Bioassay predictive values for chemical health risks in drinking water.

Bioanalytical tools can be used for assessment of the chemical quality of drinking water and its sources. For water managers it is important to know the probability that a bioassay response above an established health-based 'effect-based trigger value' (EBT) indeed implies a harmful chemical (mixture) concentration. This study presents and applies a framework, based on Bayes' theorem, to derive such risk probabilities for bioassay responses. These were evaluated under varying (in silico) chemical mixture concentrations relevant to drinking water (sources), with toxicity data for six in vitro assays from the ToxCast database. For single chemicals and in silico mixtures, the negative predictive value (NPV) was 100 % for all assays. For water managers, this means that when a bioassay response is below the EBT, a chemical risk is reliably absent, and no further action is required. The positive predictive value (PPV) increased with increasing chemical concentrations (2 µg/L) up to 40-80 %, depending on the assay. For in silico mixtures of increasing numbers of chemicals, the PPV did not increase until higher sum concentrations (>2-10 µg/L). Hence, the ability to accurately signal a harmful chemical (mixture) using bioassays will be lowest for highly diverse, low-concentration chemical mixtures. For water managers, this means in practice that further investigations after an EBT exceedance will, in many cases, not reveal chemicals at harmful concentrations. A solution offered is to increase the trigger value for positive responses to achieve a higher PPV and maintain the EBT for negative responses to ensure an optimal NPV.

Organic micropollutant removal in full-scale rapid sand filters used for drinking water treatment in The Netherlands and Belgium.

Biological treatment processes have the potential to remove organic micropollutants (OMPs) during water treatment. The OMP removal capacity of conventional drinking water treatment processes such as rapid sand filters (RSFs), however, has not been studied in detail. We investigated OMP removal and transformation product (TP) formation in seven full-scale RSFs all treating surface water, using high-resolution mass spectrometry based quantitative suspect and non-target screening (NTS). Additionally, we studied the microbial communities with 16S rRNA gene amplicon sequencing (NGS) in both influent and effluent waters as well as the filter medium, and integrated these data to comprehensively assess the processes that affect OMP removal. In the RSF influent, 9 to 30 of the 127 target OMPs were detected. The removal efficiencies ranged from 0 to 93%. A data-driven workflow was established to monitor TPs, based on the combination of NTS feature intensity profiles between influent and effluent samples and the prediction of biotic TPs. The workflow identified 10 TPs, including molecular structure. Microbial community composition analysis showed similar community composition in the influent and effluent of most RSFs, but different from the filter medium, implying that specific microorganisms proliferate in the RSFs. Some of these are able to perform typical processes in water treatment such as nitrification and iron oxidation. However, there was no clear relationship between OMP removal efficiency and microbial community composition. The innovative combination of quantitative analyses, NTS and NGS allowed to characterize real scale biological water treatments, emphasizing the potential of bio-stimulation applications in drinking water treatment.

Cyclosporin A induced toxicity in mouse liver slices is only slightly aggravated by Fxr-deficiency and co-occurs with upregulation of pro-inflammatory genes and downregulation of genes involved in mitochondrial functions.

