown. Adjustments of indi-
vidual color channels to enhance brightness and contrast were made
identically to all representative images. The analysis results for the
percentage of b Ki67-positive cells, c BrdU-positive cells, and d
cells with high DNA staining are expressed as mean+SEM (n=3).
***P<0.001 vs. DMSO; **P<0.01 vs. DMSO; *P<0.05 vs. DMSO
(one-way ANOVA followed by Tukey’s HSD or Welch ANOVA fol-
lowed by Games–Howell)
were, therefore, preselected to uncover and identify toxic-
ity-related molecular properties especially concerning the
modifications in the southern part whereas diversity in the
northern part was not preferred. Using suitably function-
alized northern and southern fragments, structural varia-
tions were easily introduced in different regions of mol-
ecules, fitting to the general structure of GATA4-targeted
compounds.
In line with our previous findings, 3i-1000 and its deriv-
atives were non-toxic to cardiomyocytes, fibroblasts and
H9c2 cardiac myoblasts. The comparison of NRVCs and
hiPSC-CMs was of particular interest, as the use of stem
cell-derived cardiomyocytes has grown exponentially over
the past couple of years upon improved differentiation and
purification techniques. There is intensive discussion on the
applicability of hiPSC-CMs for drug development research
1 3

Archives of Toxicology
Fig. 9 Effect of compounds 3i-1000 and 3i-1047 on caspase activa-
tion in hiPSCs. After a 4-h compound exposure, the cells were incu-
bated with a detection reagent for activated Caspase-3/7 for 50 min
and the nuclei were stained with Hoechst for 10 min. Fixed or live
cells were imaged with CellInsight High Content Analysis Platform.
a Representative images from four independent experiments are
shown. Adjustments of individual color channels to enhance bright-
ness and contrast were made identically to all representative images.
b Quantification of the proportion of cells positive for fluorescent
caspase 3/7 activity reporter is expressed as mean+SEM (n=6).
**P<0.01 vs. DMSO (Welch ANOVA followed by Games–Howell)
and cardiotoxicity testing due to their immature phenotype,
which is linked to differences in metabolism, sarcomere
organization, gene expression and G protein-coupled recep-
tor (GPCR) signalling (Földes et al. 2014; Magdy et al.
2018). In the present study, the GATA4-targeted compounds
were non-toxic to both cardiomyocyte types, suggesting that
they do not exhibit species-specific cardiotoxicity and sup-
port the use of hiPSC-CMs instead of primary cells in early
toxicity screening to reduce the use of experimental animals.
Conversely, the screening in stem cells and stem cell-
derived progenitor cells revealed that some of the com-
pounds exhibit potent and perhaps unexpected toxicity
resulting in apoptosis. Stem cells are known to be highly
sensitive to rapid apoptosis in response to DNA damage
(Dumitru et al. 2012; Liu et al. 2013, 2014). Multiple mecha-
nisms have been proposed to contribute to the hypersensitiv-
ity, including tumor suppressor protein p53 signalling, short
duration of cell cycle, as well as mitochondrial priming (e.g.,
disparities in the expression of pro-apoptotic and anti-apop-
totic proteins, and differences in mitochondrial structure and
activity), but the exact mechanisms are still unknown. The
compound-induced toxicity was selective to stem cells and
progenitor cells, as it was not observed in other proliferating
cells such as H9c2 cells, MEFs or CFs.
compound classes is the dihedral angle θ between the two
aromatic planes in the southern part. Moreover, the effect
of the small molecules on GATA4–NKX2-5 transcrip-
tional synergy in luciferase reporter assay did not predict
the cellular toxicity; for example, compound 3i-1229 is
an example of a toxic compound with no inhibitory effect
on GATA4–NKX2-5 interaction. Based on our previ-
ously reported protein kinase and GPCR profiling results
(Välimäki et al. 2017) we also find it unlikely that a direct
effect on protein kinases or GPCRs would contribute to
the observed stem cell toxicity. Nevertheless, based on the
current results, further lead optimization can be directed
towards the non-toxic derivatives, i.e., compounds with a
five-membered ring in the southern part of the molecule.
