StemPanTox: A fast and wide-target drug assessment system for tailor-made safety evaluations using personalized iPS cells.

An alternative model that reliably predicts human-specific toxicity is necessary because the translatability of effects on animal models for human disease is limited to context. Previously, we developed a method that accurately predicts developmental toxicity based on the gene networks of undifferentiated human embryonic stem (ES) cells. Here, we advanced this method to predict adult toxicities of 24 chemicals in six categories (neurotoxins, cardiotoxins, hepatotoxins, two types of nephrotoxins, and non-genotoxic carcinogens) and achieved high predictability (AUC = 0.90-1.00) in all categories. Moreover, we screened for an induced pluripotent stem (iPS) cell line to predict the toxicities based on the gene networks of iPS cells using transfer learning of the gene networks of ES cells, and predicted toxicities in four categories (neurotoxins, hepatotoxins, glomerular nephrotoxins, and non-genotoxic carcinogens) with high performance (AUC = 0.82-0.99). This method holds promise for tailor-made safety evaluations using personalized iPS cells.

Integration of read-across and artificial neural network-based QSAR models for predicting systemic toxicity: A case study for valproic acid.

We present a systematic, comprehensive and reproducible weight-of-evidence approach for predicting the no-observed-adverse-effect level (NOAEL) for systemic toxicity by using read-across and quantitative structure-activity relationship (QSAR) models to fill gaps in rat repeated-dose and developmental toxicity data. As a case study, we chose valproic acid, a developmental toxicant in humans and animals. High-quality in vivo oral rat repeated-dose and developmental toxicity data were available for five and nine analogues, respectively, and showed qualitative consistency, especially for developmental toxicity. Similarity between the target and analogues is readily defined computationally, and data uncertainties associated with the similarities in structural, physico-chemical and toxicological properties, including toxicophores, were low. Uncertainty associated with metabolic similarity is low-to-moderate, largely because the approach was limited to in silico prediction to enable systematic and objective data collection. Uncertainty associated with completeness of read-across was reduced by including in vitro and in silico metabolic data and expanding the experimental animal database. Taking the "worst-case" approach, the smallest NOAEL values among the analogs (i.e., 200 and 100 mg/kg/day for repeated-dose and developmental toxicity, respectively) were read-across to valproic acid. Our previous QSAR models predict repeated-dose NOAEL of 148 (males) and 228 (females) mg/kg/day, and developmental toxicity NOAEL of 390 mg/kg/day for valproic acid. Based on read-across and QSAR, the conservatively predicted NOAEL is 148 mg/kg/day for repeated-dose toxicity, and 100 mg/kg/day for developmental toxicity. Experimental values are 341 mg/kg/day and 100 mg/kg/day, respectively. The present approach appears promising for quantitative and qualitative in silico systemic toxicity prediction of untested chemicals.

Development of QSAR models using artificial neural network analysis for risk assessment of repeated-dose, reproductive, and developmental toxicities of cosmetic ingredients.

Use of laboratory animals for systemic toxicity testing is subject to strong ethical and regulatory constraints, but few alternatives are yet available. One possible approach to predict systemic toxicity of chemicals in the absence of experimental data is quantitative structure-activity relationship (QSAR) analysis. Here, we present QSAR models for prediction of maximum "no observed effect level" (NOEL) for repeated-dose, developmental and reproductive toxicities. NOEL values of 421 chemicals for repeated-dose toxicity, 315 for reproductive toxicity, and 156 for developmental toxicity were collected from Japan Existing Chemical Data Base (JECDB). Descriptors to predict toxicity were selected based on molecular orbital (MO) calculations, and QSAR models employing multiple independent descriptors as the input layer of an artificial neural network (ANN) were constructed to predict NOEL values. Robustness of the models was indicated by the root-mean-square (RMS) errors after 10-fold cross-validation (0.529 for repeated-dose, 0.508 for reproductive, and 0.558 for developmental toxicity). Evaluation of the models in terms of the percentages of predicted NOELs falling within factors of 2, 5 and 10 of the in-vivo-determined NOELs suggested that the model is applicable to both general chemicals and the subset of chemicals listed in International Nomenclature of Cosmetic Ingredients (INCI). Our results indicate that ANN models using in silico parameters have useful predictive performance, and should contribute to integrated risk assessment of systemic toxicity using a weight-of-evidence approach. Availability of predicted NOELs will allow calculation of the margin of safety, as recommended by the Scientific Committee on Consumer Safety (SCCS).

Regulation of embryonic size in early mouse development in vitro culture system.

Mammals self-regulate their body size throughout development. In the uterus, embryos are properly regulated to be a specific size at birth. Previously, size and cell number in aggregated embryos, which were made from two or more morulae, and half embryos, which were halved at the 2-cell stage, have been analysed in vivo in preimplantation and post-implantation development in mice. Here, we examined whether or not the mouse embryo has the capacity to self-regulate growth using an in vitro culture system. To elucidate embryonic histology, cells were counted in aggregated or half embryos in comparison with control embryos. Both double- and triple-aggregated embryos contained more cells than did control embryos during all culture periods, and the relative growth ratios showed no growth inhibition in an in vitro culture system. Meanwhile, half embryos contained fewer cells than control embryos, but the number grew throughout the culture period. Our data suggest that the growth of aggregated embryos is not affected and continues in an in vitro culture system. On the other hand, the growth of half embryos accelerates and continues in an in vitro culture system. This situation, in turn, implied that post-implantation mouse embryos might have some potential to regulate their own growth and size as seen by using an in vitro culture system without uterus factors. In conclusion, our results indicated that embryos have some ways in which to regulate their own size in mouse early development.

Discoidin domain receptor 2 (DDR2) regulates proliferation of endochondral cells in mice.

Discoidin domain receptor 2 (DDR2) is a receptor tyrosine kinase that is activated by fibrillar collagens. DDR2 regulates cell proliferation, cell adhesion, migration, and extracellular matrix remodeling. The decrement of endogenous DDR2 represses osteoblastic marker gene expression and osteogenic differentiation in murine preosteoblastic cells, but the functions of DDR2 in chondrogenic cellular proliferation remain unclear. To better understand the role of DDR2 signaling in cellular proliferation in endochondral ossification, we inhibited Ddr2 expression via the inhibitory effect of miRNA on Ddr2 mRNA (miDdr2) and analyzed the cellular proliferation and differentiation in the prechondrocyte ATDC5 cell lines. To investigate DDR2's molecular role in endochondral cellular proliferation in vivo, we also produced transgenic mice in which the expression of truncated, kinase dead (KD) DDR2 protein is induced, and evaluated the DDR2 function in cellular proliferation in chondrocytes. Although the miDdr2-transfected ATDC5 cell lines retained normal differentiation ability, DDR2 reduction finally promoted cellular proliferation in proportion to the decreasing ratio of Ddr2 expression, and it also promoted earlier differentiation to cartilage cells by insulin induction. The layer of hypertrophic chondrocytes in KD Ddr2 transgenic mice was not significantly thicker than that of normal littermates, but the layer of proliferative chondrocytes in KD-Ddr2 transgenic mice was significantly thicker than that of normal littermates. Taken together, our data demonstrated that DDR2 might play a local and essential role in the proliferation of chondrocytes.