BACKGROUND: The transcription factor farnesoid X receptor (FXR) governs bile acid and energy homeostasis, is involved in inflammation, and has protective functions in the liver. In the present study we investigated the effect of Fxr deficiency in mouse precision cut liver slices (PCLS) exposed to a model hepatotoxicant cyclosporin A (CsA). It was anticipated that Fxr deficiency could aggravate toxicity of CsA in PCLS and pinpoint to novel genes/processes regulated by FXR. METHODS: To test this hypothesis, PCLS obtained from livers of wild type mice (WT-PCLS) and Fxr-knockout mice (FXRKO-PCLS) were treated with 40 µM CsA for 24 h and 48 h. ATP and histological assays were applied to assess the viability of PCLS. DNA microarrays combined with bioinformatics analysis were used to identify genes and processes that were affected by CsA in WT-PCLS and/or FXRKO-PCLS. In addition, WT-PCLS and FXRKO-PCLS were exposed to the endogenous FXR ligand chenodeoxycholic acid (CDCA) and subjected to q-PCR to determine whether subsets of known FXR-targets and the identified genes were regulated upon FXR activation in an FXR-dependent manner. RESULTS: No difference in viability was observed between WT-PCLS and FXRKO-PCLS upon CsA treatment. Transcriptomics data analysis revealed that CsA significantly upregulated stress-response and inflammation and significantly downregulated processes involved in lipid and glucose metabolism in WT-PCLS and FXRKO-PCLS. However, only in FXRKO-PCLS, CsA upregulated additional pro-inflammatory genes and downregulated genes related to mitochondrial functions. Furthermore, only in WT-PCLS, CDCA upregulated a subset of known FXR-target genes as well as the regulator of inflammation and mitochondrial functions peroxisome proliferator- activated receptor delta (Ppar delta). CONCLUSIONS: Although FXR governs energy metabolism, no major differences in response to CsA could be observed between WT-PCLS and FXRKO-PCLS in regulation of processes involved in lipid and glucose metabolism. This finding indicates that CsA does not directly affect FXR functions in relation to the above mentioned processes. However, the more pronounced induction of pro-inflammatory genes and the downregulation of genes involved in mitochondrial functions only in FXRKO-PCLS suggest that FXR deficiency aggravates CsA-induced inflammation and impairs mitochondrial functions. Therefore, FXR can exert its hepatoprotective functions by controlling inflammation and mitochondrial functions, possibly involving an FXR-PPAR delta cross-talk.

A game theoretic analysis of research data sharing.

While reusing research data has evident benefits for the scientific community as a whole, decisions to archive and share these data are primarily made by individual researchers. In this paper we analyse, within a game theoretical framework, how sharing and reuse of research data affect individuals who share or do not share their datasets. We construct a model in which there is a cost associated with sharing datasets whereas reusing such sets implies a benefit. In our calculations, conflicting interests appear for researchers. Individual researchers are always better off not sharing and omitting the sharing cost, at the same time both sharing and not sharing researchers are better off if (almost) all researchers share. Namely, the more researchers share, the more benefit can be gained by the reuse of those datasets. We simulated several policy measures to increase benefits for researchers sharing or reusing datasets. Results point out that, although policies should be able to increase the rate of sharing researchers, and increased discoverability and dataset quality could partly compensate for costs, a better measure would be to directly lower the cost for sharing, or even turn it into a (citation-) benefit. Making data available would in that case become the most profitable, and therefore stable, strategy. This means researchers would willingly make their datasets available, and arguably in the best possible way to enable reuse.

Comparison of the molecular topologies of stress-activated transcription factors HSF1, AP-1, NRF2, and NF-κB in their induction kinetics of HMOX1.

For cells, reacting aptly to changes in their environment is of critical importance. The protein Heme oxygenase-1 (HMOX1) plays a critical role as a guard of cellular homeostasis and is considered as a reliable indicator of cellular oxidative stress. A better insight in the regulation of HMOX1 would assist in understanding the physiological role of HMOX1 as well as improving functional interpretation of the gene as a biomarker in toxicogenomics. Remarkably, as many as four transcription factors are known to regulate the HMOX1 gene: HSF1, AP-1, NRF2, and NF-κB. To investigate induction kinetics of these transcription factors, we constructed mathematical simulation models for each of them. We included the topology of the known interactions of molecules involved in the activation of the transcription factors, and the feedback loops resulting in their down-regulation. We evaluate how the molecular circuitries associated with the different transcription factors differ in their kinetics regarding HMOX1 induction, under different scenarios of acute and less acute stress. We also evaluate the combined effect of the four transcription factors on HMOX1 expression and the resulting alleviation of stress. Overall, the results support the assumption of different biological roles for the four transcription factors, with AP-1 being a fast acting general stress response protein at the expense of efficiency, and NRF2 being important for cellular homeostasis in maintaining low levels of oxidative stress.

Gene expression markers in the zebrafish embryo reflect a hepatotoxic response in animal models and humans.

