Growing concerns about environmental contaminants associated with adverse developmental, reproductive, neurological, and immune effects led Congress more than 20 years ago to direct the U.S. Environmental Protection Agency to develop a program to screen chemicals for their potential to mimic estrogen in humans. EPA decided to go beyond the mandate and evaluate the impacts of chemicals on androgen and thyroid hormone systems and to look at effects in wildlife as well as humans.
EPA’s resulting Endocrine Disruptor Screening Program (EDSP) got off to a slow start and sputtered for nearly two decades at a cost of about $10 million annually. To date, only a few dozen pesticides have gone through the battery of five in vitro and six in vivo assays to screen for potential endocrine activity under the program. Eighteen of those pesticides were flagged for further testing.
By 2015, it was obvious to EPA officials that the agency’s existing approach would require decades to evaluate all the chemicals EPA must examine for potential endocrine disruption. The list of chemicals targeted by EDSP includes about 1,000 active ingredients in pesticides, 5,000 inert ingredients in pesticides, and 6,000 drinking water contaminants, with some overlap between the lists of pesticides and drinking water contaminants.
EPA needed better ways to quickly identify which chemicals require more scrutiny for endocrine disruption and weed out the majority that don’t. At that point, EPA scientists had already developed a high-throughput computational model for predicting effects of chemicals on estrogen receptor (ER) bioactivity. But such models for predicting effects on the androgen receptor (AR), thyroid, and steroid hormone synthesis pathways were not as far along. EPA decided to put EDSP on hold while EPA researchers ramped up efforts to develop high-throughput methods to replace the 11 EDSP assays. C&EN visited EPA’s Office of Research & Development in Research Triangle Park, N.C., earlier this year to find out how much progress they have made.
“The team is way ahead of schedule,” says Ron Hines, associate director for health at EPA’s National Health & Environmental Effects Research Lab. Already in place are an ER model that replaces two in vitro assays for ER binding and activation of ER transcription and an in vivo uterotrophic test that examines estrogenic activity by measuring increases in uterine weight in female rats (Environ. Sci. Technol. 2015, DOI: 10.1021/acs.est.5b02641). An AR model that accurately predicts androgenic and antiandrogenic activity caused by chemicals blocking or activating the androgen receptor is also ready for prime time (Chem. Res. Toxicol. 2016, DOI: 10.1021/acs.chemrestox.6b00347).
They are now working to fill in gaps related to chemical effects on androgen synthesis and cell proliferation. EPA researchers have also developed a high-throughput steroidogenesis model that integrates 11 hormone measurements into one meaningful number that regulators can use to rank chemicals on the basis of their potential to disrupt steroid hormone synthesis (Toxicol. Sci. 2017, DOI: 10.1093/toxsci/kfx274).
Additionally, EPA scientists are testing three new thyroid assays targeting enzymes and other proteins critical for thyroid hormone production and the uptake and recycling of iodide. They are working to integrate the thyroid assays into a “thyrollicle” model to better predict the effects of chemicals on thyroid cells, which are arranged in spheres called follicles. The agency expects to have all the methods ready to begin screening again by 2020.
Estrogen receptor model paves the way
EPA scientists developed the ER model first because they had a lot of information about the pathway. “We knew every single step in ER signaling. It had been worked out in the 1980s and ’70s,” Hines says. They also knew that some environmental contaminants hit targets along the path. That made it feasible for EPA to develop an approach that hit all the key points along the ER pathway. Exposure to estrogenic chemicals has been associated with health effects such as reduced fertility and increased incidences of obesity, diabetes, endometriosis, and some cancers.
Another reason why EPA started with estrogen was that high-throughput in vitro assays were already developed either commercially for use in drug discovery by the pharmaceutical industry or through an interagency government effort called Toxicology Testing in the 21st Century, or Tox21. The researchers used 18 of those assays to screen more than 1,800 chemicals—including pesticides, food-use chemicals, and substances with known or suspected endocrine activity. In addition to receptor binding, the assays measure key events along the ER adverse-outcome pathway, such as receptor dimerization, DNA binding, transactivation, gene expression, and cell proliferation. The scientists then integrated the results into a computational model that predicts the effects of chemicals on the ER pathway.
