Definition of transcriptome-based indices for quantitative characterization of chemically disturbed stem cell development: introduction of the STOP-Toxukn and STOP-Toxukk tests
Stem cell-based in vitro test systems can recapitulate specific phases of human development. In the UKK test system, human pluripotent stem cells (hPSCs) randomly differentiate into cells of the three germ layers and their derivatives. In the UKN1 test system, hPSCs differentiate into early neural precursor cells. During the normal differentiation period (14 days) of the UKK system, 570 genes [849 probe sets (PSs)] were regulated >fivefold; in the UKN1 system (6 days), 879 genes (1238 PSs) were regulated. We refer to these genes as ‘developmental genes’. In the present study, we used genome-wide expression data of 12 test substances in the UKK and UKN1 test systems to understand the basic principles of how chemicals interfere with the spontaneous transcriptional development in both test systems. The set of test compounds included six histone deacetylase inhibitors (HDACis), six mercury-containing compounds (‘mercurials’) and thalidomide. All compounds were tested at the maximum non-cytotoxic concentration, while valproic acid and thalidomide were additionally tested over a wide range of concentrations. In total, 242 genes (252 PSs) in the UKK test system and 793 genes (1092 PSs) in the UKN1 test system were deregulated by the 12 test compounds. We identified sets of ‘diagnostic genes’ appropriate for the identification of the influence of HDACis or mercurials. Test compounds that interfered with the expression of developmental genes usually antagonized their spontaneous development, meaning that up-regulated developmental genes were suppressed and developmental genes whose expression normally decreases were induced. The fraction of compromised developmental genes varied widely between the test compounds, and it reached up to 60 %. To quantitatively describe disturbed development on a genome-wide basis, we recommend a concept of two indices, ‘developmental potency’ (Dp) and ‘developmental index’ (Di), whereby Dp is the fraction of all developmental genes that are up- or down-regulated by a test compound, and Di is the ratio of overrepresentation of developmental genes among all genes deregulated by a test compound. The use of Di makes hazard identification more sensitive because some compounds compromise the expression of only a relatively small number of genes but have a high propensity to deregulate developmental genes specifically, resulting in a low Dp but a high Di. In conclusion, the concept based on the indices Dp and Di offers the possibility to quantitatively express the propensity of test compounds to interfere with normal development.
KeywordsHuman stem cells Transcriptome Genomics biomarkers Developmental toxicity In vitro test systems
Developmental toxicity testing represents a particularly challenging field of toxicology because the currently applied animal tests, such as second-generation reproduction or developmental neurotoxicity studies, are cost as well as labour intensive and require high numbers of animals (Adler et al. 2011; Forsby and Blaauboer 2007; Leist et al. 2008, 2013). In addition, developmental toxicity animal studies are not entirely relevant to the human situation. Therefore, the currently available testing capacities for studying the high number of developmental toxicants are not sufficient (Hengstler et al. 2006; Reif 2014a; Zimmer et al. 2014). A relatively high fraction of almost 20 % of newly approved drugs in Europe were reported to have post-approval issues, and between 2009 and 2011, five drugs had to be withdrawn from the market (Mol et al. 2013). Moreover, the chemical industry has been confronted with the European regulation on Registration, Evaluation, Authorisation and Restriction of Chemicals (REACH) initiative to provide more detailed toxicological data (Hengstler et al. 2006). Therefore, large efforts have been made to develop human stem cell-based in vitro test systems (Hengstler et al. 2006; Meganathan et al. 2015; Shinde et al. 2015; Weng et al. 2014). These in vitro systems recapitulate the critical phases of development, during which they are exposed to test compounds (Krug et al. 2013). Meanwhile, these systems have been applied in numerous studies to identify and characterize developmental toxicants (Balmer et al. 2014; Meganathan et al. 2012, 2015; Sisnaiske et al. 2014; Zimmer et al. 2014). Genome-wide expression studies have been performed to classify developmental toxicants (Rempel et al. 2015), and an eight-gene classifier has been shown to distinguish compounds acting as histone deacetylase inhibitors (HDACis) from a heterogeneous group of ‘mercurials’. Moreover, concentration-dependent, genome-wide expression studies in such stem cell-based systems have revealed concentration progression principles that allow the differentiation of (1) tolerated concentrations, where no gene expression changes are induced, (2) teratogenic concentrations ranges, where critical developmental genes are deregulated but no cytotoxicity occurs and (3) cytotoxic concentrations (Waldmann et al. 2014).
Materials and methods
Gelatin, putrescine, selenium, progesterone, apotransferrin, glucose and insulin were obtained from Sigma (Steinheim, Germany). Accutase was obtained from PAA (Pasching, Austria). FGF-2 (basic fibroblast growth factor), noggin and sonic hedgehog were obtained from R&D Systems (Minneapolis, MN, USA). Y-27632, SB-43154 and dorsomorphin dihydrochloride were obtained from Tocris Bioscience (Bristol, UK). MatrigelTM was obtained from BD Biosciences (Massachusetts, USA). All cell culture reagents were obtained from Gibco/Invitrogen (Darmstadt, Germany), unless otherwise specified. The following chemicals (HDACis and mercurials) were obtained from Sigma unless otherwise specified. The vehicles used are also mentioned with the compounds: (HDACis) valproic acid (VPA, P4543; water); trichostatin (TSA, T1952; DMSO); vorinostat (SAHA, SML 0061; DMSO); belinostat (PXD101, S1085, Selleckchem; DMSO); panobinostat (LBH589, S1030, Selleckchem; DMSO); entinostat (MS-275, Cay-13284-25; Biomol; DMSO); (mercurials) methylmercury (MeHg, 442,534; 10 % ethanol); thimerosal (THM, T4687; water); mercury(II)chloride (HgCl2, 203,777; water); mercury(II)bromide (HgBr2, 437859, water); 4-chloromercuribenzoic acid (PCMB, C5913-5G; water); and phenylmercuric acetate (PMA, P27127-25G).
Human embryonic stem cells
The human pluripotent stem cell line H9 (Thomson et al. 1998) was obtained as a WA09 line from WiCell (Madison, WI, USA). The importation of the cells and the subsequent experiments were authorized (Robert Koch Institute, Berlin, Germany) under licence # 170-79-1-4-27 for the UKN1 system and licence # 1710-79-1-4-34 for the UKK system. H9 cells were cultured on irradiated mouse embryonic fibroblasts (MEFs) in H9 culture medium, as previously described (Jagtap et al. 2011; Krug et al. 2013; Shinde et al. 2015).
