Discriminating different classes of toxicants by transcript profiling

Multi-task pretrained language model with novel application domains enables more comprehensive health and ecological toxicity prediction

In silico models for screening substances of healthy and ecological concern are essential for effective chemical management. However, current data-driven toxicity prediction models confront formidable challenges related to expressive capacity, data scarcity, and reliability issues. Thus, this study introduces TOX-BERT, a SMILES-based pretrained model for screening health and ecological toxicity. Results show that masked atom recovery pretraining and multi-task learning offer promising solutions to enhance model capacity and address data scarcity issues. Two novel application domain (AD) parameters, termed PCA-AD and LDS, were proposed to improve prediction reliability of TOX-BERT with accuracy surpassing 90 % and mean absolute error (MAE) below 0.52. TOX-BERT was applied to 18,905 IECSC chemicals, revealing distinct toxicity relationships that align with experimental studies such as those between cardiotoxicity and acute ecotoxicity. In addition to previous PBT screening, 156 potential high-risk chemicals for specific endpoint were identified covering 7 categories. Furthermore, a SMILES-based toxicity site detection approach was developed for structural toxicity analysis. These advancements carry profound implications to address challenges faced by current data-driven toxicity prediction models. TOX-BERT emerges as a valuable tool for more comprehensive, reliable, and applicable predictions of health and ecological toxicity in chemical risk assessment and management.

SigCom LINCS: data and metadata search engine for a million gene expression signatures

Millions of transcriptome samples were generated by the Library of Integrated Network-based Cellular Signatures (LINCS) program. When these data are processed into searchable signatures along with signatures extracted from Genotype-Tissue Expression (GTEx) and Gene Expression Omnibus (GEO), connections between drugs, genes, pathways and diseases can be illuminated. SigCom LINCS is a webserver that serves over a million gene expression signatures processed, analyzed, and visualized from LINCS, GTEx, and GEO. SigCom LINCS is built with Signature Commons, a cloud-agnostic skeleton Data Commons with a focus on serving searchable signatures. SigCom LINCS provides a rapid signature similarity search for mimickers and reversers given sets of up and down genes, a gene set, a single gene, or any search term. Additionally, users of SigCom LINCS can perform a metadata search to find and analyze subsets of signatures and find information about genes and drugs. SigCom LINCS is findable, accessible, interoperable, and reusable (FAIR) with metadata linked to standard ontologies and vocabularies. In addition, all the data and signatures within SigCom LINCS are available via a well-documented API. In summary, SigCom LINCS, available at https://maayanlab.cloud/sigcom-lincs, is a rich webserver resource for accelerating drug and target discovery in systems pharmacology.

Combining In Vivo and Organotypic In Vitro Approaches to Assess the Human Relevance of Basimglurant (RG7090), a Potential CAR Activator

Basimglurant (RG7090), a small molecule under development to treat certain forms of depression, demonstrated foci of altered hepatocytes in a long-term rodent-toxicity study. Additional evidence pointed toward the activation of the constitutive androstane receptor (CAR), an established promoter of nongenotoxic and rodent-specific hepatic tumors. This mode of action and the potential human relevance was explored in vivo using rodent and cynomolgus monkey models and in vitro using murine and human liver spheroids. Wild type (WT) and CAR/pregnane X receptor (PXR) knockout mice (CAR/PXR KO) were exposed to RG7090 for 8 consecutive days. Analysis of liver lysates revealed induction of Cyp2b mRNA and enzyme activity, a known activation marker of CAR, in WT but not in CAR/PXR KO animals. A series of proliferative genes were upregulated in WT mice only, and immunohistochemistry data showed increased cell proliferation exclusively in WT mice. In addition, primary mouse liver spheroids were challenged with RG7090 in the presence or absence of modified antisense oligonucleotides inhibiting CAR and/or PXR mRNA, showing a concentration-dependent Cyp2b mRNA induction only if CAR was not repressed. On the contrary, neither human liver spheroids nor cynomolgus monkeys exposed to RG7090 triggered CYP2B mRNA upregulation. Our data suggested RG7090 to be a rodent-specific CAR activator, and that CAR activation and its downstream processes were involved in the foci of altered hepatocytes formation detected in vivo. Furthermore, we demonstrated the potential of a new in vitro approach using liver spheroids and antisense oligonucleotides for CAR knockdown experiments, which could eventually replace in vivo investigations using CAR/PXR KO mice.

Assessing Chemical-Induced Liver Injury In Vivo From In Vitro Gene Expression Data in the Rat: The Case of Thioacetamide Toxicity

Consumers are exposed to thousands of chemicals with potentially adverse health effects. However, these chemicals will never be tested for toxicity because of the immense resources needed for animal-based (in vivo) toxicological studies. Today, there are no viable in vitro alternatives to these types of animal studies. To develop an in vitro approach, we investigated whether we could predict in vivo organ injuries in rats with the use of RNA-seq data acquired from tissues early in the development of toxicant-induced injury, by comparing gene expression data from RNA isolated from these rat tissues with those obtained from in vitro exposure of primary liver and kidney cells. We collected RNA-seq data from the liver and kidney tissues of Sprague-Dawley rats 8 or 24 h after exposing them to vehicle (control), low (25 mg/kg), or high (100 mg/kg) doses of thioacetamide, a known liver toxicant that promotes fibrosis; we used these doses and exposure times to cause only mild toxicant-induced injury. For the in vitro study, we treated two cell types from Sprague-Dawley rats, primary hepatocytes (vehicle; low, 0.025 mM; or high, 0.125 mM dose), and renal tube epithelial cells (vehicle; low, 0.125 mM; or high, 0.500 mM) dose) with the thioacetamide metabolite, thioacetamide-S-oxide, selecting in vitro doses and exposure times to recreate the early-stage toxicant-induced injury model that we achieved in vivo. RNA-seq data were collected 9 or 24 h after application of vehicle or thioacetamide-S-oxide. We found that our modular approach for the analysis of gene expression data derived from in vivo RNA-seq strongly correlated (R² > 0.6) with the in vitro results at two different dose levels of thioacetamide/thioacetamide-S-oxide after 24 h of exposure. The top-ranked liver injury modules in vitro correctly identified the ensuing development of liver fibrosis.