The zebrafish embryo (ZFE) is a promising non-rodent model in toxicology, and initial studies suggested its applicability in detecting hepatotoxic responses. Here, we hypothesize that the detailed analysis of underlying mechanisms of hepatotoxicity in ZFE contributes to the improved identification of hepatotoxic properties of new compounds and to the reduction of rodents used for screening. ZFEs were exposed to nine reference hepatotoxicants, targeted at induction of cholestasis, steatosis and necrosis, and two non-hepatotoxic controls. Histopathology revealed various specific morphological changes in the ZFE hepatocytes indicative of cell injury. Gene expression profiles of the individual compounds were generated using microarrays. Regulation of single genes and of pathways could be linked to hepatotoxic responses in general, but phenotype-specific responses could not be distinguished. Hepatotoxicity-associated pathways included xenobiotic metabolism and oxidoreduction related pathways. Overall analysis of gene expression identified a limited set of potential biomarkers specific for a common hepatotoxicity response. This set included several cytochrome P450 genes (cyp2k19, cyp4v7, cyp2aa3), genes related to liver development (pklr) and genes important in oxidoreduction processes (zgc:163022, zgc:158614, zgc:101858 and sqrdl). In conclusion, the ZFE model allows for identification of hepatotoxicants, without discrimination into specific phenotypes.

In vivo murine hepatic microRNA and mRNA expression signatures predicting the (non-)genotoxic carcinogenic potential of chemicals.

There is a high need to improve the assessment of, especially non-genotoxic, carcinogenic features of chemicals. We therefore explored a toxicogenomics-based approach using genome-wide microRNA and mRNA expression profiles upon short-term exposure in mice. For this, wild-type mice were exposed for seven days to three different classes of chemicals, i.e., four genotoxic carcinogens (GTXC), seven non-genotoxic carcinogens (NGTXC), and five toxic non-carcinogens. Hepatic expression patterns of mRNA and microRNA transcripts were determined after exposure and used to assess the discriminative power of the in vivo transcriptome for GTXC and NGTXC. A final classifier set, discriminative for GTXC and NGTXC, was generated from the transcriptomic data using a tiered approach. This appeared to be a valid approach, since the predictive power of the final classifier set in three different classifier algorithms was very high for the original training set of chemicals. Subsequent validation in an additional set of chemicals revealed that the predictive power for GTXC remained high, in contrast to NGTXC, which appeared to be more troublesome. Our study demonstrated that the in vivo microRNA-ome has less discriminative power to correctly identify (non-)genotoxic carcinogen classes. The results generally indicate that single mRNA transcripts do have the potential to be applied in risk assessment, but that additional (genomic) strategies are necessary to correctly predict the non-genotoxic carcinogenic potential of a chemical.

Cyclosporine A treated in vitro models induce cholestasis response through comparison of phenotype-directed gene expression analysis of in vivo Cyclosporine A-induced cholestasis.

In vitro models for hepatotoxicity testing are a necessity for advancement of toxicological research. Assessing the in vitro response requires in vivo validated gene sets reflective of the hepatotoxic phenotype. Cholestasis, the impairment of bile flow, is induced in C57BL/6J mice treated with cyclosporine A (CsA) to identify phenotype reflective gene sets. CsA treatment through oral gavage for 25 days induced cholestasis, as confirmed by histopathology and serum chemistry. Over 1, 4, and 11 days of CsA exposure gradual increases in serum markers were correlated to gene expression. This phenotype-directed analysis identified gene sets specific to the onset and progression of cholestasis, such as PPAR related processes and drug metabolism, by circumventing other effects of CsA, such as immunosuppression, found in dose*time group analysis. In vivo gene sets are enriched in publicly available data sets of CsA-treated HepaRG and primary mouse hepatocytes. However, genes identified within these gene sets did not overlap between in vivo and in vitro. In vitro regulated genes represent the initial response to cholestasis, whereas in vivo genes represent the later adaptive response. We conclude that the applicability of in vitro models for hepatotoxicity testing fully depends on a solid in vivo phenotype anchored analysis.

