Current estimates project that global chemical production—currently growing 3% per year—will double every 25 years (Wilson et al., 2006). In the United States alone, excluding fuels, pesticides, pharmaceuticals, or food products, about 42 billion pounds of chemicals are produced or imported daily (U.S. EPA, 2005). Many chemicals are managed through the Toxic Substances Control Act (TSCA), but several independent analyses have concluded that these regulations seriously hinder (1) toxicity testing and hazard assessment; (2) control of chemicals of concern; and (3) investment in safer alternatives, such as those generated by the tenets of green chemistry (Wilson and Schwarzman, 2009). Combined with the widespread use and distribution of industrial chemicals, the data and safety gaps precipitated by TSCA elicit a situation in which chemical exposures to humans and ecosystems are oftentimes of unknown hazard and risk.

The field of toxicology currently employs extensive animal-based assays to evaluate chemical toxicity, a burdensome and prohibitively expensive approach that typically assesses a limited number of endpoints. Considering that tens of thousands of in use chemicals lack adequate toxicity data (Judson et al., 2009), it is unreasonable to rely upon these traditional methods to fill data gaps. The National Research Council (NRC), realizing that more innovative approaches to testing were needed, envisioned that toxicology should commit to mechanistically-based high-throughput cellular in vitro assays (NRC, 2007; Andersen and Krewski, 2010). In this way, a more complete comprehension of chemical toxicity can be achieved, while expediting testing, decreasing costs, and reducing animal usage.

Although high-throughput in vitro methods certainly signal progress in toxicity testing, they are limited to existing assays with known endpoints, such as analyses of stress response pathways induced by oxidative species, heat shock, DNA damage, hypoxia, and unfolded proteins (Simmons et al., 2009). Another approach utilizes “omics” technologies such as gene expression profiling, proteomics, lipidomics, and metabolomics to conduct targeted and untargeted investigations into chemical mechanisms of toxicity (reviewed by Hamadeh et al., 2002; Gatzidou et al., 2007). However, by associating toxicant exposure with changes in mRNA, protein, lipid, or metabolite levels, these assays are correlative and do not provide direct links between genes and their requirements in the cellular toxicant response.

Functional toxicology is based in the high-throughput use of cells/organisms harboring gene deletions or depleted proteins to systematically examine genetic requirements for toxicity tolerance. Any assayable phenotype can be measured in response to a toxicant, but viability or fitness are the most conventional endpoints (Figure 1). Functional techniques can provide information distinct from the aforementioned correlative methodologies; for example, Giaever et al. (2002) found that expression of a gene is generally unrelated to its requirement for growth under a selective condition. Functional analyses, which have been conducted in budding and fission yeast (Table 1), bacteria, nematodes, fruit flies, zebrafish, and human cell lines (Table 2), can (1) contribute novel insight into chemical mechanisms of action; (2) define more specific toxicological endpoints; and (3) inform further mechanistic-based assays.

For many reasons, the eukaryotic budding (Saccharomyces cerevisiae) and fission (Schizosaccharomyces pombe) yeasts are ideal models in which to conduct functional toxicological studies. Numerous metabolic and signaling pathways, along with basic cellular processes, are conserved with more complex organisms such as humans. Human homologs have been identified for a large number of yeast genes, with several hundred of the conserved genes linked to disease in humans (Steinmetz et al., 2002; Wood et al., 2002). A long history of genetic manipulation in yeasts confers the ability to selectively target and examine conserved genes and pathways throughout their genomes, facilitating functional analyses. The ease of culture and availability of software resources, molecular protocols, and genetic and physical interaction data collectively bolster the value of yeasts in toxicology.