In summary, by screening novel GATA4-targeted small-
molecular compounds for cytotoxicity in several cell types,
we revealed profound differences in the sensitivity of differ-
ent cell types, implicating that the model for toxicity screen-
ing should be chosen carefully. Furthermore, the discovery
that small changes in the southern part of the molecular
structure of GATA4-targeting compounds contribute to the
toxicity in stem cells and progenitor cells allows further drug
development using the non-toxic derivatives as a starting
point. We propose that toxicity screening using stem cells
in early phases of drug discovery projects can be useful in
reducing the risk of toxicity-dependent failure at later stages
of drug development, particularly when targeting transcrip-
tion factors.
The structure–toxicity analysis reveals a structurally
coherent set of GATA4-targeted compounds with a common
nominator, a phenyl ring in the southern part, as a major
cause for the observed progenitor- and stem cell-specific
toxicity. Detailed structural analysis of toxic compounds
reveals the uniform appearance of low energy conforma-
tions and consistent stereochemistry of the ring system in
southern part defined by force field- and knowledge-based
methods. Based on the present results, the characteristic con-
formational feature that distinguishes the toxic and non-toxic
Acknowledgements We thank Dr. Ingo Aumüller for the synthesis of
five compounds and comments on the manuscript, Dr. Petteri Piep-
ponen for sharing his expertise in statistical analysis, as well as Ms
Vilma Ivaska, Ms. Liisa Lappalainen and Ms. Annika Korvenpää for
technical assistance. The Light Microscopy Unit (Institute of Bio-
technology and HiLIFE, University of Helsinki) and the Biomedicum
1 3

Archives of Toxicology
Imaging Unit (HiLIFE, University of Helsinki) are acknowledged
for providing instrumentation and support for high-content analysis.
The research was funded by the Finnish Funding Agency for Inno-
vation (Tekes, 3iRegeneration, Project 40395/13), the Academy of
Finland (Project 2666621), the Finnish Foundation for Cardiovascular
Research, and the Sigrid Jusélius Foundation.
Hirvonen J, Ruskoaho H, Airaksinen AJ, Santos HA (2017) Drug-
loaded multifunctional nanoparticles targeted to the endocardial
layer of the injured heart modulate hypertrophic signaling. Small
13:1701276. https://doi.org/10.1002/smll.201701276
Földes G, Matsa E, Kriston-Vizi J, Leja T, Amisten S, Kolker L,
Kodagoda T, Dolatshad NF, Mioulane M, Vauchez K, Aranyi T,
Ketteler R, Schneider MD, Denning C, Harding SE (2014) Aber-
rant alpha-adrenergic hypertrophic response in cardiomyocytes
from human induced pluripotent cells. Stem Cell Rep 3:905–914
Garg V, Kathiriya IS, Barnes R, Schluterman MK, King IN, Butler CA,
Rothrock CR, Eapen RS, Hirayama-Yamada K, Joo K, Matsuoka
R, Cohen JC, Srivastava D (2003) GATA4 mutations cause human
congenital heart defects and reveal an interaction with TBX5.
Nature 424:443–447
Author contributions JYK, HR and VT designed and supervised the
work. MJV designed the compounds and performed computational
structure–activity and structure–toxicity analyses. MJ and SA synthe-
sized and analysed the new compounds under the supervision of GBG
and JYK. SMK carried out molecular cloning and GATA4–NKX2-5
interaction studies. STK, RSL, LP and VT carried out cell culture and
differentiation, and STK, LP and GBG performed the toxicity screen-
ing. STK, VT and GF carried out the high-content analysis. STK, MJV,
MJ, SMK, RSL, GF, GBG, JYK, HR and VT contributed to the writing
of the manuscript. All authors reviewed and accepted the manuscript.