Transcriptomic analysis in the developing zebrafish embryo after compound exposure: individual gene expression and pathway regulation.

The zebrafish embryotoxicity test is a promising alternative assay for developmental toxicity. Classically, morphological assessment of the embryos is applied to evaluate the effects of compound exposure. However, by applying differential gene expression analysis the sensitivity and predictability of the test may be increased. For defining gene expression signatures of developmental toxicity, we explored the possibility of using gene expression signatures of compound exposures based on commonly expressed individual genes as well as based on regulated gene pathways. Four developmental toxic compounds were tested in concentration-response design, caffeine, carbamazepine, retinoic acid and valproic acid, and two non-embryotoxic compounds, d-mannitol and saccharin, were included. With transcriptomic analyses we were able to identify commonly expressed genes, which were mostly development related, after exposure to the embryotoxicants. We also identified gene pathways regulated by the embryotoxicants, suggestive of their modes of action. Furthermore, whereas pathways may be regulated by all compounds, individual gene expression within these pathways can differ for each compound. Overall, the present study suggests that the use of individual gene expression signatures as well as pathway regulation may be useful starting points for defining gene biomarkers for predicting embryotoxicity.

Exploring the zebrafish embryo as an alternative model for the evaluation of liver toxicity by histopathology and expression profiling.

The whole zebrafish embryo model (ZFE) has proven its applicability in developmental toxicity testing. Since functional hepatocytes are already present from 36 h post fertilization onwards, whole ZFE have been proposed as an attractive alternative to mammalian in vivo models in hepatotoxicity testing. The goal of the present study is to further underpin the applicability of whole ZFE for hepatotoxicity testing by combining histopathology and next-generation sequencing-based gene expression profiling. To this aim, whole ZFE and adult zebrafish were exposed to a set of hepatotoxic reference compounds. Histopathology revealed compound and life-stage-specific effects indicative of toxic injury in livers of whole ZFE and adult zebrafish. Next-generation sequencing (NGS) was used to compare transcript profiles in pooled individual RNA samples of whole ZFE and livers of adult zebrafish. This revealed that hepatotoxicity-associated expression can be detected beyond the overall transcription noise in the whole embryo. In situ hybridization verified liver specificity of selected highly expressed markers in whole ZFE. Finally, cyclosporine A (CsA) was used as an illustrative case to support applicability of ZFE in hepatotoxicity testing by comparing CsA-induced gene expression between ZFE, in vivo mouse liver and HepaRG cells on the levels of single genes, pathways and transcription factors. While there was no clear overlap on single gene level between the whole ZFE and in vivo mouse liver, strong similarities were observed between whole ZFE and in vivo mouse liver in regulated pathways related to hepatotoxicity, as well as in relevant overrepresented transcription factors. In conclusion, both the use of NGS of pooled RNA extracts analysis combined with histopathology and traditional microarray in single case showed the potential to detect liver-related genes and processes within the transcriptome of a whole zebrafish embryo. This supports the applicability of the whole ZFE model for compound-induced hepatotoxicity screening.

Applicability of a keratinocyte gene signature to predict skin sensitizing potential.

There is a need to replace animal tests for the identification of skin sensitizers and currently many alternative assays are being developed that have very promising results. In this study a gene signature capable of very accurate identification of sensitizers was established in the HaCaT human keratinocyte cell line. This signature was evaluated in a separate study using six chemicals that are either local lymph node (LLNA) false-positive or false-negative chemicals in addition to nine sensitizers and four non-sensitizers. Similar studies do not apply these more difficult to classify chemicals, which show the true potential for human predictions of an assay. Although the gene signature has improved prediction accuracy compared to the LLNA, the misclassified compounds were comparable between the two assays. Gene profiling also showed a sensitizer specific response of the Nrf2-keap1 and Toll-like receptor signaling pathways. After exposure to non-sensitizing chemicals that induce either of the pathways the signature misclassified all Nrf2-inducers, while the Toll-like receptor ligands were correctly classified. In conclusion, we confirm that keratinocyte based prediction assays may provide essential information on the properties of compounds. Furthermore, chemical selection is critical for assessment of the performance of in vitro alternative assays.