Barcoded mutant collections have been generated in budding (Giaever et al., 1999, 2002) and fission yeast (Kennedy et al., 2008; Kim et al., 2010; Chen et al., 2012), allowing assessment of individual strain fitness in pooled cultures under selective conditions (reviewed by North and Vulpe, 2010; dos Santos et al., 2012). This technique, known as functional profiling, functional genomics, chemical genomics, or chemical-genetic profiling, can identify the genetic requirements for tolerance to any substance that causes measurable growth inhibition in yeast. Figure 2 demonstrates the screening process, while Table 1 provides a summary of recent functional studies. Homozygous profiling (HOP) utilizes strains deleted for non-essential genes to establish the genetic requirements for chemical tolerance, while haploinsufficiency profiling (HIP) detects strains sensitized to a chemical targeting the product of their corresponding heterozygous locus (for a review, see Smith et al., 2010a). In brief, DNA sequences (“barcodes”) uniquely identifying each deletion strain enable parallel growth analyses with thousands of pooled mutants exposed to a chemical of interest. A PCR amplification of the barcodes and their subsequent quantification via microarray hybridization or sequencing allows for discovery of strains with altered growth in the particular substance. The decreased abundance by mRNA perturbation (dAMP) collection complements heterozygote profiling by destabilizing a gene's mRNA (and thus depleting the encoded protein) via disruption of 3′-untranslated regions (Yan et al., 2008). An additional tool that may benefit functional toxicology is the barcoded yeast overexpression library (Ho et al., 2009; Douglas et al., 2012). Similar to HOP and HIP, this technique enables highly parallel and systematic investigations of overexpression phenotypes in pooled cultures. Finally, a novel “functional variomics” approach utilizes high-complexity random mutagenesis to identify genes conferring drug resistance due to mutations or overexpression (Huang et al., 2013a). The advent of multiplexed high-throughput barcode sequencing of pooled cultures (Han et al., 2010; Smith et al., 2010b) promises a future of expedited and cost efficient functional genomic analyses.

The functional tools available in yeast provide unmatched resources for inquiries into potential cellular and molecular mechanisms of toxicity. Such analyses have informed functional experimentation in more complex organisms such as zebrafish or human cell lines. For example, Ishizaki et al. (2010) utilized a yeast chemical-genetic screen to reveal that intracellular trafficking defects conferred sensitivity to copper limitation, and further reported that knockdown of zebrafish homologs to these yeast genes sensitized fish to copper-dependent hypopigmentation, a hallmark of copper deficiency in humans. Following identification of the Sas2p histone acetyltransferase as a modulator of arsenite tolerance in yeast, knockdown of its homolog MYST1 in human bladder epithelial cells was found to similarly induce arsenite sensitivity (Jo et al., 2009a,b). Another group demonstrated that the investigational cancer drug elesclomol affected electron transport mutants in yeast and extended their analysis by determining that elesclomol interacted with the electron transport chain in human cells (Blackman et al., 2012). Likewise, a functional screen in yeast identified mitochondrial translation inhibition as the lethality mechanism of the antimicrobial and antileukemic compound tigecycline, and this activity was confirmed in leukemic cells (Skrtić et al., 2011). Finally, Jo et al. (2009a) used yeast to show that a S-adenosylmethionine dependent methyltransferase conferred resistance to various arsenic species, while Ren et al. (2011) showed the corresponding gene in humans (N6AMT1) could metabolize arsenic in human urothelial cells. Harari et al. (2013) expanded upon these studies by demonstrating that N6AMT1 polymorphisms were associated with arsenic methylation in Andean women, and posited that the polymorphisms could be used as susceptibility markers for arsenic toxicity.

The recent development of the TagModule collection (Oh et al., 2010a), building upon the work of Xu et al. (2007), takes advantage of barcoded transposons to extend the yeast DNA barcoding methodology to a variety of microorganisms. In essence, in vitro transposon mutagenesis is utilized to mutagenize a genomic DNA library, and subsequent transformation of barcoded genomic fragments into a compatible unicellular organism allows for genome-wide unbiased screening of chemical-genetic interactions. Akin to the yeast functional process, the barcodes can be amplified from pooled cultures and counted by microarray hybridization or high-throughput sequencing. Oh et al. (2010a) demonstrated the versatility of the TagModule collection by generating tagged mutants in the bacterium Shewanella oneidensis MR-1 and the fungal pathogen Candida albicans. As a proof of principle, the authors identified S. oneidensis mutants with growth deficiencies in minimal media and C. albicans mutants sensitive to the antifungal drug clotrimazole (Oh et al., 2010a). The same group reports on additional haploinsufficiency screens in C. albicans (Oh et al., 2010b) and S. oneidensis (Deutschbauer et al., 2011) encompassing a wide variety of growth conditions and diverse chemical compounds. Furthermore, the method was applied to identify genes important for plant hydrolysate tolerance in Zymomonas mobilis, a bacterium with potential for commercial-scale cellulosic ethanol production (Skerker et al., 2013). By facilitating barcoding of mutant or overexpression collections for pooled functional analyses in a range of organisms, the TagModule system can be a valuable tool for toxicity testing.