Connectivity Mapping: Methods and Applications

Connectivity mapping resources consist of signatures representing changes in cellular state following systematic small-molecule, disease, gene, or other form of perturbations. Such resources enable the characterization of signatures from novel perturbations based on similarity; provide a global view of the space of many themed perturbations; and allow the ability to predict cellular, tissue, and organismal phenotypes for perturbagens. A signature search engine enables hypothesis generation by finding connections between query signatures and the database of signatures. This framework has been used to identify connections between small molecules and their targets, to discover cell-specific responses to perturbations and ways to reverse disease expression states with small molecules, and to predict small-molecule mimickers for existing drugs. This review provides a historical perspective and the current state of connectivity mapping resources with a focus on both methodology and community implementations.

Deep Neural Network Models for Predicting Chemically Induced Liver Toxicity Endpoints From Transcriptomic Responses

Improving the accuracy of toxicity prediction models for liver injuries is a key element in evaluating the safety of drugs and chemicals. Mechanism-based information derived from expression (transcriptomic) data, in combination with machine-learning methods, promises to improve the accuracy and robustness of current toxicity prediction models. Deep neural networks (DNNs) have the advantage of automatically assembling the relevant features from a large number of input features. This makes them especially suitable for modeling transcriptomic data, which typically contain thousands of features. Here, we gaged gene- and pathway-level feature selection schemes using single- and multi-task DNN approaches in predicting chemically induced liver injuries (biliary hyperplasia, fibrosis, and necrosis) from whole-genome DNA microarray data. The single-task DNN models showed high predictive accuracy and endpoint specificity, with Matthews correlation coefficients for the three endpoints on 10-fold cross validation ranging from 0.56 to 0.89, with an average of 0.74 in the best feature sets. The DNN models outperformed Random Forest models in cross validation and showed better performance than Support Vector Machine models when tested in the external validation datasets. In the cross validation studies, the effect of the feature selection scheme was negligible among the studied feature sets. Further evaluation of the models on their ability to predict the injury phenotype per se for non-chemically induced injuries revealed the robust performance of the DNN models across these additional external testing datasets. Thus, the DNN models learned features specific to the injury phenotype contained in the gene expression data.

Identification of the Toxicity Pathways Associated With Thioacetamide-Induced Injuries in Rat Liver and Kidney

Ingestion or exposure to chemicals poses a serious health risk. Early detection of cellular changes induced by such events is vital to identify appropriate countermeasures to prevent organ damage. We hypothesize that chemically induced organ injuries are uniquely associated with a set (module) of genes exhibiting significant changes in expression. We have previously identified gene modules specifically associated with organ injuries by analyzing gene expression levels in liver and kidney tissue from rats exposed to diverse chemical insults. Here, we assess and validate our injury-associated gene modules by analyzing gene expression data in liver, kidney, and heart tissues obtained from Sprague-Dawley rats exposed to thioacetamide, a known liver toxicant that promotes fibrosis. The rats were injected intraperitoneally with a low (25 mg/kg) or high (100 mg/kg) dose of thioacetamide for 8 or 24 h, and definite organ injury was diagnosed by histopathology. Injury-associated gene modules indicated organ injury specificity, with the liver being most affected by thioacetamide. The most activated liver gene modules were those associated with inflammatory cell infiltration and fibrosis. Previous studies on thioacetamide toxicity and our histological analyses supported these results, signifying the potential of gene expression data to identify organ injuries.

A statistical framework for applying RNA profiling to chemical hazard detection

Use of 'omics technologies in environmental science is expanding. However, application is mostly restricted to characterizing molecular steps leading from toxicant interaction with molecular receptors to apical endpoints in laboratory species. Use in environmental decision-making is limited, due to difficulty in elucidating mechanisms in sufficient detail to make quantitative outcome predictions in any single species or in extending predictions to aquatic communities. Here we introduce a mechanism-agnostic statistical approach, supplementing mechanistic investigation by allowing probabilistic outcome prediction even when understanding of molecular pathways is limited, and facilitating extrapolation from results in laboratory test species to predictions about aquatic communities. We use concepts familiar to environmental managers, supplemented with techniques employed for clinical interpretation of 'omics-based biomedical tests. We describe the framework in step-wise fashion, beginning with single test replicates of a single RNA variant, then extending to multi-gene RNA profiling, collections of test replicates, and integration of complementary data. In order to simplify the presentation, we focus on using RNA profiling for distinguishing presence versus absence of chemical hazards, but the principles discussed can be extended to other types of 'omics measurements, multi-class problems, and regression. We include a supplemental file demonstrating many of the concepts using the open source R statistical package.

3D bioprinting for drug discovery and development in pharmaceutics

Successful launch of a commercial drug requires significant investment of time and financial resources wherein late-stage failures become a reason for catastrophic failures in drug discovery. This calls for infusing constant innovations in technologies, which can give reliable prediction of efficacy, and more importantly, toxicology of the compound early in the drug discovery process before clinical trials. Though computational advances have resulted in more rationale in silico designing, in vitro experimental studies still require gaining industry confidence and improving in vitro-in vivo correlations. In this quest, due to their ability to mimic the spatial and chemical attributes of native tissues, three-dimensional (3D) tissue models have now proven to provide better results for drug screening compared to traditional two-dimensional (2D) models. However, in vitro fabrication of living tissues has remained a bottleneck in realizing the full potential of 3D models. Recent advances in bioprinting provide a valuable tool to fabricate biomimetic constructs, which can be applied in different stages of drug discovery research. This paper presents the first comprehensive review of bioprinting techniques applied for fabrication of 3D tissue models for pharmaceutical studies. A comparative evaluation of different bioprinting modalities is performed to assess the performance and ability of fabricating 3D tissue models for pharmaceutical use as the critical selection of bioprinting modalities indeed plays a crucial role in efficacy and toxicology testing of drugs and accelerates the drug development cycle. In addition, limitations with current tissue models are discussed thoroughly and future prospects of the role of bioprinting in pharmaceutics are provided to the reader.