Random differentiation of hESCs to germ layer cell types and their derivatives (UKK test system)
To remove MEFs, H9 cells from the maintenance culture were transferred on hESC-qualified matrix (BD Biosciences, California, USA)-coated, 60-mm tissue culture plates (Nunc, Langenselbold, Germany) in TESR1 medium (StemCell Technologies) and were maintained for 5 days prior to differentiation. The random differentiation into embryoid bodies (EBs) representing multiple lineages was performed as described previously (Meganathan et al. 2012). In brief, cell clumps were obtained by cutting and scraping the cells with a passage tool (StemPro EZPassage™ Disposable, Invitrogen) and a cell scraper. On day 0, 80 clumps were seeded in each well of a pluronic-coated, v-bottom plate in 100 µl of random differentiation (RD) medium (DMEM-F12 medium with 20 % KO serum replacement, 1 % non-essential amino acids, penicillin (100 units/ml), streptomycin (100 µg/ml) and 0.1 mM β-mercaptoethanol) containing chemical or vehicle, and the plate was then incubated (37 °C, 5 % CO2) for 4 days. The EBs were collected on day 4 and were transferred onto a 100-mm bacteriological plate in 15 ml of RD medium containing the chemical or vehicle. The medium was replenished every alternate day until day 14 of differentiation.
Neuroepithelial differentiation (UKN1 test system)
H9 cells were cultured according to standard protocols and were differentiated into neuroepithelial progenitors (NEPs) as previously described (Balmer et al. 2014; Krug et al. 2013; Shinde et al. 2015) and as shown in Fig. 1. The differentiation of the H9 cells towards NEPs was based on a dual SMAD inhibition (Chambers et al. 2009) using a combination of 35 µM noggin and 600 nM dorsomorphin together with 10 µM SB-431642. This was used to prevent BMP and TGF signalling and thus achieve a highly selective neuroectodermal lineage commitment. Cells were handled and manipulated as previously described in the supplemental methods (Balmer et al. 2014).
Benchmark concentration (BMC) determination
The BMC is defined here as the concentration at which at least 90 % of the cells survive (BMC10). For both test systems, the BMC10 was determined for six HDACis and six mercurial compounds by performing resazurin cell viability assays exactly as previously described (Krug et al. 2013; Stiegler et al. 2011). In the UKK system, the H9 cells were differentiated as described above and were treated from DoD 4 to DoD 14. In the UKN1 test system, the cells were treated from DoD 0 to DoD 6. Both test systems were treated during the indicated time periods with concentrations ranging from non-cytotoxic to cytotoxic. The analysis was performed on DoD 14 (UKK) and DoD 6 (UKN1). The experiments have been performed with five technical and three biological replicates. On the day of analysis, cells were incubated with 10 µg/ml resazurin for 1–1.5 h at 37 °C and 5 % CO2. To determine the background fluorescence of resazurin itself, a control with only resazurin in medium was included. Resorufin was measured at a wavelength of 560Ex/590Em with a fluorescence reader. The mean background fluorescence of resazurin was subtracted from all experimental data. Further data processing to identify BMC of chemicals was done as mentioned below. Data from each experiment were normalized to their respective vehicle controls (set as 100 %). The data were then displayed in semi-logarithmic plots. Data points were connected by a nonlinear regression sigmoidal dose–response curve fit. These curves were averaged, and BMC10 was then determined graphically as the data point on the average curve corresponding to 90 % viability value, or as the last real data point left of this value. The BMC10 concentrations of the compounds were considered for further Affymetrix-based studies with UKK and UKN1 protocol.
Affymetrix-based study details
For Affymetrix-based study, six HDACis and six mercurial compounds were exposed at BMC10 concentrations from DoD 0 to DoD14 or from DoD 0 to DoD6 in UKK or UKN1 test system, respectively, along with the respective vehicle controls. The samples were collected on DoD 14 or DoD 6 for RNA isolation. The samples from four biological replicates were used for further processing.
Affymetrix DNA microarray analysis
On DoD 14 (UKK) or DoD 6 (UKN1), medium was removed, and the cells were lysed in RNA protect solution (Qiagen). Affymetrix chip-based DNA microarray analysis (Human Genome U133 plus 2.0 arrays) was performed exactly as previously described (Krug et al. 2013; Meganathan et al. 2015). Briefly, total RNA was isolated using TRIzol and chloroform (Sigma, Steinheim, Germany) and purified with miRNeasy mini kit (Qiagen, Hilden, Germany). The quantification and quality control measurements were done using Nanodrop spectrophotometer (ND-1000, Thermo Fisher, Langenselbold, Germany). For microarray labelling 100 ng total RNA was taken as a starting material, and after amplification 12.5 µg amplified RNA was hybridized on Affymetrix Human Genome U133 Plus 2.0 arrays (Affymetrix, Santa Clara, CA, USA). For washing and staining Affymetrix HWS kit and Genechip Fluidics Station-450 were used according to the manufacturer’s instructions. After staining arrays were scanned with Affymetrix Gene-Chip Scanner-3000-7G and Affymetrix GCOS software has been used for quality control analysis. The corresponding raw CEL files of the Affymetrix chips are publicly available under GEO accession number GSE71127.
Batch effects, non-biological experimental variation, is a commonly observed phenomenon in the area of microarray studies. They occur due to experiments that cannot be conducted all at once, for various reasons. It was important to eliminate batch effects as otherwise relevant expression changes may remain undetected. We applied the ComBat algorithm (Johnson et al. 2007) that allowed adjusting for batch effects in datasets where the batch covariate was known. It used a nonparametric, empirical Bayes approach for the estimation of an additive and a multiplicative batch effect. First, we used extrapolation strategy (RMA +) (Harbron et al. 2007) for background correction, log2 transformation and quantile normalization. Then the data were standardized with respect to mean gene expression and treatment effect. Then, the batch effects were estimated and eliminated from the standardized data by subtracting the additive effect and dividing by the multiplicative effect. Finally, the data were back-transformed, i.e. mean gene expression and treatment effect were added. Further statistical data analysis and visualization performed by uploading ComBat-corrected files into the Partek Genomics Suite (PGS) version 6.6 software (Partek, St. Louis, MO, USA). The normalized probe sets (PSs) were used for the generation of a principal component analysis (PCA) and one-way ANOVA model that was used to identify the differentially regulated transcripts with changes of at least 1.5-fold (absolute fold change, p value ≤0.05), with Benjamini and Hochberg FDR corrections. The first 50 transcripts deregulated by each toxicant were filtered based on p value, and signals were normalized by z-score and clustered using a hierarchal cluster analysis (complete linkage method). The commonly deregulated transcripts were obtained using a Venn diagram overlap analysis (PGS). Online free software such as g:Profiler and the Database for Annotation, Visualisation and Integrated Discovery (DAVID) were used for functional annotation and gene ontology (GO) clustering of differentially expressed transcripts (p ≤ 0.05).