Alternative approaches utilizing deletion/overexpression strains, or high-throughput sequencing of tagged transposon mutants may have applications relevant to functional toxicology. A signature-tagged mutagenesis strategy permitted parallel analysis of Cryptococcus neoformans fungal mutants in experimental infections (Liu et al., 2008), while high-throughput sequencing examined the relative quantities of human gut bacterium Bacteroides thetaiotaomicron transposon mutants in wild-type and immunodeficient gnotobiotic mice (Goodman et al., 2009). Relative strain abundance has been quantified in a collection of homozygous C. albicans deletion mutants, albeit in a lower-throughput investigation (Noble et al., 2010) than allowed by the TagModule system (Oh et al., 2010a). Overexpression studies can be conducted in C. albicans, however the available ORFeome is confined to a few hundred genes (Chauvel et al., 2012). Genome-wide deletion libraries have been constructed in Escherichia coli (Baba et al., 2006), Bacillus subtilis (Kobayashi et al., 2003), and Pseudomonas aeruginosa (Jacobs et al., 2003), with a limited set available in Salmonella enterica (Santiviago et al., 2009). Thousands of specific genetic modifications were simultaneously evaluated to quantify population dynamics in various media and growth inhibitors in E. coli (Warner et al., 2010). It is conceivable that functional toxicological or even pharmaceutical inquiries can be performed using any of these tools and/or organisms.

Inherent in high-throughput functional toxicology methods is the use of computational analyses to decipher toxicant mechanisms of action. Various studies have utilized overenrichment or Cytoscape software (Shannon et al., 2003) to uncover yeast genetic networks affected by a toxicant (Zhou et al., 2009; North et al., 2011; Gaytán et al., 2013a,b; Kaiser et al., 2013). Others, by integrating data from many distinct yeast chemical-genetic datasets, have assembled chemical-phenotype networks that helped identify potential effects elicited by various compounds (Venancio et al., 2010). Without computational analyses, the high-throughput approaches in zebrafish (Truong et al., 2014) or human cell lines (Shalem et al., 2014) would not be possible.

For a variety of reasons, functional toxicology has its limitations in the range of organisms discussed within this review. Less complex organisms such as yeasts, other fungi, and bacteria possess P450 enzymes (Käppeli, 1986; Urlacher et al., 2004; McLean et al., 2005), but their role in toxicant metabolism is limited. To better understand chemical toxicity to humans, toxicant activation or deactivation may be catalyzed in these experiments by adding S-9 human liver microsomes. Additionally, in unicellular organisms, cell lines, and less complex eukaryotes such as C. elegans, one is unable to examine a toxicant's target organs or systemic effects. The discovery of human disease models through orthologous phenotypes may address these concerns. For example, McGary et al. (2010) proposed a yeast model for angiogenesis defects and a C. elegans model for breast cancer, based upon analyses of genes shared between the model system and an organism displaying the disease under study. Similar analyses of functional data from unicellular organisms, cell lines, and C. elegans may reveal toxicant mechanisms associated with more complex biological processes.

In whole organisms, throughput is the major barrier to progress in functional toxicology. In yeasts and bacteria, barcoded systems allow for the high-throughput examination of thousands of deletion mutants in parallel (Giaever et al., 2002; Oh et al., 2010a). The lack of such methods in whole organisms such as zebrafish or rodent models diminishes throughput, and complicates systematic identification of genes required for resistance to a toxicant. Until high-throughput screens are devised in whole organisms, such systems are invaluable in extending or confirming results gathered in yeasts or bacteria.

Functional toxicological screening methods, i.e., those that identify genetic requirements for chemical tolerance, are powerful, unbiased tools which provide unique mechanistic insights in the field of toxicology. High-throughput screens of chemicals of concern or unknown toxicity will allow toxicologists to formulate hypotheses related to their corresponding mechanisms and pathways of toxicity. Automation and deep parallel sequencing technologies will unquestionably increase the throughput of functional techniques, but extensive computational resources and knowledge will be required to implement screening systems and analyze the resulting data. Integration of functional assays across a variety of organisms can identify conserved modes of toxicity and direct studies most relevant to human health (Figure 3). The field of functional toxicology is primed to assist toxicologists meet the need for enhanced chemical toxicity testing.