Testing for Carcinogens: Shift From Animals to Automation Gathers Steam—Slowly

For more than 50 years, scientists have screened chemicals for carcinogenicity by testing them in rats and mice. But public health agencies in the U.S. are now moving toward what's been described as a major shift in toxicology: aiming to replace animal tests with automated, cell-based assays. The U.S. Environmental Protection Agency (EPA) is the latest to go in this direction. In its Strategic Plan for Evaluating the Toxicity of Chemicals , released in March, the EPA concedes that its present screening protocols can't keep pace with the enormous backlog of untested chemicals in commerce, which now stands at about 74,000. EPA's new goal is to look for chemically induced “perturbations” to molecular pathways involved in toxicity and disease consequences such as cancer. Robert Kavlock, Ph.D., director of the agency's National Center for Computational Toxicology, predicts that with this approach, researchers could screen chemicals for safety in a matter of weeks, for roughly $25,000 each. To fully screen just one chemical now can take up to 3 years and $6 million. “So we're talking about economies of scale that are really enticing,” Kavlock said. The agency's strategy closely mirrors recommendations made by the National Academy of Sciences (NAS), in its 2007 report Vision for Toxicity Testing in the 21st Century . Commissioned by the EPA, that report calls for transforming toxicology from “a system based mainly on whole-animal evaluations” to one that looks for “biological changes in cells, cell lines, or cellular components—preferably of human origin.” To give an oncology example, James Yager, Ph.D., senior associate dean for academic affairs at the Johns Hopkins School of Public Health in Baltimore, cited the PI3 kinase signaling pathway. Mutations in this pathway are associated with cancer. In a pathway-based system, Yager said, scientists would identify the critical mutations, or “nodes,” that irreversibly push PI3 kinase signaling toward cancerous outcomes. Chemicals that trigger these mutations could be deemed carcinogenic, without the need for animal testing. Other agencies pursuing a similar agenda include the National Toxicology Program (NTP), part of the National Institutes of Health, which recently began adding high-throughput screening and other nonrodent approaches to its chemical testing program. And the National Chemical Genomics Center (NCGC), also at NIH, has pioneered high-throughput, robotic systems that can screen thousands of chemicals simultaneously for specific endpoints in cells. Last year, EPA joined forces with NTP and NCGC in a Memorandum of Understanding known as Tox 21. With this agreement, the three agencies will collaborate on ways to identify molecular pathways that produce toxic outcomes, including cancer. Pathway-based screening differs from the current state of mechanistic toxicology, which involves dosing animals, generally at high concentrations, and then characterizing damage to tissues and cells. Because of cost and resource constraints, animal studies typically investigate no more than three dose levels. Automated screening programs, on the other hand, such as that ongoing at the NCGC, can test up to 15 dose levels. This approach provides much more information about dose–response curves for target perturbations. Validating pathway predictions will require studies in people, said Daniel Krewski, Ph.D. , a professor at the University of Ottawa and chair of the panel that produced NAS's 2007 report. Scientists will have to associate pathway perturbations with adverse effects in those exposed to chemicals that are already in the environment at levels high enough to evoke measurable effects in blood or other media, he said. These corresponding studies are needed for confidence that the perturbations identified in cell cultures have diagnostic value and pose a risk to human populations. Eventually, Krewski said, once there are enough data from corresponding studies, screening could be performed entirely “in silico,” or in computers, using virtual models of cells, organs, and tissues. Daniel Krewski, Ph.D. Krewski concedes that the new approach poses daunting challenges. “The big hurdle is going to be mapping out all the toxicity pathways,” he said. “That's as big a challenge as decoding the human genome; we're asking scientists to work out every one of the cellular and molecular mechanisms by which chemical agents affect human health.” Assays used in Tox 21 research now focus on several pathways, including those involved in cytotoxicity, upregulation of the p53 tumor suppressor gene, agonist/antagonist activity for receptors in the nucleus, and mechanisms relating to DNA damage. The possible number of toxicity pathways—which NAS considers finite—remains an open question, Krewski said. “If we aggregate across all adverse health outcomes, we're probably talking about hundreds or even thousands,” he said. Some scientists, including Lorenz Rhomberg, Ph.D., a principal investigator at Gradient Corp., a toxicology consulting firm in Cambridge, Mass., hope that the resulting data will change EPA's default view that all carcinogens—until proven otherwise—act via genotoxic mechanisms. Under this assumption, even one carcinogenic molecule can bind with DNA and trigger chain reactions leading to cancer. When plotted graphically, genotoxic carcinogens show a linear dose–response curve. In this view, any exposure, no matter how small, increases risk. According to critics of the linear approach, mounting evidence shows that some carcinogens have dose thresholds below which cancer risks are negligible. Formaldehyde, for instance, an EPA-designated “probable human carcinogen,” produces DNA–protein cross-links at low doses but spawns rodent tumors only at doses high enough to induce cell proliferation. Nevertheless, EPA regulates formaldehyde on the basis of nasal tumors in rodents, exposed during long-term inhalation studies. Nonlinear approaches could justify less stringent cleanup levels at toxic waste sites, which make them attractive to industry. But EPA officials aren't generally willing to translate dose thresholds into chemical standards that allow more human exposure. Indeed, among more than 120 EPA-regulated carcinogens, the agency has applied nonlinear assumptions only to three: chloroform; organic arsenic; and atrazine, a common herbicide. Rhomberg describes a longstanding clash at EPA between laboratory scientists and policy officials. “It's the policy maker's job and nature to be skeptical of new findings, but that skepticism can be overdone,” Rhomberg said. “Too often, EPA officials say, ‘Yes, the science is all well and good, but we're going to stick with the old methods.’ At some point, someone's going to counter, ‘If that's the case, then why are we even bothering to study these mechanisms in the first place?’ We have to see how far we can get with this, even if it's not the ultimate basis for decision making. We can show it's worth pursuing and that it will get better with time.” Gina Solomon, M.D., a senior scientist with the Natural Resources Defense Council, concedes that it's a fair point that EPA should consider relevant information about the shape of the dose–response curve. “But the willingness to examine default assumptions has to cut both ways,” she added. “When a more health-protective linear default is discarded for an individual chemical, the evidence for a threshold should be sufficiently strong that EPA is confident that human health will be protected. It makes sense that a departure from linear defaults should be the exception rather than the rule.” For EPA's part, Kavlock emphasized that with its new strategy, the agency is demonstrating a willingness to take mechanistic data seriously. “It was produced with input from senior members across all the EPA offices,” he said. “There's an explicit recognition that we're in a scientific transition and that the business part of the agency needs to come along with it.” Still, it's unlikely that Tox 21 will change existing risk assessment practices soon. Peter Pruess, Ph.D., director of EPA's National Center for Environmental Assessment, said that the agency has been overwhelmed by molecular datasets that it still can't use. “At the moment, the gathering of data is running far ahead of our ability to understand it,” he said. “We have to figure out how we can turn it all into useful information.” Moreover, the cell-based assays used in screening now suffer troubling limitations, including an inability to account for metabolic transformation: how the body converts some chemicals into reactive toxicants. Assays with isolated liver hepatocytes, for instance, could miss toxicity if they don't also address what's happening in that organ's Kupffer cells. That's because hepatocytes and Kupffer cells sometimes activate chemicals in tandem. Addressing metabolic activation is high on the list of Tox 21’s research priorities, Kavlock said. Writing in the journal Toxicological Sciences , John Doull, M.D., Ph.D., professor of toxicology at the University of Kansas Medical Center in Kansas City, claimed that the pathway approach “will not work in complex systems such as the central nervous system, where any pathway perturbation observed in animals or human cells … may be many steps away from the site of damage.” Doull also warned that the “… strategy does not adequately distinguish between effects and adverse effects in a context with which the toxicological and risk assessment communities are currently familiar…” Echoing those concerns, Solomon predicted difficult challenges in communicating pathway-derived risks to the public. “Your average person on the street understands that when something causes birth defects in a rat, there's a level of concern,” she said. “But when you talk about perturbations in thyroid hormone homeostasis, it's harder for the public to know what to think about that.” Preuss expects that the Tox 21 vision will proceed in two phases. In the near term, pathway screens will be used to prioritize the backlog of untested chemicals for further evaluation. Later—he offered a guess of several decades from now—cell-based data could guide policies to set human safety limits for chemical exposure. Doing that, he emphasized, will require that scientists redefine their concepts of adverse effect, by considering pathway perturbations in cell cultures as proxies for more obvious toxicity in animals. Meanwhile, environmental toxicologists working on Tox 21 have been collaborating with colleagues in the pharmaceutical industry. Pharma scientists are supplying compounds to EPA's testing program, according to William Pennie, Ph.D., executive director of compound safety prediction at Pfizer Global Research and Development, “so we can make sure it represents not just industrial and agricultural chemicals but also the types of chemical structures we work with in the pharmaceutical industry.” “I think we're on parallel tracks with the science,” said Pennie, a coauthor of NAS's 2007 report. “A fundamental understanding of toxicology mechanisms helps the EPA in its goal to prioritize testing, but it also helps us prioritize which compounds we take forward in product development. We're supplying compounds to EPA's testing program, so we can make sure it represents not just industrial and agricultural chemicals but also the types of chemical structures we work with in the pharmaceutical industry.” As to funding Tox 21, NAS recommended about $100 million per year for 10–20 years. That figure approximates the NTP's current annual budget, Kavlock said. NAS also recommended the creation of a stand-alone institute to coordinating the agenda, but Kavlock rejects that position. “I don't see any need for that,” he said. “We just need to work together more efficiently; we don't need any more bricks and mortar because we've already got the people.”

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.