STATEMENT OF SIGNIFICANCE

Present advances in tissue biofabrication have crucial role to play in aiding the pharmaceutical development process achieve its objectives. Advent of three-dimensional (3D) models, in particular, is viewed with immense interest by the community due to their ability to mimic in vivo hierarchical tissue architecture and heterogeneous composition. Successful realization of 3D models will not only provide greater in vitro-in vivo correlation compared to the two-dimensional (2D) models, but also eventually replace pre-clinical animal testing, which has their own shortcomings. Amongst all fabrication techniques, bioprinting- comprising all the different modalities (extrusion-, droplet- and laser-based bioprinting), is emerging as the most viable fabrication technique to create the biomimetic tissue constructs. Notwithstanding the interest in bioprinting by the pharmaceutical development researchers, it can be seen that there is a limited availability of comparative literature which can guide the proper selection of bioprinting processes and associated considerations, such as the bioink selection for a particular pharmaceutical study. Thus, this work emphasizes these aspects of bioprinting and presents them in perspective of differential requirements of different pharmaceutical studies like in vitro predictive toxicology, high-throughput screening, drug delivery and tissue-specific efficacies. Moreover, since bioprinting techniques are mostly applied in regenerative medicine and tissue engineering, a comparative analysis of similarities and differences are also expounded to help researchers make informed decisions based on contemporary literature.

Gene Expression Profiling associated with Hepatoxicity in Pregnant Rats treated with Ubi Gadong (Dioscorea hispida) Extract

Dioscorea hispida (D.hispida) is the most well-known starchy tuber in Malaysia and called 'ubi gadong'. Despite concerns over toxicity effects, the tuber is known to possess therapeutic values due to the presence of bioactive compounds such as saponins. This study was performed to identify the changes in gene expression profiles associated with hepatoxicity in pregnant rat treated with D.hispida using RT² Profiler PCR Array. The identification of steroidal saponins from D.hispida was carried out by UHPLC/MS method. Treatment of D.hispida caused mortality when dosage above 2000 mg/kg b.w. was given to pregnant rats. The PCR array showed that several genes were significantly up and down-regulated upon treatment with D.hispida. Treatment of D.hispida at 2000 mg/kg b.w leads to significant upregulation of several genes such as Btg2, Gsr, L2hgdn, S100a8, Slc17a3, Bhmt, Cd68, Cyp1a2 whereas several genes were downregulated such as Abcb1a, Aldoa, Cdc14b, Icam1, Krt18, Hpn and Maob. The consumption of D.hispida extract when taken at lower dosage of 2000 mg/kg may not be harmful to rats. D.hispida extract given at the highest dosage to pregnant rats caused alterations of several genes categorized in different hepatotoxic group functions such as necrosis, cholestasis and phospholipodisis.

ABC gene-ranking for prediction of drug-induced cholestasis in rats

As legacy toxicogenomics databases have become available, improved data mining approaches are now key to extracting and visualizing subtle relationships between toxicants and gene expression. In the present study, a novel “aggregating bundles of clusters” (ABC) procedure was applied to separate cholestatic from non-cholestatic drugs and model toxicants in the Johnson & Johnson (Janssen) rat liver toxicogenomics database [3]. Drug-induced cholestasis is an important issue, particularly when a new compound enters the market with this liability, with standard preclinical models often mispredicting this toxicity. Three well-characterized cholestasis-responsive genes (Cyp7a1, Mrp3 and Bsep) were chosen from a previous in-house Janssen gene expression signature; these three genes show differing, non-redundant responses across the 90+ paradigm compounds in our database. Using the ABC procedure, extraneous contributions were minimized in comparisons of compound gene responses. All genes were assigned weights proportional to their correlations with Cyp7a1, Mrp3 and Bsep, and a resampling technique was used to derive a stable measure of compound similarity. The compounds that were known to be associated with rat cholestasis generally had small values of this measure relative to each other but also had large values of this measure relative to non-cholestatic compounds. Visualization of the data with the ABC-derived signature showed a very tight, essentially identically behaving cluster of robust human cholestatic drugs and experimental cholestatic toxicants (ethinyl estradiol, LPS, ANIT and methylene dianiline, disulfiram, naltrexone, methapyrilene, phenacetin, alpha-methyl dopa, flutamide, the NSAIDs–—indomethacin, flurbiprofen, diclofenac, flufenamic acid, sulindac, and nimesulide, butylated hydroxytoluene, piperonyl butoxide, and bromobenzene), some slightly less active compounds (3′-acetamidofluorene, amsacrine, hydralazine, tannic acid), some drugs that behaved very differently, and were distinct from both non-cholestatic and cholestatic drugs (ketoconazole, dipyridamole, cyproheptadine and aniline), and many postulated human cholestatic drugs that in rat showed no evidence of cholestasis (chlorpromazine, erythromycin, niacin, captopril, dapsone, rifampicin, glibenclamide, simvastatin, furosemide, tamoxifen, and sulfamethoxazole). Most of these latter drugs were noted previously by other groups as showing cholestasis only in humans. The results of this work suggest that the ABC procedure and similar statistical approaches can be instrumental in combining data to compare toxicants across toxicogenomics databases, extract similarities among responses and reduce unexplained data varation.

Increased levels of urinary phenylacetylglycine associated with mitochondrial toxicity in a model of drug-induced phospholipidosis

BACKGROUND

Phospholipidosis (PLD) is a lysosomal storage disorder induced by a class of cationic amphiphilic drugs. However, drug-induced PLD is reversible. Evidence of PLD from animal studies with some compounds has led to discontinuation of development. Regulatory authorities are likely to request additional studies when PLD is linked to toxicity.

OBJECTIVE

We conducted a trial to investigate urinary phenylacetylglycine (uPAG) as a biomarker for PLD.

MATERIALS AND METHODS

Five groups of 12 male Wistar rats were dosed once with vehicle, 300 mg/kg or 1500 mg/kg of compound A (known to induce PLD), or 300 mg/kg or 1000 mg/kg of compound B (similar structure, but does not induce PLD) to achieve similar plasma exposures. Following dosing, urine and blood samples underwent nuclear magnetic resonance (NMR), proteomic, and biochemical analyses. Necropsies were performed at 48 and 168 h, organ histopathology evaluated, and gene expression in liver analyzed by microarray. Electron microscopic examination of peripheral lymphocytes was performed.