Status of EPA efforts to replace 11 Tier 1 assays with high-throughput methods under the Endocrine Disruptor Screening Program.
|TARGETED END POINTS|
|ESTROGEN RECEPTOR||ANDROGEN RECEPTOR||STEROIDOGENESIS||HYPOTHALAMIC-PITUITARY-GONADAL AXIS||HYPOTHALAMIC PITUITARY-THYROID AXIS|
|Tier 1 in vitro assays (cell line)|
|Androgen receptor binding (rat prostate cytosol)||X|
|Aromatase inhibition (human recombinant)||X|
|Estrogen receptor binding (rat uterine cytosol)||X|
|Estrogen receptor transcriptional activation
(human cell line HeLa-9903)
|Steroidogenesis (human cell line H295R) X||X|
|Tier 1 in vivo animal assays (animal model)|
|Amphibian metamorphosis (frog) X||X|
|Fish short-term reproduction (medaka)||X||X||X|
|Hershberger bioassay (rat)||X||X|
|Pubertal development and thyroid function
(intact juvenile/peripubertal female rats)
|Pubertal development and thyroid function
(intact juvenile/peripubertal male rats)
|Uterotrophic assay (rat)||X|
To evaluate whether the model accurately flagged chemicals in need of further evaluation for ER signaling disruption, EPA sought help from scientists at the NTP Interagency Center for the Evaluation of Alternative Toxicological Methods (NICEATM), housed within the National Toxicology Program at the National Institute of Environmental Health Sciences.
“We combed the literature for every possible uterotrophic study,” says Nicole Kleinstreuer, deputy director of NICEATM. Over two years, the team reviewed 700 papers: Two scientists independently read each paper and extracted all the information, such as protocol details, chemicals tested, and results. They determined that only about 15% of the papers met quality criteria based on guidelines developed by EPA and the international Organisation for Economic Co-operation & Development (OECD).
Within that 15%, however, there was “still a tremendous amount of variability in the results,” Kleinstreuer notes. NICEATM scientists further narrowed the pool of studies to those that had reproducible results. The resulting set of data for about 40 chemicals correlated well with results of the ER model, Kleinstreuer says. The ER model results also correlated well with the screening results of more than 50 pesticides that were subject to the original EDSP screening assays, including the in vivo uterotrophic rodent assay.
EPA researchers have since developed a streamlined version of the ER model that uses just four assays at each point along the receptor activation pathway. That model is more efficient and “probably more amenable to use outside of EPA,” says Katie Paul Friedman, a researcher at EPA’s National Center for Computational Toxicology (NCCT). When EPA starts ordering EDSP testing again, companies will be able to use whatever assays they want as long as they hit each part of the ER pathway.
Filling androgen data gaps
After their success with the ER model, EPA and NICEATM scientists turned to developing an AR model.
Most of the chemicals that interfere with the AR pathway are antagonists, or antiandrogenic—they block the androgen receptor and inhibit gene transcription controlled by the receptor. Such chemicals have been associated with reproductive and developmental effects, such as nondescended testes in young males and prostate cancer in men.
The team’s latest AR model integrates the results of 11 high-throughput in vitro assays, each targeting a different part of the AR pathway. Just as they did in developing the ER model, the researchers screened more than 1,800 chemicals through the 11 assays. They also sought to evaluate the model against reference chemicals from the literature. The NICEATM group is working with OECD and EPA to develop a comparison literature database for the AR model similar to the one they developed for evaluating the ER model.
The AR model does a good job of predicting whether chemicals activate or block the androgen receptor. But it is not yet ready to fully replace one of the original 11 screening assays under EDSP, the Hershberger assay, Kleinstreuer says. That test investigates androgenic or antiandrogenic activity from synthesis to cell proliferation using castrated, immature male rats.
The Hershberger assay is more complicated than the uterotrophic assay, the in vivo test for estrogenic activity in rats, Kleinstreuer says. The uterotrophic assay has only one end point—increase in uterine weight. The Hershberger assay involves measuring changes in weight in five androgen-dependent tissues of the male reproductive system.
In the long term, to fully replace the Hershberger rat assay, EPA researchers are developing an in vitro assay that simulates the effects of chemicals on the most sensitive of those five tissues, prostate tissue. The test involves aggregating human prostate cancer cells into microtissues in 96-well plates and monitoring their growth over 90 days. Images are collected every couple of days using an automated, high-content imaging fluorescence microscope to measure the growth of the tissue cultures.
