|Home | About | Journals | Submit | Contact Us | Français|
The practice of toxicology is changing rapidly, as demonstrated by the response to the 2007 NRC report on “Toxicity Testing in the 21st Century.” New assays are being developed to replace animal testing; yet the use of data from these assays in decision making is not clear. A Health and Environmental Sciences Institute committee held a May 2011 workshop to discuss approaches to identifying adverse effects in the context of the NRC report. Scientists from industry, government, academia, and NGOs discussed two case studies and explored how information from new, high data content assays developed for screening can be used to differentiate adverse effects from adaptive responses. The terms “adverse effect” and “adaptive response” were defined, as well as two new terms, the relevant pathways of toxicological concern (RPTCs) and relevant responses for regulation (RRRs). RPTCs are biochemical pathways associated with adverse events and need to be elucidated before they are used in regulatory decision making. RRRs are endpoints that are the basis for risk assessment and may or may not be at the level of pathways. Workshop participants discussed the criteria for determining whether, at the RPTC level, an effect is potentially adverse or potentially indicative of adaptability, and how the use of prototypical, data-rich compounds could lead to a greater understanding of RPTCs and their use as RRRs. Also discussed was the use of RPTCs in a weight-of-evidence approach to risk assessment. Inclusion of data at this level could decrease uncertainty in risk assessments but will require the use of detailed dosimetry and consideration of exposure context and the time and dose continuum to yield scientifically based decisions. The results of this project point to the need for an extensive effort to characterize RPTCs and their use in risk assessment to make the vision of the 2007 NRC report a reality.
The science supporting regulatory toxicology is undergoing a transformation that will change how toxicology testing, interpretation, and use of data in decision making and public health protection will be performed in the decades ahead. The developing advanced technologies and high-throughput approaches for toxicity testing will reduce animal use, permit legislative action around broader chemical testing needs leading to reform of the U.S. Toxic Substance Control Act (TSCA), and improve efficiency of drug and chemical development. The new testing approaches and techniques are designed to identify markers or endpoints that are a departure from apical endpoints associated with traditional toxicology testing (e.g., cancer, reproductive effects). However, these new early-stage endpoints present a challenge for biological interpretation as well as integration into current risk assessment practices and regulatory decision making. A series of forum articles was published in Toxicological Sciences in 2009–2010 outlining the challenges and potential solutions to the vision outlined in the 2007 NRC report (Andersen and Krewski, 2009, 2010; Boekelheide and Campion, 2010; Bus and Becker, 2009; Chapin and Stedman, 2009; Cohen Hubal, 2009; Hartung, 2009; MacDonald and Robertson, 2009; Meek and Doull, 2009; Walker and Bucher, 2009). These articles were useful for setting the stage and identifying many of the issues surrounding full implementation of the goal of the NRC report. The NRC (2007) identified one issue in particular as critical—the need to determine what makes an effect “adverse.”
Existing toxicological test designs are based on the identification of an adverse effect at a given dose, which can be used to define a point of departure for subsequent assessment of risk and regulatory decision making. The rapidly expanding availability of in vitro predictive tools and technologies will spawn an increasing number of endpoints and potential effects as well as a complex web of biological pathways that will need to be considered as viable markers for safety and risk assessment. An integrative analysis approach will identify new markers of toxicity that will need to be fully characterized for dose-response, their relation to in vivo physiological systems, and their relevance to humans before they can be used appropriately.
Differentiation between an adverse effect and an adaptive response is central to toxicology and is a critical determination in the context of these new toxicity-testing approaches. In anticipation of the need for rigorous scientific input into how new endpoints and markers of biological change may be incorporated into risk assessment, the International Life Sciences Institute (ILSI) Health and Environmental Sciences Institute (HESI) formed a committee in 2009. The committee consisted of scientists from the chemical and pharmaceutical industries, U.S. government agencies, and academic institutions (http://www.hesiglobal.org/i4a/pages/index.cfm?pageid=3440). This group first defined the terms adverse and adaptive and then examined the ways in which these terms could be used when high data content information is the sole input. Specific goals included the following: (1) develop an approach to evaluate effects from new in vitro toxicity-testing tools for integration into the safety assessment of chemicals; (2) develop criteria to assist in differentiating adverse effects from other types of biological changes; and (3) review and revise the definitions of adverse and adaptive effects based on toxicological and biological considerations relevant to regulatory decision making.