RESULTS

For compound A, uPAG increased with dose, correlating with lamellar inclusion bodies formation in peripheral lymphocytes. NMR analysis showed decreased tricarboxylic acid cycle intermediates, inferring mitochondrial toxicity. Mitochondrial dysfunction was suggested by uPAG increase, resulting from a switch to anaerobic metabolism or disruption of the urea cycle.

DISCUSSION AND CONCLUSION

uPAG shows utility as a noninvasive biomarker for mitochondrial toxicity associated with drug-induced PLD, providing a mechanistic hypothesis for toxicity associated with PLD likely resulting from combined direct and indirect mitochondrial toxicity via impairment of the proton motor force and alteration of fatty acid catabolism.

Data mining reveals a network of early-response genes as a consensus signature of drug-induced in vitro and in vivo toxicity

Gene signatures of drug-induced toxicity are of broad interest, but they are often identified from small-scale, single-time point experiments, and are therefore of limited applicability. To address this issue, we performed multivariate analysis of gene expression, cell-based assays, and histopathological data in the TG-GATEs (Toxicogenomics Project-Genomics Assisted Toxicity Evaluation system) database. Data mining highlights four genes-EGR1, ATF3, GDF15 and FGF21-that are induced 2 h after drug administration in human and rat primary hepatocytes poised to eventually undergo cytotoxicity-induced cell death. Modelling and simulation reveals that these early stress-response genes form a functional network with evolutionarily conserved structure and intrinsic dynamics. This is underlined by the fact that early induction of this network in vivo predicts drug-induced liver and kidney pathology with high accuracy. Our findings demonstrate the value of early gene-expression signatures in predicting and understanding compound-induced toxicity. The identified network can empower first-line tests that reduce animal use and costs of safety evaluation.

Identification of biomarkers that distinguish chemical contaminants based on gene expression profiles

Background

High throughput transcriptomics profiles such as those generated using microarrays have been useful in identifying biomarkers for different classification and toxicity prediction purposes. Here, we investigated the use of microarrays to predict chemical toxicants and their possible mechanisms of action.

Results

In this study, in vitro cultures of primary rat hepatocytes were exposed to 105 chemicals and vehicle controls, representing 14 compound classes. We comprehensively compared various normalization of gene expression profiles, feature selection and classification algorithms for the classification of these 105 chemicals into14 compound classes. We found that normalization had little effect on the averaged classification accuracy. Two support vector machine (SVM) methods, LibSVM and sequential minimal optimization, had better classification performance than other methods. SVM recursive feature selection (SVM-RFE) had the highest overfitting rate when an independent dataset was used for a prediction. Therefore, we developed a new feature selection algorithm called gradient method that had a relatively high training classification as well as prediction accuracy with the lowest overfitting rate of the methods tested. Analysis of biomarkers that distinguished the 14 classes of compounds identified a group of genes principally involved in cell cycle function that were significantly downregulated by metal and inflammatory compounds, but were induced by anti-microbial, cancer related drugs, pesticides, and PXR mediators.

Conclusions

Our results indicate that using microarrays and a supervised machine learning approach to predict chemical toxicants, their potential toxicity and mechanisms of action is practical and efficient. Choosing the right feature and classification algorithms for this multiple category classification and prediction is critical.

Blood transcriptomics: applications in toxicology

The number of new chemicals that are being synthesized each year has been steadily increasing. While chemicals are of immense benefit to mankind, many of them have a significant negative impact, primarily owing to their inherent chemistry and toxicity, on the environment as well as human health. In addition to chemical exposures, human exposures to numerous non-chemical toxic agents take place in the environment and workplace. Given that human exposure to toxic agents is often unavoidable and many of these agents are found to have detrimental human health effects, it is important to develop strategies to prevent the adverse health effects associated with toxic exposures. Early detection of adverse health effects as well as a clear understanding of the mechanisms, especially at the molecular level, underlying these effects are key elements in preventing the adverse health effects associated with human exposure to toxic agents. Recent developments in genomics, especially transcriptomics, have prompted investigations into this important area of toxicology. Previous studies conducted in our laboratory and elsewhere have demonstrated the potential application of blood gene expression profiling as a sensitive, mechanistically relevant and practical surrogate approach for the early detection of adverse health effects associated with exposure to toxic agents. The advantages of blood gene expression profiling as a surrogate approach to detect early target organ toxicity and the molecular mechanisms underlying the toxicity are illustrated and discussed using recent studies on hepatotoxicity and pulmonary toxicity. Furthermore, the important challenges this emerging field in toxicology faces are presented in this review article.

Combining next-generation sequencing and microarray technology into a transcriptomics approach for the non-model organism Chironomus riparius

Whole-transcriptome gene-expression analyses are commonly performed in species that have a sequenced genome and for which microarrays are commercially available. To do such analyses in species with no or limited genome data, i.e. non-model organisms, necessary transcriptomics resources, i.e. an annotated transcriptome and a validated gene-expression microarray, must first be developed. The aim of the present study was to establish an advanced approach for developing transcriptomics resources for non-model organisms by combining next-generation sequencing (NGS) and microarray technology. We applied our approach to the non-biting midge Chironomus riparius, an ecologically relevant species that is widely used in sediment ecotoxicity testing. We sampled extensively covering all C. riparius developmental stages as well as toxicant exposed larvae and obtained from a normalized cDNA library 1.5 M NGS reads totalling 501 Mbp. Using the NGS data we developed transcriptomics resources in several steps. First, we designed 844 k probes directly on the NGS reads, as well as 76 k probes targeting expressed sequence tags of related species. These probes were tested for their affinity to C. riparius DNA and mRNA, by performing two biological experiments with a 1 M probe-selection microarray that contained the entire probe-library. Subsequently, the 1.5 M NGS reads were assembled into 23,709 isotigs and 135,082 singletons, which were associated to ∼55 k, respectively, ∼61 k gene ontology terms and which corresponded together to 22,593 unique protein accessions. An algorithm was developed that took the assembly and the probe affinities to DNA and mRNA into account, what resulted in 59 k highly-reliable probes that targeted uniquely 95% of the isotigs and 18% of the singletons. Concluding, our approach allowed the development of high-quality transcriptomics resources for C. riparius, and is applicable to any non-model organism. It is expected, that these resources will advance ecotoxicity testing with C. riparius as whole-transcriptome gene-expression analysis are now possible with this species.