Construction of a transcription factor network
The TF network was constructed as previously described (Rempel et al. 2015). Briefly, we downloaded raw data for the microarray samples referenced in the manually curated CellNet tissue atlas (Cahan et al. 2014) and combined them with data from the UKN1 and UKK test systems (Balmer et al. 2014; Cahan et al. 2014; Krug et al. 2013; Waldmann et al. 2014). To obtain the expression matrix, the samples were normalized together using RMA implemented in the R package oligo. The co-expression network was constructed in two steps using functions from the parmigene package for R. First, the mutual information matrix was computed by applying the function knnmi.all with parameter k = 9, chosen based on an unpublished benchmarking, on the expression matrix. Then, we applied the clr function from the parmigene package, which implements the CLR algorithm. The co-expression network was subsequently restricted to genes annotated as transcription factors (TFs) in the Animal Transcription Factor Database (AnimalTFDB, [http://www.bioguo.org/AnimalTFDB/index.php]). The overlap of the genes detected by the Affymetrix array and the AnimalTFDB was 1300 genes. Links were drawn only for pairs of TFs with a score in the top 0.1 % of all co-expression scores. This yielded 1690 predicted interactions between 847 TFs. Nodes were arranged in the network according to the Fruchterman and Reingold’s force-directed placement algorithm provided by the R package sna with the area parameter = 109.
Representation of mercurials or HDACis consensus genes on the TF network
Communities of network nodes were determined by the fast greedy community function of the R package igraph. Only the top 18 largest communities were analysed for the enrichment of GO biological process annotations, as the others contained less than six genes. The enrichment analysis was performed with the R package topGO using the classic method and the Fisher’s test statistic. We selected representative terms for each community from the top enriched terms with an unadjusted p value <0.05. For spontaneous differentiation and regulation by compounds, TFs in the network were marked red (blue) if a probe set mapping to this TF was up-regulated (down-regulated) under the respective condition. The mapping of PSs to the Ensembl gene ids and gene symbols was determined using the BioConductor package hgu133plus2.db. Only PSs that could be mapped to a gene symbol were taken into account. TFs for which PSs mapping to them were inconsistently regulated were removed from the analysis.
Glutathione reductase (GSR) and isocitric dehydrogenase (ICDH) activity assays
ICDH (porcine, Sigma, I-2002) (10 µg/200 µl) in a Tris(hydroxymethyl)-aminomethane (Tris)-buffer (20 mM) containing MnSO4 (2 mM), pH 7.4, was incubated with the compounds to be tested at 37 °C for 20 min. ICDH activity was determined by the addition of isocitrate (4 mM) and NADP+ (0.1 mM). The enzymatic reduction of NADP+ to NADPH was monitored using photospectroscopy at 340 nm over the course of 15 min at 1-min intervals and 37 °C. The enzymatic activity was determined from the slope of the absorbance increase over time. All data were normalized to the activity of untreated enzyme (i.e. free of toxicant). GSR (human, Sigma G-9297) (10 µg/200 µl) was incubated in sodium phosphate buffer (100 mM), pH 7.5, containing ethylenediaminetetraacetic acid (EDTA; 1 mM) and the compounds to be tested for 20 min at 37 °C. To assess GSR activity, oxidized glutathione (GSSG) (5 µM), NADPH (0.4 mM) and 5,5′-dithiobis(2-nitrobenzoic acid) (DTNB) (all from Sigma) were added, and the reaction was monitored by absorbance measurements at 405 nm (37 °C) at 1-min intervals over the course of 15 min. The enzymatic activity was determined from the slope of the absorbance increase over time. All data were normalized to the activity of untreated enzyme (i.e. free of toxicant).
Identification of consensus genes
A gene was defined as significantly deregulated by a specific compound if at least one annotated probe set was significantly deregulated (absolute fold change >1.5 and FDR-corrected p value <0.05). A gene was defined as a ‘consensus’ gene if it was significantly up- or down-regulated by as many compounds of same class as possible (i.e. mercurial or HDACi).
Identification of diagnostic genes
A ranking approach was performed to identify PSs that fulfilled the following criteria: (1) deregulation occurred from as many compounds of the same class as possible (i.e. HDACi or mercurial); (2) PSs with higher fold changes compared with those of the controls were preferentially considered; (3) only the developmental genes were considered; (4) PSs were only considered when the test compounds antagonized the spontaneous development, i.e. when up-regulated developmental genes were suppressed or down-regulated developmental genes were induced; (5) only PSs with baseline expression values >6 (log2 scale) at day 0 or at the day of differentiation (day 14 in the UKK system or day 6 in the UKN1 system) were considered (the number of PSs passing this criteria are shown in Suppl. Fig. S5A &B, the cut-off value has been selected based on the frequency distribution curves provided in Suppl. Fig. S5C & D); and (6) PSs were only considered when they could be assigned to genes whose function is basically understood.
Structure of developmental genes in differentiating stem cells
In the present study, two stem cell-based test systems were used, the UKK system, where hESCs/hPSCs differentiate into cells of the three germ layers and their derivatives, and the UKN1 system, which recapitulates differentiation into early neural precursor cells (Fig. 1a). Genome-wide gene expression profiles were established from the undifferentiated hESCs and after 14 (UKK) and 6 days (UKN1) of differentiation. Additionally, hESCs of both in vitro systems were exposed to 12 test compounds at benchmark concentrations of low cytotoxicity during the differentiation periods. Among them, six mercurials and six HDACis were applied (Fig. 1b, c). All compounds were tested at their maximum non-cytotoxic concentrations, i.e. at the benchmark concentration (BMC10) resulting in a maximal viability reduction of 10 %. This biological-anchoring point has been previously found to be well suited for transcriptome analysis (Waldmann et al. 2014). Moreover, it is also well correlated with known pharmacological and toxicological molecular properties of the test compounds. For instance, the concentrations of the HDACis were all within the range known to be required for half-maximal enzyme inhibition in biochemical assays (Rempel et al. 2015). The concentrations of the mercurials were all in a similar range with respect to the inhibition of thiol-containing, redox-sensitive enzymes. Glutathione reductase (GSR) or isocitrate dehydrogenase (ICDH) were chosen as target enzymes for this rough bioequivalence test, and the BMC10 test concentrations all led to partial enzyme inhibition (Fig. 1d). Notably, none of the HDACis affected the enzyme activities at the concentrations tested here. To obtain an overview over the genome-wide data, principle component analysis (PCA) plots were established. The PCA illustrates a relatively large distance between hESCs and the cells resulting after the 14-day (UKK) and 6-day (UKN1) differentiation processes (Fig. 1c). Compared with these large distances, the influence of the test compounds appears comparatively small (Fig. 1c). However, when the PSs for the PCA analysis were limited to the 50 most regulated (defined by the lowest FDR-corrected p values) of each test substance, most compounds cluster distinctly from the controls, whereby the differentiation between compound-exposed samples and their controls appears more separated in data from the UKN1 system than in data from the UKK system (Fig. 1e, f).
Top 10 gene ontology categories overrepresented amongst up- and down-regulated probe sets during differentiation
Nervous system development
Anatomical structure dev.
Generation of neurons
Central nervous system development
Muscle tissue development
Striated muscle tissue dev.
Connective tissue development
Regulation of nervous system dev.