The HESI committee convened a workshop titled “Distinguishing Adverse from Adaptive Effects in the 21st Century” on 10–11 May 2011 at the U.S. Environmental Protection Agency (USEPA) facilities in Research Triangle Park, NC. Workshop participants (Supplementary table 1) discussed, together and in small groups, the characterization of biological responses; integration of responses within a biological pathway; interpretation of different categories of data for safety assessment; and the potential development of a framework that recognizes, prioritizes, and uses all toxicity testing approaches and data in safety assessment. Two case studies, dimethylarsinic acid (DMA) and acetaminophen, were reviewed and discussed in the context of how the data from in vitro studies, particularly toxicogenomics and the pathways implicated from those data, related to known outcomes for apical endpoints. Proposals were made for future areas of research.
The following provides an overview of committee discussions prior to and during the workshop. These discussions focused on the use of data and high data content information from in vitro studies to inform decisions about adversity and their potential to inform risk assessment and advance regulatory approaches for protection of public health.
From the time of Paracelsus, the effects of chemicals on biological systems have been characterized by the apical response, which is the observable outcome in a whole organism such as a clinical sign or pathological state that is indicative of a disease resulting from exposure to a toxicant (NRC, 2007). This has been used in toxicology as relevant for designing and interpreting studies, comparing compounds, and determining appropriate human exposure limits. As biology continues to develop, computational approaches and in vitro studies including high-throughput assays will enable the assessment of alterations of pathways and networks that have been described at the gene, protein, or metabolic level of organization. The present challenge is to determine the value and appropriate use for these data in the context of risk assessment. The adverse effect drives regulatory decisions of chemicals under U.S. legislation, including TSCA and the Federal Insecticide, Fungicide, and Rodenticide Act, and in Europe under the Registration, Evaluation, Authorisation, and Restriction of Chemicals legislation. Also, many processes that inform risk-based decisions are founded on identifying an adverse effect, including benchmark dose calculations and doses allowed in pharmaceutical clinical trials.
Definitions of adverse effect can be found in many laws, regulations, and in the scientific literature (BfR, 2009; Boekelheide and Andersen, 2010; Dorato and Engelhardt, 2005; Goodman et al., 2010; JMPR, 2006; Lewis et al., 2002; NRC, 2007; USEPA, 2009). From this extensive literature, the HESI committee developed and reached consensus on definitions of adverse effect and adaptive response for the purpose of achieving its specific goals. These working definitions were agreed to in principle by the May 2011 workshop participants:
A number of research efforts (Dix et al., 2007; Schmidt, 2009; Suter et al., 2011) have the potential to change the way toxicity testing is conducted consistent with the NRC vision. The paradigm in which a mode of action (MOA) is determined for a single chemical, followed by development of predictive assays for that MOA, is shifting toward a new paradigm where screening assays will be performed that lead to prediction of an MOA for a compound with confirmation by targeted testing in well-defined assays (Fig. 1). Fulfillment of this paradigm requires sufficient understanding of the intermediate steps between the screening assays that indicate a molecular initiating event and other assays that characterize additional key events in an MOA that will lead to an adverse outcome.
To assess the impact of in vitro high data content information on the effectiveness of the working definitions of adverse effects and adaptive responses, the HESI committee used case studies of compounds for which there was good knowledge of MOAs along with detailed mechanistic studies available in the literature. These cases were reviewed in detail by the committee and then discussed at the workshop. For illustration purposes, an overview of the DMA case is highlighted here.
The discussion of DMA began with the assumption that systems biology, including homeostasis based on repair mechanisms and other adaptive responses, will play a major role in the application of new data types. In addition, computational modeling of these pathways and networks will be critical to formulate hypotheses, design targeted tests, and establish a collection of MOAs and adverse outcome pathways associated with toxicity. With this knowledge, risk assessment of drugs and chemicals will have a stronger scientific foundation and can be conducted with greater certainty and applicability to human biology. Although this may not be achievable in the short term (Crump et al., 2010), efforts such as the USEPA’s computational toxicology program and the explosion of bioinformatics tools indicate that the vision will eventually become reality.