Alternative (non-animal) methods for cosmetics testing: current status and future prospects-2010

The 7th amendment to the EU Cosmetics Directive prohibits to put animal-tested cosmetics on the market in Europe after 2013. In that context, the European Commission invited stakeholder bodies (industry, non-governmental organisations, EU Member States, and the Commission's Scientific Committee on Consumer Safety) to identify scientific experts in five toxicological areas, i.e. toxicokinetics, repeated dose toxicity, carcinogenicity, skin sensitisation, and reproductive toxicity for which the Directive foresees that the 2013 deadline could be further extended in case alternative and validated methods would not be available in time. The selected experts were asked to analyse the status and prospects of alternative methods and to provide a scientifically sound estimate of the time necessary to achieve full replacement of animal testing. In summary, the experts confirmed that it will take at least another 7-9 years for the replacement of the current in vivo animal tests used for the safety assessment of cosmetic ingredients for skin sensitisation. However, the experts were also of the opinion that alternative methods may be able to give hazard information, i.e. to differentiate between sensitisers and non-sensitisers, ahead of 2017. This would, however, not provide the complete picture of what is a safe exposure because the relative potency of a sensitiser would not be known. For toxicokinetics, the timeframe was 5-7 years to develop the models still lacking to predict lung absorption and renal/biliary excretion, and even longer to integrate the methods to fully replace the animal toxicokinetic models. For the systemic toxicological endpoints of repeated dose toxicity, carcinogenicity and reproductive toxicity, the time horizon for full replacement could not be estimated.

Cross-platform comparison of microarray-based multiple-class prediction

High-throughput microarray technology has been widely applied in biological and medical decision-making research during the past decade. However, the diversity of platforms has made it a challenge to re-use and/or integrate datasets generated in different experiments or labs for constructing array-based diagnostic models. Using large toxicogenomics datasets generated using both Affymetrix and Agilent microarray platforms, we carried out a benchmark evaluation of cross-platform consistency in multiple-class prediction using three widely-used machine learning algorithms. After an initial assessment of model performance on different platforms, we evaluated whether predictive signature features selected in one platform could be directly used to train a model in the other platform and whether predictive models trained using data from one platform could predict datasets profiled using the other platform with comparable performance. Our results established that it is possible to successfully apply multiple-class prediction models across different commercial microarray platforms, offering a number of important benefits such as accelerating the possible translation of biomarkers identified with microarrays to clinically-validated assays. However, this investigation focuses on a technical platform comparison and is actually only the beginning of exploring cross-platform consistency. Further studies are needed to confirm the feasibility of microarray-based cross-platform prediction, especially using independent datasets.

The evolution of bioinformatics in toxicology: advancing toxicogenomics

As one reflects back through the past 50 years of scientific research, a significant accomplishment was the advance into the genomic era. Basic research scientists have uncovered the genetic code and the foundation of the most fundamental building blocks for the molecular activity that supports biological structure and function. Accompanying these structural and functional discoveries is the advance of techniques and technologies to probe molecular events, in time, across environmental and chemical exposures, within individuals, and across species. The field of toxicology has kept pace with advances in molecular study, and the past 50 years recognizes significant growth and explosive understanding of the impact of the compounds and environment to basic cellular and molecular machinery. The advancement of molecular techniques applied in a whole-genomic capacity to the study of toxicant effects, toxicogenomics, is no doubt a significant milestone for toxicological research. Toxicogenomics has also provided an avenue for advancing a joining of multidisciplinary sciences including engineering and informatics in traditional toxicological research. This review will cover the evolution of the field of toxicogenomics in the context of informatics integration its current promise, and limitations.

Promise and progress in environmental genomics: a status report on the applications of gene expression-based microarray studies in ecologically relevant fish species

The advent of any new technology is typically met with great excitement. So it was a few years ago, when the combination of advances in sequencing technology and the development of microarray technology made measurements of global gene expression in ecologically relevant species possible. Many of the review papers published around that time promised that these new technologies would revolutionize environmental biology as they had revolutionized medicine and related fields. A few years have passed since these technological advancements have been made, and the use of microarray studies in non-model fish species has been adopted in many laboratories internationally. Has the relatively widespread adoption of this technology really revolutionized the fields of environmental biology, including ecotoxicology, aquaculture and ecology, as promised? Or have these studies merely become a novelty and a potential distraction for scientists addressing environmentally relevant questions? In this review, the promises made in early review papers, in particular about the advances that the use of microarrays would enable, are summarized; these claims are compared to the results of recent studies to determine whether the forecasted changes have materialized. Some applications, as discussed in the paper, have been realized and have led to advances in their field, others are still under development.

Mapping drug physico-chemical features to pathway activity reveals molecular networks linked to toxicity outcome

The identification of predictive biomarkers is at the core of modern toxicology. So far, a number of approaches have been proposed. These rely on statistical inference of toxicity response from either compound features (i.e., QSAR), in vitro cell based assays or molecular profiling of target tissues (i.e., expression profiling). Although these approaches have already shown the potential of predictive toxicology, we still do not have a systematic approach to model the interaction between chemical features, molecular networks and toxicity outcome. Here, we describe a computational strategy designed to address this important need. Its application to a model of renal tubular degeneration has revealed a link between physico-chemical features and signalling components controlling cell communication pathways, which in turn are differentially modulated in response to toxic chemicals. Overall, our findings are consistent with the existence of a general toxicity mechanism operating in synergy with more specific single-target based mode of actions (MOAs) and provide a general framework for the development of an integrative approach to predictive toxicology.