Unraveling toxicological mechanisms and predicting toxicity classes with gene dysregulation networks.

The use of genes for distinguishing classes of toxicity has become well established. In this paper we combine the reconstruction of a gene dysregulation network (GDN) with a classifier to assign unseen compounds to their appropriate class. Gene pairs in the GDN are dysregulated in the sense that they are linked by a common expression pattern in one class and differ in this pattern in another class. The classifier gives a quantitative measure on this difference by its prediction accuracy. As an in-depth example, gene pairs were selected that were dysregulated between skin cells treated with either sensitizers or irritants. Pairs with known and novel markers were found such as HMOX1 and ZFAND2A, ATF3 and PPP1R15A, OXSR1 and HSPA1B, ZFP36 and MAFF. The resulting GDN proved biologically valid as it was well-connected and enriched in known interactions, processes and common regulatory motifs for pairs. Classification accuracy was improved when compared with conventional classifiers. As the dysregulated patterns for heat shock responding genes proved to be distinct from those of other stress genes, we were able to formulate the hypothesis that heat shock genes play a specific role in sensitization, apart from other stress genes. In conclusion, our combined approach creates added value for classification-based toxicogenomics by obtaining novel, well-distinguishing and biologically interesting measures, suitable for the formulation of hypotheses on functional relationships between genes and their relevance for toxicity class differences.

Triazole-induced gene expression changes in the zebrafish embryo.

The zebrafish embryo is considered to provide a promising alternative test model for developmental toxicity testing. Most systems use morphological assessment of the embryos, however, microarray analyses may increase sensitivity and predictability of the test by detecting more subtle and detailed responses. In this study, we investigated the possibility of relating gene expression profiles of structurally similar chemicals tested in a single concentration, to a complete transcriptomic concentration-response of flusilazole (FLU). We tested five other triazoles, hexaconazole (HEX), cyproconazole (CYP), triadimefon (TDF), myclobutanil (MYC), and triticonazole (TTC) at equipotent concentrations based on morphological evaluation. Results showed that every compound had a different degree of regulation within their anti-fungal and developmental toxicity pathways, steroid biosynthesis and retinol metabolism, respectively. Assuming that the ratio between these pathways is relevant for efficacy compared to developmental toxicity, we found TTC was more efficient and CYP was more toxic compared to the other triazoles. With the approach used in this study we demonstrated that gene expression data allow more comprehensive assessment of compound effects by discriminating relative potencies using these specific gene sets. The zebrafish embryo model can therefore be considered a useful vertebrate model providing information of relevant pathways related to anti-fungal mechanism of action and toxicological activity.

Concentration-response analysis of differential gene expression in the zebrafish embryotoxicity test following flusilazole exposure.

The zebrafish embryotoxicity test (ZET) is considered a promising alternative model in predictive toxicology. Currently, morphological assessment of the embryo is the main readout for this assay. However, implementation of transcriptomics may help to detect more subtle effects, which may increase the sensitivity and predictability of the test. In this study, we tested a concentration response of flusilazole in the ZET. After exposure for 24 h postfertilization, microarray analysis revealed a number of processes to be regulated in a concentration-dependent way. We identified development related processes, retinol metabolism and transcription, as well as processes corresponding to the antifungal mechanism of action, steroid biosynthesis, and fatty acid metabolism, to be differentially regulated. Retinol metabolism and transcription were already significantly altered at concentrations that were not inducing morphological effects. Differential expression of genes related to steroid biosynthesis and fatty acid metabolism showed a concentration response similar to morphological response. An increase in concentration was also positively associated with an increase in magnitude of expression for individual genes within functional processes. Our study shows that transcriptomics analysis in the ZET is a more sensitive readout of compound-induced effects than morphological assessment. However, the interpretation of differential gene expression in terms of predicting morphological effects is not straightforward and requires further study.

