Identifying Endocrine Disrupting Chemicals Using ToxCast and Authoritative Lists


We are constantly awash in a sea of chemicals.

Though the exact number is unknown, nearly 85,000 industrial chemicals (not including pesticides, food additives, and pharmaceuticals) have been processed or manufactured in the United States as of July of 2014 (according to the most recent Toxic Substance Control Act Inventory List). While the number of new chemicals increases steadily, government evaluation of the toxicity of these chemicals has largely lagged behind. This has resulted in a large pool of chemicals, used in many consumer products, for which we have little or no toxicity information at all.

For programs like the California Safer Consumer Products (SCP) program, that aim to reduce the number of toxic chemicals we are exposed to through the products we use, identifying the highest priority chemicals ­-- out of a sea of thousands -- can be expedited and supplemented by multiple sources of information, including emerging computational and molecular biology tools.

The California SCP program at a glance

One of the goals of the SCP is to promote the use of safer alternatives to harmful chemicals in consumer products. The SCP identifies product-chemical combinations that have potential to cause harm in widespread populations, and provides an impetus for manufacturers to find non-chemical or safer chemical alternatives. Using authoritative lists like those from the World Health Organization's International Agency for Research on Cancer, the National Toxicology Program Office of Health Assessment and Translation, the EPA Integrated Risk Information System, and California Office of Environmental Health Hazard Assessment Proposition 65 list (a list of chemicals known by the state to cause cancer or reproductive/developmental toxicity), the SCP has compiled a candidate chemical list of over 2,000 substances with known human health or ecological hazard traits.

The SCP intends to select between five and ten product-chemical combinations, per year, for the first iteration of its work plan -- an assessment process that would take over 200 years to complete for all the chemicals on the list (not taking into account the exponential number of product-chemical combinations). Determining which chemicals should be evaluated first, therefore, is particularly important. One way the SCP chemical list could be prioritized is by combining information from authoritative lists with biological information generated by programs like EPA's Toxicity Forecaster (ToxCast).

What is ToxCast?

ToxCast is an EPA-led effort to rapidly evaluate the toxicity of thousands of chemicals using automated screening technologies (also known as high-throughput screens or HTS). The HTS technologies used by EPA-contracted laboratories analyze the effects of chemicals on molecular and cellular processes that are relevant to human and animal biology. ToxCast assays measure changes in protein conformation (the way that a protein is shaped), changes in DNA binding (a way that gene expression can be altered), and changes in cell behavior (for example, how quickly a cell divides, multiplies, or dies).

ToxCast has already evaluated nearly 2,000 chemicals across a wide range of uses (e.g., food additives, pesticides, industrial chemicals, and pharmaceuticals) since it began in 2007, and has generated a vast amount of data that the public and other researchers can access. ToxCast assays have several notable limitations, however.

First, the current suite of assays in ToxCast provides only partial coverage of the many ways that a hazardous agent can cause adverse health impacts. For example, although there are nearly 30 assays for estrogen and androgen receptor binding in the ToxCast platform, the platform only contains four assays for the thyroid receptor (another important target for hormones and hormone-like chemicals in the body). Without full coverage of multiple pathways, and data on how chemicals can alter those paths in ways that don't involve receptors (e.g., changes to proper thyroid functioning can occur by blocking the receptor or by a multitude of other mechanisms), the ToxCast program could miss toxic effects of the chemicals being tested.

Second, ToxCast assays are extremely limited in their ability to determine the effects of metabolism on chemical toxicity. In other words, ToxCast assays are not able to measure how a chemical might change in toxicity (i.e., become more or less toxic) as it travels through, and is processed by, our bodies.

Third, ToxCast assays are limited in their ability to analyze the full spectrum of chemical space. For example, chemicals that are not soluble in the solvent dimethyl sulfoxide (DMSO), that are not available in sufficient quality and quantity, or have characteristics that make high-throughput analysis more difficult (e.g., high-volatility, heavy metals), are currently outside of the chemicals tested in government-led HTS programs.

Even with these significant limitations, however, the ToxCast platform could provide useful information to aide in identifying chemicals on the SCP candidate chemical list that could be prioritized for replacement.

Finding the SCP chemicals that have been tested by ToxCast

ToxCast data collection has occurred in phases, with the most recent evaluation (Phase II) containing slightly over 1,800 chemicals. Comparing the SCP candidate chemical list to the ToxCast Phase II list, there are several notable findings. First, less than 10% of the chemicals in the SCP database were in the ToxCast Phase II list (Figure 1). This means that while HT technologies can be useful for prioritizing some chemicals, this data stream is not available for all prioritization decisions. Decision-makers must therefore rely on other sources of information, like authoritative lists, animal tests, and other existing toxicity data, when finalizing a priority scheme.

Credit: Limaye et al. (2022).

Second, of the 21 hazard traits identified by SCP using authoritative lists (e.g., endocrine toxicity, reproductive toxicity, neurotoxicity, and carcinogenicity) all but neurotoxicity was represented in the overlap group. In other words, though a large portion of the SCP chemicals were not in the ToxCast database, a wide range of toxicity endpoints evaluated by SCP were in the overlap group. The abundance of multiple hazard traits within the dataset can provide a way for researchers and decision-makers to evaluate the ability of ToxCast information to support existing hazard information. The absence of hazard traits, like neurotoxicity, reinforces the usefulness of multiple streams of data in determining chemical toxicity.

Finally, more than half (54%) of the ToxCast overlap chemicals were designated as possessing more than one authoritative list-based hazard trait, with over 20 percent of chemicals having more than four authoritative list hazard traits, meaning that many chemicals can act on more than one system within a body or the environment. Three chemicals on the list (naphthalene, di(2-ethylhexyl)phthalate (DEHP), and dibutyl phthalate (DBP)) had hazard traits in ten or more categories. The most common hazard traits for the overlap chemicals were carcinogenicity, reproductive toxicity, respiratory toxicity, and developmental toxicity.