Utilization of gene profiling and proteomics to determine mineral pathogenicity in a human mesothelial cell line (LP9/TERT-1)

Identifying and understanding the early molecular events that underscore mineral pathogenicity using in vitro screening tests is imperative, especially given the large number of synthetic and natural fibers and particles being introduced into the environment. The purpose of the work described here was to examine the ability of gene profiling (Affymetrix microarrays) to predict the pathogenicity of various materials in a human mesothelial cell line (LP9/TERT-1) exposed to equal surface area concentrations (15 x 10(6) or 75 x 10(6) microm(2)/cm(2)) of crocidolite asbestos, nonfibrous talc, fine titanium dioxide (TiO(2)), or glass beads for 8 or 24 h. Since crocidolite asbestos caused the greatest number of alterations in gene expression, multiplex analysis (Bio-Plex) of proteins released from LP9/TERT-1 cells exposed to crocidolite asbestos was also assessed to reveal if this approach might also be explored in future assays comparing various mineral types. To verify that LP9/TERT-1 cells were more sensitive than other cell types to asbestos, human ovarian epithelial cells (IOSE) were also utilized in microarray studies. Upon assessing changes in gene expression via microarrays, principal component analysis (PCA) of these data was used to identify patterns of differential gene expression. PCA of microarray data confirmed that LP9/TERT-1 cells were more responsive than IOSE cells to crocidolite asbestos or nonfibrous talc, and that crocidolite asbestos elicited greater responses in both cell types when compared to nonfibrous talc, TiO(2), or glass beads. Bio-Plex analysis demonstrated that asbestos caused an increase in interleukin-13 (IL-13), basic fibroblast growth factor (bFGF), granulocyte colony-stimulating factor (G-CSF), and vascular endothelial growth factor (VEGF). These responses were generally dose-dependent (bFGF and G-CSF only) and tumor necrosis factor (TNF)-alpha independent (except for G-CSF). Thus, microarray and Bio-Plex analyses are valuable in determining early molecular responses to fibers/particles and may directly contribute to understanding the etiology of diseases caused by them. The number and magnitude of changes in gene expression or "profiles" of secreted proteins may serve as valuable metrics for determining the potential pathogenicity of various mineral types. Hence, alterations in gene expression and cytokine/chemokine changes induced by crocidolite asbestos in LP9/TERT-1 cells may be indicative of its increased potential to cause mesothelioma in comparison to the other nonfibrous materials examined.

Alterations in gene expression in human mesothelial cells correlate with mineral pathogenicity

Human mesothelial cells (LP9/TERT-1) were exposed to low and high (15 and 75 microm(2)/cm(2) dish) equal surface area concentrations of crocidolite asbestos, nonfibrous talc, fine titanium dioxide (TiO2), or glass beads for 8 or 24 hours. RNA was then isolated for Affymetrix microarrays, GeneSifter analysis and QRT-PCR. Gene changes by asbestos were concentration- and time-dependent. At low nontoxic concentrations, asbestos caused significant changes in mRNA expression of 29 genes at 8 hours and of 205 genes at 24 hours, whereas changes in mRNA levels of 236 genes occurred in cells exposed to high concentrations of asbestos for 8 hours. Human primary pleural mesothelial cells also showed the same patterns of increased gene expression by asbestos. Nonfibrous talc at low concentrations in LP9/TERT-1 mesothelial cells caused increased expression of 1 gene Activating Transcription Factor 3 (ATF3) at 8 hours and no changes at 24 hours, whereas expression levels of 30 genes were elevated at 8 hours at high talc concentrations. Fine TiO2 or glass beads caused no changes in gene expression. In human ovarian epithelial (IOSE) cells, asbestos at high concentrations elevated expression of two genes (NR4A2, MIP2) at 8 hours and 16 genes at 24 hours that were distinct from those elevated in mesothelial cells. Since ATF3 was the most highly expressed gene by asbestos, its functional importance in cytokine production by LP9/TERT-1 cells was assessed using siRNA approaches. Results reveal that ATF3 modulates production of inflammatory cytokines (IL-1 beta, IL-13, G-CSF) and growth factors (VEGF and PDGF-BB) in human mesothelial cells.

Toxicogenomic biomarkers for liver toxicity

Toxicogenomics (TGx) is a widely used technique in the preclinical stage of drug development to investigate the molecular mechanisms of toxicity. A number of candidate TGx biomarkers have now been identified and are utilized for both assessing and predicting toxicities. Further accumulation of novel TGx biomarkers will lead to more efficient, appropriate and cost effective drug risk assessment, reinforcing the paradigm of the conventional toxicology system with a more profound understanding of the molecular mechanisms of drug-induced toxicity. In this paper, we overview some practical strategies as well as obstacles for identifying and utilizing TGx biomarkers based on microarray analysis. Since clinical hepatotoxicity is one of the major causes of drug development attrition, the liver has been the best documented target organ for TGx studies to date, and we therefore focused on information from liver TGx studies. In this review, we summarize the current resources in the literature in regard to TGx studies of the liver, from which toxicologists could extract potential TGx biomarker gene sets for better hepatotoxicity risk assessment.

The role of toxicoproteomics in assessing organ specific toxicity

Aims of this chapter on the role of toxicoproteomics in assessing organ-specific toxicity are to define the field of toxicoproteomics, describe its development among global technologies, and show potential uses in experimental toxicological research, preclinical testing and mechanistic biological research. Disciplines within proteomics deployed in preclinical research are described as Tier I analysis, involving global protein mapping and protein profiling for differential expression, and Tier II proteomic analysis, including global methods for description of function, structure, interactions and post-translational modification of proteins. Proteomic platforms used in toxicoproteomics research are briefly reviewed. Preclinical toxicoproteomic studies with model liver and kidney toxicants are critically assessed for their contributions toward understanding pathophysiology and in biomarker discovery. Toxicoproteomics research conducted in other organs and tissues are briefly discussed as well. The final section suggests several key developments involving new approaches and research focus areas for the field of toxicoproteomics as a new tool for toxicological pathology.

Filling and mining the reactive metabolite target protein database

The post-translational modification of proteins is a well-known endogenous mechanism for regulating protein function and activity. Cellular proteins are also susceptible to post-translational modification by xenobiotic agents that possess, or whose metabolites possess, significant electrophilic character. Such non-physiological modifications to endogenous proteins are sometimes benign, but in other cases they are strongly associated with, and are presumed to cause, lethal cytotoxic consequences via necrosis and/or apoptosis. The Reactive Metabolite Target Protein Database (TPDB) is a searchable, freely web-accessible (http://tpdb.medchem.ku.edu:8080/protein_database/) resource that attempts to provide a comprehensive, up-to-date listing of known reactive metabolite target proteins. In this report we characterize the TPDB by reviewing briefly how the information it contains came to be known. We also compare its information to that provided by other types of "-omics" studies relevant to toxicology, and we illustrate how bioinformatic analysis of target proteins may help to elucidate mechanisms of cytotoxic responses to reactive metabolites.