Neuron projection development
Nervous system development
Single-organism dev. process
Generation of neurons
Central nervous system development
Regulation of dev. process
Anatomical structure dev.
Multicellular organismal dev.
Anatomical structure morphogenesis
Regulation of nervous system dev.
Negative reg. of nervous system dev.
Negative regulation of cell dev.
Negative reg. of neurogenesis
Regulation of organismal dev.
Anatomical structure development
Single-organism dev. process
Multicellular organismal development
Anatomical structure morphogenesis
Anatomical structure formation
Circulatory system development
Neuron projection dev.
Anatomical structure morphogenesis
Neuron projection morphogenesis
Nervous system development
Anatomical structure development
Generation of neurons
Multicellular organismal development
Single-organism dev. process
Circulatory system development
Cell morphogenesis in neuron diff.
Cardiovascular system development
Interference of chemicals with stem cell differentiation
The consensus PSs for HDACis in the UKN1 test system have been characterized previously (Rempel et al. 2015), and the mercurial consensus PSs (deregulated by at least of 3 mercurials) in the UKN1 test system are shown in Fig. 5a. For the UKK test system, no mercury consensus PSs were identified, but there was a clear group of HDACi consensus genes (deregulated by at least of 4 HDACi), and this is shown in Fig. 5b. To study whether consensus genes were derived from specific biological processes triggered by toxicants, overrepresented GO terms were identified amongst the UKN1 mercurial consensus PSs. This analysis indicated the up-regulation of two antioxidant stress response pathways (glutamate–cysteine ligase activity, required for glutathione synthesis; ascorbate metabolism), and two toxicant response pathways. Moreover, there was an indication of the regulation of the anti-apoptotic response in neurons (Fig. 5c).
To further understand the coordinated regulations leading to changes of the consensus genes on a biological system level, disturbances of the transcription factor (TF) network were visualized. For this purpose, we used a previously generated generic TF coexpression network for human cells (Rempel et al. 2015). Within the network, we previously identified groups of three highly connected TFs (communities) that are enriched for GO terms relevant for neural development (Rempel et al. 2015). On this network, we then mapped and compared the TFs changing in the respective test systems (‘developmental TFs’). Some of them were specific for either the UKK or UKN1 system (Suppl Fig. S7). However, there were also TFs regulated in both systems (overlap TFs). A high fraction of these TFs are concentrated in three communities related to neuronal development (56 %). These three communities consist of two communities enriched for neuronal development GO terms and a third community that links the two that contains stem cell factors and cell cycle regulators (Fig. 5d). Subsequently, we identified TFs amongst the toxicant consensus genes and their positioning in the network was visualized (Suppl Fig. S7). For the UKN1 test system, different consensus TFs were visualized. For HDACis, we find that about one-third (32.7 %) of TFs affected by at least three out of six compounds in the UKN1 system mapped to the three communities related to neuronal development (Fig. 5e). For the mercurial consensus (at least 3 out of 5 compounds), we find overall less affected TFs, and also a lower fraction of neuronal development TFs affected (26.5 %) (Fig. 5f). Taken together, the HDACi consensus affected TFs are more strongly concentrated in neuronal development than the mercurial consensus affected TFs. In addition, mercurial consensus TFs mapped relatively broadly across the TF network (Fig. 5f). Most of the TFs affected by mercurials were also affected by HDACis, independent of the location in the network. Such TFs are promising candidates for master regulators that are generally affected by various toxicants in the test systems (Fig. 5f). They may be used in the future as biomarkers, and they may also lead to further mechanistic studies with increased sensitivity for examining mechanisms of toxicity.
In conclusion, mercurials or HDACis usually antagonize the spontaneous development of gene expression in differentiating hESCs. This characteristic feature was also shown by an alternative approach, in which high and low gene expression was differentiated using a log2 value of 6 as a threshold (Suppl. Fig. S5; Suppl. Table 6). An overlap of chemically up-regulated genes mostly occurs for genes that are high in the undifferentiated hESCs and decrease during the differentiation period (Suppl. Fig. S5A, B), while chemically induced down-regulated genes overlap with developmental genes that are lowly expressed in stem cells and have increased expression towards the end of the in vitro differentiation period. Reciprocal scenarios were only rarely observed (Suppl. Fig. S5).
Developmental potency (Dp) and developmental index (Di) for the characterization of compromised stem cell development
The concept of these developmental indices can be further refined by differentiating biological categories of developmental genes. The example of VPA in the UKN1 system illustrates that for both Dp and Di, neural crest cell differentiation was compromised to a higher degree than were other categories, such as neurogenesis in general, sensory organ development, axonogenesis and telencephalon development (Fig. 7k, l; Suppl. Table 7 and 9). With increasing concentrations, the predominance of neural crest differentiation decreased, probably because genes of other biological categories also became increasingly deregulated. In contrast with the UKN1 system, the UKK test system shows increased Dp and Di for a much broader range of categories, including limb, heart, skin and liver development after incubation with thalidomide (thalidomide and all further exposure conditions: Suppl. Fig. S6; Suppl. Table 8 and 10).
Identification of diagnostic genes
Identification of gene–disease association
Biological or disease relevance of mercurial consensus genes
Known literature data
Relationship with known toxicity
(Mitochondrial poly(A) polymerase)
Mutated in human autosomal-recessive neurodegeneration, spastic ataxia
Crosby et al. (2010)
In utero exposure of MeHg in humans:
resulted in ataxia, spastic paraparesis, learning difficulties, speech and walking delay; histopathology—ectopic cell masses in the cortex, disorganized layers in the brain; Long-lasting effects on vision in humans, rat off springs (Burbacher et al. 1990)
(SRY (sex determining region Y)-box 6)
In mouse required for normal positioning and maturation of cortical interneuron subtypes
Batista-Brito et al. (2009)
It promotes neurites outgrowth in culture, affects migration of neuronal precursor cells in vivo in developing brain.
(Cytochrome P450, family 1, subfamily B, polypeptide 1)
Mutation in congenital glaucoma in human; role in murine embryonic axis development, neural tube patterning in chick embryo, overexpression potentiates retinal ganglion cell survival
(NAD(P)H dehydrogenase, Quinone 1)
Up-regulated/polymorphism in Alzheimer’s disease in humans. Induced by MeHg in primary rat microglial cells
(Secreted phosphoprotein 1/osteopontin)
Up-regulated in autistic children, involved in autoimmune neuroinflammatory disease & multiple sclerosis
(Gamma-aminobutyric acid (GABA) a receptor, beta 3)
Mutated in Parkinsons and Alzheimers in humans
(Ubiquitin carboxyl-terminal esterase L1)
mutated in neurodevelopmental disorder austism spectrum conditions in humans
Warrier et al. (2013)
(Neuropeptide Y receptor Y2)
It has neuroprotective effect in vitro and in vivo animal models of Parkinson’s disease
Decressac et al. (2012)
(Ferritin, light polypeptide)
Mutated in neurodegenerative disease—hereditary ferrinopathy (tremors, cerebellar signs, cognitive defects etc.)