Gupta V, Gemberling M, Karra R, Rosenfeld GE, Evans T, Poss KD
(2013) An injury-responsive gata4 program shapes the zebrafish
cardiac ventricle. Curr Biol 23:1221–1227
Han L, Zhang B, Zhu M, Yan J (2014) An environmentally benign
synthesis of isoxazolines and isoxazoles mediated by potassium
chloride in water. Tetrahedron Lett 55:2308–2311
Data availability The datasets generated during and analysed during
the current study are available from the corresponding author on rea-
sonable request.
Horvath P, Aulner N, Bickle M, Davies AM, Nery ED, Ebner D, Mon-
toya MC, Ostling P, Pietiainen V, Price LS, Shorte SL, Turcatti
G, von Schantz C, Carragher NO (2016) Screening out irrelevant
cell-based models of disease. Nat Rev Drug Discov 15:751–769
Houk KN, Sims J, Watts CR, Luskus LJ (1973) Origin of reactivity,
regioselectivity, and periselectivity in 1,3-dipolar cycloadditions.
J Am Chem Soc 95:7301–7315
Compliance with ethical standards
Conflict of interest SMK, MJV, MJ, GBG, JYK and HR are inventors
in a pending patent application “Pharmaceutical compounds” (PCT/
FI2017/050661). The authors have no other competing interests to dis-
close.
Kerbrat A, Ferre JC, Fillatre P, Ronziere T, Vannier S, Carsin-Nicol
B, Lavoue S, Verin M, Gauvrit JY, Le Tulzo Y, Edan G (2016)
Acute neurologic disorder from an inhibitor of fatty acid amide
hydrolase. N Engl J Med 375:1717–1725
Open Access This article is distributed under the terms of the Crea-
tive Commons Attribution 4.0 International License (http://creativeco
mmons.org/licenses/by/4.0/), which permits unrestricted use, distribu-
tion, and reproduction in any medium, provided you give appropriate
credit to the original author(s) and the source, provide a link to the
Creative Commons license, and indicate if changes were made.
Kesornpun C, Aree T, Mahidol C, Ruchirawat S, Kittakoop P (2016)
Water-assisted nitrile oxide cycloadditions: synthesis of isoxa-
zoles and stereoselective syntheses of isoxazolines and 1,2,4-oxa-
diazoles. Angew Chem Int Ed 55:3997–4001
Kikuchi K, Holdway JE, Werdich AA, Anderson RM, Fang Y, Egnac-
zyk GF, Evans T, Macrae CA, Stainier DY, Poss KD (2010) Pri-
mary contribution to zebrafish heart regeneration by gata4(+)
cardiomyocytes. Nature 464:601–605
Kinnunen S, Välimäki M, Tölli M, Wohlfahrt G, Darwich R, Komati H,
Nemer M, Ruskoaho H (2015) Nuclear receptor-like structure and
interaction of congenital heart disease-associated factors GATA4
and NKX2-5. PLoS One 10:e0144145. https://doi.org/10.1371/
journal.pone.0144145
References
Ang YS, Rivas RN, Ribeiro AJ, Srivas R, Rivera J, Stone NR, Pratt K,
Mohamed TM, Fu JD, Spencer CI, Tippens ND, Li M, Narasimha
A, Radzinsky E, Moon-Grady AJ, Yu H, Pruitt BL, Snyder MP,
Srivastava D (2016) Disease model of GATA4 mutation reveals
transcription factor cooperativity in human cardiogenesis. Cell
167:1734–1749
Kinnunen SM, Tölli M, Välimäki MJ, Gao E, Szabo Z, Rysä J, Ferreira
MPA, Ohukainen P, Serpi R, Correia A, Mäkilä E, Salonen J,
Hirvonen J, Santos HA, Ruskoaho H (2018) Cardiac actions of a
small molecule inhibitor targeting GATA4–NKX2-5 interaction.