Using ToxCast to identify SCP chemicals with endocrine disrupting properties.

There are multiple hazard traits upon which the SCP/ToxCast overlap chemicals could be ranked and prioritized -- one trait of particular importance is endocrine toxicity. The endocrine system regulates the finely tuned actions of hormones within the body. Hormones are necessary for different processes at different stages of life, including brain and organ development early in life, and reproduction and cellular responses to stress in adult life. Even small changes in the low levels of hormones in our bodies can disrupt the delicately balanced endocrine system, and can result in negative health consequences such as cancer, diabetes, obesity, and reproductive harm. Disruption that occurs during particularly vulnerable life stages (such as during fetal development, infanthood, childhood, and adolescence) can cause harm years, and even generations, later. Because low doses can result in significant changes over the entirety of the life course, endocrine disruption presents a significant public health concern.

Endocrine disrupting chemicals (EDCs) interfere with the normal functioning of the endocrine system in multiple ways. They can bind directly to nuclear hormone receptors, like the estrogen and androgen receptors, and either block or add to the normal response of an endogenous (made in our bodies) chemical. They can also bind to receptors outside of the nucleus and impact the ways that signals are sent through the cell. Additionally, EDCs can block the ways that hormones move and/or are stored in the body, or change the ways in which hormones are made. The endocrine system, along with the way that it can be disrupted, is extraordinarily complex, and is not easily evaluated with reductionist techniques (for additional details on tiered methods of evaluating EDCs, see the Tiered Protocol for Endocrine Disruption website). While cell- and protein-based methods like those employed by ToxCast have the potential to provide useful information on chemical toxicity, they do not reveal the entire endocrine disruption story, and can overlook harmful EDCs that act by pathways not covered in the ToxCast assays.

According to EPA's latest analysis, the ToxCast dataset has 18 assays for the estrogen receptor and nine assays for the androgen receptor that have the potential to give researchers clues about how these pathways work in whole animals (for more details on how the ToxCast assays can predict whole animal outcomes, see Rotroff, et al and Cox, et al).

Within the SCP/ToxCast overlap group, there are 32 chemicals with endocrine toxicity as one of their authoritative list hazard traits and 151 chemicals with activity in either the androgen or estrogen pathways. There are four chemicals (4-tert-octylphenol, bisphenol A, bisphenol B, and phenolphthalein) that turn on (also known as agonists) each of the six measured steps in the EPA estrogen agonist pathway (Figure 2) and three (butylbenzyl phthalate, butylparaben, and n-propylparaben) that turn on all but one piece in the measured agonist pathway.


While chemicals that turn on the estrogen pathway are more important for some health impacts (for example, in estrogen responsive cancers), chemicals that turn off (also known as antagonists) the androgen pathway can be particularly harmful to ecosystem and animal health (antiandrogens can cause harm to reproductive organs and cause delays during fetal and pubertal development). In the SCP/ToxCast overlap, there are five chemicals (3,3'-dimethoxybenzidine dihydrochloride, bisphenol A, bisphenol B, tris(1,3-dichloro-2-propyl) phosphate (TDCPP), and tris(2,3-dibromopropyl) phosphate) that turn on each of the three measured parts of the EPA androgen agonist pathway (Figure 3), and five chemicals (4-tert-octylphenol, C.I. basic violet 3, dichlorophene, tetrabromobisphenol A, and triclosan) that turn on all but one piece in the measured antagonist pathway.

Credit: Limaye et al., 2022
When comparing the 14 chemicals that turn on all or part of the estrogen receptor agonist or androgen receptor antagonist pathways to the authoritative list hazard traits, a striking observation emerges -- the lists do not always agree (Figure 4). Though there are several instances where the authoritative lists and the ToxCast data agree, there are four chemicals (3,3'-dimethoxybenzidine dihydrochloride, phenolphthalein, tris(1,3-dichloro-2-propyl) phosphate (TDCPP), and tris(2,3-dibromopropyl) phosphate) with strong activity (active in each of part of their respective pathway) that are not labeled as endocrine toxins on any of the authoritative lists used by SCP. Alternately, there are chemicals on the authoritative endocrine toxin lists that might not be considered "strong" (e.g., not fulfilling every piece of a particular pathway) in the ToxCast data set. If one were to rely on either ToxCast data or authoritative lists alone, it is possible that some potentially toxic chemicals could be missed.


At this time, the ToxCast program shows some promise in advancing chemical toxicity testing (e.g., testing of large numbers of chemicals in a short time, providing additional streams of evidence for decision makers) but it still has a number of significant pitfalls (e.g., the lack of metabolism, not all relevant pathways are covered, no information on duration of critical windows of exposure) that reduce its ability to correctly identify EDCs and other hazardous chemicals via high-throughput testing. Even with these limitations, however, ToxCast and other high-throughput platforms can provide useful information for prioritizing chemicals for evaluation in programs like SCP.

Decision makers should use caution in interpreting these data, however, particularly when trying to identify chemicals that are not likely to be EDCs. With the increasing ability of EPA and other state and federal agencies to screen thousands of chemicals quickly and relatively cheaply, chemicals in commerce should be regularly and routinely re-evaluated for toxicity, especially given the evolving nature of emerging HTS technologies. An inactive chemical in the assays of today could be an active chemical in the assays of tomorrow, and our failure to correctly identify EDCs in our air, water, food, and consumer products is not simply a matter of scientific uncertainty, but a failure of public trust and risk to public health.