It protects neurons from oxidative stress during rat neuronal development; role in brain development, expressed in hippocampus, cortex, cerebellum and olfactory bulb
Glutathione enzyme and selenoproteins are evaluated as mercury biomarkers in humans. Goodrich et al. (2013)
(Glutamate–cysteine ligase modifier subunit)
MeHg binds to thiol group regulatory protein (Keap 1) of Nrf and its activation induces up-regulation of GCLM in human neuroblastoma cells
Toyama et al. (2007)
(Solute carrier family 7 (amino acid transporter light chain, L system))
linked with autism, Encodes transporter LAT1 which increases transport of l-Cysteine conjugated MeHg in cells
(Msh homeobox 2)
Gain in function mutation-craniosynostosis (premature fusion of calvarial bones of skull in human)
(Collagen, type I, alpha 2)
Mutated in osteogenesis imperfecta type I–IV in humans
Biological or disease relevance of HDACis consensus genes
Known literature data
Relationship with known toxicity
(Asparagine synthetase (glutamine-hydrolysing))
Mutations found in congenital microcephaly human
Ruzzo et al. (2013)
VPA-induced congenital microcephaly in humans with craniofacial abnormalities: short nose, philtrum of lip, micrognathia, urogenital anomalies
Ardinger et al. (1988)
Facial dysmorphism, craniosynostosis, neural tube defects, Chandane and Shah (2014)
Atrial septum defects, ventricular septum defects, tetralogy of fallot observed in foetal valproate syndrome. Jentink et al. (2010)
Genitourinary tract defects, autism-related cerebellar anomalies in humans and rats. Ingram et al. (2000)
VPA-induced neurite outgrowth linked with autism spectrum disorder (Chomiak et al. (2013))
Multiple ocular associations—strabismus, myopia, nystagmus, epicanthic folds, infraorbital creases and dry eye and nasolacrimal duct obstruction etc. (Hornby and Welham 2003)
(Collagen, type XI, alpha 1)
Mutation found in Stickler syndrome patients
(Glucan (1,4-alpha-), branching enzyme 1)
Mutation in human patients results in adult polyglucosan body disease and glycogen storage disease IV in equines
(Endothelin receptor type B)
Mutation associated with Hirschsprung disease type II and ABCD syndrome, restricted period expression in neural crest development in mouse
(Myocardial infarction associated transcript)
SNPs found in myocardial infarction and schizophrenia in humans
(Protease, serine, 23)
knock down result in atrioventricular valve defect in zebrafish
Chen et al. (2013)
RTN1 helps in vesicular transport of Spastin and disturbance of this process probable cause of Hereditary Spastic Paraplegias
Mannan et al. (2006)
(Forkhead box D3)
variants found in Aniridia, Peters anomaly, anophthalmia in humans
Kloss et al. (2012)
(GABA-A receptor, beta 2)
Mutation results in congenital intellectual disability in human patients
Srivastava et al. (2014)
(Aryl-hydrocarbon receptor nuclear translocator 2)
Loss-of-function mutation in humans results in Webb–Dattani syndrome
Webb et al. (2013)
(Delta/Notch-like EGF repeat containing)
Overexpression inhibits proliferation of neural progenitors in zebrafish; Mutation in mouse impairs cerebellar functions
(Caspase 8, apoptosis-related cysteine peptidase)
Conditional knockout attenuates neural tube defects in mouse. Involved in Alzheimer’s, Parkinson’s and Huntington’s Disease
Recently, hESC-based in vitro systems that recapitulate specific phases of human development have become available (Balmer et al. 2014; Krug et al. 2013; Zimmer et al. 2014). In UKN1 and UKK test systems genome-wide expression data of 12 compounds (six HDACi and six mercurial) was obtained for benchmark concentrations (BMC10) resulting in viability reduced by a maximum of 10 %. The test compound VPA was additionally tested over a wide range of concentrations, from non-toxic up to severely cytotoxic concentrations. Additionally, the teratogen thalidomide was studied over a wide concentration range using the UKK test system. Based on this genome-wide data set, we studied the basic principles of how chemicals interfere with gene expression in differentiating stem cells. One key feature is that chemicals antagonize the spontaneous developments of gene expression in differentiating stem cells. Genes up-regulated during differentiation were suppressed, while down-regulated genes were induced when exposed to the test compounds. A second key feature is that compounds differ widely in the fraction of developmental genes whose expression they compromise. For example, VPA influences a higher fraction of developmental genes in both test systems than does HgCl2, although both compounds were tested at benchmark concentrations (BMC10). A third feature is the difference in susceptibility between both test systems. For example, a higher fraction of developmental genes is compromised by MeHg and VPA in the UKN1 system than in the UKK test system. Because the UKN1 system recapitulates the formation of neuronal precursor cells and the UKK system recapitulates the development of the three germ layers and their derivatives, it can be expected that the latter process is less susceptible to the analysed compounds.
Considering the above-mentioned features, we recommend two indices to quantify the developmental toxicity potential of different compounds, the developmental potency index (Dp) and the developmental index (Di). Dp gives the intersection of the genes between the deregulated genes (up- or down-regulated) by a test compound and the genes deregulated at day 6 (UKN1) or day 14 (UKK) of differentiated versus undifferentiated hESCs in the absence of a test compound. A high Dp can be interpreted as a high hazard of toxicity. It should be considered that the total number of deregulated genes is associated with Dp. If a compound deregulates only a small number of genes, e.g. n = 20, Dp will inevitably be small. However, Dp per se might not discriminate developmental toxicity effects occurring on the expression of genes required for the differentiation of hESCs to neural progenitor cells (UKN1) or for the formation of germ layer cells and their derivatives (UKK) from general toxicity effects not related to differentiation processes (e.g. toxicity effects occur also in differentiated cells). To discriminate developmental toxicity effects, we recommend additionally considering the developmental index Di. This index provides information about the ratio by which developmental genes are overrepresented among ‘toxicity genes’, which represents the number of genes deregulated by the test compound. If Di is high, developmental genes are overrepresented among ‘toxicity genes’. A low Di shows that developmental genes are underrepresented because the compound preferentially influences biological processes not related to development. The formula of Di has been designed to consider the number of developmental genes. Therefore, test systems with different numbers of developmental genes can be directly compared. The relevance of Di can be illustrated by the example of the teratogenic compound thalidomide (Fig. 7c, d). The very low concentration of 0.1 µM thalidomide up-regulates only 12 genes, resulting in a very low Dp, which does not indicate a severe hazard effect. However, Di for this test condition is approximately 16, meaning that thalidomide specifically compromises the expression of developmental genes with a ratio 16-fold higher than can be randomly expected. Therefore, 0.1 µM thalidomide may be considered as developmentally hazardous, even though the expression of only a few genes is compromised. An opposing trend to thalidomide can be illustrated by thimerosal. In the UKK system, thimerosal deregulated 1540 PSs, which similar to 0.1 µM thalidomide, resulted in a low Dp. However, a Di of 1.16 for thimerosal indicates that developmental genes are not overrepresented among the PSs deregulated by this compound (Fig. 7i, j). In conclusion, using these systems with Dp and Di gives a rapid overview of developmental toxicity. If one of the two indices is high, the test condition should be interpreted as hazardous. Moreover, hazardous compounds can be classified as developmentally and less developmentally hazardous compounds. In conclusion, based on this approach with hESCs and transcriptomic technologies, we have established two very sensitive test systems allowing the discrimination of developmental (teratogenic) toxicity from non-developmental, general toxicity hazards: the ‘STOP-Toxukn’ and the ‘STOP-Toxukk’ tests (Stem cell-based Teratogenic Omics Prediction; UKN: University of Konstanz; UKK: Universitätsklinikum Köln).