Sci Rep 8:4611
Arceci RJ, King AA, Simon MC, Orkin SH, Wilson DB (1993) Mouse
GATA-4: a retinoic acid-inducible GATA-binding transcription
factor expressed in endodermally derived tissues and heart. Mol
Cell Biol 13:2235–2246
Kohli S, Ahuja S, Rani V (2011) Transcription factors in heart: promis-
ing therapeutic targets in cardiac hypertrophy. Curr Cardiol Rev
7:262–271
Kola I, Landis J (2004) Can the pharmaceutical industry reduce attri-
tion rates? Nat Rev Drug Discov 3:711–715
Bruno IJ, Cole JC, Kessler M, Luo J, Motherwell WD, Purkis LH,
Smith BR, Taylor R, Cooper RI, Harris SE, Orpen AG (2004)
Retrieval of crystallographically-derived molecular geometry
information. J Chem Inf Comput Sci 44:2133–2144
Labute P (2010) LowModeMD–implicit low-mode velocity filtering
applied to conformational search of macrocycles and protein
loops. J Chem Inf Model 50:792–800
Butler RN, Coyne AG (2010) Water: nature’s reaction enforcer-com-
parative effects for organic synthesis “in-water” and “on-water”.
Chem Rev 110:6302–6337
Lee Y, Shioi T, Kasahara H, Jobe SM, Wiese RJ, Markham BE, Izumo
S (1998) The cardiac tissue-restricted homeobox protein Csx/
Nkx2.5 physically associates with the zinc finger protein GATA4
and cooperatively activates atrial natriuretic factor gene expres-
sion. Mol Cell Biol 18:3120–3129
Dumitru R, Gama V, Fagan BM, Bower JJ, Swahari V, Pevny LH,
Deshmukh M (2012) Human embryonic stem cells have consti-
tutively active Bax at the Golgi and are primed to undergo rapid
apoptosis. Mol Cell 46:573–583
Li RG, Li L, Qiu XB, Yuan F, Xu L, Li X, Xu YJ, Jiang WF, Jiang JQ,
Liu X, Fang WY, Zhang M, Peng LY, Qu XK, Yang YQ (2013)
GATA4 loss-of-function mutation underlies familial dilated car-
diomyopathy. Biochem Biophys Res Commun 439:591–596
Ferreira MPA, Ranjan S, Kinnunen S, Correia A, Talman V, Makila
E, Barrios-Lopez B, Kemell M, Balasubramanian V, Salonen J,
1 3

Archives of Toxicology
Liu JC, Guan X, Ryan JA, Rivera AG, Mock C, Agrawal V, Letai A,
Lerou PH, Lahav G (2013) High mitochondrial priming sensi-
tizes hESCs to DNA-damage-induced apoptosis. Cell Stem Cell
13:483–491
Rysä J, Tenhunen O, Serpi R, Soini Y, Nemer M, Leskinen H, Rusko-
aho H (2010) GATA-4 is an angiogenic survival factor of the
infarcted heart. Circ Heart Fail 3:440–450
Seiler AE, Spielmann H (2011) The validated embryonic stem cell test
to predict embryotoxicity in vitro. Nat Protoc 6:961–978
Sison-Young RL, Kia R, Heslop J, Kelly L, Rowe C, Cross MJ, Kit-
teringham NR, Hanley N, Park BK, Goldring CE (2012) Human
pluripotent stem cells for modeling toxicity. Adv Pharmacol
63:207–256
Liu JC, Lerou PH, Lahav G (2014) Stem cells: balancing resistance
and sensitivity to DNA damage. Trends Cell Biol 24:268–274
Luna-Zurita L, Stirnimann CU, Glatt S, Kaynak BL, Thomas S, Baudin
F, Samee MA, He D, Small EM, Mileikovsky M, Nagy A, Hol-
loway AK, Pollard KS, Muller CW, Bruneau BG (2016) Complex