The concept of the toxicity pathway is an important part of the NRC report (NRC, 2007). The pathway concept suggests that toxicity results when a chemical reaches and interacts with an initial key target, beginning a series of biological events that can ultimately result in the development of an adverse outcome. The toxicity pathway is a cellular response pathway that when sufficiently perturbed can result in an adverse effect (Boekelheide and Andersen, 2010; NRC, 2007). At the HESI workshop, participants referred to pathways associated with adverse effects as relevant pathways of toxicological concern (RPTCs). The number of RPTCs is unknown, although it is possible that a relatively small number may describe the majority of toxic responses. At low doses, changes in an RPTC may reflect adaptation, whereas at higher doses the changes may be adverse. A critical near-term need is to determine which pathways are RPTCs and describe them in a quantitative manner so that the changes in these pathways can be used to assess risk. Understanding the key nodes and the dose transition points in the RPTCs will be an important part of this learning. Boekelheide and Andersen (2010) provide an example of how understanding the RPTCs for carcinogenicity could change the way carcinogenicity assessments are conducted. The authors point out the need for understanding pathway dynamics and the role computational methods will play in developing these new approaches. Although some efforts toward defining specific RPTCs have been made, detailed knowledge of RPTCs for use in risk assessment is not yet available (Simmons et al., 2009; Stephens et al., 2011).
The identification of an adverse outcome after xenobiotic exposure has been a mainstay for assessing risk to inform risk management decisions. Adverse effects used for these decisions tend to be apical outcomes such as tumors, permanent changes in the target tissue, or specific transient changes in the target tissue directly associated with the ultimate outcome of concern. Another term discussed at the May 2011 HESI workshop was the relevant response for regulation (RRR), that is, denoting the endpoint, which is the basis for a risk assessment. In theory, the RPTC affected at the lowest internal dose could be considered to be the RRR, mitigating the need for in vivo studies. Clearly, for the RPTC to become the RRR, a great deal more must be known about the structure and dose-response of pathways as well as cross talk and redundancy between pathways.
As an example of the current state of the art of an MOA-based risk assessment using the most detailed biological information available, the committee considered the case of DMA. The USEPA established an MOA following the International Programme for Chemical Safety Mode of Action/Human Relevance Framework for DMA, a herbicide and inorganic arsenic metabolite (Boobis et al., 2006; USEPA, 2005). The basis for the DMA MOA was the evaluation of a series of apical endpoints as key events for the ultimate apical endpoint of transitional cell tumors of the urinary bladder in rats. These apical endpoints included transitional cell death, proliferation, and hyperplasia. Experiments characterized the effects in the target cell using transcriptional profiling (Sen et al., 2005, 2007). These studies were designed to identify key molecular changes associated with exposure to DMA and to better characterize the pathways or key events leading to the various apical outcomes. In brief, the authors identified transcriptomic changes at doses below which one sees the apical adverse endpoint of transitional cell death in the target epithelium from exposed rats as well as in vitro (Sen et al., 2005, 2007). The authors suggested that the toxicity observed at the higher doses may add to precursor effects present at the lower doses to drive the development of the apical outcome of a tumor (Sen et al., 2005). However, although the changes in gene expression indicated potential RPTCs, there was neither enough knowledge of these pathways to quantitatively describe the key events in development of the tumor nor were there sufficient species-specific descriptions of these pathways that could determine if the rat and human would have different or similar responses. Therefore, the response did not qualify as an RRR.