Comparing the results of the two compound classes, it should be considered that the selected HDACis and mercurials showed different degrees of homogeneity. HDACis that have been discussed previously (Rempel et al. 2015; Yang et al. 2013) represent a relatively homogenous compound group. This is illustrated by a consensus of 90 up- and 18 down-regulated genes that are influenced by all six HDACis in the UKN1 system. In contrast, mercurials showed a lower degree of overlap of deregulated genes. No single gene that was influenced by all six mercurials could be identified. Nevertheless, there is a set of consensus genes influenced by five of six mercurials, which are appropriate to use for the identification of developmental toxicity in the two test systems.
A further goal of this study was to identify individual genes that are particularly suitable to use for the prediction of developmental toxicity using the test systems STOP-Toxukk and STOP-Toxukn, further referred to as ‘diagnostic genes’ for severe developmental defects in humans. Prenatal exposure to teratogens results in congenital disorders in humans, which are known to be related with mutations in various genes (Webber et al. 2015). The disease association of the consensus genes de-regulated by mercurials or HDACis further helps to understand mechanistic aspects. The selection criteria were that diagnostic genes should be influenced by as many compounds of the same class (i.e. HDACis and mercurials) as possible; they should show the highest fold changes compared with the controls; they should antagonize the expression of developmental genes; and their biological relevance should be clear, based on animal knockout or overexpression models or on human disease data. Based on these criteria, several top diagnostic genes were identified.
The significance of this association can be related to the fact that mercurials can perturb neurodegenerative genes by epigenetic events, resulting in altered gene expression and late onset of neurodegenerative diseases. Mercurial consensus genes that are related to genetic disease associations are the following: MTPAP, SEPW1 (UKK); NQO1 (UKK/UKN1); SPP1, CYP1B1, GABRB3, UCHL1, SLC7A5, COL1A2 and FTL (UKN1). The only up-regulated gene found in the UKK test system was MTPAP. Interestingly, a mutated MTPAP has been reported in human autosomal-recessive spastic ataxia (Crosby et al. 2010). Moreover, the polymorphisms or overexpression of the detoxification enzyme NQO1 has been reported in Alzheimer’s patients (Raina et al. 1999) and was captured by mercurials in both systems. The up-regulation of SPP1 has been observed in autistic children (Al-ayadhi and Mostafa 2011). Mutations of GABRB3 have been identified and a role of the SLC7A5 gene has also been described in autistic patients (Anderson et al. 2009; Warrier et al. 2013). Mutations in UCHL1 have been identified in Parkinson’s patients (Liu et al. 2015), whereas a gain-of-function mutation in the corresponding mouse gene resulted in Parkinson’s disease symptoms (Yasuda et al. 2009). A mutation in FTL resulted in neurodegenerative hereditary ferrinopathy, characterized by tremor and cognitive defects (Kubota et al. 2009; Vidal et al. 2004). The NPY2R gene has been shown to serve in a neuroprotective role in animal models of Parkinson’s disease (Decressac et al. 2012). Notably, NPY2R was found to be down-regulated by mercurials.
Autopsies of children exposed to MeHg in utero showed degeneration and atrophy of cortical structures, ectopic cell masses in the cortex, astrocytes present in white matter and cell loss prominent in the cerebellum and cerebrum (Burbacher et al. 1990). The biological functions of two mercurial consensus genes SOX6 and THBS1 are in agreement with these pathologies. SOX6 was found to be down-regulated in our study, and a SOX6 mouse knockout showed that expression is required for normal positioning and maturation of cortical interneuron subtypes (Batista-Brito et al. 2009). THBS1 was up-regulated by the mercurials and has been shown to promote neurite outgrowth as well as post-natal migration of neuronal precursor cells in mice (Liu et al. 2015). Additional diagnostic mercurial consensus genes are SEPW1 and GCLM. SEPW1 has been reported to be a neuroprotectant, and it is targeted by MeHg in human neuronal cells (Kim et al. 2005). It has also been reported that MeHg up-regulates GCLM in human neuroblastoma cells (Toyama et al. 2007). Recently, a good correlation between polymorphisms of SEPW1 and GCLM and the potential of eliminating mercurials in urine and hair has been reported in humans (Goodrich et al. 2011, 2013).
In animal models, inhibition of HDAC resulted in altered gene expression during embryonic development that was accompanied by congenital malformations (Menegola et al. 2005). The data from the literature suggest that VPA-induced malformations in humans are congenital microcephaly, atrial and ventricular septal defects, craniofacial abnormalities, neural tube defect spina bifida, urogenital abnormalities and ear abnormalities (Ardinger et al. 1988). HDACis regulated consensus genes that are linked to these malformations include ASNS, COL11A1, GABRB2, and GBE1 in the UKK system and RTN1, SMOC1, RAX, ARNT2, CASP8, MIAT, BMP5 and EDNRB in the UKN1 test system. ASNS was found down-regulated by the HDACis. In this context, it has been reported that recessive loss-of-function mutations of ASNS in patients caused congenital microcephaly, intellectual disability and progressive cerebral atrophy. It has also been demonstrated that asparagine depletion due to ASNS deficiency results in neurological impairment in knockout mice (Ruzzo et al. 2013). Mutations in the COL11A1 gene have been found in patients with Marshall syndrome and Stickler syndrome, which share similar phenotypes, such as craniofacial abnormalities, a flat nasal bridge, midface hypoplasia, a short nose, palate defects and hearing loss (Annunen et al. 1999). Mutations in GBE1 have been detected in patients of adult polyglucosan body disease with symptoms such as progressive gait difficulty, urinary incontinence, and reduced reflexes in lower limbs (Sampaolo et al. 2015). In the present study, COL11A1 and GBE1 were found to be down-regulated by the incubation of the hESCs with the HDACis. Mutations in EDNRB have been identified in patients suffering from Hirschsprung’s disease, which is characterized by the absence of neural crest-derived intramural ganglia along the colon (McCallion and Chakravarti 2001). MIAT was found to be down-regulated in the present study. Although this gene is not translated to protein, it transcribes non-coding functional RNA. Recently, SNPs of MIAT have been identified in human patients with myocardial infarction and schizophrenia (Ishii et al. 2006; Rao et al. 2015).