Gene expression response in target organ and whole blood varies as a function of target organ injury phenotype

This report details the standardized experimental design and the different data streams that were collected (histopathology, clinical chemistry, hematology and gene expression from the target tissue (liver) and a bio-available tissue (blood)) after treatment with eight known hepatotoxicants (at multiple time points and doses with multiple biological replicates). The results of the study demonstrate the classification of histopathological differences, likely reflecting differences in mechanisms of cell-specific toxicity, using either liver tissue or blood transcriptomic data.

Comparative transcriptomic responses to chronic cadmium, fluoranthene, and atrazine exposure in Lumbricus rubellus

Transcriptional responses of a soil-dwelling organism (the earthworm Lumbricus rubellus) to three chemicals, cadmium (Cd), fluoranthene (FA), and atrazine (AZ), were measured following chronic exposure, with the aim of identifying the nature of any shared transcriptional response. Principal component analysis indicated full or partial separation of control and exposed samples for each compound but not for the composite set of all control and exposed samples. Partial least-squares discriminant analysis allowed separation of the control and exposed samples for each chemical and also for the composite data set, suggesting a common transcriptional response to exposure. Genes identified as changing in expression level (by the least stringent test for significance) following exposure to two chemicals indicated a substantial number of common genes (> 127). The three compound overlapping gene set, however, comprised only 25 genes. We suggest that the low commonality in transcriptional response may be linked to the chronic concentrations (approximately 10% EC50) and chronic duration (28 days) used. Annotations of the three compound overlapping gene set indicated that genes from pathways most often associated with responses to environmental stress, such as heat shock, phase I and II metabolism, antioxidant defense, and cation balance, were not represented. The strongest annotation signature was for genes important in mitochondrial function and energy metabolism.

Molecular portrait of cisplatin induced response in human testis cancer cell lines based on gene expression profiles

BACKGROUND

Testicular germ cell tumors (TGCTs) respond well to cisplatin-based chemotherapy and show a low incidence of acquired resistance compared to most somatic tumors. The reasons for these specific characteristics are not known in detail but seem to be multifactorial. We have studied gene expression profiles of testicular and colon cancer derived cell lines treated with cisplatin. The main goal of this study was to identify novel gene expression profiles with their functional categories and the biochemical pathways that are associated with TGCT cells' response to cisplatin.

RESULTS

Genes that were differentially expressed between the TGCT cell lines vs the (somatic) HCT116 cell line, after cisplatin treatment, were identified using the significance analysis of microarrays (SAM) method. The response of TGCT cells was strikingly different from that of HCT116, and we identified 1794 genes that were differentially expressed. Functional classification of these genes showed that they participate in a variety of different and widely distributed functional categories and biochemical pathways. Database mining showed significant association of genes (n = 41) induced by cisplatin in our study, and genes previously reported to by expressed in differentiated TGCT cells. We identified 37 p53-responsive genes that were altered after cisplatin exposure. We also identified 40 target genes for two microRNAs, hsa-mir-372 and 373 that may interfere with p53 signaling in TGCTs. The tumor suppressor genes NEO1 and LATS2, and the estrogen receptor gene ESR1, all have binding sites for p53 and hsa-mir-372/373. NEO1 and LATS2 were down-regulated in TGCT cells following cisplatin exposure, while ESR1 was up-regulated in TGCT cells. Cisplatin-induced genes associated with terminal growth arrest through senescence were identified, indicating associations which were not previously described for TGCT cells.

CONCLUSION

By linking our gene expression data to publicly available databases and literature, we provide a global pattern of cisplatin induced cellular response that is specific for testicular cancer cell lines. We have identified cisplatin-responsive functional classes and pathways, such as the angiogenesis, Wnt, integrin, and cadherin signaling pathways. The identification of differentially expressed genes in this study may contribute to a better understanding of the unusual sensitivity of TGCT to some DNA-damaging agents.

Predictive toxicogenomics approaches reveal underlying molecular mechanisms of nongenotoxic carcinogenicity

Toxicogenomics technology defines toxicity gene expression signatures for early predictions and hypotheses generation for mechanistic studies, which are important approaches for evaluating toxicity of drug candidate compounds. A large gene expression database built using cDNA microarrays and liver samples treated with over one hundred paradigm compounds was mined to determine gene expression signatures for nongenotoxic carcinogens (NGTCs). Data were obtained from male rats treated for 24 h. Training/testing sets of 24 NGTCs and 28 noncarcinogens were used to select genes. A semiexhaustive, nonredundant gene selection algorithm yielded six genes (nuclear transport factor 2, NUTF2; progesterone receptor membrane component 1, Pgrmc1; liver uridine diphosphate glucuronyltransferase, phenobarbital-inducible form, UDPGTr2; metallothionein 1A, MT1A; suppressor of lin-12 homolog, Sel1h; and methionine adenosyltransferase 1, alpha, Mat1a), which identified NGTCs with 88.5% prediction accuracy estimated by cross-validation. This six genes signature set also predicted NGTCs with 84% accuracy when samples were hybridized to commercially available CodeLink oligo-based microarrays. To unveil molecular mechanisms of nongenotoxic carcinogenesis, 125 differentially expressed genes (P<0.01) were selected by Student's t-test. These genes appear biologically relevant, of 71 well-annotated genes from these 125 genes, 62 were overrepresented in five biochemical pathway networks (most linked to cancer), and all of these networks were linked by one gene, c-myc. Gene expression profiling at early time points accurately predicts NGTC potential of compounds, and the same data can be mined effectively for other toxicity signatures. Predictive genes confirm prior work and suggest pathways critical for early stages of carcinogenesis.