interdependence regulates heterotypic transcription factor distri-
bution and coordinates cardiogenesis. Cell 164:999–1014
Magdy T, Schuldt AJT, Wu JC, Bernstein D, Burridge PW (2018)
Human induced pluripotent stem cell (hiPSC)-derived cells to
assess drug cardiotoxicity: opportunities and problems. Annu Rev
Pharmacol Toxicol 58:83–103
Suntharalingam G, Perry MR, Ward S, Brett SJ, Castello-Cortes A,
Brunner MD, Panoskaltsis N (2006) Cytokine storm in a phase 1
trial of the anti-CD28 monoclonal antibody TGN1412. N Engl J
Med 355:1018–1028
Takahashi K, Tanabe K, Ohnuki M, Narita M, Ichisaka T, Tomoda K,
Yamanaka S (2007) Induction of pluripotent stem cells from adult
human fibroblasts by defined factors. Cell 131:861–872
Talman V, Tuominen RK, Boije af Gennäs G, Yli-Kauhaluoma J,
Ekokoski E (2011) C1 Domain-targeted isophthalate derivatives
induce cell elongation and cell cycle arrest in HeLa cells. PLoS
One 6:e20053. https://doi.org/10.1371/journal.pone.0020053
Välimäki MJ, Tölli MA, Kinnunen SM, Aro J, Serpi R, Pohjolainen
L, Talman V, Poso A, Ruskoaho HJ (2017) Discovery of small
molecules targeting the synergy of cardiac transcription factors
GATA4 and NKX2-5. J Med Chem 60:7781–7798
Malek Mohammadi M, Kattih B, Grund A, Froese N, Korf-Klingebiel
M, Gigina A, Schrameck U, Rudat C, Liang Q, Kispert A, Wol-
lert KC, Bauersachs J, Heineke J (2017) The transcription fac-
tor GATA4 promotes myocardial regeneration in neonatal mice.
EMBO Mol Med 9:265–279
Mioulane M, Földes G, Ali NN, Schneider MD, Harding SE (2012)
Development of high content imaging methods for cell death
detection in human pluripotent stem cell-derived cardiomyocytes.
J Cardiovasc Transl Res 5:593–604
Paul SM, Mytelka DS, Dunwiddie CT, Persinger CC, Munos BH, Lind-
borg SR, Schacht AL (2010) How to improve R&D productivity:
the pharmaceutical industry’s grand challenge. Nat Rev Drug
Discov 9:203–214
Wysowski DK, Corken A, Gallo-Torres H, Talarico L, Rodriguez
EM (2001) Postmarketing reports of QT prolongation and ven-
tricular arrhythmia in association with cisapride and Food and
Drug Administration regulatory actions. Am J Gastroenterol
96:1698–1703
Pikkarainen S, Tokola H, Majalahti-Palviainen T, Kerkelä R, Hautala
N, Bhalla SS, Charron F, Nemer M, Vuolteenaho O, Ruskoaho
H (2003) GATA-4 is a nuclear mediator of mechanical stretch-
activated hypertrophic program. J Biol Chem 278:23807–23816
Pikkarainen S, Tokola H, Kerkelä R, Ruskoaho H (2004) GATA tran-
scription factors in the developing and adult heart. Cardiovasc
Res 63:196–207
Yu J, Vodyanik MA, Smuga-Otto K, Antosiewicz-Bourget J, Frane
JL, Tian S, Nie J, Jonsdottir GA, Ruotti V, Stewart R, Slukvin II,
Thomson JA (2007) Induced pluripotent stem cell lines derived
from human somatic cells. Science 318:1917–1920
Raihan MJ, Kavala V, Kuo C, Raju BR, Yao C (2010) “On-water”
synthesis of chromeno-isoxazoles mediated by [hydroxy(tosyloxy)
iodo]benzene (HTIB). Green Chem 12:1090–1096
1 3