PRSS23, which was found to be down-regulated in our study, encodes for a novel serine protease that is expressed during the development of the cardiovascular system, and it is involved in valvulogenesis of zebrafish. PRSS23 plays a significant role in the endothelial to mesenchymal transition during valvulogenesis, and the knockdown of the gene results in atrioventricular defects (Chen et al. 2013). FOXD3 was found to be up-regulated, and it is linked with congenital eye deformities such as aniridia, Peter’s anomaly and anophthalmia in humans (Kloss et al. 2012). RTN1 plays a role in the vesicular transport of Spastin, which is involved in microtubule dynamics. The disruption of Spastin-mediated vesicular transport has been discussed as a cause of hereditary spastic paraplegias, with symptoms such as the progressive stiffness of lower limbs due to nerve dysfunction, cognitive impairments and deafness (Mannan et al. 2006). Other diagnostic consensus genes with disease relevance are GABRB2, ARNT2, DNER and CASP8. The overexpression of DNER inhibits the proliferation of neural progenitor cells and induces glial cell differentiation in zebrafish (Hsieh et al. 2013). A loss-of-function mutation of DNER in mice impairs cerebellar functions (Tohgo et al. 2006). ARNT2 was found to be down-regulated in the present study. A loss-of-function mutation of ARNT2 in humans results in Webb–Dattani syndrome, with characteristic features such as microcephaly, visual impairment and urinary abnormalities (Webb et al. 2013). Caspase 8 was up-regulated by the HDACis and has been described to regulate neuronal apoptosis involved in neurodegenerative diseases such as Alzheimer’s, Parkinson’s and Huntington’s disease in humans (Ahmad et al. 2014). In conclusion, the involvement of the diagnostic genes deregulated by mercurials and HDACis by both test systems in developmental toxicity in humans and experimental animals has been well established.
A limitation of the present study is the relatively small number of test compounds. A next important step will be to include compounds with very low cytotoxicity (‘negative controls’) as well as cytotoxic compounds that are known to cause adverse effects other than developmental toxicity. The extrapolation from a hazardous test condition in vitro, e.g. defined by a high Dp or Di, to adverse developmental effects in vivo remains difficult. One challenge is to simulate the concentrations of a test compound at the target cell of toxicity in vivo. Although physiologically based pharmacokinetic (PBPK) modelling addresses this aspect, its predictions are still associated with a relatively high degree of uncertainty. A further limitation that has to be taken into account is that in vitro systems do not necessarily recapitulate all in vivo relevant mechanisms. Despite these limitations, the concept proposed here with developmental potency and ratio indices quantitatively provides information regarding the propensity of test compounds to interfere with the complex transcriptional process required for normal human development. In conclusion, this study offers a concept for the identification of potential developmental toxicity hazards by establishing Dp and Di, which predict the propensity of test compounds to interfere with transcriptional processes during human development.
The authors thank the Project SysDT funded by the BMBF (German Ministry of Education and Research) and the EU Project EU-ToxRisk21 for financial support. The authors also thank Ms. Margit Henry and Ms. Tamara Rotshteyn for technical support.
- Adler S, Basketter D, Creton S, Pelkonen O, van Benthem J, Zuang V, Andersen KE, Angers-Loustau A, Aptula A, Bal-Price A, Benfenati E, Bernauer U, Bessems J, Bois FY, Boobis A, Brandon E, Bremer S, Broschard T, Casati S, Coecke S, Corvi R, Cronin M, Daston G, Dekant W, Felter S, Grignard E, Gundert-Remy U, Heinonen T, Kimber I, Kleinjans J, Komulainen H, Kreiling R, Kreysa J, Leite SB, Loizou G, Maxwell G, Mazzatorta P, Munn S, Pfuhler S, Phrakonkham P, Piersma A, Poth A, Prieto P, Repetto G, Rogiers V, Schoeters G, Schwarz M, Serafimova R, Tahti H, Testai E, van Delft J, van Loveren H, Vinken M, Worth A, Zaldivar JM (2011) Alternative (non-animal) methods for cosmetics testing: current status and future prospects-2010. Arch Toxicol 85:367–485CrossRefPubMedGoogle Scholar
- Annunen S, Körkkö J, Czarny M, Warman ML, Brunner HG, Kääriäinen H, Mulliken JB, Tranebjaerg L, Brooks DG, Cox GF, Cruysberg JR, Curtis MA, Davenport SLH, Friedrich CA, Kaitila I, Krawczynski MR, Latos-Bielenska A, Mukai S, BrR Olsen, Shinno N, Somer M, Vikkula M, Zlotogora J, Prockop DJ, Ala-Kokko L (1999) Splicing mutations of 54-bp exons in the COL11A1 gene cause marshall syndrome, but other mutations cause overlapping marshall/stickler phenotypes. Am J Hum Genet 65:974–983CrossRefPubMedPubMedCentralGoogle Scholar
- Balmer NV, Klima S, Rempel E, Ivanova VN, Kolde R, Weng MK, Meganathan K, Henry M, Sachinidis A, Berthold MR, Hengstler JG, Rahnenfuhrer J, Waldmann T, Leist M (2014) From transient transcriptome responses to disturbed neurodevelopment: role of histone acetylation and methylation as epigenetic switch between reversible and irreversible drug effects. Arch Toxicol 88:1451–1468CrossRefPubMedPubMedCentralGoogle Scholar
- Chabas D, Baranzini SE, Mitchell D, Bernard CCA, Rittling SR, Denhardt DT, Sobel RA, Lock C, Karpuj M, Pedotti R, Heller R, Oksenberg JR, Steinman L (2001) The Influence of the Proinflammatory Cytokine, Osteopontin, on Autoimmune Demyelinating Disease. Science 294:1731–1735CrossRefPubMedGoogle Scholar
- Chandane PG, Shah I (2014) Indian J Hum Genet 20(2):187–188Google Scholar
- Crosby AH, Patel H, Chioza BA, Proukakis C, Gurtz K, Patton MA, Sharifi R, Harlalka G, Simpson MA, Dick K, Reed JA, Al-Memar A, Chrzanowska-Lightowlers ZMA, Cross HE, Lightowlers RN (2010) Defective mitochondrial mRNA maturation is associated with spastic ataxia. Am J Hum Genet 87:655–660CrossRefPubMedPubMedCentralGoogle Scholar