In the context of the discussion on adaptation and adversity, it may be that the cellular response to DMA sets up a series of activated genes, which allow the transitional cells to survive insults at the lower doses. The adaptive response may impart protection and/or enhanced susceptibility to the toxic effects. As one increases the dose of DMA to the target cell, increasingly severe responses occur, from urothelial toxicity to hyperplasia and ultimately transitional cell tumors. At the lowest doses tested in these studies, only altered gene expression was identified (Sen et al., 2005, 2007). At intermediate doses, the apical endpoint of cell death and increased cell proliferation occurred but resolved over time, suggesting an additional adaptive response to the continued exposure (Sen et al., 2005; USEPA, 2005). After extended treatment with the highest doses of DMA, irreversible tissue responses occurred resulting in the apical endpoints of cellular hyperplasia or tumors (USEPA, 2005). This example illustrates the combined significance of context, amount of exposure, and duration of exposure. The biological significance of various exposure-related effects and the determination of whether they were adverse depended on establishing a relationship among the several key events described for this MOA. Figure 2 illustrates the hypothetical dose-response relationship for putative RPTCs and how this information might be used to determine adverse effect levels. This example also illustrates the need to gain a more detailed, quantitative knowledge of molecular initiating events, toxicity pathways, and their interactions to improve the understanding of where transition points occur between adaptive changes and adverse effects. This will aid species extrapolation of effects and decrease uncertainty in risk assessments, potentially reducing reliance on uncertainty factors. A similar case can be made for data on hepatotoxicity of acetaminophen (Bushel et al., 2007; Fannin et al., 2010; Heinloth et al., 2004; Powell et al., 2006); however, this example is not described here.
At the May 2011 workshop, participants met in breakout groups to discuss issues that the HESI committee had been deliberating for the previous two years, including the following four topics:
In addition to characterizing new endpoints that emanate from emerging tools and testing approaches, interpreting these changes demands attention and continued discussion. The data must be interpreted with identification of the most relevant responses considered to be early markers of exposures leading to an adverse effect. An adverse response cannot be ascertained from a single observation. Given the upstream nature of many reported genomic and other endpoints, it is important to determine where such changes lie along the continuum of biological response and how the changes are connected to other levels of biological organization. Additionally, one would need to characterize the normal background and variability of responses of the markers and whether these markers of toxicity or exposure can be appropriately extrapolated to humans. If one assumes that RPTCs are involved in adversity, then for each pathway one would need to determine the dose and time dependence of the response and the perturbation of the pathway in the context of an affected network. The critical nodes of the network would need to be identified as well as the correspondence between affected network nodes and organ function and physiology. Some HESI workshop participants suggested that there are likely to be a finite number of MOAs that need to be investigated for the associated RPTC. These MOAs would include receptor-mediated modes, inflammation, cytotoxicity, and genotoxicity. Some key pathways involved are likely to be stress responses (Simmons et al., 2009) and nuclear receptor–mediated pathways. Although the concept is feasible, pathways have not been sufficiently elucidated for many MOAs already known to lead to an adverse outcome, such as genetic alterations leading to cancer (Boekelheide and Andersen, 2010).
The first step toward understanding adversity at the pathway level is to extensively characterize the RPTCs for a few prototypical, data-rich chemicals. As experience is gained with chemicals with a well-defined MOA, studies could be performed prospectively with novel compounds or unknowns to determine whether investigation of the RPTC was sufficient to predict adverse outcomes of apical endpoints. Ultimately, the intermediate steps could be eliminated as additional important RPTCs are identified, and confidence is gained in the validity of these predictions. A significant investment in research describing toxicity pathways and adverse outcome pathways, with identification of biomarkers associated with adversity, is needed to achieve this goal. Once RPTCs are sufficiently characterized and shown to be predictive of apical endpoints, chemicals could be tested in vitro on the key nodes of RPTCs to determine the potential for causing adverse effects without the need for animal testing.
Our present understanding of adversity is linked to the apical effect, which is typically a phenotypic response. Therefore, phenotypic anchoring of changes in gene or protein expression, or other in vitro endpoints, is critical to understanding if changes in a system are adverse or not. As more experience is gained with RPTCs, key transition points in pathways, and other details of biology, the need for phenotypic anchoring with a specific chemical should decrease. How long this paradigm switch takes will be dependent on the quality of data produced to support hypotheses. Studies will need to be designed for this specific purpose.