A comparison of gene expression responses in rat whole embryo culture and in vivo: time-dependent retinoic acid-induced teratogenic response.

The whole embryo culture (WEC) model serves as a potential alternative for classical in vivo developmental toxicity testing. In the WEC, cultured rat embryos are exposed during neurulation and early organogenesis and evaluated for morphological effects. Toxicogenomic-based approaches may improve the predictive ability of WEC by providing molecular-based markers associated with chemical exposure, which can be compared across multiple parameters (e.g., exposure duration, developmental time, experimental model). Additionally, comparisons between in vitro and in vivo models may identify objective relevant molecular responses linked with developmental toxicity endpoints in vivo. In this study, using a transcriptomic approach, we compared all-trans retinoic acid (RA)-exposed and nonexposed Wistar rat embryos derived using WEC (RA, 0.5 µg/ml) or in vivo (RA, 50 mg/kg, oral gavage) to identify overlapping and nonoverlapping effects of RA on RNA expression in parallel with morphological changes. Across six time points (gestational day 10 + 2-48 h), we observed strong similarities in RA response at the gene (directionality, significance) and functional (e.g., embryonic development, cell differentiation) level which associated with RA-induced adverse morphological effects, including growth reduction as well as alterations in neural tube, limb, branchial, and mandible development. We observed differences between models in the timing of RA-induced effects on genes related to embryonic development and RA metabolism. These observations on the gene expression level were associated with specific differential morphological outcomes. This study supports the use of WEC to examine compound-induced molecular responses relative to in vivo and, furthermore, assists in defining the applicability domain of the WEC in determining complementary windows of sensitivity for developmental toxicological investigations.

Chemical class-specific gene expression changes in the zebrafish embryo after exposure to glycol ether alkoxy acids and 1,2,4-triazole antifungals.

The zebrafish embryotoxicity test (ZET) is an alternative test to predict embryotoxicity of substances based on morphological assessment. Implementing transcriptomics may increase sensitivity and objectivity of the test system. We applied the category approach to compare effects of compounds from two chemical classes, the glycol ethers and 1,2,4-triazoles, on the embryo. At 24h post fertilization, microarray analysis revealed several thousands of responsive genes after glycol ether exposure, whereas the triazoles significantly regulated only several hundreds of genes. Principal component analysis of the genes commonly regulated per chemical class demonstrated that the two classes can be distinguished. Gene set enrichment analysis showed that after glycol ether exposure mainly gene sets related to development were downregulated. After triazole exposure, gene sets corresponding to previously described mechanisms of action, such as glycolysis and fatty acid metabolism were regulated. Our results demonstrate that transcriptomics in the ZET provides a more sensitive endpoint than standard morphological assessment. In addition, information about mechanisms of action of substances may become available, thereby facilitating the extrapolation of findings to mammalian species including man.

Effects of pooling RNA from samples treated with different compounds for determining class specific biomarkers and processes in toxicogenomics.

Pooling of RNA samples is generally applied to obtain samples that represent the average signal of biological replicates of a single treatment. For toxicogenomics, pooling RNA of samples treated by different compounds could in the same way summarize these compounds to a single sample with average signals per class. In this study, we investigated the efficiency of such an approach to establish class specific differences in gene expression. As an example we took skin sensitizing compounds as one class and irritating compounds as another. A direct comparison was made to separately hybridized RNA samples. We observed that pooling RNA from compounds of a class substantially increased power to detect significantly regulated genes between classes because variability between pooled samples was much lower. Within pools the vast majority of genes maintained patterns of expression compared to the separately hybridized samples, especially in regulated genes. Both designs yielded appropriate biomarkers. Biomarkers selected from the pooled and separate design performed equally in classification of compounds to their class and relevant processes were found enriched in both designs. Consequently, pooling of RNA of different compound treated samples can be applied to determine class specific biomarkers and processes at much reduced cost and with limited loss of accuracy.

Gene set assembly for quantitative prediction of developmental toxicity in the embryonic stem cell test.