- Godoy P, Schmidt-Heck W, Natarajan K, Lucendo-Villarin B, Szkolnicka D, Asplund A, Bjorquist P, Widera A, Stober R, Campos G, Hammad S, Sachinidis A, Chaudhari U, Damm G, Weiss TS, Nussler A, Synnergren J, Edlund K, Kuppers-Munther B, Hay DC, Hengstler JG (2015) Gene networks and transcription factor motifs defining the differentiation of stem cells into hepatocyte-like cells. J Hepatol 63:934–942CrossRefPubMedPubMedCentralGoogle Scholar
- Krug AK, Kolde R, Gaspar JA, Rempel E, Balmer NV, Meganathan K, Vojnits K, Baquie M, Waldmann T, Ensenat-Waser R, Jagtap S, Evans RM, Julien S, Peterson H, Zagoura D, Kadereit S, Gerhard D, Sotiriadou I, Heke M, Natarajan K, Henry M, Winkler J, Marchan R, Stoppini L, Bosgra S, Westerhout J, Verwei M, Vilo J, Kortenkamp A, Hescheler J, Hothorn L, Bremer S, van Thriel C, Krause KH, Hengstler JG, Rahnenfuhrer J, Leist M, Sachinidis A (2013) Human embryonic stem cell-derived test systems for developmental neurotoxicity: a transcriptomics approach. Arch Toxicol 87:123–143CrossRefPubMedGoogle Scholar
- Kubota A, Hida A, Ichikawa Y, Momose Y, Goto J, Igeta Y, Hashida H, Yoshida K, Ikeda S, Kanazawa I, Tsuji S (2009) A novel ferritin light chain gene mutation in a Japanese family with neuroferritinopathy: description of clinical features and implications for genotype-phenotype correlations. Mov Disord 24:441–445CrossRefPubMedGoogle Scholar
- Kuegler PB, Zimmer B, Waldmann T, Baudis B, Ilmjarv S, Hescheler J, Gaughwin P, Brundin P, Mundy W, Bal-Price AK, Schrattenholz A, Krause KH, van Thriel C, Rao MS, Kadereit S, Leist M (2010) Markers of murine embryonic and neural stem cells, neurons and astrocytes: reference points for developmental neurotoxicity testing. ALTEX 27:17–42PubMedGoogle Scholar
- Leist M, Bremer S, Brundin P, Hescheler J, Kirkeby A, Krause KH, Poerzgen P, Puceat M, Schmidt M, Schrattenholz A, Zak NB, Hentze H (2008) The biological and ethical basis of the use of human embryonic stem cells for in vitro test systems or cell therapy. ALTEX 25:163–190CrossRefPubMedGoogle Scholar
- Meganathan K, Jagtap S, Wagh V, Winkler J, Gaspar JA, Hildebrand D, Trusch M, Lehmann K, Hescheler J, Schluter H, Sachinidis A (2012) Identification of thalidomide-specific transcriptomics and proteomics signatures during differentiation of human embryonic stem cells. PLoS ONE 7:e44228CrossRefPubMedPubMedCentralGoogle Scholar
- Rempel E, Hoelting L, Waldmann T, Balmer NV, Schildknecht S, Grinberg M, Das Gaspar JA, Shinde V, Stober R, Marchan R, van Thriel C, Liebing J, Meisig J, Bluthgen N, Sachinidis A, Rahnenfuhrer J, Hengstler JG, Leist M (2015) A transcriptome-based classifier to identify developmental toxicants by stem cell testing: design, validation and optimization for histone deacetylase inhibitors. Arch Toxicol 89:1599–1618CrossRefPubMedPubMedCentralGoogle Scholar
- Ruzzo EK, Capo-Chichi JM, Ben-Zeev B, Chitayat D, Mao H, Pappas AL, Hitomi Y, Lu YF, Yao X, Hamdan FF, Pelak K, Reznik-Wolf H, Bar-Joseph I, Oz-Levi D, Lev D, Lerman-Sagie T, Leshinsky-Silver E, Anikster Y, Ben-Asher E, Olender T, Colleaux L, Décarie JC, Blaser S, Banwell B, Joshi RB, He XP, Patry L, Silver RJ, Dobrzeniecka S, Islam MS, Hasnat A, Samuels ME, Aryal DK, Rodriguiz RM, Jiang Y, Wetsel WC, McNamara JO, Rouleau GA, Silver DL, Lancet D, Pras E, Mitchell GA, Michaud JL, Goldstein DB (2013) Deficiency of asparagine synthetase causes congenital microcephaly and a progressive form of encephalopathy. Neuron 80:429–441CrossRefPubMedGoogle Scholar
- Shinde V, Klima S, Sureshkumar PS, Meganathan K, Jagtap S, Rempel E, Rahnenfuhrer J, Hengstler JG, Waldmann T, Hescheler J, Leist M, Sachinidis A (2015) Human pluripotent stem cell based developmental toxicity assays for chemical safety screening and systems biology data generation. J Vis Exp 110:e52333Google Scholar
- Vidal R, Ghetti B, Takao M, Brefel-Courbon C, Uro-Coste E, Glazier BS, Siani V, Benson MD, Calvas P, Miravalle L, Rascol O, Delisle MB (2004) Intracellular ferritin accumulation in neural and extraneural tissue characterizes a neurodegenerative disease associated with a mutation in the ferritin light polypeptide gene. J Neuropathol Exp Neurol 63:363–380CrossRefPubMedGoogle Scholar
- Waldmann T, Rempel E, Balmer NV, Konig A, Kolde R, Gaspar JA, Henry M, Hescheler J, Sachinidis A, Rahnenfuhrer J, Hengstler JG, Leist M (2014) Design principles of concentration-dependent transcriptome deviations in drug-exposed differentiating stem cells. Chem Res Toxicol 27:408–420CrossRefPubMedPubMedCentralGoogle Scholar
- Webb EA, AlMutair A, Kelberman D, Bacchelli C, Chanudet E, Lescai F, Andoniadou CL, Banyan A, Alsawaid A, Alrifai MT, Alahmesh MA, Balwi M, Mousavy-Gharavy SN, Lukovic B, Burke D, McCabe MJ, Kasia T, Kleta R, Stupka E, Beales PL, Thompson DA, Chong WK, Alkuraya FS, Martinez-Barbera JP, Sowden JC, Dattani MT (2013) ARNT2 mutation causes hypopituitarism, post-natal microcephaly, visual and renal anomalies. Brain 136:3096–3105CrossRefPubMedPubMedCentralGoogle Scholar
- Zimmer B, Kuegler PB, Baudis B, Genewsky A, Tanavde V, Koh W, Tan B, Waldmann T, Kadereit S, Leist M (2011) Coordinated waves of gene expression during neuronal differentiation of embryonic stem cells as basis for novel approaches to developmental neurotoxicity testing. Cell Death Differ 18:383–395CrossRefPubMedGoogle Scholar
- Zimmer B, Pallocca G, Dreser N, Foerster S, Waldmann T, Westerhout J, Julien S, Krause KH, van Thriel C, Hengstler JG, Sachinidis A, Bosgra S, Leist M (2014) Profiling of drugs and environmental chemicals for functional impairment of neural crest migration in a novel stem cell-based test battery. Arch Toxicol 88:1109–1126PubMedPubMedCentralGoogle Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, 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.