With the development of new toxicity-testing approaches and identification of biomarkers and signatures of potential adverse effects comes the need for a new framework for how these data are used in risk assessment to inform regulatory decision making. Iteration, revision, and rigorous validation will be needed to ensure that new tools and endpoints are sensitive and accurate, and these new approaches must be accepted by the scientific and regulatory communities as reflective of human biology and relevant for risk assessment. Although all data from screening assays, mechanistic studies, pathway analyses, and other in vitro and in vivo methods should be considered, a weight-of-evidence (WoE) approach allows for an analysis of the strengths and weaknesses of the data and the relative importance for animals or humans. At the present time, a WoE evaluation requires both classical in vivo toxicology and in vitro data and relies on animal and human (if available) data. As RPTCs are characterized and better understood, the data used could potentially shift toward pathway analyses, with a decreased reliance on in vivo data.
A WoE evaluation for risk assessment also needs to consider variables such as exposure, which will require the use of dosimetry, reverse dosimetry, and biomonitoring. This issue is beyond the scope of this paper but is an important element of the risk assessment process. Consideration of proper context needs to be incorporated when designing studies, as well as during evaluation, to properly interpret the data for informing regulatory decisions. For example, age-related biological differences can impact interpretation of results from exposure to a chemical at various life stages (in utero or early life exposures or for juvenile and adult later life exposures). Toxicological responses resulting from exposure across life stages must be considered along with the time- and dose-response continuum.
Adaptive responses to toxicant exposure may be characterized by reversibility (upon withdrawal of treatment or exposure). Furthermore, adaptive changes are often early homeostatic adjustments, such as metabolism or gene expression/transcriptomic changes (Goetz et al., fothcoming). These modulations are typically not considered to be precursors of functional impairment but rather a response that would return to a homeostatic condition (e.g., a return from hormone level variation(s)/cycling or blood pressure increase owing to stress). In some situations, a minor change may be sustained resulting in a “new normal” state where the cell/tissue/organism has adapted without adverse consequences, such as an induction of cytochrome P450 enzymes resulting in hepatocellular hypertrophy and increased liver weight. These adaptive changes, in a different context, may be indicators of a potentially adverse outcome. For example, short-term decrements of circulating thyroid hormone may result in an adaptive response in an adult, nonpregnant female, but the same change in early gestation could be an indicator of a potential adverse outcome on fetal brain development.
During the HESI workshop, participants discussed the usefulness of a formalized framework or a consistent series of questions to be answered for deciding if effects are adverse or not. This framework was loosely based on previously published decision trees (Dorato and Engelhardt, 2005; Lewis et al., 2002) and considered factors such as change in tissue or cellular function, reversibility, transition points in pathways, context of exposure, and species differences. Such an approach to interrogating data could prove useful both in the design of additional studies and the assessment of potential risk. It could substitute for currently used tiered-testing schemes that are designed to cover all possibilities rather than to develop targeted perspectives based on knowledge of MOAs of chemicals. Ultimately, the workshop participants agreed that although such an approach for interrogating data would be useful, there is currently not sufficient knowledge to establish a specific framework for decision making. As more data become available on the key RPTCs and critical nodes in these pathways, the development of a framework may become feasible, thus leading to better characterization of the RRR for a specific chemical and exposure scenario.
The major conclusions of the May 2011 HESI workshop are summarized here:
The HESI workshop was held in response to the development of screening assays and high data content information that are produced more rapidly than procedures established for risk assessment. At the present time, the most appropriate use of high data content information is for prioritizing chemicals for additional evaluation and is not yet directly applicable for determining a specific MOA. However, as the science continues to progress, there will be opportunities to use these new methods and types of data to inform risk assessment. It is hoped that the present effort has helped to focus the scientific community’s attention on this important area of research. Understanding the spectrum of adaptation and adversity as it applies to risk assessment will ultimately inform regulatory decisions.
ILSI Health and Environmental Sciences Institute; the National Institute of Environmental Health Sciences of the National Institutes of Health.
The authors gratefully acknowledge the U.S. EPA for providing facilities for the May 2011 workshop. The authors also acknowledge Dr Miriam Sander (Page One Editorial Services) for providing valuable assistance in summarizing workshop discussions and conclusions. This paper does not represent the policies or opinions of the USEPA. The views expressed in this paper are those of the authors and do not necessarily reflect the views of participants in the 10–11 May 2011 workshop.