The embryonic stem cell test (EST) is an in vitro method for predicting developmental toxicity based on compound-induced inhibition of embryonic stem cell (ESC) differentiation. We previously described how gene expression analysis in the EST can be used to describe normal ESC differentiation as well as identify compound developmental toxicity, by means of our differentiation track algorithm. In this study, we combined raw data from our three previous studies in a new integrated analysis, to identify a gene set that allows for improved prediction. By evaluating predictions of 100,000 randomly selected gene sets, we identified which genes contribute significantly to the prediction reliability. By additional cross-validation, we identified a set of 52 genes that allows for improved prediction of toxicity. The correlation between the predictions using this gene set and the magnitude of the EST endpoint was 0.85, and the gene set predicted developmental toxicity with 83% accuracy (area under the curve 89%). If compounds with ineffective data because of a too low tested concentration or too much variation between samples were excluded, even 100% accuracy could be reached based on 15 compounds. This novel gene set consists mainly of genes involved in the stem cell differentiation or other developmental processes. We expect that this set can be of use in future studies aimed at improving the EST for risk assessment, thus making a next step towards regulatory implementation of this method.

Identification by gene coregulation mapping of novel genes involved in embryonic stem cell differentiation.

A combined analysis of data from a series of literature studies can lead to more reliable results than that based on a single study. A common problem in performing combined analyses of literature microarray gene expression data is that the original raw data are not always available and not always easy to combine in one analysis. We propose an approach that does not require analyzing original raw data, but instead takes literature gene sets derived from (supplementary) tables as input and uses gene co-occurrence in these sets for mapping a co-regulation network. An algorithm for this method was applied to a collection of literature-derived gene sets related to embryonic stem cell (ESC) differentiation. In the resulting network, genes involved in similar biological processes or expressed at similar time points during differentiation were found to cluster together. Using this information, we identified 43 genes not previously associated with cardiac ESC differentiation for which we were able to assign a putative novel biological function. For 6 of these genes (Apobec2, Cth, Ptges, Rrad, Zfp57, and 2410146L05Rik), literature data on mouse knockout phenotypes support their putative function. Three other genes (Rcor2, Zfp503, and Hspb3) are part of major pathways within the network and therefore likely mechanistically relevant candidate genes. We anticipate that these 43 genes can help to improve the understanding of the molecular events underlying ESC differentiation. Moreover, the approach introduced here can be more widely applied to identify possible novel gene functions in biological processes.

Keratinocyte gene expression profiles discriminate sensitizing and irritating compounds.

Many chemicals can induce allergic contact dermatitis. Because evaluation of skin sensitizing potential by animal testing is prohibited for cosmetics, and screening of many chemicals is required within Registration, Evaluation, Authorisation and Restriction of Chemicals, urgent need exists for predictive in vitro assays to identify contact allergens. Keratinocytes (KC) are the first cells encountered when chemicals land on the skin. Therefore, KC form an important site of haptenization and their metabolism is likely to be important. Moreover, KC secrete mediators that affect processing and presentation of haptenized proteins by dendritic cells. To develop a KC-based in vitro assay to predict sensitizing potential of chemicals, in vitro exposure effects of eight contact sensitizers and six irritants on the KC cell line HaCaT were examined by gene profiling. Classifiers predictive of the class sensitizers or irritants were calculated, based on support vector machine (SVM) and random forest (RF) algorithms. Classifiers using high-ranking genes were 70% (SVM) and 62% (RF) accurate, based on three (SVM) and two to five (RF) features. Classifiers using oxidative stress pathway gene sets were 68-73% (SVM) and 69-71% (RF) accurate. Cross-validation showed that the top-3 of most discriminating genes added up to 13 genes and included oxidative stress gene HMOX1 irrespective of the chemical left out. Moreover, HMOX1 was the most significantly regulated gene. Gene Set Enrichment Analysis showed upregulation of "Keap1 dependent" and "oxidative stress" gene lists. In conclusion, KC expression profiling can identify contact sensitizers, providing opportunities for nonanimal testing for sensitizing potential. Moreover, our data suggest that contact sensitizers induce the oxidative stress pathway in KC.