Although it is not high throughput, the approach decreases the number of animals needed for testing while still capturing all the key events along the pathway from androgen steroid synthesis to cellular and tissue-level effects. It is also likely to be more relevant to humans than the Hershberger rat assay.
Steroidogenesis assay takes shape
Although most tests for endocrine disruption have evaluated the effects of chemicals on hormone receptor signaling, disruption of hormone synthesis can also lead to adverse reproductive and developmental effects. “We realized that only looking at the receptor and its associated signaling pathways within the cell was not going to give us all of the answers,” says Kevin Crofton, NCCT’s deputy director. “Steroidogenesis is the focus right now.”
The process of creating steroid hormones, or steroidogenesis, starts with cholesterol. “Through a series of enzymatic steps you make steroid hormones that are physiologically relevant for sexual differentiation, reproduction, metabolism,” and other critical processes, NCCT scientist Paul Friedman says.
Chemicals can perturb any step in that complex process by inhibiting various enzymes. The original EDSP steroidogenesis assay was a low-throughput in vitro method that focused on the end of the steroid synthesis pathway—the final formation of testosterone and 17β-estradiol. But “if you inhibited an enzyme at the beginning of the pathway, you could affect formation of all the steroid hormones,” says Paul Friedman.
EPA researchers have now developed a high-throughput in vitro approach to examine chemical effects on the production of 11 hormones, including progestogens, corticosteroids, androgens, and estrogens. They used the assay to screen more than 2,000 chemicals at a single high concentration, measuring the production of all 11 hormones. Of those chemicals, 656 showed potential effects on hormone levels and were further tested at multiple concentrations.
The researchers evaluated the accuracy of the high-throughput assay by comparing the results for 25 chemicals that were run through both the original EDSP low-throughput assay and the high-throughput assay. The results from the two methods correlated well.
Asking regulators to make a quick decision based on 11 different graphs for each of thousands of chemicals would be a high hurdle, so NCCT researchers developed a quantitative prioritization metric that integrates effects on the synthesis of the 11 steroid hormones into a single number. Regulators can use that number to rank chemicals on the basis of their likelihood to interfere with the synthesis of all 11 hormones.
One of the original EDSP assays not yet covered by a high-throughput method measures inhibition of aromatase, which converts androgens to estrogens to make 17β-estradiol and estrone. Instead of developing a new assay, EPA researchers are trying to capture aromatase inhibitors in the steroidogenesis assay.
The researchers are also looking to fill in gaps that existed in the original EDSP screening assays. “We know we don’t capture all aspects of androgen synthesis in our model,” Paul Friedman acknowledges. For example, EPA scientists are working to develop assays that measure production of dihydrotestosterone, a potent androgenic metabolite of testosterone. Some chemicals are known to inhibit the enzyme that converts testosterone to dihydrotestosterone.
Trio of thyroid tests
Three in vivo animal assays initially required for EDSP screening examine the effects of chemicals on thyroid function in pubertal male rats, pubertal female rats, and amphibians undergoing metamorphosis. To replace those assays, EPA researchers are developing a series of high-throughput in vitro assays targeting three key steps in the production and homeostasis of thyroid hormones.
“Thyroid hormones are essential in processes of normal development, such as brain development in humans and metamorphosis of tadpoles to adult frogs,” says Michael Hornung, branch chief of EPA’s National Health & Environmental Effects Research Lab. Chemicals that affect thyroid processes have been associated with developmental effects on the nervous system in children, attention deficit and hyperactivity in children, hearing loss, and thyroid cancer.
When EPA researchers first began to investigate how chemicals interact with the thyroid hormone system, not many high-throughput in vitro assays were available. “There were a few assays that looked at thyroid hormone transcription via the thyroid hormone receptor,” Hornung says. But very few chemicals interact with the thyroid receptor, he notes.
Evaluating effects on the thyroid system requires a broader screening approach. The thyroid network is complex, involving multiple targets crossing many tissues in the “thyroid axis,” which includes the hypothalamus, pituitary gland, and thyroid gland. That complexity makes it particularly challenging to develop a high-throughput approach for predicting effects of chemicals on thyroid function.