Immunotoxicogenomics: the potential of genomics technology in the immunotoxicity risk assessment process

Evaluation of xenobiotic-induced changes in gene expression as a method to identify and classify potential toxicants is being pursued by industry and regulatory agencies worldwide. A workshop was held at the Research Triangle Park campus of the Environmental Protection Agency to discuss the current state-of-the-science of "immunotoxicogenomics" and to explore the potential role of genomics techniques for immunotoxicity testing. The genesis of the workshop was the current lack of widely accepted triggering criteria for Tier 1 immunotoxicity testing in the context of routine toxicity testing data, the realization that traditional screening methods would require an inordinate number of animals and are inadequate to handle the number of chemicals that may need to be screened (e.g., high production volume compounds) and the absence of an organized effort to address the state-of-the-science of toxicogenomics in the identification of immunotoxic compounds. The major focus of the meeting was on the theoretical and practical utility of genomics techniques to (1) replace or supplement current immunotoxicity screening procedures, (2) provide insight into potential modes or mechanisms of action, and (3) provide data suitable for immunotoxicity hazard identification or risk assessment. The latter goal is of considerable interest to a variety of stakeholders as a means to reduce animal use and to decrease the cost of conducting and interpreting standard toxicity tests. A number of data gaps were identified that included a lack of dose response and kinetic data for known immunotoxic compounds and a general lack of data correlating genomic alterations to functional changes observed in vivo. Participants concluded that a genomics approach to screen chemicals for immunotoxic potential or to generate data useful to risk assessors holds promise but that routine use of these methods is years in the future. However, recent progress in molecular immunology has made mode and mechanism of action studies much more practical. Furthermore, a variety of published immunotoxicity studies suggest that microarray analysis is already a practical means to explore pathway-level changes that lead to altered immune function. To help move the science of immunotoxicogenomics forward, a partnership of industry, academia, and government was suggested to address data gaps, validation, quality assurance, and protocol development.

Biomarker discovery by imaging mass spectrometry: transthyretin is a biomarker for gentamicin-induced nephrotoxicity in rat

Adverse drug effects are often associated with pathological changes in tissue. An accurate depiction of the undesired affected area, possibly supported by mechanistic data, is important to classify the effects with regard to relevance for human patients. MALDI imaging MS represents a new analytical tool to directly provide the spatial distribution and the relative abundance of proteins in tissue. Here we evaluate this technique to investigate potential toxicity biomarkers in kidneys of rats that were administered gentamicin, a well known nephrotoxicant. Differential analysis of the mass spectrum profiles revealed a spectral feature at 12,959 Da that strongly correlates with histopathology alterations of the kidney. We unambiguously identified this spectral feature as transthyretin (Ser(28)-Gln(146)) using an innovative combination of tissue microextraction and fractionation by reverse-phase liquid chromatography followed by a top-down tandem mass spectrometric approach. Our findings clearly demonstrate the emerging role of imaging MS in the discovery of toxicity biomarkers and in obtaining mechanistic insights concerning toxicity mechanisms.

Meeting report: Validation of toxicogenomics-based test systems: ECVAM-ICCVAM/NICEATM considerations for regulatory use

This is the report of the first workshop "Validation of Toxicogenomics-Based Test Systems" held 11-12 December 2003 in Ispra, Italy. The workshop was hosted by the European Centre for the Validation of Alternative Methods (ECVAM) and organized jointly by ECVAM, the U.S. Interagency Coordinating Committee on the Validation of Alternative Methods (ICCVAM), and the National Toxicology Program (NTP) Interagency Center for the Evaluation of Alternative Toxicological Methods (NICEATM). The primary aim of the workshop was for participants to discuss and define principles applicable to the validation of toxicogenomics platforms as well as validation of specific toxicologic test methods that incorporate toxicogenomics technologies. The workshop was viewed as an opportunity for initiating a dialogue between technologic experts, regulators, and the principal validation bodies and for identifying those factors to which the validation process would be applicable. It was felt that to do so now, as the technology is evolving and associated challenges are identified, would be a basis for the future validation of the technology when it reaches the appropriate stage. Because of the complexity of the issue, different aspects of the validation of toxicogenomics-based test methods were covered. The three focus areas include a) biologic validation of toxicogenomics-based test methods for regulatory decision making, b) technical and bioinformatics aspects related to validation, and c) validation issues as they relate to regulatory acceptance and use of toxicogenomics-based test methods. In this report we summarize the discussions and describe in detail the recommendations for future direction and priorities.

How molecular profiling could revolutionize drug discovery

Information from genomic, proteomic and metabolomic measurements has already benefited target discovery and validation, assessment of efficacy and toxicity of compounds, identification of disease subgroups and the prediction of responses of individual patients. Greater benefits can be expected from the application of these technologies on a significantly larger scale; by simultaneously collecting diverse measurements from the same subjects or cell cultures; by exploiting the steadily improving quantitative accuracy of the technologies; and by interpreting the emerging data in the context of underlying biological models of increasing sophistication. The benefits of applying molecular profiling to drug discovery and development will include much lower failure rates at all stages of the drug development pipeline, faster progression from discovery through to clinical trials and more successful therapies for patient subgroups. Upheavals in existing organizational structures in the current 'conveyor belt' models of drug discovery might be required to take full advantage of these methods.

Preclinical Drug Safety Analysis by Chemogenomic Profiling in the Liver

The economic hurdles of drug development and the emergence of genomic technologies such as chemogenomics are combining to shift the existing paradigms in preclinical drug development. Today, the information gleaned from high content molecular data has begun to augment traditional approaches to the assessment of drug safety. The optimal approach is a hybrid strategy employing chemogenomic data and gene expression-based biomarkers of drug efficacy and toxicity to supplement low content and insensitive methods for risk assessment and mechanistic evaluation of drug candidates. Large reference databases of chemogenomic data are essential to the derivation and validation of accurate and predictive gene expression biomarkers. An example of the development of a predictive biomarker for hepatic bile duct hyperplasia is described herein. As gene expression technologies improve, biomarkers will achieve higher throughput, and become more cost effective and increasingly accurate. This will elevate the value of chemogenomics in drug development, shift attrition to earlier in the process, and reduce the overall cost of drug development. Over the past 2 to 3 years, the transition of chemogenomics from a research tool to a decision-making tool has begun and regulatory agencies are anxiously awaiting implementation of this technology to make faster and more informed evaluations of potential drugs.