EPA’s research team came up with a list of 17 potential targets where chemicals could interfere with thyroid hormone synthesis and signaling. They deemed three of the targets critical and began developing in vitro assays to screen chemicals for their ability to interfere with them. The assays measure inhibition of thyroperoxidase (TPO), an enzyme that is critical for thyroid hormone production; inhibition of iodide uptake by the sodium iodide symporter (NIS); and inhibition of deiodinase enzymes, which control thyroid hormone activation and metabolism.
EPA researchers focused on these targets, especially NIS and TPO, in part because they already had information showing environmental pollutants hit them, Crofton says. The classic example is perchlorate, which inhibits iodide uptake by NIS. Separately, some pesticides inhibit TPO, he adds.
EPA’s TPO assay uses human thyroid cells in high-throughput 96-well or 384-well plates. When the assay was used to test more than 1,800 chemicals at high concentrations, most of the chemicals—74%—did not inhibit TPO, “but 26% did produce greater than 20% inhibition,” suggesting they might be TPO inhibitors, Hornung says (Toxicol. Sci. 2016, DOI: 10.1093/toxsci/kfw034). EPA’s NIS assay measures the ability of a chemical to block the uptake of radioactive iodide by thyroid cells (Toxicol. In Vitro 2016, DOI: 10.1016/j.tiv.2016.12.006). Of 1,087 chemicals screened at high concentrations using this assay, 404, or 37%, led to greater than 20% inhibition.
The third thyroid assay being tested by EPA researchers targets inhibition of deiodinases—selenium-containing enzymes that catalyze the release of iodide, or deiodination, from thyroid hormones. Although iodine is necessary for initial thyroid hormone synthesis, deiodination is necessary to convert thyroxine (T4) into triiodothyronine (T3), the hormone with the highest affinity for the thyroid receptor. There are three deiodinase isoforms, each with different roles in producing T3.
EPA’s assay uses cell-derived deiodinase enzymes and measures a chemical’s ability to prevent deiodination. The researchers screened more than 1,800 chemicals using each of the three deiodinase enzymes. For deiodinase type 1, 225 chemicals, or 12% of those tested, showed potential for inhibiting the enzyme (Toxicol. Sci. 2017, DOI: 10.1093/toxsci/kfx279). The results for the other two enzymes have yet to be published, but Hornung says that about 17% of the chemicals tested showed potential to inhibit deiodinase type 2 and 3 enzymes.
Although the assay is performed in a 96-well plate, it is not quite high-throughput, Hornung says. Analysis requires separating free iodide from the proteins and other material in the assay media, then using a colorimetric test to quantitate the iodide. “We can run three 96-well plates a day,” Hornung says.
The researchers are now working to integrate the assays developed so far into a human thyroid follicle model that includes TPO, NIS, T4 and T3 export, and some other proteins needed for thyroid hormone synthesis.
However, to fully assess chemical effects on the thyroid system, scientists still have much to investigate beyond the biomolecules they’ve examined so far. The team recently started investigating the role of an iodide recycling enzyme, iodotyrosine deiodinase, in the thyroid gland. That enzyme “may be critical in areas where you have iodine deficiency,” Hornung says. “Animals like to conserve all the iodine they can to make sure they are able to make enough thyroid hormone.”
And there are other potentially important assay targets. “The ones with no information, especially thyroid hormone membrane transporters,” are lower priority for assay development because of R&D challenges, not because they aren’t significant, Crofton says. Researchers have also so far ignored the liver, which is key for regulating metabolism of thyroid hormones and hormone elimination if levels get too high.
EPA researchers have come a long way in the past few years, developing new assays to target various steps in the thyroid network, as well as other parts of the endocrine system. Once the current effort is complete, however, EPA will need to decide whether its second-tier assays are too slow. With so many of the test chemicals flagged as potential endocrine disruptors—495 potential TPO inhibitors, 404 potential NIS inhibitors, and 225 potential deiodinase inhibitors, among others—EPA may find that those tests will also take decades to complete. Then the race will be on to replace those five second-tier in vivo ecotoxicity tests involving quail, fish, frogs, and rats with high-throughput nonanimal alternatives.